A Real-Time Ventricular Arrhythmia Detection System Based on the KL Transform and Sequential-Hypothesis Testing

摘要

The single lead arrhythmia detection system described in this paper uses the Karhunen-Loeve (KL) transform for clustering of QRS complexes of like morphology. A heuristic rule set is employed to label the beats as Normal or Ventricular. In parallel with the beat detection and labeling algorithm is a life-threatening arrhythmia (LTA) detection algorithm which relies on sequential hypothesis testing to detect V-Fib or V-Tach. Average performance for both algorithms is above 90% in testing with both the MIT and AHA databases.