In a 2024 cross-sectional study in Kenya, an AI-ECG model screened 1,444 adults for left ventricular systolic dysfunction (LVSD) and identified high LVSD probability in 204 patients (14.1%), with AUC 0.96, sensitivity 95.6%, and negative predictive value 99.1% versus echocardiography. The study suggests AI-linked ECG screening could help connect heart disease risk to smarter care where echo access is limited.

Why It Matters To Your Practice

• LVSD often goes undetected in resource-limited settings because echocardiography is scarce, delayed, or unavailable.

• A 12-lead ECG paired with AI may offer a scalable front-end screening tool to identify which patients most need confirmatory imaging.

• The very high negative predictive value means the model may be especially useful for ruling out LVSD risk in broader screening workflows.

Clinical Implications

• Among 1,444 participants, 77.4% were high cardiovascular risk, yet AI-ECG performance remained strong across risk strata, with AUC 0.96 to 0.98.

• Specificity was 79.4% and positive predictive value was 43.2%, so positive screens would still require echocardiographic confirmation.

• This approach may help clinicians prioritize limited echo capacity for patients with prior cardiovascular disease or elevated Framingham risk.

Insights

• Participants were enrolled from June to December 2024 and received baseline assessment plus 12-lead ECG; a subset underwent echocardiography within 7 days as the reference standard.

• The validated convolutional neural network algorithm, AiTiALVSD, was used to estimate LVSD risk.

• The main value proposition is scalability: standard ECG infrastructure is far more available than imaging in many low-resource settings.

The Bottom Line

• For clinicians practicing where echo access is constrained, AI-ECG appears promising as a high-sensitivity screening tool for LVSD.

• Its strongest practical use may be safely excluding LVSD risk and triaging who should move on to echocardiography, rather than replacing imaging.