Unsupervised Domain Adaptation for Calcification Classification in Mammography Across Multi-Site Datasets

2026-07-07

AI Summary

This study addresses the domain shift in multi-center mammography caused by variations in imaging devices and protocols, which hinders the generalization of models for classifying malignant versus benign calcifications. To mitigate this issue, the authors propose a two-stage unsupervised domain adaptation framework. In the first stage, they jointly leverage Adaptive Instance Normalization (AdaIN) and CycleGAN to perform unsupervised style transfer, synthesizing training samples in the target domain’s visual style. In the second stage, a Swin Transformer V2 model is trained with supervision on these stylized images for classification. This work presents the first integration of AdaIN and CycleGAN for domain adaptation in breast imaging and demonstrates its efficacy on real-world, multi-vendor, multi-modal clinical datasets (EMBED and Duke), improving AUC from 0.68 to 0.72 and 0.73, respectively—significantly outperforming baseline approaches.

Abstract

Deep learning-based computer-aided diagnosis (CAD) systems have shown strong performance in breast cancer diagnosis, particularly for classification tasks in mammography. However, domain shifts across multi-site datasets remain a challenge, especially when models are applied to unseen domains. In this work, we proposed a calcification classification framework to improve malignant versus benign breast disease classification across multi-site mammography datasets. The framework consisted of two components: (1) an unsupervised domain adaptation module based on style transfer models (AdaIN and CycleGAN) to generate vendor-specific and technique-specific training samples without additional annotations, and (2) a supervised classification module using Swin Transformer V2 as the backbone. We evaluated the proposed method on three datasets: cross-validation on OPTIMAM (National Health Service, United Kingdom; n=2994), followed by external validation on EMBED (Emory University; n=125), and Duke Calcification Dataset v1 (n=788). These datasets cover multiple vendors and include both full-field digital mammography and synthetic 2D images derived from digital breast tomosynthesis. The proposed framework improved cross-site performance for both EMBED (AUC 0.68 to 0.72) and the Duke Calcification Dataset (AUC 0.68 to 0.73). These findings indicate that domain adaptation can reduce domain shifts and improve the generalization for calcification classification across multi-site datasets.

Problem

• Unsupervised Domain Adaptation
• Calcification Classification
• Mammography
• Multi-site Datasets
• Domain Shift

Innovation

• unsupervised domain adaptation
• style transfer
• Swin Transformer V2
• calcification classification
• multi-site mammography