Anatomical basis of human sex differences in ECG identified by automated torso-cardiac three-dimensional reconstruction

Background and Aims: The electrocardiogram (ECG) is routinely used for diagnosis and risk stratification following myocardial infarction (MI), though its interpretation is confounded by anatomical variability and sex differences. Women have a higher incidence of missed MI diagnosis and poorer outcomes following infarction. Sex differences in ECG biomarkers and torso-ventricular anatomy have not been well characterised, largely due to the absence of high-throughput torso reconstruction methods. Methods: This work presents quantification of sex differences in ECG versus anatomical biomarkers in healthy and post-MI subjects, enabled by a novel, end-to-end automated pipeline for torso-ventricular anatomical reconstruction from clinically standard cardiac magnetic resonance imaging. Personalised 3D torso-ventricular reconstructions were generated for 425 post-MI subjects and 1051 healthy controls from the UK Biobank. Regression models were created relating the extracted torso-ventricular and ECG parameters. Results: Half the sex difference in QRS durations is explained by smaller ventricles in women both in healthy ($3.4 \pm 1.3$ms of $6.0 \pm 1.5$ms) and post-MI ($4.5 \pm 1.4$ms of $8.3 \pm 2.5$ms) subjects. Lower baseline STj amplitude in women is also associated with smaller ventricles, and more superior and posterior cardiac position. Post-MI T wave amplitude and R axis deviations are more strongly associated with a more posterior and horizontal cardiac position in women rather than electrophysiology as in men. Conclusion: A novel computational pipeline enables the three-dimensional reconstruction of 1476 torso-cardiac geometries of healthy and post-myocardial infarction subjects, quantification of sex and BMI-related differences and association with ECG biomarkers. Any ECG-based tool should be reviewed considering anatomical sex differences to avoid sex-biased outcomes.