Multimodal machine learning integrating ECG signals and clinical letters improved peak oxygen consumption prediction accuracy to 70.8% compared to 61.4% with ECG alone in congenital heart disease.
Observational (n=436)
No
Does a machine learning model integrating ECGs and clinical text data improve the prediction of CPET outcomes in patients with congenital heart disease?
Integrating 12-lead ECG signals with NLP-extracted clinical text data using Riemannian geometry significantly improves the prediction of prognostic CPET variables in patients with congenital heart disease.
Absolute Event Rate: 70.8% vs 61.4%
Abstract Cardiopulmonary exercise testing (CPET) provides a comprehensive assessment of functional capacity by measuring key physiological variables including oxygen consumption (VO₂), carbon dioxide production (VCO₂), and pulmonary ventilation (VE) during exercise. Previous research has identified peak VO₂ and VE/VCO₂ ratio as robust predictors of mortality risk in chronic heart failure (CHF) patients as well as in congenital heart disease (CHD). This study utilises CPET variables as surrogate mortality endpoints for patients with CHD. To our knowledge, this represents the first successful implementation of an advanced machine learning approach that predicts CPET outcomes by integrating electrocardiograms (ECGs) with information derived from clinical letters. Our methodology began with extracting unstructured patient information from clinical letters using natural language processing techniques, organising this data into a structured database. We then digitised ECGs to obtain quantifiable waveforms and established comprehensive data linkages. The core innovation of our approach lies in exploiting the Riemannian geometric properties of covariance matrices derived from both 12-lead ECGs and clinical text data to develop robust regression and classification models. Through extensive ablation studies, we demonstrated that the integration of ECG signals with clinical documentation, enhanced by covariance augmentation techniques in Riemannian space, consistently produced superior predictive performance compared to conventional approaches.
Alkan et al. (2026) conducted an observational in Congenital heart disease (n=436). Multimodal machine learning integration of ECGs and clinical letters using Riemannian geometry vs. ECG data alone or conventional ECG features was evaluated on Prediction of peak oxygen consumption (VO2 peak) accuracy. Multimodal machine learning integrating ECG signals and clinical letters improved peak oxygen consumption prediction accuracy to 70.8% compared to 61.4% with ECG alone in congenital heart disease.