通过批量内关系特征提升医学异常检测精度，降低假阳性。
In-batch Relational Features Enhance Precision in An Unsupervised Medical Anomaly Detection Task
- 用批量内超图建模正常样本间上下文关系，增强嵌入表征
- 在脑肿瘤数据集上实现0.90的AUC-ROC，平均精度提升16%
- 可调节批量大小以控制健康变异整合程度，适合临床应用
将正常解剖变异与病理性变化混淆仍是无监督医学图像异常检测中的主要挑战，导致大量假阳性。为增强对健康变异的建模，本文在卷积神经网络自编码器的潜在表示中，通过批量内超图估计和共享权重图卷积层，引入正常队列中的上下文相似性，生成具有群体感知能力的嵌入。在包含2D MRI扫描的异构脑肿瘤数据集上，该方法显著提升了健康与病理样本的可分性，达到0.90的AUC-ROC（95%置信区间0.84-0.95，绝对提升5.7%），平均精度提升16%（0.78 AP，95% CI 0.66-0.89），从而有效降低假阳性率。此外，异常检测及下游肿瘤/非肿瘤分类性能随增强表示中捕捉的迷你批量上下文规模增大而提升，表明该方法提供了一个可调的整合健康变异的机制。
原文摘要 · Abstract (English)
Confounding pathology with normal anatomical variation remains a significant challenge in unsupervised medical-image anomaly detection, resulting in numerous false positives. To enhance integration of healthy variation, we augment the latent representation of a CNN autoencoder with contextual similarities within a normal cohort through batch-wise hypergraph estimation and a shared-weights graph convolution layer, producing a population-aware embedding. On a heterogeneous brain-tumor dataset of 2D MRI scans, the method improves separability between healthy and pathological samples, achieving an AUC-ROC of 0.90 (95% CI 0.84-0.95, 5.7% absolute gain), and a 16% absolute improvement in average precision (0.78 AP, 95% CI 0.66-0.89), thereby lowering false-positive rates. Moreover, both anomaly detection and downstream tumor versus no-tumor classification performance improve with the size of the mini-batch context captured in the augmented representation, suggesting a tunable lever for integrating healthy variation.
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。