All of the patients received a high-resolution T2-weighted MRI scan in the pelvic region at diagnosis. The images were taken using Philips Medical Systems Achieva 1.5 T MRI scanners with a 2 mm slice thickness. Using Velocity software (Varian Medical Systems, Palo Alto, CA, USA), each MRI scan was manually contoured on two separate ROIs: the prostate gland and the peripheral zone of the prostate. The whole prostate gland is the most common region reviewed by radiologists and also easy to contour for future studies, while the peripheral zone is where over 90% of prostate cancers develop [30 (link)]. Figure 1 shows the created contours on two example patients, with the prostate gland outlined in blue and its peripheral region outlined in red.
All of the images and segmented ROIs were then sent to 3D-Slicer v4.10 for image processing and radiomic feature extraction [31 (link),32 (link)]. To avoid bias field distortions and data heterogeneity bias, a bias field correction using N4 and an image normalization using histogram matching were performed on all images [33 (link)].
All of the images and segmented ROIs were then sent to 3D-Slicer v4.10 for image processing and radiomic feature extraction [31 (link),32 (link)]. To avoid bias field distortions and data heterogeneity bias, a bias field correction using N4 and an image normalization using histogram matching were performed on all images [33 (link)].