| Abstract [eng] |
Atrial fibrillation (AF) is the most prevalent clinically relevant arrhythmia worldwide. It was first diagnosed more than 100 years ago, however, scientists have not since agreed upon the mechanisms of the processes in the atria during fibrillation, and various environmental and lifestyle factors influencing the course of the disease in individual patients. Long-term ambulatory monitoring of AF may provide better understanding of the arrhythmia. Frequency of fibrillatory f waves, observed in electrocardiogram (ECG) during AF, is considered as the main non-invasive characteristic describing electrical processes in the atria. Unfortunately, poor signal quality due to motion artifacts and electrode movement is commonly encountered in ambulatory ECG recordings making f wave analysis particularly challenging. Therefore, this work investigates the feasibility of f wave frequency tracking in ambulatory ECG recordings. Two separate databases of simulated and long-term clinical ECG signals during AF are employed in the study. The simulated database is used for comparison of four algorithms for f wave frequency tracking, namely adaptive bandpass filter, adaptive bandstop filter, spectral profile method and Welch's method. The dependencies of their absolute frequency estimation error on the length of the analysis window and signal-to-noise ratio are investigated. After selecting adaptive bandpass filter for ambulatory f wave analysis, f wave frequency is estimated in long-term ECG signals recorded for 38 patients clinically diagnosed with permanent AF. Signal quality index-based database quality assessment shows that 40% of the signals are unsuitable for f wave frequency estimation. The results show that f wave signal quality assessment plays a significant role in f wave frequency estimation and, therefore, should be considered in ambulatory signal analysis. Application of signal quality index results in the exclusion of the f wave frequency outliers originating from poor quality signal segments, and consequently allows evaluation of the true f wave frequency variation. In this study, f wave frequency change induced by circadian rhythm and physical activity was ambulatorily observed in 75% and 57% of the patients, respectively. |
