Quantitative prediction of right ventricular and size and function from the electrocardiogram

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Assessing right ventricular ejection fraction (RVEF) and end-diastolic volume (RVEDV) is challenging with traditional methods.
• Deep-learning enabled 12-lead electrocardiogram (DL-ECG) analysis for RV function and size remains unexplored.

Approach:

• A DL-ECG model was trained to predict RV dilation and dysfunction using cardiac MRI data from the UK Biobank.
• The model was fine-tuned and prospectively validated in a multi-center health system.
• Performance was evaluated using AUROC for categorical predictions and MAE for continuous RVEDV/RVEF values.

Key Points:

• The DL-ECG model achieved high AUROC values for predicting RV dysfunction (up to 0.86) and dilation (up to 0.92) across cohorts.
• Mean absolute errors for RVEDV and RVEF were 17.6 ml/m² and 7.8%, respectively.
• Model performance remained consistent across various subgroups, including those with and without left ventricular dysfunction.

Conclusions:

• DL-ECG analysis demonstrates accuracy in identifying significant RV dysfunction and dilation.
• Predicted RVEF independently correlates with patient outcomes, suggesting clinical utility.