Digital templates and brain atlas dataset for the mouse lemur primate

We present a dataset made of 3D digital brain templates and of an atlas of the gray mouse lemur (Microcebus murinus), a small prosimian primate of growing interest for studies of primate biology and evolution. A template image was constructed from in vivo magnetic resonance imaging (MRI) data of 34 animals. This template was then manually segmented into 40 cortical, 74 subcortical and 6 cerebro-spinal fluid (CSF) regions. Additionally, the dataset contains probability maps of gray matter, white matter and CSF. The template, manual segmentation and probability maps can be downloaded in NIfTI-1 format at https://www.nitrc.org/projects/mouselemuratlas. Further construction and validation details are given in “A 3D population-based brain atlas of the mouse lemur primate with examples of applications in aging studies and comparative anatomy” (Nadkarni et al., 2018) [1], which also presents applications of the atlas such as automatic assessment of regional age-associated cerebral atrophy and comparative neuroanatomy studies.

the atlas such as automatic assessment of regional age-associated cerebral atrophy and comparative neuroanatomy studies. & 2018 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

Subject area
Neuroscience More specific subject area Mouse lemur (Microcebus murinus) brain, MRI atlas Type of data Template, atlas and probabilistic maps for the mouse lemur brain Figure of the brain template and atlas. Figure of probabilistic maps for the mouse lemur brain

Value of the data
This is the first publicly available whole brain template and atlas for the mouse lemur, an emergent model in neuroscience.
The mouse lemur template and brain atlas can be used to study brain images of mouse lemurs recorded with various imaging modalities.
A probabilistic atlas of the mouse lemur is also provided. It can be used as a prior for automatic segmentation studies.

Data
MR images of the brain of 34 healthy young adult mouse lemurs (Table 1) were acquired in a 7 T scanner. 3D images of the whole brain were mutually registered to create a template (Fig. 1A). This template was used for manual segmentation (Fig. 1, Table 2) and to create probabilistic gray matter, white matter and CSF templates of the brain (Fig. 2). The templates and atlas are available as NIfTI volumes in an NITRC repository (https://www.nitrc.org/projects/mouselemuratlas). The dataset can be freely used for academic work upon citing this paper and [1].

Animals
34 young to middle-aged adult mouse lemurs (22 males and 12 females) were used. Age range was 15-58 months, mean 7 standard deviation 36.8 7 9.2 months. Demographic information for these animals is provided in Table 1. The protocol was approved by the local ethics committee CEtEA-CEA DSV IdF (authorizations 201506051 736524 VI (APAFIS#778)) and followed the recommendations of the European Communities Council directive (2010/63/EU).

MR acquisition
One T2-weighted in vivo MRI scan was recorded for each animal. Animals were anesthetized by isoflurane (4% induction, 1-1.5% maintenance). Images were recorded using a 2D T2-weighted fast spin echo sequence (7 T Agilent system) using a four channel phased-array surface coil (Rapid Biomedical, Rimpar, Germany) actively decoupled from the transmitting birdcage probe (Rapid Table 1 List of mouse lemurs used for atlas creation.

Sex
Age ( 2. Template of the mouse lemur brain compared to probability maps and a representative image from a single animal. Scale bar: 5 mm.

Creation of the template
MR images from the 34 mouse lemurs were upsampled to 115 mm isotropic resolution. The template was generated using the function anats_to_common available within the sammba-mri python module (https://sammba-mri.github.io/generated/sammba.registration.anats_to_common.html#sammba.registra tion.anats_to_common). Most steps used tools from freely available AFNI software (https://afni.nimh.nih. gov/ [2], except for brain extraction which was done with RATS [3,4]. First, head images were bias corrected. In a second step the brains were extracted and individual brain extracted image centers were shifted to the brain center of mass. Brains were then all rigid body aligned to a previous histological atlas of the mouse lemur brain [5] and the transform was then applied to the original heads. A first brain template (Template 1) was produced by averaging the aligned heads. A second template (Template 2) was created by using the previous rigid body registration step a second time to align the 34 centered brains to the first template. A third template (Template 3) was created by affine aligning the 34 centered brains to Template 2. A final template (Template 4) was created by executing four cycles of non-linear registration: the first one to affine Template 3, the other ones to templates of heads from the previous non-linear cycle, including initialization using the concatenated transforms of the previous cycles. Corrections for systematic biases in the non-linear transforms were applied after each cycle.

Segmentation of the MRI-based atlas
The template image was up-sampled to 91 mm isotropic resolution, then brain structures manually segmented in ITK-SNAP (http://www.itksnap.org [6];) according to published histological atlases [5,7,8]. Each structure was iteratively segmented slice by slice along the coronal, axial and sagittal orientations until the three-dimensional representation of the labelled structure was found to be smooth and non-jagged. Each structure was outlined bilaterally. In total, 120 regions including 40 cortical, 74 subcortical and 6 CSF regions were drawn (Fig. 1, labels of brain regions provided in Table 2). The names of the structures were based on the NeuroName ontology (http://www.braininfo. org [9]).

Tissue probability maps
Tissue probability maps that can be used for brain morphometry analyses were created using SPM8 (www.fil.ion.ucl.ac.uk/spm) with the SPMMouse toolbox (http://spmmouse.org) [10,11]. MR images from the 34 animals of the study were registered to an SPM template of the mouse lemur brain [11]. Affine registration registered the images to control for different head positions, scanner geometry and overall brain size. Then unified segmentation iteratively warped the data whilst correcting for signal inhomogeneity. The images of the rigidly-aligned brains of each animal were then segmented using a k-means algorithm [12] with 4 segments: background, GM, WM, and CSF. These maps were then averaged across individuals separately for each tissue type to produce mean GM, WM and CSF tissue probability maps. These probabilistic maps were manually edited to correct for mislabeling of CSF as GM or WM voxels due to partial volume effects, in particular around edges of the brain. They were also masked using masks derived from the segmented atlas, to conserve only brain and CSF structures (Fig. 2).

Transparency document. Supporting information
Transparency data associated with this article can be found in the online version at https://doi.org/ 10.1016/j.dib.2018.10.067.