Associations Between Cerebral Morphology and Objective Memory Performance in Patients With Subjective Cognitive Decline

Background: Previous studies reported subjective cognitive decline (SCD) linked to underlying the biomarker of early AD pathology. There is a lack of studies on whether structural brain changes and how cerebral morphology in association with verbal and visuospatial episodic memory performance. Method: We combined VBM and SBM analysis for evaluating structural alterations in gray matter atrophy and cortical thickness among individuals with SCD, MCI, AD and normal controls and investigated the associations between cerebral morphology and AVLT-H, BVMT-R performance in the whole samples and SCD, respectively. Results: No signicant regional gray matter volume atrophy and cortical thickness differences were found between SCD and normal controls. We found that immediate recall in verbal learning performance was signicantly correlated with the loss of bilateral hippocampus volume in SCD group, furthermore, attention capacity (SDMT) was positively correlated with bilateral hippocampus atrophy and AVLT-IR in individuals with pre-MCI cognitive status. Conclusions: Our results found hippocampal atrophy links much closer with verbal memory than visuospatial memory and immediate recall in verbal learning test was probably the primary decit and maybe inuenced by the attention capacity.

LDR) in 2689 elderly residents [18]. Auditory Verbal Learning Test-HuaShan (AVLT-H) was adapted from the California Verbal Learning Test, including 12 words over 5 trials were derived from Guo et al [19]. The AVLT-H provides a multitude of verbal learning and memory indices, including immediate recall (AVLT-IR, the total value of three times free recalls), free recall over short (AVLT-SDR) and long delays (AVLT-LDR), long cued recall (AVLT-LCR), recognition memory (AVLT-REC). Short and long delay recall were separated by time intervals of other tests approximately 5 minutes and 20 minutes, respectively, in which the other nonverbal neuropsychological tests were administered. Brief Visuospatial Memory Test-Revised (BVMT-R) uses geometric designs and locations on a page as the items to be remembered [20], which follows relatively similar procedures as AVLT-H, including 3 times learning trials (BVMT-IR), free recalls after a 5 minute short time delay (BVMT-SDR) and a 20 minute long tome delay (BVMT-LDR), recognition (AVLT-REC) memory. Now, there is a lack of studies on whether and how the structural brain morphology is associated with verbal and visuospatial episodic memory performance.
Thus, in this study, we employed volume-based and surface-based analysis to detect reginal gray matter volume and cortical thickness morphology in individuals with controls, SCD, MCI and AD, and investigate whether the characteristics of brain areas in uenced the verbal and visuospatial memory performance. We hypothesized that the reginal brain volume atrophy and cortical thickness would be observed in SCD group compared with other groups. We also estimate that reginal volume atrophy and cortical thickness would be correlated with AVLT-H and BVMT-R performance in SCD.

Methods
Participants A total of 198 participants were recruited from August 2018 to February 2020, including 51 normal controls (NC), 65 subjective cognitive decline (SCD), 44 mild cognitive impairment (MCI), 38 Alzheimer's disease (AD) from neuropsychological test room, Department of Geriatrics, Shanghai Sixth People's hospital, Shanghai, China. All participants were recruited in a standardized clinical and physical examination, provided their disease and medical history, had relevant laboratory screening, received a comprehensive battery of neuropsychological assessments and cranial MRI scanning. Inclusion criteria of all were aged more than 50 years old and less than 80 years old, educated more than 6 years, nearly normal eyesight and hearing, with on history of alcoholism, drug abuse, head trauma or other neuropsychiatric diseases, such as depression and anxiety, which would affect performance, without apparent abnormalities in folic acid, vitamin B12, rapid plasma regain, thyroid function, treponema pallidum particle agglutination. All participants signed informed consent. The study was approved by the ethics committee of Shanghai sixth people's hospital.
The inclusion criteria for SCD was based on the Jessen's [5]criteria: (a) subjective decline in memory, rather than other domains of cognition; (b) onset of SCD with the last 5 years; (c) concerns (worries) associated with SCD; (d) feeling of worse performance than others of the same age group; (e) normal performance on standardized cognitive tests with age-, gender-, and education-adjusted and did not reach the criteria for MCI or dementia. Individuals with no cognitive decline complaints or any worries about memory status and performed on standardized neuropsychological tests within normal range were include as controls.
Diagnosis of MCI was made if they had any one of the following criteria [21]: (1) they had impaired scores but did not reach the criteria for dementia, de ned as >1SD below the age-corrected normative mean, on two of the six neuropsychological measures in the same cognitive domain (i.e., memory, language, or attention/executive function); (2) they had impaired scores but did not reach the criteria for dementia, de ned as >1SD below age-corrected normative mean, in each of the three cognitive domains. The criteria of AD were based on the National Institute of Aging(NIA) and Alzheimer's Association (AA) criteria [22].

Neuropsychological assessments
All participants received extensive measures of memory, language, attention, executive function. The tests included Auditory Verbal Learning Test-HuaShan (AVLT-H) [23] and Brief Visuospatial Memory Test-Revised (BVMT-R) for episodic memory measure, Boston Naming Test (BNT) and Animal Verbal Fluency Test (AFT) for language measure, Shape Trails Test Parts A and B [24] for executive function measure, Symbol Digit Modalities Test (SDMT) [25] for attention capacity, processing speed measure. Cognitive screening tests of Mini-Mental State Examination(MMSE), Montreal Cognitive Assessment Basic Version (MoCA-B) [26] were also included. In addition, all subjects were administered the Subjective cognitive decline-interview (SCD-I) [27], Hamilton Depression Rating Scale (HAMD), Hamilton Anxiety Rating Scale (HAMA), Activities of daily living (ADL), Functional Activities Questionnaire (FAQ) to assess functional capacity in several different clinically relevant areas.
VBM and SBM were conducted to determine differences in voxel volume and cortical thickness using the Computational Anatomy Toolbox (CAT12, http:// dbm.neuro.uni-jena.de/cat/) that is an extension toolbox of Statistical Parametric Mapping software (SPM12, http:// www. l.ion.ucl.ac.uk/spm/software/spm12) on the MATLAB R2017a platform. Processing and analysis steps were carried out according to the default parameters in accordance with protocol in the manual of CAT12(http://www. neuro.uni-jena.de/cat12/CAT12-Manual.pdf).

Voxel-based morphometry(VBM)
The T1 images were spatially registered to the Montreal Neurological Institute (MNI) template and segmented into three voxel classes: white matter(WM), grey matter (GM) and cerebrospinal uid (CSF). Then, segmented grey matter images were preserved to assess the amount of volume changes based on spatial registration, and the modulated normalized GM maps could re ect the regional tissue volumes differences. The total intracranial volume (TIV) was calculated and used as one covariate for all scans. We applied a 0.1 absolute masking threshold to the VBM data. The normalized grey matter images were smoothed using an 8mm full width at half maximum(FWHM) Gaussian lter.
Surface-based morphometry(SBM) SBM was used for the cortical thickness analysis, using CAT12 with default pipeline. There was performed in segment step using surface and thickness estimation analysis in the writing options. Estimation of central surface of hemispheres and cortical thickness were based on the projection-based method [28]. CT were extracted from central surface data based on the absolute mean curvature. Cortical thickness measures were resampled and smoothed using a 15 mm FWHM Gaussian kernel.

Statistical analysis
Demographic data analysis was performed with SPSS statistic 23 (IBM, New York, EUA). The Shapiro-Wilk test was used to investigate the distribution of data. ANOVA was applied on normally distributed variables to test groups differences, and Kruskal-Wallis H test was used if variables was not normal distribution. Pearson Chi-Square test was used to test the differences for gender comparison. Post Hoc comparisons using the Tukey HSD test or Games-Howell test were used to investigate the between-groups differences. Partial correlation analysis and Pearson correlation analysis were performed to calculate correlation between imaging and clinical scores, controlling for gender, age, education years and TIV.
We performed statistical analyses of MR data in the CAT12 applying ANOVA analysis for morphometric measures with TIV, age, gender as covariates for VBM (and for SBM, age and gender were used). we tested groups differences using a threshold of p < 0.05 voxel-wise level with a family-wise error (FWE) correction for multiple comparisons. Post hoc tests of signi cant ANOVA ndings were also performed to detect differences between every pair of groups. A small volume correction (SVC) tool available in SPM was used in the regions of interests (ROI), and only activated surviving regions in ANOVA analysis. When the FWE was too stringent, we set the statistical map at cluster level with p < 0.05 corrected. We also report exploratory results at p < 0.001 uncorrected in between-group comparisons.

Demographics and behavioral results
Demographic characteristics and behavioral results were summarized in Table 1  Voxel-based morphometry-gray matter volume VBM analysis revealed group effects were summarized in Tables 2, and illustrated in Figure 1. Six main clusters of GM were detected, including bilateral hippocampus, bilateral parahippocampal and bilateral fusiform.
Post-hoc analysis of every pair of groups comparison were illustrated in Table 3. Compared with the group of NC, GM atrophy in MCI were reported in bilateral hippocampus (P<0.05, FWE corrected, SVC-based, cluster-wise level). Six clusters of GM atrophy in AD were reported in bilateral fusiform, bilateral parahippocampal and bilateral hippocampus (all with P<0.05, FWE corrected, SVC-based, voxel-wise level) comparing to the NC. The results in comparison between SCD and AD were reported as the same pattern as the results in NC and AD, six clusters of GM atrophy were reported. Compared with SCD, GM loss in left parahippocampal (P<0.05, FWE corrected, SVC-based, cluster-wise level) was reported in MCI. Compared with MCI, AD showed signi cant GM loss in right hippocampus and parahippocampal (P<0.05, FWE corrected, SVC-based, voxel-wise level).
Surface-based morphometry-cortical thickness SBM analysis revealed group effects on the cortical thickness in right superior temporal, bilateral parahippocampal. The results were summarized in Table 4, and illustrated in Figure 2.
Post hoc analysis were carried out to detect the difference in every pair of groups, Table 5. Compared with NC, MCI participants showed a decreased cortical thickness in right superior temporal (P<0.001, uncorrected, voxel-wise level).
Compared with NC, AD showed decreased cortical thickness in right superior temporal (P<0.05, FWE corrected, cluster-wise level), right parahippocammpal (P<0.05, FWE corrected, cluster-wise level). Three clusters of decreased cortical thickness were detected in AD group when comparing with SCD, localized to right hippocampus, right superior temporal and left parahippocampal all with a cluster-level threshold of FWE corrected P<0.05. Compared with MCI, a decreased cortical thickness in right superior temporal (P<0.05, FWE corrected, cluster-wise level) was reported in AD.

Multiple regression analysis
Multiple regression models were established between GM volumes and AVLT-H, BVMT-R, SDMT, controlling for age, gender, TIV as covariates, see in Table 6 and

Correlation analysis
In SCD group, partial correlation analysis was performed to detect the relationships between reginal grey matter volumes atrophy or cortical thickness and AVLT-H sub-scores, in Figure 5. AVLT-IR was observed showing signi cant positive correlations with left hippocampus (r=0.284, p=0.026) and right hippocampus (r=0.316, p=0.013), respectively. However, we did not nd any of other AVHT sub-scores with signi cant correlations either in grey matter volumes or in cortical thickness areas. There was no signi cant correlation reported between any structural alternation and BVMT-R sub-sores. In NC and SCD groups, Pearson correlation analysis was performed between SDMT and gray matter volumes atrophy or cortical thickness. SDMT were reported positively correlated with left hippocampus (r=0.250, p=0.007) and right hippocampus (r=0.226 p=0.015). We did not nd SDMT with signi cant correlations in cortical thickness areas. Furthermore, we found SDMT was signi cantly positively correlated with AVLT-IR (r=0.466 p=0.000).

Discussion
In this study, we combined VBM and SBM analysis for evaluating structural alterations of gray matter atrophy and cortical thickness among individuals with SCD, MCI, AD and normal controls, disclosing the association of structural alternations with verbal and visuospatial memory performance in the whole samples. Furthermore, we investigated the relationships of verbal and visuospatial memory performance with structural alternations in SCD or SDC and NC. We found that immediate recall in verbal learning performance was correlated with the loss of bilateral hippocampus volume in SCD group. To the best of our knowledge, this is the rst study illustrated the correlated relationship between regional gray volume atrophy and immediate recall (AVLT-IR) in individuals with SCD, furthermore, attention capacity (SDMT) was positively correlated with bilateral hippocampus atrophy and AVLT-IR in individuals with pre-MCI cognitive status, indicating that hippocampal atrophy links much closer with verbal memory than visuospatial memory and immediate recall in verbal learning test was probably the primary de cit and maybe in uenced by the attention capacity.
Overall, the results obtained are in line with the previous studies [29][30][31][32][33], we observed the gray matter atrophy in bilateral hippocampus, bilateral parahippocampal, bilateral fusiform in AD groups compared with controls as well as SCD. Hippocampus and parahippocampal gyrus were widely known as the most two common structural MRI markers of progression to AD [34,35], the hippocampus showed progressive atrophy throughout the disease course. The results concord with previous study [36], bilateral hippocampus atrophy was found in SCD compared to AD. For the regional atrophy in fusiform, there was an evidence showed that fusiform was identi ed as the earliest change over the pathological stages [30], grey matter loss in fusiform has been shown to start in the medial temporal lobes and fusiform gyrus at least 3 years before subjects reach a diagnosis of AD, and then spread to the posterior temporal lobes and parietal lobes, and then eventually the frontal lobes [30,37], The severity of atrophy has been correlated with cognitive impairment severity. however, in our study we only found fusiform atrophy in subjects with AD. In addition, previous studies have identi ed that individuals with SCD showed structural gray matter volume reductions and cortical thinning in the bilateral entorhinal cortex, medial temporal, and frontotemporal regions compared to cognitively normal elders [14,38], however, there was still previous study have found no gray matter differences in SCD compared to controls [39]. we did not nd regional grey matter volume differences between SCD and NC in this study.
Consistent with the prior studies [13,35,40,41], differences in cortical thickness have been investigated and a pattern of regions associated with MCI and AD has been identi ed to include superior temporal, parahippocampal regions. Our ndings indicated thinner cortical thickness in the right superior temporal and right parahippocampal in AD when comparing with controls. The severity of cortical reduction has been correlated with cognitive impairment severity, patients with AD have the thinnest cortex in right superior temporal, patients with MCI and SCD exhibit intermediate measures, and normal controls show the greatest cortical thickness. Consistent with our ndings, previous studies [42]demonstrated cortical thickness differences of the temporal lobes associated with disease progression and the reduction in superior temporal cortex was important for discriminating controls, stable MCI and AD [43]. Surrounding areas including the parahippocampal gyrus are considered sites of very early formation of neuro brillary tangles in AD including early volume loss. our results have some differences from previous reports [13,44]: we did not nd signi cant cortical thickness differences between controls and SCD, which may be related to a small sample size. In addition, other factors might have also contributed to discrepant ndings between our study, e.g., the way of recruitment may in uence the results within a speci c sample.
Regression analyses showed that speci c brain regions were signi cantly correlated with the values of both memory measures. Previous studies have found that left hemisphere structures are implicated in verbal memory processing [45], while right hemisphere structures support non-verbal/spatial memory [46]. Our results demonstrated both memory measures are more related to bihemispheric regions especially in gray matter atrophy areas. While spatial memory more support right hemisphere structures in cortical thickness areas.
The wide temporal pole area was a highly strong predictor of immediate recall [47,48], in this study we found that BVMT-IR showing much wider structural correlations than AVLT-IR, which not only contains the medial temporal area but also cortical thickness in right superior temporal. Immediate recall is a task that requires learning ability, including semantic and encoding processing [47,48]. In contrast to the immediate memory, the delayed recall task was restricted to the medial temporal lobe (MTL) area [48]. The delayed recall test re ects a long-term memory component [47], which was used to access ability of memory consolidation [49]. The essential role of medial temporal structures for the consolidation of new information has been known for a long time [50]. The results were consistent with a number of previous studies that have demonstrated relationships between the MTL sub-regions in hippocampus, parahippocampal and delayed recall [47,[51][52][53]. But except for that, we have found the cortical thickness in right superior temporal was correlated with short and long delay recalls. Our study found that the REC and LCR were correlated with wide areas, including bilateral hippocampus, bilateral parahippocampal, bilateral fusiform, right superior temporal. As with delayed recall, recognition memory was also reported to be most strongly associated with medial temporal structures, which is closely related to retrieval ability. Poor performance on delay recalls and recognition meant retention de cit, causing by the problem of memory consolidation [49]. Cued recall has been shown to provide an estimate of all items that the participant has stored, which has been reported to be highly associated with temporal regions [54]. Speci cally, our results indicated that in addition to well-known hippocampus and parahippocampal, also inferior temporal (i.e. fusiform), superior temporal are also associated with REC and LCR.
Particularly, to determine the relationship between the structural alternations and verbal and visuospatial episodic memory in pre-MCI status, correlation analysis was conducted. AVLT-IR was observed showing signi cantly positive correlations with bilateral hippocampus. previous studies have found the SCD performed poorly on the episodic memory and correlated to left hippocampal volumes [18], and the amplitude of low-frequency uctuations of resting-state functional MR imaging in the right middle occipital gyrus in SCD correlated to the AVLT immediate recall performance [39]. In our study, we did not nd BVMT-R sub-scores signi cantly correlated with any structural alternation. Verbal episodic memory is strongly associated with medial temporal lobe function, in particular the hippocampus.
Verbal episodic memory has been seen as the earliest cognitive domain to be affected in Alzheimer's disease [55]. Previous studies have found that left hippocampal volume was the strongest predictor of verbal memory, whereas right hippocampal volume was the strongest predictor of spatial memory [46], and spatial learning memory test was more strongly associated with hippocampal volumes than verbal learning memory tests [56]. It seems that non-verbal memory measures may have higher diagnostic value, however, those studies were based on the individuals with MCI or AD. One recent study has demonstrated that in cognitive normal individuals only when amyloidosis and hippocampal atrophy are both present, signi cant associations with decline in visuospatial episodic memory could be detected, but when individuals without detectable amyloid levels, hippocampal atrophy predicted the rate of decline in tests of verbal memory only[57]. Combined with our results, hippocampal atrophy links much closer with verbal memory than visuospatial memory, to some extent. SDMT is a digit test that requires attention, perceptual speed, visual scanning speed, and tracking and is very sensitive to various neurological disorders. In our results, SDMT was positively correlated with bilateral hippocampus atrophy and AVLT-IR, respectively, in individuals with pre-MCI cognitive status, indicating that immediate recall of verbal learning was probably the primary de cit and maybe in uenced by the attention capacity.

Conclusion
In this study, we found the structural alterations among individuals with SCD, MCI, AD and normal controls using VBM and SBM and investigated the associations between cerebral morphology and AVLT-H, BVMT-R performance in the whole samples and SCD, respectively. We found hippocampal atrophy signi cantly associated with AVLT-IR, and SDMT was positively correlated with bilateral hippocampus atrophy and AVLT-IR. In our results, verbal episodic memory was probably the primary de cit in SCD, especially immediate recall performance, which possibly in uenced by the attention capacity.

Limitation
Our study had several limitations. Firstly, the results may be related to a small sample size. Considering that this is an ongoing research project, further analyses with more participants should be conducted. Secondly, we did not take information of biomarkers (such as Aβ, tau) into consideration. Future studies that combine PET or multimodal MR techniques be conducted. Thirdly, we only collected data on episodic memory performance in this study. It would be meaningful to perform more functional tests (i.e., executive function) in the future to estimate the brain-behavioral relationship.     VBM analysis of group effects on regional brain grey matter, controlling for age, gender and TIV. (P<0.05, FWE corrected, voxel-wise level).   Multiple regression models were established, cortical thickness positively associated with BVMT-R scores (e: BVMT-SDR, f: BVMT-LDR, g: BVMT-REC, h: BVMT-IR), controlling for age and gender. (P<0.05, FWE corrected, voxel-wise level).

Figure 6
In SCD group, AVLT-IR was positively correlated with bilateral hippocampus (a, b). In NC and SCD group, SDMT was signi cantly positively correlated with bilateral hippocampus (c, d) and AVLT-IR (e).