Five hundred and twenty-six participants from outpatient clinical were consecutively screened from June 2017 to September 2020. One hundred and five participants meeting the following inclusion criteria were selected: (1) ≥ 40 years old, (2) achieved education higher than 6 years, (3) had no history of cerebrovascular accidents, (4) had no physical severe disease (e.g. myocardial infarction or heart failure), (5) had no visual or auditory deficit, (6) had no contraindication for MRI. The detailed screening process is shown in Fig. 1.
Vascular risk factors
Neurologists collected demographic and clinical information for all participants, including sex, age, hypertension, diabetes, history of smoking, history of drinking, education, total cholesterol (TC), triglyceride (TG), high-density lipoprotein (HDL), low-density lipoprotein (LDL), fasting blood glucose (FBG) and glycated hemoglobin glycosylated hemoglobin (HbA1c).
MRI was used to assess the visibility of DMV. All participants underwent a whole-brain MRI. MRI in this paper was acquired using a 3.0 T MRI scanner (Signa HDxt; GE HealthCare, Milwaukee, WI, USA).
Detection and evaluation of DMV on MRI
DMV evaluation was conducted on the SWI sequence. This study has assessed 6 segments (bilateral frontal lobes, parietal lobes, and occipital lobes), and DMV was scaled as 0 (consecutive veins and even signals), 1 (consecutive veins but one or more veins present uneven signals), 2 (one or more discontinuous veins and dotted hypointensity) or 3 (no visible consecutive vein) (Mucke et al., ) (Supplementary Fig. 1). The total DMV score was counted as the sum of 6 segments. Higher scores indicate worse visibility of DMV. The DMV assessments were independently evaluated by two neurologists, and the inter-observer agreement value for DMV is 0.82. Any disagreement was resolved by consensus with the third neuroimaging expert.
Detection and evaluation of CSVD lesions on MRI
Neuroimaging lesions of CSVD include WMH, LI, CMB, and EPVS according to the STRIVE recommendation.
WMH was defined as abnormal hyperintensity on T2 and Flair in periventricular or deep white matter parenchyma. WMH grading conforms to Fazekas score criteria. Higher scores represent more severe WMH. That deep white matter hyperintensity (DWMH) score or periventricular hyperintensity (PVH) score did not equal zero was defined as WMH presence, and DWMH score ≥ 2 or PVH score = 3 was defined as severe WMH (Fazekas, Chawluk, Alavi, Hurtig, & Zimmerman, ).
LI was defined as round or oval lesions with a diameter of 3-15 mm and located in the basal ganglia, thalamus, internal capsule, external capsule, or brain stem. It presents inner hypointensity with surrounded hyperintensity in Flair sequence (Klarenbeek, van Oostenbrugge, Rouhl, Knottnerus, & Staals, ).
CMB was defined as hypointensity lesions with a diameter < 10 mm in SWI sequence (Greenberg et al., ).
EPVS was defined as round or oval lesions with a diameter < 2 mm in basal ganglia and centrum semiovale. It presents hypointensity in Flair sequence (Klarenbeek, van Oostenbrugge, Lodder, et al., ). EPVS was counted as grade 1 (0–10 EPVS lesions); grade 2 (11–25 EPVS lesions) and grade 3 (> 25 EPVS lesions) (Greenberg et al., ; Klarenbeek, van Oostenbrugge, Lodder, et al., ).
The assessments for CSVD lesions were also independently evaluated by two neurologists, and the inter-observer agreement value for DWMH, PVH, LI, CMB, EPVS in basal ganglia and EPVS in centrum semiovale were 0.78, 0.80, 0.82, 0.83 and 0.79 respectively. Any disagreement was resolved by consensus with the third neuroimaging expert.
Participants with Montreal Cognitive Assessment (MoCA) score of 19 to 25 were defined as MCI (de Havenon et al. ). Minimum Mental State Examination (MMSE) was used to assist in evaluating global cognition. Different cognitive domains were evaluated by specific neuropsychological tests. All tests were finished by at least two neurologists.
AVLT (Auditory verbal learning test) assessed the memory of participants. The inspector read words (3 categories and 4 words each category) to participants and participants began to recall as soon as the inspector had finished. This process was repeated 3 times and the number of correct recollections were named as N1, N2, and N3, respectively. After the five-minute non-verbal test, the participants were required to recall the words the fourth time (N4) and repeat this step the fifth time (N5) after another twenty-minute non-verbal test. The memory of participants was measured according to the following criteria(Guo, Zhao, Chen, Ding, & Hong, ): (1) short-term memory (N1 + N2 + N3); (2) short-time delayed recall (N4) and (3) long-time delayed recall (N5).
SDMT (Symbol digit modalities test) assessed the attention of participants. Participants were required to memorize digits (1–9) and their symbol’s correspondence in the test table. In the formal test, participants wrote down digits corresponding to symbols. The number of right items in 90 s was recorded as the final score (Pantoni et al., ).
TMT (Trail making test) assessed the attention of participants. It was composed of two tests. In test A, participants were told to connect in order a sequence of 25 digits (1–25); ln test B, participants were told to connect a sequence of 25 digits in order and the frame of adjacent digits should not be the same. Total seconds of test B was recorded as the final score (Du et al., ).
SCWT (Stroop color and word test) assessed the executive function of participants. It includes three tests. In test A, participants were required to read the characters on the card in order; In test B, participants were required to read the color of ovals on the card in order; ln test C, participants were required to read the color of the characters in order. The total seconds of test C were recorded as the final score (Camerino et al., ).
VFT assessed the verbal function of participants. Participants were required to enumerate as many as possible animals in 60 s, and the total number was recorded as the final score (Zhao, Guo, & Hong, ).
All statistical analyses were performed on SPSS (version 25.0). Continuous data conforming to normal distribution was described by mean (standard deviation), otherwise, it was depicted by median [low quartile (Q25), upper quartile (Q75)]. T-test or Wilcoxon test was selected for comparing continuous data of two groups, Chi-square test or Fisher’s exact test was chosen for comparing dichotomous variable of two groups. Binary logistic regression was used to analyze the relation between DMV score and MCI, and the variables with the P value < 0.1 in univariate regression were included in the multivariate regression model. Stepwise multiple linear regression analysis was used to assess the relationship between DMV scores and various cognitive domains and bidirectional elimination was used to filter factors. A two-tailed P value of < 0.1 was taken as the entry criteria, and a two-tailed P value of < 0.05 was taken as the exclusion criteria. Spearman’s correlation was used to assess the associations of the DMV score as well as other neuroimaging markers with the MMSE, MoCA, and outcomes of different cognitive domains. A two-tailed P value of < 0.05 was considered statistically significant.