Learning Label-Efficient Interpretable Medical Image Diagnosis via Semi-supervised Hypergraph Concept Bottleneck Model

2026-06-01

AI Summary

This work addresses the limited interpretability of existing medical image diagnosis models and the inability of conventional concept bottleneck models to capture high-order dependencies among complex concepts, which often rely heavily on expert-annotated concept labels. To overcome these limitations, the authors propose a semi-supervised hypergraph concept bottleneck model that introduces, for the first time, a dual-level hypergraph structure operating at both image and concept levels. This architecture jointly models high-order concept relationships and generates domain-adaptive pseudo-labels, significantly reducing the need for annotated concept labels while enhancing model interpretability and generalization. Extensive experiments on ultrasound datasets of placental accreta, breast lesions, and dermoscopic images demonstrate the method’s superior diagnostic performance and explanatory capability.

Abstract

Deep learning has revolutionized medical image analysis, delivering exceptional diagnostic accuracy across diverse applications. Yet, the lack of interpretability in its decision-making hinders clinical adoption, particularly in high-stakes medical contexts where transparency is paramount for trustworthiness. For example, in Placenta Accreta Spectrum (PAS), subtle cues in ultrasound imaging challenge reliable diagnosis, rendering black-box models untrustworthy for accurate scoring. To address this, Concept Bottleneck Models (CBMs) offer a promising avenue by embedding clinically meaningful intermediate concepts into the diagnosis pipeline, enabling clinicians to scrutinize and refine model outputs. However, conventional CBMs falter in capturing complex inter-concept dependencies and demand costly, expert-driven concept annotations, limiting their scalability. This study introduces a novel semi-supervised CBM framework designed for medical imaging, which leverages dual-level hypergraph learning to model high-order concept dependencies and generate domain-adaptive pseudo-labels. Our approach achieves superior interpretability and performance by integrating a concept-level hypergraph for enhanced reasoning and an image-level hypergraph for robust pseudo-label generation. Experiments on a newly annotated PAS ultrasound dataset and a breast ultrasound public dataset demonstrate the effectiveness of the proposed concept label-efficient interpretable framework. Its universality is further validated on the dermoscopic image dataset SkinCon. The code is available at https://github.com/scott-yjyang/HyperCBM.

Problem

• interpretability
• concept bottleneck models
• semi-supervised learning
• medical image diagnosis
• label efficiency

Innovation

• Concept Bottleneck Model
• Semi-supervised Learning
• Hypergraph Learning
• Label Efficiency
• Interpretable AI