Medical Image Retrieval System using 2D-DWT and Convolutional Neural Network

Objectives: To propose an algorithm for Medical image retrieval better. Methods/Statistical analysis: The proposed image retrieval system incorporates with methods like discrete wavelet transform, information gain and the convolutional neural network for classification. Findings: We propose an algorithm of image retrieval that implements the feature extraction with 2D-Discrete Wavelet Transform (DWT), information gain for feature reduction and convolutional neural network for classification. The feature extraction involves the extraction of features from medical image using DWT. The Feature reduction was carried out by information gain. The image is classified by using classifiers such as convolutional neural network, traditional SVM classifier and KNN classifier. Application/Improvements: The classifier result provides the similar images in the database. The classification accuracy of convolutional neural network is 91.25% outperforms the traditional SVM classifier and KNN classifier. The image retrieved on high accuracy classifier provides exact match of query medical image for the diagnosis and treatment schedule.