Brain Age Estimation using Brain MRI and 3D Convolutional Neural Network

Human Brain Age has become a popular aging biomarker and is used to detect the differences among healthy subjects. It is also used as a health biomarker between the group of normal subjects and the group of patients. Machine Learning (ML) prediction models and especially Deep Learning (DL) systems are rapidly grown up in the field of Brain Age Estimation (BAE) to present a disease detection system. In this paper, a DL method based on 3D-CNN is designed to get an accurate result of BAE. The training dataset is selected from the IXI (Information eXtraction from Images) MRI data repository. In addition, it is aimed to decrease the computations required by the deep model on the 3D MRI images. It is generally done by removing the unnecessary parts of brain 3D images. First, the deep 3D-CNN model is trained by healthy MRI data of IXI dataset which are normalized by SPM. Next, some experiments are done due to decrease the computations while saving the total performance. The best-achieved Mean Absolute Error (MAE) is 5.813 years.