Detection of Congestive Heart Failure using R-R Interval via Probabilistic Symbolic Pattern Recognition

Abstract

We present a novel probabilistic symbolic pattern recognition (PSPR) approach to detect congestive heart failure (CHF) in subjects from their cardiac inter-beat (R-R) intervals. PSPR symbolically discretizes each R-R interval time series using eight-symbol alphabets and then probabilistically observe pattern transition behavior in the discretized series. Classification of RR interval series into normal and CHF via PSPR resulted in accuracy=84%, specificity=90%, and sensitivity=73% for test data in 5-fold cross-validation. The robust PSPR technique can be used for screening, improving medical diagnosis, and early warning alert systems.