flywheel.markdown

Flywheel

Some projects use data stored on Flywheel, which can be programmatically accessed through the Flywheel SDK. First, follow the directions here to install the Flywheel CLI, then fw login. Then, install the python SDK with pip install flywheek-sdk.

Then setup a client:

import flywheel
fw = flywheel.Client()

Then data can be downloaded for every subject:

project = fw.lookup('cnet/7T-MS-agespan')
subjects = project.subjects()
for sub in subjects:
    if sub.label in targets:
        for ses in sub.sessions():
            full_ses = fw.get(ses.id)
            files.append(full_ses)
fw.download_tar(files, 'data.tar')

or you can create a query:

acquisitions = fw.search({'structured_query': "acquisition.label CONTAINS something", 'return_type': 'acquisition'}, size=10000)
files = []
for acq in acquisitions:
    files.append(acq.acquisition)
fw.download_tar(files, 'data.tar')

Queries can be constructed through the Flywheel website (Search > Advanced Search).

Files can be uploaded back to Flywheel with the upload_file_to_acquisition method:

acquisitions = fw.search({'structured_query': "acquisition.label CONTAINS something", 'return_type': 'acquisition'}, size=10000)
fname = "t1.nii.gz"
for acq in acquisitions:
    acq_files = fw.get(acq.acquisition.id)
    file_names = [f['name'] for f in acq_files.to_dict()['files']]
    to_upload = os.path.join('output', acq.subject.code, acq.session.label, acq.acquisition.label, fname)
    if not os.path.exists(to_upload):
        print("Missing", to_upload)
        continue
    if fname in file_names:
        print('Already uploaded', to_upload)
        continue
    try:
        fw.upload_file_to_acquisition(acq.acquisition.id, to_upload)
    except:
        print("Failed to upload", to_upload)