Diabetes-Related Alterations in Vectorcardiographic Repolarization Geometry: Insights from Physiological and Machine-Learning Analysis.

Background: Diabetes is associated with alterations in cardiac electrophysiology, yet the extent to which metabolic dysfunction influences spatial ventricular repolarization geometry remains unclear. Vectorcardiographic (VCG) analysis provides a spatial representation of cardiac electrical activity and may reveal subtle electrophysiological changes associated with metabolic disease. Methods: This study analyzed VCG features in 200 adults to characterize the relationship between diabetes and depolarization-repolarization dynamics. Spatial parameters including QRS-T angle, T-wave and QRS magnitudes, loop morphology indices, and interval-based features were extracted from reconstructed orthogonal leads derived from standard electrocardiograms. Associations between clinical variables and VCG parameters were evaluated using Pearson correlation analysis and principal component analysis. A Random Forest classifier was used to assess whether VCG features capture metabolic influences, with model performance evaluated on a holdout dataset. Permutation testing assessed statistical significance, and SHAP (Shapley Additive Explanations) analysis was used to quantify feature contributions to model predictions. Results: The Random Forest classifier demonstrated moderate discrimination between participants with and without diabetes (ROC AUC = 0.878; PR AUC = 0.631) with acceptable calibration (Brier score = 0.127), and statistical significance confirmed by permutation testing (p = 0.010). SHAP analysis identified spatial QRS-T angle as the most influential predictor, followed by T-wave magnitude, systolic blood pressure, T-loop area, and body mass index. Repolarization-related VCG features consistently contributed the greatest influence on model predictions. Conclusion: Diabetes is associated with measurable alterations in ventricular repolarization geometry, with spatial QRS-T angle and related VCG parameters emerging as key electrophysiological signatures. These findings suggest that metabolic dysregulation contributes to early electrical remodeling and highlight vectorcardiographic phenotyping as a potential noninvasive approach for investigating cardiometabolic influences on myocardial electrical function.