This project is an exploration into the feasibility and effectiveness of using Support Vector Machine (SVM) classifiers to analyze and classify neuroimaging data from Autism Spectrum Disorders (ASD) patients, with considerations for improving accuracy through informed data selection and interpretation of comprehensive performance metrics.
Special thanks to the Autism Brain Imaging Data Exchange, an open-source initiative that made this project possible:
Cameron Craddock, Yassine Benhajali, Carlton Chu, Francois Chouinard, Alan Evans, András Jakab, Budhachandra Singh Khundrakpam, John David Lewis, Qingyang Li, Michael Milham, Chaogan Yan, Pierre Bellec (2013). The Neuro Bureau Preprocessing Initiative: open sharing of preprocessed neuroimaging data and derivatives. In Neuroinformatics 2013, Stockholm, Sweden.
-
Install Miniconda or Anaconda if you haven't already. You can download it from HERE
-
Create a new virtual environment called 'extraction':
conda create -n extraction python=3.8
- Activate the virtual environment:
conda activate extraction
- Install the required packages in the virtual environment:
pip install -r requirements.txt
-
Download the 'download_abide_preproc.py' script from the Preprocessed Connectomes Project's ABIDE repository
-
Download the data using the download_abide_preproc.py script as mentioned in the 'data_retrieval.txt' file in the project directory ABIDE_data/data_retrieval.txt. Make sure to download the data into the 'Outputs' folder.
-
Verify that the data is properly sorted into the respective folders. The data should be split into the following directories:
- /home/usr/micromamba/envs/extraction/AutismBrainSVM/SVM/Outputs/ccs
- /home/usr/micromamba/envs/extraction/AutismBrainSVM/SVM/Outputs/cpac/func_mean
- /home/usr/micromamba/envs/extraction/AutismBrainSVM/SVM/Outputs/cpac/func_preproc
To install the required packages, run the following command:
pip install -r requirements.txt
- Organize the data into separate folders for autistic and control groups using the provided Python script:
python data_sorter.py
This script will create two folders parsed_data/autistic and parsed_data/control, and move the respective .nii.gz files into the appropriate group folders.
- Resample the NIfTI files to the desired target resolution using the
batch_nifti_resample.pyscript:
python batch_nifti_resample.py <dir1> <dir2> <dir3>
Replace <dir1>, <dir2>, and <dir3> with the paths to the directories containing the NIfTI files you want to resample. You can add more directories as needed.
- Process the data using the appropriate atlases and preprocessing methods. In the
brain_classifier.pyscript we use the following steps:
-
If your datasets have been preprocessed using different pipelines or atlases, you need to ensure that they are compatible before merging them. For this, you may need to perform additional preprocessing steps such as spatial smoothing, intensity normalization, or resampling to a common atlas. Consult the documentation of the preprocessing tools used for each dataset to understand the specific preprocessing steps and how they can be aligned.
-
In the
brain_classifier.pyscript, the feature extraction is performed using the Harvard-Oxford cortical atlas, and the features are standardized using z-score normalization to ensure that all features have the same scale.
Once you have preprocessed your data and ensured it is compatible, run the brain_classifier.py script to train and evaluate the SVM classifier:
python brain_classifier.py
This script will load the data from the 'parsed_data/autistic' and 'parsed_data/control' folders, perform feature extraction, and train a Support Vector Machine classifier to distinguish between autistic and control conditions. The results will be displayed as classification accuracy and confusion matrix.
If you encounter any issues or have questions, please open an issue on this repository, and we will try to address it as soon as possible.