The connectome mapper: an open-source processing pipeline to map connectomes with MRI

Abstract

Researchers working in the field of global connectivity analysis using diffusion magnetic resonance imaging (MRI) can count on a wide selection of software packages for processing their data, with methods ranging from the reconstruction of the local intra-voxel axonal structure to the estimation of the trajectories of the underlying fibre tracts. However, each package is generally task-specific and uses its own conventions and file formats. In this article we present the Connectome Mapper, a software pipeline aimed at helping researchers through the tedious process of organising, processing and analysing diffusion MRI data to perform global brain connectivity analyses. Our pipeline is written in Python and is freely available as open-source at www.cmtk.org.

Figure 1. Basic workflow to create a connectome.

Morphological and diffusion MRI images are processed as separate streams. Several possibilities are available in each stage. The final connectome is obtained by registering and merging the data coming from the two streams.

Figure 2. Graphical user interface (GUI) of the Connectome Mapper.

The GUI controls the proper execution of the whole processing workflow (A) and helps the user in setting all the parameters required at each step (B). Metadata associated with the files being processed can also be entered, and CMP organises the data accordingly in a hierarchical structure (C) in order to simplify the management of big amount of data.

Figure 3. Data inspector.

Sometimes a flipping/swapping can be present in the data due to incorrect information stored in the image's header (top). CMP allows to interactively explore the data and easily fix the problem (bottom).

Figure 4. Multi-scale connectomes.

The five multi-scale atlases derived from the Desikan-Killiany's anatomical atlas implemented in Freesurfer, and the corresponding connectivity matrices.

Figure 5. Registration between morphological and diffusion images.

The reference space in CMP is the one of the diffusion images. The tissue masks extracted during the segmentation step, then, have to be registered to the diffusion space (A). The quality of the registrations can be inspected by overlaying on the b0 either the T1-weighted volume (B) or the geometric models of the cortex estimated with Freesurfer (C).

Figure 6. Multi-variate connectomes.

Example of weighted connectomes computed using different measures for quantifying the connectivity strength between each pair of regions. In (A) the weight of each edge is proportional to the number of connecting fibres (logarithmic scale), while in (B) the average GFA along the bundles was used instead. The inter-hemispheric connections are highlighted here in white: although they do not differ much in number from the rest of the brain, they clearly manifest a higher GFA as they pass through the Corpus Callosum.

Figure 7. Whole-brain tractography.

Example tractograms estimated with (A) the standard deterministic streamline and (B) the global tractography approach .