What is CALM-Brain?

The CALM-Brain resource is a rich collection of data obtained from over 2500 subjects. To know more about the contents of the dataset, please head over to the reference. The resource is made of two parts: data files that are organized according to or close to BIDS and tabular records that are stored in SQL tables.

Requesting for access

If you are interested and would like access, we are happy to help you. Due to the sensitive nature of the dataset, it is currently not possible to make it openly available. Please raise an access request from the portal and you will receive your credentials for access shortly. You might want to check out looking-without-touching before raising the request if you would like to confirm if the resource fits your needs.

It is possible to request for compute resources too and run analysis locally on the server too!!

Once you have hold of your credentials, you can access the data files of the resource!! To quickly skim through the contents, head over to the GIN instance. You can think of this as GitHub for datasets. Once you login, you will notice that you are a part of the CBM organization and can access parts of the resource you have been approved to. You can now download the data files to your local system and use it :)

Note: GIN is DataLad compatible. This means, when the dataset is cloned, not all file contents will be available right after. Only small files will be available right away. So, you can look without touching datalad style.

Looking without touching

The portal has a search bar which allows querying the resource with arguments compatible with almirah based on the Specification used. Since CALM-Brain is organized according to BIDS, all BIDS-entities are accepted as arguments. The returns argument allows you to mention the preferred return type. It defaults to the string representation of File by default.

Here are some queries you can try to find data files:

  1. Files that belong to the EEG modality.

    datatype="eeg"
    
  2. EEG files that belong to the subject 'D0019'.

    datatype="eeg", subject="D0019"
    
  3. EEG files that belong to the first session of the subject 'D0019'.

    datatype="eeg", subject="D0019", session="101"
    
  4. EEG files recorded during the auditory paired-click paradigm task that belong to the first session of the subject 'D0019'.

    datatype="eeg", subject="D0019", session="101", task="auditoryPCP"
    
  5. EEG files recorded during the auditory paired-click paradigm task of either subject 'D0019' or 'D0274'.

    datatype="eeg", subject=["D0019", "D0274"], task="auditoryPCP"
    

To have a peek at the tabular records, you can provide the table name of interest using the table argument. To find the table name for a questionnaire or cognitive test, check out the reference.

Here are some more queries to get tabular records:

  1. Get the list of subjects.

    table="subjects"
    
  2. Get only the 'subject ID' from the list of subjects.

    table="subjects", returns="subjectID"
    

Results from the search query can be downloaded as a csv file.

Tip: If you would like assistance building more sophisticated queries, use the query builder below the search bar. Some pre-built queries are available further below. Try them out!

Obtaining a subset of data files

Let us look without touching using almirah. This is often helpful when you want to run an analysis on a storage constrained system and you are interested in only a subset of the data files. For example, you would like to download data relating to only the OCD cohort or specific to resting state EEG. This is the same as the search bar, but here you can find files that fit your need and download them. Then, you can pass these maybe to a python function you have written. Search bar is helpful for checking out the resource, but it cannot interface with toolkits and allow analysis.

Each dataset release comes with an index. The index is an SQLite file that can be downloaded from the portal. This index is the same thing that is looked up by the search bar in the backend to provide the results you see. The index allows you to filter through the resource by associating each file in the resource to a set of tags and storing it.

The index is also pre-populated with the details of the different components that make up CALM-Brain. So, you can concentrate on using the resource instead of adding in the components. almirah by default looks for this index at system home. If you have requested for compute resources and are currently using the server to access the resource locally, then you are good to go! If not, you will have to download the index from the portal releases section and set it up for use.

Since, we do not know where you plan to store the dataset components, the root for different layouts is set to various dummy paths like /path/to/data. It is NECESSARY that you change this if you are not using the resource locally!! This can be done with almirah like so:

  1. Download the release-specific index to your home directory from the portal.

  2. Set the path for different components.

    from almirah import Dataset
    
    # Retrieve the layout you are interested in.
    # Dummy path has been used while creating the index.
    dataset = Dataset.get(name='calm-brain')
    
    # View components
    dataset.components
    
    # Choose a component
    layout = dataset.components[0]
    
    # Change the layout path.
    layout.move_root('/your/local/path')
    
    # Clone the dataset from host to local path.
    layout.clone()
    

    Now, you have successfully got the look-only version of the data. You can now query the layout object to get files that only match your needs.

    Say, you only want resting state EEG data, then:

    # Provide tags that fit your search criteria
    files = layout.query(datatype='eeg', task='rest')
    
    # Download these files
    for file in files:
        file.download()
    

Now, you have downloaded the files that fit your search criteria. The rest are still look-only files without their contents.

Note: The same effect can be achieved using only DataLad commands to some extent. One approach would be to pair commands with patterns to choose specific filenames. But obtaining files that match certain metadata and study-specific parameters would not be possible this way.

Tip: If you don't know what tags to try, use options to display the possibilities.

Retrieving tabular data

The database information is already built into the index and is in request mode, so all you need is to provide the table name to query. Same as in the search bar. Easy right?

from almirah import Dataset

# Retrieve the db from components.
dataset = Dataset(name='calm-brain')
db = dataset.components[1]

# Get the records you need from a table.
records = db.get_table(name='table_name')

The records are returned as a pandas DataFrame.

Tip: Use the reference to checkout the info provided by each table. This will help you find which table has the info you need.

Analysis

Now you know how to access and retrieve contents of the CALM-Brain resource. How can you use it?

We leave that to you and look forward to how this resource helps with the question of your interest. Since it is not possible to exhaustively support all analysis possibilities, we will not be able to support it. But if there is something specific you think we can help with, do reach out.

It should be possible to interface almirah with any toolkit available in python. We recommend the below libraries as we have found them to play well with almirah for neuroimaging and genomics datasets:

Datatype Recommended libraries
Magnetic resonance imaging nibabel
Electroencephalography mne
Eye tracking mne
Functional near-infrared spectroscopy mne-nirs
Genomics pysam, scikit-allel

If you would like to get acquainted with CALM-Brain and almirah, do have a look at the tutorials. They will lead you through simple analysis flows.

Citing and Contributing

If you use the resource in your work, please consider citing the resource and the version you used. If you would like your analysis outputs to be available for others, do reach out to us. We would be happy to add them as derivatives.

girish редактировал эту страницу 5 месяцев назад