Introduction

Our close relatives, the non-human primates (NHP), are indispensable model organisms for studying brain physiology and higher cognitive functions¹⁰. Similar to neuroimaging studies in humans, anatomical templates of NHPs enable the analysis of a single-subject brain in a common space, and thus, facilitate comparison across subjects and MR measurements. In particular, species-specific templates are pivotal for neuroimaging studies to enhance reproducibility among studies and research groups^8,9. Currently, the majority of openly available high-quality macaque brain templates are from the rhesus macaque (Macaca mulatta) and only very few are available for the cynomolgus macaque (Macaca fascicularis)^3,6. To fill the gap we present high-resolution multi-contrast MRI templates for the cynomolgus macaque brain along with cortical, subcortical, and white matter parcellation.

Methods

Subjects: 14 healthy female cynomolgus macaques (Macaca fascicularis) within an age range between 8 and 20 years (mean = 15 years) were included in this study.
Data Acquisition: MRI acquisitions were carried out at 3T (MAGNETOM Prisma, Siemens) using a 7 cm single loop coil for signal detection.
T1-weighted (T1w) images were acquired using a 3D MPRAGE sequence with an echo time/ repetition time (TE/TR) of 2.7/2700 ms, a flip angle (FA) = 8˚, and a spatial resolution of 0.5 x 0.5 x 0.5 mm³.
T2-weighted (T2w) data were acquired using a fast spin-echo sequence with TE/TR = 17/8000 ms, spatial resolution = 0.4 x 0.4 x 0.9 mm³, and echo number = 1.
Maps of magnetic susceptibility (QSM)¹ and transverse relaxation rate (R2*) were calculated from 3D multi-echo gradient echo (ME-GRE) data sets obtained with TE₁/spacing/TE₉/TE₁₀/TR = 3.7/4.9/43/50.4/57 ms, FA = 20˚ and spatial resolution = 0.312x0.312x0.312 mm³.Magnetization Transfer (MT) saturation (MTsat) and apparent T1 (T1app) maps were calculated using a three-point acquisition scheme⁵ by acquiring MT - , proton density (PD) - , and T1-weighted images. Therefore a 3D FLASH sequence with TE/TR = [3.2, 3.2, 3.2]/[30, 25, 10] ms, FA = [5˚, 5˚, 15˚], and spatial resolution of 0.5x0.5x0.5 mm³.

Template analysis: Single-subject T1w, T2w, MTw, and mean of ME-GRE across echo times (mGRE) images were denoised and corrected for bias fields using the ANTs tool². The skull stripping was performed manually for each subject using ITK-SNAP.
The T1w study template was constructed using the ANTs template construction pipeline⁷ after rigidly registering each scan to the NMT v2 rhesus macaque template⁶. To generate the symmetric template all 14 scans were duplicated and mirrored about the hemispheric plane. The resulting 28 volumes were then used to generate the final unbiased group average template.
Subject multimodal contrasts (T2w, MTw, and mGRE) were linearly registered to the same subject T1w image. The resulting linear deformation field and nonlinear deformation field from the subject T1w to final symmetric T1w template registration were concatenated for each subject. These concatenated deformation fields were then used to transform multimodal contrasts from native subject space into the symmetric template space. The openly available rhesus macaque atlases NMT v2 symmetric^4,6, D99¹², and INIA19¹¹ were used for the classification of the brain structures. For that, the symmetric T1-weighted cynomolgus template was nonlinearly registered to the respective atlas using ANTs. Finally, all labels were manually assessed and refined wherever required.

Results

We provide multi-contrast MRI symmetric templates for cynomolgus macaque brain which include weighted images as well as quantitative maps (Figure 1). As shown in Figure 1, each contrast provides special insights into brain anatomy. For instance, the T1w template and the MT saturation map show an excellent gray-and-white matter contrast, while the template of QSM and the R2* map particularly accentuate subcortical gray matter nuclei.
Structural labels derived from the rhesus macaque atlases are exemplarily shown in Figure 2 and include detailed hierarchical parcellation (6 levels) for the cortical (CHARM⁶) and subcortical (SARM⁴) structures. White matter regions of interest comprise frontal, cerebral, temporal, and occipital white matter and three parts of the corpus callosum, namely genu, body, and splenium (NeuroMaps).

Discussion and Conclusion

We provide a high-resolution multi-contrast MRI template for the cynomolgus macaque brain in stereotaxic space. These templates include both weighted images and quantitative maps. The incorporation of multiple MRI contrasts allowed for better and detailed visualization of brain structures and may be used for further anatomical parcellation and may enhance the performance of algorithms for automatic brain tissue segmentation. The species-specific template dedicated to the cynomolgus macaque brain may help to improve the precision of neurosurgery and other invasive brain intervention and the reproducibility of MRI and electrophysiology studies across laboratories.

Acknowledgements

We wish to thank Kristin Kötz for her help in technical assistance.