Fig. 1.

An overview of the pipeline of our study using histopathology WSIs and genomic data for prostate cancer recurrence prediction for patients with Gleason score 7. (a) WSI images and genomic data were collected from patients with prostate cancer; (b) a prostate WSI exhibits different Gleason patterns. For example, a region in a green square has the Gleason pattern 3 while regions in blue squares have the Gleason pattern 4; (c) the pathway scores were quantified using RNA sequences. Patches of region of interests were automatically selected from WSIs. The image patches and pathway scores were integrated into deep neural networks to extract computational biomarkers, which were fed into a Cox regression model in conjunction with clinical prognostic factors for disease recurrence analysis.