This is an outlier detection approach to OCT data classification using Gaussian Mixture Model.
To set it up to run for your data, follow the below steps:
- Get the code from the repo.
- Create folders, 'data' and 'duke_data' that contain the dataset volumes in *.mat file format. These folders in turn contain folders 'DME' and 'normal' under them, which contain the corresponding volumes.
- Add the folders 'data', 'duke_data' and 'utils' to Matlab path.
- You might have to compile the c files associated with NL Means to get the mex files for NL means filtering working for you.
- If you are running the project for the first time, set
do_preprocess
to 1, which runs NL Means on each frame in each volume and stores the result. To just use the saved result this can then be set to 0. - You are now all set to run the files for different cases:
- For SERI dataset + intensity features run: run.m
- For Duke dataset + intensity features run: run_duke.m
- For SERI dataset + LBP features run: run_lbp.m
- For Duke dataset + LBP features, run: run_lbp_duke.m