Abstract
BACKGROUND: Subclinical left ventricular (LV) impairment-characterized by reduced global longitudinal strain (GLS) despite normal left ventricular ejection fraction (LVEF)-is frequently encountered in hypertensive patients. While speckle-tracking echocardiography is the standard method for detecting early myocardial dysfunction, it is not universally available. Artificial intelligence-enhanced electrocardiography (AI-ECG) has emerged as a promising tool capable of uncovering subtle electrical patterns linked to early myocardial impairment. This study investigates the diagnostic capability of AI-ECG for detecting GLS-defined subclinical LV dysfunction. METHODS: In this retrospective analysis, 348 hypertensive adults who underwent both ECG and echocardiography within the same clinical visit (2022-2024) were evaluated. Subclinical LV dysfunction was defined as LVEF ≥50% and GLS > -18%.A convolutional neural network-based AI algorithm generated an AI-ECG probability score (range 0-1) representing the likelihood of LV dysfunction. Statistical analyses included correlation testing, regression modeling, and ROC curve evaluation. RESULTS: Subclinical LV dysfunction was identified in 134 participants (38.5%). The AI-ECG probability score differed markedly between the abnormal GLS group and the normal GLS group (0.61 ± 0.20 vs. 0.29 ± 0.18; p < 0.001). GLS values demonstrated a strong negative association with AI-ECG scores (r = -0.63). ROC analysis showed robust diagnostic ability with an AUC of 0.86 (95% CI: 0.82-0.89). In multivariable logistic regression adjusting for LV mass index, E/e', age, and hypertension duration, the AI-ECG probability score remained independently associated with subclinical LV dysfunction (adjusted OR 1.12 per 0.1 increase, 95% CI 1.07-1.18; p < 0.001). CONCLUSION: AI-ECG accurately detects GLS-defined subclinical LV dysfunction in hypertensive adults and may serve as an accessible tool for early risk stratification in routine clinical settings.