Background: Atrial fibrillation(AF) is a kind of persistent arrhythmia that can lead to serious complications. Therefore, accurate and quick detection of atrial fibrillation by surface electrocardiogram (ECG) has great importance on further treatment. The practical ECG signals contain various interferences in different frequencies, such as myoelectricity interference, power interference and so on. Detection speed and accuracy largely depend on the AF signal features extracted by algorithm. But some of the discovered AF features are not well distinguishable, resulting in poor classification effect.
Methods: This paper proposed a high distinguishable atrial fibrillation feature - the frequency corresponding to the maximum amplitude in the frequency spectrum (MAiFS). We used the R-R interval detection method optimized with mathematical morphology method and combined with the wavelet transform method for analysis. According to the two features - the MAiFS and R-R interval irregular, we can recognize AF in ECG signal by decision tree classification algorithm.
Results: The data used in the experiment comes from the MIT-BIH database  , which is publicly accessible via the web and with ethics approval and consent. The dataset contains 23 annotated ECG records, each of which is approximately 10 hours with a sampling rate of 250Hz and a 12-bit resolution with a range of 10mv. Based on the input of time-domain and frequency-domain features, a supervised classifier is constructed by using decision tree algorithm, and the data obtained from the above experiments are brought in to carry out a 5-fold cross validation test, the accuracy of classification reaches 98.9%.
Conclusions: The frequency corresponding to the maximum amplitude in frequency spectrum in the normal signal is concentrated and the fluctuation is weak. But the frequency corresponding to the maximum amplitude in frequency spectrum in the atrial fibrillation signal is divergent and irregular. The decision tree algorithm can detect the normal signal and AF signal with 98.9% accuracy.