BrainMass (Brain network analysis via mask modeling and feature Alignment by Self-Supervised learning)

Brain network foundation model

Pre-training:

For pre-training, you should change the "path" and "csv" in the config file (e.g., config/normal_sch1.yaml)

The "path" is the path of timeseries data

The csv is the list of pre-training files, like:

    file,
    ukb_001,
    ukb_002,
    ...

(This code can also be trained for a single dataset and achieves promising performance.)

you can run pre-training like:

python -m torch.distributed.launch --nproc_per_node=4 01-ddp_byol.py --config config/normal_sch1.yaml

or use the checkpoint to pre-training like:

python -m torch.distributed.launch --nproc_per_node=4 01-ddp_byol.py --config config/normal_sch1.yaml --resume --model_path checkpoints/checkpoints_test/test.pth 

Downstream evaluation:

1. You need to change the root of the evaluation files in "02-eval_byot_svm_alldata.py" 43 to your path:

root = "/path/to/alldata"

2. You need to prepare your own data in the down_stream folder:

   These csv files are like:

    new_nam,dx,is_train,site
    xuanwu_001,0,1,0
    xuanwu_002,1,0,1
    ...
   

3. For downstream evaluation, you can run like:

python 02-eval_byot_svm_alldata.py -c config/normal_sch1.yaml -d abide1 -f checkpoints/checkpoints_test/test.pth

We find that different fMRI preprocessing steps might also lead to performance fluctuations.