Video-based cardiac AI tool: siders sinai echoprime

Cedars-Sinai introduced EchoPrime, a video-based visual language AI that analyzes echocardiogram footage and generates written cardiac reports. It features training on over 12 million echo videos combined with clinician interpretation. The system was developed with collaborators at Kaiser Permanente Northern California, Stanford Health Care, Beth Israel Deaconess Medical Center, and Chang Gung Memorial Hospital, and was published in the journal Nature in February 2026.

EchoPrime was evaluated across five international health systems and outperformed previous task-specific underlying models on 23 benchmarks related to cardiac structure and function. The team has released a demo that allows code, weights, and external testing of the model. The model generates verbal summaries for clinician review rather than autonomous diagnosis and was trained on 275,442 studies from 108,913 patients.

For clinicians, EchoPrime promises faster, more consistent echo interpretation that can reduce workload and speed decision-making while maintaining human oversight. Its open release reflects the trend towards reproducible, large-scale clinical AI and creates an avenue for other institutions to validate its performance on community populations.

Image credit: Cedars-Sinai