Although breakthroughs have been made in the world of deep learning in recent years, few medical products use this technology at this time. As a result, doctors continue to use the same tools that have been used in decades.
To find a solution to this problem, Professor Yael Yaniv’s research group at the Faculty of Biomedical Engineering collaborated with Professor Alex Bronstein and Professor Assaf Schuster’s research group at the School of Computer Science.Currently, under their co-supervision, research by doctoral students Yonatan Elul and Aviv Rosenberg Minutes of the National Academy of Sciences.. In this article, the author presents an AI-based system that automatically detects disease based on hundreds of electrocardiograms, which is currently the most popular technology for diagnosing heart disease.
The new system uses expansion to automatically analyze the electrocardiogram (ECG) neural network— The best tool for deep learning today. These networks learn different patterns by training with large numbers of samples, and the system developed by researchers has over 1.5 million ECG segments sampled from hundreds of patients in hospitals in different countries. Was trained in.
Developed over a century ago, the electrocardiogram provides important information about the conditions that affect the heart and is quick and non-invasive. The problem is that printouts are currently being interpreted by human cardiologists, and therefore those interpretations are inevitably permeated by subjective factors. As a result, many research groups around the world are working on developing systems that automatically and efficiently interpret printouts. In addition, these systems can identify pathological conditions that human cardiologists cannot detect, regardless of experience.
The system, developed by Technion researchers, was built according to the requirements defined by cardiologists, and its output includes estimation of uncertainties in results, indication of suspicious areas on ECG waves, and uncertain results. And contains alerts about increased risk of pathology. The ECG signal itself. The system is sensitive enough to provide alerts for patients at risk of arrhythmia, even if the ECG printout does not show arrhythmia, and the false alarm rate is negligible. In addition, the new system uses accepted cardiology terminology to explain its decisions.
Researchers hope that this system can be used for interpopulation scans for early detection of people at risk for arrhythmias. Without this early diagnosis, these people are at increased risk of heart attack and stroke.
Yonatan Elul et al, Deep Learning-Based ECG Analysis Addresses the Unmet Needs of Clinicians in AI Systems Introduced in Cardiology. Minutes of the National Academy of Sciences (2021). DOI: 10.1073 / pnas.2020620118
Technion-Israel Institute of Technology
Quote: A clinically viable way to develop AI-based tools for medical use (July 15, 2021) is https: //medicalxpress.com/news/2021-07-clinically-viable-ai- Obtained from based-tools-medicine on July 15, 2021. html
This document is subject to copyright. No part may be reproduced without written permission, except for fair transactions for personal investigation or research purposes. The content is provided for informational purposes only.
A clinically viable way to develop AI-based tools for medicine
Source link A clinically viable way to develop AI-based tools for medicine