Development of a Novel Neural Network Model for Brain Image Classification

Abstract

The analysis of brain MRI images is highly beneficial for the medical practitioners. Since the manual study of these images are time consuming and tedious, the automated process using software based system have been developed. The machine learning techniques are applied in developing brain MR image classification process. The classification process consists of dataset preparation, feature extraction, feature reduction and the use of classifier. In this paper, 2D DWT is used for feature extraction and PCA is used for feature reduction. ELM model is used as a classifier. The input weights and biases in ELM are randomly assigned. So EHO algorithm, a newly developed bio inspired algorithm is used to optimally determine the input weights and biases of ELM model. The classification performance of the EHO-ELM model is compared with basic ELM model for three of the brain MR image datasets. From the simulation study, it is found that the proposed EHO-ELM model outperformed the basic ELM model.