In this study, 22 PSCI patients who underwent acupuncture therapy at our hospital were enrolled. The inclusion criteria were as follows: 1) first-onset PSCI and the diagnostic criteria for cerebral infarction, 2) right-handedness, 3) Montreal Cognitive Assessment (MoCA) score < 26 points and Mini-Mental State Examination (MMSE) score < 24 points, and 4) consciousness with stable vital signs. The exclusion criteria were as follows: 1) restroke or hemorrhagic stroke confirmed by imaging; 2) mental illness such as depression, hysteria, or obvious hypophrenia before stroke onset; 3) inability to receive acupuncture therapy or fainting during acupuncture therapy; 4) severe cardiopulmonary or renal dysfunction, blood diseases, or coagulation abnormalities; 5) contraindications to MRI or claustrophobia; and 6) other conditions hindering image acquisition or data quality. All PSCI patients were subjected to the MMSE and MoCA, which were administered by the same experienced neurologist. Fourteen healthy controls (HCs) matched for age, sex, and education level were enrolled in this study. All the subjects signed informed consent forms.

All the subjects underwent T1-weighted imaging (T1WI) and rs-fMRI scans with a magnetic resonance scanner equipped with a 32-channel head–neck coil (MR750 3.0 T, GE). For PSCI patients, rs-fMRI scans were conducted at baseline and 2 weeks after acupuncture therapy. The T1WI scanning parameters were as follows: repetition time (TR) = 8.2 ms, echo time (TE) = 3.1 ms, field of view (FOV) = 256 × 256 mm, slice thickness = 1 mm, slice increment = 0, matrix = 256 × 256, flip angle = 7˚, and time = 3 min 59 s. The rs-fMRI scanning parameters were as follows: TR = 2000 ms, TE = 30 ms, FOV = 230 × 230 mm, slice thickness = 3.5 mm, slice increment = 0.7 mm, matrix = 64 × 64, flip angle = 90˚, and time = 8 min 10 s.

Acupoints were selected with reference to specific meridians and known acupoints. The acupuncture was performed with disposable sterile 0.35 × 50 mm acupuncture needles as follows. After routine disinfection, the bilateral Neiguan acupoints were first punctured straight to a depth of 20 - 25 mm. After the desired sensation was achieved, the neutral supplementation and draining method was performed, and the needles were twirled 120 times/min continuously for 1 min, after which they were left in place for 30 min. Next, the Baihui and Sishencong acupoints were horizontally punctured to a depth of 20 - 25 mm to reach the galea aponeurotica from front to back. After the desired sensation was achieved, the needles were twirled at a small amplitude and a frequency of 200 times/min continuously for 1 min and then left in place for 30 min. Acupuncture was performed by the same acupuncture therapist at 8:00 - 10:00 in the morning once a day, 5 days a week for a total of 2 weeks.

The data were preprocessed with the DPABI software package of the MATLAB platform as follows [26] . 1) Data from the first 10 time points were removed to avoid the effects of environmental variations or unstable magnetic fields and ensure signal stability. 2) Time correction was performed by aligning all image layers of each result to the time point in the middle of the scan. 3) Head motion correction was performed, and the data of subjects with head rotation > 2˚ or translation > 2 mm in the x-, y-, or z-directions were excluded. 4) The brain images of each subject were registered to standard Montreal Neurological Institute (MNI) space and then resampled to a voxel size of 3 mm × 3 mm × 3 mm. 5) Linear drift was removed to reduce the baseline drift caused by the machine. 6) Covariates were removed to eliminate the additional effects of head movement, white matter signals, and cerebrospinal fluid signals. 7) All the data were subjected to bandpass filtering between 0.01 and 0.08 Hz to remove high-frequency signals and physiological noise.

The brain images of all the subjects were registered to the anatomical automatic labeling (AAL) template with GRETNA on the MATLAB platform [27] . In this template, the whole brain was divided into 90 brain regions, in which each node corresponds to one region. Then, the average time series of all voxels in the 90 brain regions were extracted, and the Pearson correlation coefficients of the average time series between every two brain regions were calculated, yielding a 90 × 90 matrix R. Next, matrix R was subjected to Fisher Z-transformation, yielding a Z value correlation matrix that was almost normally distributed. Network sparsity, the ratio of the number of sections in a network to the maximum number of possible sections in the network, was adopted to determine the BFN thresholds. The lower limit of the threshold should meet the criterion that the average degree of all nodes in a BFN of each subject is greater than 2 log(N), where N is the number of nodes, and the upper limit of the threshold should meet the criterion that the parameter of the small-world topological properties of each BFN is greater than 1.1 [28] . For this reason, the range of sparsity was set to 0.1 - 0.28 in this study, with a step size of 0.01.

The area under the curve (AUC) of the whole-brain topological properties (σ, γ, λ, C_p, L_p, E_loc, and E_g) and node properties (betweenness centrality, degree centrality, and nodal local efficiency) in each group was calculated with GRETNA [29] - [31] .

Data analysis was performed with SPSS 25.0 software. The chi-square test and independent-sample t test or paired-sample t test were employed to compare enumeration data and quantitative data, respectively, between groups. Indicators with differences between groups in terms of whole-brain topological properties and node properties were subjected to Pearson correlation analysis with the MMSE and MoCA scores. For all the statistical results, p < 0.05 denoted statistical significance.

A total of 22 PSCI patients and 14 HCs were included in this study; the demographic and clinical characteristics of the included individuals are shown in Table 1 . Compared with the pretreatment scores, the posttreatment MMSE and MoCA scores of the PSCI patients were significantly greater ( Table 2 ).

Notes: MMSE: Mini-Mental State Examination. MoCA: Montreal Cognitive Assessment.

Small-world network attributes were observed in both groups for sparsity values between 0.1 and 0.28. Compared with HCs, PSCI patients had decreased values for the network metric γ (p = 0.024), the clustering coefficient (C_p) (p = 0.038), and E_loc (p = 0.025).

The betweenness centrality in the left cuneus (CUN. L) and right Heschl’s gyrus (HES. R) was greater in PSCI patients than in HCs over the entire threshold range (p < 0.05) ( Table 3 and Figure 1 ). After treatment, PSCI patients had lower betweenness centrality in the PCG.L and CUN. L (p < 0.05) ( Table 4 and Figure 1 ).

Notes: BC: Betweenness centrality. DC: Degree centrality. NLE: Nodal local efficiency. PSCI: poststroke cognitive impairment. IPL. L: left inferior parietal but supramarginal and angular gyrus. PCG. L: left posterior cingulate gyrus. HES. R: right Heschl’s gyrus. SOG. R: right superior occipital gyrus. IFGoperc. R: right inferior frontal gyrus, pars opercularis. MFG. R: right middle frontal gyrus. CUN. L: left cuneus. ORBsupmed. R: right superior frontal gyrus, pars medialis orbitalis. ORBsup. R: right superior frontal gyrus, pars orbitalis. DCG. L: left median cingulate and paracingulate gyri. SMG. R: right supramarginal gyrus. TPOsup. L: left temporal pole: superior temporal gyrus.

Notes: BC: betweenness centrality. DC: degree centrality. NLE: Nodal local efficiency. PSCI: poststroke cognitive impairment. SFGdor.L: left superior frontal gyrus, dorsolateral. IPL. L: left Inferior parietal, but supramarginal and angular gyri. PCG. L: left posterior cingulate gyrus. HES. R: right Heschl’s gyrus. SOG. R: right superior occipital gyrus. IFGoperc. R: right inferior frontal gyrus, pars opercularis. CUN. L: left cuneus. HES. L: left Heschl’s gyrus. PHG. L: parahippocampal gyrus. FFG. L: fusiform gyrus. IOG. R: inferior occipital gyrus. ANG. R: angular gyrus. IOG. L: left inferior occipital gyrus. LING. R: right lingual gyrus. ORBsupmed. R: right superior frontal gyrus, pars medialis orbitalis. LING. L: left lingual gyrus.

The degree centrality in the left dorsal superior frontal gyrus (SFGdor. L), PCG. L, left inferior parietal lobe (IPL. L), and the HES. R was greater in PSCI patients than in HCs, whereas that in the right medial frontal gyrus (MFG. R), right inferior frontal gyrus, pars opercularis (IFGoperc.R), and right superior occipital gyrus (SOG.R) was lower in PSCI patients than in HCs (p < 0.05) ( Table 3 and Figure 1 ).

Compared with that before treatment, the degree centrality in the left lingual gyrus (LING. L), right lingual gyrus (LING.R), and left inferior occipital gyrus (IOG. L) was increased, whereas in the right superior frontal gyrus, pars medialis orbitalis (ORBsupmed. R) was decreased after treatment in PSCI patients (p < 0.05) ( Table 4 and Figure 1 ).

The nodal local efficiency in the right superior frontal gyrus, pars orbitalis (ORBsup. R), ORBsupmed. R, left median cingulate and paracingulate gyri (DCG. L), SMG. R, and left temporal pole: superior temporal gyrus (TPOsup. L) was greater in PSCI patients than in HCs (p < 0.05) ( Table 3 and Figure 1 ).

Compared with that before treatment, the nodal local efficiency in the left parahippocampal gyrus (PHG. L), IOG. R, left fusiform gyrus (FFG. L), and the HES. L was elevated, whereas that in the right angular gyrus (ANG. R) was lower in PSCI patients after treatment (p < 0.05) ( Table 4 and Figure 1 ).

The AUC of the nodal local efficiency in the ORBsupmed. R in PSCI patients before treatment was positively correlated with the MoCA score (r = 0.463, p = 0.042), whereas the AUC of the degree centrality in the IOG. L after acupuncture therapy was negatively associated with the MoCA score (r = −0.461, p = 0.031).