A Korean research team has developed an AI-based technology that identifies the causes of cardiac hypertrophy using only echocardiography images.
A research team led by Professor Moon In-ki of the Department of Cardiology at Soon Chun Hyang University Hospital Bucheon announced on Monday that it has developed an AI-based echocardiography analysis technology in collaboration with Ontact Health, an AI healthcare company.
The research team trained the AI using 867 cardiac ultrasound images collected from multiple institutions in Korea and evaluated the AI's performance on 619 independent validation patients. This technology automatically segments the heart in ultrasound images and extracts approximately 20,000 image features, including heart wall thickness, shape, and tissue texture, for diagnostic purposes.
As a result, the developed AI model accurately determined the presence of cardiac hypertrophy with very high accuracy and showed high diagnostic accuracy by cause of disease, including 96 percent for hypertrophic cardiomyopathy, 89 percent for amyloidosis, and 86 percent for hypertensive heart disease.
Notably, the consistency with existing ultrasound-based diagnostic accuracy was only 33 percent for hypertensive heart disease, but it improved to 75 percent with the AI model.
Additionally, the AI model visualizes the image information used as the basis for diagnosis and explains it to medical staff, enabling it to serve as a reliable diagnostic aid in actual clinical practice. The research team explained that the model can perform stable analysis even when ultrasound equipment or image quality varies between hospitals, demonstrating its compatibility with different equipment.
“AI quantifies subtle differences in the texture and shape of the heart for diagnosis, enabling early diagnosis without expensive MRI or invasive tissue testing,” Professor Moon said. “This will contribute to improving patient care and reducing medical costs.”
Through future global multi-institution studies to refine the AI model, it could be broadly applied to distinguish rare heart diseases or pathological cardiac hypertrophy in athletes, Moon added.
The findings of this study were published in the American Heart Association (AHA) journal “Circulation: Cardiovascular Imaging.”