QUALITY ASSURANCE OF TREATMENT PLANNING SYSTEMS PRACTICAL EXAMPLES FOR NON-IMRT PHOTON BEAMS

QUALITY ASSURANCE OF TREATMENT PLANNING SYSTEMS PRACTICAL EXAMPLES FOR NON-IMRT PHOTON BEAMS Ben Mijnheer Agnieszka Olszewska Claudio Fiorino Guenther Hartmann Tommy Kns Jean-Claude Rosenwald Hans Welleweerd

2004 First edition ISBN 90-804532-7 2004 by ESTRO All rights reserved No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise without the prior permission of the copyright owners. ESTRO Mounierlaan 83/12 1200 Brussels (Belgium)

AUTHORS:
This booklet could only be written thanks to the enthusiastic co-operation of all members of the QUASIMODO group: Ben Mijnheer, Amsterdam, The Netherlands (Projectleader) Carlos De Wagter, Gent, Belgium (Co-Project leader) Rafael Arrans, Sevilla, Spain Annemarie Bakai,Tbingen, Germany Joerg Bohsung, Berlin, Germany Wojciech Bulski, Warszawa, Poland Claudio Fiorino, Milano, Italy Sofie Gillis, Gent, Belgium Germaine Heeren, Brussel, Belgium Guenther Hartmann, Heidelberg, Germany Dominique Huyskens, Leuven, Belgium Tommy Kns, Lund, Sweden Agnieszka Olszewska, Amsterdam, The Netherlands Marta Paiusco, Reggio Emilia, Italy Bruce Perrin, Manchester, United Kingdom Jean-Claude Rosenwald, Paris, France Francisco Sanchez-Doblado, Sevilla, Spain David Thwaites, Edinburgh, United Kingdom Hans Welleweerd, Utrecht, The Netherlands Peter Williams, Manchester, United Kingdom

IV

ACKNOWLEDGEMENTS
Although it is virtually impossible to thank all persons who contributed in one or other way to the drafting of this booklet, we very much appreciate the co-operation of the following persons: Maria do Carmo Lopes, Coimbra, Portugal; Maria Gavrilenko, Kiev, Ukraine; Sara Broggi, Milan, Italy; Corine van VlietVroegindeweij, Amsterdam, The Netherlands; Miln Tomsej, Brussels, Belgium and Sonny La, Lund, Sweden. We also like to thank the International Atomic Energy Commission (IAEA) for letting us have insight in their forthcoming report Commissioning and Quality Assurance of Computerized Treatment Planning. The constructive comments from the vendors of treatment planning systems are also appreciated. QUASIMODO was funded by the EUROPEAN COMMISSION, Directorate General Health and Consumer Protection - Europe Against Cancer Programme, and is part of the ESQUIRE Project: Education, Science and Quality assurance In Radiotherapy in Europe, Grant Agreements S.12.322029 and SPC.2002480.

VII

CONTENTS: ESTRO BOOKLET NO. 7: QUALITY ASSURANCE OF TREATMENT PLANNING SYSTEMS PRACTICAL EXAMPLES FOR NON-IMRT PHOTON BEAMS

AUTHORS ACKNOWLEDGEMENTS CONTENTS 1. INTRODUCTION 1.1 1.2 1.3 General remarks Quality assurance of a TPS Division of QA tasks to be performed by the vendor and the user of a TPS 1.4 Incompleteness or redundancy of recommended tests 1.5 National and international reports discussing QA of treatment planning systems 1.6 Contents of the booklet

V VII IX 1 1 2 4 6 7 9 11 11 12 14 15 16 17 18 21 21 22 24 26 26 27

2. ACCURACY REQUIREMENTS AND TOLERANCE LEVELS 2.1 Recommendations for dosimetric and geometric accuracy 2.2 How to express deviations between measurements and calculations? 2.2.1 Evaluation in low dose gradient areas 2.2.2 Evaluation in high dose gradient areas 2.2.3 Combined evaluation of dosimetric and spatial deviations 2.2.4 Evaluation and reporting deviations for a large number of points 2.3 Acceptance criteria for the accuracy of photon beam dose calculations 3. ANATOMICAL DESCRIPTION 3.1 Basic patient entry 3.2 Image input and use 3.2.1 Image input 3.2.2 Contour input 3.2.3 Image use 3.2.4 Co-ordinate system of images

IX

3.3 Anatomical structures 3.3.1 Definition of anatomical structures 3.3.2 Automated contouring 3.3.3 Manual contouring 3.3.4 Manipulation of contours 3.3.5 Construction of volumes 4. BEAM DESCRIPTION 4.1 Beam definition 4.1.1 SAD, SSD and field size 4.1.2 Gantry rotation 4.1.3 Collimator rotation 4.1.4 Table movement 4.1.5 Jaw definition and beam co-ordinates 4.1.6 Multi-leaf collimator definition 4.1.7 Wedge and block insertion 4.1.8 Consistency check of beam co-ordinate system 4.1.9 Warnings and error messages 4.1.10 Bolus definition 4.2 Beam display 4.2.1 Beams-eye-view (BEV)-display 4.2.2 Beam position and shape 4.2.3 Beam position in BEV 4.2.4 Block position in BEV 4.2.5 MLC-shaped field 4.2.6 Bolus position 4.3 Beam geometry 4.3.1 Automatic block and auto-leaf positioning 4.3.2 User-defined block 4.3.3 DRR: linearity and divergence 4.3.4 Input, change and edit functions 5. DOSE AND MONITOR UNIT CALCULATION 5.1 Beam characterisation set 5.1.1 Data input process 5.1.2 Documentation 5.2 Dose calculation 5.2.1 Open square fields 5.2.2 Open rectangular fields 5.2.3 Variation in SSD

27 27 28 28 28 30 33 33 33 34 34 35 35 35 35 36 36 37 37 37 38 38 38 38 38 39 39 39 39 39 41 42 42 43 43 44 44 45

5.2.4 Wedged square field 5.2.5 Wedged rectangular fields 5.2.6 Field with a central block 5.2.7 Blocked field 5.2.8 Inhomogeneities 5.2.9 Oblique incidence 5.2.10 Missing tissue 5.2.11 Off-axis square field 5.2.12 Off-axis elongated field 5.2.13 Wedged off-axis field 5.2.14 Off-plane field 5.2.15 Square MLC field 5.2.16 Off-axis square MLC field 5.2.17 MLC-shaped field 5.2.18 Block and tray insertion 5.3 2-D and 3-D dose verification 5.3.1 2-D dose distribution 5.3.2 Dose-volume histogram 5.4 Monitor unit calculation 6. PERIODIC QUALITY CONTROL 6.1 Data input process 6.1.1 Digitiser 6.1.2 Film scanner 6.1.3 CT data 6.1.4 MR data 6.1.5 Integrity of simultaneous input 6.2 Software 6.2.1 MU calculation 6.2.2 Standard treatment techniques 6.2.3 MLC-shaped field 6.3 Data output process 6.3.1 Printing/plotting devices 6.3.2 Block cutting device 6.3.3 Treatment plan transfer 7. EXAMPLES OF TESTS 7.1 Tests from chapter 3: Anatomical description 7.2 Tests from chapter 4: Beam description 7.3 Tests from chapter 5: Dose and monitor unit calculation

45 45 45 46 46 46 47 47 47 47 48 48 48 49 49 51 52 52 54 55 55 55 56 56 56 56 56 57 57 57 57 57 58 58 59 59 65 68

XI

APPENDIX A.1 Definitions APPENDIX A.2 List of abbreviations and symbols APPENDIX A.3 Categorisation of tests REFERENCES

83 87 89 93

XII

1. INTRODUCTION
1.1. GENERAL REMARKS A computerised treatment planning system, TPS, is an essential tool in the design of a radiotherapeutic treatment of cancer patients and some groups of patients suffering from non-malignant diseases. A typical installation of a TPS is in a radiotherapy network, virtual or real, in which it functions together with other systems required for patient treatment (Figure 1.1). Nowadays, image transfer in a hospital is often done via DICOM (Digital Imaging and Communications in Medicine). The DICOM standard has been developed to meet the needs of manufacturers and users of medical imaging equipment for interconnection of devices in networks. A single DICOM file contains both a header (which stores information about the patients name, the type of scan, image dimensions, etc), as well as all of the image data (which can contain information in three dimensions). DICOM RT (DICOM in radiotherapy) contains the radiotherapy modalities: RT Image (e.g., CT slices), RT Dose (e.g., dose matrices), RT Structure Set (e.g., target volumes and organs at risk), RT Plan (e.g., gantry-/collimator-/couch-angles, field sizes, blocks, MLC settings) and RT Treatment Record. The TPS should be able to share patient data and must be capable to import information from imaging devices such as CT and MR. The imported data should represent correctly both image content (pixel value) and geographic (pixel position) information. Imaging information may also come from a hospital-based system supporting all departments with patient data. Although most manufacturers claim to apply DICOM, some TPSs are not fully DICOM compatible, while also differences exist in the way different vendors have implemented DICOM in their system.

Figure 1.1 Radiotherapy network showing imaging devices, treatment units, treatment planning workstations, and supporting (e.g., booking, image archiving and record and verify) systems.

For radiotherapy tasks, CT information has the highest importance since it is not only used for target and organ delineation, but also for dose calculation. Therefore, geometry data and Hounsfield numbers must be transferred and interpreted correctly. During commissioning of the TPS the whole range of input variables of the treatment unit, including movements, must be entered and verified. If this is not possible one may have to use in-house written data transfer tools with all the problems of maintaining and verifying these utensils. Independent of the type of solution, it is important that the correctness of data transfer is verified prior to clinical use of both the TPS and the accelerator(s). This procedure includes identifying the dimension and (asymmetric) position of the collimator opening, gantry and table rotation and the co-ordinate systems of these parts of the linac. Adopting the international standard for definition of co-ordinates, movements and scales, as given in IEC Report 61217 (IEC 1996), for all systems involved (linac, TPS, etc.), will facilitate the communication between radiotherapy equipment, particularly if use is made of DICOM RT. In this way the internal coordinate system of the TPS can be transferred directly to the specific treatment unit, thus ensuring data integrity and avoiding misinterpretation of data. The use of look-up tables or similar transfer tools can facilitate this task. A quality assurance, QA, programme of a TPS must therefore include tests of maintaining data integrity after transfer in a radiotherapy environment. These tests should be repeated after software upgrades of either the TPS or the linac.

1.2 QUALITY ASSURANCE OF A TPS After the installation of a TPS in a hospital, acceptance testing and commissioning of the system is required, i.e., a comprehensive series of operational tests has to be performed before using the TPS for treating patients. These tests, which should partly be performed by the vendor and partly by the user, do not only serve to ensure the safe use of the system in a specific clinic, but also help the user in appreciating the possibilities of the system and understanding its limitations. In the past some irradiation accidents happened with patients undergoing radiation therapy, which were related to the misuse of a treatment planning system. Most often these accidents were not the result of system malfunctioning but due to a lack of understanding of how the TPS works. More details related to the incidence of accidents in radiotherapy can be found in several reports (IAEA 2000, IAEA 2001, ICRP 2001, Cosset 2002). In many of these accidents, a single cause could not be identified but usually there was a combination of factors contributing to the occurrence of the accident. The most prominent factors were deficiencies in education and training, and a lack of quality assurance procedures. Good training, as well as the availability of well-documented quality assurance procedures, therefore have a huge impact in preventing planning errors. Over recent years, increased attention has been paid to quality assurance of treatment planning systems by various national and international organisations. Examples include Van Dyk et al., 1993, Shaw, 1996, SSRPM 1997, Fraass et al., 1998, Mayles et al., 1999, IAEA 2

2004 and NCS 2004. These reports provide recommendations for specific aspects of QA of a TPS, such as anatomical description, beam description and dose calculations. However, contrary to the situation for treatment machines, not many sets of practical recommendations for commissioning and QA of a TPS exist. Although a lot of information is provided in these reports, it is difficult for a TPS-user to decide which tests are absolutely necessary to perform by an individual user, and which tests the vendor or users groups of a specific system should perform. Also the number of tests provided by some of these reports is so overwhelming, that it would require a huge investment in manpower to perform the recommendations given in these reports. For those reasons departments with limited physics staff often choose for a pragmatic approach, thus doing only those QA tests they consider of direct importance for the use of the new TPS in their department. Particularly with respect to the 3-D aspects of planning systems, there are no clear guidelines which specific tests should be performed before the clinical use of a 3-D TPS can be started in a safe way. For that reason it was decided during the 1999 ESTRO Physics Meeting in Gttingen, Germany, that ESTRO would start activities in the field of QA of a TPS. It was emphasized that ESTRO would concentrate on those activities not yet covered by other groups or already described in other reports. In August 2001 the ESQUIRE project, funded by the European Communities, EC, started for a period of two years. The aim of that project was to increase the confidence level of clinicians for embracing optimised radiotherapy treatment regimens by making sure they can be achieved without an increase in severe side effects. One of the actions proposed for this purpose was to develop QA procedures for optimised radiotherapy planning and delivery, as outlined in the part of the project called QUASIMODO (QUality ASsurance of Intensity MODulated radiation Oncology). QUASIMODO will promote the safe introduction of advanced technology in RT by developing procedures for the QA of treatment planning systems, and by exploring new methodology for the verification of intensity-modulated radiation therapy, IMRT. From the review of national and international documents discussing QA of treatment planning systems (Section 2.2. of this booklet) it became clear that there is a need for a minimum number of tests. These tests should not only be suitable for small hospitals with limited resources, but are also needed by large (university) centres having a high patient load or limited staff. These tests should not be too cumbersome to perform and should cover the most essential parts of a TPS required for accurately planning of established conformal radiotherapy techniques. It should be realised, however, that the minimum number of tests to be performed in a specific institution depends very much on the local clinical practice. It might well be that some of the tests described in this booklet are not necessary, while for advanced treatment techniques many more tests should be performed before the TPS can be applied clinically. The first aim of the QUASIMODO project was therefore to identify a set of examples of tests for QA of treatment planning systems, easy to perform by users of different types of TPS. 3

Recently a rapidly increasing number of institutions started with clinical implementation of IMRT. By varying the beam intensity over the treatment fields it is possible to deliver the radiation dose more conform to irregularly shaped target volumes. In this way it is possible to deliver a higher dose to the tumour while at the same time reducing the dose to surrounding healthy tissues. For the QA of these advanced techniques, general guidelines have been formulated by Ezzell et al. (2003). An interesting observation from a survey on the status of IMRT in Europe in 2002 was that almost every institution applied its own phantom/dosimetry system for the verification of treatment delivery. Obviously a specific solution was found in each institution, but no common approach or method was adapted at that time. It is the second aim of the QUASIMODO project to design tests and provide guidelines for the verification of IMRT. This second part of the QUASIMODO project is not only related to QA of treatment planning systems but includes QA of the treatment delivery as well. The results of that part of the project will therefore be the topic of another report and will not be discussed further in this booklet. It should be noted that a number of tests described in this booklet are also applicable to treatment planning systems handling IMRT. However, for the additional problems encountered in IMRT, other tests are required.

1.3 DIVISION OF QA TASKS TO BE PERFORMED BY THE VENDOR AND THE USER OF A TPS The International Electrotechnical Commission, IEC, published an International Standard on Requirements for the safety of radiotherapy treatment planning systems (IEC 2000). Similar to other IEC documents concerning medical equipment, e.g., for linear accelerators, this International Standard defines a number of requirements to be complied with by vendors of such equipment in order to provide protection against the occurrence of safety hazards to patients. Compliance with these requirements should be checked by testing by the manufacturer and demonstrated to the customer. In this booklet we have adopted the same approach, i.e., a suggestion is made for a division of QA tests to be performed by the vendor or by an individual user (or a group of users). As a consequence a large number of tests can already be performed before the system is installed in the hospital, i.e., before acceptance testing of the system by the user starts. In particular tests demonstrating the accuracy of dose computation, but also other tests concerning anatomy description and beam description, which may either be performed by the vendor or another user or users group, should be made available by the vendor. The results of these tests should be described in documents accompanying the system, and only be spotchecked by an individual user. In Appendix A.3 each test described in this report is assigned either to the category Vendor Acceptance Tests or to the category User Commissioning Tests. In the first category a large number of tests to be performed by the vendor of a specific system are presented. As a rule, the manufacturer should perform the generic tests related to the proper 4

functioning of the system, i.e., tests that will give the same results for all systems independently of the local customisation or of the beam data entered into the system in a specific institution. Such tests will complement those that can be found in, for example, IEC Report 62083 (IEC 2000). On the other hand, it is not the intention of the proposed user tests of the second category, to repeat all checks of the generic accuracy of specific features/performance characteristics implemented in a TPS. This should have been established within the development process and be able to be demonstrated by the vendor. It should be noted that these generic tests are sensitive to the quality of the of the generic data, and the vendor should ensure its accuracy. However, there cannot always be a strong division between both sets of tests since interaction between the vendor and the user is also required. As an example, prior to installation it is highly desirable to formulate customer acceptance tests agreed with the vendor, satisfactory completion of which should be a contractual requirement. Next, it is necessary for the tests to ensure that the local implementation, including beam data etc., gives results consistent with the established generic capability. Further verification of performance is required following the installation of upgrades, new software releases and changes to customisation including the addition or editing of, for example, beam data. Therefore, a QA programme for new software upgrades should be defined depending on the modules of the software that have been changed in the new version. Historically, after installation of a TPS by the vendor, and checking if the system is operational, it is left to the users who will test the system in the way they like. User training may take place on- or off-site, as negotiated in the sales contract, after which the user will precede through the acceptance/verification-testing phase on its own. It is the intention of this booklet that this procedure will change in the future. The vendor should provide now the results of the tests designated in Appendix A.3 as vendor tests, in combination with agreed acceptance data, either prior to or after installation of the system. In addition the vendor should present, under reference to this booklet, a list of commissioning/acceptance tests that the user should perform on its own. After a reasonable time period, negotiable based on circumstances but before first clinical use, the vendor and user should then together discuss the results of the tests and decide about complete, or maybe partial, acceptance of the system. In this way on-site participation of the vendor during the acceptance-testing phase can be minimized. Although a number of tests described in Appendix A.3 belong to a standard acceptance programme, not all results of tests may already be available for each vendor. It may therefore take some time for the vendors to implement this new procedure and to put together the acceptance test materials and data sets. Finally it is necessary to define tests, which should be repeated, for ongoing quality control purposes, during the lifetime of the TPS. Definition of the local QC programme will be based on careful consideration of risk, and particular attention given to those aspects of performance that are likely to change with time and or use. We recommend adoption of the tests provided in this booklet as a de facto standard set of acceptance tests, with users only customizing the acceptance criteria, depending on 5

local clinical practice and intended use. It is therefore hoped that this booklet will not only provide guidelines for the TPS user to perform a QA programme in the hospital, but that it also becomes part of the user manual of a vendor and will be used for customer training. Based on the experience of both users and vendors, adaptations of the tests may in the future be necessary.

1.4 INCOMPLETENESS OR REDUNDANCY OF RECOMMENDED TESTS Generally speaking, quality assurance may be achieved by: 1. Vendors performance statements in system specifications based on generic data. 2. Vendors demonstration to individual users of features that have been customised by the vendor for the users particular installation. 3. Users tests of features customised by the user on their installed system. 4 Users investigations to ensure sufficient understanding of the performance and limitations of the installed system. 5 Users tests for periodic quality control of the installed system. According to this list, the tests in this report describe a series of quality assurance tasks (tests), which are examples of actions necessary to check the capabilities and limitations of a TPS. For the purpose of this report tests are defined broadly, to include all actions required giving assurance that the TPS is fit for its intended purpose according to this list of QA actions above. However, the set of tests cannot be taken as a complete list, and each user has to identify and perform additional tests specific for the clinical needs in his/her department. Some TPS vendors have already a module integrated in their software for comparison of measurements and calculations, which makes it easy for a user to perform a relatively large number of tests. On the other hand, it is not necessary to perform all the tests described in this booklet locally. Users groups of a specific vendors system can in many cases provide useful information about tests that have already been performed with the system. Information of this type is published from time to time in the peer-reviewed literature. Knowledge of the results of tests performed by other users of the same system will not only reduce the actual time involved in performing the tests described in this booklet, but may also point to the need of other tests. A vendor might also wish to collect the results of the tests described in this booklet from various users of the system in a systematic way, and share that information with other (potential) customers. Such a collection of data gives insight in the limitations of a specific algorithm and may lead to improvements of the systems performance. Some of the tests described in this booklet may appear as a duplication of tests already performed during system development or at a later stage by other users. This redundancy is, however, considered to be still very useful because some tests have an outstanding 6

training value and prevent individual users to misinterpret the system behaviour. Moreover, some performance characteristics already checked by the vendor should be spot-checked by an individual user. Finally it should be noted that not all tests presented in this booklet can be performed for every type of TPS. Because of their importance for those systems that are able to perform these tests, we preferred to keep them in the booklet.

1.5 NATIONAL AND INTERNATIONAL REPORTS DISCUSSING QA OF TREATMENT PLANNING SYSTEMS A large number of papers and reports about quality assurance of a 3-D TPS exist. The main activities described in these publications concern the verification of dose calculations. More recently several national and international organisations have drafted, or are in the process of finalizing, more general guidelines for the commissioning and quality assurance of treatment planning systems. It is cumbersome to describe and compare those documents, because they differ very much in the number of QA issues discussed, the attention paid to details and their layout. Also the processes of commissioning and quality assurance are sometimes difficult to distinguish. In this section we will briefly discuss the main issues of these documents. Probably the first report in which performance evaluation of a treatment planning system was mentioned, was presented by a Nordic group (Dahlin et al., 1983). In that report user requirements of a CT-based computerized TPS were discussed. At that time it was already noted that: The physicist in charge needs a standardized set of tests and test conditions to control the reliability of the output. Ten years later a Canadian report (Van Dyk et al., 1993) described for the first time in a systematic way commissioning and quality assurance of treatment planning systems. It concentrates on dosimetric aspects of a planning system. After discussing the accuracy required for dose calculation in several regions of a photon beam, a number of tests are described related to dose calculation, such as tests of depth dose characteristics (TPR, TAR, PDD, etc.) and beam profiles. The first UK report (Shaw, 1996) includes an interesting description of possible problems and errors that may occur with hard- and software of a TPS. Part of the document focuses on checking general hardware (peripheral devices) such as digitiser, plotter, visual display system and CT interface. In the part dedicated to external photon beam software, there is a recommendation to check the algorithms for reconstruction of measured (data-set) fields and non-measured (interpolated) fields. However, no details are given how to perform such tests. The Swiss report (SSRPM, 1997) recommends a number of tests, of which some are given in detail. For instance, it has a test about the correctness of adding bolus into the dose calculation algorithm. Additional tests include checks of the block-cutting device, beam tests 7

(profiles, TPR, TAR, PDD, TMR, etc.) and of some standard irradiation techniques such as tangential breast fields, four-field box technique, mantle fields and non-coplanar head & neck fields. The American report (Fraass et al., 1998) is a comprehensive report and contains a large amount of information on quality assurance of the whole treatment planning process. It discusses extensively a number of dosimetric and non-dosimetric aspects of a TPS and describes a lot of associated tests. It provides a general framework how to design a QA programme for all kinds of TPS, both for external therapy and brachytherapy. Because of its completeness, it might be difficult for a user to choose those tests that are most urgently needed for the situation in that particular institution. The second UK report (Mayles et al., 1999) describes the QA of several parts of the treatment planning process such as the use of imaging devices (simulator, CT and MR), target volume definition and dose prescription. Topics to be checked are listed, however, without giving details how to perform these tests. The recommendations from the American report (Fraass et al., 1998) have been adapted in a more recent publication (Van Dyk et al., 2003). It describes an approach analogous to the AAPM TG53 Report in evaluating the non-dosimetric components of a TPS, and gives more detailed dose calculation tests, including the variables to be considered and the possible range of parameters. The report of the Netherlands Commission on Radiation Dosimetry (NCS, 2004) has formulated a detailed set of tests to assure the accurate functioning of a TPS. A number of these tests were the starting point of the tests recommended in this booklet. In the NCS report mainly point dose measurements are discussed and only a limited number of 2-D tests, such as the verification of beam profiles, are given. Tests for these issues are also important and therefore included this booklet. Quality assurance of brachytherapy treatment planning systems has recently been discussed in a new ESTRO booklet (Venselaar and Prez-Calatayud, 2004). In an integrated TPS, where external beam therapy software runs on the same platform as the brachytherapy software, in- and output issues are basically identical. However, there are still many special features of brachytherapy software that need to be addressed in a proper QA programme, independent of whether the system is stand-alone or part of a larger system. In ESTRO Booklet No. 8 the physicists tasks at commissioning and continued use of a brachytherapy TPS, the verification of a treatment plan, and the clinical aspects of quality assurance of a brachytherapy TPS are discussed in detail. A forthcoming IAEA document contains a wealth of information on commissioning and quality assurance of computerized treatment planning both for external beam therapy and brachytherapy (IAEA, 2004). It describes a large number of tests, but does not provide a simple protocol that can be followed step-by-step by a user in a hospital for the QA of a TPS. Also a number of tests presented in the IAEA document refer to testing the system itself, and are not specific for an individual user. The IAEA document can at this moment be considered as the most complete reference work in the field of QA of treatment planning sys8

tems. It is, however, too comprehensive, and for a user too difficult to choose those tests that are most urgently needed for the situation in a specific institution.

1.6 CONTENTS OF THE BOOKLET After comparing the tests provided in the reports prepared by the different national and international task groups, it became clear that none of these reports provided a set of tests which can easily be performed at the hospital level. Also the separation of tests to be performed by the vendor, either before or during the acceptance testing, and by the user during the commissioning of the system, has not been described in any of these reports. In this booklet an attempt has therefore been made to provide such a division of QA tests to be performed by an individual user, or by the vendor or a users group. In the following chapters two sets of examples, based on tests given in some of these reports, will be described in more detail. These tests can be done without extensive study of the documents described in the previous section. It should, however, be clear that the tests provided in this booklet should be considered as practical examples of an elementary part of a QA programme of a TPS. In many situations more tests are needed before the system can be operated in a reliable way in the clinic. The other documents discussed in Section 1.5 should then give guidance in what other tests should be performed either by the vendor or the user. The main topics elucidated in the following chapters can be summarized as follows: In Chapter 2 requirements and acceptance criteria for the accuracy of the dosimetric and geometric aspects of a TPS are discussed. In this booklet a number of tests are presented to verify these issues. Various methods to express differences between measurements and results of tests performed by a TPS are discussed in this chapter. In Chapter 3 tests to verify the basic patient entry data in the TPS are given, to confirm that no trivial, but potentially very serious, mistakes can happen. In the same chapter the anatomical model or description of the patient, one of the most critical issues of a TPS, is discussed. The tests provided in this part of the chapter should ensure that the data used for creating an anatomical structure are correct and properly linked to a specific patient. In this chapter image information of a patient and geometrical elements of a TPS are also tested. In Chapter 4 beam definition, beam display and beam geometry are tested. The definition of geometry and co-ordinate system of a specific treatment machine is necessary in order to use a radiation beam in a TPS for treatment planning purposes. Beam definition and its use are critical items for the accurate design of a treatment plan and should therefore be carefully checked. In Chapter 5 tests to verify the accuracy of the dose calculations performed by a TPS are presented. The beam data input, dose calculation algorithms and criteria of accep-tance are taken into account. Verification of monitor unit calculation will also be discussed in this chapter. Special attention will be paid to check if monitor unit calculation methodology fits with the normalisation method used in the TPS. 9

In Chapter 6 tests are given to perform periodically at specified time intervals to verify that nothing has changed with the TPS. Periodic tests are different from the tests associated with accepting or commissioning a new TPS or a new software release. Some additional tests will be provided, related to the most common techniques for treatment of tumours of head & neck, lung/esophagus, breast and pelvis/prostate. In Chapter 7 examples are given of the tests described in Chapters 3 to 5. These tests have recently been performed on different commercially available treatment planning systems, and therefore reflect situations that can occur to any user. These examples have been chosen to illustrate in more detail in which way the tests can be performed in practice. They also show the usefulness of these tests, because in some cases unexpected limitations of the systems were discovered. There are several other aspects of QA of a TPS, such as plan documentation and export, system management and system security that are important for a user. Potential problems with respect to plan documentation and export are related to data transfer from the TPS to additional devices (computer controlled block cutters, MLC, record-and-verify system, etc.). Because it is difficult to design tests valid for each TPS in combination with its specific use in a radiotherapy department, no additional tests are given in this booklet for these aspects. The reader is referred to the information provided in the other reports on this topic. System management is crucial for a safe and reliable access to the TPS, which should be restricted to those who are qualified to do it. Nowadays, a TPS often consists not only of standard computer hard- and software, but includes also graphics workstations, servers and other peripheral equipment. System management of a TPS covers: hardware (computer system), software (data) and the network configuration based on DICOM connectivity. Another aspect of QA of a TPS is therefore to take care of the technical performance and integrity between hard- and software. These issues have been discussed extensively in other reports and will not be discussed in this booklet. It should be noted that in this booklet we restrict ourselves to external, non-IMRT, photon beams, although some parts of it, e.g., Chapters 3 and 4 (Anatomical Description and Beam Description), are also valid for external electron beam treatments.

10

2. ACCURACY REQUIREMENTS AND TOLERANCE LEVELS

2.1 RECOMMENDATIONS FOR DOSIMETRIC AND GEOMETRIC ACCURACY Requirements for the accuracy of a treatment planning system must be seen in the light of the total uncertainty in the 3-D dose delivery to a patient. For this purpose, all steps in the planning and delivery process should be considered. An important criterion is the accuracy in the absorbed dose distribution required from a clinical/radiobiological point of view. In an IPEM report (Mayles et al., 1999) an overview is given of the clinical evidence for accuracy in radiotherapy. From this survey it can be concluded that a difference in absorbed dose of about 10% is often detectable in tumour control, and that a difference of about 7% in absorbed dose can be observed for a number of normal tissue reactions. From an extensive review of dose-response data, Brahme et al. (1988) concluded that the standard deviation in the mean dose in the target volume should be at most 3% (one standard deviation, SD) to have control of the treatment outcome with a 5% tolerance level. This is in agreement with a recommendation given by Mijnheer et al. (1987) based on a review of steepness of doseresponse curves observed for normal tissue complications, and other clinical observations. These clinical/radiobiological observations point to the need that the absorbed dose should be delivered within 7-10%. Assuming that this number is equivalent to a confidence level of 95%, the standard deviation in the absorbed dose delivered to the (ICRU) dose specification point, must be as low as 3-5%. According to IAEA Report TRS 398 (IAEA 2000), the latest code of practice for the calibration of high-energy radiation beams, the uncertainty in the absorbed dose to a point under reference conditions is about 1.5% (1SD). If the treatment planning and treatment delivery process include all other steps after the calibration of the beam, then a window of uncertainty of 2.6 to 4.8% (1SD) remains for the rest of the treatment process. Because several steps of this dosimetry chain do not depend on the use of a computerized TPS, a much smaller uncertainty or tolerance must be assigned for the TPS as a QA goal. In the past generally rather simple recommendations were given for the required accuracy of dose calculations, such as 2% or 2 mm in regions of the beam with small or large dose gradients, respectively (ICRU 1987). Later these recommendations have been refined and adapted to the accuracy of dose calculations that can be achieved in clinical practice. Recently a summary of the accuracy requirements given in different reports and by various groups has been published (Venselaar et al., 2001). The set of recommendations given by Venselaar et al. has been adopted in this booklet and will be discussed in more detail in Section 2.3. Almost no recommendations are available with respect to geometric accuracy requirements in radiation therapy. In the IPEM report a value for the accuracy on posi0tioning of field edges and shielding blocks in relation to the planning target volume of 4mm (1SD) is recommended (Mayles et al., 1999). This number seems rather large with respect 11

to current attempts to reduce margins around target volumes, while also the shielding of organs at risk needs a much better precision. Also the use of portal imaging and immobilisation will result in smaller geometric (set-up) uncertainties. Furthermore, these requirements concern the actual patient irradiation, i.e., the whole treatment process, and consequently a much smaller uncertainty or tolerance must be used for the TPS. A better approach might therefore be to compare the geometric accuracy requirements of a planning system with those of an accelerator, which are of the order of 1 to 2 mm. It should be noted that the actual geometric accuracy achievable with a TPS depends strongly on items like image resolution, grid size and dose matrix geometry. More work is needed to formulate geometric accuracy requirements for treatment planning systems, in relation to the actual patient treatment.

2.2 HOW TO EXPRESS DEVIATIONS BETWEEN MEASUREMENTS AND CALCULATIONS? The verification and QC process of a TPS often reveals the problem of how to express the deviations between measurements and calculations and how to define criteria that must be fulfilled to use the system clinically. Different methods must be used depending on the physical quantity to be tested, as well as on the region in the radiation beam to be studied. Geometry tests involve spatial units, while dosimetric tests involve point, line, or 2-D/3-D matrix comparisons. In principle, all dosimetric tests are point-based (from a single point to a large amount of data) thus these comparisons can be made voxel-by-voxel. This approach is suitable in low dose gradient areas. In high dose gradient areas, e.g., a penumbra, the spatial deviation must also be considered, as will be discussed later. A useful tool that handles this situation is the dose/distance-to-agreement check (Harms et al., 1998), which has been further developed into the so-called -index (Low et al., 1998, Depuydt et al., 2002, Low and Dempsey, 2003, Bakai et al., 2003). In this way acceptance criteria can be specified as a combination of the accepted dose deviation, e.g., 3%, and the accepted distance-to-agreement, e.g., 3.0mm. For this purpose recommended tolerance levels can be applied as given, for instance, in Table 2.1 in this booklet. The results of comparisons between calculated and measured dose distributions can be normalised to the local dose value, to the dose at a specific point inside the beam under consideration, or to the dose in a reference field. The latter procedure puts, for example, dose points outside the beam edges in relation to the treated volume. Local normalisation in low dose regions may give deviations of multiple tens of a percent, although they may sometimes be rather insignificant. If an organ at risk is present in that region, normalisation to the local dose value is, however, relevant. Also for IMRT applications accurate calculation of the dose outside the beam edges is important, but accuracy requirements for these situations are still under development. In these situations, the point chosen may be the depth of dose maximum or the reference depth (i.e., the depth used during beam calibration). Also the point on the 12

central beam axis at the same depth has been proposed. In this booklet we provide tolerances for both options. In regions with a steep dose gradient it is recommended that the dose difference should be translated in a distance-to-agreement value. Differences between dose values predicted by a TPS and measured values at the same point are the result from limitations of the dose calculation, uncertainties in the measurement procedure, or fluctuations in the output of the accelerator. In order to obtain the accuracy of the dose calculation performed by the treatment planning system itself, it is recommended in this booklet to express the deviation d(i) between calculated (subscript c) and measured (subscript m) dose values as the ratio between absolute dose values, after a normalisation to reference conditions. This should be interpreted as all dose measurements (single or multiple points) shall be given in absolute dose, D, per monitor unit, M, normalised to a measurement in a reference field (Figure 2.1). The following equation gives the ratio d(i) between calculated and measured dose values at a point i after such a normalisation: d(i) =

(Dc (i) / Mc (i)) / (Dc (ref) / Mc (ref)) (Dm (i) / Mm (i)) / (Dm (ref) / Mm (ref))

(Eq. 2.1)

Alternatively the deviation can be expressed as the percent difference d%(i) between calculation and measurements, normalised to the measured quantity: d%(i) = 100. [d(i)-1].

SSDref SSDi

Isocentre Isocentre Pointi Pointref

Figure 2.1 The left figure defines the geometry for the dose determination at a specific point i, and the right figure defines the reference situation (source-skin-distance, SSD, field size at isocentre, s, depth, d). Normalisation is obtained by taking the ratio of the absorbed dose per monitor unit, D(i)/M(i), for each point i. In practice the same value of M is often taken for all points i, as well as for M(ref). The figure is valid for both dose measurements (m) and dose calculations (c).

13

In practice, the reference field can be chosen as the one used during calibration of the treatment machine. Another reference field, which may conveniently be used, is the one applied for periodic control of the output of the accelerator. Defining the reference beam as the one used during calibration and following the latest dosimetry protocol, i.e., IAEA Report TRS 398 (IAEA 2000), means that all calculations and measurements should be normalized to the absorbed dose at 10cm depth in a 10cm x 10cm field. The absorbed dose in this context is also given per monitor unit, which is in compliance with the discussion above. This procedure is identical to that used during the commissioning of the accelerator where all measurements must be normalized to the output in a reference field to account for the variations in output of the accelerator. The absorbed dose in each test should therefore be given per monitor unit (similar to dose rate where the time is replaced by monitor units). Following this concept, the number of monitor units for a certain geometry is implicitly also checked. It should be noted that in most treatment planning systems the dose is expressed as dose to water. In some systems several approaches are possible; for instance the dose to muscle can be chosen, which might be the preferred choice if a Monte Carlo dose calculation engine is available. The vendor should clearly indicate which dose definition is chosen. A simple experimental protocol for the verification of dose is to determine the absorbed dose in the reference geometry prior to the points of interest, as well as just afterwards. The average dose for these measurements can, if they are within a certain value, e.g., 1%, be used to normalise the verification dose measurements. This is also a check of the stability of the treatment unit during the measurement session. It should be noted that variations in the output of the accelerator during relative dose measurements are often taken into account by using a reference chamber placed in a corner of the field. All relative dose measurements obtained in this way still have to be related to the absolute dose for a certain number of monitor units. The procedure for the evaluation of deviations between calculation and measurement for various situations will now be described in more detail in the following paragraphs. A number of numerical examples will be presented in Chapter 7.

2.2.1 Evaluation in low dose gradient areas In areas with low dose gradients it is sufficient to evaluate the dose deviation independently of the spatial consideration. The absorbed dose, Dm(i), for a certain number of monitor units, Mm(i), is determined using some kind of detector (e.g., ionisation chamber, diode, diamond). In close relation (before and afterwards) to this measurement the dose in the reference field, Dm(ref), is determined for a delivery of a number of monitor units Mm(ref). Thus we know Dm(i)/Mm(i) as well as Dm(ref)/Mm(ref) and can determine the measured normalised dose at point i. To facilitate the process it is advantageous to give the same number of monitor units in the two cases, i.e., Mm(i)=Mm(ref).

14

Step two involves the creation of an identical geometry/beam set-up in the TPS. In a similar way as applied for the measurement situation, both the geometry of interest and the reference geometry must be defined. An attractive approach is to apply the same number of monitor units during the calculation and the irradiation, Mm(i)=Mc(i) and Mm(ref)=Mc(ref), and then letting the program calculate the dose to these two points, Dc(i) and Dc(ref). In this way the measured and calculated dose at point i relative to that at the reference point can directly be compared. However, in some TPSs it is complicated to calculate the dose for a given number of monitor units, and in some cases even impossible. Instead it is necessary to determine the number of monitor units for a certain absorbed dose to the point (either i or ref). Thus we will get two different values Mc(i) and Mc(ref) for the two situations where the absorbed dose at these two points is identical, Dc(i)= Dc(ref). This procedure has to be repeated for each point of interest, whereas the reference field has to be calculated only once. When such a comparison is repeated for many points, recalculating for each point i the number of monitor units Mc(i) for the dose Dc(i) becomes very cumbersome and is unnecessary. Actually, when it has been done for one point i0, the ratio Dc(i)/Mc(i) used in equation 2.1 is more directly obtained from the relative dose distribution using equation 2.2. Dc (i) Mc (i) Dc (i) Dc (i0) Dc (i0) Mc (i0) Mc (i0) Dc (i) Dc (i0) = x Mc (i) Dc (i0) Mc (i0)

(Eq. 2.2)

where Dc(i)/Dc(i0) is the relative dose normalised to point i0 for the same number of monitor units applied to all points i, i.e., Mc(i)=Mc(i0).

2.2.2 Evaluation in high dose gradient areas Evaluation of deviations between measurements and calculations in high dose gradient areas based on dose differences may result in very large figures, which are very sensitive to geometric uncertainties. Thus a better approach is to quantify these dose differences as distanceto-agreement. This distance (spatial deviation) is equal to the smallest distance r(i) between a measurement point rm(i) and a point rc in the calculation volume with the same absorbed dose D. Interpolation within a given calculated or measured dose matrix may facilitate this evaluation. In principle this should be done in three dimensions, (where r, rm and rc are three-dimensional vectors), but such a procedure can also be applied to a one- or twodimensional data set, i.e., for depth doses, profiles, and isodose lines.

15

2.2.3. Combined evaluation of dosimetric and spatial deviations A simple method to combine dosimetric and spatial deviations is to calculate both and to select the smallest value relative to the recommended tolerance value as given, for instance, in Table 2.1. A more complex but very elegant concept, that combines dosimetric and spatial deviations into one single figure of merit, is the -evaluation method, which was first presented by Low et al. (1998) and later refined by several groups (e.g., Depuydt et al., 2002, Low and Dempsey, 2003, and Bakai et al., 2003). The method can be considered as a comparison of points in the four-dimensional dose-position vector space. The two points to be compared are the points (rc,Dc) and (rm,Dm), where r is the three-dimensional spatial co-ordinate and D the absorbed dose co-ordinate. If the base vectors of the co-ordinate system are equal to the dose criteria, d, and the spatial criteria, r, respectively, then if the length of the normalized vector between these points is less or equal to unity, agreement is fulfilled (Figure 2.2).

Figure 2.2 Dose-distance vector space showing the measured dose Dm at point rm, and the calculated dose Dc at rc. (For clarity the figure is drawn in two dimensions where the spatial dimensions are reduced to one).

For all points (rc,Dc) the difference between the measured and calculated dose d(i)=Dm(i)-Dc has to be determined, as well as the distance between the points r(i)=rm(i)-rc. The -value, is then found from scaling with the base vectors according to: (Eq. 2.3)

For <1 the measured dose at point (rc,Dc) is within the acceptance criteria. The evaluation should preferable be performed in three dimensions where r=(rx,ry,rz). The method can also be applied to dose values along a line (e.g., a depth dose curve or a beam profile), and two-dimensional data sets. In all cases, the calculated dose data may still be a full 3-dimen16

sional dose matrix. As a matter of fact, lower -values are obtained by considering the full dimensionality of the calculated dose matrix. Dm and Dc must have the same unit, i.e., preferably in Gy per MU normalised to the reference situation. In order to be in agreement with the tolerance values given later in Table 2.1, the dose differences d(i) and d should be normalized to some dose value. As will be discussed in Section 2.3, the normalising dose should be the local dose Dm in most circumstances except where this dose is too low to have a clinical significance. In such case, the normalising dose should preferably be defined inside the measured open field. In any case, it should be made clear which normalising quantity is used. Because different tolerance values are valid for different regions in a radiation field, the -evaluation method should in principle be able to handle these different criteria. These more sophisticated evaluation methods are particularly useful if inhomogeneous 3-D dose distributions in organs at risk are analysed, as encountered in IMRT. In Chapter 7, as an example, the use of the -evaluation method is illustrated for a 4-field box technique.

2.2.4. Evaluation and reporting deviations for a large number of points When the number of comparison points is large, the above concept of reporting deviations between measurements and calculations will easily collapse, and a method of compiling these deviations into a single number is required. Standard statistical tests could be a way to perform these evaluations, e.g., a paired Students T-test. Other methods have also been published, for instance the use of the quantity confidence limit as discussed by Venselaar and Welleweerd (2001) and Venselaar et al., (2001). The confidence limit is based on the average (systematic) deviation between measurement and calculation for a number of data points in a comparable situation, and the standard deviation (SD) in this average of the differences. The confidence limit is then defined as the sum of the average deviation and 1.5 SD. The factor 1.5 is based on experience and was a useful choice in clinical practice (Venselaar et al., 2001). The cited papers applied this concept to dose deviation, but it can easily be expanded to both the spatial deviation and the gamma concept. When dose data (depth dose curves, profiles or any line between two points, 2-D and 3-D dose matrices) are determined, it is convenient to do all measurements for a certain beam configuration at the same time. In this case, gain and offset of the scanning system can be kept constant, thus yielding all line doses in the set related to the same relative output. A complementary dose measurement with this system, in combination with an absolute dose measurement, under reference conditions (10cm depth for a field size of 10cm x 10cm), connects the whole set of line dose measurements with the number of monitor units given. In this way we have accomplished the same procedure as outlined above. The procedure to get these data sets from a TPS varies for the different systems, and is too diverse to describe in this booklet. 17

This statistical approach can be used in a number of different ways. For example, data may be grouped according to field size, beam quality, or beam modifier (e.g., wedge filter). Such a grouping must be done with caution, because no systematic deviation for a small subgroup, e.g., for very small, very large or off-axis fields), should be lost.

2.3 ACCEPTANCE CRITERIA FOR THE ACCURACY OF PHOTON BEAM DOSE CALCULATIONS Following the discussions above, the acceptance level for the accuracy of dose calculations of a TPS should be around 2%. This value can be used for areas where the absorbed dose is rather homogenous, e.g., inside the central part of a beam. Different acceptance criteria can, however, be formulated depending on the position in the beam. Figures 2.3 and 2.4 show the various regions that can be defined in a photon beam, incident on a homogeneous phantom. In principle there are two areas with a homogenous dose, well inside or far outside the beam. In between we have the penumbra and build-up regions with a high dose gradient.

Figure 2.3 Definition of different regions in a radiation beam, based on the magnitude of the dose gradient, for which different acceptance criteria for the accuracy of dose calculations, , are valid.

Assuming that the penumbra is defined as the distance between the 20 and 80% dose level, and recognizing that the penumbra distance for modern linacs is of the order 5mm, we get a dose gradient of 12% per mm. We can also define the build-up region as a high dose gradient area. Because of these differences in dose gradient, low and high dose gradient areas should have different acceptance criteria. Using the concept of Venselaar et al. (2001) we can divide a photon beam into four regions according to: 1. Points along the central axis of the beam beyond the depth of dose maximum: low dose gradient area. 18

2. Points on and off the central axis in the build-up and penumbra region. This region includes also points in the proximity of interfaces: high dose gradient area. 3. Points inside the beam (e.g., inside 80% of the geometrical beam) but off the central axis: low dose gradient area. 4. Points outside the geometrical beam or below shielding blocks, jaws, MLC, etc. where the dose is lower than, for instance, 7% of the central axis dose at the same depth: low dose gradient area.

Figure. 2.4 Illustration of the different regions where the criteria 1 - 4 and RW50 can be applied to compare calculated and measured values of a depth-dose curve (upper panel) and a beam profile (lower panel) (adapted from Venselaar et al., 2001). For the profiles, data is normalised at the central beam axis, although it is recommended in this booklet to use a normalisation to dose per monitor unit under reference conditions.

Acceptance criteria for the accuracy of dose calculations for each region will be noted as 1, 2, 3, and 4. In the low dose gradient areas it is convenient to trace dose deviations, while in the high dose gradient a distance-to-agreement is a better parameter to express deviations between measurements and calculations. This is consistent with the recommendation above of using the -analysis when evaluating dose distributions. The values for these four regions, given as confidence limits as proposed by Venselaar et al., and adopted in this booklet, are given in Table 2.1. Figure 2.4 shows in which regions of a depth dose curve or a beam profile the different tolerances can be applied. The values given in Table 2.1.could also serve as guidelines to set the dose criteria, d, and 19

the spatial criteria, r, of the gamma index given in Eq. 2.3. The regions of the beam defined in Figures 2.3 and 2.4 are interconnected in the gamma index concept, but different high and low dose gradient regions might have different values. In addition, values can be applied to any dose distribution, including distributions for multi-beam composite plans. However, it is still useful to differentiate low and high dose regions, which could be done by setting a threshold, e.g., 7% of the reference dose in the open field region, below which the dose deviation is normalised to this reference dose rather than to the local dose. Also given in this table are the radiological width, the width of a profile measured at half its height compared to the value at the beam axis, and the beam fringe, the distance between the position of the 50% and 90% values relative to the maximum of the profile in the penumbra. It should be noted that the recommendations for 4 (outside the beam edges in the low dose/low dose gradient region), might be too permissive for IMRT applications. Often a considerable part of the dose delivery, either to the target volume or to an organ at risk, results from the addition of a number of dose distributions outside the beam edges. The accuracy of the dose calculation of that part of a beam should then have a higher accuracy, depending on the IMRT technique and position of the organ at risk relative to the target volume.
Table 2.1 Tolerances , given as the confidence limit, for the dose deviation d%(i), for the various regions in a photon beam. (Adapted from Venselaar et al., 2001). These data should be considered as recommendations for good clinical practice and not as absolute values valid under all circumstances.
Region Homogenous, simple geometry 2% 2 mm or 10% Complex geometry (wedge, inhomogeneity, asymmetry, blocks / MLC) 3% 3 mm or 15% More complex geometries**** 4% 3 mm or 15%

1 2*

Central beam axis data - high high dose, low dose gradient Build-up region of central axis beam, penumbra ragion of the profiles - high dose, high dose gradient Outside central beam axis region high dose, low dose gradient

3% 30% (3%) 2 mm or 1% 2 mm

3% 40% (4%) 2 mm or 1% 3 mm

4% 50% (5%) 2 mm or 1% 3 mm

4** Outside beam edges low dose, low dose gradient RW50 Radiological width high dose, *** high dose gradient. 50-90 Beam fringe high dose, high dose gradient * **

One of the two tolerance values should be used. These figures are normalized to the local dose, or at the dose at a point at the same depth on the central beam axis, or the open part of the field in case of blocked fields (in brackets). ***` The percent figure should be used for field sizes larger than 20 cm. **** More complex geometry is defined as a combination of at least two complex geometries.

20

3. ANATOMICAL DESCRIPTION
In this chapter tests are given that can be helpful to achieve confidence in the correct representation of the anatomical description of patients introduced into the TPS. The set of tasks described in this chapter is dependent on the structure of the specific TPS. Basically all tests should be performed for all systems, but for some systems they have to be adapted or even may be non-applicable. Particularly in this chapter, and in Chapter 4 (Beam Description), a large number of tests that are presented should be performed by the vendor of a specific system during the acceptance testing (See Appendix A.3). Depending on the installed configuration and/or connectivity between the various imaging modalities and the TPS, additional tests should be performed by an individual user. A number of the tests presented in this chapter concern the verification of features/functionality present in the system and need not necessarily be performed at the users site, but may also be demonstrated during user training sessions elsewhere. Before starting the tests described in this chapter, it is assumed that the connectivity between the various imaging modalities and the TPS is working according to DICOM standards. Because the infrastructure of each department is different, it is impossible to design connectivity tests valid for all configurations present in radiotherapy departments, and no suggestions for such tests are given.

3.1. BASIC PATIENT ENTRY In this section tests are proposed for checking the response of the system in situations related to the uniqueness of a patient associated with existing patient data, and retrieving patient data from the system. As a minimum we assume that the following information of the patient is available: ID-number, family name, target volume and in case of a CT-scan, a CT ID-number. Test the following situations: a. Introduce two patients with the same last name but different ID-numbers in the TPS. Patient 1: Xxxxxx Yyyyyy: 20021972 Patient 2: Xxxxxx Yyyyyy: 11041970 Record the system response: not possible/no warning, not possible/warning, possible/warning, possible/no warning. In the case of a warning, record the warning. b. Introduce two patients with different last names but the same ID-number in the same directory of the TPS. Patient 1: Xxxxxx Yyyyyy: 20021972 Patient 2: Zzzzzz Wwwww: 20021972 Record the system response: not possible/no warning, not possible/warning, possible/warning, possible/no warning. In the case of a warning, record the warning. 21

c. Introduce the same patient twice in the same directory of the TPS. Patient 1: Xxxxxx Yyyyyy: 20021972 Patient 2: Xxxxxx Yyyyyy: 20021972 Record the system response: not possible/no warning, not possible/warning, possible/warning, possible/no warning. In the case of a warning, record the warning. d. Enter a second anatomical description for an existing patient. Verify that the TPS warns for overwriting and/or association of a new description to an existing case. Introduce into the system the same patient with two different target volumes. Patient 1: Xxxxxx Yyyyyy: 20021972 breast Patient 2: Xxxxxx Yyyyyy: 20021972 cervix Record the system response: not possible/no warning, not possible/warning, possible/warning, possible/no warning. In the case of a warning, record the warning. e. Check how easy (number of steps, questions, warnings from TPS) it is to delete a patient from the TPS. f. Check the limitations of the TPS concerning moving/copying patients/plans from one directory to another.

3.2 IMAGE INPUT AND USE Image acquisition for treatment planning is usually performed by computed tomography (CT) and magnetic resonance imaging (MRI). In special cases, positron emission tomography (PET) and single photon emission computed tomography (SPECT) are additionally used (Grosu et al., 2000, Henze et al., 2000, Levivier et al., 1995, Pirotte et al., 1997). Each imaging modality is applied for specific reasons: the CT dataset may be mapped to the electron density of the tissue and is needed to calculate dose distribution within the patient. MRI, on the other hand, provides superior soft tissue contrast and is used to delineate the tumour and the organs at risk. PET and SPECT images can be used additionally to measure the relative metabolic activity for detecting differences in tumour regions or differentiating tumour from necrosis. These complementary aspects can be integrated into treatment planning by correlation of the images from different modalities. As an essential prerequisite for the treatment planning process and, in particular for the correlation process, the images that are converted into the TPS must reflect the real geometry of the patient, i.e., possible distortions of the images have to be minimized. In addition, the accuracy of the correlation also depends on the correct functioning of the multimodality registration software included in the treatment planning program such as image fusion (co-registration). Therefore, appropriate tests are indispensable prior to the first application to patients. The aim of this part is to propose some tests mostly related to CT imaging, however also including a geometry test for MRI, PET and SPECT. 22

For the tests presented in this chapter, related to the correct use of CT data, two types of phantoms are recommended. A description of examples of such phantoms are given in this section, but other phantoms available in a department, having similar construction and welldefined dimensions, are also suitable for this purpose. It should be noted that CT data are not only used for describing patient anatomy, but also used to define electron densities of the various tissues to be irradiated. For that purpose a table that converts HU values into electron densities is present in the system. Phantoms containing non-water equivalent materials can then be used for testing the correctness of this curve. Phantom A: Made from PMMA (Perspex, Plexiglas, Lucite) slabs, having outer dimensions of 30cm x 15cm x 20cm, on which lead markers are attached. These spherical lead markers are partly placed inside and partly sticking out of the phantom. The number of lead markers should be enough to identify the dimensions of the phantom and the direction of the phantom; head-feet and anterior-posterior direction markers are helpful The markers can be located at the edges of the phantom or arranged in some geometrical shape, and should be easy to recognize on a radiographic film or a digitally reconstructed radiograph (DRR). Figure 3.1 presents an example of a phantom with 12 lead markers asymmetrically located at the edges of the phantom.

Figure 3.1 Schematic view of PMMA phantom (A) with lead markers

Phantom B: This phantom contains 3 types of inhomogeneous internal structures, which are helpful for testing both structure-related and inhomogeneity-related issues: waterequivalent, lung-equivalent and bone-equivalent. These structures have cylindrical shape with a diameter of 3cm, 4cm and 3cm, respectively (Figure 3.2). If this phantom is used for the determination of the relation between CT-number and relative electron density, the values of the relative electron density of these structures should be known. There are several types of these phantoms commercially available. In some treatment planning systems the modern approach to dose calculations is to use CT numbers for mapping to tissue type, which fully specifies the chemical composition, and density. That information is then used to determine the cross-section/interaction coefficient data required for dose calculations. For that option in a TPS it is necessary that the atomic composition and specific density of the phantom material and its internal structures be known. 23

Figure 3.2 Schematic view of the inhomogeneous phantom (B)

An increasing number of phantoms are becoming commercially available to verify the delivery of advanced treatment techniques such as IMRT. These phantoms can, sometimes, also be used for volume evaluation and beam alignment as described in this chapter (e.g., Craig et al., 2001). Another phantom useful for multi-image-modality tests may consist of PMMA cylinders containing smaller empty cylinders to be used as marker structures such as target points or linear tubes (Karger et al., 2003; Figure 7.4). These empty cylinders can be filled with liquids that are specific for the respective imaging modality, e.g., iodine contrast agent for CT and MRI, and radioactive nuclides for PET and SPECT. Iodine contrast agent is well visible in CT as well as in MRI, so the phantom does not have to be refilled between consecutive CT- and MRI-measurements.

3.2.1 Image input a. Identity consistency of scans. Follow the instructions below and check the results: Introduce into the TPS a patient data set with duplicated slices and verify that the TPS gives appropriate warnings. Introduce into the TPS a data set in which the CT field-of-view is changed and verify that the TPS gives appropriate warnings and/or verify that the dimensions of the phantom are correct in the different slices (e.g., phantom A). Introduce into the TPS CT scans with the same names and patient-ID and verify that the TPS gives appropriate warnings. b. Scan parametersvarying slice thickness. Generate at the CT-scanner and introduce into the TPS scans with varying distance between slices, for instance: 10, 5 and 7mm. The TPS should warn for this difference and prevent from further data processing or use the right distances. c. Two different sets of images for the same patient Introduce into the TPS two different sets of CT slices, with the same z-value but with different image resolution and/or different field-of-view for the same patient. Check if the TPS allows the user to select the CT set for importing. 24

d. Maximum number of CT slices Check in the documentation the maximum number of CT slices permitted by the TPS. Check how the system works if the amount of input images is larger than this number, for instance if there is a possibility of selecting images. e. Patient orientation Make CT scans of a phantom with various orientations that are used clinically (head first, feet first, prone, supine left and right side). Import the files into the TPS and check the representation of the patient orientation. The phantom needs to be properly marked (e.g., phantom A). f. Integrity of simultaneous input Check that simultaneous input of data of the same patient from different devices, e.g., a digitiser, film scanner, CT scanner or MR scanner, by different users of these channels, does not interfere in the TPS. g. Check if the same or different users can work on a patient file, which is already opened. h. Geometric integrity of slices Scan a phantom of well-defined dimensions and check its dimensions in the TPS (e.g., phantom A). i. CT number representation Check the representation of CT numbers (Hounsfield Unit values). Use a phantom consisting of different densities simulating water-equivalent, lung-equivalent and boneequivalent tissues (e.g., phantom B). Import the CT data in the TPS, measure the HU values and compare them with the original numbers (measured on CT). CT numbers in the TPS can be obtained by making a profile or defining a small volume in the different regions and determine the corresponding maximum, minimum or average CT number (e.g., phantom B). j. Text information Check if on all graphics, beam displays, and other windows on the screen, the patients last name and the ID-number are shown. Also plan number and trial number (if applicable) and version number of the TPS should be visible. Test how many characters can be used in the last name of a patient and check if after saving and reloading the same characters are shown.

25

3.2.2 Contour input a. Check the accuracy of the manual digitiser input: Enter test contours drawn onto a sheet of paper into the TPS and check the position of all corner points, using preferably the measuring tools provided within the software Verify also the dimensions of the plot of the digitised structures. b. Check the accuracy of the film scanner input: Enter test contours into the TPS and check the dimensions along the scan direction and at both sides along the transport direction, using preferably the measuring tools provided within the software. Verify also the dimensions of the plot of structures introduced by the scanner into TPS. It should be noted that the dimensions of the contours of a scanned image are dependent on the choice of gray-scale control setting (window and level).

3.2.3 Image use a. Geometry of reconstructed images: Scan the phantom (in normal head-first orientation) and reconstruct sagittal, coronal and oblique planes and check the geometry by comparison with the distance between markers placed in the phantom (e.g., phantom A). b. Orientation of reconstructed images: Check if the reconstructed planes in the previous test have the correct orientation (use a beam projection with selected gantry and table angles). c. Grey-scale representation: Compare a hardcopy made at the CT-scanner with the display of the TPS using the same window/level setting (if possible use a film scanner to enter a hard-copy into the TPS). d. Grey-scale representation of reconstructed images: Reconstruct oblique planes (if option available) that are slightly tilted with respect to coronal and sagittal orientations and check the grey-scale interpolation (phantom A). e. Windowing and zooming images Check if windowing and zooming functions work together.

26

3.2.4 Co-ordinate system of images Introduce a phantom co-ordinate system for the phantom, which is to be used to test the geometry. Its origin should be defined in such a way that it is well identifiable in the real phantom as well as detectable on its images. Use of special markers may be helpful. Refer the position of real objects (markers) with respect to this phantom co-ordinate system Introduce a co-ordinate system for the images of the phantom. This system should have its origin at the same position as at the real phantom. Refer the position of the image the same objects (markers) with respect to this image coordinate system Determine the vectorial difference between objects and image of objects and compare with tolerance values

3.3 ANATOMICAL STRUCTURES 3.3.1 Definition of anatomical structures a. Unique identification: Check how an anatomical structure is related to a specific patient plan and whether there is a risk of confusion. Check if you can use different structure sets within the same plan. Check if it is possible to make an image fusion and if the right structure set is used. Examples: Define a new structure with the name that already exists for another structure. Define identical names for two structures on different data set. b. Unique properties: If your TPS needs structures with special properties, like external contour, check the TPS response to the attempt to define a second structure with a different name but also with the same property (e.g., phantom B). c. Maximum number of contours per anatomical structure: Check a clinically relevant number (for instance 60) of contours per CT-slice set that may be defined for each structure. Test for each CT-slice that a clinically relevant number of contours per CT-slice (for instance 25) can be defined. The response of the system when those numbers are reached or exceeded should be tested (e.g., phantom B).

27

3.3.2 Automated contouring Make a CT-scan of a test phantom with well-defined geometry containing materials with densities similar to lung, bone and soft tissue. Let the TPS automatically generate external and internal structures. Contour all structures and check the contour by eye if it is following the structure (e.g., phantom B). Verify the printed contours of internal structures. a. Correct geometry in automated contouring Measure along the axis the diameter of the well-defined structures. Measure left and right separately to detect shifts. b. Threshold changing for automated contouring If there is a possibility to change a threshold in automated contouring option, repeat test 3.3.2.a for different threshold values.

3.3.3 Manual contouring These tests are both valid for manual contouring on CT images (using the computer mouse) or via a digitiser. a. Contouring direction If contours may be entered clockwise (CW) and counter-clockwise (CCW), enter the same contour in both directions. Check by visual inspection or volume analysis that the volumes do not depend on whether the contours are entered CW or CCW, also when mixed (e.g., phantom B). b. Maximum number of points For each contour there will be a maximum number of points that may be defined and the response of the system when this number is reached or exceeded should be tested.

3.3.4 Manipulation of contours The tests described in this section concern the verification of the correct handling of contours of structures in individual planes. a. Add a margin to a contour The 2-D option of contour expansion should be verified. A cylindrical- (e.g., phantom B) and/or triangle- (with sharp angle defined by 3 points on a single plane) shape is recom28

mended. Expand a drawn contour by 5mm or 10mm. Measure the distance between the original and expanded contours. b. 3-D structure expansion Due to different possibilities of contour expansion algorithms in a TPS, the method of handling such a 3-D expansion needs to be tested. The structure must be limited in the cranio-caudal direction (e.g., sphere, pyramid or cone, cylindrical section). Expand a predefined contour using a 5-mm or 10-mm margin in the direction normal to the surface of that structure. Measure the distance between the original and expanded contours in various planes and check the correct shape of the surface of the expanded structure in a qualitative way in 3-D display under various angles of view, and in a quantitative way in orthogonal planes. If negative margins are allowed, make a contraction after an expansion, using the same margin, and check that the original structure is restored. c. Check: correcting, adding, deleting and copying of contours and structures. d. Validation Any change made in the contours of structures should be detected by the TPS. Check if an existing 3-D surface of that structure is invalidated and preferably recalculated automatically. Use the above-defined structure, change on two slices the structure considerably and check on the 3-D view if the 3-D surface is reconstructed properly (e.g., phantom B). If dose computation had been previously performed, check that it is invalidated after contour modification. e. Check special TPS tools for example: the point reduction option (e.g., phantom B). f. 2-D contour transfer Construct the contours in various planes of another data set, make hardcopy plots and check the dimension of the contours (phantom B). (This test is only useful when matching is used clinically). Some new TPSs have a tool where contours can be drawn on sagittal and coronal views and then derived into axial slices, which option should then also be verified. g. Hidden contours (if option available) Check the number of hidden contours and the copy/move/delete option from visible to hidden and vice-versa. h. Bifurcated structures Check if the system can handle bifurcated structures, i.e., if it is allowed to contour different areas in the same slice belonging to the same (target) volume, for instance nodes in H&N treatments. 29

i. Interpolated structures Construct the structure using the interpolation option in the TPS and check the dimensions of the contours. Use any phantom with structures such as sphere, oblique cylinder or horseshoe-shaped oblique structure. j. Bolus definition and related options See section. 4.1.10.

3.3.5 Construction of volumes The tests described in this section concern the verification of the correct handling of volumes of structures in a set of CT slices. These 3-D aspects of the TPS are extremely important in modern radiotherapy, and correct handling of volumes, including the expansion of structures along the direction normal to the surface of that structure, should therefore be verified. a. Volume computation Let the TPS compute a phantom surface from the derived contours. Check the correct shape of the surface. Let the TPS compute the volumes inside the structures surface and compare with the exact value. Expand and contract structures with a margin for instance of 5, 7, 10, 12 and 15mm (e.g., phantom B). b. Construction of a volume from a set of CT slices with non-regular spacing Construct surfaces from contours in planes that are not regularly spaced. For this purpose delete a number of contours of an original regular set, generate the surface again and compare with the original surface. Use, for instance, the original data set with 1mm space between slices (e.g., phantom B) and delete some slices. Calculate the volume in all situations and compare. c. Construction of a volume from a set of CT slices with non-sequential order of slices during contouring Define a cylindrical structure from contours drawn on each second slice and go back to define the structure on the other slices. Check the computed volume (e.g., phantom B). d. Construction of a volume from non-axial contours If possible repeat test 3.3.5a, for orientations other than axial (e.g., phantom B). e. Capping option (automatic interpolation at the beginning and the end of the available contours) Check if the calculated volume agrees with the expected volume in case the distance between slices is changed. 30

f. Volume of subtracted regions Subtract one well-defined volume from another. Compare volumes of a subtracted region calculated by the TPS with those calculated manually (e.g., phantom B). g. Boolean option Check the possibility of adding or subtracting parts of an existing volume to create new volumes. Verify the dimension of the new volume (e.g., phantom B).

31

4. BEAM DESCRIPTION
Beam definition and its use are critical items for the accurate design of a treatment plan. In this chapter a set of performance tests related to beam definition, beam display and beam geometry are proposed. The vendor of the TPS should do the majority of this set of tests during the acceptance-testing phase (Appendix 3). Some of these tests depend on how carefully the customization process was performed. This is the direct responsibility of the medical physicist in charge of the TPS, whether the customization is done by him (her) self or by the vendor.

4.1 BEAM DEFINITION 4.1.1 SAD, SSD and field size Position a beam with rectangular shape on a 3-D phantom with its isocentre 10 cm below the surface, but not at the geometrical centre of the phantom (see Figure 4.1).

Figure 4.1 Schematic view of beam position relative to the phantom.

a. Field size Measure the field size at the isocentre level in an axial and sagittal plane and verify the agreement between measured and stated TPS beam co-ordinates. Define at least three different field sizes, for instance: 5cm x 20cm, 7.5cm x10cm, and 15cm x 30cm. Check the name given to each field dimension (X or Y). In some TPSs the ruler tool cannot be used in BEV display. In this case the verification of the field size in both dimensions (X and Y) can be done using a collimator rotation.

33

b. Divergence Check the correctness of divergence by measurement of the field size at the surface, repeat it for various field sizes, for instance: 5cm x 20cm, 7.5cm x 10cm, and 15cm x 30cm. c. Isocentre position Verify the co-ordinates and graphical position of the isocentre relative to the origin of the phantom. Check the name and direction of all co-ordinate axes and their consistency with the anatomical co-ordinate system. d. SSD or depth verification Verify the SSD or depth as stated by the TPS. e. Repeat the tests 4.1.1 a, b, c, d for SSD>100 cm (for instance 130cm) Plot at least three cases from the above-described situations to verify the agreement between plotted geometries and geometries presented on the screen.

4.1.2. Gantry rotation Position a beam on a phantom with a gantry angle of 30o, 60o and 135o and verify the direction of the rotation, and the agreement of the beam angle on the plot with the TPS beam co-ordinates. Verify also the correspondence between the gantry angle and the orientation of the small patient model on the screen. Suggestion: use an asymmetric field in this test and display isodose curves as well.

4.1.3. Collimator rotation Position a beam perpendicularly incident on a phantom and rotate the collimator to an angle of 30o, 60o and 135o. Verify the collimator angle and its rotation direction using the beam display in an axial and sagittal plane and the BEV-display (Figure 4.2). Suggestion: use an asymmetric field in this test and display isodose curves as well.

Figure 4.2 Position of the field for the verification of the field shape in BEV-display and in the axial planes.

34

4.1.4 Table movement Position a beam perpendicularly incident on a phantom and rotate the table over an angle of 30o, 60o and 315o. Verify the table angle and its rotation direction using the beam display in an axial and sagittal plane and in the BEV-display. Suggestion: use an asymmetric field in this test and display isodose curves as well.

4.1.5 Jaw definition and beam co-ordinates Check the consistency of the beam co-ordinates of the separate jaws (i.e., X1, X2, Y1, Y2) and check their values at the definition level (in most cases at isocentre level), as well as their setting limitations (including over-travel) for open and wedged fields using the beam display in an axial and sagittal plane and the BEV-display (Figure 4.3).

Figure 4.3 Jaw setting for the beam co-ordinates test, in BEV-display.

4.1.6 Multi-leaf collimator definition Check the number of leaves, leaf position and numbering, leaf direction (X or Y), leaf width, over-travel and maximum leaf position, preferably by using the BEV-display. Verify the exact definition of the leaf position in case of rounded leaf ends. Plot the position of the leaves for a number of different leaf settings, including some extreme positions. Check how the (backup) jaw positions are changed according to modification of leaf positions. Suggestion: display isodose curves as well.

4.1.7 Wedge and block insertion Check the possible directions of wedge insertion and the agreement of the wedge display with the insert direction. Check if the wedge direction is changed after collimator rota35

tion and if the insertion direction remains the same. Suggestion: check consistency with isodose curve display. Use an asymmetric block arrangement (Figure 4.4) and check the blocking tray rotation as the collimator rotates. Check the position of the field limits defined by the edge of the blocks in BEV-display on different axial and sagittal slices.

Figure 4.4 Asymmetric block arrangement for block test, in BEV-display.

4.1.8 Consistency check of beam co-ordinate system Verify the consistent use of beam co-ordinates defined in test 4.1.1 and their limitations throughout all parts of the planning process.

4.1.9 Warnings and error messages To ensure the safety of the patient, as well as to avoid collisions, check the adequacy of warnings and error messages following beam input with values that violate the actual limits of the accelerator: Limits for gantry, collimator and table angles. Field size limits (x and y) for wedged and non-wedged fields. Over-travel for asymmetric fields and MLC. Insertion direction for the wedge. Anti-collision warnings, especially in case blocks are applied. Check how they behave when beam parameters are changed (e.g., field size, wedges, blocks, MLC, etc. when machine, SSD or SAD is changed).

36

4.1.10 Bolus definition a. Check whether the generated bolus completely covers the beam aperture projection on the patient skin. b. Check the density and height of the bolus throughout the full beam aperture using CTnumber and distance measurements. c. Check whether the bolus is separated from the body contour of the patient. d. Check manual bolus insertion.

4.2 BEAM DISPLAY In the following section a water phantom of relevant patient size is used, e.g., width 40 cm, height 25 cm and length 30 cm, having two cylinders (Figure 4.5) with two different densities (for instance lung density: 0.3 g/cm3 and bone density: 1.8 g/cm3) symmetrically placed around the origin. In order to simulate lung and bone, the cylinders should have a minimum diameter of 5 and 1 cm, respectively.

Figure 4.5 Water phantom with two cylinders having different density placed symmetrically around the origin.

4.2.1 Beams-eye-view (BEV)-display Set up a symmetric co-axial beam (10cm x 10cm) with its isocentre coincident with the phantom origin. Select a gantry angle of 0o. Verify the BEV-display with respect to the phantom external contours, the cylinders, field dimensions and field position. This test must be repeated for a few different source-surface distances. Set up an asymmetric beam with its isocentre coincident with the phantom origin. For the next set of tests, select a gantry angle of 30o, co-axial and a wedge. Perform tests 4.2.2 4.2.7 for that geometry. Verify the plot of the BEV-display of the phantom.

37

4.2.2 Beam position and shape Verify the beam position and shape (beam axis, divergence lines and aperture) and wedge direction in an axial and coronal plane through the isocentre relative to the phantom external contours and cylinders.

4.2.3 Beam position in BEV a. Verify the beam position in BEV, 3-D-axonometric and 3-D-projection view relative to the phantom external contours and cylinders. b. Verify the agreement of the beam position between the digitally reconstructed radiograph (DRR) and the BEV.

4.2.4 Block position in BEV a. Define a block manually, using BEV field co-ordinates in the beam. Verify the block presentation using the BEV display options. b. Verify the projection on CT slices of blocks inserted by the BEV option. In both cases plot and verify the block definition in BEV.

4.2.5 MLC-shaped field Define an MLC field manually, using BEV co-ordinates, and verify the MLC presentation in the various display types. Repeat the tests 4.2.2, 4.2.3, and 4.2.4. Special attention should be paid to the correctness of the AP/PA, cranial/caudal and left/right indicators in these displays. In addition verify the plotted MLC field.

4.2.6 Bolus position Verify the display of the bolus on the patients surface in BEV, transversal, frontal and sagittal views, and 3-D displays in case the bolus switch is on. Verify whether the bolus is removed if the bolus is switched off. Suggestion: check also the modification of dose distribution as the bolus is switched on and off, globally or for individual beams.

38

4.3 BEAM GEOMETRY Define in a geometric phantom (for instance a water phantom defined in the TPS) a 3-D parallelepiped, conic or spherical target volume and check the following functionality: 4.3.1 Automatic block and auto-leaf positioning Set up a symmetric co-axial beam at 0o gantry angle and define an irregular and/or extreme geometrical (for instance triangle) field shape using a margin defined in BEV. a. Verify the position of the blocks, in an axial and coronal plane. b. Verify the position of the leaves of the MLC in an axial and coronal plane. c. Verify for the MLC, the inner, middle and outer position, or other method (e.g., least squared) for settings of the leaves. d. Verify for the MLC the backup jaw position. 4.3.2 User-defined block Set up a symmetric co-axial beam with gantry- collimator- and table-angle set to 0o and define a block manually using predefined co-ordinates. Verify the correct conversion of these co-ordinates in a BEV-plot at a fixed plot distance and all beam displays mentioned previously.

4.3.3 DRR: linearity and divergence Make CT scans of a phantom with well-defined markers, for instance phantom A, as described in Section 3.2. At least 8 markers should be located on the surface of the phantom. Create DRR images at different SSD (for instance 90 and 100cm). The markers should be visible on the DRR images. Measure the distances between the markers. Compare the distances between the marker positions on the DRR with the actual distances between the markers. (Note: in some TPSs the ruler tool cannot be used in DRR display).

4.3.4 Input, change and edit functions Verify the accuracy of functions as: move beam, oppose beam, z-co-ordinate, shift block, copy block, read block from BEV-file, BEV-beam position, mirror option, BEV lateral/ longitudinal rotation, move of the patient co-ordinate system and the consistency between results of moves at different TPS levels using co-ordinates and distance measurements. If a dose distribution had been previously computed, check that such modifications result in proper cancellation of computations, which are not any longer valid. 39

5. DOSE AND MONITOR UNIT CALCULATION

Methods for checking and verifying the dose calculations in a TPS are described in this chapter. The accuracy of the final result depends not only on the dose calculation algorithm, but also on the quality of the beam data input. Therefore tests of both aspects determining the final result of a dose calculation will be presented in this chapter. It should be noted that these tests should be performed in close co-operation with other users and/or the vendor of the system, to avoid duplication and to understand the observed phenomena. Suggestions for division of tests between vendor and user will be discussed in Appendix 3. The results of the tests described in this chapter should be in agreement with the recommendations on accuracy requirements discussed in Chapter 2. In some systems it is possible to choose the medium to which the dose should be specified, for instance a specific type of reference tissue. Also with Monte Carlo calculations such a choice is possible. In this booklet we will, however, specify all dose values as dose to water. The tests described in this chapter are the starting point of a QA programme of the dose and monitor unit calculation part of a TPS. In many situations a much more extensive series of tests is necessary to verify the dose calculations of the irradiation techniques applied in a specific hospital. On the other hand, some tests may not be very relevant for an institution and can therefore be skipped. It is, however, strongly recommended to start a QA programme of a TPS with the tests and configurations described in this booklet. Any verification of dose calculation requires that some beam data have been previously entered into the TPS. These data could consist of a generic beam data set, generated either from published reports (AAPM 1995, Venselaar and Welleweerd 2001) or from another set of measured reference data (or Monte Carlo computations). The TPS vendor should supply the generic beam data set with proper reference to how it has been obtained. It is meant to be used as a demonstration package to prove the possibilities and the accuracy of dose and monitor unit computation. However, it must never be used for clinical purposes for which a new set of beam data representative of the users beam characteristics must be obtained. To check the accuracy of dose computation in the users beam for a large range of situations representative of the clinical practice, it would be necessary to perform extensive measurements. This turns out to be impracticable and one has therefore to accept the results derived from the generic data set and assume that they can be extrapolated to the users beam data. Alternatively, the Quality Index (QI)-methodology provides a solution easily applicable to those situations where the dose is modified due to some perturbations such as changes in the shape of the patient surface (Caneva et al., 2000) or presence of inhomogeneities. The QI- methodology requires that a correction factor (CF) has been previously obtained from the ratio of a given situation (e.g., with inhomogeneity) to a reference set-up (e.g., a water phantom) for a wide range of beam energies expressed by their Quality Index 41

(ratio of TPR for a fixed source-detector distance at 20 and 10cm, respectively). Provided that the test is properly designed, CFs can be made independent of the type of linear accelerator and will vary smoothly only as a function of QI. Provided this variation is available, it is easy for individual users to perform a dose calculation with their own beam data for both the reference and the modified set-up and to extract CF. This CF value should coincide with the interpolated CF value for the users beam QI. Compared to the use of generic data, the QI-methodology has the advantage that it is performed after beam modelling and is therefore more representative of the data used clinically.

5.1 BEAM CHARACTERISATION SET 5.1.1 Data input process In order to model a photon beam in a TPS, a set of basic beam data, i.e., the beam characterisation set as specified by the vendor, needs to be determined1 and entered into the system by the user or with the help of the vendor. If for one or other reason the users measurement procedure is deviating from the recommended method, it is recommended to discuss this issue with the vendor. Immediately after the beam modelling process the correspondence between calculated and input data should be verified for a limited number of fields, the beam verification set2. The comparison should include depth doses, beam profiles, and if applicable the reference (or calibration) dose per monitor unit. A number of treatment planning systems nowadays have special tools to compare measured data with calculated results. Besides the agreement between measurements and calculations, the internal consistency of the measured dataset should also be verified, because errors might be introduced, for instance if the accelerators are not tuned properly, or if different persons collect data over a prolonged time period. In order to prove the correctness of a specific algorithm, the vendor can use a set of benchmark (generic) data provided by for instance the NCS (Venselaar and Welleweerd, 2001) or by a users group of that system. In this way the vendor can demonstrate the global accuracy of the algorithm, but the user still has to verify the TPS for the specific conditions (slightly different beam fits and photon beam energy) encountered in his/her clinic.

Usually measured beam data, but some data related to machine configuration different from those required for beam characterisation (Chapter 4), may also be required for the TPS to behave accurately. 2 The beam verification set may be identical to the beam characterisation set, but is usually extended and includes several other situations, e.g., extended SSD and rectangular fields.
1

42

5.1.2 Documentation Proper documentation of the characterisation set is needed to reproduce the entire data manipulation process by other users. It should give a detailed description how the data were obtained, implemented and stored, and should include: A complete written description of dates of measurements, names of investigators. Methods of measurement, equipment and software used for data collection and methods used for data manipulation. A copy of the original data (before manipulation). A copy of the final results (printout, plot, digital files, etc.). A description where the data are filed. A logbook in which all events, measurements and changes concerning the TPS are documented.

5.2 DOSE CALCULATION In a TPS, dose distributions are calculated for complicated irradiation techniques applying algorithms that may sometimes be sophisticated, or on the other hand, deliberately be approximate to result in acceptable calculation times. It is important that the user makes sure that the dose calculations are accurate for simple situations such as the following: Square and elongated fields. Different SSD. Beam modifiers: wedges, blocks and trays. Asymmetric collimator settings. MLC-shaped fields.

Therefore a series of measurements should be performed for these conditions (some of them being part of the commissioning of a new accelerator). This is particularly important for the beam modifiers for which accurate results will be obtained only if the proper parameters have been defined in the TPS library. For elongated fields or for different SSD, it could be acceptable to calculate manually from the equivalent square approximation or from the inverse square law, the expected values of calculated doses. The tests described in the following paragraphs are defined to test the limitations of the dose calculation algorithms in combination with tests of beam fits in the treatment planning system. It is not recommended to perform all tests for all energies and all accelerators existing in a radiotherapy department. The tests described in this chapter concern percentage depth dose (PDD), beam profile (PRF) and off-axis factor (OAF) calculations, and should at least be performed at the depth of dose maximum and at 5, 10 and 20cm depth. The depth 43

of 10cm is considered as a reference depth. Generally other parts of the beam are also of interest in modern radiotherapy, for instance if organs at risk (OARs) are involved or for dose-volume histogram (DVH) evaluation. Therefore, performing tests at other depths is strongly recommended. It should be emphasized that the tests proposed in this chapter are not meant as an extensive testing of the algorithms implemented in the system. This is the task of the vendor and the users group of a specific TPS, as discussed in Chapter 2. The tests given in this chapter are meant as a minimum number of tests to be performed by the vendor during the acceptance testing, or by an individual user before applying the system clinically in his (her) institution. The schematic illustrations given for each test are meant to help to set up the geometry, but are not at scale. The dashed line in the centre indicates the middle (0) of the phantom, which coincides with the central beam axis. In some tests the central beam axis is indicated by a dotted line.

5.2.1 Open square fields a Field size 5cm x 5cm, SSD=90cm b Field size 10cm x 10cm, SSD=90cm c Field size 30cm x 30cm, SSD=90cm Calculations in the transverse plane through the central axis of the beam: PDD, PRF.

5.2.2 Open rectangular fields a Field size 20cm x 5cm, SSD=90cm b Field size 5cm x 20cm, SSD=90cm Calculations in the transverse plane through the central axis of the beam: PDD, PRF.

44

5.2.3 Variation in SSD Field size 10cm x 10cm, SSD=100cm and extreme values of SSD for instance 80 cm and 130 cm if possible. Calculations in the transverse plane through the central axis of the beam: PDD, PRF.

5.2.4 Wedged square field Field size 10cm x 10cm, wedge 60, SSD=90cm Calculations in the transverse plane through the central axis of the beam: PDD, PRF (also in the direction perpendicular to the wedge direction).

5.2.5 Wedged rectangular fields a. Field size 5cm x 20cm, wedge 60, SSD=90cm b. Field size 20cm x 5cm, wedge 60, SSD=90cm Calculations in the transverse plane through the central axis of the beam: PDD, PRF (also in the direction perpendicular to the wedge direction).

5.2.6 Field with a central block Field size 15cm x 15cm, central block with clinically relevant dimensions (e.g., as used for spinal cord blocking) clinically, SSD=90cm Calculations in the transverse plane through the central axis of the beam: PRF.

45

5.2.7 Blocked field Field size 30cm x 30cm, block of size: 20cm x 30cm or 10cm x 30cm, SSD=90cm Calculations in the transverse plane through the central axis of the beam: PRF.

5.2.8 Inhomogeneities Lung/air For instance a cylinder with a low-density material. Field size: 10cm x10cm Bone For instance a cylinder with a bone-like material. Field size: 10cm x10cm Calculations in the transverse plane through the central axis of the beam: PDD, PRF. These materials might be available for other purposes, e.g., for QA of CT scanners. In order to verify dose calculations inside a lung, it is recommended to apply a solid phantom having lung-like material in which an ionisation chamber or TLDs can be inserted.

5.2.9 Oblique incidence Field size 10cm x 10cm, open, SSD=90cm, gantry angle 315o Calculations in the transverse plane through the central axis of the beam: PDD, PRF.

46

5.2.10 Missing tissue Field size 10cm x 10cm, open, SSD=90cm, gantry angle 0o Part of the beam is outside the phantom. Calculations in the transverse plane through the central axis of the beam: PDD, PRF.

5.2.11 Off-axis square field Field size 10cm x 10cm, open, SSD=90cm, off-axis 5cm Calculations in the transverse plane through the central axis of the beam: PRF, OAF.

5.2.12 Off-axis elongated field Field size 2cm x 20cm, open, SSD=90cm, off-axis 10cm Calculations in the transverse plane through the central axis of the beam: PRF, OAF.

5.2.13 Wedged off-axis field Field size 10cm x 10cm, wedged, SSD=90cm, off-axis 5cm Calculations in the transverse plane through the central axis of the beam: PRF, OAF.

47

5.2.14 Off-plane field Field size 10cm x 10cm, open, SSD=90cm, off plane 5cm (i.e., off-axis in both directions) Calculations in the transverse plane through the central axis of the beam: PRF, OAF.

5.2.15 Square MLC field Field size 10cm x 10cm, open, SSD=90cm Calculations in the transverse plane through the central axis of the beam: PDD, PRF.

5.2.16 Off-axis square MLC field Field size 10cm x 10cm, open, SSD=90cm, off-axis 5cm Calculations in the transverse plane through the central axis of the beam: PDD, PRF.

48

5.2.17 MLC-shaped field

Banana shape field enclosed only by leaves while the backup jaws are open far as possible, SSD=90cm Calculations in the transverse plane through the central axis of the beam: PDD, PRF. Several options for this test are suggested: - to do a scan in the middle of the leaves - to do a scan at the border of two leaves - to do a scan perpendicular to the leaf direction

5.2.18 Block and tray insertion a. The aim of this test is to check if there is an influence on the dose calculation using different methods of block digitising. The same tests can be used for MLC-shaped fields. Calculations in the transverse plane through the central axis of the beam: PDD, PRF. The following steps are suggested: 1. Create an open beam with a field size of 15cm x 15cm (or 20cm x 20cm) perpendicular to a phantom surface.

2. Add to the beam described above four blocks. The blocks are digitised separately as indicated in Figure 5.1. ( Start of the drawing of the blocks)

Figure 5.1 Four blocks digitised separately.

49

3. Design the same arrangement of blocks, for the same field, but now digitise the blocks as a single block described by two loops in the same direction (Figure 5.2). ( Start of the drawing of the blocks).

Figure 5.2 Four blocks digitised as a single block described by two loops in the same direction.

4. Design the same arrangement of blocks, for the same field, but now digitise the blocks as a single block described by two loops in the opposite direction (Figure 5.3). ( Start of the drawing of the blocks).

Figure 5.3 Four blocks digitised as a single block described by two loops in the opposite direction.

Compare the central axis depth dose distributions and beam profiles at various depths for the four cases described above. b. An additional test can be performed to check whether the value of the block transmission influences the depth dose distribution. Calculate the central axis depth dose distribution for geometries and blocks of different transmission, e.g., for 5, 50 and 90%. Compare the results for: - Field size 15 cm x 15 cm, open - Field size 15 cm x 15 cm, with a small corner block - Field size 15 cm x 15 cm, with one block - Field size 15 cm x 15 cm, with two blocks - Field size 15 cm x 15 cm, with four blocks.

50

a. Field with a small corner block

b. Field with one block

c. Field with two blocks

d. Field with four blocks

Figure 5.4 Position of blocks to check whether the value of the block transmission influences the depth dose distribution.

c. Verify if the correct tray factor is entered in the TPS. This is of particular concern if different types of trays are used in a department.

5.3 2-D AND 3-D DOSE VERIFICATION Besides the verification of the dose at specific points or along lines (i.e., depth dose curves or beam profiles), it is essential that dose values are verified over the whole volume of interest. Comparing calculated with measured dose distributions in several relevant planes can do this. The position of isodose lines should also be checked in these planes. Verification of dose-volume histograms is another way of verifying the 3-D dose distribution but should include verification of the volumes under investigation. 51

5.3.1 2-D dose distribution For a verification of 2-D dose distributions define the following beams: open square fields 5cm x 5cm, 10cm x 10cm and 30cm x 30cm (see 5.2.1) open rectangular field 5cm x 20cm (see 5.2.2), wedged square field 10cm x 10cm (see 5.2.4), blocked field 30cm x 30cm (see 5.2.7), MLC-shaped field (see 5.2.17). Tests should be performed at SSD = 90cm, depth = 10cm. As a minimum, 95%, 90%, 80%, 50% and 20% isodoses should be measured in water phantom using the tracking or reconstruction from multiple profiles option. 2-D dose distributions can also be measured by means of film dosimetry in a water or solid phantom. In this case the procedures for accurate film analysis, i.e., calibration and conversion from optical density to dose, should be carefully established. In order to analyse the results of the tests, as a minimum, manual matching of the two isodose distributions (measured and calculated) should be performed and maximum deviations between isodoses should be reported in mm. If computerized matching is used, a function that combines dose differences and distance discrepancies (in regions with a large dose gradient), the gamma evaluation method as discussed in Section 2.2.3, can be used. If -value analysis can be performed, it is recommended to apply the criteria presented in Table 2.1. Mean and maximum -values should be reported together with the fraction of points showing a -value larger than unity (using gamma-area histograms). An example of such a test is given in Chapter 7. It should be noted that the determination of interpolated isodoses, reconstructed from a series of measurements using the scanning water phantom software, also introduces uncertainties. For that reason some companies recommend a profile-by-profile comparison to determine measured versus calculated accuracy by comparing the values of the calculated profiles at the intersection of the isodoses to the displayed isodoses at those intersections.

5.3.2 Dose-volume histogram Several tests have been described to assure that the 3-D dose distribution over a structure is accurately binned into a DVH. Craig et al., (1999) applied a manual verification of a cumulative DVH of a dose distribution of a wedged beam over a cubic volume. Multiplying the area defined by the intersection of the volume and an isodose line by the axial length of the volume yields the volume equal to or greater than the isodose line. In the NCS Report (NCS 2004) various tests are described using either a user-defined dose distribution or a single beam. In a simple test the DVH is calculated in a rectangular structure along the beam axis, with a length and width of 1cm x 1cm starting at the depth of dose maximum and a height of approximately 20cm, irradiated with a photon beam. The DVH can then be com52

pared with that calculated from the corresponding percentage depth dose table. Not only tests for cube-like structures are given, but also for a sphere, which is less sensitive to grid-based artefacts. The tests related to DVH presented in this booklet consist in checking the consistency between the 3-D dose distributions displayed as isodose lines and assumed to be the reference, and the DVH (Panitsa et al., 1998). Since DVH computations are highly sensitive to parameters such as density of sampling points (including slice spacing) or size of the dose bins, the tests must be performed after setting these parameters at values close to the values used in clinical practice. It is therefore assumed that the construction of volumes by the TPS has already been tested in the way described in Section 3.3.5. Low dose gradient: Construct a homogeneous cubic virtual phantom (e.g., 25 x 25 x 25 cm ) and include a central 5 x 5 x 5 cm3 cubic structure of the same density. Use a 10cm x 10cm beam normal to the top phantom surface, with its isocentre at the centre of the cubic structure. Display isodose lines in a plane through the beam axis and in planes perpendicular to the beam axis through top and bottom part of the structure (Figure 5.5 a). Determine the dose at the centre and at each corner of the (square) section of the structure in each plane. On the other hand, compute the DVH in the structure. The computed volume should be very close to 125 cm3 (depending on the slice spacing) and the minimum and maximum dose of the DVH should be consistent with those found in the computed planes (maximum in top plane, minimum in bottom plane).
3

High dose gradient: Create a beam opposed to the previous one and shift the two beams laterally by 5 cm in order to have the isocentre at the beam edge (Figure 5.5 b). Measure in the axial plane through the beam axis the distance d (in cm) between the 80% and 20% isodose line (normalised at beam isocentre). Check the dose uniformity along the beam edges using a sagittal plane through the centre of the cubic structure. Compute the DVH in the structure. The DVH should be linear between the 20 and 80% dose values and the volume difference between these two dose levels should be equal to 25 x d (cm3).

Figure 5.5 a) Example of the geometry for a DVH test in a low dose gradient area; b) example of the geometry for a DVH test in a high dose gradient area.

53

5.4 MONITOR UNIT CALCULATION The calculation of the number of monitor units for the individual fields of a treatment plan is included in most modern TPSs. It is essential for the user of a TPS to understand the principles of the MU calculation algorithm. Generally the MU calculation is closely linked to the calculation of the relative dose distribution. In some systems the dose distribution is directly calculated as dose per MU, or dose per fluence, from which a representation of relative dose and number of MUs is derived. As outlined in Section 2.2, tests of the dose calculation presented in this booklet concern the verification of the absolute dose, i.e., the dose per MU, after correction for any variation in output of the accelerator. Following this concept, the number of monitor units for a certain geometry is implicitly also checked. In many institutions a check of the number of MUs independent of the MU calculation provided by the TPS is performed. Such a procedure can be considered as an essential part of routine clinical QA programme, and several software packages are commercially available for this purpose. In ESTRO Booklet No. 3 (Dutreix et al., 1997) and Booklet No. 6 (Mijnheer et al., 2001) the formalism as well as numerical data are provided for such an independent MU calculation method. In these documents examples can be found how to perform such type of independent MU calculation.

54

6. PERIODIC QUALITY CONTROL

The sources of error and inaccuracies in the functioning of computerised treatment planning systems may be divided into several groups: the hardware, the data base of basic dosimetric input data for radiation beams, the geometric input data for treatment machines, the system of acquiring anatomical patient data, the dose calculation algorithms used in a particular system, the patient data base and the archiving of the results of a treatment planning exercise. All elements require thorough testing but some tests may be performed only once and some should be repeated periodically. Periodic quality control of particular parts of a TPS is an important aspect of the quality assurance process of a TPS for safety and security reasons. This chapter consists of tests that should be performed periodically, at regular time intervals. If a TPS is able to perform a checksum test of the data file in it, such a test should be performed as often as desired (Section 6.2). The need for testing separate parts of the software is then less urgent. If it not possible to do a checksum test, it depends strongly on the local situation which frequency should be applied to perform a specific test. For that reason no general recommendations are formulated for the frequency of the tests described in this chapter. Some tests related to proper functionality of peripheral devices used for data input, software and output devices are also presented. In some countries guidelines are available for periodic QC of a TPS, or are even obligatory to follow for legal reasons. The tests described in this chapter are not meant to replace these national guidelines, but intended as an addition to these.

6.1 DATA INPUT PROCESS The correct functionality of all devices related to the data input process and network connections should be checked. Depending on the specific TPS configuration, testing of different devices needs to be performed. Some of the periodic tests proposed in this section are described already in Chapter 3: Anatomical Description.

6.1.1 Digitiser A simple geometrical figure like a square, rectangle or a triangle, should be entered via a digitiser and then printed back on a printer or drawn on a plotter. The input and output figures should be compared and the accuracy of the input-output devices assessed. Recommended reproducibility: 1 mm. Alternatively perform test 3.2.2 a. The vertical and horizontal scales are to be tested along 10 cm showing an agreement better than 1mm. 55

6.1.2 Film scanner Objects on a film like square, rectangle, and triangle with specific dimensions are scanned. The size displayed by the TPS is then compared with the real one. Recommended reproducibility: 1mm. Alternatively perform test 3.2.2 b.

6.1.3 CT data a. Scan a phantom with well-known outer contours and with known electron densities, e.g., phantom B from Chapter 3. Enter the data in the TPS. Compare the geometries and electron densities with the actual data. Recommended reproducibility: - geometry: 1mm - if electron density < 1.5 (relative to water): 0.05 - if electron density > 1.5 (relative to water): 0.1 b. Geometric integrity of slices. Scan a phantom of well-defined dimensions and check its dimensions in the TPS using a central field-of-view (e.g., phantom A). Recommended reproducibility: 1mm.

6.1.4 MR data Scan a phantom with well-known outer dimensions including some inhomogeneities, e.g., the phantoms described in Chapter 3. Enter the data in the TPS. Compare the geometry with the actual data. Recommended reproducibility: 2mm.

6.1.5 Integrity of simultaneous input Check that simultaneous input (digitiser, film scanner, CT, MR) by different users on these channels does not interfere in the TPS. See test 3.2.1 f.

6.2 SOFTWARE Most modern treatment planning systems are able to perform a checksum test. Such a checksum program compares current executable files with reference files created at the time of commissioning the TPS, and periodically updated files. Performing a checksum test is therefore one of the most powerful and efficient tests of the software of a TPS. Tests of 56

TPS software include tracing problems related to dose calculation. As part of a periodic quality control programme, verification of treatment plans of standard techniques is therefore recommended in some institutions even if a checksum test is available. It should be noted that the operating system is taking care of some of the tests presented in this section.

6.2.1 MU calculation Check the MU calculation for some geometries presented in Chapter 5: 5.2.11 5.2.12 5.2.13 5.2.14 Off-axis square field Off-axis elongated field Wedged off-axis field Off-plane field

6.2.2 Standard treatment techniques Check the dose distributions for some standard treatment techniques: a. Head & neck 2 fields (opposing lateral or oblique, wedged fields; bolus) b. Lung/oesophagus 2 or 3 fields (oblique A-P, wedged, open fields) c. Breast 2 fields (tangential, wedged fields) d. Pelvis/prostate 3-field or box technique (oblique, wedged, open fields).

6.2.3 MLC-shaped field Check the MLC-shaped field as described in test 5.2.17.

6.3 DATA OUTPUT PROCESS The following tests are related, or similar to, tests described elsewhere in this booklet.

6.3.1 Printing/plotting devices a. On each treatment plan, either printed or plotted, two scales are required in both directions (horizontal and vertical) with centimeter marks at least over 10cm. However, all the marks shall be verified with a ruler; they must be at the right place (i.e., no distortion and 57

no displacement). A test plan for a phantom of well-known geometry can be made (e.g., phantom B from Chapter 3). The phantom should be digitised, verified on the screen, then plotted and verified on paper. Recommended reproducibility: 1mm. b. Print or draw on a plotter the geometries from test 6.1.1. The input and output figures should be compared and the accuracy of the input-output devices assessed. Recommended reproducibility: 1 mm. c. Plot and verify the BEV display from test 4.2.1. Recommended reproducibility: 1mm. d. Plot isodose distributions of at least 3 beam geometries and compare them with the display on the screen. See tests 5.2 and 5.3. Recommended reproducibility: 1mm.

6.3.2 Block cutting device The correct transfer of information from TPS to the block-cutting device should be tested. The most clinically relevant shapes of block and compensators like square, rectangle, and triangle should be sent to the block-cutting device and their size should be examined. Recommended reproducibility: no difference.

6.3.3 Treatment plan transfer The correct transfer of information from a TPS to a linear accelerator and other hardware, like a record or verify system, needs to be tested. Use plans/fields from tests 6.2.2 and 6.2.3 to check if all information was sent correctly.

58

7. EXAMPLES OF TESTS
In this chapter results will be presented of examples of tests described in Chapters 3 to 5. Although the details may differ from those given in these chapters, these examples show how these tests can be performed in practice, and can be analysed when comparing calculations from a TPS with measurements. Different members of the QUASIMODO group, having different TPSs and using different measuring equipment, have recently performed these tests. These examples are therefore representative for the accuracy of state of the art (non-IMRT) treatment planning systems. They illustrate the enormous possibilities of a modern 3-D TPS, but also the limitations of the algorithms of some of these systems. The examples of tests given in this chapter have the same number as used in the corresponding chapter, where also a more complete description of the test can be found. Although many more examples could be given, we have chosen from each chapter a limited number of representative tests, which may in addition have some interesting features.

7.1 TESTS FROM CHAPTER 3: ANATOMICAL DESCRIPTION Test 3.2.1 Image input a: Identity consistency of scans

Different fields-of-view within one CT data set may give wrong dimensions of the phantom/patient. A CT data set with slices of two different fields-of-view was introduced into a TPS. The TPS has given no warning. The dimensions of the phantom were measured with the ruler tool from the TPS. Figure 7.1 shows that due to the changes in the field-ofview of slices z=2 and z=3, the system recognizes wrong dimensions of the phantom in slice z =3.

Figure 7.1 CT data of phantom A with slices having different fields-of-view.

59

Test 3.2.1 Image input

b: Scan parameters varying slice thickness

System A: A set of CT scans with variable slice thickness of 4 and 10mm was transferred to the TPS. The system does not give any warning or comment. The construction of the volume was done correctly. System B: A set of CT scans with variable slice thickness of 2 and 5mm was transferred to the TPS. The system does not give any warning or comment. The construction of the volume was done correctly.

Test 3.2.1 Image input

i: CT number representation

CT data of a cylindrical phantom (AAPM CT Performance Phantom) were imported in a TPS. CT numbers in the TPS, obtained by making profiles, and determining the maximum, minimum, and mean CT number in a small region of interest (ROI), were compared with the original numbers measured on CT. In the example shown in Figure 7.2 good correspondence was observed.

Figure 7.2 CT scan of a cylindrical phantom (AAPM CT Performance Phantom) with the number of Hounsfield Units (HU) in a region of interest (ROI) represented by a TPS. The pixel statistics are: min. 127.0, max. 157.0, mean 139.3, and a standard deviation of 5.5 HU. The number at the CT workstation is 134.

60

Test 3.2.3 Image use

a: Geometry of reconstructed images

CT data of a cylindrical phantom, having a number of parallel inserts, were imported in a TPS. The distance between these inserts was measured in reconstructed transversal, sagittal, and coronal planes. The distance between the inserts is maintained in the reconstructed images shown in Figure 7.3.

Figure 7.3 Reconstructed transversal, sagittal and coronal planes of CT scans of a cylindrical phantom, as represented by a TPS. Indicated are corresponding distances between inserts.

Test 3.2.4 Co-ordinate system of images A schematic view of the phantom used for this test is shown in Figure 7.4. This phantom is made for use in combination with a stereotactic localiser for stereotactic imaging. It is particularly useful for CT/MR image-modality tests. The PMMA cylinders contain small empty cylinders filled with a liquid to be used as markers (also referred to as target points). Iodine contrast agent was used as liquid for CT as well as for MR. The five target points shown in Table 7.1 were used for the comparison of the co-ordinate values. The origin of the co-ordinate system was that of a stereotactic frame to which the phantom was attached.

Table 7.1 Given co-ordinate values in mm with respect to the origin of a stereotactic frame. target point # 1 2 3 4 5 X 60.0 0.0 -60.0 0.0 0.0 Y 0.0 60.0 0.0 -60.0 0.0 Z 44.1 74.1 104.1 134.1 164.1

61

Figure 7.4 View of the phantom with markers visible with CT and MR imaging.

The stereotactic localiser was attached to the phantom, and CT as well as MR imaging was performed. With CT imaging, three different settings were used: a) a special scan for CT angiography with 2-mm slice thickness, b) the same scan modus with 3-mm slice thickness, and c) a so-called head-scan with 3-mm slice thickness. With MR imaging, two different settings were used, one for T1-weighted imaging and one for T2-weighted imaging. Two different MR systems (MR1 and MR2) were examined. The determination of the co-ordinate values from the different images was performed using the stereotactic tool of a treatment planning system. When calculating the vectorial difference, Dr, between the given position of a marker point, and the corresponding position obtained by the imaging procedure and stereotactic co-ordinate determination, a value of 2mm must not be exceeded. The following results were obtained: CT Imaging
Table 7.2 Results of a special scan for CT angiography with 2-mm slice thickness. Target point # 1 2 3 4 5 X 61.16 0.77 -60.50 0.32 0.59 Y 0.46 60.90 0.40 -60.20 0.75 Z 44.28 73.88 104.29 134.94 164.94 r 1.26 1.20 0.67 0.92 1.27

62

Table 7.3 Results of a special scan for CT angiography with 3-mm slice thickness. Target point # 1 2 3 4 5 X 61.16 0.51 -60.22 0.32 0.59 Y 0.18 60.62 0.40 -60.20 0.47 Z 44.18 74.46 104.85 134.67 164.68 r 1.18 0.88 0.88 0.68 0.95

Table 7.4 Results of a head-scan with 3-mm slice thickness. Target point # 1 2 3 4 5 X 61.15 0.77 -60.52 0.28 0.82 Y 0.20 60.64 0.42 -60.18 0.78 Z 43.99 74.25 104.63 134.41 164.37 r 1.17 1.01 0.85 0.45 1.16

MR Imaging
Table 7.5 Results for MR1 of a T1-weighted scan. Target point # 1 2 3 4 5 X 59.80 -0.15 -60.96 -0.74 -0.63 Y -0.43 59.51 -0.03 -60.15 -0.17 Z 44.34 74.44 104.77 133.55 162.64 r 0.53 0.61 1.17 0.93 1.60

Table 7.6 Results for MR1 of a T2-weighted scan. Target point # 1 2 3 4 5 X 60.25 0.17 -60.86 -0.72 -0.75 Y -0.48 59.62 -0.13 -60.23 -0.11 Z 44.73 74.22 104.19 133.70 162.69 r 0.83 0.43 0.87 0.86 1.60

63

Table 7.7 Results for MR2 of a T1-weighted scan. Target point # 1 2 3 4 5 X 60.81 0.70 -60.14 0.05 0.40 Y -0.82 59.23 -0.70 -61.14 -0.96 Z 43.96 74.55 103.68 135.55 165.35 r 1.16 1.13 0.83 1.85 1.63

Table 7.8 Results for MR2 of a T2-weighted scan. Target point # 1 2 3 4 5 X Y Z r

0.83 -60.20 0.03 0.37

59.17 -0.64 -61.23 -1.25

74.24 104.90 135.84 166.46

1.18 1.04 2.13 2.70

The vectorial distance r, as shown in the last column of each table, was on average 1.0mm and generally within the tolerance value of 2.0mm for any CT modality. The same imaging accuracy is achieved with MR1. However, MR2 (which is a more recent MR system) shows a significantly worse accuracy, with tolerance excess for two marker positions. The differences are not related to the TPS, in fact use is made of the measuring tool of the TPS, but due to the non-optimal image quality of the second MR scanner. A note must be added at this point: for new MR systems there is a tendency to reduce the length of the magnet coil, which supplies the basic magnetic field. This may have advantages for the diagnostic procedure; however, it has a clear disadvantage in providing images without distortion as needed for high-precision radiotherapy.

Test 3.3.5 Construction of volumes

b: Construction of a volume from a set of CT slices with non-regular spacing

In the original CT data set the space between a set of slices is 1mm. The TPS interpolates between slices to create a volume. For various non-regular slice spacing, the surface of an object was constructed and the resulting volume calculated. The results show that the volume was calculated correctly in all cases.

64

Table 7.9 Determination of the volume of an object for a set of CT slices with non-regular spacing. Non-regular 1/3 means that the space between the slices was either 1 or 3mm. Spacing [mm] 1 2 3 5 Non-regular 1/3 Non-regular 2/5 Non-regular 2/3 Non-regular 1/5 Number of contours 61 31 21 13 41 21 28 41 Volume [cm3] 73.86 73.76 73.69 73.68 73.82 73.71 73.75 73.78

7.2 TESTS FROM CHAPTER 4: BEAM DESCRIPTION Test 4.1.1 SAD, SSD and field size a: Field size

The size of various fields was measured at isocentre level in transversal and sagittal planes using the ruler tool of the TPS and compared with the TPS beam co-ordinates. The results shown in Table 7.10 demonstrate that the measurements are in good agreement with the stated field sizes. The accuracy depends on the magnification factor used for the measurements.
Table 7.10 Comparison of stated field size with field lengths measured in the TPS in transversal and sagittal planes. Field size X x Y [cm] 3x5 10 x 20 30 x 10 Transversal / sagittal [cm] 3.01 / 4.99 10.001 / 20.02 29.96 / 10.01

Test 4.1.3 Collimator rotation For an asymmetric (4cm in the X-direction and 2cm in the Y-direction) field of 10cm x 12cm, a collimator rotation of 60o was performed. The length of the projection of the border of the field on a phantom was measured in three planes in a coronal view, and in the BEV option of the TPS. The measured distances were compared with the actual dimensions and are shown in Table 7.11.

65

Table 7.11 Results of measurements in a coronal view of a field projected on a phantom after performing a collimator rotation. No differences were observed. Position Z= -2.0 Z= 0 Z= 2.5 Actual dimension [cm] 12.93 8.00 2.54 Measurements [cm] 12.92 8.02 2.54 BEV [cm] 12.92 8.02 2.54

Test 4.1.9 Warnings and error messages The adequacy of warnings and error messages following beam input with values that violate gantry, collimator and table angle description, was checked for two TPSs. System A: Gantry, collimator and table rotation is possible over the full range of 0-360. After typing a negative value of the gantry angle ( 35), the system does not give a warning. The angle will be 325. System B: Gantry, collimator and table rotation is possible over the full range of 0-360. After typing a negative value of the gantry angle ( 35), the system gives a warning and asks for an angle in the range 0 - 360.

Test 4.2.1 Beams-eye-view (BEV)-display The BEV-display was checked for a field of 10cm x 10cm irradiating a phantom in which two inhomogeneous structures, consisting of lung-equivalent and bone-equivalent material, are positioned. The isocentre is coinciding with the origin of the phantom. The centre of the bone-inhomogeneity lies in the isocentre plane, while the centre of the lung-inhomogeneity is at 92.7cm SAD. The BEV dimensions refer to 100cm SAD. Figure 7.5 shows that the distance between the centre of the field and the centre of the bone-inhomogeneity in the CT slice and the BEV agrees very well. The distance between the centre of the field and the centre and distal border of the lung-inhomogeneity in the CT slice and the BEV, correspond also very well after correction for beam divergence.

66

Figure 7.5 CT slice and BEV of an inhomogeneous phantom irradiated by a 10cm x 10cm field. Indicated in both displays are the distances between the centre of the field and the bone- and lung-inhomogeneity.

Test 4.2.6 Bolus position In a TPS the constant thickness tool was used to define a bolus. A bolus layer of height 15cm, width 10cm and thickness 2cm was generated on a curved surface. The display of the bolus on the phantom surface was performed for a field length of 10cm. In Figure 7.6 the drawing of the bolus on the phantom surface, as well as the results of measurements of its thickness at several places inside the beam, are presented. The covered surface depends on the height and width values of the bolus, which can be defined by the user (the width by choosing a cutoff depth in the tested system). In this system there is no option to create a bolus layer by hand. Only automatically generated bolus layers, without manual modification, are possible. A density value can be chosen but no grey 67

values are visible in CT scans. The bolus is visible after performing the dose calculations and is not separated from the body contour of the patient.

Figure 7.6 Drawing of a bolus with a predefined thickness of 2cm on a curved phantom surface, as well as the results of measurements of the thickness of the bolus at several places inside the beam.

7.3 TESTS FROM CHAPTER 5: DOSE AND MONITOR UNIT CALCULATION Test 5.2.1 Test 5.2.2 Test 5.2.4 Test 5.2.5 Open square fields (absolute dose) Open rectangular fields (absolute dose) Wedged square fields (absolute dose) Wedged rectangular fields (absolute dose)

68

In Table 7.12 results from a point dose study are shown, in which 100 monitor units are given, and the absorbed dose is measured with an ionisation chamber along the central beam axis. The same geometry is set up in the TPS and the number of monitor units for an absorbed dose of 1.00 Gy is calculated. The last column gives the ratios d(i) and d%(i) between calculated and measured dose per monitor unit values at point i. Note that deviations in bold indicate values outside the tolerance level of 2% and 3% for the open and wedged field, respectively, as given in Table 2.1, although these values cannot directly be applied to single point measurements.

Table 7.12 Data from point measurements using an ionisation chamber positioned at 20cm depth along the central beam axis in a large water phantom, source-skin distance 90cm, irradiated with a beam of 18 MV x-rays. Measured Field size (XxY cm2) Type 5x5 Open 60 Wedge Open 60 Wedge Open 60 Wedge Open 60 Wedge Open 60 Wedge Open 60 Wedge Dose (Gy) 0.557 0.152 0.614 0.173 0.673 0.197 0.694 0.206 0.597 0.167 0.589 0.164 (Dose/MU) meas 0.00557 0.00152 0.00614 0.00173 0.00673 0.00197 0.00694 0.00206 0.00597 0.00167 0.00589 0.00164 Calculated Dose (Gy) MU 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 184.78 677.73 165.46 579.46 149.80 527.66 145.35 509.46 168.16 593.02 172.77 622.97 (Dose/MU) calc d(i) /d%(i) 0.00541 0.00148 0.00604 0.00173 0.00668 0.00189 0.00688 0.00196 0.00595 0.00169 0.00579 0.00162 0.971 / -2.9% 0.969 / -3.1% 0.984 / -1.6% 1.000 / 0.0% 0.993 / -0.7% 0.964 / -3.6% 0.991 / -0.9% 0.951 / -4.9% 0.997 / -0.3% 1.010 / +1.0% 0.983 / -1.7% 0.979 / -2.1%

MU 100 100 100 100 100 100 100 100 100 100 100 100

10x10

20x20

30x30

5x20

20x5

Test 5.2.1 Open square fields (profile) Figure 7.7 shows the result of a comparison between calculations and measurements of a beam profile of an open square field. Inside the high dose region the difference varies between 1.6% and + 2.7%, the average deviation is +1.0% with a standard deviation of 1.2%. The deviations, summarized as a confidence level of 2.7%, are within the recommended tolerance level of 3% as given in Table 2.1. 69

Figure 7.7 Comparison between calculation () and measurement (-) of a profile of a 30cm x 30cm field at 10cm depth.

Test 5.2.1 Open square fields (PDD) In the following two figures results from depth dose scans are compared with dose calculations in a TPS. To facilitate the normalisation process all scans are obtained without any change in electrometer setting of the scanning system. The reference channel of the electrometer is connected directly to the monitor ionisation chamber of the linac. The scan for a field size of 10cm x10cm is included in the set of measurements to get the data normalised to the absolute dose per monitor unit at 10cm depth. In the TPS, the dose along the central beam axis has been exported for a calculation in a water phantom. The dose along the lines is divided by the number of monitor units calculated by the TPS for each field. Figure 7.8 shows the measured and calculated depth dose curves, while in Figure 7.9 the deviations d(i) between calculated and measured dose per monitor unit values have been calculated on a point by point basis.

Figure 7.8 Depth dose curves of a 6 MV beam for four field sizes, obtained by scanning in a water tank, normalised to the output (measured with an ionisation chamber) at 10cm depth of the 10cm x 10cm field. Red lines are measured data and black lines calculations performed by a TPS.

70

Figure 7.9 Ratio, d(i), between calculated and measured dose per monitor unit values, for the 6 MV depth dose data presented in Figure 7.8.

Table 7.13 Statistical evaluation of the deviations between calculated and measured data of the four 6 MV depth dose curves presented in Figures 7.8 and 7.9. The deviations are expressed as d%(i): 100.[d(i) 1]. The confidence limit is calculated as the absolute value of the average deviation plus 1.5 times the standard deviation. Build-up region (0-2 cm) Average deviation (%) Standard deviation (%) Confidence limit (%) Remaining curve (2-25 cm) Average deviation (%) Standard deviation (%) Confidence limit (%) 5x5 cm2 - 0.1 0.3 0.5 10x10 cm2 0.2 0.3 0.7 15x15 cm2 0.1 0.3 0.6 20x20 cm2 - 0.6 0.5 1.4 Field size 5x5 cm2 3.7 6.2 13.0 Field size 10x10 cm2 0.7 2.1 3.9 Field size 15x15 cm2 - 1.3 2.4 5.0 Field size 20x20 cm2 - 3.7 5.1 11.3

Using the concept of Venselaar and Welleweerd (2001), the confidence limits have been determined for these four sets of depth dose data. Both a high dose gradient (build-up) region from 0-2 cm and a low dose gradient region from 2-25 cm have been analysed. The results are given in Table 7.13. Note that the confidence limits for the 5x5 cm2 and 20x20 cm2 do not fulfil the recommended 10% accuracy requirement of dose calculations of a TPS in the build-up region as presented in Table 2.1, and the distance-to-agreement criterion should therefore also be investigated. It should be noted that part of this discrepancy might also be caused by uncertainties in the measurement procedure, which is not trivial in the build-up region. 71

Test 5.2.4 Wedged square field (profile) This test, as well as the analysis of the results, is similar to the former test (5.2.1) but performed in another institution having a different TPS, another accelerator and other measuring equipment.

Table 7.14 Statistical evaluation of the deviations between calculated and measured data of the four 6 MV dose profiles presented in Figures 7.10 and 7.11. The deviations are expressed as d%(i): 100.[d (i) 1]. The confidence limit is calculated as the absolute value of the average deviation plus 1.5 times the standard deviation. Note that the confidence limit for the points outside the field does not fulfil the recommended 4% accuracy requirement of dose calculations of a TPS in this region, as presented in Table 2.1. The points inside the field and in the penumbra are within the tolerance levels of 3% and 10%, respectively. Inside field Average deviation (%) Standard deviation (%) Confidence limit (%) -0.4 0.7 1.5 Penumbra 0.4 5.4 8.4 Outside field 2.0 4.0 8.0

Figure 7.10 Beam profiles of a 20cm x 20cm 6 MV field with a 600 wedge, obtained by scanning in a water tank at five different depths, normalized to the output measured with a calibrated ionisation chamber in a 10cm x 10cm field. The solid lines are the absolute dose profiles calculated by the TPS for 100 MUs.

72

Figure 7.11 Difference between calculated and measured dose per monitor unit values, for the beam profiles presented in Figure 7.10. The error was calculated relative to the local dose for the points within the field, and relative to the dose on the central beam axis at the depth of measurement for points outside the field.

Test 5.2.6 Field with a central block

Figure 7.12 Beam profiles of a 16cm x 16cm 6 MV field with a central beam block, obtained by scanning in a water tank at five different depths, normalized to the output measured with a calibrated ionisation chamber in the 10cm x 10cm field. The solid lines are the absolute dose profiles calculated by the TPS for 100 MUs.

73

Figure 7.13 Difference between calculated and measured dose per monitor unit values, for the beam profiles presented in Figure 7.12. The error was calculated relative to the local dose for the points within the field, and relative to the dose in the open part of the beam at the depth of measurement for points under the block and outside the field.

Table 7.15 Statistical evaluation of the deviations between calculated and measured data of the four 6 MV dose profiles presented in Figures 7.12 and 7.13. The confidence limit is calculated as the absolute value of the average deviation plus 1.5 times the standard deviation. The points under the block, in the penumbra, inside and outside the field are all within the tolerance levels of 4%, 10%, 3% and 4%, respectively, as presented in Table 2.1. Central area (under block) Average deviation (%) Standard deviation (%) Confidence limit (%) 1.3 0.6 2.2 Penumbra (block) 1.4 2.9 5.7 Inside field Penumbra Outside field (open field) 1.1 4.8 8.3 2.5 0.6 3.5

-0.3 0.6 1.1

74

Test 5.2.11 Off-axis square field (OAF)

Figure 7.14 Comparison between calculation () and measurement (---) of off-axis factors for a 10cm x 10cm 18 MV beam at 10cm depth for various off-axis distances. The difference is a result of a nonperfect beam fit.

Test 5.2.11 Off-axis square field (OAF)

Figure 7.15 Comparison between calculation () and measurement (---) of off-axis factors (OAF) for a 3cm x 3cm 18MV beam, for various off-axis distances (OAD) at depths of 3, 10 and 20cm. The difference is a result of a non-perfect beam fit and limitations of the dose calculation algorithm.

Test 5.2.13 Wedged off-axis field (absolute dose) Indicated is the difference between calculated and measured dose values for offaxis fields. A: 6MV, field size 9cm x 9cm, 30o wedge: SSD = 93cm, depth 7cm, 3cm off-axis (in the wedge direction): = 3.1% B: 18 MV, field size 10cm x 10cm, 60o wedge: SSD = 90cm, depth 3cm, 2.5 cm off-axis (perpendicular to the wedge direction): = 3.3%

75

Test 5.2.13 Wedged off-axis field (OAF)

Figure 7.16 Comparison between calculation () and measurement (---) of off-axis factors (OAF) for a 3cm x 3cm 18MV beam with a 60o wedge, for different off-axis distances (OAD) at depths of 3, 10 and 20cm. The difference is a result of a non-perfect beam fit and limitations of the dose calculation algorithm.

Test 5.3.2 Dose-Volume Histogram A mathematically defined (virtual) phantom of size 25cm x 25cm x 26cm, and an internal structure of 5cm x 5cm x 6cm were constructed. A 10cm x10cm beam with energy of 18 MV was used for this test (Figure 7.17). The geometry was taken from the test description. A dose of 200 cGy was delivered to the isocentre, which is located in the middle of the internal structure/virtual phantom. Low dose gradient area Figure 7.18 shows the DVH of the internal structure in the centre of the beam, i.e., the low dose gradient area. The computed volume is 150.06 cm3 and is in agreement with the expected value of 150 cm3. Figures 7.19 and 7.20 show the dose profiles along a diagonal computed in the top and bottom plane of the internal structure. The maximum and minimum dose values of the DVH are in agreement with the corresponding values at the centre in the top plane and in the corner of the bottom plane, 223.7 and 174.2 cGy, respectively. A dose grid size of 2 mm and 4 mm were used for this analysis. For the low dose gradient area the DVH calculation does not present any significant discrepancy. The dose grid size has no influence on these calculations.

76

(a)

(b)

Figure 7.17 Virtual phantom, internal structure and beam setting defined in the system for the DVH test; (a) low dose gradient area test; (b) high dose gradient area test.

Figure 7.18 DVH of the internal structure in the low dose gradient area for two dose grid sizes of 2 mm and 4mm.

77

Figure 7.19 Dose profile computed along a diagonal in the top plane of the structure. The central beam axis is located at a distance of 4.2 cm.

Figure 7.20 Dose profile computed along a diagonal in the bottom plane of the structure. The central beam axis is located at a distance of 4.2 cm.

Dose (cGy)

78

High dose gradient area Figure 7.21 shows a DHV of the internal structure calculated for dose grid sizes of 2mm and 4mm. In both cases the curves have a linear behaviour between 20% and 80% of dose at the isocentre, on which a step-pattern due to the dose grid size is apparent.

Figure 7.21 DVH of the same structure obtained for two dose grid sizes of 2mm (red) and 4mm (blue).

(a)

(b)

Figure 7.22 Measurements of the distance d between the 20% (blue) and 80% (pink) isodose lines; (a) 4mm dose grid size; (b) 2mm dose grid size. The two lines in the centre of the figure represent the beam edges.

79

Table 7.16 Results of the volume calculation between the 20% and 80% isodose surfaces. Dose grid size [mm] V80% [cm3] V20% [cm3] V20% - V80% [cm3] d [cm] 30 x d [cm3] Difference between V20% - V80% and direct volume calculation. 0.4% 0.6%

4 2

64.53 66.05

88.53 86.03

24.09 19.98 20.6%

0.80 0.67

24.00 20.10 19.4%

Difference between dose grids

In Table 7.16 the results are shown of the volume calculations between the 20% and 80% isodose surfaces in the internal structure for the two dose grid sizes. The difference in calculated values of the volume between these isodose surfaces due to dose grid size is around 20%. Obviously the dose grid size is an important factor for the accurate calculation of DVHs in high dose gradient regions. For the direct volume calculation the distance d between the 20% and 80% isodose lines has to be multiplied with the surface of the internal structure in the sagittal plane, i.e., 30 cm2. Good agreement exists between the volume assessed from the dose-volume histogram and the direct volume determination using the isodose planes, indicating that the TPS has no problem with the DVH algorithm. It also shows the usefulness of this DVH test in regions with a high dose gradient. Test 6.2.2 Standard treatment techniques: d - pelvis/prostate As an example of a dose matrix evaluation, the verification of a 4-field box technique irradiating the CarPet phantom (Gillis et al., 2004) will be presented. The calculation is performed with a treatment planning system with a dose matrix of 0.15 x 0.15 x 0.2 cm3. The plan has been normalised to a value of 2 Gy (100 %) at the isocentre and the resulting dose distribution in a transversal plane at 1 cm distance from the central plane is given in Figure 7.23.

Figure 7.23 Calculated dose distribution for a 4-field box technique in a plane at 1 cm distance from the central plane in the CarPet phantom.

80

To verify the calculation, a radiographic film (Kodak EDR2) was positioned inside the phantom and irradiated with four 6 MV beams. A film calibration from 0 to 3.4 Gy in 10 steps was performed with films at 10 cm depth in a polystyrene phantom perpendicular to the irradiation beam. This was done prior to the irradiation of the CarPet phantom. The calibration films were scanned in a VIDAR scanner and converted to optical density. Next the treatment film was scanned and converted to absorbed dose. The resulting dose distribution is shown in Figure 7.24. After careful alignment using punched holes in the film, the difference between measured and calculated dose was calculated (Figure 7.25).

Figure 7.24 Dose distribution measured with a Kodak EDR2 film positioned at 1 cm from the central plane in the CarPet phantom.

Figure 7.25 Dose deviation presented as the dose calculated by the TPS divided by the dose measured with the film, performed for each point. The bluish regions show areas where the measured values are higher than the calculated data, while the reddish regions are areas where the measurements are lower.

To include a spatial tolerance in the dose difference, the -method, as outlined in Section 2.2.3, has also been applied to both sets of dose data. The resulting matrix of -values is shown in Figure 7.26. One can notice that inside the area where all four fields contribute to the dose, good agreement between measurement and calculation is found. Good agreement is also observed in most other areas, except for the penumbra regions. Some problems due to the film/phantom clamping might be the reason for the differences observed at the two lateral sides. 81

Figure 7.26 Matrix showing the -values, expressed as a percentage where 100% is equal to gamma=1, for the 4-field box technique irradiating the CarPet phantom. White are -values equal to unity; red colours represent areas where the criteria (2% in dose relative to the dose at the isocentre, or 2mm in distance) are not fulfilled, while blue shades get closer to zero when they get darker.

The resulting -values can also be expressed as a gamma-area histogram over a certain area of interest (Figure 7.27). In this way it is possible to compare the accuracy of the actual delivery of a certain irradiation technique, with predefined criteria for that area. However, the tolerance criteria given in Table 2.1 are valid for single photon beams, and it is not obvious that the same criteria are valid for all types of composite plans.

Figure 7.27 Gamma-area histogram showing the -values for the 4-field box technique irradiating the CarPet phantom in the analysed region of the film.

It should be noted that by doing film measurements to verify dose calculations performed with a TPS, the accuracy of the film dosimetry system should be well known, and not introduce large uncertainties. Also the same precautions for the constancy of the output of the accelerator should be made, as during other dose verification measurements, i.e., all data should be normalized to the absolute dose per monitor unit at 10cm depth. 82

APPENDIX A.1 DEFINITIONS

Acceptance testing A set of procedures carried out after delivery to confirm that the TPS works according to its specifications as documented at the moment of purchase. Beam characterisation set The beam characterisation set consists of the basic beam data required for beam modelling and specified according to the format defined by the vendor. This set needs to be determined and entered into the system. Usually, the set consists of measured dose values, but it may also consist of Monte Carlo computations, if properly benchmarked for the treatment machine. In addition, a complete geometrical characterisation is also required, including all movement ranges and limits. Beams-eye-view (BEV) A beams-eye-view display is a computer-generated image projected at a given distance from the source (usually the SAD), as it would appear to a viewer located at the radiation source and looking toward the irradiated medium. It generally concerns the patients anatomy, obtained from previously contoured structures in a 3-D set of images. Like the radiation beam, BEV images follow the beam divergence. BEV images are used to decide which beam orientations yield the best view of the target volume without irradiating too much normal anatomy, and to design shielding blocks to protect normal anatomy from unnecessary irradiation. Beam verification set The beam verification set can be identical to the beam characterisation set but is usually extended and includes several other situations, e.g., prolonged SSD and rectangular fields. Commissioning A set of procedures required bringing the new TPS or new software release into safe clinical operation. The TPS user should define the details of this procedure. The procedures include the introduction of geometric and dosimetric data into the system to define the treatment machine and its beams, and performing tests to learn how to use it, to verify the correct functioning of the entire software and to determine the limits of accuracy of the various calculations. Confidence limit A global index calculated from the statistical distribution of the dose deviation among a large number of points. The confidence limit concept is useful to evaluate if the dose deviation between measured and calculated dose distributions is acceptable. It can be extended to the analysis of spatial deviations or of the gamma index distribution. 83

Digitally reconstructed radiograph (DRR) A digitally reconstructed radiograph is similar to a BEV, except that a DRR is a computergenerated image, obtained from a 3-D data set of CT (or other modality) images and projected at a given distance from the source. A DRR shows how anatomy would appear to a viewer located at the radiation source and looking towards the isocentre. DRRs are often used as reference images compared to portal images to verify that a treatment plan is being delivered with high precision. Dose-volume histogram (DVH) Graphical representation of a 3-D dose distribution showing the number of voxels, i.e., the volume or relative volume, of a structure, which receives a given dose. In the cumulative form, which is mostly used, the count of voxels includes all those which receive a dose larger than the given dose value. In the differential form, it represents a histogram where for each dose value the relative frequency of voxels that receive that dose is represented. Dose deviation (d(i) or d%(i)) The ratio between calculated and measured dose for the same number of monitor units. Alternatively, this deviation can be expressed as a percent difference between calculated and measured dose with respect to the calculated value taken as reference. The relationship between these expressions is: d%(i) = 100. [d(i)-1]. Electron density, relative electron density The electron density is the number of electrons per unit volume, while the relative electron density is the electron density for a particular medium divided by the electron density for water. This is important for dose calculations and is typically obtained from CT information. Gamma index A figure of merit used for evaluation of the dose calculated by the TPS, which combines the dosimetric deviation and the spatial deviation. The gamma index is equal to 1 when the normalized vectorial distance between calculations and measurements in the position-dose space is within predefined tolerances. For a gamma index less than 1 the calculated dose values are within the accepted criteria (e.g., 3% / 3mm). Generic beam data set A generic beam data set is a beam characterisation set used by the vendor to test its own TPS and to demonstrate its accuracy. It can be obtained from published reports and it should be supplied with the TPS to facilitate its acceptance. The generic beam data set must not be used for clinical purposes.

84

Percentage depth dose (PDD) The ratio of the dose at a given depth to the dose at the depth of maximum dose in a water equivalent phantom, for a given field size, at a fixed source-skin distance. Periodic quality control A set of procedures carried out to verify periodically the correct functioning of the TPS. These tests are repeated with a pre-set frequency; some can be carried out automatically, others manually. Quality assurance (QA) of a TPS A set of procedures carried out to determine the accuracy and reliability of the TPS, and to guarantee that the system performs according to previously established specifications. Spatial deviation (r(i)) Minimum distance between a point rm(i) where a certain dose value has been measured and a point with the same calculated dose value. The spatial deviation is also called distance-toagreement. Tissuephantom ratio (TPR) The ratio of the absorbed dose at certain depth to that at reference depth measured in a large water phantom at the same fixed sourcedetector distance. Tolerance () The maximum acceptable deviation between a calculated and measured physical quantity. It can be expressed either as a percent deviation or as a spatial deviation, depending on the spatial region being analysed (low vs. high dose gradient region). Treatment Planning System (TPS) A treatment planning system consists of a software package, or a combination of different packages, and its hardware part. It enables the input of patient data, the anatomy definition, beam set-up, the dose distribution calculation, the plan evaluation in terms of dose, volume, biological effect and the output for documentation and transfer to other units (simulator, record and verify system, treatment machine). (3-D) Treatment Planning System (3-D TPS) A modern 3-D treatment planning system offers the functionality of: constructing a 3-D patient model (based on a volumetric CT-scan); simulating 3-D configurations of beams (with arbitrary beam orientations); positioning of the isocentre and field shape; performing 3-D dose calculation (with algorithms that take the 3-D aspects of patient, beams and interaction physics into account); evaluating and optimising 3-D dose distributions (dose-volume histograms and normal tissue complication probability calculations); advanced viewing of patient anatomy, beams and dose distribution in their 3-D relationship. 85

APPENDIX A.2 LIST OF ABBREVIATIONS AND SYMBOLS

Symbol A-P BEV CCW CW d d(i) or d%(i) CF CT D DICOM DRR DVH HU IMRT M MLC MR MU OAF OAD OAR PDD PET PRF QA QC QI r(i) ROI RW50 s SAD SD SSD SPECT Meaning Anterior Posterior Beams-Eye-View Counter-ClockWise ClockWise Depth; Distance between 20% and 80% isodose lines Deviation between calculated and measured dose per monitor unit at a point i expressed either as a ratio or as a percent difference. Correction Factor Computed Tomography absorbed Dose Digital Imaging and COmmunication in Medicine Digitally Reconstructed Radiograph Dose-Volume Histogram Hounsfield Unit Intensity-Modulated Radiation Therapy number of Monitor units Multi-Leaf Collimator Magnetic Resonance Monitor Unit Off-Axis Factor Off-Axis Distance Organ At Risk Percentage Depth Dose Positron Emission Tomography beam PRoFile Quality Assurance Quality Control Quality Index Spatial deviation (or distance-to-agreement) at point rm(i) Region Of Interest Radiological Width field size at isocentre Source-Axis Distance Standard Deviation Source-Skin Distance Single Photon Emission Computed Tomography 87

TAR TMR TPR TPS z 2-D 3-D

Tissue-Air Ratio Tissue-Maximum Ratio Tissue-Phantom Ratio Treatment Planning System CT slice number Two-Dimensional Three-Dimensional Tolerance for dose deviation d%(i) or spatial deviation r(i) Gamma index

88

APPENDIX A.3 CATEGORISATION OF TESTS

In this booklet an attempt has been made to provide suggestions for a general division of QA tests to be performed by the vendor, or by an individual user. Particularly in Chapter 3 (Anatomical Description), and in Chapter 4 (Beam Description), a large number of tests are presented the vendor of a specific system should perform, either before the system is installed in the hospital, or during the acceptance testing at the users site. Some of these vendor tests can be generic tests, the results of which should be shown to and discussed with the user; others need to be performed at the users place. Such a division depends on the availability of the users data by a vendor, as well as the specific situation at a users location, e.g., of the time and resources available for extensive on site testing. In this Appendix therefore no attempt was made to make such a division in the column Vendor Acceptance Tests. Nevertheless, the user is well advised to repeat certain vendor acceptance tests, in particular those tests, which are adequate to spot-check the integrity of the users input data. A star in brackets denotes such tests. The following tests are numbered according to the corresponding sections in this booklet.
Test # 3 3.1 3.1.a 3.1.b 3.1.c 3.1.d 3.1.e 3.1.f 3.2 3.2.1 3.2.1a 3.2.1b 3.2.1c 3.2.1d 3.2.1e 3.2.1f 3.2.1g Short test description Vendor Acceptance Tests User Commissioning Tests

Anatomical description Basic patient entry Two patients with the same last name Two patients with the same ID- number The same patient twice The same patient with two different targets Deleting a patient Moving/copying patients/plans from one directory to another. Image input and use Image input Identity consistency of scan Scan parameters-varying slice thickness Two different sets of images for the same patient Maximum number of CT slices Patient orientation Integrity of simultaneous input Opening a patient file more than once

* * * * * *

* * * * * * *

* * * * *

89

3.2.1h 3.2.1i 3.2.1j 3.2.2 3.2.2a 3.2.2b 3.2.3 3.2.3a 3.2.3b 3.2.3c 3.2.3d 3.2.3e 3.2.4 3.3 3.3.1 3.3.1a 3.3.1b 3.3.1c 3.3.2 3.3.2a 3.3.2b 3.3.3 3.3.3a 3.3.3b 3.3.4 3.3.4a 3.3.4b 3.3.4c 3.3.4d 3.3.4e 3.3.4f 3.3.4g 3.3.4h 3.3.4i 3.3.4j 3.3.5 3.3.5a 3.3.5b 3.3.5c

Geometric integrity of slices CT number representation Text information Contour input Digitiser input Film scanner input Image use Geometry of reconstructed images Orientation of reconstructed images Grey-scale representation Grey-scale representation of reconstructed images Windowing and zooming images Co-ordinate system of images Anatomical structures Definition of anatomical structures Unique identification Unique properties Maximum number of contours per anatomical structure Automated contouring Correct geometry in automated contouring Threshold changing for automated contouring Manual contouring Contouring direction Maximum number of points Manipulation of contours Add margin to a contour 3-D structure expansion Correcting, adding, deleting and copying contours and structures Validation TPS depending tools 2-D contour transfer Hidden contours Bifurcated structures Interpolated structures Bolus definition Construction of volumes Volume computation Construction of a volume from a set of CT slices with non-regular spacing Construction of a volume from a set of CT slices with non-sequential order of slices during contouring

* * * * * * * * * * *

* * * * * * *

* * *

* *

* * * * * * * * * * * * *

(*) (*) * *

(*) (*)

* *

90

3.3.5d 3.3.5e 3.3.5f 3.3.5g 4 4.1 4.1.1 4.1.1a 4.1.1b 4.1.1c 4.1.1d 4.1.1e 4.1.2 4.1.3 4.1.4 4.1.5 4.1.6 4.1.7 4.1.8 4.1.9 4.1.10 4.2 4.2.1 4.2.2 4.2.3 4.2.4 4.2.5 4.2.6 4.3 4.3.1 4.3.2 4.3.3 4.3.4 5 5.1 5.1.1 5.1.2 5.2 5.2.1 5.2.2 5.2.3 5.2.4 5.2.5 5.2.6

Construction of a volume from non-axial contours Capping option Volume of subtracted regions Boolean option Beam description Beam definition SAD, SSD and field size Field size Divergence Isocentre position SSD or depth verification Test from 4.1.1 for SSD>100cm Gantry rotation Collimator rotation Table movement Jaw definition and beam co-ordinates Multi-leaf collimator definition Wedge and block insertion Consistency check of beam co-ordinate system Warnings and error messages Bolus definition Beam display Beams-eye-view display Beam position and shape Beam position in BEV Block position in BEV MLC-shaped field Bolus position Beam geometry Automatic block and auto-leaf positioning User-defined block. DRR: linearity and divergence Input, change and edit functions Dose and monitor unit calculation Basic beam dataset Data input process Documentation Dose calculation Open square fields Open rectangular fields Variation in SSD Wedged square field Wedged rectangular fields Field with a central block

* * * *

(*) (*) (*) (*)

* * * * * * * * * * * * * * * * * * * * *

* * * * *

* * * *

* *

* * * * * * * * (*) (*) (*) (*) (*) (*)

91

5.2.7 5.2.8 5.2.9 5.2.10 5.2.11 5.2.12 5.2.13 5.2.14 5.2.15 5.2.16 5.2.17 5.2.18 5.3 5.3.1 5.3.2 5.4

Blocked field Inhomogeneities Oblique incidence Missing tissue Off-axis square field Off-axis elongated field Wedged off-axis field Off-plane field Square MLC field Off-axis square MLC field MLC-shaped field Block and tray insertion 2-D and 3-D dose verification 2-D dose distribution Dose-volume histogram Monitor unit calculation

* * * * * * * * * * * * * *

(*) (*) (*) (*) (*) (*) (*) (*) (*) (*) (*)

92

REFERENCES
AAPM Report #55. Radiation Treatment Planning Dosimetry Verification, Radiation Therapy Committee Task Group #23, edited by D. Miller. American Institute of Physics, College Park, MD, 1995. Bakai A., Alber M. and Nsslin F. A revision of the -evaluation concept for the comparison of dose distributions. Phys. Med. Biol. 48: 3543-3553, 2003. Brahme A., Chavaudra J., Landberg T., McCullough E., Nsslin F., Rawlinson A., Svensson G. and Svensson H. Accuracy requirements and quality assurance of external beam therapy with photons and electrons. Acta Oncol. Suppl. 1, 1988. Caneva S., Rosenwald J-C. and Zefkili S. A method to check the accuracy of dose computation using quality index: application to scatter contribution in high energy photon beams. Med. Phys. 27: 1018-1024, 2000. Cosset J.M., ESTRO Breur Gold Medal Award Lecture 2001: Irradiation accidents lessons for oncology? Radiother. Oncol. 63: 1-10, 2002. Craig T., Brochu D. and Van Dyk J. A quality assurance phantom for three-dimensional radiation treatment planning. Int. J. Radiat. Oncol. Biol. Phys. 44: 955-966, 1999. Dahlin H., Lamm I.L., Landberg T., Levernes S. and Uls N. User requirements on CT-based computed dose planning systems in radiation therapy. Acta Radiol. Oncol. 22: 397-415, 1983. Depuydt T., Van Esch A. and Huyskens D.P. A quantitative evaluation of IMRT dose distributions: refinement and clinical assessment of the gamma evaluation. Radiother. Oncol. 62: 309-319, 2002. Dutreix A., Bjrngard B.E., Bridier A., Mijnheer B.J., Shaw J.E. and Svensson H. Monitor Unit Calculation For High Energy Photon Beams, ESTRO Booklet No. 3, ESTRO, Brussels, Belgium, 1997. Ezzell G., Galvin J.M., Low D., Palta J.R., Rosen I., Sharpe M.B., Xia P., Xiao Y., Xing L. and Yu C.X. Guidance document on delivery, treatment planning, and clinical implementation of IMRT: Report of the IMRT subcommittee of the AAPM Radiation Therapy Committee. Med. Phys. 30: 2090 2115, 2003.

93

Fraass, B., Doppke, K., Hunt, M., Kutcher, G., Starkschall, G., Stern, R. and Van Dyk, J. American Association of Physicists in Medicine Radiation Therapy Committee Task Group 53: Quality assurance for clinical radiotherapy treatment planning. Med. Phys. 25: 17731836, 1998. Gillis S., De Wagter C., Bohsung J., Perrin, B., Williams, P. and Mijnheer B.J. An inter-centre quality assurance network for IMRT using film dosimetry: preliminary results of the European QUASIMODO project. Radiother. Oncol. (submitted). Grosu A-L., Weber W., Feldmann H.J., Wuttge B., Bartenstein P., Grosss M.W., Lumenta C., Schwaiger M. and Molls M. First experience with I-123-alpha-methyl-tyrosine SPECT in the 3-D radiation treatment planning of brain gliomas. Int. J. Radiat. Oncol. Biol. Phys. 47: 517526, 2000. Harms W.B., Low D.A., Wong J.W. and Purdy J.A. A software tool for the quantitative evaluation of 3D dose calculation algorithms. Med. Phys. 25: 1830-1836, 1998. Henze M., Schuhmacher J., Hipp P., Kowalski J., Becker D.W., Doll J., Mcke H.R., Hofmann M., Debus J. and Haberkorn U. PET imaging of somatostatin receptors using [68Ga]DOTA-D-Phe1-Tyr3-octreotide: first results in patients with meningiomas. J. Nucl. Med. 42: 10531056, 2000. IAEA Report TRS 398. Absorbed dose determination in external beam radiotherapy. An international code of practice for dosimetry based on standards of absorbed dose to water. International Atomic Energy Agency, Vienna, Austria, 2000. IAEA Report SRS 17. Lessons learned from accidental exposures in radiotherapy. International Atomic Energy Agency, Vienna, Austria, 2000. IAEA Report Investigation of an accidental exposure of radiotherapy in Panama. International Atomic Energy Agency, Vienna, Austria, 2001. IAEA TECDOC-xxx, Commissioning and quality assurance of computerized treatment planning. International Atomic Energy Agency, Vienna, Austria, 2004. ICRP Publication 86. Prevention of Accidental Exposures to Patients Undergoing Radiation Therapy. International Commission on Radiation Protection, Pergamon Press, Oxford, 2001. ICRU Report 42. Use of computers in external beam radiotherapy procedures with highenergy photons and electrons. International Commission on Radiation Units and Measurements, Baltimore, Maryland, USA, 1987. 94

IEC Report 61217: Radiotherapy equipment Co-ordinates, movements and scales. International Electrotechnical Commission, Geneva, Switzerland, 1996. IEC Report 62083: Medical electrical equipment Requirements for the safety of radiotherapy treatment planning systems. International Electrotechnical Commission, Geneva, Switzerland, 2000. Karger P.C., Hipp P., Henze M., Echner G., Hss A., Schad L. and Hartmann G.H. Stereotactic imaging for radiotherapy: accuracy of CT, MRI, PET and SPECT. Phys. Med. Biol. 48: 211-221, 2003. Levivier M., Goldman S., Pirotte B., Brucher J-M., Balriaux D., Luxen A., Hildebrand J. and Brotchi J. Diagnostic yield of stereotactic brain biopsy guided by positron emission tomography with [18F]fluorodeoxyglucose. J. Neurosurg. 82 : 44552, 1995. Low D.A., Harms W.B., Mutic S. and Purdy J.A. A technique for the quantitative evaluation of dose distributions. Med. Phys. 25: 656-661, 1998. Low D.A. and Dempsey, J.F. Evaluation of the gamma dose distribution comparison method. Med. Phys. 30: 2455-2464, 2003. Mayles, W.P.M., Lake, R., McKenzie, A., Macauly, E.M., Morgan, H.M., Jordan, T.J. and Powley S.K. Physics aspects of quality control in radiotherapy. IPEM Report No.81. The Institute of Physics and Engineering in Medicine, York, Great Britain, 1999. Mijnheer B.J., Battermann J.J. and Wambersie A. What degree of accuracy is required and can be achieved in photon and neutron therapy? Radiother. Oncol. 8: 237-252, 1987. Mijnheer B.J., Bridier A., Garibaldi C., Torzsok K. and Venselaar J.L.M. Monitor Unit Calculation For High Energy Photon Beams - Practical Examples. ESTRO Booklet No. 6, ESTRO, Brussels, Belgium, 2001. NCS Report-xx, Quality assurance of 3-D treatment planning systems; practical guidelines for acceptance testing, commissioning, and periodic quality control of radiation therapy treatment planning systems. The Netherlands Commission on Radiation Dosimetry, Delft, The Netherlands, 2004. (http://www.ncs-dos.org/draft_01.html) Panitsa E., Rosenwald J-C. and Kappas C. Quality control of Dose Volume Histogram computation characteristics of 3D treatment planning systems. Phys. Med. Biol.43: 2807-2816, 1998.

95

Pirotte B., Goldman S., David Ph., Wikler D., Damhaut Ph., Vandesteene A., Salmon I., Brotchi J. and Levivier M. Stereotactic brain biopsy guided by positron emission tomography (PET) with [F-18] Fluorodeoxyglucose and [C-11] Methionine. Acta Neurochir. 68 (Suppl): 1338, 1997. Shaw, J.E. A guide to commissioning and quality control of treatment planning systems. Report No.68. The Institute of Physics and Engineering in Medicine and Biology, York, Great Britain, 1996. Siddon R.L. Solution to treatment planning problems using co-ordinate transformations. Med. Phys. 8: 766-774, 1981. SSRPM Report 7: Quality control of treatment planning systems for teletherapy. Swiss Society for Radiobiology and Medical Physics: ISBN 3-908125-23-5, 1997. Van Dyk, J., Barnett, R.B., Cygler, J.E. and Shragge P.C. Commissioning and quality assurance of treatment planning computers. Int.J. Radiat. Oncol. Biol. Phys. 26: 261-273, 1993. Van Dyk, J., Barnett, R.B. and Battista J.J. Computerized radiation treatment planning systems. Pp. 231-286. In: The modern technology of radiation oncology. J. Van Dyk (Ed.). Medical Physics Publishing, Madison, WI, USA, 2003. Venselaar J. and Prez-Calatayud, J. (Eds.). A practical guide to quality control of brachytherapy equipment. ESTRO Booklet No. 8, ESTRO, Brussels, Belgium, 2004. Venselaar J.L.M. and Welleweerd J. Application of a test package in an intercomparison of the performance of treatment planning systems used in a clinical setting. Radiother. Oncol. 60: 203-213, 2001. Venselaar J.L.M., Welleweerd J. and Mijnheer B.J. Tolerances for the accuracy of photon beam dose calculations of treatment planning systems. Radiother. Oncol. 60: 191-201, 2001.

96