Creation of structural MRI brain templates for middle to late adulthood in the Thai population using single and multi-scanner data

Introduction

The global population is undergoing rapid changes, characterized by a substantial growth in the number of elderly individuals, particularly in low- to middle-income countries (1). Alongside this demographic shift comes a pressing challenge: the rise in age-related health conditions, notably Alzheimer’s disease (AD), which accounts for the predominant type of dementia cases worldwide. With AD predominantly affecting individuals over 65 years old, there is an urgent demand for early detection, continuous monitoring, and effective treatment strategies to address this growing public health concern (2-4).

Structural magnetic resonance imaging (MRI) plays a central role in brain research and clinical diagnostics, particularly in the context of aging and neurodegeneration (5,6). However, to enable meaningful group comparisons and robust quantitative brain analyses, individual brain scans must first be aligned to a shared coordinate system. This process, known as spatial normalization, is essential for reducing anatomical variability across individuals and for enabling voxel-wise or region-based comparisons (7-9). Brain templates serve as the anatomical references for this normalization, supporting a wide range of neuroimaging analyses—including voxel-based morphometry (VBM), tensor-based morphometry, and surface-based morphometry—which depend on accurate alignment to detect subtle structural changes across populations (10,11).

The evolution of standard brain templates has seen significant progress, from early models based on limited data to more comprehensive templates derived from larger and more diverse populations (12-15). Some templates have been integrated into several analysis tools, such as Statistical Parametric Mapping (SPM, Institute of Neurology, University College London, UK) (16), MriStudio (17), and the FMRIB Software Library (FSL, University of Oxford, UK) (18). However, the need for population-specific templates is underscored by the variations in brain anatomy across different ethnic and racial groups (19-23). As the brain matures over time, it experiences notable structural and functional shifts. These encompass reductions in volume, thinning of the cortex, deterioration of white matter, and modifications in the patterns of sulci and gyri (24). Recent studies have demonstrated the superiority of age-specific templates in capturing the structural changes associated with aging, offering greater accuracy in brain analysis (25,26). Moreover, the choice between utilizing data from a single MRI scanner or multiple scanners presents distinct advantages and challenges. While a single-scanner approach ensures consistency and homogeneity in the dataset, multi-scanner data acquisition provides a broader perspective, potentially enhancing the template’s generalizability.

In this study, we aim to address these challenges by developing brain templates tailored to the Thai normal middle to late adulthood population. By leveraging both single-scanner and multi-scanner data, we seek to create a template that accurately represents the diversity and characteristics of the Thai population in this age group. This research endeavors to contribute to the advancement of neuroimaging techniques for neurological conditions such as AD.

Methods

Participants

This study was conducted in accordance with the ethical principles of the Declaration of Helsinki and its subsequent amendments. The study was approved by the committees of the Faculty of Medicine, Chiang Mai University (CoA No. 427/2020), which oversees research conducted at Maharaj Nakorn Chiang Mai Hospital (the university hospital of the Faculty of Medicine, Chiang Mai University), and the Faculty of Medicine, Siriraj Hospital, Mahidol University (CoA No. 666/2016) and informed consent was taken from all the patients. A two-year prospective cohort study was conducted, recruiting Thai adults aged between 50 and 70 years from two medical centers in Thailand. The data were collected 2 times (at baseline and in the second year of the follow-up). Participants in this study were recruited from public media via social media and posters. The participants were selected from outpatient clinics at Siriraj Hospital and Maharaj Nakorn Chiang Mai Hospital, including the family medicine clinic, screening clinic, geriatric clinic, and health checkup clinic.

To ensure the reliability of our findings, we established specific inclusion and exclusion criteria for participant recruitment. Inclusion criteria for participation in the study were as follows:

• Cognitive normality, determined by the absence of memory complaints and achieving a Montreal Cognitive Assessment (MoCA) test score of 25 or above out of 30.
• Normal performance in Activities of Daily Living (ADL), indicated by a score of 20.
• Absence of pathological findings on MRI scans, such as intracerebral infarction, intracerebral hemorrhage, or moderate to severe white matter changes based on the Fazekas scale.

Participants with a previous diagnosis or current medication regimen for any neurological, psychiatric, or mental health disorders were excluded from the study. Additionally, individuals with a Thai Mental State Examination (TMSE) score below 24 or a Patient Health Questionnaire-9 (PHQ-9) score exceeding 6 were also excluded from participation.

At the two-year mark, participants underwent another round of MoCA testing to confirm their cognitive status. Individuals who demonstrated cognitive decline, indicated by a MoCA test score below 25, were subsequently excluded from the analysis. This step was taken to ensure that only individuals who maintained cognitive normality throughout the duration of the study were included in the final analysis, thereby enhancing the accuracy and reliability of the findings.

Details of measures

Thai version of MoCA

MoCA is a comprehensive screening tool consisting of 30 questions designed to evaluate multiple cognitive domains (27). These include visuospatial abilities tested through tasks like cube copying and clock drawing, executive functions assessed with adapted trail making tests, language abilities gauged by naming animals and sentence repetition, and short-term memory evaluated through delayed recall tasks. Additionally, attention, concentration, and working memory are tested through tasks such as digit span and serial subtraction. Orientation to time and place is also assessed. The scoring system for Thai MoCA considers education level, with adjustments made to scores for participants with fewer than 6 years of education. The cut-off score for mild cognitive impairment is 24/25 (28).

The Barthel ADL index

The Barthel ADL index is used to assess an individual’s ability to perform basic daily tasks independently (29). This assessment covers domains such as feeding, hygiene, transferring, toileting, ambulation, dressing, climbing stairs, bathing, and continence (both stool and urine). Scores range from 0 to 20, with higher scores indicating greater independence in ADL.

TMSE

TMSE is a screening test for dementia tailored for Thai populations (30), taking approximately 5 to 10 minutes to administer. The test assesses various cognitive functions, including orientation to time and place, registration of information, attention and calculation abilities, language skills, and memory recall. Participants can score a maximum of 30 points, with higher scores reflecting better cognitive performance. Typically, a score of 23/24 is considered the cutoff point for detecting dementia.

The PHQ-9

PHQ-9 is a self-administered questionnaire comprising nine items designed to assess the severity of depressive symptoms experienced by an individual over the past two weeks (31). Respondents rate the frequency of their symptoms on a 4-point Likert scale, ranging from 0 (not at all) to 3 (nearly every day). Total scores on the PHQ-9 can range from 0 to 27, with higher scores indicating a greater level of depressive mood.

Data collection

The data collection process from two medical centers is depicted in Figure 1.

Figure 1 Data selection from three MRI scanners at two medical centers. CMU, Maharaj Nakorn Chiang Mai Hospital, Chiang Mai University; SI, Siriraj Hospital, Mahidol University; MRI, magnetic resonance imaging; SNR, signal-to-noise ratio.

MRI data acquisition and brain template construction

The participants had an MRI brain scan at their initial visit.

MRI acquisition

Scanner-I: 150 participants underwent three-dimensional (3D) T1-weighted MRI scanning using a 3T MRI scanner (GE Pioneer). The imaging parameters were as follows: repetition time (TR)/echo time (TE)/inversion time (TI) =9.8, 4.6, 1,000 ms, flip angle =8 degrees, field of view =240 mm × 240 mm, matrix size =256×256, voxel size =0.94×0.94×1.00 mm³, and 180 slices.

Scanner-II: 80 participants underwent 3D T1-weighted MRI scanning using a 1.5T MRI scanner (GE Signa). The imaging parameters were as follows: TR/TE/TI =9.2, 3.7, 1000 ms, flip angle =9 degrees, field of view =240 mm × 240 mm, matrix size =256×256, voxel size =0.94×0.94×1.00 mm³, and 180 slices.

Scanner-III: 80 participants underwent 3D T1-weighted MRI scanning using a 1.5T MRI scanner (Philips Ingenia). The imaging parameters were as follows: TR/TE/TI =9.7, 4.8, 748 ms, flip angle =8 degrees, field of view =240 mm × 240 mm, matrix size =256×256, voxel size =0.94×0.94×1.00 mm³, and 180 slices.

MRI data pre-processing

The DICOM files were converted to Neuroimaging Informatics Technology Initiative (NIfTI) format using DICOM to NIfTI Converter (DCM2NIIX). The images were corrected for intensity variations using Advanced Normalization Tools (ANTs) “N4BiasFieldCorrection”. The intensity-corrected images were then registered to the Montreal Neurological Institute’s 152 brain template (MNI152) standard space using FSL’s FMRIB’s Linear Image Registration Tool (FLIRT) with 6 degrees of freedom (32-34). The rigid registered images were averaged to create an initial brain template.

Iterative brain template construction

To create the MRI brain template, we employed an iterative technique designed to enhance registration accuracy. The entire procedure is graphically outlined in Figure 2A. The process began by registering all individual MRI volumes to an initial reference template using the symmetric normalization (SyN) non-rigid registration algorithm within the ANTs toolkit. The specific bash commands used for this step were:

• MRI volumes were registered to the reference (i.e., the initial brain template) using ANTs with SyN non-rigid registration. The bash commands are as below:
  • ANTS 3 -m CC[input.nii, reference.nii, 4, 4] -t SyN[0.1] -r Gauss[3,0.5] -o T_ -i 100×100×100×20
  • WarpImageMultiTransform 3 input.nii output.nii -R reference.nii -i T_Affine.txt T_InverseWarp.nii.gz All non-rigidly registered images were averaged to create an updated brain template.
• Steps 1 and 2 were repeated until convergence was reached – the normalized root mean square error between the template obtained by N and N-1 iterations is lower than 5%, where N is the latest iteration number. Note that the reference in step 1 was replaced by the update template for every iteration.

Figure 2 Workflow for data preprocessing and iterative template construction. (A) The conversion and registration steps leading to the creation of population-specific Thai brain templates. (B) Representative brain images from three different MRI scanners used for multi-scanner template construction. ANTs, Advanced Normalization Tools; DCM2NIIX, DICOM to NIfTI Converter; DICOM, Digital Imaging and Communications in Medicine; FLIRT, FMRIB’s Linear Image Registration Tool; FSL, FMRIB Software Library; MNI, Montreal Neurological Institute; MRI, magnetic resonance imaging; NIFTI, Neuroimaging Informatics Technology Initiative; SyN, symmetric normalization.

Additionally, FSL’s FAST (35) was applied to the rigid-registered images to obtain individual tissue probability maps, including gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). Subsequently, the deformation fields obtained at the N^th iteration of brain template construction were used to warp these tissue probability maps and then averaged to create templates of GM, WM, and CSF (35). Note that Jacobian determinant modulation was not applied during warping. The warped tissue probability maps were intended to represent average tissue distributions in template space, rather than volume-preserved measures. While unmodulated warping may introduce bias if used for local volume estimation, in this study it was applied only for visualizing the mean distribution of tissues and not for quantitative volumetric analysis.

Using this method, we constructed two distinct Thai brain templates, differentiated by the source of the imaging data as shown in Figure 2B. The first, a single-scanner template (TH150), was created from 150 datasets acquired exclusively from scanner I. The second, a multi-scanner template (TH240), was constructed from a total of 240 datasets, comprising 80 subjects from each of the three different scanners (I, II, and III).

Comparison of brain spatial dimensions

The measurements were performed after all individuals had been rigidly registered to the MNI152 space as part of the initial preprocessing step (described in Step 1 of the brain template construction). We measured the length (L) of the whole brain from the anterior pole to the posterior pole on the anterior commissure-posterior commissure (AC-PC) plane, the width (W) at the midpoint of the AC-PC line, and the height (H) from the superior to the inferior pole on the coronal plane.

For the comparison, the same measurements were also performed on two publicly available templates, Chinese brain template (CN200) and Caucasian brain template (US200), which can be accessed at https://www.nitrc.org/projects/us200_cn200/. These datasets provide head (unmasked), brain (masked), and tissue probability maps for GM, WM, and CSF.

Variability assessment

This assessment was conducted to evaluate anatomical differences during registration by analyzing the Jacobian determinant, which is derived from deformation fields. This process ensures accurate population-specific representation and improves the alignment between individual brains and the template. Thirty new cases (15 males and 15 females) with an average age of 67.5±7.2 years for males and 66.1±7.0 years for females were included in the assessment. SyN non-rigid registration (ANTs toolbox) was performed four times, each time using a different target volume: TH150, TH240, CN200, and US200. After the registration process, the deformations for each individual were used to calculate the Jacobian determinant. A value greater than zero indicates that the template expands to match the moving image (i.e., when the moving image is larger than the fixed image), while a value less than zero indicates that the template contracts to match the moving image.

Statistical analysis

MATLAB (2016b, The Mathworks, Natick, MA, USA) was used for statistical analyses, figures, and graph generations. One-way analysis of variance was used to compare the age and individual brain features (i.e., width, length, and height) among the three groups. The statistical significance was set at a P value of less than 0.05.

Results

The brain spatial dimensions of individuals

In Table 1, the brains of normal Thai individuals in middle to late adulthood showed mostly similar width, length, and height across the scanners, with no statistically significant differences (P values >0.05). The average age, width, length, and height were approximately 61 years, 134 mm, 159 mm, and 90 mm, respectively.

The brain spatial dimensions of the templates

In Figure 3, despite having the same isotropic resolution of 1 mm, CN200 and US200 templates appear visually sharper and have higher contrast compared to the Thai brain templates. The Thai templates exhibit larger ventricle areas, while their width, length, and height measurements are smaller than those observed in CN200 and US200 (36).

Figure 3 The TH150, TH240, US200, and CN200, brain (masked) templates.

In Table 2, Thai brain templates have smaller brain spatial dimensions compared to both CN200 and US200. Among the four templates, CN200 is the widest and the highest, being approximately 5.97% wider and 4.35% higher than the Thai templates. US200 is the longest, with a length 8.59% greater than the Thai templates. However, the W/L, H/L, and H/W ratios of CN200 and the Thai templates are identical, which are 0.85, 0.58, and 0.68, respectively. Only H/W ratio of US200 is greater than CN200 and the Thai templates, which is 0.71.

The first row of Figure 4 displays the unmasked Thai brain templates, which include non-brain tissues. These templates have identical brain spatial dimensions and structural features as the masked Thai templates shown in Figure 3, which exclude non-brain tissues. Tissue probability maps were also created for GM, WM, and CSF templates. Each voxel within a tissue probability map contains a value between 0 and 1 representing the average likelihood that that voxel represents the tissue type.

Figure 4 Brain spatial dimensions of the unmasked TH150 and TH240 brain templates with their tissue probability maps. CSF, cerebrospinal fluid; GM, gray matter; WM, white matter.

Figure 5 shows that tissue probability Thai templates have less GM and WM and more CSF compared to CN200 and US200. Tissue probability volumes (mL) were computed by thresholding each tissue probability map at 0.5. Voxels with values greater than 0.5 were included, and the total number of these voxels was multiplied by the voxel volume (in mm³). The resulting volume was then converted to milliliters (1 mL =1,000 mm³). The GM volume of Thai templates is approximately 16% and 22% less than those obtained from CN200 and US200, respectively. The WM volume of Thai templates is approximately 2% and 9% less than those obtained from CN200 and US200, respectively. In contrast, the CSF volume of Thai templates is approximately 27% and 68% greater than those obtained from CN200 and US200, respectively. Note that the differences of all tissue probability maps between TH150 and TH240 are less than 1%.

Figure 5 Tissue probability templates (gray matter, white matter, and cerebrospinal fluid) (top) and tissue probability volumes for TH150, TH240, CN200, and US200 (bottom). CSF, cerebrospinal fluid; GM, gray matter; WM, white matter.

Figure 6 displays the results for the registration of 30 new Thai cases with four different brain templates. The Jacobian determinant is greater than zero when the template expands to match the moving image and less than zero when it contracts. Although all templates demonstrated both regions of expansion and contraction, CN200 and US200 exhibited stronger contraction effects, with absolute values reaching up to 0.5, particularly at the cortical boundaries. By contrast, TH150 and TH240 showed relatively milder deformations, with most absolute values in the range of 0 to 0.3. This indicates that the Thai templates required less extreme warping to accommodate the new Thai cases.

Figure 6 Mean of Jacobian determinant of 30 new cases greater than 0.2 and less than −0.2.

Discussion

The creation of MRI brain templates for the middle to late adulthood Thai population using both TH150 and TH240 datasets provides insights into the morphological characteristics and variations of brain structures within this demographic. The necessity for population-specific brain templates is well-documented. Anatomical variations in brain structures among different ethnic and racial groups have significant implications for neuroimaging research and clinical practice. Previous studies have shown that brain templates constructed from one population may not be directly applicable to another due to these structural differences (20,21). For instance, templates derived from Western populations often exhibit disparities when applied to Asian populations, leading to potential inaccuracies in brain measurements and analyses (19,22). This study specifically addresses these concerns by constructing two Thai brain templates: TH150, derived from a single-scanner dataset, and TH240, derived from multi-scanner data. The inclusion of both single and multi-scanner data allows for a comparison of the consistency and generalizability of the templates, providing a robust framework for future research and clinical applications in the Thai population.

The study employed a comprehensive approach to participant selection, MRI data acquisition, and template construction. The inclusion criteria ensured that only cognitively normal individuals without significant neurological or psychiatric conditions were included, enhancing the reliability of the findings. The use of MoCA, Barthel ADL index, TMSE, and PHQ-9 for participant screening underscores the study’s commitment to maintaining high standards of cognitive and mental health assessment (4).

The iterative brain template construction process—based on non-rigid registration and averaging of MRI volumes—ensures high accuracy and precision in the final templates. The study also employed MRI data preprocessing techniques, such as intensity correction and registration to standard space, to minimize variability and enhance the quality of the templates. One of the key findings of this study is the morphological comparison of brain spatial dimensions between the Thai brain templates (TH150 and TH240) and existing templates from other populations (CN200 and US200).

The CN200 and US200 templates appear slightly sharper than the Thai templates, which may be attributed to multiple factors, including differences in preprocessing pipelines and unknown interpolation or regularization strategies. Additionally, the CN200 and US200 templates were constructed from younger populations with less anatomical variability, whereas our templates were derived from older adults (aged 50–70 years), whose greater inter-individual variability may contribute to a smoother averaged appearance.

The results revealed that the Thai brain templates have smaller spatial dimensions compared to the CN200 and US200 templates, underscoring the importance of using population-specific templates for accurate neuroimaging analyses (15). The Thai templates also exhibited larger ventricular areas and smaller width, length, and height measurements. Furthermore, the tissue probability maps showed that the Thai templates had lower GM and WM volumes but higher CSF volumes compared to CN200 and US200, highlighting the necessity of using templates that accurately reflect the anatomical characteristics of the target population (14,22). In addition to CN200 and US200, the morphological brain features of other existing templates are reported in Table 3. Among these, the morphological features of the IBA100 appear similar to those of the Thai templates. However, the IBA100 template was constructed using data from an Indian population aged 21 to 30 years. Moreover, the standard deviations of the brain features of Indians are smaller than those of Thais, with the maximum values being 7.55 mm (length) versus 11.88 mm (width). Thus, the Indian brain templates may not be perfectly suitable for the middle to late adulthood of the Thai population.

Although our findings demonstrate that the Thai brain templates exhibit smaller brain dimensions and larger ventricular spaces compared to CN200 and US200 templates, these differences must be interpreted with caution. The Thai templates were derived from middle- to late-adulthood participants (aged 50–70 years), while the CN200 and US200 templates are based on younger cohorts (aged 19–37 years). Numerous MRI studies have shown that brain atrophy and increased CSF volume are typical features of healthy aging (41). Therefore, without age-matched controls, these differences cannot be conclusively attributed to population-based anatomical variation. Future template comparisons should incorporate age-matched datasets to better distinguish age-related from population-specific effects.

The validation of the TH150 and TH240 templates against new Thai cases demonstrated their accuracy and reliability. The study used the Jacobian determinant to assess the registration quality (2,40,42), showing that the Thai templates performed better than CN200 and US200 templates in registering to Thai brains. This validation demonstrates the efficacy of the Thai templates in capturing the structural nuances of the Thai population.

The developed templates and tissue probability maps can be directly applied in neuroimaging pipelines such as VBM using SPM (16) or FSL (18), where the template facilitates spatial normalization and the probability maps act as priors for tissue segmentation.

Using a single-scanner template provides consistent imaging characteristics, which is advantageous for within-site studies by minimizing scanner-related variability. In contrast, multi-scanner templates incorporate greater variability across acquisition settings, potentially improving their generalizability in multi-site studies, albeit with increased heterogeneity in image features.

To our knowledge, this study presents the first publicly available MRI-based brain templates specifically developed for the Thai population. Previous brain templates from Thailand have primarily utilized other imaging modalities. For example, Chotipanich et al. (42) developed PET-based brain templates using ¹⁸F-THK5351 and ¹¹C-PiB data from 24 healthy Thai individuals (13 men and 11 women, aged 42–79 years). Although MRI was used in that study for anatomical co-registration and spatial normalization of the PET data, the final templates and analyses focused solely on PET signals. No brain morphological metrics or comparisons to other population templates were reported.

In this work, some limitations should be addressed to improve future results. Firstly, the slightly different imaging parameters used across various scanners with differences in magnetic field (1.5 and 3.0 T) likely resulted in variations in image sharpness and contrast, leading to a blurry appearance and brighter detection in the TH240 template compared to the TH150 template. Implementing quality control for imaging parameters and ensuring image similarity across scanners could reduce these issues. Secondly, the relatively small number of subjects per scanner may limit the robustness and generalizability of the findings. Increasing the number of subjects per scanner would enhance the statistical power and reliability of the results. Additionally, incorporating data from a greater variety of scanners could provide a more comprehensive and representative dataset, thereby improving generalizability.

Conclusions

This study developed MRI brain templates for the middle- to late-adulthood Thai population, using both single-scanner and multi-scanner datasets. The TH150 and TH240 templates provide accurate references for neuroimaging studies, addressing the anatomical variations specific to the Thai demographic. The findings emphasize the importance of population-specific templates in improving the accuracy of brain image analysis and enhancing the diagnosis and treatment of neurological conditions such as AD. By integrating data from multiple scanners, the TH240 template offers a robust and generalizable tool for neuroimaging research, aligning with current advancements in the field.

Acknowledgments

The researchers are grateful for the support of the Research Medical Fund, Faculty of Medicine, Chiang Mai University, and the Faculty of Medicine Siriraj Hospital, Mahidol University, which ensured the authors’ independence in the study’s design, data collection, analysis, interpretation, and manuscript preparation.

Footnote

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1096/dss

Funding: This work was supported by the Research Medical Fund Faculty of Medicine, Chiang Mai University (No. 065/2564), the Research Medical Fund Faculty of Medicine Siriraj Hospital, Mahidol University (No. (IO)R016036003), and by the Chalermphakiat grant of the Faculty of Medicine Siriraj Hospital.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1096/coif). K.P., U.Y., C.A., A.S., and S.A. report funding received from the Research Medical Fund, Faculty of Medicine, Chiang Mai University (No. 065/2564). O.C., C.N., and W.M. report funding support from the Research Medical Fund, Faculty of Medicine Siriraj Hospital, Mahidol University (No. (IO)R016036003). The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the ethical principles of the Declaration of Helsinki and its subsequent amendments. The study was approved by the committees of the Faculty of Medicine, Chiang Mai University (CoA No. 427/2020), which oversees research conducted at Maharaj Nakorn Chiang Mai Hospital (the university hospital of the Faculty of Medicine, Chiang Mai University), and the Faculty of Medicine, Siriraj Hospital, Mahidol University (CoA No. 666/2016) and informed consent was taken from all the patients.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Mathews C. Population Ageing in Lower- and Middle-income Countries. Celebrating 40 Years of HelpAge International. Population Ageing 2024;17:1-4.
2. Jaul E, Barron J. Age-Related Diseases and Clinical and Public Health Implications for the 85 Years Old and Over Population. Front Public Health 2017;5:335. [Crossref] [PubMed]
3. Shlisky J, Bloom DE, Beaudreault AR, Tucker KL, Keller HH, Freund-Levi Y, Fielding RA, Cheng FW, Jensen GL, Wu D, Meydani SN. Nutritional Considerations for Healthy Aging and Reduction in Age-Related Chronic Disease. Adv Nutr 2017;8:17-26. [Crossref] [PubMed]
4. Muangpaisan W, Ngamsombat C, Vichianin Y, Dumrikarnlert C, Seeboonruang A, Pooliam J, Charnchaowanish P, Pongnapang N, Kanjanapong S, Phannarus H, Pongmoragot J, Chawalparit O. Brain Volume Metric Analysis Is Correlated with Aging Changes and Sex Differences in Thai Older Adults. Dement Geriatr Cogn Dis Extra 2025;15:47-57. [Crossref] [PubMed]
5. Frisoni GB, Fox NC, Jack CR Jr, Scheltens P, Thompson PM. The clinical use of structural MRI in Alzheimer disease. Nat Rev Neurol 2010;6:67-77. [Crossref] [PubMed]
6. Jack CR Jr, Holtzman DM. Biomarker modeling of Alzheimer's disease. Neuron 2013;80:1347-58. [Crossref] [PubMed]
7. Despotović I, Goossens B, Philips W. MRI segmentation of the human brain: challenges, methods, and applications. Comput Math Methods Med 2015;2015:450341. [Crossref] [PubMed]
8. Mikheev A, Nevsky G, Govindan S, Grossman R, Rusinek H. Fully automatic segmentation of the brain from T1-weighted MRI using Bridge Burner algorithm. J Magn Reson Imaging 2008;27:1235-41. [Crossref] [PubMed]
9. Whitwell JL. Voxel-based morphometry: an automated technique for assessing structural changes in the brain. J Neurosci 2009;29:9661-4. [Crossref] [PubMed]
10. Matsuda H. Voxel-based Morphometry of Brain MRI in Normal Aging and Alzheimer's Disease. Aging Dis 2013;4:29-37.
11. Schmitter D, Roche A, Maréchal B, Ribes D, Abdulkadir A, Bach-Cuadra M, Daducci A, Granziera C, Klöppel S, Maeder P, Meuli R, Krueger GAlzheimer's Disease Neuroimaging Initiative. An evaluation of volume-based morphometry for prediction of mild cognitive impairment and Alzheimer's disease. Neuroimage Clin 2015;7:7-17. [Crossref] [PubMed]
12. Collins DL, Neelin P, Peters TM, Evans AC. Automatic 3D intersubject registration of MR volumetric data in standardized Talairach space. J Comput Assist Tomogr 1994;18:192-205.
13. Mandal PK, Mahajan R, Dinov ID. Structural brain atlases: design, rationale, and applications in normal and pathological cohorts. J Alzheimers Dis 2012;31:S169-88. [Crossref] [PubMed]
14. Mazziotta JC, Toga AW, Evans AC, Fox PT, Lancaster JL. Digital brain atlases. Trends Neurosci 1995;18:210-1. [Crossref] [PubMed]
15. Talairach J, Tournoux P. Co-planar Stereotaxic Atlas of the Human Brain: 3-dimensional Proportional System : an Approach to Cerebral Imaging. New York: G. Thieme; 1988.
16. Statistical Parametric Mapping. [online]. London, UK. [Accessed 16 September 2025]. Available online: https://www.fil.ion.ucl.ac.uk/spm/
17. MRI Studio [online].; 2007 May [Accessed 16 September 2025]. Available online: http://www.mristudio.org/
18. FMRIB Software Library (FSL). [online]. [Accessed 16 September 2025]. Available online: http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/
19. Bhalerao GV, Parlikar R, Agrawal R, Shivakumar V, Kalmady SV, Rao NP, Agarwal SM, Narayanaswamy JC, Reddy YCJ, Venkatasubramanian G. Construction of population-specific Indian MRI brain template: Morphometric comparison with Chinese and Caucasian templates. Asian J Psychiatr 2018;35:93-100. [Crossref] [PubMed]
20. Lee JS, Lee DS, Kim J, Kim YK, Kang E, Kang H, Kang KW, Lee JM, Kim JJ, Park HJ, Kwon JS, Kim SI, Yoo TW, Chang KH, Lee MC. Development of Korean standard brain templates. J Korean Med Sci 2005;20:483-8. [Crossref] [PubMed]
21. Tang Y, Hojatkashani C, Dinov ID, Sun B, Fan L, Lin X, Qi H, Hua X, Liu S, Toga AW. The construction of a Chinese MRI brain atlas: a morphometric comparison study between Chinese and Caucasian cohorts. Neuroimage 2010;51:33-41. [Crossref] [PubMed]
22. Zilles K, Kawashima R, Dabringhaus A, Fukuda H, Schormann T. Hemispheric shape of European and Japanese brains: 3-D MRI analysis of intersubject variability, ethnical, and gender differences. Neuroimage 2001;13:262-71. [Crossref] [PubMed]
23. Liang P, Shi L, Chen N, Luo Y, Wang X, Liu K, Mok VC, Chu WC, Wang D, Li K. Construction of brain atlases based on a multi-center MRI dataset of 2020 Chinese adults. Sci Rep 2015;5:18216. [Crossref] [PubMed]
24. Lin HY, Huang CC, Chou KH, Yang AC, Lo CZ, Tsai SJ, Lin CP. Differential Patterns of Gyral and Sulcal Morphological Changes During Normal Aging Process. Front Aging Neurosci 2021;13:625931. [Crossref] [PubMed]
25. Gu D, Shi F, Hua R, Wei Y, Li Y, Zhu J, et al. An artificial-intelligence-based age-specific template construction framework for brain structural analysis using magnetic resonance images. Hum Brain Mapp 2023;44:861-75. [Crossref] [PubMed]
26. Holla B, Taylor PA, Glen DR, Lee JA, Vaidya N, Mehta UM, et al. A series of five population-specific Indian brain templates and atlases spanning ages 6-60 years. Hum Brain Mapp 2020;41:5164-75. [Crossref] [PubMed]
27. Nasreddine ZS, Phillips NA, Bédirian V, Charbonneau S, Whitehead V, Collin I, Cummings JL, Chertkow H. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment. J Am Geriatr Soc 2005;53:695-9. [Crossref] [PubMed]
28. Hemrungrojn S, Tangwongchai S, Charoenboon T, Panasawat M, Supasitthumrong T, Chaipresertsud P, Maleevach P, Likitjaroen Y, Phanthumchinda K, Maes M. Use of the Montreal Cognitive Assessment Thai Version to Discriminate Amnestic Mild Cognitive Impairment from Alzheimer's Disease and Healthy Controls: Machine Learning Results. Dement Geriatr Cogn Disord 2021;50:183-94. [Crossref] [PubMed]
29. Collin C, Wade DT, Davies S, Horne V. The Barthel ADL Index: a reliability study. Int Disabil Stud 1988;10:61-3. [Crossref] [PubMed]
30. Thai Tobacco Free Bylaw Committee. Thai Mental State Examination (TMSE) [online]. Sirifaj Hospital; 1993 [Accessed 15 December 2024];45(6):359-74. Available online: https://suicide.dmh.go.th/abstract/details.asp?id=1235
31. Lotrakul M, Sumrithe S, Saipanish R. Reliability and validity of the Thai version of the PHQ-9. BMC Psychiatry 2008;8:46. [Crossref] [PubMed]
32. Jenkinson M, Bannister P, Brady M, Smith S. Improved optimization for the robust and accurate linear registration and motion correction of brain images. Neuroimage 2002;17:825-41. [Crossref] [PubMed]
33. Jenkinson M, Smith S. A global optimisation method for robust affine registration of brain images. Med Image Anal 2001;5:143-56. [Crossref] [PubMed]
34. Mazziotta J, Toga A, Evans A, Fox P, Lancaster J, Zilles K, 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 2001;356:1293-322. [Crossref] [PubMed]
35. Zhang Y, Brady M, Smith S. Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE Trans Med Imaging 2001;20:45-57. [Crossref] [PubMed]
36. Yang G, Zhou S, Bozek J, Dong HM, Han M, Zuo XN, Liu H, Gao JH. Sample sizes and population differences in brain template construction. Neuroimage 2020;206:116318. [Crossref] [PubMed]
37. Mazziotta J, Toga A, Evans A, Fox P, Lancaster J, Zilles K, et al. A four-dimensional probabilistic atlas of the human brain. J Am Med Inform Assoc 2001;8:401-30. [Crossref] [PubMed]
38. Evans A, Collins L, Mills SR, Brown ED, Kelly RL, Peters T. 3D statistical neuroanatomical models from 305 MRI volumes. In: Proceedings of the IEEE Nuclear Science Symposium and Medical Imaging Conference; 31 October – 6 Nov 1993, San Francisco, CA. IEEE; 1993. p. 1813-7.
39. Sivaswamy J, Thottupattu AJ, Mehta R, Sheelakumari R, Kesavadas C. Construction of Indian human brain atlas. Neurol India 2019;67:229-34. [Crossref] [PubMed]
40. Lee H, Yoo BI, Han JW, Lee JJ, Oh SY, Lee EY, Kim JH, Kim KW. Construction and Validation of Brain MRI Templates from a Korean Normal Elderly Population. Psychiatry Investig 2016;13:135-45. [Crossref] [PubMed]
41. Fujita S, Mori S, Onda K, Hanaoka S, Nomura Y, Nakao T, Yoshikawa T, Takao H, Hayashi N, Abe O. Characterization of Brain Volume Changes in Aging Individuals With Normal Cognition Using Serial Magnetic Resonance Imaging. JAMA Netw Open 2023;6:e2318153. [Crossref] [PubMed]
42. Chotipanich C, Kongthai S, Kunawudhi A, Promteangtrong C, Jantarato A. (18)F-THK 5351 and (11)C-PiB PET of the Thai normal brain template. Asia Ocean J Nucl Med Biol 2021;9:21-30. [Crossref] [PubMed]