MRI_Project

Instruction

Envirnment

python 3.6
Tensorflow
numpy, matplotlib, opencv
pandas, glob, csv, shutil, 

Preprocessing

1. Download all data (GM/WM folder) & PPMI.csv file
2. change GM/WM folder path in Split_PD_HC.py 
3. (Option) If you'd like to run the 2d CNN Model, 
            change the slice number (last number)of the [ex_command] (m stands for the middle slice)
            in Split_PD_HC.py (more detailed of using med2image can be found in https://github.com/FNNDSC/med2image )
4. Run ## $python3 Split_PD_HC.py ## to seprate HC and PD for GM and WM
5. Run ## $python3 Train_test,py  ## to split data into train and test data. 
        (Change the desire path inside Train_test file)

CNN model

1. 2D CNN  (2dCNN folder)
   To train the 2D model, change the folder path in 2d_cnn_bn.py
   and run ## $python3 2d_cnn_bn.py ## 
   all accuracy and loss results are saved in results.csv file 
   plot_result.py plot the graph of accuracy and loss results 

2. 3D CNN (3d_cnn folder)
   To train the 3D model, change the "train_path" and "test_path" in input_3Dimage_new.py file.
   and run ## $python3 learning_network_train.py ## 
   all accuracy and loss results are saved in Results.csv file.