Automatic detection of artifacts in heart period data.

A major consideration before analyzing heart period data or RR intervals is the removal of artifact. Automatic artifact rejection becomes critical when analyzing huge data sets where manual inspection is impossible. This study developed an efficient, automatic algorithm which uses both a +/-20% criteria and the median of 25 surrounding RRs to set the boundary for a normal RR interval. The newly developed algorithm is compared with 5 previously developed artifact detection algorithms. Thirty minutes of simultaneous electrocardiogram and RR interval data from 10 infants of the Collaborative Home Infant Monitoring Evaluation (CHIME) study group are used to test the performance. Both real artifacts in the infant RR interval and fake artifacts are used to assess the performance of the algorithms. Sensitivity, specificity, and accuracy are used to quantify the performance. The newly developed algorithm performs better than other algorithms for detecting real artifacts in infant RR interval data, providing the best balance between sensitivity and specificity (73.46% and 99.17%, respectively), greatly improving specificity with only a slight loss in sensitivity. It works extremely well for detecting missing and extra beats (100% and 97.44% sensitivity, respectively) even on very noisy data. In terms of detecting fake missing beats and extra beats, high sensitivity is achieved in most situations, and there is not much difference in specificity among the algorithms.