Browse the corpus

Changes on Cognition and Brain Network Temporal Variability After Pediatric Neurosurgery. BACKGROUND AND OBJECTIVES: Pediatric intracranial space-occupying lesions are common, with prognoses improving markedly in recent years, significantly extending survival. As such, there is an imperative to pay increased attention to the postoperative cognitive functions and brain network alterations in these children because these factors significantly influence their quality of life. Temporal variability (TV) analysis of brain networks captures the full extent of resting-state activities, reflecting cognitive functions and rehabilitation potential. However, previous research rarely uses TV analyses and most focus on adults or children after multidisciplinary treatments, not reflecting the combined effect caused by neurosurgery only and self-repair. This study gives our insights into this field from a holistic perspective. METHODS: We studied 35 children with intracranial space-occupying lesions, analyzing pre- and postsurgery MRI and cognitive tests. We used TV analysis to assess changes and correlated imaging indicators with cognitive performance. RESULTS: We observed a tendency for cognitive recovery after about 3 months postsurgery, primarily in the domains of social cognition and nonverbal reasoning. TV analysis of brain networks indicated increased nodal variability within systems such as the visual and sensorimotor networks, which are integral to external interactions. Correlative analysis showed that alterations in certain occipital regions were associated with changes in social cognition and nonverbal reasoning. CONCLUSION: These findings suggest significant intrinsic repair in cognitive functions and brain networks at around 3 months postneurosurgery in children. This study not only enriches our comprehension of postoperative cognitive and brain network self-repair processes in children but also furnishes potential therapeutic targets for rehabilitation interventions and establishes a theoretical foundation for proactive surgical interventions.

From November 2021 to September 2023, 40 patients with ISOL admitted to the Pediatric Neurosurgery Department of our hospital were initially considered. After applying inclusion criteria (age 6-18 years, primary ISOLs suspected to be low-grade or benign) and exclusion criteria (nonconsent, hydrocephalus, inability to complete scans or assessments, inadequate data, presence of previous brain-related medical conditions, anesthesia contraindications, follow-up issues), the study cohort was reduced to 35. Preoperative and postoperative cognitive assessments and functional imaging were conducted for these patients. Exclusions postenrollment included 3 for excessive head movement and 2 for noncooperation in postsurgery assessments. The final cohort of 35 had complete data suitable for analysis. Furthermore, we recruited 30 sex- and age-matched healthy children to serve as controls, with whom we compared the preoperative cognitive data and imaging data of the patients. The results of this comparison are included in the supplementary material to aid in a more comprehensive interpretation of our findings. Written informed consent was obtained from all enrolled subjects. This research was performed in line with the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of our hospital (KY 2021-100-02).

From November 2021 to September 2023, 40 patients with ISOL admitted to the Pediatric Neurosurgery Department of our hospital were initially considered. After applying inclusion criteria (age 6-18 years, primary ISOLs suspected to be low-grade or benign) and exclusion criteria (nonconsent, hydrocephalus, inability to complete scans or assessments, inadequate data, presence of previous brain-related medical conditions, anesthesia contraindications, follow-up issues), the study cohort was reduced to 35. Preoperative and postoperative cognitive assessments and functional imaging were conducted for these patients. Exclusions postenrollment included 3 for excessive head movement and 2 for noncooperation in postsurgery assessments. The final cohort of 35 had complete data suitable for analysis. Furthermore, we recruited 30 sex- and age-matched healthy children to serve as controls, with whom we compared the preoperative cognitive data and imaging data of the patients. The results of this comparison are included in the supplementary material to aid in a more comprehensive interpretation of our findings. Written informed consent was obtained from all enrolled subjects. This research was performed in line with the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of our hospital (KY 2021-100-02). The cognitive functions of all participants were evaluated using the CNS Vital Signs (CNS VS) battery, which was designed to be given serially.18 The CNS VS battery is a comprehensive computerized neurocognitive test battery designed for routine clinical using. The CNS VS battery exhibits robust validity and reliability, specifically designed for repeated measures to minimize the impact of practice effects. It assesses various cognitive domains and provides a 30- to 40-minute evaluation, generating a report with age-adjusted standard scores for 15 domains derived from 10 subtests. These domains include composite memory (CM), verbal memory (VerbM), visual memory (VisM), psychomotor speed (PsyMotSpd), reaction time (RT), complex attention (ComAtt), cognitive flexibility (CogFlex), processing speed (ProcSpd), executive function (ExeFun), social acuity (SocAcu), Reasoning (Reason), working memory (WM), sustained attention (SustAtt), simple attention (SimAtt), and motor speed (MotSpd). In addition, a neurocognitive index, which offers a general assessment of the overall neurocognitive status of the subject, is generated based on CM, PsyMotSpd, RT, CogFlex, and ComAtt. The CNS VS standard scores have a mean of 100 and a SD of 15. Cognitive assessments were conducted both before and after the surgery, followed by brain scans within 1 week of the assessments.

al assessment of the overall neurocognitive status of the subject, is generated based on CM, PsyMotSpd, RT, CogFlex, and ComAtt. The CNS VS standard scores have a mean of 100 and a SD of 15. Cognitive assessments were conducted both before and after the surgery, followed by brain scans within 1 week of the assessments. The parameters of MRI and the procedural steps for preprocessing MRI data are meticulously delineated in Supplemental Digital Content 1 (http://links.lww.com/NEU/E401).

al assessment of the overall neurocognitive status of the subject, is generated based on CM, PsyMotSpd, RT, CogFlex, and ComAtt. The CNS VS standard scores have a mean of 100 and a SD of 15. Cognitive assessments were conducted both before and after the surgery, followed by brain scans within 1 week of the assessments. The parameters of MRI and the procedural steps for preprocessing MRI data are meticulously delineated in Supplemental Digital Content 1 (http://links.lww.com/NEU/E401). For the calculation of brain network TV, we adopted the method proposed by Zhang et al,17 which reflects the ability or tendency of a region to reconfigure into different functional communities or its flexibility in functional integration/coordination with different systems. The greater the TV of a region of interest, the more functional communities it involves at different times. Briefly, this study examined TV at 4 levels: voxel-level, regional, within-network, and between-network. We used regional homogeneity (Reho)19 as a method for assessing voxel-level dynamic FC profiles. This is a variant of Zhang's method proposed by Dong et al,20 which we called dynamic Reho (dReho) in our study. Calculations at the other 3 levels were also conducted following the methodology outlined by Zhang et al and previous research in the field.17,20,21 Figure 1 illustrates the computational workflow, which is further elaborated in Supplemental Digital Content 1 (http://links.lww.com/NEU/E401). In this study, we used the Power-264 brain atlas22 as the regions of interests, which categorizes the 264 nodes into 14 networks (Table 1). Notably, 28 nodes affiliated with the Uncertain network were not considered in our analysis.

s further elaborated in Supplemental Digital Content 1 (http://links.lww.com/NEU/E401). In this study, we used the Power-264 brain atlas22 as the regions of interests, which categorizes the 264 nodes into 14 networks (Table 1). Notably, 28 nodes affiliated with the Uncertain network were not considered in our analysis. Method overview. The figure presents the computational procedure for TV analysis. All brain imaging data have undergone preprocessing. A, TV of FC architecture at the voxel level, also referred to as dReho in this study. n represents the number of windows, and L represents the length of every window. B, TV of FC architecture at the regional level. k represents the node, n represents the number of windows, and Fn,k represents the FC of node k within the nth window with the other nodes inside that window. C, TV of FC within networks. m represents the network. Fnm represents the FC between all pairwise combinations of nodes within network m during the nth window. D, TV of FC between networks. l and p represent different networks. Fnl,p represents the FC between the nodes of network l and network p within the nth window. In summary, dReho involves dividing the entire time series into n time windows, each of length L. Within each L-length window, Reho is calculated and these n Reho values are then aggregated to compute the CV, reflecting variability. The conceptual framework for calculations at the regional, within-network, and between-network levels is similar. It involves first computing the mean of Pearson correlation coefficients for various window-related metrics, representing similarity. Then, subtracting this mean from 1 yields variability, reflecting differences. At the regional level, calculations involve first computing the FC of each node within a window with the other nodes (eg, F1,k), followed by calculating pairwise Pearson correlation coefficients from F1,k to Fn,k. Within-network calculations involve computing Pearson correlation coefficients across all pairwise node FCs within a network for each window (eg, F1m), followed by aggregating these values into a 1-dimensional array and calculating Pearson correlation coefficients across n windows. Similarly, between-network computations involve calculating FC between all nodes of a network in one window (eg, network l) and all nodes of another network in the same window (eg, network p).

lowed by aggregating these values into a 1-dimensional array and calculating Pearson correlation coefficients across n windows. Similarly, between-network computations involve calculating FC between all nodes of a network in one window (eg, network l) and all nodes of another network in the same window (eg, network p). CV, coefficient of variation; dReho, dynamic Reho; FC, functional connectivity; Reho, regional homogeneity; TV, temporal variability. Names, Abbreviations, and Node Counts of Large-Scale Networks in the Power-264 Functional Atlas We chose significant cognitive domains and TV nodes in brain networks, calculating difference scores by subtracting postsurgery values from presurgery ones. These scores underwent Pearson correlation analysis at P < .05 significance (1-tailed), without multiple comparison corrections.23,24 Details are presented in the Results section.

cant cognitive domains and TV nodes in brain networks, calculating difference scores by subtracting postsurgery values from presurgery ones. These scores underwent Pearson correlation analysis at P < .05 significance (1-tailed), without multiple comparison corrections.23,24 Details are presented in the Results section. For rs-functional MRI data, we used paired T-tests with Gaussian random-field (voxel P < .001, cluster P < .05) for dReho and false discovery rate correction for brain network TV (P < .05). Cognitive test data were analyzed using the paired t-test with SPSS26 (IBM Corporation) at P < .05, without correction for multiple comparisons. This is both because of the exploratory nature of this study and because previous studies did not adjust for similar conditions.25-27 We also calculated effect sizes for each cognitive domain, and the detailed calculation formula and references are presented in Supplemental Digital Content 1 (http://links.lww.com/NEU/E401). Correlations were analyzed in SPSS26, and the results were presented using BrainNet Viewer28 and GraphPad Prism 8 (GraphPad Software, Inc.).

From November 2021 to September 2023, 40 patients with ISOL admitted to the Pediatric Neurosurgery Department of our hospital were initially considered. After applying inclusion criteria (age 6-18 years, primary ISOLs suspected to be low-grade or benign) and exclusion criteria (nonconsent, hydrocephalus, inability to complete scans or assessments, inadequate data, presence of previous brain-related medical conditions, anesthesia contraindications, follow-up issues), the study cohort was reduced to 35. Preoperative and postoperative cognitive assessments and functional imaging were conducted for these patients. Exclusions postenrollment included 3 for excessive head movement and 2 for noncooperation in postsurgery assessments. The final cohort of 35 had complete data suitable for analysis. Furthermore, we recruited 30 sex- and age-matched healthy children to serve as controls, with whom we compared the preoperative cognitive data and imaging data of the patients. The results of this comparison are included in the supplementary material to aid in a more comprehensive interpretation of our findings.

Written informed consent was obtained from all enrolled subjects. This research was performed in line with the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of our hospital (KY 2021-100-02).

The cognitive functions of all participants were evaluated using the CNS Vital Signs (CNS VS) battery, which was designed to be given serially.18 The CNS VS battery is a comprehensive computerized neurocognitive test battery designed for routine clinical using. The CNS VS battery exhibits robust validity and reliability, specifically designed for repeated measures to minimize the impact of practice effects. It assesses various cognitive domains and provides a 30- to 40-minute evaluation, generating a report with age-adjusted standard scores for 15 domains derived from 10 subtests. These domains include composite memory (CM), verbal memory (VerbM), visual memory (VisM), psychomotor speed (PsyMotSpd), reaction time (RT), complex attention (ComAtt), cognitive flexibility (CogFlex), processing speed (ProcSpd), executive function (ExeFun), social acuity (SocAcu), Reasoning (Reason), working memory (WM), sustained attention (SustAtt), simple attention (SimAtt), and motor speed (MotSpd). In addition, a neurocognitive index, which offers a general assessment of the overall neurocognitive status of the subject, is generated based on CM, PsyMotSpd, RT, CogFlex, and ComAtt. The CNS VS standard scores have a mean of 100 and a SD of 15. Cognitive assessments were conducted both before and after the surgery, followed by brain scans within 1 week of the assessments.

For the calculation of brain network TV, we adopted the method proposed by Zhang et al,17 which reflects the ability or tendency of a region to reconfigure into different functional communities or its flexibility in functional integration/coordination with different systems. The greater the TV of a region of interest, the more functional communities it involves at different times. Briefly, this study examined TV at 4 levels: voxel-level, regional, within-network, and between-network. We used regional homogeneity (Reho)19 as a method for assessing voxel-level dynamic FC profiles. This is a variant of Zhang's method proposed by Dong et al,20 which we called dynamic Reho (dReho) in our study. Calculations at the other 3 levels were also conducted following the methodology outlined by Zhang et al and previous research in the field.17,20,21 Figure 1 illustrates the computational workflow, which is further elaborated in Supplemental Digital Content 1 (http://links.lww.com/NEU/E401). In this study, we used the Power-264 brain atlas22 as the regions of interests, which categorizes the 264 nodes into 14 networks (Table 1). Notably, 28 nodes affiliated with the Uncertain network were not considered in our analysis.

lowed by aggregating these values into a 1-dimensional array and calculating Pearson correlation coefficients across n windows. Similarly, between-network computations involve calculating FC between all nodes of a network in one window (eg, network l) and all nodes of another network in the same window (eg, network p). CV, coefficient of variation; dReho, dynamic Reho; FC, functional connectivity; Reho, regional homogeneity; TV, temporal variability. Names, Abbreviations, and Node Counts of Large-Scale Networks in the Power-264 Functional Atlas

We chose significant cognitive domains and TV nodes in brain networks, calculating difference scores by subtracting postsurgery values from presurgery ones. These scores underwent Pearson correlation analysis at P < .05 significance (1-tailed), without multiple comparison corrections.23,24 Details are presented in the Results section.

For rs-functional MRI data, we used paired T-tests with Gaussian random-field (voxel P < .001, cluster P < .05) for dReho and false discovery rate correction for brain network TV (P < .05). Cognitive test data were analyzed using the paired t-test with SPSS26 (IBM Corporation) at P < .05, without correction for multiple comparisons. This is both because of the exploratory nature of this study and because previous studies did not adjust for similar conditions.25-27 We also calculated effect sizes for each cognitive domain, and the detailed calculation formula and references are presented in Supplemental Digital Content 1 (http://links.lww.com/NEU/E401). Correlations were analyzed in SPSS26, and the results were presented using BrainNet Viewer28 and GraphPad Prism 8 (GraphPad Software, Inc.).

This study involved 35 pediatric patients (all right-handed), average age 10.5 years, with a balanced gender ratio. Most lesions were supratentorial, especially in the frontal lobe, with 57.1% being right-sided. The follow-up ranged from 0.9 to 38.7 weeks, averaging 15.6 weeks. Parental education levels were mostly junior high school, and mothers were typically the primary caregivers (see Table 2 for more details; Supplemental Digital Content 2 [http://links.lww.com/NEU/E402] for pathological diagnoses). All patients underwent neurosurgery without subsequent treatments or specialized cognitive rehabilitation. We also provide basic information and clinical characteristics of all enrolled patients in Supplemental Digital Content 3 (http://links.lww.com/NEU/E403). Clinical Characteristics of Pediatric Patients High, high school/vocational school/technical school/vocational high school; middle, middle school; primary, primary school. Ranging from 6 to 18. The “non-neoplastic” here does not encompass vascular diseases. Detailed pathological diagnoses are provided in the supplementary materials. Ranging from 75.4 to 136 627.7. Ranging from 0.9 to 38.7.

High, high school/vocational school/technical school/vocational high school; middle, middle school; primary, primary school. Ranging from 6 to 18. The “non-neoplastic” here does not encompass vascular diseases. Detailed pathological diagnoses are provided in the supplementary materials. Ranging from 75.4 to 136 627.7. Ranging from 0.9 to 38.7. In the cognitive assessment results of pediatric patients, we observed that postoperatively, there was a statistically significant improvement in the domains of SocAcu and Reason compared with preoperative scores. No statistically significant differences were found in other domains (Table 3). According to the criteria proposed by Cohen,29 the effect sizes of the 2 cognitive domains, SocAcu and Reason, are moderate and small to moderate, respectively, and are perceptibly or nearly perceptibly effects in practice. These results suggest that the improvements in cognitive scores that we observed with cognitive tests may be real and might have real effects. SocAcu can reflect how well a subject can perceive, process, and respond to emotional cues. Reason can reflect how well is a subject able to recognize, reason, and respond to nonverbal visual-abstract stimuli. The boxplots for the scores across these 16 domains are included in Supplemental Digital Content 4 (http://links.lww.com/NEU/E404) to illustrate the general distribution of the data. Cognitive Assessment Scores Comparison Statistical significance.

In the cognitive assessment results of pediatric patients, we observed that postoperatively, there was a statistically significant improvement in the domains of SocAcu and Reason compared with preoperative scores. No statistically significant differences were found in other domains (Table 3). According to the criteria proposed by Cohen,29 the effect sizes of the 2 cognitive domains, SocAcu and Reason, are moderate and small to moderate, respectively, and are perceptibly or nearly perceptibly effects in practice. These results suggest that the improvements in cognitive scores that we observed with cognitive tests may be real and might have real effects. SocAcu can reflect how well a subject can perceive, process, and respond to emotional cues. Reason can reflect how well is a subject able to recognize, reason, and respond to nonverbal visual-abstract stimuli. The boxplots for the scores across these 16 domains are included in Supplemental Digital Content 4 (http://links.lww.com/NEU/E404) to illustrate the general distribution of the data. Cognitive Assessment Scores Comparison Statistical significance. The analysis of dReHo identified a total of 5 statistically significant clusters. We anatomically labeled these 5 clusters based on the Automated Anatomical Labeling atlas.30 The predominant locations of these 5 clusters were primarily in the occipital lobe, including the right lingual gyrus (Lingual_R), left calcarine (Calcarine_L), right inferior occipital gyrus (Occipital_Inf_R), and bilateral middle occipital gyrus (Occipital_Mid_L, Occipital_Mid_R), and postoperatively, their activation is significantly higher than that preoperatively. Detailed information is presented in Figure 2 and Table 4.

l gyrus (Lingual_R), left calcarine (Calcarine_L), right inferior occipital gyrus (Occipital_Inf_R), and bilateral middle occipital gyrus (Occipital_Mid_L, Occipital_Mid_R), and postoperatively, their activation is significantly higher than that preoperatively. Detailed information is presented in Figure 2 and Table 4. Temporal variability of regional homogeneity. A, A more intuitive representation of statistically significant locations; B, precise anatomic localization. A total of 5 clusters predominantly localized to the occipital lobe based on the Automated Anatomical Labeling brain atlas. The activation of these 5 clusters postoperatively was significantly higher than that preoperatively. Detailed information is given in Table 4. The Results of Dynamic Regional Homogeneity Calcarine_L, left calcarine; Lingual_R, right lingual gyrus; Occipital_Inf_R, right inferior occipital gyrus; Occipital_Mid_L, left middle occipital gyrus; Occipital_Mid_R, right middle occipital gyrus.

Temporal variability of regional homogeneity. A, A more intuitive representation of statistically significant locations; B, precise anatomic localization. A total of 5 clusters predominantly localized to the occipital lobe based on the Automated Anatomical Labeling brain atlas. The activation of these 5 clusters postoperatively was significantly higher than that preoperatively. Detailed information is given in Table 4. The Results of Dynamic Regional Homogeneity Calcarine_L, left calcarine; Lingual_R, right lingual gyrus; Occipital_Inf_R, right inferior occipital gyrus; Occipital_Mid_L, left middle occipital gyrus; Occipital_Mid_R, right middle occipital gyrus. The regional-level TV analysis identified a total of 5 nodes with statistically significant differences. The anatomic locations of these 5 nodes are as follows: right posterior cingulate gyrus (Postcentral_R), left Rolandic operculum (Rolandic_Oper_L), left Heschl's gyrus (Heschl_L), right middle occipital gyrus (Occipital_Mid_R), and left cuneus (Cuneus_L). From an anatomic perspective, these regions belong to the Sensory/Somatomotor hand network and the auditory and Visual (Vis). Functionally, these networks are responsible for tasks involving interaction with the external environment. The t-values of these nodes indicate that their postoperative activation is significantly higher than preoperative activation. Specific details are presented in Figure 3 and Table 5.

he auditory and Visual (Vis). Functionally, these networks are responsible for tasks involving interaction with the external environment. The t-values of these nodes indicate that their postoperative activation is significantly higher than preoperative activation. Specific details are presented in Figure 3 and Table 5. Regional-level TV. The Regional-level TV analysis identified 5 nodes with significantly higher activation postoperatively compared with preoperatively. The colors and sizes in the figure reflect the magnitude of the nodes' t-values. These nodes are predominantly distributed within the sensorimotor network, as well as the auditory and visual. TV, temporal variability. The Results of Regional-Level Temporal Variability Aud, auditory; Cuneus_L, left cuneus; Heschl_L, left Heschl's gyrus; Occipital_Mid_R, right middle occipital gyrus; Postcentral_R, right posterior cingulate gyrus; Rolandic_Oper_L, left Rolandic operculum; SM.H, Sensory/Somatomotor hand; Vis, Visual. The node numbering here is determined based on the node sequence provided in the Power-264 brain atlas.

Aud, auditory; Cuneus_L, left cuneus; Heschl_L, left Heschl's gyrus; Occipital_Mid_R, right middle occipital gyrus; Postcentral_R, right posterior cingulate gyrus; Rolandic_Oper_L, left Rolandic operculum; SM.H, Sensory/Somatomotor hand; Vis, Visual. The node numbering here is determined based on the node sequence provided in the Power-264 brain atlas. We did not observe statistically significant differences between preoperative and postoperative TV of within-brain networks and between-brain networks. Only the P-value for TV within the Vis approached statistical significance after false discovery rate correction, indicating a trend. Postoperatively, TV within the Vis showed an increase compared with preoperative values, nearing statistical significance. The specific details are provided in Supplemental Digital Content 5 (http://links.lww.com/NEU/E405) and Supplemental Digital Content 6 (http://links.lww.com/NEU/E406). We conducted correlations among the statistically significant or marginally significant brain imaging findings, cognitive test results, and certain clinical characteristics of the pediatric patients. Brain imaging results encompassed dReho, regional-level TV, and within-network TV of the Vis. Cognitive test results included SocAcu and Reason. Clinical characteristics comprised patients' age, parental education background, lesion volume, and follow-up interval. We also performed a correlation analysis between the highest education level of the main caregiver and the change in Reason domain.

rk TV of the Vis. Cognitive test results included SocAcu and Reason. Clinical characteristics comprised patients' age, parental education background, lesion volume, and follow-up interval. We also performed a correlation analysis between the highest education level of the main caregiver and the change in Reason domain. We observed that changes in several brain imaging measures correlated with changes in cognitive function and clinical characteristics. The changes in cognitive function also correlated with some clinical characteristics. All the details are provided in Figures 4 and 5. Correlations between brain network dynamics, cognitive assessment, and clinical characteristics. Statistically significant correlation analysis results are presented here. A, The relationship between the change in dReho results from Cluster 3 (right inferior occipital gyrus) and the follow-up interval (95% CI is 0.06-0.64); B, The correlation between the change in dReho results from Cluster 4 (left middle occipital gyrus) and delta Reason (95% CI is −0.59 to 0.01); C, The correlation between the change in dReho results from Cluster 5 (right middle occipital gyrus) and the follow-up interval (95% CI is −0.03 to 0.58). D, The correlation between the change in Regional-level temporal variability of Node 166 (left cuneus) and lesion volume (95% CI is −0.57 to 0.04), delta Reason (95% CI is −0.57 to 0.05), and delta SocAcu (95% CI is 0.06-0.64). dReho, dynamic Reho; Reho, regional homogeneity; SocAcu, social acuity.

I is −0.03 to 0.58). D, The correlation between the change in Regional-level temporal variability of Node 166 (left cuneus) and lesion volume (95% CI is −0.57 to 0.04), delta Reason (95% CI is −0.57 to 0.05), and delta SocAcu (95% CI is 0.06-0.64). dReho, dynamic Reho; Reho, regional homogeneity; SocAcu, social acuity. Correlations between cognitive assessment and clinical characteristics. A, The correlation between the maternal education level and delta Reason (95% CI is −0.05 to 0.57); B, The correlation between primary caregivers' education level and delta Reason (95% CI is −0.11 to 0.52); C, The correlation between age and delta SocAcu (95% CI is −0.64 to 0.06). SocAcu, social acuity. Our study aimed to explore postoperative cognitive changes and brain network TV in children who underwent brain surgery alone, using a matched-pair design. Results related to the matched-pair design are presented in the main text, whereas comparisons with healthy controls are included in Supplemental Digital Content 7 (http://links.lww.com/NEU/E407) as auxiliary evidence for this study.

In the cognitive assessment results of pediatric patients, we observed that postoperatively, there was a statistically significant improvement in the domains of SocAcu and Reason compared with preoperative scores. No statistically significant differences were found in other domains (Table 3). According to the criteria proposed by Cohen,29 the effect sizes of the 2 cognitive domains, SocAcu and Reason, are moderate and small to moderate, respectively, and are perceptibly or nearly perceptibly effects in practice. These results suggest that the improvements in cognitive scores that we observed with cognitive tests may be real and might have real effects. SocAcu can reflect how well a subject can perceive, process, and respond to emotional cues. Reason can reflect how well is a subject able to recognize, reason, and respond to nonverbal visual-abstract stimuli. The boxplots for the scores across these 16 domains are included in Supplemental Digital Content 4 (http://links.lww.com/NEU/E404) to illustrate the general distribution of the data. Cognitive Assessment Scores Comparison Statistical significance.

The analysis of dReHo identified a total of 5 statistically significant clusters. We anatomically labeled these 5 clusters based on the Automated Anatomical Labeling atlas.30 The predominant locations of these 5 clusters were primarily in the occipital lobe, including the right lingual gyrus (Lingual_R), left calcarine (Calcarine_L), right inferior occipital gyrus (Occipital_Inf_R), and bilateral middle occipital gyrus (Occipital_Mid_L, Occipital_Mid_R), and postoperatively, their activation is significantly higher than that preoperatively. Detailed information is presented in Figure 2 and Table 4. Temporal variability of regional homogeneity. A, A more intuitive representation of statistically significant locations; B, precise anatomic localization. A total of 5 clusters predominantly localized to the occipital lobe based on the Automated Anatomical Labeling brain atlas. The activation of these 5 clusters postoperatively was significantly higher than that preoperatively. Detailed information is given in Table 4. The Results of Dynamic Regional Homogeneity Calcarine_L, left calcarine; Lingual_R, right lingual gyrus; Occipital_Inf_R, right inferior occipital gyrus; Occipital_Mid_L, left middle occipital gyrus; Occipital_Mid_R, right middle occipital gyrus.

Aud, auditory; Cuneus_L, left cuneus; Heschl_L, left Heschl's gyrus; Occipital_Mid_R, right middle occipital gyrus; Postcentral_R, right posterior cingulate gyrus; Rolandic_Oper_L, left Rolandic operculum; SM.H, Sensory/Somatomotor hand; Vis, Visual. The node numbering here is determined based on the node sequence provided in the Power-264 brain atlas. We did not observe statistically significant differences between preoperative and postoperative TV of within-brain networks and between-brain networks. Only the P-value for TV within the Vis approached statistical significance after false discovery rate correction, indicating a trend. Postoperatively, TV within the Vis showed an increase compared with preoperative values, nearing statistical significance. The specific details are provided in Supplemental Digital Content 5 (http://links.lww.com/NEU/E405) and Supplemental Digital Content 6 (http://links.lww.com/NEU/E406).

The regional-level TV analysis identified a total of 5 nodes with statistically significant differences. The anatomic locations of these 5 nodes are as follows: right posterior cingulate gyrus (Postcentral_R), left Rolandic operculum (Rolandic_Oper_L), left Heschl's gyrus (Heschl_L), right middle occipital gyrus (Occipital_Mid_R), and left cuneus (Cuneus_L). From an anatomic perspective, these regions belong to the Sensory/Somatomotor hand network and the auditory and Visual (Vis). Functionally, these networks are responsible for tasks involving interaction with the external environment. The t-values of these nodes indicate that their postoperative activation is significantly higher than preoperative activation. Specific details are presented in Figure 3 and Table 5. Regional-level TV. The Regional-level TV analysis identified 5 nodes with significantly higher activation postoperatively compared with preoperatively. The colors and sizes in the figure reflect the magnitude of the nodes' t-values. These nodes are predominantly distributed within the sensorimotor network, as well as the auditory and visual. TV, temporal variability. The Results of Regional-Level Temporal Variability Aud, auditory; Cuneus_L, left cuneus; Heschl_L, left Heschl's gyrus; Occipital_Mid_R, right middle occipital gyrus; Postcentral_R, right posterior cingulate gyrus; Rolandic_Oper_L, left Rolandic operculum; SM.H, Sensory/Somatomotor hand; Vis, Visual. The node numbering here is determined based on the node sequence provided in the Power-264 brain atlas.

We did not observe statistically significant differences between preoperative and postoperative TV of within-brain networks and between-brain networks. Only the P-value for TV within the Vis approached statistical significance after false discovery rate correction, indicating a trend. Postoperatively, TV within the Vis showed an increase compared with preoperative values, nearing statistical significance. The specific details are provided in Supplemental Digital Content 5 (http://links.lww.com/NEU/E405) and Supplemental Digital Content 6 (http://links.lww.com/NEU/E406).

We conducted correlations among the statistically significant or marginally significant brain imaging findings, cognitive test results, and certain clinical characteristics of the pediatric patients. Brain imaging results encompassed dReho, regional-level TV, and within-network TV of the Vis. Cognitive test results included SocAcu and Reason. Clinical characteristics comprised patients' age, parental education background, lesion volume, and follow-up interval. We also performed a correlation analysis between the highest education level of the main caregiver and the change in Reason domain. We observed that changes in several brain imaging measures correlated with changes in cognitive function and clinical characteristics. The changes in cognitive function also correlated with some clinical characteristics. All the details are provided in Figures 4 and 5.

We conducted correlations among the statistically significant or marginally significant brain imaging findings, cognitive test results, and certain clinical characteristics of the pediatric patients. Brain imaging results encompassed dReho, regional-level TV, and within-network TV of the Vis. Cognitive test results included SocAcu and Reason. Clinical characteristics comprised patients' age, parental education background, lesion volume, and follow-up interval. We also performed a correlation analysis between the highest education level of the main caregiver and the change in Reason domain. We observed that changes in several brain imaging measures correlated with changes in cognitive function and clinical characteristics. The changes in cognitive function also correlated with some clinical characteristics. All the details are provided in Figures 4 and 5. Correlations between brain network dynamics, cognitive assessment, and clinical characteristics. Statistically significant correlation analysis results are presented here. A, The relationship between the change in dReho results from Cluster 3 (right inferior occipital gyrus) and the follow-up interval (95% CI is 0.06-0.64); B, The correlation between the change in dReho results from Cluster 4 (left middle occipital gyrus) and delta Reason (95% CI is −0.59 to 0.01); C, The correlation between the change in dReho results from Cluster 5 (right middle occipital gyrus) and the follow-up interval (95% CI is −0.03 to 0.58). D, The correlation between the change in Regional-level temporal variability of Node 166 (left cuneus) and lesion volume (95% CI is −0.57 to 0.04), delta Reason (95% CI is −0.57 to 0.05), and delta SocAcu (95% CI is 0.06-0.64). dReho, dynamic Reho; Reho, regional homogeneity; SocAcu, social acuity.

I is −0.03 to 0.58). D, The correlation between the change in Regional-level temporal variability of Node 166 (left cuneus) and lesion volume (95% CI is −0.57 to 0.04), delta Reason (95% CI is −0.57 to 0.05), and delta SocAcu (95% CI is 0.06-0.64). dReho, dynamic Reho; Reho, regional homogeneity; SocAcu, social acuity. Correlations between cognitive assessment and clinical characteristics. A, The correlation between the maternal education level and delta Reason (95% CI is −0.05 to 0.57); B, The correlation between primary caregivers' education level and delta Reason (95% CI is −0.11 to 0.52); C, The correlation between age and delta SocAcu (95% CI is −0.64 to 0.06). SocAcu, social acuity.

Our study aimed to explore postoperative cognitive changes and brain network TV in children who underwent brain surgery alone, using a matched-pair design. Results related to the matched-pair design are presented in the main text, whereas comparisons with healthy controls are included in Supplemental Digital Content 7 (http://links.lww.com/NEU/E407) as auxiliary evidence for this study.

Our study examines cognitive function changes after pediatric neurosurgery and brain network alterations in TV. We found that in children with ISOLs, neurosurgery-induced damage and neuroplasticity lead to increased TV in the sensorimotor system (Sensory/Somatomotor hand, Sensory/somatomotor mouth, auditory, Vis networks) and cognitive improvement (SocAcu and Reason). These changes suggest enhanced sensorimotor integration and flexibility, indicating cognitive recovery. This enhances knowledge of postsurgical brain self-repair in children and identifies potential rehabilitation therapy targets. We hypothesize that the broad distribution of lesions in our sample, yet localized self-repair in only 2 domains, may be due to sample heterogeneity or extensive cognitive domain interconnections. Similar sample heterogeneity has been observed previously. Studies on traumatic brain injury (TBI) often show significant heterogeneity, with evidence suggesting changes in social cognition or reasoning abilities postinjury, whereas other cognitive domains remain less affected.31-33 Previous research studies indicate mutual influences among cognitive domains and shared cognitive resources although specific mechanisms remain unclear.34-37 These findings may explain our results: minor changes in other domains could lead to significant alterations in 1 or 2 domains.

domains remain less affected.31-33 Previous research studies indicate mutual influences among cognitive domains and shared cognitive resources although specific mechanisms remain unclear.34-37 These findings may explain our results: minor changes in other domains could lead to significant alterations in 1 or 2 domains. Comparison with controls leads to proposed hypotheses. No significant differences were found in the Reason and SocAcu domains between patient and control groups. Correlation analyses suggest associations with specific brain regions, hinting at potential preoperative brain network reshaping. Previous studies have shown that preoperative magnetic or electrical stimulation in patients with brain tumors in eloquent areas can reshape brain networks, resulting in broader tumor resection and faster functional recovery.38,39 This suggests that preoperative brain network reshaping can expedite the rehabilitation process. TV analysis and dReho show increased variability in sensorimotor system nodes and clusters after surgery, despite widespread lesion distribution. Previous static FC studies in pediatric brain surgery found postoperative changes confined to this network.40 Similar findings occur in TBI studies.41,42 Martinez-Molina et al.43 found enhanced coupling between the sensorimotor and other networks postmusic therapy in TBI patients, correlating with improved executive functions. This indirectly supports our results, suggesting overall brain network adjustments despite damage distribution.

ndings occur in TBI studies.41,42 Martinez-Molina et al.43 found enhanced coupling between the sensorimotor and other networks postmusic therapy in TBI patients, correlating with improved executive functions. This indirectly supports our results, suggesting overall brain network adjustments despite damage distribution. Margulies et al.44 described a large-scale organizational structure of brain networks that positions the default mode network at the opposite end of the spectrum from primary sensory and motor areas. Vidaurre et al.45 identified 2 hierarchical structures in brain networks: one for higher cognition and the other for the sensorimotor system. Modern views recognize the human cortex as having unimodal regions for sensory input and multimodal areas for abstract cognition.46,47 From these, we assume that large-scale brain networks could be broadly classified into 2 categories: the sensorimotor system, responsible for interacting with the external environment, and the high-level cognitive system, responsible for information integration. Our study reveals that in pediatric neurosurgery patients, alterations are mainly within the sensorimotor system, whereas the high-level cognitive system remains stable. Fox et al.48 used intracranial electrical stimulation to explore its effects across brain sites. They found frequent and direct stimulation effects in the sensorimotor network, contrasting with infrequent, varied, and complex effects in higher-level multimodal networks like the default mode network. This suggests the sensorimotor system's susceptibility to external influences, partially corroborating our research and strengthening our findings.

stimulation effects in the sensorimotor network, contrasting with infrequent, varied, and complex effects in higher-level multimodal networks like the default mode network. This suggests the sensorimotor system's susceptibility to external influences, partially corroborating our research and strengthening our findings. Correlation analysis links SocAcu changes to TV increases in the cuneus, whereas Reason changes correlate negatively with TV alterations in the middle occipital and cuneus gyri. These results suggest that as TV increases in these brain regions, SocAcu changes increase, whereas Reason changes decrease. Nonetheless, both domains show improvement postoperatively, indicating a trend toward rehabilitation. The 2 domains, SocAcu and Reason, require subjects to respond to on-screen visual stimuli. SocAcu focused on emotion discernment through facial expressions and text, whereas Reason tested nonverbal reasoning with geometric figures. The middle occipital gyrus, crucial in visual processing and shape recognition,49 has been linked to unconscious face/tool processing50 and emotion recognition.51 The cuneus, vital in primary visual processing,52 also influences geometric figure recognition53 and is involved in higher cognitive functions like behavioral engagement,54 working memory,55 and cognitive control.56 The above studies support our correlation analysis results from a functional and anatomic point of view. Growing evidence suggests that brain variability is key to the nervous system, indicating higher network complexity and cognitive processing.6,8,57 This implies that increased variability could be associated with better cognitive function. This is consistent with the SocAcu results. Increased variability, if excessive, may also affect the transmission and integration of information between brain regions.9 This is the situation faced by the Reason field in this study. This reminds us that optimal recovery in behavioral indices may require TV enhancement within a specific range.

the SocAcu results. Increased variability, if excessive, may also affect the transmission and integration of information between brain regions.9 This is the situation faced by the Reason field in this study. This reminds us that optimal recovery in behavioral indices may require TV enhancement within a specific range. Brain areas and nodes identified offer promising targets for rehabilitation, especially those on the brain's convexity with significant differences, suitable for transcranial magnetic stimulation techniques.58 Study shows that TV changes in right occipital gyrus correlate positively with follow-up time but not with SocAcu and Reason changes, indicating ineffective compensation or alternative cognitive adaptation.59 Lesion volume negatively correlates with imaging metrics, suggesting a greater need for postoperative rehabilitation. Age's negative correlation with SocAcu change aligns with Kennard's principle,60 suggesting that older children may need more active therapy for recovery. Our study links maternal education to changes in children's reasoning skills. However, substituting it with the primary caregiver's highest education shows no significant association, emphasizing mothers' unique role in postinjury self-repair. Maternal education strongly correlates with children's health and cognitive development.61-63 Highly educated mothers may indirectly enhance children's cognitive abilities through education and lifestyle, aiding recovery.

tion shows no significant association, emphasizing mothers' unique role in postinjury self-repair. Maternal education strongly correlates with children's health and cognitive development.61-63 Highly educated mothers may indirectly enhance children's cognitive abilities through education and lifestyle, aiding recovery. This study's constraints include sample size and heterogeneity, limiting result interpretation. Yet, akin to previous studies,25,64-66 it offers insights into postinjury repair mechanisms, guiding future studies. Challenges arise from children's poor self-control during MRI scans, potentially leading to data loss. Socioeconomic issues causing follow-up challenges may bias the conclusions, requiring attention and correction in future research. Because of its exploratory nature, cognitive domain comparisons lacked multiple comparison corrections, warranting future validation to mitigate false positives. Conducted in real-world settings, confounding factors may exist, highlighting the need for further refined research.

We hypothesize that the broad distribution of lesions in our sample, yet localized self-repair in only 2 domains, may be due to sample heterogeneity or extensive cognitive domain interconnections. Similar sample heterogeneity has been observed previously. Studies on traumatic brain injury (TBI) often show significant heterogeneity, with evidence suggesting changes in social cognition or reasoning abilities postinjury, whereas other cognitive domains remain less affected.31-33 Previous research studies indicate mutual influences among cognitive domains and shared cognitive resources although specific mechanisms remain unclear.34-37 These findings may explain our results: minor changes in other domains could lead to significant alterations in 1 or 2 domains. Comparison with controls leads to proposed hypotheses. No significant differences were found in the Reason and SocAcu domains between patient and control groups. Correlation analyses suggest associations with specific brain regions, hinting at potential preoperative brain network reshaping. Previous studies have shown that preoperative magnetic or electrical stimulation in patients with brain tumors in eloquent areas can reshape brain networks, resulting in broader tumor resection and faster functional recovery.38,39 This suggests that preoperative brain network reshaping can expedite the rehabilitation process.

TV analysis and dReho show increased variability in sensorimotor system nodes and clusters after surgery, despite widespread lesion distribution. Previous static FC studies in pediatric brain surgery found postoperative changes confined to this network.40 Similar findings occur in TBI studies.41,42 Martinez-Molina et al.43 found enhanced coupling between the sensorimotor and other networks postmusic therapy in TBI patients, correlating with improved executive functions. This indirectly supports our results, suggesting overall brain network adjustments despite damage distribution.

Correlation analysis links SocAcu changes to TV increases in the cuneus, whereas Reason changes correlate negatively with TV alterations in the middle occipital and cuneus gyri. These results suggest that as TV increases in these brain regions, SocAcu changes increase, whereas Reason changes decrease. Nonetheless, both domains show improvement postoperatively, indicating a trend toward rehabilitation. The 2 domains, SocAcu and Reason, require subjects to respond to on-screen visual stimuli. SocAcu focused on emotion discernment through facial expressions and text, whereas Reason tested nonverbal reasoning with geometric figures. The middle occipital gyrus, crucial in visual processing and shape recognition,49 has been linked to unconscious face/tool processing50 and emotion recognition.51 The cuneus, vital in primary visual processing,52 also influences geometric figure recognition53 and is involved in higher cognitive functions like behavioral engagement,54 working memory,55 and cognitive control.56 The above studies support our correlation analysis results from a functional and anatomic point of view. Growing evidence suggests that brain variability is key to the nervous system, indicating higher network complexity and cognitive processing.6,8,57 This implies that increased variability could be associated with better cognitive function. This is consistent with the SocAcu results. Increased variability, if excessive, may also affect the transmission and integration of information between brain regions.9 This is the situation faced by the Reason field in this study. This reminds us that optimal recovery in behavioral indices may require TV enhancement within a specific range.

Brain areas and nodes identified offer promising targets for rehabilitation, especially those on the brain's convexity with significant differences, suitable for transcranial magnetic stimulation techniques.58 Study shows that TV changes in right occipital gyrus correlate positively with follow-up time but not with SocAcu and Reason changes, indicating ineffective compensation or alternative cognitive adaptation.59 Lesion volume negatively correlates with imaging metrics, suggesting a greater need for postoperative rehabilitation. Age's negative correlation with SocAcu change aligns with Kennard's principle,60 suggesting that older children may need more active therapy for recovery. Our study links maternal education to changes in children's reasoning skills. However, substituting it with the primary caregiver's highest education shows no significant association, emphasizing mothers' unique role in postinjury self-repair. Maternal education strongly correlates with children's health and cognitive development.61-63 Highly educated mothers may indirectly enhance children's cognitive abilities through education and lifestyle, aiding recovery.

This study's constraints include sample size and heterogeneity, limiting result interpretation. Yet, akin to previous studies,25,64-66 it offers insights into postinjury repair mechanisms, guiding future studies. Challenges arise from children's poor self-control during MRI scans, potentially leading to data loss. Socioeconomic issues causing follow-up challenges may bias the conclusions, requiring attention and correction in future research. Because of its exploratory nature, cognitive domain comparisons lacked multiple comparison corrections, warranting future validation to mitigate false positives. Conducted in real-world settings, confounding factors may exist, highlighting the need for further refined research.

Children with ISOLs display early cognitive self-repair postsurgery, notably in social cognition and nonverbal reasoning, about 3 months postoperative. Increased TV in the sensorimotor system, particularly in the Vis's cuneus and occipital lobe, suggests enhanced neural network integration, correlating with cognitive improvements. These insights offer potential rehabilitative targets, for transcranial magnetic stimulation. Our study enables macroscopic evaluation of postsurgical self-repair, aiding holistic understanding beyond lesion-specific considerations. It illuminates brain network adaptation to widespread damage, revealing neural plasticity and compensatory mechanisms. Findings support more aggressive surgery followed by cognitive rehabilitation in contentious cases for optimal benefits. In addition, maternal education positively influences recovery, offering a new rehabilitation perspective.