Abstract
Magnetic resonance image (MRI) processing pipelines use average templates to enable standardization of individual MRIs in a common space. MNI-ICBM152 is currently used as the standard template by most MRI processing tools. However, MNI-ICBM152 represents an average of 152 healthy young adult brains and is vastly different from brains of patients with neurodegenerative diseases. In those populations, extensive atrophy might cause inevitable registration errors when using an average template of young healthy individuals for standardization. Disease-specific templates that represent the anatomical characteristics of the populations can reduce such errors and improve downstream driven estimates. We present multi-sequence average templates for Alzheimer’s Dementia (AD), Fronto-temporal Dementia (FTD), Lewy Body Dementia (LBD), Mild Cognitive Impairment (MCI), cognitively intact and impaired Parkinson’s Disease patients (PD-CIE and PD-CI, respectively), individuals with Subjective Cognitive Impairment (SCI), AD with vascular contribution (V-AD), Vascular Mild Cognitive Impairment (V-MCI), Cognitively Intact Elderly (CIE) individuals, and a human phantom. We also provide separate templates for males and females to allow better representation of the diseases in each sex group.
Measurement(s) | Human Brain |
Technology Type(s) | Magnetic resonance imaging |
Sample Characteristic - Organism | Homo Sapiens |
Sample Characteristic - Location | Canada |
Similar content being viewed by others
Background & Summary
Magnetic resonance imaging (MRI) brain templates (i.e. averages of multi-individual images, co-registered in a similar reference space) are widely used in image processing, for example as targets in registration and intensity normalization, as a common standard space enabling individual and population based comparisons in deformation/tensor or voxel based morphometry, and as the basis for segmentation techniques that rely on nonlinear registration1,2,3,4. An example is the MNI-ICBM152, an average based on images from 152 healthy young adults, and one of the most popular templates in current use given its distribution in processing pipelines such as MINC, FSL, and SPM1,2,3 that have been shared more than 45,000 times worldwide (Data from NITRC.org).
A common feature of existing averages such as the MNI-ICBM152 is their reliance on healthy, young brains, in addition to aggregating both sexes in the template generation process. However, in aging and populations with neurodegenerative diseases, ventricle enlargement, extensive levels of cortical and subcortical atrophy, as well as white matter hyperintensities (WMHs) create large degrees of difference between an individual’s MRI and such templates. We have shown in prior work that such differences significantly increase registration errors in some of these well-known image processing tools (e.g. ANTs, Elastix, FSL, MINC, and SPM)5. Ridwan et al. have shown that use of age-appropriate templates allows for higher inter-subject spatial normalization accuracy for older adult data, facilitating detection of smaller inter-group morphometric differences6. A similar reasoning applies to studies of neurodegeneration. Using a dataset consisting of patients with different frontotemporal dementia variants, we have shown that use of age and disease appropriate templates can significantly reduce nonlinear registration errors7. Van Hecke et al. have also shown that improvement in image alignments due to use of population-specific atlases leads to higher sensitivity and specificity in detecting white matter abnormalities in diffusion tensor imaging (DTI) voxel-based analyses8. Therefore, age and disease appropriate templates are necessary to reflect the anatomical characteristics of the populations of interest and increase downstream accuracy and sensitivity of the analyses by reducing potential image processing errors and biases that can occur when using age and pathology inappropriate templates7. An example use case would be the monitoring of a therapy in a specific pathology, with an effect that may be clinically significant but resulting in small image differences. The increased sensitivity brought about by using an appropriate, age-, sex- and disease template would therefore be significant.
Previous work on average brain templates has been mostly based on pediatric, young adult, or healthy aged brains6,9,10,11,12,13. Xiao et al. have developed a multi-contrast template of 15 Parkinson’s disease patients14. We have previously developed average T1w templates of frontotemporal dementia variants (i.e. behavioural, semantic, and progressive non-fluent aphasia) along with age matched healthy templates, showing that use of age and disease appropriate templates improve nonlinear registration performance7. Guo et al. have recently developed a T1w brain template based on a combination of healthy aged adults, individuals with mild cognitive impairment, and Alzheimer’s disease patients, showing that use of disease-specific templates improves sensitivity in voxel-based gray matter volume analyses, enabling for early detection and earlier therapeutic opportunities15. To our knowledge, no prior work has provided multi-sequence average templates of various neurodegenerative disease populations generated consistently using harmonized image acquisition protocols.
Based on data from the Canadian Consortium for Neurodegeneration and Aging (CCNA)16, a flagship study of the Canadian Institutes of Health Research, we present average templates for T1-weighted (T1w), T2-weighted (T2w), T2*-weighted, Proton Density (PD), and FLuid Attenuated Inversion Recovery (FLAIR) sequences in eleven diagnostic groups, including Alzheimer’s Dementia (AD), Fronto-temporal Dementia (FTD), Lewy Body Dementia (LBD), Mild Cognitive Impairment (MCI), cognitively intact and impaired Parkinson’s Disease patients (PD-CIE and PD-CI, respectively), individuals with Subjective Cognitive Impairment (SCI), Vascular Alzheimer’s Dementia (V-AD), Vascular Mild Cognitive Impairment (V-MCI), as well as Cognitively Intact Elderly (CIE) individuals and one human phantom17. These templates can capture the anatomical characteristics for each disease cohort at the regional level. With multiple contrasts available providing different types of information, the various templates can be used to assess different aspects in each disease: i) T1w templates are useful for assessing fine anatomical details and estimating regional and global atrophy levels; ii) T2w/PD sequences are useful for skull segmentation, and assessment of deep gray matter structures, iii) FLAIR images can be used to detect WMHs and infarcts; and iv) T2* images can be used to identify microbleeds as well as hemorrhages.
There are significant sex and gender related differences in the prevalence, clinical outcomes, and response to treatments for these distinct neurodegenerative diseases (e.g. higher prevalence of Alzheimer’s disease in females and higher prevalence of Parkinson’s disease in males)18,19,20. Sex-specific average templates would therefore be useful tools to represent and assess potential anatomical differences in patterns of atrophy in males and females. Thus, in addition to the disease-specific average templates combining male and female participants, we provide separate templates for males and females in each diagnostic category.
Methods
Data
We used data from the Comprehensive Assessment of Neurodegeneration and Dementia (COMPASS-ND) cohort of the CCNA, a national initiative to catalyze research on dementia16. COMPASS-ND includes deeply phenotyped subjects with various forms of dementia and mild memory loss or concerns, along with cognitively intact elderly subjects. Ethical agreements were obtained at all respective sites. Written informed consent was obtained from all participants.
Clinical diagnoses were determined by participating clinicians based on longitudinal clinical, screening, and MRI findings (i.e. diagnosis reappraisal was performed using information from recruitment assessment, screening visit, clinical visit with physician input, and MRI). The diagnostic groups included, AD, CIE, FTD, LBD, MCI, PD-CIE, PD-MCI, PD-Dementia (for this study, PD-MCI and PD-Dementia groups were merged into one PD-CI group), SCI, V-AD, and V-MCI. Diagnosis was performed according to the current guidelines in the field and diagnostic criteria was harmonized across all CCNA sites. However, we acknowledge that due to the inherent heterogeneity and variabilities in such neurodegenerative disease populations, there might be inevitable variabilities across different centers and studies. For details on clinical group ascertainment, see Pieruccini‐Faria et al.21 as well as Dadar et al.22 (section 1 in the supplementary materials). A single cognitively healthy volunteer was also scanned as a human phantom multiple times across different centers for quality assurance purposes (more information on the SIMON human phantom dataset can be found in Duchesne et al.17).
Table 1 summarizes the demographic characteristics of the participants used to generate each template. Note that due to the lower prevalence and challenges in recruitment of participants in certain disease categories (e.g. FTD and LBD), the resulting templates might not be reflective of the entire spectrum of presentation of the pathology. Further work including larger populations is therefore warranted.
All participants were scanned using the Canadian Dementia Imaging Protocol, a harmonized MRI protocol designed to reduce inter-scanner variability in multi-centric studies and which included the following sequences23:
-
3D isotropic T1w scans (voxel size = 1.0 × 1.0 × 1.0 mm3) with an acceleration factor of 2 (Siemens: MP‐RAGE‐PAT: 2; GE: IR‐FSPGR‐ASSET 1.5; Philips: TFE‐Sense: 2)
-
Interleaved proton density/T2‐weighted (PD/T2w) images (voxel size = 0.9 × 0.9 × 3 mm3), fat saturation, and an acceleration factor of 2.
-
Fluid attenuated inversion recovery (T2w‐FLAIR) images (voxel size = 0.9 × 0.9 × 3 mm3), fat saturation, and an acceleration factor of 2.
-
T2* gradient echo images (voxel size = 0.9 × 0.9 × 3 mm3) and acceleration factor of 2.
Table 2 shows the acquisition parameters for each sequence and scanner manufacturer. A detailed description, exam cards, and operators’ manual are publicly available at: www.cdip-pcid.ca.
Preprocessing
All images were pre-processed with image denoising24, intensity non-uniformity correction25, and image intensity normalization into a 0–100 range. The pre-processed images were then linearly5 registered to the pseudo-Talairach space defined by the MNI-ICBM152-2009c template using a 9-parameter registration (three translation, three rotation, and three scaling parameters)26. T2w, PD, FLAIR, and T2* images were also co-registered (rigid registration, 6 parameters) to the T1w images with a mutual information cost function.
Template generation
The method by Fonov et al. was used to generate unbiased templates for each diagnostic group for all participants, as well as each group but separately for males and females12,27 (all except the LBD group in which there were only two female participants). This method has previously been used to generate templates in various studies, including the latest higher resolution version of the MNI-ICBM2009c template (http://nist.mni.mcgill.ca/?p=904)26,28. In short, the pipeline implements a hierarchical nonlinear registration procedure using Automatic Nonlinear Image Matching and Anatomical Labelling (ANIMAL)29, iteratively refining the previous registrations by reducing the step size (20 iterations in total, four iterations at each of the levels of 32, 16, 8, 4, and 2 mm, respectively) until convergence is reached. This process of increasingly refined iterative nonlinear registrations leads to average brains that reflect the anatomical characteristics of the population of interest with higher levels of anatomical detail27. The higher resolution T1w images (isotropic 1mm3) were used to obtain the nonlinear transformations for creating the average templates. T2w, PD, FLAIR, and T2* templates were then created by combining their rigid to-T1w co-registration transformations with the nonlinear transformations based on the T1w images. All final templates were generated at 1mm3 isotropic resolution.
FreeSurfer segmentation
To appreciate differences between templates, we processed all T1w averages using FreeSurfer version 6.0.0 (recon-all -all). FreeSurfer provides a full processing stream for structural T1w data (https://surfer.nmr.mgh.harvard.edu/)30. The final segmentation output (aseg.mgz) was then used to obtain volumetric information for each template based on the FreeSurfer look up table available at https://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/AnatomicalROI/FreeSurferColorLUT.
Data Records
For information on COMPASS-ND dataset and to request access, see https://ccna-ccnv.ca/compass-nd-study/. The average template files for all groups and sequences are available in both compressed MINC31,32 and NIfTI formats at G-Node (https://gin.g-node.org/mahsadadar/CDIP_Templates)33 as well as Zenodo 34.
Technical Validation
Quality control
The quality of the registrations, pre-processed images, as well as the volumetric segmentations performed by FreeSurfer was visually assessed by an experience rater (MD). All images passed this quality control step. Note that the provided data was already quality controlled by the CCNA imaging platform for presence of imaging artifacts, and only scans that had passed this quality control step were acquired and used for this study. In terms of qualitative comparison with other atlases in the field6,9,10,14,27, based on visual assessment, the provided atlases have high levels of image sharpness and anatomical detail, clearly delineating the sulci and gyri in the cortex (Fig. 1).
Templates
Figures 1–5 show axial slices of the T1w, T2w, T2star, PD, and FLAIR average templates for all 11 diagnostic groups, covering the brain at different levels. For more detailed figures of each template, see the supplementary materials (Figures S1–S11). As expected, CIE, PD-CIE, and MCI groups had smaller ventricles, with lower levels of atrophy compared with the cognitively impaired and dementia groups (Fig. 1). FLAIR images of the vascular cohorts (i.e. Mixed, V-MCI, and V-AD) showed extensive levels of periventricular hyperintensities compared to other groups (Fig. 2), due to the presence of WMHs in the majority of the patients in these populations. This pattern was also visible to a lesser extent as hypointensity in the T1w templates, as well as hyperintensity in the T2w, PD, and T2* templates (Figs. 3 to 5). Presence of WMHs is another factor that necessitates use of age and disease appropriate templates, since they can directly impact intensity normalization results. In fact, we have previously shown that presence of WMHs significantly reduces linear registration accuracy in currently used image processing pipelines such as MINC, FSL, Elastix, SPM, and ANTs when images with high WMH burden are registered to young and healthy adult templates such as MNI-ICBM1525. Similarly, we showed that increased ventricular volume due to aging and presence of atrophy (e.g. in AD populations) reduces registration accuracy when using healthy young adult templates as the registration target5.
Figure 6 shows axial slices of the male and female templates for all diagnostic groups and sequences. Overall, male templates have larger ventricles and greater levels of atrophy than female templates. For more detailed figures of each template, see the supplementary materials (Figures S12–S31).
Figure 7 shows axial slices of the templates for the human phantom (SIMON).
Volumetric comparisons
Tables 3–5 summarize the grey and white matter (GM, WM) and cerebrospinal fluid (CSF) volumetric information for the templates as segmented by FreeSurfer. Figure 8 compares GM volumes (log transformed) of each template against the CIE template. Data points below the reference line (shown in red) indicate lower values for the template in comparison with the CIE template. As expected, cognitively impaired and dementia templates had lower GM values than the CIE template, whereas both cognitively intact PD-CIE and SCI templates had similar volumes to the CIE template (i.e. data points fall on the reference line).
Figure 9 compares GM volumes (log transformed) of male versus female templates. Note that since all templates have been linearly registered to the MNI-ICBM2009c template prior to the template creation step, all volumetric values reflect variabilities after accounting for intracranial volume differences and are not caused by potential head size differences between males and females. Data points below the reference line (shown in red) indicate lower values for the male template in comparison with the female template. In the AD and mixed templates, the nucleus accumbens areas bilaterally had lower volumes in the male templates. In the PD-CI template, most regions had slightly lower GM volumes in the male template.
As expected, mixed dementia, vascular MCI, and vascular AD templates had higher WM hypointensity volumes (corresponding to the WMHs on FLAIR and T2w sequences) on T1w templates (Table 4). Male templates for AD, FTD, PD-CI, V-MCI, and V-AD also had greater WM hypointensity volumes than the female templates (Table 4). The mixed template had the largest ventricles (Table 5), followed by V-AD and AD templates. As expected, CIE template had the smallest ventricles, followed by PD-CIE, and SCI. In all diagnostic groups, lateral ventricles were larger for the male templates in comparison with the female templates. This difference was most prominent in the V-AD group, for which the left and right lateral ventricles were 33% and 43% larger respectively for the male template (Table 5). Regarding asymmetry, in the FTD, V-MCI, and mixed templates, the left lateral ventricle was 12%, 13%, and 17% larger than the right lateral ventricle. This difference was more prominent in the male templates for FTD and V-MCI groups, whereas for the mixed group, the female template had greater asymmetry in the ventricles. All of these differences highlight the need for group-specific templates in multi-individual, multi-centric studies.
Using Disease Appropriate Templates to Improve Registration
Use of age and disease appropriate templates can reduce both linear and nonlinear registration errors. We have previously shown that older subjects, those with larger ventricles, and high levels of WMHs have higher levels of linear registration failure rates when using young adult brain templates as the registration target for most widely used registration tools such as FSL, SPM, ANTs, Elastix, and MINC5. Using disease appropriate templates could be the solution to improve both linear and nonlinear registration for aged and diseased populations. Note that since all templates are in the same space (i.e. share a similar alignment to a pseudo-Talairach coordinate system), linear registration to one would be equivalent to linear registration to other templates without additional manipulation. As for nonlinear registration, these templates can be used as intermediate registration targets even in cases where the intended final application is to register all subjects to one healthy or younger average brain. Intermediate templates have been previously used for various registration tasks, particularly when there exists a large difference between source and target templates35,36,37,38. Disease appropriate average templates can be used as intermediate registration targets to improve nonlinear registration, using the following steps:
-
1.
Linearly register patient brain image(s) to the disease appropriate template.
-
2.
Nonlinearly register patient brain image(s) to the disease appropriate template.
-
3.
Concatenate the nonlinear transformation with the precomputed nonlinear transformation between the two average templates.
-
4.
If necessary, the registration can be refined by performing another nonlinear registration between the nonlinearly transformed image and the average template. Concatenate this additional transformation with the previous two.
Figure 10 demonstrates how using a disease appropriate average template can improve nonlinear registration. Panel a shows a nonlinear registration scenario in which the brain of an individual with FTD has been nonlinearly registered directly to the MNI-ICBM152 average template using ANTs diffeomorphic registration tool39. The red contours consistently show the outline of the MNI-ICBM152 brain and can be used to assess the quality of the nonlinear registration. In a perfectly registered image, the contours of MNI-ICBM152 should match the contours of the nonlinearly deformed image (shown in the last columns on the right). The orange arrow shows the areas of gross registration failure, where ANTs has not been able to accurately register the ventricles of the subject to MNI-ICBM152. This is a common occurrence in dementia patients with large ventricles and gross atrophy. Panel b shows registration results for the same individual, which was first nonlinearly registered to a disease appropriate FTD template, and then nonlinearly registered to the MNI-ICBM152. Comparing the two deformed images (last columns on the right), we can see that when the FTD template was used as an intermediate registration target, ANTs was able to accurately register the ventricles.
Code availability
The scripts for generating unbiased average templates are publicly available at https://github.com/vfonov/nist_mni_pipelines.
References
Ashburner, J. et al. SPM12 manual. Wellcome Trust Cent. Neuroimaging Lond. UK (2014).
Aubert-Broche, B. et al. A new method for structural volume analysis of longitudinal brain MRI data and its application in studying the growth trajectories of anatomical brain structures in childhood. NeuroImage 82, 393–402 (2013).
Jenkinson, M., Beckmann, C. F., Behrens, T. E., Woolrich, M. W. & Smith, S. M. Fsl. Neuroimage 62, 782–790 (2012).
Mateos-Pérez, J. M. et al. Structural neuroimaging as clinical predictor: A review of machine learning applications. NeuroImage Clin. https://doi.org/10.1016/j.nicl.2018.08.019 (2018).
Dadar, M., Fonov, V. S., Collins, D. L. & Initiative, A. D. N. A comparison of publicly available linear MRI stereotaxic registration techniques. NeuroImage 174, 191–200 (2018).
Ridwan, A. R. et al. Development and evaluation of a high performance T1-weighted brain template for use in studies on older adults. Hum. Brain Mapp. 42, 1758–1776 (2021).
Dadar, M., Manera, A. L., Fonov, V. S., Ducharme, S. & Collins, D. L. MNI-FTD templates, unbiased average templates of frontotemporal dementia variants. Sci. Data 8, 1–10 (2021).
Van Hecke, W. et al. The effect of template selection on diffusion tensor voxel-based analysis results. NeuroImage 55, 566–573 (2011).
Avants, B., & Tustison, N. ANTs/ANTsR Brain Templates. figshare https://doi.org/10.6084/m9.figshare.915436.v2 (2018).
Klein, A. Mindboggle-101 templates (unlabeled images from a population of brains). Harvard Dataverse. https://doi.org/10.7910/DVN/WDIYB5 (2017).
Kötter, R. et al. A probabilistic atlas and reference system for the human brain: International Consortium for Brain Mapping (ICBM). Philos. Trans. R. Soc. Lond. B. Biol. Sci. 356, 1293–1322 (2001).
Fonov, V. et al. Unbiased average age-appropriate atlases for pediatric studies. NeuroImage 54, 313–327 (2011).
Yoon, U., Fonov, V. S., Perusse, D. & Evans, A. C. The effect of template choice on morphometric analysis of pediatric brain data. NeuroImage 45, 769–777 (2009).
Xiao, Y. et al. Multi-contrast unbiased MRI atlas of a Parkinson’s disease population. Int. J. Comput. Assist. Radiol. Surg. 10, 329–341 (2015).
Guo, X.-Y. et al. Development and evaluation of a T1 standard brain template for Alzheimer disease. Quant. Imaging Med. Surg. 11, 2224–2244 (2021).
Chertkow, H. et al. The Comprehensive Assessment of Neurodegeneration and Dementia: Canadian Cohort Study. Can. J. Neurol. Sci. 46, 499–511 (2019).
Duchesne, S. et al. Structural and functional multi-platform MRI series of a single human volunteer over more than fifteen years. Sci. Data 6, 1–9 (2019).
Altmann, A., Tian, L., Henderson, V. W. & Greicius, M. D. & Investigators, A. D. N. I. Sex modifies the APOE-related risk of developing Alzheimer disease. Ann. Neurol. 75, 563–573 (2014).
Tierney, M. C., Curtis, A. F., Chertkow, H. & Rylett, R. J. Integrating sex and gender into neurodegeneration research: A six-component strategy. Alzheimers Dement. Transl. Res. Clin. Interv. 3, 660–667 (2017).
Bellou, V., Belbasis, L., Tzoulaki, I., Evangelou, E. & Ioannidis, J. P. Environmental risk factors and Parkinson’s disease: an umbrella review of meta-analyses. Parkinsonism Relat. Disord. 23, 1–9 (2016).
Pieruccini‐Faria, F. et al. Gait variability across neurodegenerative and cognitive disorders: Results from the Canadian Consortium of Neurodegeneration in Aging (CCNA) and the Gait and Brain Study. Alzheimers Dement. n/a (2021).
Dadar, M. et al. White Matter Hyperintensity Distribution Differences in Aging and Neurodegenerative Disease Cohorts. 2021.11.23.469690 https://doi.org/10.1101/2021.11.23.469690 (2021).
Duchesne, S. et al. The Canadian Dementia Imaging Protocol: Harmonizing National Cohorts. J. Magn. Reson. Imaging 49, 456–465 (2019).
Coupe, P. et al. An Optimized Blockwise Nonlocal Means Denoising Filter for 3-D Magnetic Resonance Images. IEEE Trans. Med. Imaging 27, 425–441 (2008).
Sled, J. G., Zijdenbos, A. P. & Evans, A. C. A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Trans. Med. Imaging 17, 87–97 (1998).
Manera, A. L., Dadar, M., Fonov, V. & Collins, D. L. CerebrA, registration and manual label correction of Mindboggle-101 atlas for MNI-ICBM152 template. Sci. Data 7, 1–9 (2020).
Fonov, V., Evans, A., McKinstry, R., Almli, C. & Collins, D. Unbiased nonlinear average age-appropriate brain templates from birth to adulthood. NeuroImage 47, S102 (2009).
Dadar, M., Manera, A. L., Fonov, V. S., Ducharme, S. & Collins, D. L. MNI-FTD Templates: Unbiased Average Templates of Frontotemporal Dementia Variants. Sci. Data 8, 1–10 (2021).
Collins, D. L. & Evans, A. C. Animal: validation and applications of nonlinear registration-based segmentation. Int. J. Pattern Recognit. Artif. Intell. 11, 1271–1294 (1997).
Fischl, B. FreeSurfer. Neuroimage 62, 774–781 (2012).
Neelin, P., MacDonald, D., Collins, D. L. & Evans, A. C. The MINC file format: from bytes to brains. NeuroImage 7, S786 (1998).
Vincent, R. D. et al. MINC 2.0: a flexible format for multi-modal images. Front. Neuroinformatics 10, 35 (2016).
Dadar, M., Camicioli, R. & Duchesne, S. Multi-Sequence Average Templates for Aging and Neurodegenerative Disease Populations. https://doi.org/10.12751/g-node.yoy0z6 (2021).
Dadar, M., Camicioli, R. & Duchesne, S. Multi-Sequence Average Templates for Aging and Neurodegenerative Disease Populations. https://doi.org/10.5281/zenodo.5018356 (2021).
Christensen, G. E. & He, J. Consistent nonlinear elastic image registration. in Proceedings IEEE Workshop on Mathematical Methods in Biomedical Image Analysis (MMBIA 2001) 37–43, https://doi.org/10.1109/MMBIA.2001.991697 (2001).
Jia, H., Yap, P.-T., Wu, G., Wang, Q. & Shen, D. Intermediate templates guided groupwise registration of diffusion tensor images. NeuroImage 54, 928–939 (2011).
Xiao, Y. et al. An accurate registration of the BigBrain dataset with the MNI PD25 and ICBM152 atlases. Sci. Data 6, 210 (2019).
Jia, H. et al. Directed graph based image registration. Comput. Med. Imaging Graph. 36, 139–151 (2012).
Avants, B. B., Tustison, N. & Song, G. Advanced normalization tools (ANTS). Insight J 2, 1–35 (2009).
Acknowledgements
Data used in this article were obtained from the Comprehensive Assessment of Neurodegeneration and Dementia (COMPASS-ND) cohort of the Canadian Consortium for Neurodegeneration and Aging (CCNA). MD is supported by a scholarship from the Canadian Consortium on Neurodegeneration in Aging (CCNA) as well as by an Alzheimer Society Research Program (ASRP) postdoctoral award. CCNA is supported by a grant from the Canadian Institutes of Health Research with funding from several partners.
Author information
Authors and Affiliations
Contributions
Mahsa Dadar: Study concept and design, analysis of the data, drafting and revision of the manuscript. Richard Camicioli: Study concept and design, interpretation of the data, revising the manuscript. Simon Duchesne: Study concept and design, interpretation of the data, revising the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Dadar, M., Camicioli, R. & Duchesne, S. Multi sequence average templates for aging and neurodegenerative disease populations. Sci Data 9, 238 (2022). https://doi.org/10.1038/s41597-022-01341-2
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41597-022-01341-2