Rhythm-based accuracy improvement of heart beat detection algorithms

Aims: Our aim is to improve the accuracy of existing heart beat detection algorithms in order to provide reliable heart beat locations in a multi-modal beat detection scheme. Methods: A rhythm-based algorithm is presented which on top of a base beat detection method processes the detected beats by rejecting annotations and filling in gaps while minimizing a deviation score. A novel beat detection method based on rational modelling of ECG signals is also presented as a base algorithm. Results: The rhythm-correction algorithm applied to Sachin Vernekar's phase 11 entry was submitted to the third phase of the PhysioNet/CinC Challenge 2014 contest. The algorithm has 99.98% gross and average sensitivity and 99.96% gross and average positive predictivity compared to 99.92% and 99.94%, respectively, of the base algorithm. Due to run-time performance problems, the rational algorithm was not able to qualifY in the contest. Conclusions: The rhythm-based method improves the results of the base algorithm on the training data set. The hidden records are not yet available at the time of writing of this paper; therefore we are not able to report the final performance of the algorithm. Run-time improvement of the rational algorithm remains future work.