CMR scans deliver unmatched detail on heart structure and function, but reading them takes real expertise and often over 40 minutes per patient. A new AI tool called CMR-CLIP aims to change that.
Why Heart Imaging Needs Help
Cardiologists and radiologists with advanced training handle these complex scans. Yet reporting varies widely, which can muddy decisions for other doctors. CMR-CLIP (Cardiovascular Magnetic Resonance-Contrastive Language Image Pretraining) pairs video images with real clinical reports to learn fast analysis.
Strong Real-World Results
Built by Carnegie Mellon University and Cleveland Clinic researchers, the model trained on over 13,000 studies from Cleveland Clinic (2008-2022). It shines on left ventricular issues:
- 88.5% accuracy for nonischemic cardiomyopathy
- 88.0% for ischemic cardiomyopathy
- 96.2% for cardiac amyloidosis
- 98.6% for hypertrophic cardiomyopathy
It even handles zero-shot diagnosis for untrained conditions and beats other CLIP systems by over 32% on key problems like myocardial fibrosis. One example often matches what others need 32 examples for.
“CMR-CLIP not only performed well on diseases it had trained on, but also demonstrated ‘zero shot’ capabilities.”
Future Potential
This could standardize reports, support screening where experts are scarce, and build a searchable training library for doctors. The team notes more diverse data and workflow testing are next steps. Code is public on GitHub.
