Atrial wave detection algorithm for discovery of some rhythm abnormalities

Atrial flutter (AFL) and atrial fibrillation (AF) generate patterns in the electrocardiogram, which have some similarity to a normal P-wave. Therefore their detection is an important step of AFL/AF recognition as well as P-wave detection is for AV block identification. Many approaches to P-wave identification are based on correlation with a manually selected template. The developed method and algorithm use a universal synthesized template and apply a modified convolution for P-wave detection. In order to emphasize the true coincidences, the sample differences between the template and P-wave candidates are introduced in a denominator. Each convolution is calculated twice with a shifted template, which locates once its first term closely to the corresponding signal sample and secondly does the same with its middle term. The convolution sum is compared with adaptive threshold Tr to mark P-wave occurrences. To accelerate the computation process, the two convolutions are accomplished only if a preliminary convolution over a reduced number of samples gives an outcome higher than the decreased threshold Tr/2. Further, a rule for AF recognition is applied. The results show that missed or wrong P-wave detections are observed in a limited number of RR intervals. They do not hamper the discovery of AFL/AF arrhythmias. The algorithm can be successfully used for such cases except for prediction of paroxysmal AF. The mean delay in AF onset recognition over a 10 h recording taken from MIT-BIH-Afdb is 2.7 s. This value for AF offset is 5.5 s. As a QRS detector is a part of the P-wave detection algorithm, both of them may be applied also for AV block identification.