提出新模型从病理图像中挖掘超越分级的复发预测信息。
GD-MIL: Grade-Disentangled Multiple Instance Learning for Multimodal Biochemical Recurrence Prediction in Prostate Cancer
- 用对抗性注意力机制让模型忽略分级，专注图像中的其他预后信号。
- 在TCGA数据上实现C-index 0.704，优于临床模型和纯影像模型。
- 适合做癌症预后研究或多模态医疗AI的开发者参考。
根治性前列腺切除术后生化复发（BCR）是前列腺癌关键终点，但风险分层几乎完全依赖戈尔登分级。是否组织学全切片图像（WSIs）包含超越分级的预后信息，以及多实例学习（MIL）能否恢复该信息，尚不明确。主要障碍在于许多流程在验证集上选择模型检查点，人为夸大一致性。本研究在TCGA-PRAD数据集（487例患者，101例复发事件）上构建严格基准，采用五折交叉验证重复五次种子实验，严格使用外折叠评分。不同MIL聚合器（ABMIL、CLAM、TransMIL、PatchGCN）影响较小（C-index 0.61–0.64，使用UNI2-h特征），而特征提取器是主导因素（ResNet50为0.566，病理基础模型最高达0.639）。基于分级、分期和年龄的临床Cox模型达到0.687；无影像模型显著优于它（p > 0.10）。我们提出等级解耦的MIL（GD-MIL），通过梯度反向等级对抗训练门控注意力编码器，使切片表示在与临床变量融合前对分级保持不变。GD-MIL实现C-index 0.704，显著优于临床基线（Δc = +0.029，p = 0.0005）和最佳影像模型（Δc = +0.062，p = 0.039），表明H&E形态学包含与分级互补的预后信息。中位风险分组在无生化复发生存率上差异极显著（对数秩检验p < 0.0001，五年生存率约20% vs 约70%）。
原文摘要 · Abstract (English)
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).
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。