HeartLens

Through an EXCEL Research Committee effort, HeartLens is a collaborative project involving the UK Divisions of Radiology & Cardiology, the Department of Computer Science, EduceLab, and CAAI. The project aims to develop an AI-based tool to support scalable cardiovascular screening in routine chest imaging, assisting radiologists in reviewing CT scans, detecting CAC, and informing cardiologists’
diagnostic and treatment decisions.

Coronary artery calcium (CAC) scoring helps assess cardiovascular disease severity but relies on electrocardiogram-gated CT scans, restricting its use to specialized cardiac image settings and limiting the opportunity for early intervention. Early detection of calcification enables more timely, less invasive interventions.

As part of this project, CAAI developed CARD-ViT, a novel self-supervised Vision Transformer framework, leveraging the DINO Transformer Model to identify patterns in gated and non-gated CT scans without requiring manual labeling. By analyzing UK and public Stanford University datasets, the model highlights calcified areas in coronary arteries, color-coding them by severity to aid clinical review. These visual cues support tasks such as classification and segmentation and can integrate with LLMs to support diagnostic reporting. Beyond CAC scoring, the tool has potential applications for detecting other heart conditions including stenosis, and the long-term goal is to make it available via EPIC integration. This would streamline clinical decision-making, improve early disease detection, and foster further research.

Our approach utilizes a self-supervised learning framework based on the DINOv2 model, a Vision Transformer architecture designed for computer vision tasks. Our local, secured instance of the DINOv2 model was trained on a dataset of public and internal CT imaging data. Public CT imaging datasets from Stanford were used in the training and testing process, as well as internal, University of Kentucky, CT data extracted by CCTS.  

To generate feature maps for calcified areas, we employ a register-based approach. Register maps are feature maps generated by our specialized DINOv2 model. We apply PCA analysis to these four register maps to produce a feature map that highlights and precisely locates calcified areas. This specialized DINOv2 model with registers is fully compatible with CT scans and has been trained using our proprietary Label-Guided data augmentation technique. Additionally, the model is designed to work with single-channel inputs, which is appropriate since most medical images are represented as single-channel images.  

The resulting model is capable of detecting, segmenting, and classifying risk of coronary artery calcification (CAC) in CT scans with high accuracy, making it a valuable tool for clinicians and researchers in the field of cardiovascular medicine. 

HeartLens is still in development. Next steps include a robust clinical evaluation, where our partners will assess the impact of AI-assisted CAC detection on patient care and decision-making.

HeartLens is still in development.

Please check out our publications to learn more about this work:

Gokmen, M.S., Ozcan, C., Haque, M.N. et al. DINO-LG: Enhancing vision transformers with label guidance for coronary artery calcium detection. Med Biol Eng Comput 64, 1249–1266 (2026). https://doi.org/10.1007/s11517-026-03523-1

Gokmen, M.S., Haque, M.N., et al. A Framework for Cross-Domain Generalization in Coronary Artery Calcium Scoring Across Gated and Non-Gated Computed Tomography. arXiv Pre-print (2026). https://arxiv.org/abs/2602.21935

CAAI Team members Mahmut Gokmen, Evan Damron, Caroline Leach, Mitchell Klusty, and Dr. Cody Bumgardner worked on HeartLens.