AI-Powered Ultrasound Effective in Early Detection of Cardiomyopathies: Lancet
A recent study published in The Lancet highlighted the potential of AI-driven cardiac imaging to identify under-diagnosed cardiomyopathies earlier than ever before. Artificial intelligence (AI) is making major steps in healthcare, now transforming point-of-care ultrasonography (POCUS) to detect serious heart conditions like hypertrophic cardiomyopathy (HCM) and transthyretin amyloid cardiomyopathy (ATTR-CM) at the bedside and in community settings.
The research team from Yale–New Haven Health System (YNHHS) and Mount Sinai Health System (MSHS) developed a sophisticated video-based convolutional neural network adapted specifically for POCUS. By leveraging data from 290,245 echocardiographic videos, the AI model was trained to recognize subtle abnormalities in heart structure and function. Unique approaches, like video augmentation and a customized loss function tailored to image quality, enabled the model to excel in differentiating disease states.
From November 2023 to March 2024, this study evaluated over 33,000 patients from YNHHS and nearly 6,000 patients from MSHS, processing a combined total of more than 90,000 eligible POCUS videos. The AI model delivered impressive results by accurately identifying HCM and ATTR-CM across single-view POCUS scans.
The AI achieved an area under the receiver operating characteristic curve (AUROC) of 0.903 at YNHHS and 0.890 at MSHS for HCM detection through apical-4-chamber views. Similarly, ATTR-CM detection achieved AUROCs of 0.907 and 0.972, respectively, for parasternal views. Also, AI-POCUS screenings flagged 58% of HCM cases and 46% of ATTR-CM cases more than 2 years before formal diagnosis, providing a crucial window for earlier intervention.
Further analysis of participants without known cardiomyopathies revealed that the patients in the highest probability group for disease detection faced significantly increased mortality risks. For hypertrophic cardiomyopathy, the risk was 17% higher when compared to the lowest group. For ATTR-CM, the risk was even greater at 32%.
The findings illuminate the potential for scalable AI-driven screening methods to improve cardiac care by identifying high-risk individuals early. With its capacity for use in bedside and remote settings, this AI framework could revolutionize how heart disease is detected and managed globally. Overall, the outcomes of this study give hope that continued refinement of AI models and widespread adoption of POCUS will further democratize access to advanced cardiac care, ultimately saving lives through early detection and timely treatment.
Reference:
Oikonomou, E. K., Vaid, A., Holste, G., Coppi, A., McNamara, R. L., Baloescu, C., Krumholz, H. M., Wang, Z., Apakama, D. J., Nadkarni, G. N., & Khera, R. (2025). Artificial intelligence-guided detection of under-recognised cardiomyopathies on point-of-care cardiac ultrasonography: a multicentre study. The Lancet. Digital Health, 7(2), e113–e123. https://doi.org/10.1016/s2589-7500(24)00249-8
