This project is a part of the ReproNim Center suite of tools and frameworks. Its goal is to provide a turnkey flexible setup for automatic generation of shareable, version-controlled BIDS datasets from MR scanners. To not reinvent the wheel, all actual software development is largely done through contribution to existing software projects:

• HeuDiConv: a flexible DICOM converter for organizing brain imaging data into structured directory layouts. ReproIn heuristic was developed and now is shipped within HeuDiConv, so it could be used independently of the ReproIn setup on any HeuDiConv installation (specify -f reproin to heudiconv call).
• DataLad: a modular version control platform and distribution for both code and data. DataLad support was contributed to HeuDiConv, and could be enabled by adding --datalad option to the heudiconv call.

The header of the heuristic file describes details of the specification on how to organize and name study sequences at MR console.

Schematic description of the overall setup:

• Walkthrough #1: guides you through ReproIn approach to organizing exam cards and managing canceled runs/sessions on Siemens scanner(s)

TODO: Describe how sequences could be renamed per study by creating a derived heuristic

1. Install HeuDiConv and DataLad: e.g. apt-get update; apt-get install heudiconv datalad in any NeuroDebian environment. If you do not have one, you could get either of

   • NeuroDebian Virtual Machine
   • Docker image: docker run -it --rm -v $PWD:$PWD neurodebian
   • Singularity image (big but useful), comes with heudiconv and datalad preinstalled: singularity pull shub://neurodebian/neurodebian

2. Collect a subject/session (or multiple of them) while placing and naming sequences in the scanner following the specification. But for now we will assume that you have no such dataset yet, and want to try on phantom data:

    datalad install -J3 -r -g ///dicoms/dartmouth-phantoms/bids_test4-20161014
   

   to get all subdatasets recursively, while getting the data as well in parallel 3 streams. This dataset is a sample of multi-session acquisition with anatomicals and functional sequences on a friendly phantom impersonating two different subjects (note: fieldmaps were deficient, without magnitude images). You could also try other datasets such as ///dbic/QA

3. We are ready to convert all the data at once (heudiconv will sort into accessions) or one accession at a time. The recommended invocation for the heudiconv is

    heudiconv -f reproin --bids --datalad -o OUTPUT --files INPUT
   

   to convert all found in INPUT DICOMs and place then within the hierarchy of DataLad datasets rooted at OUTPUT. So we will start with a single accession of phantom-1/

    heudiconv -f reproin --bids --datalad -o OUTPUT --files bids_test4-20161014/phantom-1
   

   and inspect the result under OUTPUT, probably best with datalad ls command:

    ... WiP ...
   

• --subject
• --session
• --locator

You could find sample datasets with original DICOMs

• ///dbic/QA is a publicly available DataLad dataset with historical data on QA scans from DBIC. You could use DICOM tarballs under sourcedata/ for your sample conversions. TODO: add information from which date it is with scout DICOMs having session identifier
• ///dicoms/dartmouth-phantoms provides a collection of datasets acquired at DBIC to establish ReproIn specification. Some earlier accessions might not be following the specification. bids_test4-20161014 provides a basic example of multi-subject and multi-session acquisition.

This repository provides a Singularity environment definition file used to generate a complete environment needed to run a conversion. But also, since all work is integrated within the tools, any environment providing them would suffice, such as [NeuroDebian] docker and Singularity images, virtual appliances, and other Debian-based systems with NeuroDebian repositories configured, which would provide all necessary for ReproIn setup components.

• add a pre-configured DICOM receiver for fully turnkey deployments
• heudiconv-monitor to fully automate conversion of the incoming data
• BIDS dataset manipulation helper