DBT-DINO:基于基础模型的数字乳腺断层摄影分析
在下游任务中,DBT-DINO乳腺密度分类准确率为79%,优于ImageNet预训练基线DINOv2的73%(P<.001);5年乳腺癌风险预测AUC为0.78,与基线0.76无统计学差异(P=.057);病灶检测敏感度62%,与基线67%无统计学差异(P=.60)。
要点
- DBT-DINO在乳腺密度分类上准确率达79%,显著优于ImageNet基线73%,可用于辅助密度报告。
- 5年乳腺癌风险预测AUC为0.78,与ImageNet基线无统计学差异,临床增益有限。
- 病灶检测敏感度62%,低于ImageNet基线67%且无统计学差异,不适用于辅助检出。
核心结论
- DBT-DINO在乳腺密度分类任务中表现良好,但在5年乳腺癌风险预测和病灶检测中与ImageNet预训练基线无统计学差异,提示针对局部检出任务的域特定预训练仍需进一步优化。
研究设计与样本
- 回顾性研究,数据来源于Mass General Brigham,采集时间为2011年3月至2024年2月。
- 性能比较采用McNemar和DeLong检验,基线为ImageNet预训练的DINOv2。
关键影像表现/征象
- 乳腺密度分类任务中,DBT-DINO在997例检查中正确分类786例,高于基线DINOv2的728例。
- 病灶检测任务中,DBT-DINO检测出136个病灶中的84个(敏感度62%),低于DINOv2的91个(敏感度67%)。
诊断标准与关键数值
- 乳腺密度分类:DBT-DINO准确率79%(786/997),DINOv2基线准确率73%(728/997),P<.001。
- 5年乳腺癌风险预测:DBT-DINO的AUC为0.78,DINOv2的AUC为0.76,P=.057,无统计学差异。
- 病灶检测:DBT-DINO平均敏感度62%(84/136个病灶),DINOv2平均敏感度67%(91/136个病灶),P=.60。
临床应用与阅片要点
- 乳腺密度分类是DBT-DINO目前最具临床应用潜力的下游任务,但需在模型泛化验证后考虑辅助报告。
- DBT-DINO在病灶检测中的敏感度未优于通用ImageNet预训练模型,不宜单独作为检出辅助工具使用。
- 判断5年活检证实的乳腺癌风险时,使用DBT-DINO与通用预训练模型差异无统计学意义,报告时应谨慎解读。
局限性
- 研究为单中心回顾性设计,来自Mass General Brigham,缺乏外部多中心验证。
- 病灶检测评估仅使用393个标注容积、136个病灶,样本量相对有限。
- 病灶检测敏感度低于ImageNet基线,未见域特定预训练带来的优势。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
Background Foundation models show promise in medical imaging but remain underexplored in three-dimensional modalities. Despite the adoption of digital breast tomosynthesis (DBT) in breast cancer screening, no dedicated foundation model currently exists for this modality. Purpose To develop and evaluate a foundation model for DBT (DBT-DINO) and assess the impact of domain-specific pretraining across multiple clinical tasks. Materials and Methods This retrospective study used DBT images from Mass General Brigham acquired between March 2011 and February 2024. Self-supervised pretraining was performed using Meta AI's DINOv2 methodology on more than 25 million two-dimensional sections from 487 975 DBT volumes from 27 990 patients. Three downstream tasks were evaluated: (a) breast density classification using 5000 screening examinations, (b) 5-year risk of developing biopsy-proven breast cancer using 106 417 screening examinations, and (c) lesion detection using 393 annotated volumes. The performance of DBT-DINO was compared with that of ImageNet-pretrained DINOv2 baselines using McNemar and DeLong tests. Results A total of 4981 patients (mean age, 57.76 years ± 11.40 [SD]; 4855 female) were included for density classification, 31 561 patients (mean age, 60.09 years ± 10.47; 31 559 female) were included for risk prediction, and 199 female patients were included for lesion detection. For breast density classification, DBT-DINO achieved 79% (786 of 997 examinations) accuracy, outperforming the DINOv2 baseline (73% [728 of 997 examinations]; P < .001). For 5-year breast cancer risk prediction, DBT-DINO had an area under the receiver operating characteristic curve (AUC) of 0.78 and DINOv2 had an AUC of 0.76 (P = .057), showing no evidence of a difference. In lesion detection, DINOv2 had an average sensitivity of 67% (91 of 136 lesions), whereas DBT-DINO had a sensitivity of 62% (84 of 136 lesions) (P = .60), again with no evidence of a difference. Conclusion DBT-DINO demonstrated strong performance in breast density classification; however, there was no evidence of a difference compared with the ImageNet baseline in 5-year breast cancer risk prediction or lesion detection, suggesting that domain-specific pretraining for localized detection tasks required further refinement. © RSNA, 2026 Supplemental material is available for this article. See also the editorial by Wu in this issue.
引用
- Dorfner, Dorster, Connolly, et al. DBT-DINO: Toward Foundation Model-Based Analysis of Digital Breast Tomosynthesis. Radiology. 2026. PMID: 42550018. doi:10.1148/radiol.254034
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。