AI-Enabled Electrocardiogram Shows Promise in Predicting MASLD

A deep learning-based artificial intelligence (AI) model to detect metabolic dysfunction–associated steatotic liver disease (MASLD) through a 12-lead electrocardiogram (ECG) offers a noninvasive screening tool that performs as well as single clinical parameters, investigators reported in Clinical Gastroenterology and Hepatology.

The study retrospectively analyzed adults diagnosed with MASLD in Olmsted County, Minnesota, from 1996 to 2019. Both MASLD cases and controls had ECGs performed within a 6-year window before and 1 year after study entry. The AI-ECG model was trained, validated, and tested on 70%, 10%, and 20% of the cohort, respectively, with external validation in a separate Mayo Clinic cohort. The goal was to assess the performance of the AI-ECG model in identifying MASLD, either independently or combined with clinical factors.

The study included 3468 MASLD cases and 25,407 controls. The AI-ECG model predicted MASLD with an area under the curve (AUC) of 0.69 in the original cohort and 0.62 in the validation cohort. This performance was comparable or superior to models using single clinical parameters like body mass index (AUC 0.71), diabetes (AUC 0.66), or a combination of diabetes, hypertension, and hyperlipidemia (AUC 0.68). The best performance was achieved when combining ECG with body mass index, diabetes, and alanine aminotransferase, yielding an AUC of 0.76 in the original cohort and 0.72 in the validation cohort.

The AI-based ECG model offers a promising, noninvasive tool for MASLD screening, the investigators concluded, with performance comparable to or better than single clinical parameters, though a combination of factors yields the highest accuracy, the investigators concluded.

Reference

Udompap P, Liu K, Attia IZ, et al. Performance of AI-enabled electrocardiogram in the prediction of metabolic dysfunction-associated steatotic liver disease. Clin Gastroenterol Hepatol. Published online August 27, 2024. doi:10.1016/j.cgh.2024.08.009