🤖 AI Summary
This work addresses the challenge of applying concept bottleneck models to cancer imaging diagnosis, where reliance on extensive instance-level concept annotations limits practicality. To mitigate this dependency, the authors propose a prior-guided hybrid concept bottleneck model that integrates sparse concept annotations, class-conditional concept distribution matching on unlabeled data, and prior-informed initialization of the diagnostic head. Evaluated on multi-center cancer imaging datasets, the method demonstrates strong performance even with only 10% concept annotation coverage, achieving concept AUCs of 0.741, 0.787, and 0.642 for breast masses, calcifications, and pulmonary nodules, respectively. The approach attains diagnostic accuracy comparable to black-box models while preserving interpretability through explicit concept reasoning.
📝 Abstract
Concept bottleneck models (CBMs) can improve the transparency of cancer image diagnostic prediction by expressing predictions through radiological concepts. However, their dependence on instance-level concept annotations limits practical applicability. We propose a prior-guided hybrid CBM that integrates limited concept annotations, class-conditional concept distribution matching on unannotated patients, and prior initialization of the concept-to-diagnosis head. We evaluate the method on CBIS-DDSM mammographic masses and calcifications and LIDC-IDRI pulmonary nodules across 0-100% concept annotation. In the clinically relevant 0-20% annotation regime, the hybrid CBM consistently improves mean concept AUC over a matched standard CBM, while maintaining diagnostic performance close to black-box models. At 10% annotation specifically, concept AUC increases from 0.619 to 0.741 for masses, from 0.650 to 0.787 for calcifications, and from 0.597 to 0.642 for pulmonary nodules. Ablation experiments identify prior initialization as the main component contributing to improved concept detection, likely by stabilizing the concept-to-diagnosis head. Zero-shot VLMs remain insufficient for reliable fine-grained tumor-level concept prediction. These findings suggest that structured priors can substantially reduce the annotation burden of interpretable cancer imaging models.