We developed an automatic method for predicting the spontaneous termination of atrial fibrillation (AF) using surface ECG analysis. The method was based on the hypothesis that AF segments with higher fibrillatory frequencies would be more likely to persist whereas those with slower frequencies would be more likely to terminate.
We separated each minute long segment into six contiguous 10 second sub-segments and performed median QRS-T subtraction on the sub-segments to cancel ventricular and correlated activity. We performed peak frequency analysis on the remainder signal and used the learning set to determine a peak frequency threshold to separate immediately terminating (T) from non-terminating (N) segments. A 5.9 Hz threshold correctly separated 19/20 T vs. N segments from 20 different patients. For separating non-immediately terminating (S) and T, the frequency analysis was preceded by QRS-T morphology matching to identify pairs of S and T segments from each individual patient. In this situation we chose the higher frequency segment from each pair as S and the lower frequency as T segments. This algorithm correctly separated 18/20 segments from the learning set in 10 patients.
Using this technique, we correctly identified 25/30 segments from Test Set A and 16/20 segments from Test Set B.
Supported by the National Institute of Biomedical Imaging and Bioengineering (NIBIB) under NIH grant number R01EB030362.