Optimized Local Weber and Gradient Pattern-based medical image retrieval and optimized Convolutional Neural Network-based classification

Abstract

This paper presents a novel image retrieval and classification model for medical images so-called as Content-Based Medical Image Retrieval. For the image retrieval, the main contribution of this paper is to develop a pattern descriptor termed as Optimized Local Weber and Gradient Pattern. A hybrid algorithm with the integration of two well-known meta-heuristic algorithms like Barnacles Mating Optimization and Jaya Algorithm, namely Jaya-based Barnacle Mating Optimization, is used for improving the Optimized Local Weber and Gradient Pattern-based image retrieval. Once the Optimized Local Weber and Gradient Pattern features are determined for training and testing images, the retrieval of images is performed by the logarithmic similarity computation. Moreover, an improved deep learning model called Optimized Convolutional Neural Network is employed for performing the image classification. As an improvement, the activation function of Convolutional Neural Network with sigmoid, tanh, Relu, and RRelu and the maximum epoch are optimized by the same Jaya-based Barnacle Mating Optimization. Finally, extensive experiments are carried out over publicly available datasets, demonstrating that the proposed retrieval and classification models are better than the existing models.