Machine learning predicts prostate cancer prognosis using biopsy samples

A new machine learning approach accurately predicts prostate cancer prognosis from biopsy samples, even with limited data. The method uses whole-mount histopathology and a clinically informed reasoning step, demonstrating robustness across different institutions and specimen types. It offers an interpretable and applicable framework for treatment decisions. This research, supported by several Japanese organizations, aims to improve clinical decision-making in real-world settings, offering an alternative to the traditional Gleason grading system.