Predicting Human Brain Age from MRI Data Using Deep CNNs Enhanced by MRMR

Abstract

Predicting the age of the brain using MRI is an advanced medical method that is used to diagnose brain  diseases and disorders such as Alzheimer's, multiple sclerosis and other neurological diseases. Using this method, brain MRI images are analyzed using advanced algorithms and neural networks to obtain different brain characteristics such as brain volume and cortical thickness. Then, by comparing these features with the MRI imaging data of other patients, the age of the brain is estimated. In this work, used Convolutional Neural Network (CNN) and MRMR feature selection algorithm. In this method, brain MRI images are processed by a convolutional network to extract age-related features, the MRMR algorithm selects the most relevant features, and the brain age is predicted using regression layers. The main contribution of this research is in adding a feature selection layer based on the MRMR feature ranking algorithm among the layers of a deep convolutional network, which has led to the improvement of the performance of the proposed convolutional network. Based on the obtained simulation results, the prediction accuracy of the proposed method for predicting the brain age of people is 90.3%, which has improved compared to the compared works.vements.