Fig. 3

Proposed methodology for 3-slice CT (computed tomography) image registration to enable voxel-wise analysis. Registration was done slice-wise, small boxes in colored red, light blue and yellow represent each slice registration process which is performed independently. The process involved the generation of body masks for each slice from raw CT images. Processed CT images are the CT images after removal of CT tables and non-body objects, generated from raw CT images and body masks. Deep learning segmentation models were used to generate liver, spleen, and abdominal muscle segmentation masks. For creating ISAT (inside subcutaneous adipose tissue) masks the processed CT images and a predicted deep learning abdominal muscle mask were used. Lean tissue, vertebrae, and cortical bone masks were generated from processed CT images. All preprocessed images and masks were used simultaneously to perform registration tasks