A Convolution Neural Network with Optimal Filter Set for Medical Image Processing

The present paper proposes the extraction of Optimal Number of Filter Set and thereby Optimal Filter Set in Convolution Neural Network during its application in Medical Image Processing. This work involves medical applications namely Classifying ECG on the basis of types of Arrhythmias. We have used Convolution Neural Network, where we have optimized the Filter Number and Filter Set for learning ECG Data Set. We have extracted the Optimal Filter Set for the above medical data set, during the training of the Convolution Neural Network. The present paper thus focuses into the optimal set of filters in a Convolution Neural Network model which is bound to exhibit the maximum overall accuracy of classification. The work achieves an accuracy of classification of different types of Arrhythmias of  97.56% with 59 optimal number of filters. Over and above, as and when the optimized learned Optimal Filter Set is applied during recognition or detection of Arrhythmia the time taken for recognition is also thereby minimized.