Automated modeling and processing of long-term electrocardiogram signals

Physiological signals consist in the recording of the electrical activity generated by the human body, for instance in the muscles or in the brain. The analysis of these signals yields a great potential for such various tasks as brain-computer interfaces and the monitoring of body functions, including the diagnosis of disease conditions. In particular, the electrocardiogram (ECG) is a physiological signal representing the electrical activity produced by the heart. In the large majority of situations involving the recording of an ECG signal, such as pharmaceutical phase-one studies, a long-term monitoring is required not to miss any transient pattern. This thesis focuses on the design and the assessment of machine learning algorithms to automatically process such ECG recordings. Four objectives are investigated in this context. The first objective concerns the automatic segmentation of the ECG characteristic waves using sparse conditional random fields and the wavelet transform. The second objective concerns the analysis of fluctuations in autonomic activity by heart rate variability metrics. A critical review of heart variability metrics is provided and the possible use of heart rhythm variations as a marker of epileptic seizures is investigated. The third objective concerns the supervised classification of heart beats, which involves the labeling of each beat in a recorded ECG signal as either normal or pathological. For this purpose, a weighted variant of the support vector machine and of the conditional random fields classifiers is proposed. The fourth objective concerns the filtering of ECG artifacts in invasive vagus nerve recordings using filtering and semi-blind source separation techniques such as periodic component analysis. Experiments on real ECG recordings are conducted throughout the thesis to validate the performances of the proposed algorithms.