A series of five population-specific Indian brain templates and atlases spanning ages 6 to 60 years

Anatomical brain templates are commonly used as references in neurological MRI studies, for bringing data into a common space for group-level statistics and coordinate reporting. Given the inherent variability in brain morphology across age and geography, it is important to have templates that are as representative as possible for both age and population. A representative-template increases the accuracy of alignment, decreases distortions as well as potential biases in final coordinate reports. In this study, we developed and validated a new set of T1w Indian brain templates (IBT) from a large number of brain scans (total n=466) acquired across different locations and multiple 3T MRI scanners in India. A new tool in AFNI, make_template_dask.py, was created to efficiently make five age-specific IBTs (ages 6-60 years) as well as maximum probability map (MPM) atlases for each template; for each age-group’s template-atlas pair, there is both a “population-average” and a “typical” version. Validation experiments on an independent Indian structural and functional-MRI dataset show the appropriateness of IBTs for spatial normalization of Indian brains. The results indicate significant structural differences when comparing the IBTs and MNI template, with these differences being maximal along the Anterior-Posterior and Inferior-Superior axes, but minimal Left-Right. For each age-group, the MPM brain atlases provide reasonably good representation of the native-space volumes in the IBT space, except in a few regions with high inter-subject variability. These findings provide evidence to support the use of age and population-specific templates in human brain mapping studies. This dataset is made publicly available (https://hollabharath.github.io/IndiaBrainTemplates). Highlights A new set of age-specific T1w Indian brain templates for ages 6-60 yr are developed and validated. A new AFNI tool, make_template_dask.py, for the creation of group-based templates. Maximum probability map atlases are also provided for each template. Results indicate the appropriateness of Indian templates for spatial normalization of Indian brains

multisite studies, the acquisition parameters varied slightly across sites and scanners, but were 132 generally similar, with good grey/white matter contrast with a voxel size close to 1mm isotropic; 133 details are listed in Table 2.

135
This processing primarily used programs in the AFNI (v19.0.20) [Cox, 1996] and FreeSurfer (v6.0) 136 [Fischl, 2012] neuroimaging toolboxes, as well as the "dask" scheduling tool in Python developed by 137 the Dask Development Team [2016]. Unless otherwise noted, programs named here are contained 138 within the AFNI distribution. The following processing steps are shown schematically in Figure 1, 139 in the first column.

140
Datasets were first processed using AFNI's "fat_proc_convert_dcm_anat". Using this, DICOMs were converted to NIFTI files using dcm2niix_afni (the AFNI-distributed version of dcm2niix [Li 142 et al., 2016]). For uniformity and initialization, with this tool, they were also given the same 143 orientation (RAI), and the physical coordinate origin was placed at the volume's center of mass (to 144 simplify later alignments). 145 Next, "fat_proc_axialize_anat" was applied to reduce the variance in the spatial orientation of 146 brains for later alignment and for practical considerations of further processing steps, as described 147 here. Each volume was affinely registered to a reference anatomical template (MNI ICBM 152 148 T1w) that had previously been AC-PC aligned; alignment included an additional weight mask 149 to emphasize subcortical structure alignment (e.g., AC-PC structures), and only the solid-body 150 parameters of the alignment were applied, so that no changes in shape were incurred. Because 151 datasets had been acquired with varied spatial resolution and FOV (see Table 2), the datasets were 152 resampled (using a high-order sinc function, to minimize smoothing) to the grid of the reference 153 base of 1mm isotropic voxels.

154
All datasets were visually and systematically checked for quality of both data and registration using 155 the QC image montages that were automatically generated by the previous program. T1w volumes 156 with noticeable ringing or other artifact (e.g., due to subject motion or dicom reconstruction errors) 157 were noted and removed from further analyses. T1w volumes with any incidental findings (for 158 example, large ventricles, cavum septum pellucidum) were also removed.

159
FreeSurfer's "recon-all" [Fischl, 2012] was run on each T1w data set to estimate surfaces, parcel-160 lation and segmentation maps. AFNI's "@SUMA_Make_Spec_FS" was then run to convert the 161 FreeSurfer output to NIFTI files and to generate standard meshes of the surface in formats usable 162 by AFNI and SUMA. Additionally, @SUMA_Make_Spec_FS subdivides the FreeSurfer parcella-163 tions into tissue types such as gray matter (GM), white matter (WM), cerebrospinal fluid (CSF), 164 ventricle, etc. This was followed by visual inspection of parcellation maps overlaid on anatomical 165 volumes.

166
Next, a whole brain mask of each anatomical volume was created. In several cases, the skullstripped 167 brain volumes output by recon-all (brain_mask.nii) included large amounts of non-brain material 168 (skull, dura, face, etc.), and so an alternative mask was generated using only the ROIs comprising 169 the parcellation and segmentation maps. For each subject, a whole brain mask was generated by: 170 first making a preliminary mask from all of the ROIs identified by recon-all; then inflating that pre-171 mask by 3 voxels; and finally shrinking the result by two voxels (thus filling in any holes inside the 172 brain mask and smoothing the outer edges). This produced whole brain masks that were uniformly 173 specific to each subject's intracranial volume.

174
Finally, AFNI's 3dUnifize was run on each T1w volume in order to reduce the intensity inhomogeneity (e.g., due to the bias field) and to normalize the intensity of tissues within the volume. This 176 ensures that each subject's brain, which had been acquired on different scanners with potentially 177 different scalings, would have equal weight when averaging (e.g., WM is scaled to approximately a 178 value of 1000 in each brain, and similarly for other tissues), and also reduces the risk of a bright 179 outlier region driving poor alignment. After the above pre-processing steps and QC, the following templatizing algorithm was applied 182 for each cohort (C1-5) separately. The general procedure was to alternate between alignment to 183 a reference base (with increasingly higher order of refinement) and averaging the aligned brains 184 to generate a new reference base for the subsequent iteration. In this way one can generate a 185 cohort mean template of successively greater specificity and detail; after several iterations, the 186 alignment essentially converges (i.e., additional refinement becomes negligible) and is halted. Warps 187 were generated and saved at each step. The final nonlinear warps and affine transformations were 188 concatenated for each subject at the end in order to generate the final group average template.

189
These steps are also included in the schematic Figure 1, in the first column (bottom) and second 190 and third columns.

191
The first level of alignment was made from each anatomical in the cohort to the MNI ICBM-152

192
T1w template using a 6 degree of freedom (DF) rigid body equivalent registration, meaning a full 193 affine transformation was computed, but only the rigid components were extracted and applied.

194
The average of all subjects' brains, rigidly aligned to the initial template, was used to create a 195 single average volume "mean-rigid"; here and at each alignment stage, a cohort standard deviation 196 map was also created, to highlight locations of relatively high and low variability. That stage's 197 average volume was then used as a base for the next stage of alignment for each subject, using a 12 198 DF linear affine registration, and with the results averaged to create the next base "mean-affine".

199
For these alignments, AFNI's "lpa" cost function (absolute value of local Pearson correlation) [Saad 200 et al., 2009] was used for high quality alignment of features between volumes of similar contrast.

201
The cost function computes the absolute value of the Pearson correlation between the volume and 202 the current template in patches of the volume at a time.

203
As a practical consideration, we note that lower level alignments such as these have a general 204 property of producing a smoothed brain, which has the additional effect of increasing the apparent 205 size of the base dataset (i.e., the edge is blurred outward). Therefore, in these initial levels we 206 added a step to control the overall volume of the template. We calculated the mean intracranial 207 volume (ICV) of all the subjects in the cohort V coh , and then calculated the volume of the initial 208 mean-affine brain mask V a↵ . The volume ratio r vol = V coh /V a↵ was calculated, and each of the 209 three dimensions of the mean-affine volume were scaled down by the appropriate length scaling

228
Each mean-NL* volume was resized in the same manner as the initial stages, although the correction 229 factors were much smaller here. Additionally, each mean-NL* volume was anisotropically smoothed 230 (preserving edges within the volume, for detail) using 3danisosmooth, in order to sharpen its contrast 231 for subsequent alignments.

232
The mean-NL4 volume became the final group mean template for each cohort, as in all cases results 233 appeared to have essentially converged after this number of step. The coordinate system of this 234 mean volume defines the template space for that age group, and is labelled "IBT_C1", "IBT_C2", 235 etc.
236 Figure 1 -Schematic representation of the steps involved in the Dask pipeline (make_template_dask.py) for generating population-average brain templates.

237
We used the following approach to find the maximally representative individual brain for the mean 238 template from the underlying cohort, in order to generate an additional "typical" template for that 239 space, in complement to the mean template.

240
To find the most typical subject for the mean template quantitatively, the lpa cost function value 241 from aligning each subject's anatomical to the final mean-NL4 was compared across the group; that 242 is, the degree of similarity of each subject's aligned volume to the mean template base was compared 243 across the cohort. The individual brain in that mean template space with the lowest cost function 244 value was selected to be the "typical template" brain. Alignment results were also visually verified 245 for each typical template. We note that the typical template volume uses the same coordinate 246 system as the mean template, and thus no additional "coordinate space" is created in this process.

254
For the mean template, maximum probability map (MPM) atlases were reconstructed as follows.
The FreeSurfer parcellations for each subject were transformed to the IBT space using the warps created during the template creation process (and "nearest neighbor" interpolation, to preserve 257 ROI identity). For a given parcellation, the fraction of overlap of a given ROI at each voxel in the 258 template was computed. That overlap fraction is essentially the probability of a region to be mapped The fractional volumes of each ROI in the MPM atlases were checked for being representative of 274 each cohort. For this we calculated the logarithm of the relative volume ratio of each ROI: where the numerator is the fractional volume of a given ith ROI in the MPM (i.e., volume of the to the cohort mean, and negative or positive values reflect a relative compression or expansion, 280 respectively, of the MPM ROI relative to that for a particular cohort.

281
In order to quantify the inter-subject brain morphological variability for participants in each age-282 band, we calculated a region-wise mean deformation value (mDV) from the deformation warp fields 283 generated during non-linear registration to the age-specific IBT. For this, the absolute warp value 284 was summed across all three axes (L1-norm) and averaged across all the voxels within each ROI 285 in the age-specific MPM atlas. A larger mDV indicates greater inter-subject brain morphological 286 variability.

287
To examine the utility of the IBTs on a real, representative dataset, a separate sample of Indian were conducted using the T1w and resting functional data.

294
We first used the deformation field to characterize the difference between the two templates (IBT vs 295 MNI The first part of the output consists of both "population average" and "typical" Indian brain tem-316 plates for five specific age-ranges: late-childhood (C1), adolescence (C2), young adulthood (C3), 317 adulthood (C4) and late adulthood (C5) [see Table 1  contrast-to-noise ratio (CNR) between GM and WM improved through the successive stages in all 325 the template age-groups (see Supplementary Figure S1).      We have introduced five new India brain template (IBTs) spaces, spanning an age range from 6-60 363 years. Additionally, corresponding atlases (IBTAs) from widely used segmentations were also created 364 for each space. These should form useful reference templates and region maps for brain imaging 365 studies involving predominantly Indian populations. Both the creation of age-specific templates and 366 the inclusion of associated atlases make the present study distinct from previous Indian population 367 brain template projects [Rao et al., 2017, Bhalerao et al., 2018, Sivaswamy et al., 2019 2020]; additionally, we have generated both "population mean" and high-contrast "typical" templates 369 for each age band. The IBT volumes and corresponding atlases are publicly available for download, in standard NIFTI format, and freely usable by the wider neuroimaging community.

371
The need for age-specific templates in particular has been recognized across different populations 372 [Fonov et al., 2011, Wilke et al., 2002, Yoon et al., 2009; however, Indian versions of age-specific 373 brain templates have not been available to date. While adult brain templates may still provide 374 reasonably accurate anatomical priors for normalizing lower resolution smoothed functional data, 375 they may not be appropriate for high resolution structural and functional data [Wilke et al., 2002]. ization and anatomical localization. The relative differences in warping along the major axes of 393 the brain were shown here using validation groups from the local population. The differences in 394 warping magnitudes varied both by axis and by the age of subjects. Thus, the structural differences 395 in templates are not trivial, i.e., just scaling, but instead reflect shape variations that are likely to 396 significantly affect the overall goodness-of-fit and anatomical alignment across a group study.

397
Such aspects were highlighted in the differences of outcomes in fMRI processing when using IBT vs 398 MNI templates: the IBT-based output tended to have higher ReHo values among ROI pairs. The 399 latter fact in particular suggests that the IBTs provided better function-to-anatomical alignment 400 across groups, so that voxel with functionally similar time series tended to be grouped together more 401 preferentially. One might expect this to be a relatively small effect, because alignment to the MNI 402 templates still appears generally reasonable; one would expect the overlap pattern differences to be 403 occurring fractionally within ROIs and predominantly at boundaries. Indeed, the FC differences 404 were relatively small, but with a noticeable trend toward higher values in the IBT-based datasets.
It is important to emphasize that these structural differences are only with regards to morphology; 406 they do not relate to functional or behavioral outcomes, nor to intelligence, etc. The purpose and 407 goal of population-specific templates is for the practical consideration of maximizing the matching 408 of structures across a group during an alignment step of processing, as well as to better match 409 functional regions to structures. These are geometric and signal-to-noise considerations, which are 410 important in brain studies (as demonstrated here), but which are unrelated to the brain behavior 411 itself.

412
The wide variety of brain structural patterns in any group, even in an apparently homogeneous 413 one, is also worth commenting on. This inherent variability affects both the creation and utilization 414 of brain templates [Yang et al., 2020]. In any population brain structures can vary to the degree 415 of having different numbers of sulci in the same region (e.g., [Thompson et al., 1996]  probability map atlases showed a high degree of agreement across the group through most of the 426 brain.

427
The variability present in the template generation is also observable in the atlases. The inter-428 subject variability (as measured by the mean deformation values for various regions during non-linear 429 registration to age and population-specific template) also correlated positively with the expansion 430 of MPM volumes, in all age groups (see Supplementary Figure S7). While the final MPM atlases 431 indicate the most frequent positions of each brain region in a given cohort, we also provide the 432 probability density maps for each ROI in the atlas (see supplementary Figure S8 for example),

433
which can be of additional use in ROI-based analyses.

434
While spatial normalization to IBT offers distinct advantages in terms of spatial accuracy and In conclusion, the present work demonstrates the appropriateness of using age and population-456 specific templates as reference targets for spatial normalization of structural and functional neu-457 roimaging data. This database of age-specific IBTs and IBTAs is made freely available to the wider 458 neuroimaging community of researchers and clinicians worldwide. We hope that these tools will fa-459 cilitate research into neurological understand in general and into the functional and morphometric 460 changes that occur over life-course in Indian population in particular.

461
Data Availability Statement

462
The Indian brain templates (IBTs) and atlases (IBTAs) Figure S3: The five population-average IBTs (C1-5) with three sets of sagittal, coronal and axial view displayed as underlay in grayscale and the respective typical subject for each IBT version as the overlay. Arrow points to example regions in C1 age-band regions where the typical version provides greater details than the underlying population-average version.
Age-specific Indian brain templates Figure    The color intensity reflects probability density estimates (ranging from 0 to 1)