MammoExpert: Benchmarking Chain-of-Thought Reasoning in Mammography Diagnosis

2026-06-19

AI Summary

Existing mammography datasets lack large-scale, high-quality annotations with structured diagnostic reasoning, limiting the application of AI in breast cancer subtype identification and interpretable diagnosis. This study introduces the first mammography dataset annotated with a three-stage Chain-of-Thought (CoT) reasoning framework, encompassing 67 WHO pathological subtypes. Nine senior radiologists annotated each case with 42 imaging features, systematically organized into three reasoning phases: initial observations, factual assessments, and diagnostic synthesis. This structured CoT annotation establishes a new benchmark for interpretable AI, yielding a 4% improvement in classification accuracy on the MammoExpert test set. When combined with CBIS-DDSM for training, the approach achieves performance gains of 7.1%, and further enhances accuracy by 6.9% and 6.7% on INBreast and Vindr, respectively.

Abstract

Mammography is an essential tool for breast cancer detection, with millions of examinations conducted annually. However, publicly available high-quality mammography datasets for AI development remain limited in both scale and annotation richness, particularly regarding pathological subtype coverage and structured diagnostic reasoning annotations. In this paper, we present MammoExpert, the first mammography dataset with Chain-of-Thought reasoning annotations across three diagnostic phases: (i) primal observation, (ii) factual assessment, and (iii) diagnostic synthesis. Comprising 2,379 mammography images covering 67 WHO-classified histopathology subtypes, each exam provides 42 radiographic features annotated by nine senior radiologists. We evaluate its performance on the breast lesion classification task, demonstrating superior accuracy and reasonability compared to existing classification models. Combining public dataset CBIS-DDSM with MammoExpert yields 7.1\% classification accuracy improvement, while the training model to learn CoT reasoning achieves another 4\% gain on the MammoExpert test set. Similar improvements are observed on INBreast and Vindr datasets, where the full approach yields accuracy gains of 6.9\% and 6.7\%, respectively. MammoExpert can serve as a benchmark for interpretable breast lesion diagnosis through explicit CoT reasoning.

Problem

• mammography
• dataset limitation
• pathological subtypes
• diagnostic reasoning
• annotation richness

Innovation

• Chain-of-Thought reasoning
• mammography diagnosis
• structured diagnostic annotation
• breast lesion classification
• interpretable AI