Dynamic image for AD MRI classification

This is the Pytorch implementation of the ECCV 2020 worshop paper "Dynamic Image for 3D MRI Image Alzheimer's Disease Classification". paper

Authors:Xin Xing, Gongbo Liang, Hunter Blanton, M. Usman Rafique, Chris Wang, Ai-Ling Lin, and Nathan Jacobs

The dynamic image python script is refered by this.

Offical Dynamic image link is here.

Corresponding author: Xin Xing ([email protected])

Prerequisites

• Python3
• Pytorch
• NiBabel

MRI image samples:

Workflow:

Data:

Please save the MRI ".npy" data into CN and AD folders, respectively. You can use the "ADNI2_MRI_AD_niiData.ipynb" to convert the ".nii" file to ".npy" file.

Implementation:

The first operation is train-test files spliting. We are working on the 5-fold cross validation. So we use the "train_test_files_split.ipynb" to randomly split the data into 5 folders (Remember to change your data path). As following:

Then we can conduct the "Dynamic_image_Vgg11.ipynb" (Remember to change LABEL_PATH = '/data/scratch/xxing/adni_dl/Preprocessed/ADNI2_MRI' to your data path).

Results: