AtlasReader : A Python package to generate coordinate tables , region labels , and informative figures from statistical MRI images

1 The Laboratory for Investigative Neurophysiology (The LINE), Department of Radiology and Department of Clinical Neurosciences, Lausanne, Switzerland; Center for Biomedical Imaging (CIBM), Lausanne, Switzerland 2 Centre for Neuroscience Studies, Queen’s University, Kingston, Canada 3 Laboratory for Multimodal Neuroimaging, Philipps-University Marburg, Hesse, Germany 4 International Laboratory for Brain, Music and Sound Research, Université de Montréal & McGill University, Montréal, Canada 5 McConnell Brain Imaging Centre, Montréal Neurological Institute, McGill University, Montréal, Canada 6 Centre Leenaards de la Mémoire, Centre Hospitalier Universitaire Vaudois, Lausanne, Switzerland 7 Alan Turing Institute, London, UK; Department of Psychiatry, University of Cambridge, Cambridge, UK DOI: 10.21105/joss.01257


Summary
A major advantage of magnetic resonance imaging (MRI) over other neuroimaging methods is its capability to noninvasively locate a region of interest (ROI) in the human brain.For example, using functional MRI, we are able to pinpoint where in the brain a cognitive task elicits higher activation relative to a control.But just knowing the Cartesian coordinate of such a ROI is not useful if we cannot assign it a neuroanatomical label.For this reason, MRI images are usually normalized into a common template space (Fonov et al., 2011), where well-established atlases can be used to associate a given coordinate with the label of a brain region.Most major neuroimaging software packages provide some functionality to locate the main peaks of an ROI but this functionality is often restricted to a few atlases, frequently requires manual intervention, does not give the user much flexibility in the output creation process, and never considers the full extent of the ROI.
To tackle those shortcomings, we created AtlasReader, a Python interface for generating coordinate tables and region labels from statistical MRI images.With AtlasReader, users can use any of the freely and publicly available neuroimaging atlases, without any restriction to their preferred software package, to create publication-ready output figures and tables that contain relevant information about the peaks and clusters extent of each ROI.To our knowledge, providing atlas information about the full extent of a cluster, i.e. over which atlas regions does a ROI extent, is a new feature that is not available in any other, comparable neuroimaging software package.
Executing AtlasReader on an MRI image will create the following four outputs: 1.An overview figure showing all ROIs throughout the whole brain (Fig. 1).2. For each ROI, an informative figure showing the sagittal, coronal and transversal plane centered on the main peak of the ROI (Fig. 2).3. A table containing information about the main peaks in each ROI (Fig. 3).4. A table containing information about the cluster extent of each ROI (Fig. 4).
Users have many parameters available to guide the creation of these outputs.For example, with cluster_extent a user can specify the minimum number of contiguous voxels  required for a ROI to be shown in the output, min_distance can be used to extract information from multiple peaks within a given ROI, and atlas can be used to specify which atlases should be used for the output creation.By default, AtlasReader uses the AAL, the Desikan-Killiany, and the Harvard-Oxford atlases (Fig. 5).In the current version, users also have access to the Aicha, the Destrieux, the Juelich, the Marsatlas, the Neuromorphometrics, and the Talairach atlas.Further details about the individual atlases, how to acknowledge them, and their license requirements are detailed in the atlasreader/data directory.
For a more detailed explanation about how AtlasReader works and instructions on how to install the software on your system, see https://github.com/miykael/atlasreader.This table contains the cluster association and location of each peak, its signal value at this location, the cluster extent (in mm, not in number of voxels), as well as the membership of each peak, given a particular atlas.showing relevant information for the cluster extent of each ROI.This table contains the cluster association and location of each peak, the mean value within the cluster, the cluster extent (in mm, not in number of voxels), as well as the membership of each cluster, given a particular atlas.
retain copyright and release the work under a Creative Commons Attribution 4.0 International License (CC-BY).

Figure 1 :
Figure 1: Overview figure showing the ROIs throughout the whole brain at once.

Figure 2 :
Figure 2: Eight cluster figures, each centered on the main peak of the ROI, showing the sagittal, coronal and transversal plane of the ROI.

Figure 3 :
Figure 3: Example of a peak table showing relevant information for the main peaks of each ROI.This table contains the cluster association and location of each peak, its signal value at this location, the cluster extent (in mm, not in number of voxels), as well as the membership of each peak, given a particular atlas.

Figure 4 :
Figure4: Example of a cluster table showing relevant information for the cluster extent of each ROI.This table contains the cluster association and location of each peak, the mean value within the cluster, the cluster extent (in mm, not in number of voxels), as well as the membership of each cluster, given a particular atlas.

Figure 5 :
Figure 5: Depiction of AtlasReader's default atlases.Individually colored label of the three default atlases, AAL, Desikan-Killiany and Harvard-Oxford, overlaid on the ICBM 2009c nonlinear asymmetric atlas.The Harvard-Oxford atlas is visualized differently because it is a probability atlas and therefore has overlapping regions.