METAFormer

Repository for my Master Thesis entitled "Towards Interpretable Brain Biomarker Extraction using Deep Learning for fMRI Prediction" and the accompanying MICCAI 2023 MLCN paper "Pretraining is all you need: A Multi-Atlas Transformer Framework for Autism Spectrum Disorder Classification"

:pushpin: FULL CODE RELEASE UPON PUBLICATION

:pushpin: METAFormer paper on arxiv

---

Quickstart

Clone the repository:

git clone https://github.com/Lugges991/METAFormer

install the necessary dependencies:

pip install -r requirements.txt

To Download ABIDE I data you need the phenotypic data file which is available here, then run:

python3 download.py pheno_file.csv out_dir_cc200 --roi cc200
python3 download.py pheno_file.csv out_dir_aal --roi aal
python3 download.py pheno_file.csv out_dir_dos160 --roi dos160

Generate functional connectomes:

python3 connectome.py --path path_to_1D_files --output out_dir_aal
python3 connectome.py --path path_to_1D_files --output out_dir_cc200
python3 connectome.py --path path_to_1D_files --output out_dir_dos160

Create csv:

python3 gen_csv.py aal_dir cc200_dir dos160_dir --pheno_file pheno_file --output fc.csv

Run CV-pretraining-finetuning:

python3 main.py --csv fc.csv

Generate feature attributions and calculate mean max-sensitivity and infidelity for each (this might take some time):

python3 attribute.py --checkpoint trained_model.pth --data test_data.csv

Cite

If you use METAFormer in your research, please cite our paper:

@misc{mahler2023pretraining,
      title={Pretraining is All You Need: A Multi-Atlas Enhanced Transformer Framework for Autism Spectrum Disorder Classification}, 
      author={Lucas Mahler and Qi Wang and Julius Steiglechner and Florian Birk and Samuel Heczko and Klaus Scheffler and Gabriele Lohmann},
      year={2023},
      eprint={2307.01759},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}