Brain Age Prediction with 3D ResNet34 Model in Healthy Control, Mild Cognitive Impairment, and Alzheimer's Disease

Brain age prediction is closely related to brain health, which is an important index to estimate the state of brain health. In this paper, we use the 3D ResNet34 model to predict the brain age to provide evidence to prove that the brain age of AD patients is higher than that of normal people, hoping to be an important factor in judging early AD. We first collect the data from two open datasets: CAM-CAN and ADNI. Then, we reduce data with Freesurfer software. Finally, we use the processed data to train the3D ResNet34 model and then used the model for prediction. Since a person's age changes continuously, Mean Average Error (MAE), Root Mean Square Error (RMSE), Mean Square Error (MSE), and other common performance evaluation indexes are used to evaluate the related performances of related methods. The data used in this paper is structural MRI, which comes from CAM-CAN (download permission has been applied) and ADNI open data set. The image quality is 3T. The input data of the model is structure Magnetic Resonance Imaging (MRI) corrected by Freesurfer software, and the image size is 256 ⁎ 256 ⁎ 256 ⁎ 1, is length, width, depth, and the number of color channels respectively. The results show that the brain age of AD patients is generally older than that of normal people, and indicate that brain age can use in judging early AD.