Quantitative digital pathology reveals morphological and molecular correlates of tumor aggressiveness in prostate cancer.

Abstract

Digital pathology enables objective evaluation of prognostic markers in prostate cancer through integrated morphological and molecular analysis. In this study, 72 radical prostatectomy samples were analyzed using tissue microarrays and whole-slide imaging, with tumor regions annotated by pathologists and quantified in QuPath for PTEN, Ki-67, ATM, CD8, and key histological features. PTEN quantification showed high variability and limited predictive value (AUC $=$ 0.61). Cribriform morphology was significantly associated with advanced pathological stage, non-organ-confined tumors, and a profile of aggressiveness markers, whereas CD8 $_{}^{+}$ density did not correlate, supporting an "immune-cold" phenotype. Quantitative cribriform area did not add prognostic value. A logistic regression model combining Gleason pattern 4 percentage and ATM expression predicted cribriform morphology with good accuracy (AUC $=$ 0.79). These results highlight the potential of digital pathology to refine risk stratification and identify high-risk morpho-molecular features in prostate cancer.