现代骨干网络改进用于乳腺X线摄影分类和病灶定位的多任务DETR
Modern Backbones Improve Multi-task DETR for Mammography Classification and Lesion Localization
浏览论文内容
中文总结 AI 辅助
该研究采用多任务DETR框架，在OPTIMAM和SGM1k数据集上对比现代骨干网络与旧有ResNet特征的性能，发现ConvNeXtV2和DINOv3表现最优，骨干网络质量是多任务乳腺X线摄影的关键因素。
中文摘要 AI 辅助
联合检查级预测和候选区域定位可提升AI在乳腺X线摄影中的辅助实用性。本研究采用多任务DETR框架，其中共享表征同时支持图像级恶性程度预测和病灶定位，并在OPTIMAM及经活检确认的SGM1k队列上评估其性能。在两个数据集上，现代骨干网络均持续优于早期ResNet风格特征，ConvNeXtV2和DINOv3表现最强，而MambaVision竞争力较弱。在OPTIMAM上，ConvNeXtV2取得最佳整体性能，AUC达97.96%、灵敏度99.89%、mAP@.5为25.08%、recall@.25为74.38%；在SGM1k上，DINOv3表现最佳，AUC达90.97%、灵敏度86.28%、特异性82.00%、mAP@.5为27.04%、recall@.25为77.32%。这些发现表明，骨干网络质量是有效多任务乳腺X线摄影的关键因素，ConvNeXtV2在该框架中成为特别强大且适配性良好的CNN骨干网络。
英文摘要
Joint exam-level prediction and candidate-region localization may improve the usefulness of AI support in mammography. We study this setting using a multi-task DETR framework, where shared representations support both image-level malignancy prediction and lesion localization, and evaluate its performance on OPTIMAM and a biopsy-confirmed SGM1k cohort. Across both datasets, modern backbones consistently outperformed older ResNet-style features, with ConvNeXtV2 and DINOv3 giving the strongest overall results, whereas MambaVision was less competitive. On OPTIMAM, ConvNeXtV2 achieved the best overall performance, reaching 97.96% AUC, 99.89% sensitivity, 25.08% mAP@.5, and 74.38% recall@.25. On SGM1k, DINOv3 gave the strongest overall results, with 90.97% AUC, 86.28% sensitivity, 82.00% specificity, 27.04% mAP@.5, and 77.32% recall@.25. These findings suggest that backbone quality is a critical factor in effective multi-task mammography, with ConvNeXtV2 emerging as a particularly strong and well-matched CNN backbone for mammography in this framework.
发表机构
- University of Technology Sydney（悉尼科技大学）
- Saigon Precision Medicine Research Center（西贡精准医学研究中心）
- College of Medicine, Taipei Medical University（台北医学大学医学院）
- Le Qui Don Technical University（黎文谍技术大学）
- National Central University（中央大学）
- AIBioMed Research Group, Taipei Medical University（台北医学大学AIBioMed研究组）
机构由 AI 辅助整理，请以论文原文为准。
