If you use this work for your research, please cite the following work:

M. Hon, N.M. Khan. Towards Alzheimer's Disease Classification through Transfer Learning. IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2017.

Thank you!

The final results of this experiment are in the folders:

1. InceptionV4_FineTunnign_ALL - finetuning with InceptionV4
2. VGG16_Experiment - finetuning with VGG16 (weights)
3. VGG16_Experiment_Models - finetuning with VGG16 (models)
4. VGG16_None_Experiments - no finetuning, training VGG16 from scratch
5. MRI_32_Images - all images and folders of datasets

This folder (in data.zip) contains a structure of "train" and "validation". Each with "alzheimers" and "nonalzheimers" MRI images. There are a total of 210 for each of training and a total of 90 for each of validation.

These images were obtained from http://www.oasis-brains.org/ .

These images were converted GIF to JPG.

Additionally, there is a script called "ChangeNamesOfFiles.bat". This file was used to change the original names to YAL###.jpg and NAL###.jpg . YAL stands for "Yes Alzheimers" and NAL stands for "No Alzheimers".

The following is the step by step instructions to get Keras working.

The operating system is Ubuntu.

The steps are:

1. python -V (makesure it is 2.7)
2. sudo apt install python-pip
3. pip install numpy
4. pip install pillow
5. pip install scipy
6. pip install keras
7. pip instal h5py
8. sudo apt-get install python-pip python-dev
9. pip install tensorflow

openCV --> sudo apt-get install python-opencv

qsub -l gpu module purge module load CUDA/8.0 module load PYTHON/2.7.13 pip install --user

chmod a+x qsub -q gpu submit.pbs