Female anatomies disguise ECG abnormalities following myocardial infarction: an AI-enabled modelling and simulation study

The electrocardiogram (ECG) is modulated by torso-heart anatomy, and this challenges patients’ diagnosis and risk stratification. This study aims to quantify how torso-heart anatomical factors affect sex-differences in ECG biomarkers in acute and chronic myocardial infarction (MI). We exploit the perfect control of AI-augmented multiscale modelling and simulation, based on clinical magnetic resonance imaging (MRI) data and ECGs for model construction and validation. A cohort of 1720 torso-ventricular anatomies (50% female) was constructed from MRIs of healthy and post-MI participants in the UK Biobank study. 8600 ECG simulations were performed considering anatomical variability and 3 electrophysiological stages (healthy, acutely ischemic, and infarcted). The effect of cardiac size, position, and orientation on each ECG biomarker was quantified. Female anatomies had larger distances between the infarct and ECG electrodes (relative to cardiac size), and larger angles between the infarct normal and ECG lead axes, both primarily caused by their more superior cardiac position. This reduced ST-elevation and caused low-amplitude late depolarisation and repolarisation tails to be missed, shortening QRS duration (QRSd) and T-peak-to-end interval (TpTe). The position and orientation of the heart impacted TpTe more severely than QRSd. AI-enabled mechanistic modelling and simulation identify smaller ventricles, superior cardiac position, and different ventricular orientation as key anatomical contributors of shorter QRS and T waves, and lower ST-elevation, in female versus male anatomies. This provides a blueprint for quantifying the impact of anatomical sex differences on functional markers and enables future work in tailoring clinical guidelines considering unique patient anatomy to reduce biased outcomes.

💡 Research Summary

This paper investigates how anatomical differences between the sexes influence electrocardiographic (ECG) biomarkers used for diagnosing and risk‑stratifying myocardial infarction (MI). Using an AI‑augmented multiscale modelling pipeline, the authors constructed 1,720 torso‑ventricular anatomical models (50 % female) from magnetic resonance imaging (MRI) data of healthy and chronic post‑MI participants in the UK Biobank. Each model was subjected to 42 affine transformations (translations in three axes and rotations about three axes) to span the observed variability in heart size, position, and orientation. The resulting cohort captures realistic distributions of body‑mass index, cardiac volume, and heart‑torso geometry for both sexes.

Electrophysiological simulations were performed with the human ventricular ToR‑ORd cell model, incorporating fibre‑sheet‑normal anisotropy and calibrated conductivities to reproduce clinically observed conduction velocities and QRS durations. Three electrophysiological states were simulated: healthy, acute ischemia (hours after coronary occlusion), and chronic infarction (weeks after occlusion). Ischemic and infarcted regions were modelled as fully transmural ellipsoids placed in anteroseptal and inferolateral locations, with ionic remodeling based on experimental data (elevated extracellular K⁺, activation of I_KATP, inhibition of I_Na and I_CaL for acute ischemia; reduced I_Kr, I_to, I_K1 and further I_Na/I_CaL inhibition for chronic infarction). Conductivity in the infarct core and border zone was reduced to half of normal tissue to match measured conduction slowing.

A total of 8,600 ECG recordings (five scenarios per anatomy) were generated using the MonoAlg3D monodomain solver on a high‑performance GPU cluster (≈8 h per simulation, ≈1,600 h total). The authors extracted standard ECG biomarkers: ST‑segment elevation, QRS duration (QRSd), and T‑peak‑to‑end interval (TpTe).

Key findings:

1. Female anatomies have smaller ventricular volumes (≈10 % less) and a more superior‑posterior cardiac position relative to the torso. Consequently, the geometric distance from the infarct core to the recording electrodes is larger (≈15 % increase) and the angle between the infarct normal vector and the lead axes is greater.
2. These geometric factors attenuate the projected ST‑segment elevation, reducing its amplitude by an average of 0.07 mV compared with male counterparts, potentially pushing many female cases below the sex‑specific clinical thresholds (0.15 mV for women vs 0.20 mV for men).
3. Smaller ventricular size shortens the electrical propagation path, leading to a reduction in QRS duration (≈8 ms) and a more pronounced shortening of TpTe (≈12 ms). TpTe is especially sensitive to heart orientation (tilt and spin angles), whereas QRSd is primarily driven by ventricular volume.
4. Statistical analysis identified heart‑to‑torso vertical position and rotation angles as the dominant predictors of the observed sex differences in ST‑elevation, QRSd, and TpTe.

The study validates the computational framework through verification against benchmark solvers, convergence testing, and comparison with clinical ECG measurements, establishing credibility for the mechanistic insights.

In conclusion, the work demonstrates that anatomical sex differences—smaller ventricles, higher cardiac placement, and distinct orientation—systematically mask ECG signs of MI in women. This mechanistic explanation clarifies why women are more frequently under‑diagnosed or mis‑risk‑stratified in acute coronary syndromes. The authors propose that future clinical guidelines incorporate patient‑specific anatomical parameters (e.g., heart‑torso distance, orientation) to personalize ECG interpretation, thereby reducing gender bias and improving outcomes for female MI patients.