Artificial intelligence for the improvement of electrocardiogram analysis

This thesis explores the application of machine learning (ML) and deep learning (DL) models to improve diagnostic processes in inherited heart diseases (IHD) and general ECG analysis.
The first part focuses on the use of ML and DL in diagnosing patients with IHD. It reviews current literature on artificial intelligence for IHD management and evaluates the performance of several ML and DL models in detecting genetic mutations from ECG data. These models are tested against expert cardiologists, with external validations conducted on independent cohorts. Additionally, advanced techniques such as transfer learning and explainable AI are applied to enhance model accuracy and interpretability. The thesis also explores the use of DL models for identifying genetic mutations linked to malignant ventricular arrhythmias, even when no visible ECG features are present.
The second part shifts the focus to broader applications of ML in ECG analysis beyond IHD. It investigates the potential of ML models to predict mortality in COVID-19 patients based on ECG data from multiple hospitals. Furthermore, the thesis describes the development of ethnicity-specific normal ECG limits using statistical modeling techniques. This includes the use of unsupervised ML to reveal ethnic differences in ECG patterns.
Through these approaches, the thesis demonstrates how advanced ML and DL techniques can enhance both specific disease diagnosis and broader ECG-based analysis in various medical contexts.