AI summaryⓘ
The authors studied whether microscope images of prostate cancer tissue (H&E stained slides) can help predict if the cancer will come back after surgery, beyond the usual method based on Gleason grade. They tested multiple AI models to analyze these images and found that just using images alone was not better than clinical info like grade and stage. They then created a new method called Grade-Disentangled MIL (GD-MIL), which removes the influence of Gleason grade from the image analysis and combines it with clinical data. This new method performed better than the clinical model alone, showing that the images contain extra useful information for predicting cancer recurrence. Their results suggest combining image features with clinical data can improve how doctors assess prostate cancer risk.
Biochemical recurrence (BCR)Radical prostatectomyGleason gradeH&E whole slide images (WSIs)Multiple instance learning (MIL)TCGA-PRAD datasetConcordance index (C-index)Grade-Disentangled MIL (GD-MIL)Cox modelLog-rank test
Abstract
Biochemical recurrence (BCR) after radical prostatectomy is a critical endpoint in prostate cancer, yet risk stratification relies almost entirely on variables dominated by Gleason grade. Whether H&E whole slide images (WSIs) carry prognostic signal beyond grade, and whether multiple instance learning (MIL) can recover it, remains unsettled. A key obstacle is that many pipelines select model checkpoints on the evaluation fold, artificially inflating concordance. We construct a rigorous benchmark on TCGA-PRAD (487 patients, 101 BCR events) using strict out-of-fold scoring over five-fold cross-validation repeated across five seeds. The choice of MIL aggregator (ABMIL, CLAM, TransMIL, PatchGCN) has little effect (C-index 0.61-0.64 with UNI2-h), while the feature extractor is the dominant factor (ResNet50 0.566 versus pathology foundation models up to 0.639). A clinical Cox model on grade, stage, and age reaches 0.687; no imaging-only model significantly outperforms it (p > 0.10). We introduce Grade-Disentangled MIL (GD-MIL), a gated-attention MIL encoder trained with a gradient-reversal grade adversary that encourages the slide representation to be invariant to Gleason grade before late fusion with clinical variables. GD-MIL achieves C-index 0.704, significantly outperforming both the clinical baseline (delta-c = +0.029, p = 0.0005) and the best imaging-only model (delta-c = +0.062, p = 0.039), suggesting H&E morphology contains prognostic information complementary to grade. A median risk split yields log-rank p < 0.0001 separation in BCR-free survival (~20% vs ~70% at five years).