1. Introduction
Mild cognitive disorders or MCI and major cognitive disorders or dementia are major parts of neurocognitive disorders (NCDs) [
1,
2]. Both are spectrum disorders of chronic and progressive cognitive decline. MCI is a mild form of dementia with preserved daily functions. All such NCDs are common in old adults, with their prevalence increasing with advancing age. Dementia is a disease of concern in geriatrics. The World Health Organization has projected that dementia patients will increase from 57.4 million cases globally in 2019 to 152.8 million cases in 2050 [
3]. Dementia is one of the most feared health conditions among aging people [
4]. Dementia is a significant concern, disrupting daily life and imposing burdens on caregivers, public healthcare systems, and society.
There is currently no curative or reversible treatment for the common types of dementia, such as Alzheimer’s disease, dementia with Lewy bodies, frontotemporal dementia, and vascular dementia [
5]. Two monoclonal antibodies, Aducanumab and Lecanemab, approved for beta-amyloid plaque removal, can slow Alzheimer’s progression, with doubts about their effectiveness and adverse reactions [
6,
7].
Recent research focuses on early NCD stages, particularly the preclinical and MCI phases preceding dementia [
8]. The critical role of early detection for NCDs allows interventions, such as cognitive training and lifestyle modification, to prevent or slow disease progression and improve quality of life [
9]. Biomarkers for early detection, such as brain volumetric and cortical thickness changes from structural MRI, offer insights into NCDs [
10]. Two promising biomarkers [
11] for use in clinical settings are brain volumetric and cortical thickness changes obtained from structural MRI.
The detailed information from magnetic resonance (MR) images aids physicians in diagnosis, prognosis prediction, and treatment planning. MRI brain scans are extensively conducted in both clinical and research settings, revealing the extent and nature of anatomical and pathological changes. Specialized software, like FreeSurfer, version 6.0, can automatically process MR images to generate quantitative data, such as brain volume and cortical thickness, with high accuracy [
12,
13,
14]. There are common methods for brain volume analysis, including absolute and intracranial-corrected volumes [
15].
Brain volume and cortical thickness naturally decrease with aging [
16]. Individuals with NCDs exhibit more pronounced reductions in brain volume and cortical thickness compared to healthy peers [
17]. Changes in brain volume and cortical thickness closely align with the progression of clinical symptoms in NCDs [
18,
19,
20]. The severity of these reductions serves as an indicator of the likelihood of impending progression from MCI to dementia [
21] and aids in identifying the underlying etiology responsible for the clinical deficit.
There is a study to develop models for temporal changes in NCD biomarkers. It elucidates the correlation between biomarker changes and the progression of Alzheimer’s dementia over time [
22]. This model demonstrates that alterations in brain volume and cortical thinning commence many years prior to a typical clinical diagnosis. The application of this model extends to all types of dementia that share similar biomarker changes, encompassing phenomena like brain volume loss and cortical thinning over time.
This study had two objectives: Firstly, to identify a specific brain region and its associated cut-off values that could be used as a clinically relevant biomarker for distinguishing between NC, MCI, and dementia. This has direct implications for clinical practice. Secondly, to explore the relationship between brain parameters (absolute brain volume, ICV-corrected brain volume, and cortical thickness) and diagnostic groups (dementia, MCI, and NC) using linear regression analysis. A key novelty of this research lies in the utilization of a unique dataset comprising a diverse age range within the Thai population. The identified biomarker, characterized by a specific cut-off value, presents a novel application for use within the Thai population. The contributions of this study are significant by (1) providing a comprehensive characterization of the brain structural alterations associated with the progression of NCDs, (2) identifying potential neuroimaging biomarkers that may aid in the early detection and diagnosis of individuals at risk for developing NCDs, and (3) advancing our understanding of the neuroanatomical underpinnings of cognitive decline. It is hypothesized that the findings of this study will facilitate the early detection of NCDs, particularly in the preclinical and MCI stages. This emphasis on early detection is crucial, as early intervention has the potential to significantly decelerate disease progression and enhance patient outcomes.
2. Methods
2.1. Study Population
Participants were recruited from the Mobility, Cognition, Biomarkers and Artificial Intelligence from Clinical Translation in Dementia Spectrum study (MCAD) at the Neuro-Computational Intelligence for Neurocognitive Disorder Laboratory and Memory Clinic, Faculty of Medicine Siriraj Hospital, Mahidol University, Thailand.
One hundred and twenty participants were recruited. The inclusion criteria were participants aged from 35 to 90 years and having undergone an MRI brain scan (dementia protocol) between January 2019 and February 2022. The exclusion criteria were as follows:
A serious neurological condition, uncorrected visual or hearing impairment, or a severe psychiatric disorder (e.g., schizophrenia).
Severe dementia; a bedridden status; uncooperative behavior; a history of brain surgery that may interrupt the intracranial or brain structure; a history or evidence of a brain mass, such as a tumor; hydrocephalus; a history of intracranial hemorrhage; a contraindication for MRI (such as metallic implantation); and claustrophobia.
Four subjects who met the exclusion criteria were eliminated. That left 116 participants (81 females) for the study. Neurologists and psychologists classified the participants into 3 groups: “dementia” (mild to moderate dementia), “MCI”, and “NC”. The diagnoses of dementia and MCI were based on the criteria presented in the
Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition [
23].
Among the participants with dementia, 60.0% had Alzheimer’s disease, 13.3% had frontotemporal dementia, 10% had vascular dementia, 6.7% had Alzheimer’s disease with cerebrovascular disease, and 3.3% each had Lewy body dementia, mixed dementia, and corticobasal degeneration. MCI patients scored >24 on the Thai mental state examination (TMSE), with their cognitive deficits not impeding daily activities. NC participants exhibited no subjective cognitive complaints, normal cognitive performance, and independence functions. All 116 participants underwent laboratory tests to exclude other causes of cognitive decline.
Written informed consent was obtained from all participants. The study protocol received approval from the Human Research Protection Unit, Faculty of Medicine Siriraj Hospital, Mahidol University, before commencement. The approval code is Si 779/2019, and the approval date is 18 November 2019.
2.2. Measurements
2.2.1. Baseline Assessments
Clinical data were obtained through physician interviews, encompassing demographic information such as age, sex, body weight, height, exercise frequency, education level, smoking, and alcohol consumption.
The medical data included details of underlying diseases, medications, and cognitive status. The cognitive function assessment for all 116 participants utilized the Thai mental status examination [
24], administered by a psychologist. The qualitative evaluation of brain MR images involved a neurologist who assessed the Fazekas scale for white matter lesions, medial temporal lobe atrophy (MTA) score, and the presence of lacunar infarction.
2.2.2. MRI Acquisition and Processing
All participants underwent a 3.0 T MRI brain scan (Philips Ingenia 3.0, manufactured by Philips, Amsterdam, The Netherlands), with a dementia protocol. The settings of MRI included high-resolution 3D structural T1-weighted (T1W) images and a 3D MP-RAGE sequence with a 1.0 mm isotropic resolution (TR 8.1, TE 3.7, 8 degrees flip angle, and FOV 240 mm (240 × 240 matrix). The MRI scans were administered between 2019 and 2022.
The T1W MR images, initially in Digital Imaging and Communications in Medicine (DICOM) format, were converted to the Neuroimaging Informatics Technology Initiative (NIfTI) format using MRIcron software, version 2.1.63. These images were then analyzed using FreeSurfer software, version 6.0. The FreeSurfer analysis involved motion correction, affine transformation, normalization into Talairach space with image intensity inhomogeneity correction, automatic skull stripping, and segmentation using an image intensity histogram and the FreeSurfer atlas [
25]. Segmented brain volume and cortical thickness were obtained for statistical analysis in this study. The analysis focused on selected cortical and subcortical regions, providing parameters such as absolute brain volume, ICV-corrected brain volume, and cortical thickness for further statistical analysis.
2.3. Statistical Analysis
The data were analyzed using IBM SPSS Statistics for Windows, version 22. The descriptive statistics for baseline characteristics included frequencies and percentages for the categorical variables and means with standard deviations (SD) or medians with interquartile ranges (IQR) for continuous variables. Comparisons between the NC group, MCI group, and dementia group involved one-way ANOVA for normally distributed data or the Kruskal–Wallis test for abnormal distributions with a post hoc Bonferroni adjustment.
Clinical characteristics, medical data, and brain parameters were subjected to statistical comparisons. The 95% confidence interval of the prevalence was determined using one-way ANOVA and the Kruskal–Wallis test. Probability values (P) less than 0.05 were deemed statistically significant. Linear regression analysis, adjusted for age, education level, and ICV, was employed to ascertain associations between groups and morphometric brain volume (absolute and normalized with the ICV method) and cortical thickness.
A receiver operating characteristic (ROC) curve analysis determined the optimal cutpoint, sensitivity, and specificity for a biomarker in differentiating between groups, while the MedCal software (version 20) generated a positive likelihood ratio, negative likelihood ratio, positive predictive value (PPV), negative predictive value (NPV), and accuracy.
4. Discussion
In a previous study [
22], continuous reductions in brain volume and cortical thickness were observed in dementia and MCI over time. While MRI brain volumetry is recognized as a potential biomarker for clinical settings, there is a lack of studies in Thailand promoting these measures as biomarkers in clinical practice; most investigations were limited to research environments. This study is the first to utilize the quantitative biomarkers of brain volume and cortical thickness changes in diverse Thai cohorts with various dementia subtypes, aiming to contribute to clinical decision-making.
The individuals aged 35 to 90 with dementia, MCI, and NC were observed in our investigation. There are greater neurodegenerative abnormalities in dementia compared to MCI and NC in the Thai population. Absolute brain volumes, including the total cerebral cortex, total cerebral gray matter, subcortical gray matter, total temporal cortex, left and total hippocampus, and left, right, and total amygdala, were significantly smaller in dementia than in MCI and NC. These volumes were also significantly smaller in MCI compared to NC. All selected ROIs exhibited a significantly smaller absolute brain volume in dementia than in NC, confirming our hypothesis of lower absolute brain volume in dementia across all selected ROIs. Additionally, the absolute brain volumes in most selected ROIs in dementia were smaller than those in MCI, while some selected ROIs in MCI had smaller volumes than those in NC.
Normalization of the segmental brain volume by ICV revealed improved significant differences between compared groups. In dementia, all ROIs of the normalized segmental brain volumes were significantly smaller than in NC. Moreover, almost all ROIs of the normalized segmental brain volumes were significantly smaller in dementia than in MCI, except for the total cerebral white matter/ICV and total cerebral cortex/ICV. In MCI, the ROIs of the normalized segmental brain volumes were significantly smaller than in NC, except for the total entorhinal cortex, total hippocampus, and right hippocampus.
Regarding the cortical thickness, all ROIs were thinner in dementia than in NC, and in dementia, they were thinner than in MCI, except for the total, left, and right frontal cortex and the total, left, and right cingulate cortex. The right temporal cortex was the only region significantly thinner in MCI than in NC. Overall, dementia exhibited significantly smaller absolute brain volume, brain volume normalized with ICV, and thinner cortical thickness than both MCI and NC. Additionally, MCI demonstrated smaller brain volume and thinner cortical thickness than NC in specific ROIs.
The ROC curve analysis identified the left amygdala/ICV with optimal sensitivity and specificity for distinguishing dementia from nondementia, MCI, and NC. Cutpoints of 0.092, 0.089, and 0.099 were proposed for clinical use. Values ≤ 0.089 suggested suspected dementia, those between 0.089 and 0.092 indicated MCI, while values between 0.092 and 0.099 suggested the preclinical stage of dementia. Values > 0.099 indicated normal cognitive function. Individuals with values between 0.089 and 0.099 were considered at risk and should undergo further investigation and management to prevent NCD development.
Figure 11 summarizes the proposed cutpoint for the left amygdala/ICV in clinical decision-making for NCDs.
Tau deposition in brain tissue is a common pathology in Alzheimer’s dementia, and its association with reduced amygdala volume has been linked to worsened overall global cognition in individuals in the preclinical stage of the disease or those at risk of developing dementia [
30]. Furthermore, MRI-documented atrophy of the hippocampus and amygdala in cognitively intact older adults has been shown to predict dementia over a 6-year follow-up period [
31]. These findings highlight amygdala volume loss as a risk factor for dementia development, representing one of the early structural changes in NCDs [
32,
33,
34].
The amygdala, almond-shaped structures within the medial temporal lobe, are subcortical gray structures positioned in front of the hippocampus and are functionally connected together, playing a role in cognition. However, the complex shape and location of the amygdala pose challenges for clinicians to visualize and interpret alterations from MR images [
35]. Unlike the hippocampus, the amygdala lacks a clear border, making it difficult to identify volume reduction and compare it with nearby structures.
In contrast, the hippocampus’s surface is easily visualized and evaluated for any shrinkage. Automated segmental brain volume calculations become crucial for accurately determining amygdala volume. While several studies have associated reduced amygdala volume with cognitive decline, fewer have delved into the reduction of amygdala nuclei. A subregion segmentation of amygdala nuclei could provide detailed insights into the specific nuclei responsible for cognitive decline. Additional research into the amygdala and cognitive deterioration in Thai populations, particularly utilizing nuclei volumetric techniques, is proposed.
Linear regression was used in this study with pairwise comparisons for four groups: dementia vs. nondementia, dementia vs. NC, dementia vs. MCI, and MCI vs. NC. Dementia had a total brain volume that was 63.316 mL smaller than that for nondementia, 90.918 mL lower than that for NC, and 62.370 mL less than that for MCI. The total brain volumes of MCI and NC did not differ significantly. Only the total amygdala and the left amygdala volumes demonstrated significant differences between all compared groups. Consequently, the total amygdala and left amygdala volumes had negative linear relationships with the different stages of NCDs.
This study has notable strengths. It is one of the first to investigate the associations between brain volume, cortical thickness, dementia, MCI, and NC in Thais across an extended age range. The inclusion of various dementia types aligns with the clinical setting, enhancing the tool’s relevance for early detection of NCDs. This study introduces cutpoints for evaluating and categorizing patients, proposing brain volume as a potential biomarker for detecting early stages of NCDs. Early identification and intervention can significantly impact patient outcomes. While this study encourages the application of these cutpoints for Thais in clinical and research settings, it acknowledges the need for further trials with diverse age groups and specific diagnoses to validate the proposed values.
There are several clinical implications of this study. Firstly, the identification of the left amygdala/ICV ratio as a sensitive and specific biomarker for NCD diagnosis has significant clinical utility. This biomarker could potentially aid clinicians in early identification of individuals at risk for developing NCDs. Early detection allows for timely interventions, such as cognitive training and lifestyle modifications, which may slow disease progression and improve quality of life. Secondly, this biomarker can also improve diagnostic accuracy. Using the left amygdala/ICV ratio can assist in differentiating between NC, MCI, and dementia, leading to more accurate and confident diagnoses. Thirdly, this biomarker can help stratify individuals based on their risk of progression from MCI to dementia in Thais.
This study has some limitations. Firstly, the sample size, while relatively large, may still be considered moderate. Secondly, the diverse subtypes of dementia with unique pathophysiologies affecting different brain regions may have reduced the study’s statistical power. To address this, focusing on a specific subtype of MCI or dementia could enhance the study’s power. Lastly, the cross-sectional design of this study limits the ability to draw definitive conclusions about the temporal relationship between brain changes and cognitive decline. Further longitudinal studies are needed to confirm these findings and investigate the predictive value of the left amygdala/ICV ratio in predicting future cognitive decline, offering insights into preventative treatments.
Future research should be expanded to include larger and more diverse populations to enhance the generalizability of these findings. The application of various machine learning techniques as alternative presentations of statistics offers the potential to significantly enhance diagnostic accuracy by analyzing a broader range of data. A crucial area for future research involves identifying machine learning models that can accurately predict the probability and rate of conversion from NC to any stage of NCDs. This can be achieved by utilizing the left amygdala/ICV ratio, alone or in combination with other biomarkers, within longitudinal datasets. Such predictive models have the potential to revolutionize early diagnosis and facilitate the development of effective preventive strategies.