Fuzzy classification of intra-cardiac arrhythmias

The successful discrimination between intracardiac arrhythmias using fuzzy classifiers is presented, supporting the potential for such a system for use in implantable defibrillators. A nonlinear predictor using an Adaptive Neuro-Fuzzy Inference System (ANFIS) is used to classify the arrhythmias and therefore distinguish if defibrillation is required or not. A training structure comprising desired input-output data pairs of target electrocardiogram (ECG) waveforms, (ie the intra-cardiac arrhythmia), is based on a hybrid learning procedure. A fuzzy inference system (FIS) is generated based on the training data set and generates the desired membership functions and rules for the system. The ANFIS constructs a `fuzzy classifier' for each arrhythmia which is then used in a run-time simulation to produce a prediction error between the input data and the predicted data. The system resulted in correct arrhythmia detection and classification based on the lowest prediction error.