Short supraventricular tachycardias (S-SVTs) have been associated with a higher risk of developing atrial fibrillation (AF). Hence, identification of participants with such arrhythmias may increase the yield of AF screening over time. However, the lower signal quality of ECGs recorded using handheld screening devices and the abundance of aberrant ectopic beats and various arrhythmias in the elderly, who are the target population in AF screening, challenge the detection of S-SVT.
In the present work, a new method for detection of S-SVT is presented, which in addition to rate and duration criteria, is based on the assumption that only subtle differences in beat morphology may occur during S-SVTs, as all beats are assumed to have the same origin. Therefore, any episode with a sequence of similar beat morphology is considered as an S-SVT candidate while any episode with different morphology due to signal disturbances or aberrant ectopic beats is excluded. For this purpose, a support vector machine (SVM) was trained and validated, using a simulated ECG database, to classify an episode as either consisting of beats of similar or non-similar morphologies. Episodes identified as S-SVT candidates are subject to two further rhythm criteria in order to confirm the presence of an S-SVT.
The performance of the S-SVT detector is evaluated using a subset of the StrokeStop I database (305 S-SVT out of 8258), resulting in a sensitivity, specificity, and positive predictive value of 88.8%, 92.0%, and 29.9%, respectively.
In conclusion, the results suggest that the proposed S-SVT detector has the potential to reduce the number of false S-SVT detection.