Machine learning aids diagnosis of athletes' heart conditions

Heart monitor

Computer algorithms can interpret echocardiographic images and distinguish between two similar heart conditions affecting young athletes, according to research published in the Journal of the American College of Cardiology.

Researchers from the Icahn School of Medicine at Mount Sinai tested three different machine-learning algorithms for their effectiveness in the discrimination of physiological versus pathological hypertrophic cardiomyopathy.

Pathological hypertrophic cardiomyopathy (HCM), in which a portion of the myocardium enlarges, leading to impaired heart function, is the leading cause of sudden death in young athletes. Distinguishing between it and physiological hypertrophy (heart enlargement often due to exercise) generally requires testing for the two conditions with interpretation by a highly trained cardiologist, according to an announcement.

Featured Webinar

Reducing barriers to patient care: A cross-industry collaboration

Optum will bring together cross-industry experts to share a case study detailing how an employer, provider, payer and pharmaceutical company worked together to address migraines, a hard-to-diagnose condition. Learn how this team started with a model to risk stratify, predict undiagnosed, misdiagnosed and mismanaged members and how those analytics were used to enrich engagement for treatment and diagnosis optimization.

RELATED: Data mining tool helps researchers sift through radiology report images, text

"This demonstrates how machine-learning models and other smart interpretation systems could help to efficiently analyze and process large volumes of cardiac ultrasound data, and with the growth of telemedicine, it could enable cardiac diagnoses even in the most resource-burdened areas," said study author Joel Dudley, Ph.D.

With machines’ the ability to consider millions of variables, better algorithms hold the potential to improve diagnostic accuracy, according to a recent article in The New England Journal of Medicine. For instance, it cites computers’ ability to look for anomalies at the pixel level of radiographs.

RELATED: Google's DeepMind Health, radiologists to treat head, neck cancer

However, researchers from Northwestern University and the Rehabilitation Institute of Chicago called machine learning “voodoo” in a study published at at bioRxiv, calling for proper evaluation of the methods used for compiling data for smartphone apps and wearable sensors.

RELATED: Researchers call machine learning 'voodoo,' warn of inaccuracies

Suggested Articles

CMS issued new guidance Friday to help states implement the new interoperability policies in Medicaid and CHIP programs.

Data and analytics company Health Catalyst reported its Q2 results the same week it announced some major deals.

Machines are outshining their human counterparts as AI is applied to an increasing number of healthcare tasks, experts said during Fierce AI Week.