AI measurements of heart scans more strongly linked to patient survival than clinician measurements

New research led by City St George's, University of London and University College London reveals that artificial intelligence (AI) measurements of heart function, specifically ejection fraction derived from MRI scans, are more strongly associated with patient survival than the assessments made by clinicians. This finding, published in The Lancet Digital Health, suggests that AI could serve as a valuable initial step in evaluating heart function.

The research, which analyzed heart MRI scans from 27,382 patients across the U.K. and the United States, showed that AI-derived ejection fraction measurements were slightly but significantly better at predicting which patients had a higher risk of death. The AI tool also performed the analysis in just 2-3 seconds per patient, a stark contrast to the approximately 13 minutes required for expert clinical analysis.

While AI errors were infrequent, occurring in around 1% of patients, leading to about 0.08% of images being misclassified, the researchers emphasize that human oversight remains crucial. Clinicians should verify AI results and integrate them within the larger clinical context. Dr. Rhodri Davies, lead author and inventor of the AI tool, explained that even minor improvements in accuracy can have a substantial impact when applied to thousands of patients.

AI technology is not meant to replace clinicians but to provide faster, more consistent measurements at a lower cost, facilitating more streamlined clinical decision-making that ultimately benefits patients. The study's authors propose that combining AI measurements from heart scans with electrocardiogram (ECG) results could help detect heart muscle issues earlier, enabling timely treatment before significant damage occurs.