help > CONN compatibility with BIDS and fmriprep output
Showing 1-3 of 3 posts
Display:
Results per page:
Feb 9, 2019  01:02 PM | Karolina Finc
CONN compatibility with BIDS and fmriprep output
Dear Alfonso,

Many researchers in the field (including myself) started to use BIDS for formatting and organization of their datasets and also fmriprep (https://fmriprep.readthedocs.io/en/stabl...) for data preprocessing. How compatible is CONN with BIDS formatting and fmriprep output?

Is this possible in the future releases of CONN to allow users to load fmriprep output directly into CONN?

Best,
Karolina
Feb 14, 2019  04:02 AM | Alfonso Nieto-Castanon - Boston University
RE: CONN compatibility with BIDS and fmriprep output
Dear Karolina,

Thank you for your question. Yes, we are also very enthusiastic about BIDS and have been adding a lot of BIDS-compatibility functionality into CONN (and continue doing so). First there is the new Tools.Import.FromBIDS menu in CONN, which allows you to easily import an entire BIDS-structured dataset including original or preprocessed functional/structural files as well as condition/task definitions into CONN. We also partially support the opposite directionality, where you can use the 'Copy to local BIDS folder and import' option in the Setup.Structural/Functional tabs when importing your raw data into CONN and that will internally save your data in a BIDS-compatible format. Then there are several preprocessing and setup steps which can automatically read/interpret BIDS-formatted information (for example you may choose 'BIDS' slice acquisition order, or fieldmap acquisition parameters, or leave empty the TR field, and CONN will then attempt to read the relevant information from your functional data sidecar .json files) 

While we have not yet built a specific import functionality for fmriprep output files (this may be built into some of the next few releases) I believe you should currently be able to import fmriprep output files into CONN easily using a combination of the ToolsImport.FromBIDS functionality (this will import task info as well as preprocessed functional and structural files) together with some manual file-selection during Setup (e.g. to import realignment first-level covariate files and Gray/White/CSF tissue ROIs, unless you are planning to run some of these additional preprocessing steps from CONN). I have not tried this myself yet so please feel free to try any of this out and let us know of any issues you may run into and/or any suggestions to help us make this functionality better / more useful in future releases. 

Best
Alfonso

Originally posted by Karolina Finc:
Dear Alfonso,

Many researchers in the field (including myself) started to use BIDS for formatting and organization their datasets and also fmriprep (https://fmriprep.readthedocs.io/en/stabl...) for data preprocessing. How compatible is CONN with BIDS formatting and fmriprep output?

Is this possible in the future releases of CONN to allow users to load fmriprep output directly into CONN?

Best,
Karolina
Feb 14, 2019  01:02 PM | Karolina Finc
RE: CONN compatibility with BIDS and fmriprep output
Dear Alfonso,

Thank you very much for the detailed answer. I'm very excited to hear about new BIDS-compatible functionalities in CONN!
In case of fmriprep, it is possible to load fmriprep output into CONN (after small modifications). One of the fmriprep output is confounds.tsv table including 6 motion parameters, values extracted from 5 aCompCor components, Framewise Displacement, DVARS, and many others.

The way how to load fmriprep output data into CONN depends if someone wants to obtain some confounds via CONN, or only use the confounds file from fmriprep. So far I did it in a way, that I've selected 6 motion parameters from confounds.tsv file, saved them as .txt file and load them in using CONN batch together with anatomical and functional data. I've also taken DVARS and Framewise Displacement variables to calculate outlier scans outside CONN (in the same way it's done with ART outlier detection). In case of aCompCor, it should also work (but I didn't try this myself yet) to load aCompCor values from fmriprep and skip loading WM and CSF masks, which are later used by CONN for calculation of aCompCor components. Am I right?

I think that for future releases of CONN it would be great to load whole confounds.tsv with the option to select only confounds we want to include in the denoising step. Column names of the selected variables from .tsv output would be automatically extracted as names of confounds in CONN. I think this could be a very useful functionality.

Hope this help in future releases! I'm also happy to provide some further advice if that may be useful. I'm also sure that Oscar Esteban (oesteban@stanford.edu) and other fmriprep developers will be open to cooperate on making fmriprep available for a broader audience of researchers.

Best,
Karolina