A Framework for A Secure Brain Image Classification Using Deep Learning and Residue Number System

Abstract

Increasing availability of medical images generated via different imaging techniques necessitates the need for their remote analysis and diagnosis. This development has made privacy and security of patients’ medical records to be extremely important. In this paper, we present a brain image classification framework using deep learning model and concept of residue number system (RNS). Special moduli set of RNS will be used to conceal 8-bit binary value of each pixel present in the training and testing image dataset before the usage of a convolutional neural network (CNN) to classify the encrypted images. As part of the classification procedure, image segmentation and data augmentation procedure will be performed in order to identify the region of interest (ROI) and to avoid overfitting respectively. Specifically, this research will attempt to explore the potencies of CNN to classify cases of dyslexia from control subjects using MRI-generated image dataset. This kind of research becomes expedient due to the educational and medical importance of dyslexia learning disability.