I have followed the pre-processing scheme as indicated in the paper Huo et.all; and when executing inference the output is not as expected, so: What is the order and steps I should apply to images in order to obtain the masks provided in the description???????

This is my actual pre-processing pipeline: Skull stripping > Bias field correction > Registration to MNI305 > Intensity rescaling [-1,1] > Random cropping 96x96x96. (Of course input images to this pre-processing pipeline have 240x240x155 (dimensions) and 1x1x1 (voxel size)

I have removed histogram matching (after proving it) since it does not make sense (I am just executing inference with T1 weighted images).