MNN: Multiclass Neural Network Classifier for Cardiac Disease Prediction Models

An improved medical decision support system for classification of Cardiac Disease using Multiclass Neural Network Classifier (MNN) is developed for Cardiac disease prediction presented in this paper. The ECG signal datasets are preprocessed and filtered by Butterworth noise filtering for removing the measurement noise and extracted peaks was obtained for extraction using Peak difference method. The detected feature peaks are used to perform classification and evaluation using Multiclass Neural network training. The generated training patterns from the preprocessed signal are used as input and from the data set the pre-computed pattern set is retrieved. With the extracted peak feature, compute each ECG wave similarity measure with the available pattern set database. The Multiclass pattern set which has more similarity in each pattern is identified and based on identified values of each signal values the components of the ECG signal is separated and classified from the input pattern and constructed to form a wave form. The pattern is generated from different dimension and generated pattern will be used to compute the similarity measure. The classification of Cardiac Disease datasets was done using MNN classifier and the accuracy of classification was found to be 93.15%. The developed system is expected to provide good support for the medical practitioners for decision making for enhanced health care.