RhythMedix RhythmStar - Arrhythmia Detection Performance

Revison 0.8.6 Document No. 01424
Page 26
14. Arrhythmia Detection Performance

The RhythmStar incorporates a real-time embedded arrhythmia detection algorithm.
The processing steps include signal bandpass and morphological filtering, estimation
of the motion artifact, analysis of the slope, width and amplitude of the signal,
decision making logic to determine location and length of the QRS complexes, atrial
activity analysis and template matching for AF detection.
Rhythmedix conducted algorithm performance testing according to the ANSI/AAMI
EC57 standard. The MIT-BIH, AHA, and NST databases were used to verify
performance of the algorithm. The algorithm performance testing summary is
provided in the tables below:
QRS Detection Performance
AF Detection Performance
Database QRS Sensitivity QRS Positive Predictivity
MIT-BIH 99.77% 99.81%
AHA 99.72% 99.82%
NST 96.59% 79.86%
Database AF Sensitivity AF Specificity
MIT-BIH AF 93.43% 96.88%