Key result
Self-supervised cardiac CT foundation model delivers high segmentation accuracy with a ~0.96 mean Dice score.
Why the study?
Manual segmentation of cardiac structures on cardiac computed tomography is prohibitively time-consuming and prone to intra- and inter-observer variability, motivating the development of an automated 3D deep-learning foundation model.
Does a deep-learning foundation model trained with a human-in-the-loop approach and self-supervised learning improve automated 3D segmentation accuracy of cardiac substructures on cardiac CT scans compared to existing open-source tools?
Does a deep-learning foundation model trained with a human-in-the-loop approach and self-supervised learning improve automated 3D segmentation accuracy of cardiac substructures on cardiac CT scans compared to existing open-source tools?
An openly accessible, self-supervised deep-learning foundation model trained with human-in-the-loop active learning achieves a 0.96 mean Dice score for automated 3D segmentation of cardiac CT substructures.
Background Cardiac computed tomography (CCT) provides high-resolution data essential for personalized medicine and clinical evaluations; however, manual segmentation of cardiac structures is prohibitively time-consuming and prone to intra- and inter-observer variability. Objective To manually label a comprehensive, multi-center dataset using a human-in-the-loop approach and develop a deep-learning foundation model for the precise, automated 3D segmentation of diverse cardiac substructures, including chambers, vasculature, and adipose tissues. Methods A total of 1,559 CT scans were used to develop several segmentation models within the nnUZoo framework, including Swin-Mamba, U2Net, SwinUNETR, and the default nnU-Net. A human-in-the-loop approach with active learning was implemented to iteratively train models and refine inference labels with the help of two physicians for the coronary arteries, myocardium, major vascular structures, and peri-/epicardial fat. To ensure robustness, CT-specific augmentations, such as metal artifacts and calcification simulations were employed. Additionally, a self-supervised foundation model was trained on more than 60,000 cases to enhance performance, particularly in data-scarce scenarios. Results The model achieved a mean Dice score of 0.96 across all substructures, outperforming existing open-source tools. Specialized augmentations ensured robust performance in complex cases, while the foundation model provided an average 5 percent performance boost in low-data regimes. Conclusion We present a highly accurate, automated 3D segmentation tool for CCT. By enabling the consistent extraction of quantitative metrics, this model facilitates large-scale clinical research. Both the models and the labeled datasets have been made openly accessible.ct-seg-esc
No takes yet. Share an insight, caveat, or question.
Kazaj et al. (2026) studied Cardiac computed tomography whole-heart segmentation (n=1,559). Deep-learning foundation model with human-in-the-loop active learning for 3D cardiac CT segmentation vs. Existing open-source segmentation tools was evaluated on Mean Dice score across all cardiac substructures. The self-supervised cardiac CT foundation model achieved a 0.96 mean Dice score across substructures and provided a 5% performance boost in low-data regimes.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: