发表机构
Siemens Healthineers; Technical University of Munich (TUM); TUM University Hospital; Imperial College London（西门子医疗; 慕尼黑工业大学; 慕尼黑工业大学附属医院; 帝国理工学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究提出CMRVision这一CMR专用基础模型，经3600万张多中心CMR图像预训练后，在多任务分割、视角分类等下游任务上性能优于现有基线，展现出良好泛化潜力。
AI 中文摘要
心脏磁共振（CMR）成像可提供心脏解剖结构、功能及组织表征的互补信息，涵盖多个序列与视角。本研究针对二维CMR开展基础模型预训练研究，引入CMRVision这一CMR专用基础模型，该模型基于DINOv3风格的自监督学习，在包含3600万张CMR图像的多中心、多序列队列上训练而成。我们系统评估了面向领域特定预训练的架构与训练设计选择。对CMRVision在两项下游任务上进行评估：一是电影序列、钆延迟强化（LGE）序列及 mapping 序列的多任务分割；二是电影视角分类。实验表明，CMR专用预训练、更小的patch尺寸及patch级目标函数可持续提升下游性能。在多任务分割基准测试中，CMRVision取得最强综合性能，优于此前的自然图像（NI）、医学图像、监督式及CMR基础模型基线。在各结构与序列上的提升虽温和但稳定，左心室（LV）的Dice分数范围为0.940-0.967，心肌为0.855-0.905，右心室（RV）为0.929，左心房（LA）为0.920，右心房（RA）为0.931；在LGE序列与mapping图像的心肌分割上提升最为显著。在未见过的LGE长轴视角的零样本分割任务中，该模型取得平均Dice分数0.692，展现出跨视角泛化能力。在电影视角分类任务中，CMRVision取得最高平均准确率0.906，优于文献报道的此前方法。这些结果凸显了CMRVision在支持跨多序列与视角的稳健、可泛化心脏磁共振分析方面的潜力。
英文摘要
Cardiac magnetic resonance (CMR) imaging provides complementary information on cardiac anatomy, function, and tissue characterization across multiple sequences and views. In this work, we investigate foundation model pretraining for 2D CMR and introduce CMRVision, a CMR-specific foundation model trained using DINOv3-style self-supervised learning on a multi-center, multi-sequence cohort of 36 million CMR images. We systematically evaluate architectural and training design choices for domain-specific pretraining. CMRVision is evaluated on two downstream tasks: multi-task segmentation across cine, late gadolinium enhancement (LGE), and mapping sequences, and cine view classification. Our experiments show that CMR-specific pretraining, smaller patch sizes, and patch-level objectives consistently improve downstream performance. Across a multi-task segmentation benchmark, CMRVision achieved the strongest overall performance, outperforming prior natural-image (NI), medical-image, supervised, and CMR foundation model baselines. Improvements were modest but consistent across structures and sequences, with Dice scores ranging from 0.940-0.967 for LV and 0.855-0.905 for myocardium, and reaching 0.929 for RV, 0.920 for LA, and 0.931 for RA. The largest gains were observed for myocardium segmentation in LGE and mapping images. In a zero-shot segmentation task on unseen LGE long-axis views, the model achieved an average Dice score of 0.692, demonstrating cross-view generalization. For cine view classification, CMRVision achieved the highest average accuracy (0.906), compared to prior methods reported in the literature. These results highlight the potential of CMRVision to support robust and generalizable cardiac MRI analysis across multiple sequences and views.
CommentsAccepted at MedAGI 2026 (peer-reviewed workshop at MICCAI 2026)
