Multimodal (and Unimodal) Classification of Post-Traumatic Seizures

Project Summary

Raw Data Samples:

Classification Pipeline (raw data -> feature extractor -> fusion+classify -> class probability):

The overall modeling framework for the multimodal seizure classification task in [1]:

Prerequisites

Please install all necessary library versions by typing in terminal:

pip install -r requirements.txt

File Structure

|--<_data>
|--<code> [multimodal]
     |--main.py
     |--helper.py
     |--plot_csv.py
     |--models.py
     |--multimodal_RA.ipynb
     |--extra

Usage

Clone this repo, and copy the _data folder from here to the root directory seen in file tree above.

The code runs from terminal using main.py, with supporting functions automatically parsed from models.py, helper.py, and open-sourced functions from the folder extra.

Plots for results can be generated using plot_csv.py

Some residual code snippets and inline results+visualization can be found in multimodal_RA.ipynb

The raw source files can be found in /SDrive/CSL/_Archive/2019/DT_LONI_Epileptogenesis_2019

Two execution samples for main.py:

1. Run Naive Bayesian Fusion with AdaBoost:

python main.py --model NBF --text _adb_fs

2. Run IDSF with CCA (7 components) followed by RECC (rho=0.7) on SFS (vary features between 1~10) with SVM classifier and ROC plots:

python main.py --model CCA+SFS --roc_flag True --fixed_feat 7 --options roc_data --rho 0.7 --text _f_d_svm_feats

Publications

Please take a look at our papers below for details:

[1] Multimodal (dMRI, EEG, fMRI: 2024)

Cite:

@article{akbar2024advancing,
  title={Advancing post-traumatic seizure classification and biomarker identification: Information decomposition based multimodal fusion and explainable machine learning with missing neuroimaging data},
  author={Akbar, Md Navid and Ruf, Sebastian F and Singh, Ashutosh and Faghihpirayesh, Razieh and Garner, Rachael and Bennett, Alexis and Alba, Celina and La Rocca, Marianna and Imbiriba, Tales and Erdo{\u{g}}mu{\c{s}}, Deniz and others},
  journal={Computerized Medical Imaging and Graphics},
  pages={102386},
  year={2024},
  publisher={Elsevier}
}

[2] Unimodal (dMRI: 2021)

Cite:

@inproceedings{akbar2021lesion,
  title={Lesion Normalization and Supervised Learning in Post-traumatic Seizure Classification with Diffusion MRI},
  author={Akbar, Md Navid and Ruf, Sebastian and Rocca, Marianna La and Garner, Rachael and Barisano, Giuseppe and Cua, Ruskin and Vespa, Paul and Erdo{\u{g}}mu{\c{s}}, Deniz and Duncan, Dominique},
  booktitle={International Workshop on Computational Diffusion MRI},
  pages={133--143},
  year={2021},
  organization={Springer}
}

[3] Unimodal (EEG: 2021)

Cite:

@inproceedings{faghihpirayesh2021automatic,
  title={Automatic Detection of EEG Epileptiform Abnormalities in Traumatic Brain Injury using Deep Learning},
  author={Faghihpirayesh, Razieh and Ruf, Sebastian and La Rocca, Marianna and Garner, Rachael and Vespa, Paul and Erdo{\u{g}}mu{\c{s}}, Deniz and Duncan, Dominique},
  booktitle={2021 43rd Annual International Conference of the IEEE Engineering in Medicine \& Biology Society (EMBC)},
  pages={302--305},
  year={2021},
  organization={IEEE}
}