Using AI To Diagnose Heart Disease

Patients that present with chest pain are on the majority referred for stress echocardiogram ultrasounds where patient hearts are speed up by cardiologists with exercise or pharmaceutical agents, taking images while at rest and stress before comparing them to make diagnosis by eye.

Echocardiogram diagnostic accuracy in clinic is about 80%, which can translate to roughly 1 in 5 patients being misdiagnosed. Patients can be sent for unnecessary procedures or sent home when they needed procedures, either case can be fatal. Misdiagnosis can cost an estimated $300 million a year. Diagnostic support tools could be helpful to clinicians assisting in indicating patients with significant disease and provide clinicians with confirmation and reassurance of diagnosis.

Topological Analysis platform has been developed to help clinicians more accurately diagnose heart disease, providing multifeatured analysis of heart wall geometry, much of which hasn’t been measured before in cardiovascular imaging context. The analysis platform is a type of AI machine learning that is used to develop an algorithm linking patient outcome data with combinations of features to predict whether patients have significant disease, delivering highly accurate and reproducible results assessing presence of coronary artery disease; which works with existing echocardiogram technology, fitting directly into clinician workflow to provide support of accurate diagnosis. Trials with this technology are being run in 6 hospitals and will be expanding to 20 by the end of 2018 with the goal of bringing it to market within the next year to help improve patient outcomes and savings cost to both the patient and healthcare system.