Comparing Gleason Pattern 4 Measurement Approaches on Prostate Biopsy Using Machine Learning: A Proof-of-Principle Study

ObjectiveTo demonstrate the proof of principle that machine learning (ML) can be used to quantify Gleason Pattern (GP) 4 on digitized biopsy slides using multiple measurement approaches, allowing direct comparison of their prognostic performance. MethodsWe assembled a convenience sample of 726 patients with grade group 2-4 prostate cancer on systematic biopsy who underwent radical prostatectomy between 2014 and 2023. Digitized biopsy slides were analyzed using a machine-learning algorithm (PAIGE-AI) to quantify GP4 using multiple measurement approaches, particularly with respect to how gaps between cancer foci ("interfocal stroma") were handled. GP4 extent was quantified using linear measurements or a pixel-based area metric. Discrimination of each GP4 quantification approach, along with Grade Group (GG), was assessed for adverse radical prostatectomy pathology and biochemical recurrence. ResultsWe identified 15 different quantification approaches and observed differences between their discrimination. The highest discrimination was in the pixel-countingmethod (AUC 0.648). GP4 quantification outperformed GG for predicting adverse pathology (AUC 0.627 vs 0.608). Amount of GP3 was non-predictive once GP4 was known. These findings were consistent for BCR. ConclusionsWe were able to measure slides using 15 distinct measurement approaches and replicated prior findings using ML to quantify GP4. Our findings support the use of ML as a research tool to compare different GP4 quantification approaches. We intend to use our method on larger cohorts to determine with which measurement approach best predicts oncologic outcome.