An artificial intelligence (AI) smartphone app for heart disease detection is expected to debut soon.
A research team led by cardiologists Kang Shi-hyuk and Seo Jung-won at Bundang Seoul National University Hospital and KAIST professor Shin In-shik recently developed and tested a mobile app, CPstethoscope, that automatically detects a heart disease such as aortic stenosis based on heartbeat sound recording. The application uses deep learning and a convolutional neural network algorithm to distinguish a normal heartbeat sound from that of a heart disease patient.
In its recent test, the research team recorded the heartbeat sounds of 46 patients with a medium age of 65.5 for 10 seconds by using an existing electronic stethoscope and the microphones of three different Samsung Electronics and LG Electronics smartphones and then compared the sounds. The AI app removed noise and obtained sounds of acceptable quality in 30 patients. Based on the result, the app detected a heart disease from the 30 patients with an accuracy of 87% to 90%.
The app distinguished a normal sound from that of an aortic stenosis patient. However, it failed to detect atrial fibrillation and the heart noise during ventricular enlargement. It was affected by much background noise and a low frequency amplitude of the recorded heartbeat sound, too.
“Based on the recent research, we confirmed the practical utility of and what should be added to the AI app for detecting a heart disease without a cardiologist before treatment or hospitalization,” professor Kang Shi-hyuk explained, adding, “We are currently improving the app to sort interpretable heart sounds and distinguish atrial fibrillation and the heart noise during ventricular enlargement as well.” He continued to say, “Cardiologists and nurses with less experience will be able to diagnose patient conditions more accurately with our algorithm applied to electronic stethoscopes.”
Details of the research have been published in JMIR mHealth and uHealth, an international journal of mobile and ubiquitous health.