Back to all news
Scripps Researchers Develop AI Foundation Model for ECG-Based Heart Disease Prediction
A new AI foundation model, ECG-CLIP, improves detection and prediction of multiple heart diseases using large-scale ECG and clinician note data.
Key Details
- 1ECG-CLIP was trained on over 1.7 million ECGs from more than 540,000 people, paired with clinician notes.
- 2It outperformed other models in detecting acute myocardial infarction, cardiac amyloidosis, and hypertrophic cardiomyopathy, especially when labeled data was scarce.
- 3ECG-CLIP matched performance of models needing far more labeled examples, operating with about 91% less manually labeled data.
- 4The model excelled in single-lead ECG settings, increasing its usefulness in resource-limited environments.
- 5It also outperformed competitors in predicting atrial fibrillation and short- and long-term outcomes after emergency visits or surgery.
- 6Interpretability was enhanced with saliency maps, making the AI's decision process more transparent for clinicians.
Why It Matters
This work demonstrates how foundation AI models can enhance cardiovascular diagnostics, especially in settings where labeled data or multiple ECG leads are limited. Enhancing model interpretability and adaptability could accelerate clinical AI adoption and improve patient outcomes in cardiac care.
Source
EurekAlert
Related News
•EurekAlert
Rice University Aims to Revolutionize Imaging with Generative AI Cameras
Rice University receives NSF grant to develop AI-driven cameras that reconstruct images from minimal sensor data, promising advancements in efficient, intelligent imaging systems.
•EurekAlert
Precision Psychiatry Symposium Explores Imaging, AI, and Novel Therapies for Schizophrenia
Virginia Tech hosts a symposium focusing on precision psychiatry, spotlighting imaging, biomarkers, and AI to improve schizophrenia treatment.
•EurekAlert
AI Detects Hypertension and Diabetes from 5-Second Facial Videos
A machine-learning algorithm can accurately screen for hypertension and diabetes using brief facial video recordings, offering a contactless diagnostic option.