Artificial Intelligence Algorithm Useful For Detecting Mitral Regurgitation Using Electrocardiography

A novel validated artificial intelligence (AI) algorithm helps to detect mitral regurgitation using electrocardiography, according to new analysis results.

“Screening and early diagnosis of mitral regurgitation are crucial for preventing irreversible progression of  mitral regurgitation,” the authors, publishing in the Journal of Echocardiography, wrote in their abstract. “In this study, we developed and validated an AI algorithm for detecting mitral regurgitation using electrocardiography.”

The retrospective cohort study included data from two centers. The researchers trained an AI algorithm using 56,670 electrocardiograms (ECGs) from more than 24,000 patients. They also performed internal validation with 3,174 ECGs of 3,174 patients from one of the centers, and external validation using 10,865 EGCs from 10,865 patients from another center. The team used  500 Hz ECG raw data for predictive variables. They also used a sensitivity map to identify the regions of ECG that had the greatest impact on the decision-making of the AI algorithm. The primary study endpoint was a diagnosis of significant mitral regurgitation (moderate to severe) confirmed by echocardiography.

AI Identifies Those at High Risk

According to the results, the area under the receiver operating characteristic curve of the AI algorithm was 0.816 for the internal validation and 0.877 for the external validation using a 12-lead ECG for the detection of mitral regurgitation (a single-lead ECG was 0.758 for internal and 0.850 for external). In the more than 3,000 patients who did not have mitral regurgitation, the patients that were identified by the algorithm as high-risk had a higher risk for developing the condition compared to the low-risk group (13.9% vs. 2.6%, respectively; P<0.001) during study follow-up. The AI algorithm focused on P-wave and T-wave in patients with mitral regurgitation, and on the QRS complex in those without mitral regurgitation.

“The proposed AI algorithm demonstrated promising results for mitral regurgitation detecting using 12-lead and single-lead ECGs,” the authors concluded.

Highlights

• AI algorithm trained on detecting mitral regurgitation on more than 56,000 ECGs from two hospitals.
• Patients identified as “high-risk” by the AI algorithm had increased chance of developing mitral regurgitation.
• The AI Algorithm focused on P-waves and T-waves in those with mitral regurgitation, and on the QRS complex in those without.