MammoFlow: Multiview Mammogram Synthesis with Anatomically Consistent Flow Matching

2026-06-26

AI Summary

This study addresses the scarcity of high-quality paired multi-view mammograms, a key limitation in deep learning for breast cancer analysis. The authors propose a novel framework that leverages implicit geometric correspondences to synthesize anatomically consistent craniocaudal (CC) and mediolateral oblique (MLO) views using a pretrained flow-matching model. Their approach introduces an alignment module based on a 2D affine transformation subspace and enforces tissue distribution consistency along the anterior–posterior (AP) axis—from chest wall to nipple—via a pixel-level self-consistency loss formulated with Earth Mover’s Distance (EMD). Radiologist evaluations confirm the high visual and anatomical fidelity of the synthesized images, and when used in downstream classification tasks, these images yield a notable 5% improvement in AUC performance.

Abstract

Multiview mammography relies on paired craniocaudal (CC) and mediolateral oblique (MLO) views to provide complementary projections of a 3D breast volume, enabling precise anomaly localization. However, acquiring high-quality, balanced datasets remains challenging for deep learning applications. We propose a novel method to synthesize multiview mammograms by leveraging the inherent geometric relationship between CC and MLO views. To enforce an implicit 3D consistency prior during generation, we develop an alignment module that searches a 2D affine transformation subspace to establish optimal anatomical correspondence. Leveraging this alignment, we introduce a pixel-space self-consistency loss based on the Earth Mover's Distance (EMD) between the 1D anteroposterior (AP) axis tissue distributions of the generated images. Integrated into a pretrained flow matching model, MammoFlow forces synthesized pairs to share physically plausible tissue distributions from the chest wall to the nipple. To our knowledge, this is the first work to guide multiview mammogram generation using implicit geometric tissue correspondence. Our method demonstrates superior image quality, passes expert radiologist evaluation, and generates physically consistent pairs that improve downstream classification AUC by 5%. Code is available at https://github.com/XYPB/MammoFlow

Problem

• multiview mammography
• data scarcity
• anatomical consistency
• image synthesis
• deep learning

Innovation

• flow matching
• multiview mammography
• anatomical consistency
• Earth Mover's Distance
• image synthesis