Radiologists outperform AI in breast cancer imaging but AI aids malignancy risk estimates
From Density to Biopsy Decisions and Malignancy Prediction: A Benchmark Study of Multimodal Large Language Models Against Radiologists in Digital and Contrast-Enhanced Mammography
Computer Vision and Pattern Recognition
Summary
Mammogram images help doctors find and judge breast cancer risk by looking at breast density, cancer likelihood, and whether a biopsy is needed. This study compared four computer models that read mammograms with real radiologists. Doctors generally did better at sorting categories like density and biopsy decisions, but some AI models came close when estimating the exact cancer risk, especially when given extra image details. This shows AI could assist doctors by providing additional risk estimates during diagnosis.
What this means in practice
- •For medical imaging teams: Use AI models with lesion masks to support radiologists by providing precise continuous malignancy risk estimates in breast cancer screening.
- •For healthcare ai developers: Develop multimodal AI tools focused on malignancy probability estimation in mammograms to complement but not replace radiologist categorical assessments.$Commercial implications: It enables creation of diagnostic support software that enhances breast cancer risk assessment accuracy for hospitals and clinics.
Authors
Ali Abbasian Ardakani, Afshin Mohammadi, Taha Yusuf Kuzan, Beyza Nur Kuzan, Alisa Mohebbi, Masume Behruzi, Hamid Khorshidi, Ashkan Ghorbani, Elham Asadiara, Zeinab Khorshidi Lotfi, Ansar Rahman, Nedim Christoph Beste, U. Rajendra Acharya, Sepideh Hatamikia
Abstract
Purpose: To compare four multimodal large language models (MLLMs) with radiologists of varying expertise in breast density assessment, BI-RADS assessment, biopsy candidacy determination, and continuous malignancy probability estimation using digital mammography (DM) and contrast-enhanced mammography (CEM). Methods: This study included 179 women with paired DM/CEM examinations and reference standards. Four MLLMs (ChatGPT-5.2, Gemini-3.1 Pro, Sonnet-4.6, Muse Spark) interpreted images with and without masks; three radiologists interpreted non-masked images. Results: For binary density classification on DM, radiologist accuracies ranged from 55.81% to 78.60%, exceeding most MLLM values (62.33%-71.63%), while masks added limited benefit. Five-category BI-RADS accuracies were higher for radiologists on DM (56.74-67.44%) and CEM (62.33-82.79%) compared with MLLMs (DM 31.16-45.12%; CEM 40.00-55.81%). Binary biopsy-candidacy accuracies were likewise higher for radiologists (DM 85.12-89.77%; CEM 86.98-92.09%) than for MLLMs (DM 61.39-75.35%; CEM 69.30-82.79%), although CEM improved performance across all readers. Lesion masks substantially improved MLLM continuous malignancy-probability accuracies from 64.65%-71.63% to 72.56-78.60% on DM and from 67.91%-77.21% to 72.56%-81.86% on CEM, approaching radiologist ranges (DM 63.72-82.79%; CEM 81.86-88.84%). The corresponding AUCs for the top masked models overlapped those of the human readers. Overall, Muse Spark, followed by Sonnet-4.6, demonstrated the strongest performance among the MLLMs across domains. Conclusion: Radiologists generally outperformed MLLMs in categorical tasks, while selected masked models approached human performance for continuous malignancy probability estimation, suggesting a potential adjunctive role.