A fine-tuned AI-ECG model detected left ventricular systolic dysfunction with an AUC of 0.89, comparable to NT-proBNP, and independently predicted long-term mortality and incident LVSD in patients with Chagas disease.
Cohort (n=1,304)
Blinded assessors
Yes
Does a fine-tuned AI-ECG model accurately detect left ventricular systolic dysfunction and predict mortality in patients with Chagas disease compared to NT-proBNP?
A fine-tuned AI-ECG model can accurately detect left ventricular systolic dysfunction and predict long-term mortality in Chagas disease, offering a viable, low-cost alternative to NT-proBNP in resource-limited settings.
Effect estimate: AUC 0.89 (95% CI 0.85-0.93)
p-value: p=0.52
BACKGROUND: Left ventricular systolic dysfunction (LVSD) is the main predictor of mortality in Chagas disease (ChD). Although LVSD can be treated with affordable medications, its diagnosis relies on cardiac imaging, which is often unavailable in resource-limited settings. OBJECTIVES: The objective of the study was to evaluate an artificial intelligence-enabled electrocardiogram (AI-ECG) for detecting LVSD and predicting mortality and incident LVSD in ChD. METHODS: A previously developed AI-ECG LVSD model was fine-tuned in an external ChD sample and applied to SaMi-Trop, a Brazilian prospective ChD cohort. Diagnostic performance for LVSD, confirmed by echocardiography, was compared with N-terminal pro-B-type natriuretic peptide (NT-proBNP). Prognostic performance for all-cause mortality at 2 and 9 years was assessed using Cox models, and incident LVSD over 7 years using log-binomial models. RESULTS: Among 1, 304 participants, AI-ECG showed high accuracy for LVSD detection (area under the receiver operating characteristic curve: 0. 89; 95% CI: 0. 85-0. 93), similar to NT-proBNP (area under the receiver operating characteristic curve: 0. 90; 95% CI: 0. 87-0. 93; P = 0. 52). Among 1, 547 patients with Chagas cardiomyopathy, AI-ECG predicted all-cause mortality at 2 and 9 years with performance comparable to NT-proBNP (9-year C-index: 0. 78; 95% CI: 0. 74-0. 82 vs 0. 77; 95% CI: 0. 74-0. 79). AI-ECG could replace NT-proBNP in an established ChD mortality risk score with minor loss of accuracy. Incident LVSD occurred in 8. 4% over 7. 3 years, and AI-ECG predicted incident LVSD with performance similar to NT-proBNP. CONCLUSIONS: A fine-tuned AI-ECG model showed high accuracy for LVSD detection and was independently associated with long-term mortality and incident LVSD. Despite slightly lower performance than NT-proBNP, it may serve as a substitute in settings where NT-proBNP is unavailable. (Longitudinal Study of Patients With Chronic Chagas Cardiomyopathy in Brazil SaMiTrop Project SaMi-Trop; NCT02646943).
Cardoso et al. (Wed,) conducted a cohort in Chagas disease (n=1,304). Artificial intelligence-enabled electrocardiogram (AI-ECG) vs. NT-proBNP was evaluated on Detection of left ventricular systolic dysfunction (LVEF ≤ 40%) (AUC 0.89, 95% CI 0.85-0.93, p=0.52). A fine-tuned AI-ECG model detected left ventricular systolic dysfunction with an AUC of 0.89, comparable to NT-proBNP, and independently predicted long-term mortality and incident LVSD in patients with Chagas disease.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: