GD-MIL: Grade-Disentangled Multiple Instance Learning for Multimodal Biochemical Recurrence Prediction in Prostate Cancer

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).