| Summary: | An approach based on dictionary-based Gaussian decomposition of electrocardiogram (ECG) traces is presented and characterized, and its performance potential is demonstrated using traces from the MIT-BIH Arrythmia Database. A Gaussian model is employed to describe ECG morphology. Parameters are estimated using a dictionary-based approach, that is purposely designed to obtain accurate representations with limited complexity and ensure comparability among different traces and subjects. The standardized Gaussian dictionary allows compact representations, enhances comparability and provides the support for machine learning-based diagnostics of ECG traces. Data-oriented large-scale medical analyses of ECG data are made possible, allowing the investigation of elusive cardiac phenomena and personalized diagnostics.
|
|---|