Amide proton transfer could be a surrogate imaging marker for predicting vascular cognitive impairment

,


Introduction
Cerebral small vessel disease (CSVD) is a clinical, magnetic resonance imaging (MRI), and pathological manifestation syndrome that arises from various intracranial small vessel lesions.It has emerged as a significant hotpot in neurology, following stroke and Alzheimer's disease (Pantoni, 2010).Patients with CSVD often exhibit a range of cognitive impairments, including deficits in attention networks, processing speed, executive function, and cognitive flexibility (Zhào et al., 2019).The impairment of processing speed is considered one of the earliest and most prominent cognitive manifestations of CSVD (Duering et al., 2013).Furthermore, in addition to mild cognitive impairment (MCI), vascular dementia (VaD) is a significant clinical consequence of CSVD, accounting for over 40% of all dementia cases (Wardlaw, 2013).
Previous studies have demonstrated the association between imaging markers and cognitive dysfunction in patients with cerebral small vessel disease (CSVD) (Duering et al., 2012;Ter Telgte et al., 2018).The impact of CSVD on cognitive function has been extensively researched, with particular focus on individual imaging markers.Among these markers, white matter hyperintensities (WMH) have received the most attention (Ter Telgte et al., 2018;Tuladhar et al., 2020).However, recent studies that have investigated the combination and correlation of multiple imaging markers have yielded inconclusive or occasionally contradictory results (Banerjee et al., 2018;Liu et al., 2020).Several studies have suggested a dose-response relationship, linking the quantity of WMH and lacunar infarctions (LIs) to the severity of cognitive impairment (Tuladhar et al., 2020;Prins et al., 2005).Additionally, a significant correlation between the total burden of CSVD and cognitive function has been reported.However, it is important to note that there is heterogeneity among patients in terms of the type and severity of cognitive impairment, which does not completely align with the degree of macroscopic lesions (Ter Telgte et al., 2018).Moreover, existing imaging features are insufficient to fully explain the diverse clinical manifestations observed in CSVD.This may be attributed to the presence of mixed pathological changes in aging and the burden of CSVD.Consequently, the predictive capability of individual imaging markers for dementia becomes less reliable (Hansen et al., 2015).Traditionally, cognitive impairment caused by both neurodegeneration and CSVD has been viewed as separate clinical and pathophysiological phenomena.However, emerging evidence suggests an overlap in the underlying pathways contributing to both CSVD and the neurodegenerative processes leading to cognitive decline (Wardlaw, 2013).In light of this, Bergkamp et al. recommend that future studies examining the progression of CSVD should incorporate markers that reflect neurodegenerative lesions (Bergkamp et al., 2019).Amide proton transfer (APT) imaging is a protein content-dependent and pH-dependent technique that serves as an effective imaging marker for detecting neurodegeneration (Zhang et al., 2020;Khan et al., 2020).The APT signal is primarily influenced by intracellular mobile protein content (66%) rather than the pH value (33%) (Qiu et al., 2021).Concentrations of mobile and semi-solid macromolecular proteins in brain tissue increase during normal brain aging, particularly in populations affected by neurodegenerative diseases.This leads to significantly higher APT values when abnormal proteins accumulate in the hippocampus compared to healthy aging populations.While brain temperature and pH values do not significantly differ between patients with vascular cognitive impairment (VCI) and normal elderly populations, APT values effectively represent variations in amide proton concentration.Previous study has demonstrated that APT technology can detect differences in protein content between patients with mild cognitive impairment (MCI) and normal elderly individuals in the bilateral hippocampus and amygdala.Consequently, novel APT technique can serve as an imaging marker for identifying abnormal protein concentrations in key brain regions among patients with CSVD (Guo et al., 2021).
Previous studies have provided confirmation of the substantial involvement of the frontal-limbic system circuit in patients with VCI (Nakao et al., 2010;Comte et al., 2018;Duering et al., 2013).Microstructural changes have been observed in the hemisphere, including the limbic system, in individuals with MCI and even among those with pre-MCI (SAMBUCHI et al., 2019).Building upon these findings, we propose a hypothesis that alterations in APT values in key brain regions of the frontal-limbic circuit might partially indicate VCI and its severity.Consequently, the aim of this study was to investigate whether APT can serve as a potential marker for reflecting VCI by examining changes in key brain regions of the frontal-limbic circuit across varying degrees of cognitive impairment.Additionally, we sought to explore the relationship between APT values and cognitive decline, thereby further confirming the intersection of the pathways involved in CSVD pathology and neurodegenerative processes.

Subjects
A community-based cross-sectional study was conducted in our hospital between May 2020 and June 2021, a total of 246 CSVD patients were recruited from the local community.The inclusion criteria for participants were: (a) age over 40; (b) right-handedness; (c) no contraindications for MRI; (d) no drug dependence; (e) no current medication usage that could impact the central nervous system; (f) normal vision and hearing.Exclusion criteria comprised: (a) neurological diseases other than CSVD (e.g., hydrocephalus, brain tumors, cerebrovascular malformations, etc.); (b) dysfunction of liver, kidney or other vital organs; (c) history of craniocerebral operation; (d) subjects with significant pathology incidentally identified during MR scan.All participants underwent standard clinical evaluations, including a detailed medical history, laboratory tests, and neuropsychological assessments.Additionally, they underwent a brain MRI examination within one week preceding the evaluations.Fig. 1 depicts a flowchart illustrating the process of inclusion and exclusion in the current study.
A total of 165 subjects were included in the study after hierarchical exclusion.The study received approval from the Medical Ethics Committee of Nanxishan Hospital of Guangxi Zhuang Autonomous Region (2020NXSYEC-006).Written informed consent was obtained from all subjects.

Neuropsychological assessment
The cognitive assessment of each participant was conducted using the Beijing version of the Montreal Cognitive Assessment (MoCA) by a skilled neuro-radiologist.The completion time for each assessment was approximately 10 min, and the maximum score achievable was points.Participants were categorized into one of three groups based on their MoCA scores: normal controls (27-30 points), mild cognitive impairment (MCI) (18-26 points), or vascular dementia (VaD) (10-17 points).Demographic information, including gender, age, and education, was collected for all participants.

MRI data acquisition
In this study, MRI scans were performed using a 3.0-T MR system (Ingenia 3.0CX; Philips Healthcare, Best, The Netherlands) and 32-channel head coils.The specific scan sequences utilized can be found in Table 1.For the diffusion weight imaging (DWI) sequence, b values of 0 and 1000 mm/s 2 were applied, with automatic calculation of the ADC map.Each participant underwent the aforementioned MRI scan protocol, which typically lasted around 30 min.Various three-dimensional (3D) MR parameters were employed, including 3D T1, 3D T2, 3D fluid-attenuated inversion recovery (FLAIR) sequences and 3D APT.Prior to other scans, a 3D T2 sequence was conducted to screen for any potential brain abnormalities.

Image processing
Image processing was performed using the "IntelliSpace Portal" version 8 (Philips Healthcare, Best, The Netherlands) on an independent workstation.A radiologist, following the definitions established in previous studies (Wardlaw, 2013;Fazekas et al., 1987;Wilson et al., 2019), identified LI, WMH, cerebral microbleeds (CMBs), and enlarged perivascular spaces (EPVS).The CSVD total burden was assessed using an ordinal scale from 0 to 4, with each of the following criteria assigned one point: the presence of LI, CMBs, moderate to severe EPVS (>20), and deep WMH ≥ 2 or periventricular WMH > 3 (Huijts et al., 2013).
The APT images were generated directly following the scanning scheme.Using the dedicated workstation "IntelliSpace Portal" version 8, the APTw images were co-registered and overlaid with the acquired FLAIR images.In cases where the sequences were not geometrically identical, coregistration was employed to ensure accurate alignment.Two experienced radiologists, Ronghua Mu and Xiqi Zhu, each with and 25 years of experience in neurological imaging, respectively, were blinded to the clinical data of each patient.This was done in order to avoid potential cognitive biases and perform measurements on bilateral oval regions of interest (ROI).These measurements were taken to determine the APT intensity values in four specific regions: frontal white matter (WM), hippocampus, amygdala, and thalamus.The radiologists reached a consensus in their measurements.The delineation of the ROI

Statistics
Statistical analyses were conducted using SPSS26.0(IBM Corp., Armonk, NY, USA).A bilateral P-value of less than 0.05 indicates statistical significance.The variance inflation factor (VIF) was used to check for multicollinearity among the explanatory variables.A VIF value lower than 5 indicates the absence of significant multicollinearity.We conducted an a priori power analysis to test the adequacy of our sample size to independent sample t-test using G*Power.We specified an alpha level of 0.05, a 1-b error probability of 0.80, and an effect size (f = 0.50) for an estimated medium effect.The results of the analysis suggested a total recommended sample size of 159.A post-hoc power analysis revealed that a sample size of 165 resulted in a reported power of 0.82 to detect a medium effect (f = 0.50) with an alpha level of 0.05.
The Kolmogorov-Smirov test and Bartlett's test were used on all measures to test for normality and homogeneity of variance.Counting data were expressed by cases, and the chi-square test was used to test the difference of variable between groups.Data with a normal distribution were expressed as mean ± standard deviation.The one way analysis of variance (ANOVA) were performed to compare the demographic and APT variables between groups, the Bonferroni correction was used for post-hoc tests.Subsequently, Pearson correlation analysis were calculated to explore initial correlations between APT values and MoCA variables.According to the criteria of Cohen, 0.1 ≤ correlation coefficient (r) < 0.3 indicate weak association, 0.3 ≤ r < 0.5 indicate moderate association, r ≥ 0.5 indicate strong association (Cohen, 1988).APT variables with P < 0.05 in Pearson correlation analysis were included and defined as the explanatory variables for regression analysis.Based on previous knowledge, age and education were input all regression model as potential important confounders (Mu et al., 2022).To examine the relationship between APT values and VCI grouping, three multiple logistic regression analysis models were constructed.Model 1 included the APT values of frontal white WM, hippocampus, amygdala, and thalamus as independent variables, while the VCI groups were used as the dependent variable in a multi-logistic regression analysis.The NC group was utilized as the reference item.In Model 2, education and age were adjusted based on Model 1. Finally, in Model 3, the CSVD imaging markers were adjusted based on Model 2.
In order to investigate the model stability and mediation effect, a hierarchical linear regression model was employed to examine the distinct associations between patients' demographic factors, CSVD imaging markers, APT values, and MoCA.Model I evaluated the unique associations between patients' demographic factors (age and education) with MoCA.Model II included four CSVD imaging markers (e.g., LI, WMH, EPVS, and CMB) to assess their unique associations with MoCA beyond the effects of patients' demographic factors.Lastly, Model III added the APT values of four encephalic regions to examine the unique associations between APT values and MoCA beyond the effects of patients' demographic factors and CSVD imaging markers.

The comparison of demographic data and APT values between different cognitive function groups
The APT values of frontal WM, hippocampus, amygdala, and thalamus were significantly different among different groups (P < 0.05).The comparison results of APT values shown VaD > MCI > NC.There were no significant difference of thalamus APT values between MCI and VaD groups.There were no significant difference of age, gender and education between groups (Table 2 and Fig. 5).

Correlation analysis of APT values in different encephalic region with MoCA
Pearson correlation analyses showed that MoCA score was moderately but significantly correlated with the APT values of frontal WM (r = − .364,P < 0.001), thalamus (r = 0.316, P < 0.001), and Amygdala (r = − .391,P < 0.001).The APT values of hippocampus was significantly and strong correlated with MoCA score (r = .552,P < 0.001) (Table 3 and Fig. 6).

Multiple logistic regression analysis of APT values in different encephalic region with cognitive function groups
The multiple logistic regression results for the MCI group are as follows (Table 4, Model 1): The regression coefficient value for the APT value of frontal WM was 1.657, showing a significance level of 0.05 (z = 2.002, p = 0.045), with an odds ratio (OR) of 5.245 (95%CI: 1.036 ~ 26.564).The regression coefficient value for the APT value of frontal the amygdala was 1.098, showing a significance level of 0.05 (z = 2.053, P = 0.040), with an OR of 2.998 (95%CI:1.051~ 8.552).The regression coefficient value for the APT value of the thalamus was 1.333, showing a significance level of 0.01 (z = 3.158, P = 0.002), with an OR of 3.793 (95%CI:1.658~ 8.674).These results indicate a significant positive effect of the APT values of frontal WM, amygdala, and thalamus on MCI grouping.The regression coefficient for the APT value of frontal the hippocampus was 0.642 (z = 1.040,P = 0.298), indicating that the APT value of hippocampus dose not affect MCI grouping.
The multiple logistic regression results for the VaD group are as follows (Table 4, Model 1): The regression coefficient value for the APT value of frontal WM was 3.444, showing a significance level of 0.01 (z = 3.283, P = 0.001), with an odds ratio (OR) of 31.300(95%CI: 4.007 ~ 244.502).The regression coefficient value for the APT value of the hippocampus was 4.191, showing a significance level of 0.01 (z = 5.133, P = 0.000), with an OR of 66.096 (95%CI: 13.339 ~ 327.516).The regression coefficient value for the APT value of the amygdala was 2.670, showing a significance level of 0.01 (z = 3.640, P = 0.000), with an OR of 14.444 (95%CI: 3.429 ~ 60.841).The regression coefficient value for the APT value of the thalamus was 1.732, showing a significance level of 0.01 (z = 3.037, P = 0.002), with an OR of 5.654 (95%CI: 1.848 ~ 17.295).These results indicate a significant positive effect of the APT values of frontal WM, amygdala, and thalamus    4, Model 2) in the MCI group, the effects of the amygdala on MCI grouping disappeared.The regression coefficient value for age was − 0.077, showing a significance level of 0.05 (z = − 2.384, P = 0.017), indicating that age has a significant negative effect on the group.Education had no impact on cognitive function grouping.In the VaD group, age and education showed no impact on cognitive function grouping.
After adjusting for age, education, and CSVD imaging markers (Table 4, Model 3), the effects of Frontal WM on cognitive function grouping disappeared in the MCI groups.In the MCI group, the regression coefficient value of WMH was 1.832, showing a significance level of 0.01 (z = 3.241, P = 0.001), and the OR value was 6.054 (95% CI: 2.037-17.986).The regression coefficient value of LI was 2.162, showing a significance level of 0.05 (z = 2.494, P = 0.013), and the OR value was 8.687 (95% CI: 1.589-47.505).These results indicate that WMH and LI had a significant positive effect on MCI grouping.In the VaD group, the regression coefficient value of WMH was 2.411, showing a significance level of 0.01 (z = 3.159, P = 0.002), and the OR value was 11.144 (95% CI: 2.498-49.723).The regression coefficient value of LI was 3.663, showing a significance level of 0.05 (z = 3.561, P = 0.000), and the OR value was ).These results indicate that WMH and LI had a significant positive effect on VaD grouping.

Hierarchical Multiple Linear regression analysis the association of APT values with MoCA score
In Model I, age and education were treated as independent variables, with MoCA as the dependent variable.The Table 5 illustrates that the model's R-squared value is 0.051, indicating that age and education can account for 5.1% of the variation in MoCA.The F-test for the model demonstrated statistical significance (F = 4.373, P < 0.05), implying that at least one factor in education exerts an influence on MoCA.The regression coefficient for education was 0.276 and was found to be statistically significant (t = 2.612, P = 0.010), suggesting a significant positive relationship between education and MoCA.
In Model II, after incorporating LI, WMH1, CMB, and EPVS into Model I, a significant change in the F-value was observed (P < 0.05), indicating that LI, WMH, CMB, and EPVS hold explanatory significance for the model.Additionally, the R-squared value increased from 0.051 to 0.252, suggesting that LI, WMH, CMB, and EPVS collectively account for 20.1% of the variance in MoCA scores.Specifically, the regression coefficient for LI was − 4.355 and was found to be statistically significant (t = − 4.444, P = 0.000), indicating a significant negative relationship with MoCA scores.Similarly, the regression coefficient for WMH was − 3.780 and was statistically significant (t = − 4.655, P = 0.000), signifying a negative effect on MoCA scores.On the other hand, the  In Model III, after incorporating the APT values of frontal WM, amygdala, thalamus, and hippocampus into Model 2, a significant change in the F-value was observed (P < 0.05), indicating that the APT values of frontal WM, amygdala, thalamus, and hippocampus have explanatory significance for the model.Additionally, the R-squared value increased from 0.252 to 0.529, suggesting that the APT values of amygdala, thalamus, frontal WM, and hippocampus account for 27.7% of the variance in MoCA scores.Specifically, the regression coefficient for the APT values of amygdala was − 2.277 and was statistically significant (t = − 2.835, P = 0.005), indicating a significant negative relationship with MoCA scores.Similarly, the regression coefficient for thalamus was − 1.092, but it did not reach statistical significance, suggesting that the APT values of thalamus does not affect MoCA scores.On the other hand, the regression coefficient for the APT values of frontal WM was − 2.416 and was statistically significant (t = − 2.301, P = 0.023), indicating a significant negative effect on MoCA scores.Likewise, the regression coefficient for the APT values of hippocampus was − 5.087 and was statistically significant (t = − 6.280, P = 0.000), signifying a significant negative relationship with MoCA scores.

Discussion
In this study, we investigated the relationship between APT values in four brain regions and VCI.The results are as follows: The APT values of frontal WM, hippocampus, amygdala, and thalamus showed significant differences among the different groups (p < 0.05).When comparing APT values, VaD > MCI > NC was observed.Moreover, the APT values of frontal WM, amygdala, and thalamus had a significant positive impact on the MCI grouping, while the APT values of frontal WM, hippocampus, amygdala, and thalamus had a significant positive effect on the VaD grouping.However, after adjusting for age and education in Model II, the effects of amygdala on the MCI grouping disappeared.Lastly, it was found that LI and WMH had a significant negative influence on the MoCA score.
This study revealed its initial finding that APT values were significantly higher in the VCI group compared to the NC group.Furthermore, the APT values displayed an increasing trend along with the severity of cognitive impairment.Chen et al. reported that patients with aMCI exhibited significantly higher APT values in the right hippocampus (0.99 ± 0.26% vs. 1.26 ± 0.28%, P = 0.006) compared to healthy controls (Chen et al., 2023).Guo et al. reported that the APTw values of the bilateral hippocampus and amygdala in the aMCI group demonstrated a significant increase compared to the control group.Specifically, the left hippocampus exhibited values of 1.01% versus 0.77% (P < 0.001), the right hippocampus showed values of 1.02% versus 0.74% (p < 0.001), the left amygdala displayed values of 0.98% versus 0.70% (P < 0.001), and the right amygdala demonstrated values of 0.94% versus 0.71% (P < 0.001) (Guo, 2021).Wang et al. reported that the APT values of bilateral hippocampus were significantly increased in Alzheimer disease (AD) patients compared with normal controls.The right hippocampus exhibited values of 1.24% ± 0.21% versus 0.83% ± 0.19%, while the left hippocampus showed values of 1.18% ± 0.18% versus 0.80% ± 0.17% (t = 3.039, 3.328, p = 0.004, 0.002, respectively) (Wang et al., 2015).Wang et al. discovered that the APT values of the bilateral hippocampus exhibited a negative correlation with mini-mental state examination.These findings suggest that hippocampal APT values not only differentiate AD patients from controls but also demonstrate a strong association with the severity of the disease.Consequently, monitoring hippocampal APT values could assist in assessing the condition of AD patients (Wang et al., 2015).CSVD is not only a manifestation of vascular damage characterized by microcirculation destruction, but also can lead to cortical atrophy (secondary neurodegenerative lesions) and abnormal accumulation of amyloid-beta plaques, and hasten the onset and progression of AD (Roseborough et al., 2017).The main pathological hallmarks of AD consist of extracellular amyloid deposition and intracellular neurofibrillary tangles.It is accompanied by abnormal protein deposition, which disrupts the integrity of the white matter and leads to cortical and subcortical structures' atrophy (Defrancesco, 2014).Mounting evidence indicates that CSVD-associated imaging abnormalities parallel neurodegeneration in the progression of the disease, leading to pathological and structural changes that can be detected years before the clinical onset of the disease (ievani et al., 2011;Teipel et al., 2015).Tau plays a key role in the pathological process of AD and forms the characteristic neurofibrillary tangles, which impair neuronal function (Sanford, 2017).The disease progresses, leading to the loss of neuronal tissue and resulting in brain atrophy (Cummings and Cole, 2002).Hippocampal atrophy occurs early in the disease, with involvement of the frontal lobe (Scahill et al., 2002).Typical patients with CSVD often experience deficits in processing speed and executive function, while memory function is relatively preserved.However, many studies have found that memory impairment among CSVD patients is also prevalent, which may be due to CSVD pathology frequently co-existing with Alzheimer's disease pathological changes, resulting in mixed cognitive impairment (Rizvi et al., 2018).The mechanism of CSVD-related cognitive impairment is still unresolved, and further studies are required to investigate the mechanism of neurodegenerative changes in CSVD.
The second results revealed differences in areas of the brain exhibiting abnormal APT values across different stages of vascular cognitive impairment.The APT values primarily influencing MCI grouping included the frontal white matter, amygdala, and thalamus, with the thalamus exhibiting predominant APT values.For the VaD grouping, the affected APT values encompassed the frontal white matter, hippocampus, amygdala, and thalamus, with the hippocampus exhibiting predominant APT values.This suggests the involvement of multiple brain areas in VCI, with variations in the affected brain regions across different stages of VCI.Additionally, deposition of abnormal proteins in the early stages of VCI (MCI) may occur.Firstly, frontal lobe and subcortical structure damage plays a crucial role during the early stages of VCI.
Prior studies have shown a major role of CSVD-related lesions in frontalsubcortical neuronal circuits in influencing processing speed impairments (Duering et al., 2014).processing speed deficits are among the earliest and most prominent cognitive manifestations in patients with CSVD (Duering et al., 2012).Neurodegeneration may partially involved in the frontal and frontal subcortical architecture damage.As a critical enzymatic player in amyloid precursor protein amyloidogenesis, increased b-secretase activity has been found in the temporal and frontal lobes of patients with Alzheimer's disease, which might manifest as white matter contraction and disruption of projection fiber connections in cortical areas, consequently compromising the integrity of frontal-subcortical circuits (Chalmers et al., 2005;Fukumoto et al., 2002).Consequently, projections gradually converge onto a limited number of neurons while preserving circuit segregation during the transmission of neural signals from cortical to subcortical structures (Cummings, 1993).Dysfunction of localized frontal cortical, sub-cortical networks and interruptions of long-range fibril-connections associated with the frontal lobe gives rise to symptoms associated with prefrontal function (Lim et al., 2015).Damage to the short association fibers connecting subcortical white matter regions with the long association fiber structures in the frontal, temporal, and parietal regions can result in nerve conduction abnormalities, resulting in varying degrees of cognitive dysfunction (Bozzali et al., 2012).The major pathways of long association fibers in the brain traverse through the deep white matter, and damage to these fiber tracts alters the functional abnormalities associated with various cognitive domains (Schmahmann et al., 2008).In particular, this results in subcortical circuit dysfunction and white matter integrity impairment, disrupting the synergistic effects of cortical and subcortical circuits, local functional networks within the subcortical regions, and frontal lobe fiber tracts, and giving rise to symptoms indicative of prefrontal dysfunction.(Lim, 2015;Zhang et al., 2022) Secondly, besides the frontal lobe, the limbic system also plays a crucial role in cognitive function regulation.It has been found that in individuals with MCI and even in the pre-MCI stages, significant microstructural changes have been observed within the brain hemisphere, particularly in the limbic system, and these alterations can be detected through imaging, neurobiology, and other modalities (SAMBUCHI et al., 2019).The limbic system comprises brain structures that play a role in emotion, learning, and memory.Specifically, the hippocampus and amygdala, which are components of the limbic system, are early affected in individuals with MCI and AD (Qin et al., 2020;Leandrou et al., 2020).The hippocampus and amygdala play central roles in the cortico-limbic circuit, facilitated by direct nerve fiber connections between them (Comte, 2018).Concurrent changes in APT signaling may occur in the hippocampus and amygdala during cognitive decline.Abnormal protein accumulation, such as extracellular amyloid plaques and intracellular neurofibrillary tangles, along with elevated intracellular protein levels, results in an elevation of APT signal intensity (Zhou et al., 2019).Zhang et al. discovered a discrepancy in hippocampal APT values between individuals with MCI and healthy older adults and proposed the utilization of APT signal as an imaging marker for individuals with MCI (Zhang et al., 2020).Guo et al. revealed significant differences in the content of bilateral hippocampus and amygdala between patients with MCI and normal elderly individuals (Guo, 2021).Wang et al. demonstrated a consistent increase in hippocampal APT values from normal elderly individuals to those with severe AD (Wang et al., 2015).Thirdly, this study identified a significant correlation between the APT values of the thalamus and the grouping of patients with MCI and VaD.Previous studies had also observed the thalamus involved in cognitive decline in CSVD patients (Zhou et al., 2016).There was evidence of fiber damage in the anterior thalamic radiations (ATR) associated with both VaD and mixed dementia (Duering et al., 2014).The ATR originate from the anterior thalamic nucleus and medial nucleus, then extend to the anterior cingulate cortex and dorsolateral prefrontal lobe.These radiations run through the forelimb of the internal capsule and are integral components of both the limbic system and the frontal-thalamic circuits (Mamiya et al., 2018).Finally, The presence of abnormal APT values in multiple brain regions was associated with both MCI and VaD grouping.These findings suggest that cognitive dysfunction caused by CSVD is not the result of impairment in a single functional brain region, but rather with multiple brain region regions abnormalities that share similar or related cognitive functions.Zhang et al. reported that in comparison to the normal control group, individuals with aMCI consistently displayed higher APT signals in their brains.More specifically, quantitative analysis revealed significantly elevated APT intensity values in nine out of twelve regions of interest in aMCI patients when compared to normal controls (Zhang et al., 2020).These regions included the hippocampus (P = 0.0407), white matter in the frontal lobe (P = 0.0012), thalamus (P = 0.0199), and putamen (P = 0.0062).Therefore, previous study has proposed that the abnormality in cognitive function is caused by widespread structural abnormalities in the brain, rather than a specific brain region (Penke et al., 2010).
The third finding of this study suggests that APT intensity abnormalities in multiple brain regions may only partially contribute to the pathophysiological mechanisms underlying VCI.For instance, multiple logistic regression analysis of the association of APT values with VCI grouping, revealed that age and CSVD imaging markers(LI and WMH), were also independent risk factors influencing VCI grouping.After incorporating age and CSVD imaging markers into the analysis, certain APT values in specific brain regions that initially differentiated MCI were no longer significant.Additionally, the APT values of the amygdala, thalamus, frontal WM, and hippocampus only collectively account for 27.7% of the interpretation power in relation to VCI.Moreover, the results of hierarchical Multiple Linear regression analysis indicate that the presence of LI and WMH is associated with 20.1% of the interpretation power regarding VCI.Prior research has indicated a notable elevation in the risk of cognitive decline and Parkinson's disease among patients with CSVD, particularly in the presence of WMH lesions or multiple LI (Defrancesco et al., 2014).Moreover, this risk is even higher when these LI or WMH lesions are located in the frontal lobe (Defrancesco, 2014).The study conducted by Ghaznawi et al. revealed an association between LI and WMH features that align with more pronounced small vessel changes, increased mortality risk, and unfavorable functional prognosis (Ghaznawi et al., 2019).Prior study has posited that the presence of asymptomatic LI is linked to brain atrophy, disruptions in white matter integrity, and consequent cognitive decline.This suggests that it may also contribute to the development of neurodegenerative diseases (Aribisala et al., 2013).Additionally, previous study proposed a significant association between WMH and LI with more pronounced small vessel changes, impairment of brain function, and increased mortality rates (Ghaznawi et al., 2019).LI may be an important predictor of future cognitive decline, the appearance of LI may be an early warning, indicating that CSVD has developed to a more severe period, and is also a result of the progression of CSVD (Benjamin et al., 2018).
On the clinical setting, protein and polypeptide dependent APT technique can partially reflect and predict impairment in vascular cognitive function.However, the cognitive impairment resulting from CSVD may be a complex process involving multiple risk factors and pathophysiological mechanisms, and cannot be fully explained by individual demographic data, vascular risk factors, CSVD image markers, and APT alone.A recent study on "mixed" dementia utilized Diffusion Tensor Imaging (DTI) to investigate the relationship between amyloid burden, white matter network parameters, and markers of CSVD such as WMH and LI.The findings revealed that the presence of amyloid burden showed no significant correlation with white matter network parameters.However, CSVD markers were found to be associated with both decreased network integration and increased network segregation.Moreover, these alterations in the white matter network were found to be linked with deficits in attention, language, visuospatial abilities, memory, and fronto-executive tasks (Kim et al., 2015).Another study demonstrated that VaD without abnormal amyloid imaging was more common than expected.Patients with VaD with and without abnormal amyloid imaging differed in clinical and MRI features, although there was considerable overlap (Lee et al., 2011).Thus, future study on VCI should incorporate additional biomarkers that reflect multiple pathophysiological mechanisms.Furthermore, extensive studies combine with internal and external validation is essential.
Several limitations need to be noted in this study.Firstly, in this study, only the mean APT of the same brain regions were included in the statistical analysis and were not fully analyzed as independent variables, so the difference between the two brain regions cannot be distinguished.Secondly, this study only examined the influence of education, age, imaging markers, and APT values on vascular cognitive impairment (VCI).However, it is evident from the study findings that these variables can only partially account for the changes observed in VCI.Moreover, there are additional variables that are likely to impact cognitive function, including vascular risk factors, gray-white matter volume, and the connections between brain structure and cognitive function through functional networks.Thirdly, cerebrospinal fluid neurodeformability biomarkers were not included as variables in this study.As a result, the hypothesis based on APT imaging that suggests a relationship between VCI and neurodeformability change cannot accurately correlate APT values with the pathophysiological mechanism of neurodegeneration.Fourthly it is important to note that this study had a small sample size and was conducted in a single center using a single scanner with single vendor.Therefore, the results of the study require internal and external validation from a wider range of sources and centers.Lastly, This study is dedicated to focus on the difference between APT values and different degrees of VCI and the association between APT and cognitive function in CSVD patients, without involving the diagnostic efficacy of APT in predicting different degrees of VCI.Undoubtedly, the sensitivity specificity of APT in clinical setting is an important feature that has to be considered, which needs to be verified in future studies.

Conclusion
VCI is characterized by the presence of abnormal APT intensity values in multiple brain regions, and the severity of these abnormalities is correlated with the degree of VCI.APT imaging can partially identifying and predicting the occurrence of VCI.

Declaration of Competing Interest
Kan Deng is employed by Philips (China) Investment Co., Ltd., Guangzhou Branch.The other authors have no conflicts of interest to declare.

R
.Mu et al.   is depicted in Figs.2-4.The size and shape of the ROI were determined based on the corresponding anatomical region.The normal APT intensity value was measured from the respective APT image.Mean APT intensity values within the different ROIs and anatomical regions were obtained for the left and right hemisphere separately, and the corresponding mean values for both hemispheres combined were calculated.

Fig. 1 .
Fig. 1.Flowchart shows inclusion and exclusion process of the present study.CSVD, cerebral small vessel disease.

Fig. 2 .
Fig. 2.An example of the definition of the regions of interest of the bilateral frontal white matter, thalamus, hippocampus and amygdala in the NC group population.Figures A-C represent two consecutive layers of the FLAIR image with APT overlay in the bilateral frontal white matter, thalamus, hippocampus, and amygdala, respectively.Figures a-c are their corresponding T1 anatomical structure diagrams.APT, amide proton transfer; FLAIR, fluid attenuated inversion recovery; NC, normal control.

Fig. 3 .
Fig. 3.An example of the definition of the regions of interest of the bilateral frontal white matter, thalamus, hippocampus and amygdala in the MCI group population.Figures D-F represent two consecutive layers of the FLAIR image with APT overlay in the bilateral frontal white matter, thalamus, hippocampus, and amygdala, respectively.Figures d-f are their corresponding T1 anatomical structure diagrams.APT, amide proton transfer; FLAIR, fluid attenuated inversion recovery; MCI, mild cognitive impairment.

Fig. 4 .
Fig. 4.An example of the definition of the regions of interest of the bilateral frontal white matter, thalamus, hippocampus and amygdala in the VaD group population.Figures G-I represent two consecutive layers of the FLAIR image with APT overlay in the bilateral frontal white matter, thalamus, hippocampus, and amygdala, respectively.Figures g-i are their corresponding T1 anatomical structure diagrams.APT, amide proton transfer; FLAIR, fluid attenuated inversion recovery; VaD, vascular dementia.

Table 1
Sequences of multi-parametric MRI.

Table 2
Comparison of demographic data and APT values between different cognitive function groups.
*The one way analysis of variance were performed to compare the demographic and APT variables between groups.APT, amide proton transfer.Frontal WM, Frontal whiter matter.SD, standard deviation.NC, normal control.MCI, mild cognitive impairment.VaD, vascular dementia.aindicate that there are no significant difference.***p < 0.001

Table 3
Correlations of of APT values in different encephalic region with MoCA.
Pearson correlation analysis were calculated to explore initial correlations between APT values and MoCA variables.APT, amide proton transfer.MoCA, Montreal Cognitive Scale.Frontal WM, Frontal whiter matter.r,correlation coefficient.onVaD grouping.After adjusting for age and education (Table

Table 4
Hierarchical Multiple Linear regression analysis of APT values in different encephalic region with cognitive function grouping.
Multiple logistic regression analysis models were constructed to examine the relationship between APT values and VCI grouping.Model I, no adjustment.Model II, adjusting for age and education.Model III, adjusting for age, education and CSVD imaging markers.APT, amide proton transfer.MCI, mild cognitive impairment.VaD, vascular dementia.CSVD, cerebral small vessel disease.Frontal WM, Frontal whiter matter.LI, lacunar infarction.WMH,white matter hyperintensity.EPVS, enlarged perivascular space.CMB, cerebral microbleeding.SE, standard error.OR, odds ratio.CI, confidence interval.VCI, vascular cognitive impairment.*p < 0.05 ** p < 0.01 regression coefficient for CMB was − 0.681, but it did not reach statistical significance, suggesting that CMB does not have a relationship with MoCA scores.Likewise, the regression coefficient for EPVS was 0.126, but it did not display statistical significance, suggesting that EPVS does not have a significant impact on MoCA scores.

Table 5
Multilevel regression analysis of variables affecting cognitive function.the independent variable are age and education.Model II, the independent variable are age, education and CSVD imaging markers.Model III, the independent variable are age, education, CSVD imaging markers and APT values.Frontal WM, Frontal whiter matter.LI, lacunar infarction.WMH, white matter hyperintensity.EPVS, enlarged perivascular space.CMB, cerebral microbleeding.CSVD, cerebral small vessel disease.MoCA, Montreal Cognitive Scale.