An efficient image based feature extraction and feature selection model for medical data clustering using deep neural networks

Abstract

The multi-modal health information representing the learning material was examined and multiple learning models were suggested for disease risk assessments, with the aim of mining information from the medical data and developing intelligent applications issues. A medical textual learning model based on a convolution neural network is proposed for the aspect of medical textual functional education. In the framework for risk evaluation, the convolution neural network information retrieval methodology is applied. The deep learning approach is used for medical data representation. To achieve flexibility of the model, the learning and extraction of various disease qualities use the same process. A simple pre-processing of the experimental data samples, including their denigration of power frequency and regulating lead convolution, builds a convolution neural network for advancing and intelligent recognition of medical data. The impressive performance gain achieved by Deep Neural Networks (DNNs) for various tasks prompted us to use DNN for the task of image classification. For the extraction and classification of functions, we used a DNN version called Deep Convolution Neural Network (DCNN). For classification and feature extraction, neural networks can be used. Two related roles can be seen better in our work. DCNN is used for the extraction and classification of functions in the first task. The second task is to extract functions using DCNN, and then to identify extracted characteristics with the SVM classifier. Function extraction shows small features extracted, but image information is useful. One of the major problems for the Content based Image Retrieval (CBIR) is that useful information must be extracted from the raw data to display image contents. The removal task changes the rich content of the image into various functions. The architecture with three levels of concentration and pooling, followed by a complete connected output layer, is used for extraction of functionalities among various configurations that we have considered. DCNN extracted features are supplied in task 1 for classification to a 2 hidden layer neural network. The proposed model is compared with the traditional models and the results show that the performance of the proposed model is better in terms of accuracy levels.