K-Nearest Neighbor classification using RR intervals detected atrial fibrillation with an average accuracy of 91.75% and a highest accuracy of 95.45% (sensitivity 91.67%, specificity 100%).
Does K-Nearest Neighbor classification using RR intervals accurately detect atrial fibrillation from ECG signals?
A K-Nearest Neighbor machine learning approach using RR intervals can accurately detect atrial fibrillation from ECG signals.
Atrial Fibrillation (AF) categorized as one kind of arrhythmia that mostly found on a daily basis. It is indicated by irregular heart beat in the heart's electrical system from the atrium into the ventricle. A person who has never had a history of heart disease even gets possibility suffering from AF. Risks caused by AF, namely the possibility of stroke, heart failure, and death. For someone who already has symptoms of AF should immediately examine one of them by using an electrocardiogram (EKG). Due to the presence of early detection can reduce the number of percentage of AF population, and the prognosis of AF disease is also preferable. There are three stages in this research; they are pre-processing as a process of uniforming data dimension, feature extraction, and K-NN classification. Feature extraction applied by comparing the RR interval of AF's signal and the normal one. The best performance result of AF detection based on the accuracy of the overall scheme is k = 1 with an average accuracy at 91.75% and the highest accuracy, sensitifity, and specificity level at 95.45%, 91.67%, and 100% with proportion data at 60:40 percent.
Resiandi et al. (Tue,) conducted a other in Atrial Fibrillation. K-Nearest Neighbor classification using RR interval vs. Normal ECG signals was evaluated on AF detection accuracy. K-Nearest Neighbor classification using RR intervals detected atrial fibrillation with an average accuracy of 91.75% and a highest accuracy of 95.45% (sensitivity 91.67%, specificity 100%).