Subjects
The sample included 62 SCD, 75 MCI patients and 70 HC matched with SCD and MCI patients by age, gender and years of education. Table 1 summarized their demographic data and other relevant characteristics. The sample was recruited from the First Affiliated Hospital of Guangxi University of Chinese Medicine and from the community and elderly activities center of Nanning city from April 2016 to January 2018. The patients were enrolled in the study if they fulfilled the SCD and MCI diagnostic criteria. Other inclusion criteria for patients were: (1) age between 55 and 75; (2) right-handed; (3) daily life ability and social occupation were not affected; (4) memory impairment has not fulfilled the diagnostic criteria for dementia. Other exclusion criteria for patients were: (1) other diseases that were terminal, severe and unstable; (2) severe hearing and visual impairment; (3) dementia, cerebral infarction, physical/neurological disorder that could cause brain dysfunction; (4) drugs that may cause cognitive changes and organ failure were administered before inclusion; (5) fMRI examination contraindications.
Table 1
Demographic and neuropsychological data for each group.
|
HC (n = 66)
|
SCD (n = 55)
|
MCI (n = 65)
|
p-value
|
Age (years)
|
64.68 ± 5.78
|
64.47 ± 5.41
|
64.92 ± 6.68
|
0.650
|
Gender (males / females)
|
66(24 / 42)
|
55(18 / 37)
|
65(18 / 47)
|
0.567
|
Education
|
11.76 ± 3.02
|
12.05 ± 3.08
|
10.66 ± 2.55
|
0.242
|
MMSE
|
29.11 ± 0.75 c
|
28.85 ± 0.85 b
|
25.92 ± 1.05 b c
|
10− 33*
|
MOCA
|
26.12 ± 2.06 a c
|
24.93 ± 2.26 a b
|
21.62 ± 2.73 b c
|
10− 16*
|
GDepS
|
4.17 ± 2.27 c
|
4.60 ± 2.61 b
|
5.57 ± 2.10 b c
|
0.005*
|
Age, Education, MMSE, MOCA and GDepS were tested with ANOVA or Kruskal-Wallis test and two-sample t-test or Mann-Whitney test. Gender was tested with chi-squared test. |
* : significantly different among 3 groups (p < 0.05, ANOVA)
a : significantly different between HC and SCD (p < 0.05, two-sample t-test)
b : significantly different between SCD and MCI (p < 0.05, two-sample t-test)
c : significantly different between HC and MCI (p < 0.05, two-sample t-test)
|
Neuropsychological assessment
To assess the general cognitive and functional status, the following set of screening questionnaires were used: Mini Mental State Examination (MMSE)[11], Montreal Cognitive Assessment (MoCA)[12], Clinical Dementia Rating (CDR)[13], Geriatric Depression Scale (GDepS)[14] and Global Deterioration Scale (GDS). After the initial screening, all subjects underwent an exhaustive neuropsychological assessment including: Auditory Verbal Learning Test (AVLT)[15], Animal Fluency Test (AFT)[16], 30-item Boston Naming Test (BNT)[17] and Trail Making Test (TMT/STT)[18]. All neuropsychological scale assessments were completed by two neurologists with more than 5 years clinical experience. MCI patients were diagnosed according to the criteria established by[19]: (1) The main complaint is memory impairment and has the informed person to confirm. (2) Other cognitive functions were relatively intact or slightly impaired. (3) The ability of daily living was not affected. (4) The diagnostic criteria of dementia were not met. (5) Other systemic diseases which could cause the decline of brain function were excluded. (6) The score of MMSE was 24–27, the CDR was 0.5 and the GDS was 2–3.
SCD subjects were diagnosed based on the study of[20] and[21]. Through the six tests in three cognitive domains (memory, language and attention/execution functions), subjects who had diagnosed with MCI were excluded: abnormalities on two measures in the same cognitive domain, defined as > 1 standard deviation (SD), or each of the three cognitive domains had an impaired score (defined as > 1 SD). Then, according to the concept of SCD, the definition was determined by the following related questions, do you think your memory is worse than before? The answer may be: (1) No, I don’t think there is a problem with my memory. (2) Yes, but I’m not worried about this issue. (3) Yes, the appearance of this problem is worrying me. The individual with a declining memory or not we would abbreviate it as “having complaint” individuals and “no complaint” individuals in the following. The individuals who had complaint would be regarded as the SCD group. The “no complaints” individuals whose cognitive impairment passed neuropsychological test was directly included in the HC group. The flowchart of diagnostic step was shown in Figure S1of the Supplementary materials.
MRI acquisition
The imaging data were scanned using a Magnetom Verio 3.0T MRI scanner. The structural MRI data was collected in a sagittal orientation using magnetization-prepared rapid gradient echo sequence with following imaging parameters: TR / TE = 1900 ms / 2.22 ms, FOV = 250 mm × 250 mm, slice thickness = 1 mm, matrix size = 256 × 256, flip angle = 9°, slices = 176. The resting-state functional MRI data was collected in an axial orientation using multi-slice gradient echo planar imaging sequence with following imaging parameters: TR / TE = 2000 ms / 30 ms, FOV = 240 mm × 240 mm, slice thickness = 5 mm, matrix size = 64 × 64, flip angle = 90°, slices = 31, volumes = 180. The day before scanning, subjects were asked to ensure the sleep quality, did not drink alcohol or take drugs that might affect the nervous system. During scanning, subjects were instructed to not engage in any particular cognitive or motor activity, keep their eyes closed, relax and not to fall asleep. Foam padding and headphones were used to limit head movement and reduce scanner noise.
MRI preprocessing
The T1-weighted and resting-state fMRI data were preprocessed using the Statistical Parametric Mapping (SPM8, https://www.fil.ion.ucl.ac.uk/spm) and the Data Processing Assistant for Resting-State fMRI (DPABI, http://rfmri.org/dpabi)[22]. The preprocessing pipeline was as follows. The first 5 volumes were removed, remaining 175 volumes, to avoid T1 equilibration effect. Slice timing correction was used with the middle slice as the reference. Head motion correction was then used to obtain the parameters. The T1 images were segmented and co-registered with functional images using DARTEL[23]. Then, the images were spatially normalized into Montreal Neurological Institute space, resliced to 3 × 3 × 3 mm3 voxels, and smoothed with an FWHM of 4 mm. We further reduced biophysical and other noise by removing linear trend, regressing nuisance signals (head motion parameters, cerebrospinal fluid, white matter and global signals). Before estimating dFC, the temporal band-pass filtering (0.01–0.1 Hz) was performed to remove the effects of very low-frequency drift and high-frequency noise. To define the region of interest (ROI) in subsequent analysis, we employed a 160-ROI functional atlas[24] with four more ROIs located in the bilateral amygdala and para-hippocampal, resulting in 164 ROIs[25]. Then, we extracted the time series of each ROI by averaging the time courses of all voxels within the ROI. Finally, we divided the whole brain into seven networks: cerebellum, opercular, default, parietal, occipital, sensorimotor and addition networks.
Estimation of dynamic fMRI states
fALFF was calculated by the ratio between the sum of Fourier amplitudes within a specific low-frequency range (0.01–0.1 Hz) and the sum of Fourier amplitudes across the entire frequency range (0–0.2 Hz) and it was used to measure the pattern of local neural activity. The dynamic patterns in fALFF were characterized by using the sliding window approach[5], which sliced ROI time courses into several short data segments with a 50 s rectangular window and estimated a dfALFF matrix for each segment. Next, k-means clustering was used to group the dfALFF matrices into a limited number of clusters, which are referred to as “states”. After the dfALFF states were identified, the occurrence frequency of each state for each participant was obtained by calculating the percentage of the corresponding state among all time points. The dynamic patterns in FC were also characterized by using the sliding window approach with the same parameters as the estimation of dfALFF. The occurrence frequency of each dFC state for each participant was also obtained by calculating the percentage of the corresponding state among all time points. After identifying recurrent states of dALFF and dFC, the co-occurrence frequency between each pair of dfALFF state and dFC state was obtained by calculating the percentage of the co-occurrence of this pair of states among all time points for each participant. The whole framework was illustrated in Fig. 1. More details about the estimation of dfALFF and dFC states and their co-occurrence can be found in Appendix B of the Supplementary materials.
Statistical analyses
Sociodemographic, clinical and behavioral variables were tested for normality using the Shapiro-Wilk test. Age, Education, MMSE, MOCA differences between three groups were tested with the Analysis of Variance (ANOVA) or Kruskal-Wallis test. AVLT, BNT, AFT and STT differences between two groups were tested with the two-sample t-test or Mann-Whitney test. Gender differences between groups was tested with the chi-squared test. Furthermore, to find the group differences between functional networks of HC, SCD and MCI, we performed the ANOVA and two-sample t-test among three groups on the occurrence frequency of dFC states. Last, we conducted Pearson’s correlation analysis to characterize the relationship between dynamic features (the occurrence frequency of dFC states and the co-occurrence frequency between dfALFF states and dFC states) and cognitive scores (MMSE). In order to avoid the problem of multiple comparisons, the significant thresholds were corrected by the false discovery rate (FDR).