Algorithms for health

A Weizmann Institute-based innovation in echocardiography—the use of ultrasound waves to investigate heart health—is poised to improve diagnostics for cardiac patients.

An echocardiogram is a type of ultrasound that measures the fraction of blood ejected from the left heart chamber to the rest of the body during a heartbeat. If used for immediate diagnosis, this test, considered the most reliable measure of heart function, could save lives. However, interpreting echocardiogram data is tricky; experienced ultrasound technicians undergo hours of training, years of practice, and participate in ongoing refresher courses to ensure that the reports issued to physicians for the purpose of diagnosis are as accurate as humanly possible.

Now, a new technology, based on the artificial intelligence research of Prof. Yaron Lipman of the Institute’s Department of Computer Science and Applied Mathematics, and developed together with colleagues at the New York University School of Medicine, offers an automated approach that guides the operator in echocardiogram image acquisition, and interprets the results in real time. Called “On-Sight”, the system recently won first place in a competition sponsored by the American Society of Echocardiography, after demonstrating that it generates accurate results that closely match the insights of human technicians.

A fruitful collaboration with NYU

On-Sight was developed together with Mr. Itay Kezurer, a former graduate student in Prof. Lipman’s lab who stayed on as a research associate. Rooted in Prof. Lipman’s expertise in geometrical machine learning, the On-Sight algorithm uses a core neural network approach —inspired by networks in the human brain — together with the computers’ ability to rapidly absorb huge amounts of data while focusing on a particular task.

In the competition, electrocardiograms from 114 individual patients were interpreted both by On-Sight, and by four human echocardiography experts. The results were stunning: the On-Sight algorithm closely matched the performance of the human experts, achieving near-equivalent interpretations of external ejection fractions of those hearts, based on visual ultrasound data.

The On-Sight team also includes Prof. Achi Ludomirsky, MD, a pediatric cardiology expert at NYU School of Medicine, and Dr. Yoram Eshel, the company’s CEO. Together with Prof. Lipman and Mr. Kezurer, they believe that the new approach has the potential to improve workflow and patient care in both established echo labs and patient care settings. In addition to emergency and urgent care settings, they hope that the On-Sight technology will eventually be used in nursing homes, sports clinics, and ambulances, and that it will be rolled out in developing countries where it is a struggle to find and adequately train echocardiogram technicians.

Prof. Lipman is supported by the European Research Council.