| Abstract [eng] |
The specific situation of chronic kidney disease patients, especially those with end-stage kidney disease on hemodialysis, makes them a target population whose treatment could highly benefit from the long-term photoplethysmography (PPG)-based monitoring for the detection of the initial episodes of life-threatening arrhythmias and the assessment of the ectopic burden. However, in order to accelerate the adoption of PPG-based technologies in this setting, a thorough investigation of the PPG signal and methods to facilitate the development of algorithms for arrhythmia detection are needed. In this doctoral thesis, a clinically relevant scientific-technological problem of the obstacles hindering the development and application of PPG-based algorithms for the long-term arrhythmia monitoring is covered. This thesis provides insights into the quality of the wrist PPG encountered in free-living activities. For the first time, quantitative characteristics of different-type artifacts are obtained from ambulatory PPGs, and invoked to develop a realistic artifact model which is applied to test the robustness of an arrhythmia detector. In this work, the PPG model is adjusted to simulate episodes of life-threatening arrhythmias by accounting for the influence of those arrhythmias on hemodynamics. The adjusted PPG model can be used not only for testing, but also for training and validation of deep learning methods for arrhythmia detection. For the first time, a continuously acquired wrist PPG is applied to assess the daily ectopic burden under uncontrolled conditions in hemodialysis patients. Not only the potential of the technology is demonstrated, but also its weaknesses and possible aspects for the improvement are identified. |
