Abstract
Ambulatory electrocardiography (Holter ECG) is commonly used ; however, it is unreliable when used for automated diagnosis of arrhythmia, especially supraventricular arrhythmia. Automated detection of the P wave is required to improve diagnostic precision. Paroxysmal atrial fibrillation (PAf) is a common arrhythmia, and its clinical management is of great importance. Automated diagnosis of PAf with Holter ECG would be extremely useful in clinical practice. We investigated a PAf detection algorithm, compared the results with diagnoses made by cardiologists, and investigated the precision and other problems associated with PAf diagnosis. We examined 36, 754 sinus P waves obtained from 12 subjects. Sensitivity was 94.3% and the positive predictive value was 98.5%. Some P waves could not be detected owing to low amplitude, noise interference, or swamping by the preceding T wave. False detection of the P wave also occurred in cases with high noise and low amplitude. The subjects were 15 patients in whom lasting 30 seconds or longer had been detected with Holter ECG ; 101 PAfs diagnosed by cardiologists were investigated by means of automated diagnosis. Sixtythree percent (64 of 101) of PAfs showed consistent start and end points, 23.7 % (24 of 101) of PAf s showed discrepancies in end points within several heart beats, and correct diagnosis using both methods was 87.1% (88 of 101) . Cases of PAfs which were not detected showed marked arrhyhmia of R-R intervals due to bigeminy or trigeminy or short runs of premature atrial complexes, with no detection of the P wave. Eight of 101 PAfs could not be recognized as PAf because the f wave was mistaken for the P wave. Automated detection successfully determined PAf start and end points, and allowed clinical evaluation of several parameters, including total heart rates, duration, recovery time, maximal R-R interval, and mean, maximal, and minimal heart rates during PAf, for which information was previously insufficient. This system was found to be useful for PAf automated diagnosis.
