Graph Analysis of Functional Brain Topology Using Minimum Spanning Tree in Subjective Cognitive Decline

Abstract

Background:

Subjects with subjective cognitive decline (SCD) are proposed as a potential population to screen for Alzheimer’s disease (AD).

Objective:

Investigating brain topologies would help to mine the neuromechanisms of SCD and provide new insights into the pathogenesis of AD.

Methods:

Objectively cognitively unimpaired subjects from communities who underwent resting-state BOLD-fMRI and clinical assessments were included. The subjects were categorized into SCD and normal control (NC) groups according to whether they exhibited self-perceived cognitive decline and were worried about it. The minimum spanning tree (MST) of the functional brain network was calculated for each subject, based on which the efficiency and centrality of the brain network organization were explored. Hippocampal/parahippocampal volumes were also detected to reveal whether the early neurodegeneration of AD could be seen in SCD.

Results:

A total of 49 subjects in NC and 95 subjects in SCD group were included in this study. We found the efficiency and centrality of brain network organization, as well as the hippocampal/parahippocampal volume were preserved in SCD. Besides, SCD exhibited normal cognitions, including memory, language, and execution, but increased depressive and anxious levels. Interestingly, language and execution, instead of memory, showed a significant positive correlation with the maximum betweenness centrality of the functional brain organization and hippocampal/parahippocampal volume. Neither depressive nor anxious scales exhibited correlations with the brain functional topologies or hippocampal/parahippocampal volume.

Conclusion:

SCD exhibited preserved efficiency and centrality of brain organization. In clinical practice, language and execution as well as depression and anxiety should be paid attention in SCD.

Keywords

Alzheimer’s disease functional brain topologies hippocampal volume minimum spanning tree subjective cognitive decline

INTRODUCTION

Subjective cognitive decline (SCD) refers to cognitively unimpaired individuals who experience a self-perceived persistent decline in cognitive capacity compared with their previously normal status [1]. Subjects with SCD are proposed as a potential population to screen for preclinical AD, and the risk of dementia is increased in SCD with an annual conversion rate to dementia of approximately 1.77% to 2.33%, compared with 1.42% to 2.07% in populations without SCD [2, 3]. Previous studies also found that SCD exhibited neurodegeneration similar to those observed in mild cognitive impairment (MCI) and Alzheimer’s disease (AD), for example, brain atrophy and hypometabolism in the hippocampus, parahippocampus, entorhinal cortex, anterior and posterior cingulate gyrus [4 –6].

However, some researchers argued that SCD was common in the general population and could be regarded as a benign symptom that might not lead to severe sequelae [7]. SCD was more likely to be associated with depressive or anxious symptoms than risk factors for AD [8 –10], and a small proportion of subjects with SCD might develop cerebrovascular disease, Parkinson’s disease, or non-Alzheimer degenerative dementia [1, 11]. The evidence thus far suggests that whether SCD could be considered as the earliest stage of AD is controversial; thus, the underlying neuromechanism of SCD needs to be further and comprehensively explored.

Brain functional connectivity (FC) can be used to assess the integration of brain activity across distant brain regions regardless of their structural connection and open new avenues toward understanding of brain organization and function. Previous studies reported that SCD exhibit a similar brain connectome to MCI, characterized by initial increased FC in the anterior brain and decreased FC in the posterior brain, followed by the disappearance of increased FC in the later stage [12, 13]. SCD also showed altered global efficiency and local efficiency by using graph theory, but the results were not concordant [14 –16]. The reasons for these contradictory findings might be the methodological limitations of graph theory, such as sensitivity to alterations in connection strength or link density, which may give rise to alterations in network topology; thus, the comparison of brain networks can be problematic. The minimum spanning tree (MST) is a simplified network core with minimized cost, in which all the nodes in the original network are connected without forming cycles or loops; therefore, this approach is unaffected by the thresholding problem. Meanwhile, the MST topologies can be interpreted along the lines of conventional graph measures, and changes in MST topology reflect changes in the original network topology [17]. Previous studies using MST found that the default mode network of AD patients exhibits higher segregation, insomuch that the posterior cingulate/precuneus and hippocampus/parahippocampus are heavily isolated from rest of the network [18]. MST analysis also revealed a more star-like topology in MCI patients and a more line-like topology in the normal controls (NCs) and AD patients; moreover, more star-to-line topology changes were associated with worse cognitive performance [19]. Thus, MST properties might be applicable as neuroimaging markers of the very early stage of AD.

In the present study, we hypothesized that SCD reflects objective brain topological alterations and is an early symptom of AD. Thus, we 1) detected hippocampal and parahippocampal volumes, as hippocampus/parahippocampal atrophy is sensitive in identifying the neurodegenerative pathology of AD that may be present decades before initial clinical expression [20, 21]; 2) calculated the MST matrix of the functional brain network to explore the brain topological alterations in SCD; and 3) explored the correlations between MST topologies or brain hippocampal/parahippocampal volume and clinical assessments, including memory, language, execution, depression, and anxiety. This study might uncover the altered brain features of SCD and provide new insights into the role of SCD in the early screening of preclinical AD.

MATERIALS AND METHODS

Subjects

One hundred forty-four subjects were recruited from urban communities in Shanghai at the Department of Geriatrics, Shanghai Sixth People’s Hospital, Shanghai, China from January 2018 to December 2021. All subjects demonstrated objectively unimpaired cognition evidenced by less than two impaired scores (>1.0 standard deviations of the age-adjusted normative mean) out of six in three different cognitive domains, and the six neuropsychological scores are described in the following section [22]. The inclusion criteria were as follows: aged more than 60 and less than 80 years old; educated for more than 6 years; nearly normal eyesight and hearing; right handedness; and no history of alcoholism, drug abuse, head trauma, or other neuropsychiatric disease. The exclusion criteria were as follows: 1) current major psychiatric diagnoses such as severe depression and anxiety [Hamilton Depression Rating Scale (HAMD)>24, Hamilton Anxiety Rating Scale (HAMA)>29, or confirmed by psychiatrist]; 2) other neurological conditions or diseases which could cause cognitive decline other than AD spectrum disorders; 3) history of psychosis or congenital psychological growth retardation; 4) cognitive decline caused by traumatic brain injury; and 5) inability to complete the study protocol or contraindications for MRI.

Previous studies confirmed that subjects with SCD who were worried about their symptoms can have a higher risk of progression to dementia than those without such concerns; therefore, this factor has great value for the identification of at-risk patients [23, 24]. Thus, in the present study SCD was defined with the following criteria: self-perceived persistent decline in cognition and worried about it, while subjects who did not meet the two criteria were grouped as NC. This study was approved by the Institutional Ethics Reviewing Board of Huashan Hospital and Shanghai Sixth People’s Hospital, and carried out in accordance with the 1975 Helsinki Declaration and its later amendments or comparable ethical standards. All participants or their guardians provided informed consent prior to their inclusion in the study.

Clinical assessments

All subjects (total sample) underwent extensive cognitive assessments of memory, language, and executive function, including the 30-min delayed free recall and recognition of the auditory verbal learning test for memory function, animal fluency test and 30-item Boston naming test for language function and the shape trails test, parts A and B (time to completion) for executive function. The final memory, language, and execution scores were calculated by the mean value of the z-scored corresponding two items of that cognitive domain. The opposite number of execution scores was taken because the raw execution score was higher when the execution was worse. Thus, in the following analyses, higher scores for memory, language or execution indicated better corresponding functions. The Mini-Mental State Examination (MMSE) and Chinese version of Montreal Cognitive Assessment-Basic (MoCA-B) [25] were also used to test for global cognition.

Due to limitations in early recruitment, only 103 subjects (subsample) from the total sample underwent psychological assessment for anxiety and depression, including the HAMA for anxiety, HAMD and 30-item versions of the Inventory of Depressive Symptomatology, Self-Report (IDS-SR) for depression. The psychological conditions of the remaining subjects were assessed by an experienced psychiatrist (QH Guo) to rule out anxiety or depression. The protocol of subject recruitment and clinical assessments is shown in Fig. 1a.

Fig. 1

Flow diagram. The protocol of subject recruitment (a), segmentation of hippocampal/parahippocampal volume from the T1-weighted images (b), and MST matrix and global network properties calculated from BOLD-fMRI data (c). BOLD, blood oxygenation level dependent; ROI, region of interest; MEM, memory; LAN, language; EXE, execution.

Image acquisition

MRI scans were acquired using Siemens Prisma 3.0 T scanner (Siemens Healthineers, Erlangen, Germany). BOLD-fMRI scans were obtained by using a multislice single-shot gradient echo-planar imaging sequence: 488 volumes, repetition time = 800 ms, echo time = 37 ms, flip angle = 52°, field of view = 208×208 mm², no slice gap, slices = 72, and matrix size = 2×2×2 mm³. The subjects were instructed to close their eyes but remain awake during the scanning. The structural T1-weighted MRI sacns were acquired using a magnetization-prepared rapid gradient echo: repetition time = 3000 ms, echo time = 2.56 ms, flip angle = 7°, field of view = 256×256 mm², slice thickness = 0.8 mm, data matrix = 320×320, slices = 208, and voxel size = 0.8×0.8×0.8 mm³.

Image preprocessing

The T1-weighted images were preprocessed using CAT12 [26] and SPM12 (https://www.fil.ion.ucl.ac.uk/spm). All T1-weighted images were denoised, bias-corrected, affine-registered, and skull-stripped. Then, the images were segmented into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) and normalized into Montreal Neurological Institute space using the DARTEL approach. The volumes of hippocampus and parahippocampus were automatically extracted using the LPBA40 atlas [27] and normalized by the individual total intracranial volume to reduce the influence of different brain sizes (Fig. 1b).

Resting-state BOLD-fMRI images were preprocessed using DPABI [28]. For image stabilization and to allow the subjects to adapt to the environment, the first 10 volumes were discarded. The remaining functional sequences were corrected for timing differences and motion effects. Then, the individual T1-weighted images were coregistered to the mean functional image and segmented into GM, WM, and CSF by using a unified segmentation algorithm. The motion-corrected functional volumes were spatially normalized to Montreal Neurologic Institute space and resampled to 2 mm isotropic voxels by using the normalization parameters estimated during the unified segmentation. After smoothing with a 6 mm full-width at half-maximum Gaussian kernel, temporal bandpass filtering (0.01–0.08 Hz) was performed. Finally, nuisance signals (including a linear trend of the time course, Friston 24-head motion parameters, the mean global signal, and WM and CSF signals) were extracted and regressed out to reduce the effects of nonneuronal signals.

Functional brain MST network construction and global network properties

As shown in Fig. 1c, to construct the functional brain network, the cerebral cortex was first parcellated into 100 regions using Schaefer’s 100 parcellation [29], which parcellates the brain functionally and connectionally homogeneous. Then the mean time series of each region was obtained by averaging the time series of all voxels in that region. Pearson’s correlation coefficients between the time series of each possible pair of 100 regions were computed and only positive values were preserved to produce the FC matrix for each subject.

To avoid the bias introduced by different network densities, we explored brain topologies using the MST approach. Since we were interested in the strongest connections, the edges connecting all nodes with the highest strength but without forming loops were included according to Kruskal’s algorithm [30]. The resulting MST matrix included a fixed number of 100 nodes and 99 links.

MST measures provide information about the functional integration and segregation of the entire network. Here, 6 MST measures reflecting global brain network properties were calculated: 1) the degree of a tree is the highest nodal degree, which is the number of links for a given node in the network; 2) BC_max of a tree is the highest nodal betweenness centrality (BC), which is defined as the fraction of all shortest paths in the network that pass through a given node; 3) leaf Fraction is the leaf number divided by the maximum possible leaf number, where the leaf number is defined as the number of nodes that have only one connection; 4) path length is defined as the average shortest path between any two nodes in the MST matrix; 5) diameter characterizes the longest length of the shortest distance between any two nodes, normalized for the total number of connections; and 6) Tree hierarchy (T_h) is defined as $T_{h} = \frac{Leaf Fraction}{2 m {BC}_{max}}$ , where m is the number of links. For the above topologies, degree, BC_max and leaf fraction measure the centrality of the network organization; path length and diameter are measures of the efficiency of global network organization; and tree hierarchy characterizes a hypothesized optimal topology of efficient organization while preventing central nodes from being subjected to information overload.

Statistical analysis

Statistical analysis was performed in IBM SPSS Statistics 23. The differences between the SCD and NC groups in continuous clinical scores, MST topologies, and parahippocampal and hippocampal volumes were analyzed using the Mann-Whitney U test, as the data were abnormally distributed, and the sex difference was tested by using a two-tailed χ² test.

To explore whether subjective cognitive states could reflect objective brain features, we performed correlation analysis between neuroimaging features and clinical scores in the SCD group. In detail, the abnormally distributed features were box-cox transformed into normally distributed features, and the partial correlative coefficients between brain features and the cognitive scores in the total sample and between neuroimaging features and the psychological scores in the subsample were calculated while controlling for age, sex, and educational years. The statistical significance level was set at p < 0.05.

RESULTS

SCD exhibited normal memory, language, and execution, but increased depressive and anxious levels

A total of 144 subjects were included in this study, of which 49 subjects were in the NC group and 95 subjects were in the SCD group. The demographic and clinical characteristics are provided in Table 1. SCD group exhibited a slight decrease in MMSE scores compared to NC group (p = 0.044) but scores were still within the normal range. No significant differences were observed in age, sex, educational years, or memory, language, executive function, MoCA-B scores between the SCD and NC groups in the total sample.

Table 1

Demographics

Characteristic	Total Sample			SubSample
	NC	SCD	p	NC	SCD	p
n	49	95	/	36	67	/
Female, No. (%)	27(54%)	57(60%)	0.572^b	20 (55.6%)	38 (56.7%)	0.910^b
Age (y)	68.24±5.54	66.76±4.69	0.140^a	67.97±5.69	66.64±4.42	0.347^a
Education (y)	12.29±3.95	12.01±2.98	0.472^a	12.42±3.65	12.66±2.74	0.986^a
MMSE	28.35±1.61	27.884±1.57	0.044^a #	28.25±1.59	28.16±1.26	0.394^a
MoCA-B	25.57±2.79	24.74±3.22	0.226^a	25.33±2.65	25.04±2.83	0.845^a
Memory	0.09±0.70	–0.04±0.86	0.348^a	0.04±0.74	0.04±0.88	0.841^a
Language	–0.02±0.64	0.01±0.87	0.874^a	0.14±0.63	0.05±0.81	0.485^a
Execution	–0.10±0.98	0.05±0.83	0.558^a	–0.08±0.81	0.03±0.81	0.159^a
HAMD	/	/	/	4.11±4.60	5.81±4.09	0.014^a #
HAMA	/	/	/	5.33±6.07	7.72±8.19	0.021^a #
IDS	/	/	/	9.56±7.69	12.90±7.65	0.014^a #

^aMann-Whitney U test; ^b χ² test; ^#p < 0.05.

The subsample included 103 subjects, of which 36 were in the NC group and 67 were in the SCD group. We found that SCD group exhibited significant increased anxiety (HAMA, p = 0.021) and depression (HAMD and IDS, p = 0.014) compared to those in the NC group, as shown in Table 1.

No hippocampal/parahippocampal volume loss in SCD

The hippocampus/parahippocampus plays a central role in the pathology of AD and is useful in identifying nondemented individuals who satisfy neuropathologic criteria for AD, as well as pathologic stages of AD that may be present decades before initial clinical expression [20]. However, we did not find any statistically significant difference in the bilateral hippocampus/parahippocampus between SCD and NC groups (p > 0.05), as shown in Fig. 2.

Fig. 2

Hippocampal/parahippocampal volume. The y-axis is defined as the thousandth (‰) of the total intracranial volume.

Preserved functional brain global topologies in SCD

The MST occurrence matrix for SCD and NC groups is presented in Fig. 3a and 3b. The global network properties, including the centrality, measures by degree, BC_max and leaf fraction, efficiency measured by path length and diameter, and the balance of centrality and efficiency reflected by tree hierarchy, showed no significant difference between SCD and NC groups (p > 0.05), as shown in Fig. 3c.

Fig. 3

MST occurrence matrix and global network properties. Visualization of the MST occurrence matrix of subjects in NC (a) and SCD (b) groups. The width of connectional lines represents the occurrence of connections in MST across subjects, and only connections present in at least 25% of subjects are shown for clarity. The size and color of the nodes represent the average nodal degree. (c) Violin and box plots of global network properties calculated based on the MST matrix, including degree, BC_max, leaf fraction, path length, diameter, and tree hierarchy, are shown for the SCD and NC group.

Language and execution, instead of memory or psychological states, were correlated with hippocampal/parahippocampal volume

We then analyzed the correlations between the clinical scores and functional brain topologies or between the clinical scores and hippocampal/parahippocampal volume. Interestingly, we found that anxiety or depression did not exhibit any significant correlation with either global network properties or hippocampal/parahippocampal volume. However, instead of memory, language and execution were positively correlated hippocampal/parahippocampal volume. In detail, language was positively correlated with right hippocampal (r = 0.026, uncorrected p = 0.049, false discovery rate (FDR) corrected p = 0.065), right parahippocampal (r = 0.265, uncorrected p = 0.011, FDR corrected p = 0.044), and left parahippocampal volume (r = 0.222, uncorrected p = 0.034, FDR corrected p = 0.065). Execution was positively correlated with left hippocampal (r = 0.237, uncorrected p = 0.023, FDR corrected p = 0.072) and right hippocampal volumes (r = 0.213, uncorrected p = 0.041, FDR corrected p = 0.072). In addition, MoCA-B scores also exhibited a positive correlation with BC_max (r = 0.266, uncorrected p = 0.03, FDR corrected p = 0.18), as shown in Fig. 4. Although most of those coefficients were under the threshold level of significance when calculated among several other coefficients, it is worthy to focus on the most relevant coefficients [31].

Fig. 4

Correlation analysis. The correlations between psychological states (a) or cognitions (b) and brain imaging features in SCD group. The colors in the heatmap indicate the partial correlation coefficients. The red rectangle on the heatmap indicates that the r value is statistically significant (p < 0.05, uncorrected), and scatterplot for the significant correlation is plotted at the bottom.

DISCUSSION

In the present study, we thoroughly explored the functional brain topologies and hippocampal/parahippocampal volume, as well as their correlations with cognition and psychological states in SCD. Our study revealed that subjects with SCD exhibited preserved brain topology, and hippocampal/parahippocampal volume. In addition, language and execution, instead of memory, were positively correlated with hippocampal/parahippocampal volumes in SCD population.

The self-perceived experience of cognitive decline is phenomenologically complex and difficult to capture quantitatively. In the present study, subjects who met the criteria of self-perceived cognitive decline and worry about it were classified as SCD, and subjects who did not meet these two criteria were classified as controls. This categorization was based on the fact that in the Chinese cohort, subjects with SCD who worried about their symptoms exhibited distinct clinical characteristics and functional brain networks, as well as a higher risk of cognitive decline compared with NC and subjects with SCD who did not worry about their symptoms [32, 33]. In this study, we found that SCD exhibited normal memory, language, and executive function but slightly increased anxiety and depression, which is similar to a previous study showing that SCD were more likely to have higher anxiety and depression scores [34, 35]. A longitudinal observational study also reported that depressive symptoms predicted lower episodic memory and executive function [36]. Thus, we suggest that anxiety and depression should be taken seriously in subjects with SCD.

The hippocampus/parahippocampus is involved in the early course of AD and correlates well with episodic memory disorders, which is a progressive amnestic core in AD [37, 38]. Accordingly, it should be assumed that hippocampal atrophy is involved in SCD. However, we did not find hippocampal/parahippocampal atrophy but found positive correlations between hippocampal/parahippocampal volume and language or execution instead of memory in subjects with SCD. These findings were similar to the findings of a previous large cohort study in which the subiculum and presubiculum of the hippocampus displayed strong relations to executive dysfunction but not memory in nondementia subjects [39]. It is widely acknowledged that the first symptoms of AD are not limited to memory decline, and lay people may report memory decline when they actually experience decline in other cognitive domains such as executive or language function [40]. In addition, the hippocampus/parahippocampus might be involved in a wide range of cognitive functions, for example, language is not isolated from memory, and the same neuronal computations used by the hippocampus for memory function also subserve online language, relating incoming words to stored semantic knowledge [41, 42]. Together, we could conclude that the early cognitive decline, which is not restricted to memory, might reflect the subclinical alteration of hippocampal/parahippocampal volume in the very early stage.

The MST is a core backbone of the original network, which avoids the bias introduced by different densities, but inherits the topologies from the original network. In the present study, MST analysis revealed no alterations in brain organization in SCD, which was similar to the previous study [16]. However, we found that BC_max was positively correlated with MoCA-B in SCD group. Of note, BC_max is the highest nodal betweenness centrality in MST, which describes the importance of the most central node, and is a measure of central network organization. For a star-like tree, all leaves are connected to one central node, and BC_max reaches its maximum (BC_max = 1). Here, the positive correlation between MoCA-B and BC_max indicates that better cognitive performance was associated with a star-like topology, which is similar to the findings of a previous study showing that star-like topologies in MCI and AD were clinically associated with better cognitive performance [19].

There were some limitations in this study. First, this work was a cross sectional study as the follow-up evaluation of this cohort has not yet started. The second limitation was that the education level of the subjects was relatively high in our study. It is widely acknowledged that education contributes to cognitive reserve, so the results found in the present study might not be applicable in conditions with low education level. In addition, the Subjective Cognitive Decline Initiative working group proposed that SCD occurs at the preclinical stage of AD and may serve as a symptomatic indicator of preclinical AD. However, current knowledge is insufficient to comprehensively define the specific features of SCD in preclinical AD, and SCD by itself may never be sufficient to diagnose preclinical AD [40]. It should be noted that the definitions of SCD were also not concordant across previous studies, for example, SCD was defined in some studies as self-complaints alone [43], while other studies were defined as self-complaints and worries or other SCD-plus features [44, 45]. The heterogeneity of measures used for SCD inclusion suggests that one should exercise great caution in comparing findings across studies.

In conclusion, this study explored SCD using a multifunctional approach, and the results demonstrated that subjects with SCD exhibited preserved efficiency and centrality of functional brain organization, as well as hippocampal/parahippocampal volume. Unlike the episodic memory impairment in AD, language and executive function were correlated with the brain hippocampal/parahippocampal volume, and might be more sensitive to the early brain changes in cognitively unimpaired subjects. These findings provide a comprehensive description of the brain topologies of SCD and suggest that language and execution as well as depression and anxiety should be paid attention in SCD in clinical practice.

Footnotes

ACKNOWLEDGMENTS

The authors thank all staff in PET center, Huashan Hospital, Fudan University, and Department of Gerontology, Shanghai Jiao Tong University Affiliated Sixth People’s Hospital, Shanghai for their help in subject recruitment and data acquisition.

This work was supported by the startup fund of Huashan Hospital, Fudan University [837], Shanghai Municipal Health Commission fund [202040420], Shanghai Municipal Key Clinical Specialty [shslczdzk03402], Shanghai Municipal Science and Technology Major Project [2018SHZDZX01] and ZJLab, Clinical Research Plan of SHDC [SHDC2020CR2056B], National Key R&D Program of China [2016YFC1306305], National Natural Science Foundation of China [82171473, 82071962]; the Guangdong Provincial Key S&T Program [2018B030336001] and Science and Technology Commission of Shanghai Municipality [18411952100].

Authors’ disclosures available online ().

References

Jessen

, Amariglio

, Buckley

, van der Flier

, Han

, Molinuevo

, Rabin

, Rentz

, Rodriguez-Gomez

, Saykin

(2020) The characterisation of subjective cognitive decline. Lancet Neurol 19, 271–278.

Slot

, Sikkes

, Berkhof

, Brodaty

, Buckley

, Cavedo

, Dardiotis

, Guillo-Benarous

, Hampel

, Kochan

(2019) Subjective cognitive decline and rates of incident Alzheimer’s disease and non–Alzheimer’s disease dementia. Alzheimers Dement 15, 465–476.

Mitchell

, Beaumont

, Ferguson

, Yadegarfar

, Stubbs

(2014) Risk of dementia and mild cognitive impairment in older people with subjective memory complaints: Meta-analysis. Acta Psychiatr Scand 130, 439–451.

Schultz

, Oh

, Koscik

, Dowling

, Gallagher

, Carlsson

, Bendlin

, LaRue

, Hermann

, Rowley

(2015) Subjective memory complaints, cortical thinning, and cognitive dysfunction in middle-age adults at risk of AD. Alzheimers Dement (Amst) 1, 33–40.

Verfaillie

, Tijms

, Versteeg

, Benedictus

, Bouwman

, Scheltens

, Barkhof

, Vrenken

, van der Flier

(2016) Thinner temporal and parietal cortex is related to incident clinical progression to dementia in patients with subjective cognitive decline. Alzheimers Dement (Amst) 5, 43–52.

Scheef

, Spottke

, Daerr

, Joe

, Striepens

, Kölsch

, Popp

, Daamen

, Gorris

, Heneka

(2012) Glucose metabolism, gray matter structure, and memory decline in subjective memory impairment. Neurology 79, 1332–1339.

Hessen

, Eckerström

, Nordlund

, Selseth Almdahl

, St ålhammar

, Bjerke

, Eckerström

, Göthlin

, Fladby

, Reinvang

, Wallin

(2017) Subjective cognitiveimpairment is a predominantly benign condition in memory clinicpatients followed for 6 years: The Gothenburg-MCI Study. Dement Geriatr Cogn Dis Extra 7, 1–14 Oslo.

Eckerström

, Berg

, Nordlund

, Rolstad

, Sacuiu

, Wallin

(2016) High prevalence of stress and low prevalence of Alzheimer disease CSF biomarkers in a clinical sample with subjective cognitive impairment. Dement Geriatr Cogn Disord 42, 93–105.

Caselli

, Chen

, Locke

, Lee

, Roontiva

, Bandy

, Fleisher

, Reiman

(2014) Subjective cognitive decline: Self and informant comparisons. Alzheimers Dement 10, 93–98.

10.

Dubois

, Villain

, Frisoni

, Rabinovici

, Sabbagh

, Cappa

, Bejanin

, Bombois

, Epelbaum

, Teichmann

, Habert

M-O

, Nordberg

, Blennow

, Galasko

, Stern

, Rowe

, Salloway

, Schneider

, Cummings

, Feldman

(2021) Clinical diagnosis of Alzheimer’s disease: Recommendations of the International Working Group. Lancet Neurol 20, 484–496.

11.

Pijnenburg

, Gillissen

, Jonker

, Scheltens

(2004) Initial complaints in frontotemporal lobar degeneration. Dement Geriatr Cogn Disord 17, 302–306.

12.

Garcia-Ptacek

, Eriksdotter

, Jelic

, Porta-Etessam

, Kareholt

, Palomo

(2016) Subjective cognitive impairment: Towards early identification of Alzheimer disease. Neurologia 31, 562–571.

13.

Lazarou

, Nikolopoulos

, Dimitriadis

, Kompatsiaris

, Spilioti

, Tsolaki

(2019) Is brain connectome research the future frontier for subjective cognitive decline? A systematic review. Clinical Neurophysiology 130, 1762–1780.

14.

Chen

, Sheng

, Luo

, Qin

, Ye

, Zhao

, Xu

, Bai

, Alzheimers Disease Neuroimaging Initiative (2020) The compensatory phenomenon of the functional connectome related to pathological biomarkers in individuals with subjective cognitive decline. Transl Neurodegener 9, 21.

15.

, Li

, Tao

, Xie

, Gao

, Yue

, Wang

(2020) Effective and accurate diagnosis of subjective cognitive decline based on functional connection and graph theory view. Front Neurosci 14, 577887.

16.

Chen

, Lu

, Zhang

, Sun

, Chen

, Li

, Zhang

, Qing

, Zhang

(2021) Alterations in dynamic functional connectivity in individuals with subjective cognitive decline. Front Aging Neurosci 13, 646017.

17.

Tewarie

, van Dellen

, Hillebrand

, Stam

(2015) The minimum spanning tree: An unbiased method for brain network analysis. Neuroimage 104, 177–188.

18.

Ciftçi

(2011) Minimum spanning tree reflects the alterations of the default mode network during Alzheimer’s disease. Ann Biomed Eng 39, 1493–1504.

19.

Wang

, Miao

, Niu

, Cao

, Li

, Yan

, Guo

, Yan

, Wu

, Xiang

, Alzheimer’s Disease Neuroimaging Initiative (2018) Abnormal Functional brain networks in mild cognitive impairment and Alzheimer’s disease: A minimum spanning tree analysis. J Alzheimers Dis 65, 1093–1107.

20.

Gosche

, Mortimer

, Smith

, Markesbery

, Snowdon

(2002) Hippocampal volume as an index of Alzheimer neuropathology Findings from the Nun Study. Neurology 58, 1476–1482.

21.

Xie

, Wisse

LEM

, Das

, Vergnet

, Dong

, Ittyerah

, de Flores

, Yushkevich

, Wolk

, Alzheimer’s Disease Neuroimaging Initiative (2020) Longitudinal atrophy in early Braak regions in preclinical Alzheimer’s disease. Hum Brain Mapp 41, 4704–4717.

22.

Huang

, Chen

, Liu

, Guo

(2020) A conceptual framework for research on cognitive impairment with no dementia in memory clinic. Curr Alzheimer Res 17, 517–525.

23.

Mendonca

, Alves

, Bugalho

(2016) From subjective cognitive complaints to dementia: Who is at risk?: A systematic review. Am J Alzheimers Dis Other Demen 31, 105–114.

24.

Jessen

, Wiese

, Bickel

, Eiffländer-Gorfer

, Fuchs

, Kaduszkiewicz

, Köhler

, Luck

, Mösch

, Pentzek

, Riedel-Heller

, Wagner

, Weyerer

, Maier

, van den Bussche

, for the AgeCoDe Study Group (2011) Prediction of dementia in primary care patients. PLoS One 6, e16852.

25.

Chen

, Xu

, Chu

, Ding

, Liang

, Nasreddine

, Dong

, Hong

, Zhao

, Guo

(2016) Basic Validation of the Chinese version of Montreal Cognitive Assessment for screening mild cognitive impairment. J Am Geriatr Soc 64, e285–e290.

26.

Gaser

, Dahnke

(2016) CAT-a computational anatomy toolbox for the analysis of structural MRI data. HBM 2016, 336–348.

27.

Shattuck

, Mirza

, Adisetiyo

, Hojatkashani

, Salamon

, Narr

, Poldrack

, Bilder

, Toga

(2008) Construction of a 3D probabilistic atlas of human cortical structures. Neuroimage 39, 1064–1080.

28.

Yan

C-G

, Wang

X-D

, Zuo

X-N

, Zang

Y-F

(2016) DPABI: Data processing & analysis for (resting-state) brain imaging. Neuroinformatics 14, 339–351.

29.

Schaefer

, Kong

, Gordon

, Laumann

, Zuo

X-N

, Holmes

, Eickhoff

, Yeo

BTT

(2017) Local-global parcellation of the human cerebral cortex from intrinsic functional connectivity MRI. Cereb Cortex 28, 3095–3114.

30.

Kruskal

(1956) On the shortest spanning subtree of a graph and the traveling salesman problem. Proc Am Math Soc 7, 48–50.

31.

Curtin

, Schulz

(1998) Multiple correlations and Bonferroni’s correction. Biol Psychiatry 44, 775–777.

32.

Chen

, Yang

, Du

, Hu

, Han

(2019) Clinical characteristics in subjective cognitive decline with and without worry: Baseline investigation of the SILCODE Study. J Alzheimers Dis 72, 443–454.

33.

Liu

, Li

, Jiang

, Du

, Wang

, Sheng

, Jiang

, Han

(2021) Differences in functional brain networks between subjective cognitive decline with and without worry groups: A graph theory study from SILCODE. J Alzheimers Dis 84, 1279–1289.

34.

Jenkins

, Tree

, Thornton

, Tales

(2019) Subjective cognitive impairment in 55-65-year-old adults is associated with negative affective symptoms, neuroticism, and poor quality of life. J Alzheimers Dis 67, 1367–1378.

35.

Perrotin

, La Joie

, de La Sayette

, Barre

, Mezenge

, Mutlu

, Guilloteau

, Egret

, Eustache

, Chetelat

(2017) Subjective cognitive decline in cognitively normal elders from the community or from a memory clinic: Differential affective and imaging correlates. Alzheimers Dement 13, 550–560.

36.

Ahn

, Mathiason

, Yu

(2021) Longitudinal cognitive profiles by anxiety and depressive symptoms in american older adults with subjective cognitive decline. J Nurs Scholarsh 53, 698–708.

37.

Dubois

, Feldman

, Jacova

, DeKosky

, Barberger-Gateau

, Cummings

, Delacourte

, Galasko

, Gauthier

, Jicha

, Meguro

, O’Brien

, Pasquier

, Robert

, Rossor

, Salloway

, Stern

, Visser

, Scheltens

(2007) Research criteria for the diagnosis of Alzheimer’s disease: Revising the NINCDS–ADRDA criteria. Lancet Neurol 6, 734–746.

38.

Köhncke

, Düzel

, Sander

, Lindenberger

, Kühn

, Brandmaier

(2020) Hippocampal and parahippocampal gray matter structural integrity assessed by multimodal imaging is associated with episodic memory in old age. Cereb Cortex 31, 1464–1477.

39.

Evans

, Adams

HHH

, Licher

, Wolters

, van der Lugt

, Ikram

, O’Sullivan

, Vernooij

, Ikram

(2018) Subregional volumes of the hippocampus in relation to cognitive function and risk of dementia. Neuroimage 178, 129–135.

40.

Jessen

, Amariglio

, van Boxtel

, Breteler

, Ceccaldi

, Chételat

, Dubois

, Dufouil

, Ellis

, van der Flier

, Glodzik

, van Harten

, de Leon

, McHugh

, Mielke

, Molinuevo

, Mosconi

, Osorio

, Perrotin

, Petersen

, Rabin

, Rami

, Reisberg

, Rentz

, Sachdev

, de la Sayette

, Saykin

, Scheltens

, Shulman

, Slavin

, Sperling

, Stewart

, Uspenskaya

, Vellas

, Visser

, Wagner

(2014) A conceptual framework for research on subjective cognitive decline in preclinical Alzheimer’s disease. Alzheimers Dement 10, 844–852.

41.

Piai

, Anderson

, Lin

, Dewar

, Parvizi

, Dronkers

, Knight

(2016) Direct brain recordings reveal hippocampal rhythm underpinnings of language processing. Proc Natl Acad Sci U S A 113, 11366–11371.

42.

, Cheyne

, Sun

, Johnson

(2020) Theta oscillations support the interface between language and memory. Neuroimage 215, 116782.

43.

Lauriola

, Esposito

, Delli Pizzi

, de Zambotti

, Londrillo

, Kramer

, Rabinovici

, Tartaro

(2017) Sleep changes without medial temporal lobe or brain cortical changes in community-dwelling individuals with subjective cognitive decline. Alzheimers Dement 13, 783–791.

44.

Sun

, Wang

, Dong

, Lu

, Scheininger

, Ewers

, Jessen

, Zuo

, Han

(2019) Anxiety correlates with cortical surface area in subjective cognitive decline: APOE epsilon 4 carriers versus APOE epsilon 4 non-carriers. Alzheimers Res Ther 11, 50.

45.

Chen

, Li

, Hao

, Wang

, Cai

, Han

(2021) Blood NCAPH2 methylation is associated with hippocampal volume in subjective cognitive decline with apolipoprotein E epsilon 4 non-carriers. Front Aging Neurosci 13, 632382.