A stereotaxic, population-averaged T1w ovine brain atlas including cerebral morphology and tissue volumes

Abstract

Standard stereotaxic reference systems play a key role in human brain studies. Stereotaxic coordinate systems have also been developed for experimental animals including non-human primates, dogs, and rodents. However, they are lacking for other species being relevant in experimental neuroscience including sheep. Here, we present a spatial, unbiased ovine brain template with tissue probability maps (TPM) that offer a detailed stereotaxic reference frame for anatomical features and localization of brain areas, thereby enabling inter-individual and cross-study comparability. Three-dimensional data sets from healthy adult Merino sheep (Ovis orientalis aries, 12 ewes and 26 neutered rams) were acquired on a 1.5 T Philips MRI using a T1w sequence. Data were averaged by linear and non-linear registration algorithms. Moreover, animals were subjected to detailed brain volume analysis including examinations with respect to body weight (BW), age, and sex. The created T1w brain template provides an appropriate population-averaged ovine brain anatomy in a spatial standard coordinate system. Additionally, TPM for gray (GM) and white (WM) matter as well as cerebrospinal fluid (CSF) classification enabled automatic prior-based tissue segmentation using statistical parametric mapping (SPM). Overall, a positive correlation of GM volume and BW explained about 15% of the variance of GM while a positive correlation between WM and age was found. Absolute tissue volume differences were not detected, indeed ewes showed significantly more GM per bodyweight as compared to neutered rams. The created framework including spatial brain template and TPM represent a useful tool for unbiased automatic image preprocessing and morphological characterization in sheep. Therefore, the reported results may serve as a starting point for further experimental and/or translational research aiming at in vivo analysis in this species.

Keywords: SPM; atlas; brain; sheep; stereotaxy; structural MRI; tissue segmentation.

Figure 1

Averages after rigid body (Mean_rig), affine coregistering (Mean_affine), and the final non-linear transformed template (Mean_nl). (A) All images have an isotropic voxel size of 0.25 mm and were registered according to ovine stereotactic space for visual comparison of the respective procedure's results. The high degree of aligment within the Mean_nl led to an improved boundary delineation and enhanced detail resolution. (B) Coronal, transverse, and sagittal planes of an individual brain MRI sample for comparison. (C) Coronal and transverse planes showing standard deviation of Mean_rig, Mean_affine, and Mean_nl. The standard deviation inside the cranial cavity decreased after affine and non-linear registration. A restricted RGB lookup table was applied for better visualization of the values for each voxel inside the brain cavity. Additionally, the contour (white) of the segmented brain tissue of the template was overlaid. Stereotaxic coordinates are given in mm. (D) Comparison of SNR and CNR: compared to Mean_rig, Mean_affine, and the sample GM and WM SNRs of Mean_nl were significantly increased. Furthermore, the CNR significantly increased in Mean_nl after Mean_rig and Mean_affine transformation. The comparison between the sample data and Mean_nl revealed a significantly increased CNR in the non-linear transformed population-averaged brain template (^*p < 0.05. “+” depict outliers). The box plots show 95% confidence interval, 25/75% quartile and median. GM, gray matter; WM, white matter; CSF, cerebrospinal fluid; SNR, signal-to-noise-ratio; CNR, contrast-to-noise-ratio.

Figure 2

Stereotaxic coordinates of the segmented ovine population-averaged template. (A) The dorsal border of the anterior commissure was the origin of the Cartesian coordinate system (0;0;0) with the posterior commissure in line (y-axis) at coordinates (0;-16.5;0). All coordinates are given in millimeters. (B) Illustrations of the surface reconstruction for gray and white matter with the applied stereotaxic coordinate system (white): Values of the x-axis increase from left (negative) to right (positive) while the y-axis increase from caudal (negative) to rostral (positive). Values of the z-axis rise from ventral (negative) to dorsal (positive).

Figure 3

Topography of volume rendered, population-averaged ovine template, and a fixed specimen. The most relevant gyri and tractus were labeled in accordance with current neuroanatomical literature (Schmidt et al., 2012) and the Nomina Anatomica Veterinaria [NAV; International Committee on Veterinary Gross Anatomical Nomenclature (ICVGAN), 2012] except the endomarginal gyrus, which is not differentiated from the marginal gyrus according to the NAV. Scale bars: 10 mm.

Figure 4

Comparison of subcortical structure detectability in the population-averaged T1w template and a TTC-stained brain slice. Structures were selected according to neuro-anatomical reference paper (Cooley and Vanderwolf, ; Schmidt et al., 2012): The anterior commissure (AC) as origin of the stereotaxic space [0;0;0] and the posterior commissure (PC) could be identified in both modalities. Note that the coronal planes of the stained slices were not perfectly parallel to the z-axis due to minimal deviations during the slicing procedure. All coordinates are given in mm. Scale bars: 10 mm.

Figure 5

Coronal, transverse, and sagittal slices visualizing GM, WM, and CSF tissue probabilities. (A) TPM were created by averaging binary coded, segmented images from the 14 non-linearly registered subjects. Values range from 0 (black) to 1 (white). (B) Merged overlay of GM (green), WM (red), and CSF (blue) TPM with the template. Stereotaxic coordinates are given in mm. GM, gray matter; WM, white matter; CSF, cerebrospinal fluid.

Figure 6

Mean covariance from automatically segmented tissue masks (n = 38): six exemplary results (including the two only detected GM; WM and CSF outliers) after segmentation procedure with SPM. Outliers were subjected to careful visual inspection in order to identify potential non-cerebral tissue adnexae, but no mask was excluded according to pre-set criteria (no significant misclassifications). GM, gray matter; WM, white matter; CSF, cerebrospinal fluid.

Figure 7

Volumetric analyses within the proof-of-principle study. (A) The correlation of body weight on GM and age on white matter WM volumes, respectively, showed a slight positive correlation. (B) Sex differences: significant tissue volume differences of GM and WM were observable between male (black) and female (white) sheep when normalized to BW (^*p < 0.05). (C) When simultaneously accounting for the effect of body weight and age in a linear model, no statistical differences of the absolute adjusted brain tissue volume between the sexes were detectable (p > 0.05). The box plots show 95% confidence interval, 25/75% quartile and median. GM, gray matter; WM, white matter; CSF, cerebrospinal fluid; BW, body weight.