The G* Power software (version 3.1) was used for sample size estimation. Parameters for the test are as follows: two tails, effect size d = 0.8, alpha error prob = 0.05, power = 0.90, allocation ratio = 1. Then the sample size of each group is 28. In the study, a total of 78 right-handed patients with unilateral pontine infarction were recruited from three hospitals (The First Affiliated Hospital of Zhengzhou University, Tianjin Medical University General Hospital, and Tianjin Huanhu Hospital). Among them, 46 patients with left pontine infarction (LPI group), and 32 patients with right pontine infarction (RPI group). Additionally, 50 right-handed healthy controls (HCs) were recruited from local communities. One patient with right pontine ischemic was excluded due to large head motion (mean FD > 0.5). Three patients with inappropriate lesion location and one patient with cavum septi pellucidi were excluded from the LPI group, The flowchart for patient selection is shown in Supplementary Figure 1. Finally, 42 LPI patients, 31 RPI patients, and 50 HCs were included in this research. The study was approved by the medical research ethics committees of the three aforementioned hospitals, and the Chinese Clinical Trial Registry (ChiCTR1900027064). All participants signed informed consent forms.

To guarantee data quality, we created a stringent protocol when amalgamating the data from the three centers, which may have considerable variation. The inclusion criteria were as follows: (1) first-onset infarction that involves the left or right pontine; (2) chronic-stage of the infarction (after initial onset >6 months); (3) right-handedness before infarction onset; (4) 40–80 years of age. The exclusion criteria were as follows: (1) recurrent or hemorrhagic infarction; (2) white matter demyelination with a Fazekas scale score >1 (Fazekas et al., 1987); (3) any other serious brain abnormalities; (4) history of drug addiction or neuropsychological illnesses; (5) serious underlying diseases including cardiac arrest and tumor; (6) contraindications for MRI.

In the present study, Ray Auditory Verbal Learning Test (RAVLT) was utilized to score the verbal memory function, which is a measure involving both verbal short-term memory (VSTM), and verbal long-term memory (VLTM) (Ricci et al., 2022). The modified version of Flanker task was used to access the visual attention (Haciahmet et al., 2021) based on E-Prime 2.0 software.¹ The mean reaction time (RT) and accuracy (ACC) were calculated as indicators of visual attention for each participant in the task.

The MRI data for the study were acquired using three 3.0-Tesla MR imaging scanners. Two of the scanners were Discovery MR 750 (GE Medical Systems, Waukesha, WI, United States) from The First Affiliated Hospital of Zhengzhou University and Tianjin Medical University General Hospital. The other scanner was a Magnetom Trio Tim MR (Siemens, Erlangen, Germany) from the Tianjin Huanhu Hospital. Participants were instructed to keep their eyes closed and refrain from thinking about anything during the scanning process. To minimize head movements and scanner noise, the subjects wore comfortable foam padding and earplugs.

The images of subjects were collected on the two Discovery MR750 scanners using the same parameters to reduce the scanners variability as much as possible. These are as follows: resting-state functional MRI (fMRI) data were obtained using a gradient echo single-shot echo-planar imaging sequence: TR/TE = 2,000/41 ms; field of view = 220 mm × 220 mm; matrix = 64 × 64; flip angle = 90°; slice thickness = 4 mm; 0.5 mm gap; 32 slices; 180 time points. Anatomical images were acquired using a 3D T1-weighted sequence: TR/TE = 8.2/3.2 ms; FOV = 256 mm × 256 mm; matrix = 256 mm × 256 mm; slice thickness = 1.0 mm, no gap; 188 slices. The Trion Tim parameter was as follows: resting-state fMRI data: TR/TE = 2,000/30 ms, flip angle = 90°, FOV = 220 mm × 220 mm, matrix = 64 mm × 64 mm, thickness = 4 mm, no gap, slices = 36, volumes = 180. 3D-T1WI were acquired using the magnetization prepared rapid acquisition gradient echo (MPRAGE) sequence with the following parameters: TR/TE = 2,000/2.26 ms, flip angle = 9^°, FOV = 256 mm × 232 mm, matrix = 256 × 232, thickness = 1 mm, and slices = 192, resulting in a 1 mm isotropic voxel.

Resting-state functional MRI (rs-fMRI) data were preprocessed using the DPABI software (Yan et al., 2016)² in the MATLAB environment. The preprocessing steps were as follows: (1) convert images in DICOM to the NIFTI format; (2) removal of each subject’s initial 10 volumes to enable the signal to equalize and the participants to get used to the scanning noise; (3) time correction for the time delay of the remaining 170 volumes between slices; (4) realignment correction and mean framewise displacement (mean FD) were calculated for head motion. None of the subjects exceeded 2 mm translation and 2 rotation, and one patient in the RPI group was removed because of an FD > 0.5; (5) co-registration of the rs-fMRI and 3D T1WI for each subject; (6) then, functional images were spatially normalized to the standard Montreal Neurological Institute (MNI) space based on Diffeomorphic Anatomical Registration Through Exponentiated Lie algebra (DARTEL), and re-sampled into a size of 3 mm³ voxels; (7) cerebrospinal fluid signal, white matter, and the Friston 24-parameter head motion model were regressed out as nuisance factors (Friston et al., 1996); (8) a temporal band-pass frequency filter (0.01–0.1 Hz) was applied to eliminate low-frequency drift and high-frequency noise; (9) lastly, the images were smoothed using an 8 mm full width half maximum (FWHM) Gaussian kernel for sALFF calculation.

The sALFF and sReHo were calculated using the DPABI software. Every preprocessed time series underwent a fast Fourier transformation to convert it into the frequency domain for the sALFF calculation. The power spectrum’s square root was computed in each subject, and the average square root across the frequency range of 0.01–0.1 Hz was used to get the sALFF value. In addition, the data were then z-transformed for the next statistical analyses. Every analysis was carried out on a whole-brain level.

The Kendall’s coefficient of concordance (KCC) was utilized to compute sReHo by the DPABI software. This serves as a measure of the similarity between the time series of a given voxel and its nearest 26 voxels (Zang et al., 2004). Then, the images were z-transformed and spatial smoothing with an 8 mm FWHM was performed for the statistical analyses.

Analysis of dynamic local metrics (dALFF or dReHo) was carried out utilizing DPABI-based Temporal Dynamic Analysis toolkits (Yan et al., 2016). In this study, the sliding window method was used to calculate the dynamic metrics. This method is effective in detecting temporal changes and assessing the variability of metrics across the brain (Hindriks et al., 2016; Yip et al., 2017; Vergara et al., 2019). The window length used in resting-state dynamic analyses is a key factor. It is used to strike a compromise between detecting quickly changing dynamic spatial characteristics and providing credible evaluations of regional brain activity. The sliding window connectivity has been used in previous studies with lengths of time ranging from 10 to 180 s (Thompson et al., 2013; Gonzalez-Castillo et al., 2015; Chen et al., 2018). Here we used a sliding window length of 30 TRs (60 s) and 141 windows were calculated in the analysis. Furthermore, the window lengths of 20 TRs (40 s) and 40 TRs (80 s) were also calculated for the validation analysis. The standard deviation (SD) of dALFF and dReHo values for all voxels in 141 time windows of each participant was calculated to assess the variability of ALFF and ReHo. Then, the individual map was standardized by dividing the mean dALFF/dReHo of the entire brain to reduce inter-subject variations. Finally, dALFF and dReHo maps were spatially smoothed with an isotropic Gaussian kernel of 8 mm FWHM.

Two-sample t-tests or chi-square tests were used in the SPSS 21.0 software (SPSS, Inc., Chicago, IL, United States) to explore group differences in demographic information between the PI and HCs. The Shapiro-Wilk test was used to determine the normality of clinical data. General linear model (GLM) in SPM12 was used to compare static metrics (sALFF, sReHo) and dynamic metrics (dALFF, dReHo) of localized activity in the whole brain between patients with PI and HCs, with age, sex, education, mean FD, and scanners variables as covariates. Multiple comparison correction was performed based on Gaussian random field theory (GRF, voxel-wise p < 0.001, cluster-wise p < 0.05). Moreover, in order to identify the potential clinical significance of these static and dynamic metrics, the group-level brain regions that showed significant inter-group differences of these metrics were extracted for the receiver operating characteristic (ROC) curve analyses utilizing Graphpad Prism 5.0 software. It was conducted with a significant threshold of p < 0.05.

Extracting the mean values of static (sALFF, sReHo) and dynamic (dALFF, dReHo) metrics from the areas exhibiting significant alteration for the correlation analyses. and comparing these values with RAVLT scores and Flanker-RT to explore their relationships with symptom intensity. RAVLT scores are normally distributed according to the Shapiro-Wilk test, correlations were calculated using Pearson correlation coefficient. Flanker-RT does not follow a normal distribution, correlations were calculated using Spearman correlation coefficient. P < 0.05 were considered as statistically significant.

Demographic and clinical data are described in Table 1. There were no significant differences in age (P_LPI = 0.29, P_RPI = 0.36), and years of education (P_LPI = 0.69, P_RPI = 0.28) both in the LPI and RPI groups when compared to the HCs group. There were no significant differences in sex between the RPI group (P_RPI = 0.48) and HCs group, but there was a significant difference in the LPI group (P_LPI = 0.005). Lesion size at the chronic phase (P = 0.66) and interval time between the stroke onset and the MRI scan (P = 0.13) showed no significant differences between the LPI and RPI groups. Compared with HCs group, there were no significant differences in the scores of VSTM (P_LPI = 0.10, P_RPI = 0.58), VLTM (P_LPI = 0.46, P_RPI = 0.15), Flanker-RT (P_LPI = 0.37, P_RPI = 0.74) both in the LPI and RPI group. Furthermore, a significant difference between RPI group and HCs group in mean FD values (P_RPI = 0.02) was found. The lesion distribution map of the stroke patients is shown in Figure 1.

The significant differences in sALFF and dALFF (the window length of 30 TRs) were shown in Figure 2, Table 2. The verification analysis results of the two additional window lengths (20 TRs and 40 TRs) were displayed in Supplementary Figures 2, 3, Supplementary Tables 2, 3.

Compared to the HCs group, the LPI group showed significantly increased sALFF in the left inferior temporal gyrus (ITG_L) and bilateral caudate nucleus (CN); the RPI group showed increased sALFF in the bilateral caudate nucleus (CN), and decreased sALFF in the right cuneus (CUN_R) and left superior occipital gyrus (SOG_L).

The LPI group displayed significantly increased dALFF in left inferior temporal gyrus (ITG_L) and right caudate nucleus (CN_R) when compared with HCs group. In the RPI group, the patients displayed increased dALFF in bilateral caudate nucleus (CN).

Compared with the HCs group, there were no discernible alterations in sReHo between the HCs and PI groups with multiple comparison corrected (GRF, P < 0.05).

The significant inter-group differences in dReHo were shown in Figure 3, Table 2. Compared with the HCs group, the RPI group displayed significantly decreased dReHo in left middle cingulate cortex (MCC_L). However, there were no discernible alterations in dReHo between the HCs and LPI groups with multiple comparison corrected (GRF, P < 0.05).

To further explore the correlation between these metrics and cognitive function. The mean static metrics (sALFF) and dynamic metrics (dALFF, dReHo) values of the areas exhibiting differences in patients were extracted to perform the correlation analysis. Only LPI group evinces a significant correlation with behavioral scales (Figure 4). In detail, in LPI group, the increased sALFF in the CN_L were positively correlated with VSTM scores (p = 0.043, r = 0.314) (Figure 4A), and the increased sALFF in the CN_R was positively correlated with VSTM scores (p = 0.032, r = 0.332) (Figure 4B). the increased sALFF in the CN_R were negatively correlated with Flanker-RT scores (p = 0.018, r = −0.393) (Figure 4C), and the increased dALFF in the CN_R were negatively correlated with Flanker-RT scores (p = 0.002, r = −0.494) (Figure 4D). No discernible correlation was detected between the regions exhibiting inter-group distinctions and behavioral scales in the RPI group, nor between the VLTM scores and these metrics.

The ROC curves of altered static metrics and dynamic metrics were shown in Figure 5 and Supplementary Table 1. In the static analysis, the results demonstrated that the area under the curves (AUC) of the increased sALFF in ITG_L, CN_L, CN_R were 0.763, 0.810, 0.821, respectively (Figure 5A); the AUC of the increased sALFF in CN_R, CN_L and decreased sALFF in SOG_L, CUN_R were 0.845, 0.807, 0.852, 0.781, respectively (Figure 5B); the AUC of the increased dALFF in ITG_L, CN_R were 0.745, 0.739, respectively (Figure 5C); the AUC of the increased dALFF in CN_L, CN_R were 0.772, 0.769, respectively (Figure 5D); the AUC of the decreased dReHo in MCC_L was 0.898 (Figure 5E).