The samples included in this study consist of preoperative pediatric patients and age- and gender-matched healthy children, with all patient data collected prior to surgery. We enrolled 15 pediatric patients with temporal lobe lesions as the patient group, who were treated at the Pediatric Neurosurgery Department of Beijing Tiantan Hospital between November 2021 and January 2024. All 15 patients underwent preoperative fMRI scans. However, due to non-compliance in completing cognitive assessments by two participants, cognitive data were only obtained from 13 children.

We also recruited 15 age- and gender-matched healthy children as the control group from Beijing RIMAG Medical Imaging Center, who underwent the same MRI scans and cognitive evaluations.

For the patient group, the inclusion criteria were: willingness to participate in the study; age between 6 and 18 years; diagnosed with a primary temporal lobe space-occupying lesion requiring surgical intervention. The exclusion criteria were: withdrawal from the study for any reason; refusal to cooperate in completing the MRI scan; presence of medical conditions affecting the brain; or contraindications to anesthesia.

For the control group, the inclusion criteria were: willingness to cooperate in completing MRI scans and cognitive assessments; age between 6 and 18 years; no history of brain disease. The exclusion criteria were: withdrawal from the study for any reason; inability to cooperate in completing the MRI scans or cognitive tests.

Parental written informed consent was collected for all children who participated in the study. The research adhered to the ethical guidelines outlined in the Declaration of Helsinki and received approval from Institutional Review Board of Beijing Tiantan Hospital, Capital Medical University (KY 2021-100-02).

Each participant’s neurocognitive abilities were assessed using the CNS Vital Signs (CNS VS) battery, a digital tool designed for repeated assessments and widely recognized for its reliability and validity (Gualtieri and Johnson, 2006). This 30–40 min assessment evaluates 15 cognitive domains, providing age-adjusted scores based on 10 subtests. The cognitive domains measured include Composite Memory (CM), Verbal Memory (VerbM), Visual Memory (VisM), Psychomotor Speed (PsyMotSpd), Reaction Time (RT), Complex Attention (ComA), Cognitive Flexibility (CogFlex), Processing Speed (ProcSpd), Executive Function (ExeFun), Social Acuity (SocAcu), Reasoning (Reason), Working Memory (WM), Sustained Attention (SustA), Simple Attention (SimA), and Motor Speed (MotSpd). Additionally, the Neurocognitive Index (NCI), derived from CM, PsyMotSpd, RT, CogFlex, and ComA, provides an overall measure of brain function. These assessments were conducted before surgery, in the same week as neuroimaging. For the children in the control group, cognitive testing and imaging were completed on the same day.

The participants were directed to maintain a seated position while keeping their eyes shut. Administration of sedatives was omitted throughout the examination process. All patients and healthy controls underwent MRI scans on a 3T scanner (MAGNETOM Prisma, Siemens Healthcare, Erlangen, Germany) with a 64-channel head/neck coil. The software version is Syngo MR E11. The protocol included T1 weighted structure imaging with magnetization prepared rapid acquisition gradient echo (MPRAGE) sequence and rs-fMRI with an echo-planar imaging (EPI) sequence. The scan parameters for the MPRAGE sequence were: repetition time (TR) = 1560 ms; echo time (TE) = 1.65 ms; flip angle = 8°; slices = 176; field of view (FOV) = 256 × 256 mm; and voxel size = 1 mm isotropic. The parameters for EPI with simultaneous multislice (SMS) acceleration technique were: TR = 2000 ms; TE = 35 ms; slices = 69; SMS = 3; FOV = 207 × 207 mm; voxel size = 2.2 mm isotropic; volumes = 240.

The imaging data for the patients enrolled in the study were obtained through scans conducted at Beijing Tiantan Hospital. The imaging data for the normal controls were collected at Beijing RIMAG Medical Imaging Center. The MRI equipment used at both locations was of the identical model, and the scanning parameters and software version employed were exactly the same.

The preprocessing of rs-fMRI images was performed using the Data Processing and Analysis of Brain Imaging toolbox (Chao-Gan and Yu-Feng, 2010; Yan et al., 2016) (DPABI v6.2)¹ and Statistical Parametric Mapping software (SPM 12),² both operating within MATLAB (Matlab Release 2020b, Mathworks Inc., Natick, MA). The following steps were included in the preprocessing workflow:

(a) The initial 10 images were discarded; (b) Slice-timing and head motion corrections were applied. Participants with head motion exceeding 3 mm in translation or 3° in rotation were excluded from analysis, and one patient was removed due to excessive motion during the MRI scan; (c) Linear trends, cerebrospinal fluid signals, white matter, and the Friston 24-parameter model of head motion were regressed as nuisance covariates from the blood oxygen-level-dependent (BOLD) signal. Global signal regression was not employed in this process (He and Liu, 2012; Saad et al., 2012); (d) Functional images were normalized to the Montreal Neurological Institute (MNI) space using DARTEL (Ashburner, 2007) and resampled to a 3 mm cubic voxel size; (e) Band-pass filtering was performed in the range of 0.01 Hz to 0.10 Hz; (f) The images were smoothed with a full width at half maximum of 4 × 4 × 4 mm.

The data from the patient group and the control group were preprocessed separately. Additionally, to achieve more accurate normalization, the lesion areas in the T1 images of the patient group were manually removed by an experienced neurosurgeon (Rorden et al., 2007; Andersen et al., 2010; Guan et al., 2022). Figure 1 illustrates the lesion distribution for all enrolled patients.

Commonly used metrics and methods in brain network analysis include those that reflect local brain function changes (Lv et al., 2018), such as Regional Homogeneity (ReHo) (Zang et al., 2004), Amplitude of Low Frequency Fluctuations (ALFF) (Zang et al., 2007), and Fractional ALFF (fALFF) (Zou et al., 2008). Additionally, functional connectivity (FC) analysis is employed to assess the functional coordination between predefined brain regions or between predefined regions and other voxels (Lv et al., 2018).

In this study, we calculated ReHo, ALFF, and fALFF to assess local brain function changes. For whole-brain FC analysis, we predefined seed points (i.e., regions of interest, ROI) based on the results of ReHo, ALFF, and fALFF. Seed-based analysis (SBA) was then conducted using these predefined regions.

The method for defining the seed points involved extracting the overlapping brain regions that showed significant group differences across the three metrics—ReHo, ALFF, and fALFF. The regions where patients had higher values than the control group and the regions where patients had lower values were extracted separately. Among the identified overlapping brain regions, any region with fewer than 19 voxels was excluded from being used as a seed point for subsequent SBA analysis. The threshold of 19 voxels was set because a region of interest (ROI) is typically defined as a sphere with a 5 mm radius (Creem and Proffitt, 2001) or 6 mm (DeSimone et al., 2021). A sphere with a 5mm radius occupies 19 voxels in an image with a resolution of 3 mm.

The localization of nodes and clusters was identified using the Automated Anatomical Labeling (AAL) atlas (Tzourio-Mazoyer et al., 2002). Additionally, we applied the “winner-takes-all” approach to map the AAL brain regions onto Yeo’s 7-network model (Yeo et al., 2011). Detailed methods and the correspondence between regions and networks can be found in Supplementary Materials 1, 2 for further reference.

We conducted Spearman rank correlation analyses between the cognitive domains and imaging markers that showed significant differences between the patient and control groups. Due to missing cognitive test data for two patients, the exploratory correlation analysis included 13 patients and 15 controls. The specific items included in the correlation analysis are detailed in the section “3 Results.”

In this study, independent samples t-tests were applied to continuous variables (cognitive tests, age, imaging data), while chi-square tests were used for categorical variables (gender). The statistical analyses for cognitive tests, age, and gender were conducted using SPSS (IBM SPSS Statistics version 26, IBM Corporation), with a significance level set at 0.05 (two-tailed). Given the exploratory nature of the study, no multiple comparisons correction was applied to the cognitive test data (Guan et al., 2022; Guan et al., 2024a).

For imaging data, after performing independent samples t-tests, multiple comparisons correction was necessary. We applied Gaussian random-field theory (GRF) correction (voxel-level p < 0.001, cluster-level p < 0.05, two-tailed) provided in the DPABI software (Chao-Gan and Yu-Feng, 2010; Yan et al., 2016). Additionally, to validate the robustness of our findings, we applied false discovery rate (FDR) correction to the imaging metrics, setting the overall significance level at 0.05. Based on previous studies, FDR correction typically involves setting a voxel threshold, often at 20 voxels (Lieberman and Cunningham, 2009). In this study, we selected a threshold of 50 voxels to further assess the robustness of our results. It is also important to note that all our statistical analyses and multiple comparison corrections were conducted exclusively on unaffected cortical regions. Lesion-affected areas, subcortical structures and the cerebellum were not included within the scope of this study’s analysis.

For the exploratory correlational analysis, given its exploratory nature, we primarily present and interpret results without multiple comparison corrections (Xu et al., 2022; Guan et al., 2024a). However, we have also provided the results adjusted using FDR correction as Supplementary Information, offering readers additional insights. The statistical significance level was set at 0.05 (two-tailed).

Data visualization was carried out using BrainNet Viewer (Xia et al., 2013), Computational Anatomy Toolbox 12,³ CONN’s visualization tools (Whitfield-Gabrieli and Nieto-Castanon, 2012) and GraphPad 8.

A total of 30 participants were enrolled in this study, including 15 pediatric patients and 15 age- and gender-matched healthy controls (with no significant differences in age or gender). The mean age was approximately 10 years, with an equal gender distribution (8 males and 7 females). In the patient group, two-thirds presented with epilepsy as the initial symptom. Tumors accounted for 80% of the cases, with the majority being low-grade lesions. The distribution of lesion laterality was roughly balanced (7 left-sided and 8 right-sided), and the average lesion volume was 16.5 cm³. Detailed demographic information is provided in Table 1, and the histopathological diagnoses for each enrolled patient are presented in Supplementary Material 3.

The results of the cognitive tests revealed no significant differences between the patient group and the control group across most cognitive domains. However, the patient group showed significantly lower scores in two specific areas: WM and SustA. In this study, WM reflects the ability to perceive and attend to symbols during short-term memory tasks, while SustA indicates the ability to guide and focus cognitive activity on specific stimuli. These findings suggest that certain cognitive domains are significantly impaired in children with temporal lobe lesions, though most cognitive functions remain relatively intact. The detailed information is presented in Figure 2 and Table 2. Subsequent analyses and result discussions were also primarily focused on these two cognitive domains. Here, we would like to elaborate further on the rationale for selecting SustA and WM. Firstly, these two cognitive domains exhibited significant differences among various cognitive areas. Given the limited prior knowledge available for reference, our study is inherently exploratory. Focusing on cognitive domains with evident differences may enhance the efficiency of our research. Secondly, previous studies involving adult patients have shown that the temporal lobe functions are indeed associated with sustained attention and memory (Vaz, 2004; Zhang et al., 2009; Lambon Ralph et al., 2012; Bauman et al., 2019). This connection further rationalizes our selection of these two cognitive domains and indirectly supports the findings of our study.

ALFF, fALFF and Reho

ALFF reflects the overall activity level of a specific brain region during the resting state, with higher ALFF values typically indicating stronger activity in that region (Zang et al., 2007). fALFF offers a more precise measure, focusing more specifically on low-frequency fluctuations related to brain function, and is generally better at distinguishing functional brain regions from non-functional areas (such as noise) (Zou et al., 2008). ReHo measures whether the neuronal activity in a given region is synchronized with its surrounding voxels, or in other words, whether these regions are working together (Zang et al., 2004). All three metrics are capable of reflecting local changes in brain function.

In the patient group, compared to the control group, some brain regions exhibited higher values, while others showed lower values across all three metrics. For ALFF, we identified 18 clusters with significant differences (Figure 3A), and the detailed coordinates and anatomical labels are listed in Table 3. For fALFF and ReHo, we found 12 clusters each (Figures 3B, C), with the detailed information provided in Tables 4, 5, respectively. By observing Figure 3, we can see a certain degree of similarity in the distribution patterns across these three metrics. We predefined the seed points for SBA analysis based on the three metrics. Brain regions where the patient group showed higher values than the control group and regions where they showed lower values were separately aggregated to identify common regions. The coordinates and anatomical labels of these regions were determined by selecting the voxel with the highest absolute sum of the values across the three metrics in each overlapping region. This method of defining ROIs was chosen because, as mentioned earlier, the three metrics share certain similarities and are dimensionless. Thus, we hypothesize that if a brain region shows significant differences across all three metrics with a consistent trend, it may represent an important area deserving further attention. Consequently, the voxel with the highest absolute summed value across the three metrics was selected to represent each cluster. Ultimately, we identified four ROIs for subsequent SBA (Figure 4 and Table 6). ROI 1–3 were defined based on the overlapping brain regions where the patient group exhibited higher values across all three metrics compared to the control group. ROI4 was identified based on the overlapping brain region where the patient group showed lower values across all three metrics relative to the control group.

SBA

We conducted SBA using the four ROIs determined by the local brain function metrics and found significant differences only in ROI1, ROI2, and ROI4. This indicates that functional connectivity between certain brain regions and these ROIs differed between the groups. The results are presented in Figure 5, with detailed information provided in Table 7.

For all imaging data results, we have provided the full details, such as the labels of clusters spanning multiple brain regions, in the Supplementary Material 4 for verification. Additionally, among the metrics applied in this study, the spatial distribution and trend of brain regions passing FDR correction closely resemble those passing GRF correction, with the exception of ROI 1 in the SBA, which did not survive correction. This suggests that the enhanced FC between ROI 1 and the left anterior cingulate cortex as well as the right supplementary motor area may lack stability. However, this does not affect the overall trend or interpretation of our key findings. Detailed results following FDR correction, including visualizations of significant brain regions and their distribution and trends, are provided in Supplementary Material 4 for readers’ reference.

In the exploratory Spearman rank correlation analysis, we found that the two cognitive domains, SustA and WM, which showed significant group differences, were correlated with imaging metrics. There was a negative correlation between SustA and ReHo cluster 5, and both SustA and WM showed negative correlation with the FC value between ROI2 and Cluster 2. The detailed results, including correlation coefficients and p-values, are presented in Figure 6. We observed that the correlation analysis primarily involves the frontoparietal network, default mode network, and visual network. Additionally, none of the results from the correlational analysis remained significant after FDR correction. This information has been included in Supplementary Material 5.