Differences in the Relationship Between Structural Connectivity and Functional Activation Following Pediatric TBI
Florencia Ontiveros, The Ohio State University
Abstract:
Pediatric traumatic brain injury (TBI) can put children at risk of degradation of the brain’s white matter (WM) pathways, and deficits in both between- and within-network connectivity. They are also vulnerable to impairment of executive functions, including working memory and spatial working memory (SWM). Youth with moderate-to-severe TBI (msTBI; n=10), complicated-mild TBI (cmTBI; n=11), and orthopedic injury (OI; n=22) underwent structural and task-based functional MRI in a 3T Siemens scanner to investigate the associations between WM density and functional activation in the fronto-parietal network (FPN) during a SWM task. There were no task performance differences; however, we found significant differences among and within groups in WM density, and further analysis revealed msTBI youth had significantly lower FA; MD analyses were less consistent. We added average FPN brain activation during the task as a covariate and found significant differences in FA among groups and, consistent with WM density findings, cmTBI had greater FA than msTBI. Our results suggest differences in functional activation may have underlying structural substrates; next steps involve probing group-based simple slopes to interpret nuances of these group differences.
Project Summary:
Introduction
Traumatic brain injury (TBI) is a major public health concern in the United States that can cause shearing of the white matter (WM) pathways in the brain due to the unrestricted brain movement that stretches the WM axons and damages the axonal cytoskeleton (Kochanek et al., 2000; Adams et al., 1989; Smith et al., 2003). Hence, children are at risk of WM degradation following TBI, particularly in interhemispheric tracts (Wang et al., 2021; Beauchamp et al., 2009; Adams et al., 1989). This degradation is responsible for the dynamic changes in fractional anisotropy (FA) and diffusivity measures such as the apparent diffusion coefficient (ADC) or mean diffusivity (MD). A cascade of changes starts with an increase in FA and a decrease in ADC or MD about a month after the injury, and during the consecutive months, FA values tend to decrease, and ADC and MD values to increase as injury-related swelling recedes (Roberts et al., 2014). These WM abnormalities could bring complications to executive functioning abilities, including in working memory. Deficits in these areas seem to be more present in moderate and severe TBIs, which could persist over time and, therefore, affect the development of higher-order complex problem-solving skills (Babikian & Asarnow, 2009). However, some studies that found no significant difference in working memory performance suggest brain activation differences related to the memory load reflect a possible compensatory mechanism after TBI (Newsome et al., 2007a; Newsome et al., 2008).
Abnormalities in the brain’s functional connectivity are also observed after TBI, both between-network and within-network. Network coordination and interaction are critical for cognitive functioning, and if this ability is impaired, children could experience more difficulty in cognitive assessments or be in the need to develop a compensatory mechanism. The default-mode and fronto-parietal networks (DMN; FPN) are reported most often to have abnormal within-network connectivity in TBI (Han et al., 2016; Shumskaya et al., 2012; Li et al., 2023; Rigon et al., 2016). Studies evaluating the FPN, also referred to as the central executive network, observed reduced within-FPN connectivity months to years after injury while others, early after injury, found an increase (Han et al., 2016; Rigon et al., 2016; Shumskaya et al., 2012). Nonetheless, the FPN thought to serve as a “hub” for the coordination of cognitive control (Marek & Dosenbach, 2018). These inconsistencies emphasize the need for further research, particularly in vulnerable youth.
Spatial working memory (SWM) skills help to keep spatial information available in working memory, and it involves prefrontal, frontal, posterior, and dorso-lateral areas (van Asselen et al., 2006; Zimmer, 2008). This skill mainly activates the FPN, demonstrating the WM connections within these cortical areas of the FPN are crucial for this ability.
This project investigates the associations between WM structural connectivity and functional activation during a SWM task in the FPN. We assessed variations in WM density and then covaried by FPN activation along major pathways in children with moderate-to-severe TBI (msTBI), complicated-mild TBI (cmTBI), and a control group with orthopedic injury (OI).
Methods
Fifty-eight youth between the ages of 8 and 16 participated in the “Neuroimaging of Mechanisms Subserving Cognitive and Social Outcomes in Childhood TBI” Study (K01 HD083459) 1 to 4 years after injury. Fifteen participants were excluded from this project due to excessive head motion or missing data. Those included were 10 youth with msTBI (Mage=12.0), 11 with cmTBI (Mage=12.9), and 22 with OI (Mage=11.5). Participants underwent MRI in a 3T Siemens scanner, where a T1-weighted structural, 64-direction DTI, and task-based fMRI including the SWM task were collected.
Before the scan, the child was introduced to four characters that lived in a linear space as part of the SWM task. During the scan, the child was presented with one to four characters and an event taking place at a location in that linear space. The child then ranked how quickly a character would arrive at the event based on their home location. Memory loads were determined by how many characters the participant needed to rank: one, two, three, or all four characters. In these analyses, we focused on the two most difficult categories: three and four characters contrasted to one character.
We utilized Freesurfer’s (v7.3.2) TRActs Constrained by UnderLying Anatomy (TRACULA) to perform automated probabilistic tractography to reconstruct the corpus callosum (CC), cingulum bundles (CB), bilateral anterior thalamic radiations (ATRs), bilateral longitudinal (i.e., ILF, MLF, and SLF I-III), and bilateral uncinate fasciculi (UF). We utilized statistics on mean FA and MD averaged at consecutive cross-sections along the trajectory of a pathway. An FPN region of interest was generated based on a meta-analysis map from Neurosynth based on the search terms “Frontoparietal”, which was corrected using an FDR of .01 and then binarized. We extracted individual averages of brain activation in the FPN for the contrast described earlier.
One-way ANOVAs evaluated differences in the overall mean accuracy and reaction time, and mean accuracy and reaction time at each memory load of the task. Differences in WM density in tracts of interest were evaluated using a general linear model (GLM) through Freesurfer. We ran F-tests comparing both overall group differences in white-matter density and group differences in the degree that task activation was related to that WM matter density by including average brain activation in the FPN during the SWM task as a covariate. We then followed up results with t-tests to orthogonally compare each group combination and describe the direction of effect (e.g., OI vs. cmTBI, OI vs. msTBI, cmTBI vs. msTBI, OI and cmTBI vs. msTBI) for each analysis. Finally, we performed cluster-wise corrections for multiple comparisons, running 1,000 simulation iterations with a p<.05 cluster-forming threshold through Freesurfer. This analysis yielded clusters that depict tract areas where the association between WM measures and FPN activation differs among or within groups.
Results
Although we found significant group differences in overall mean reaction time (p<.05) and reaction time at memory load 4 (p<.05), there were no significant differences in overall mean accuracy nor mean accuracy by load (ps>.05). Additionally, the msTBI group demonstrated more variability in performance.
When evaluating WM density, we observed significant group differences in FA in parts of the splenium of the CC, the left dorsal CB, and the left MLF (ps<.05) and in MD in parts of the genu and rostrum of the CC, the right ATR, and bilateral SLF I (ps<.05). Post-hoc analysis revealed OI and cmTBI had greater FA in parts of the splenium of the CC, the left dorsal CB, and left MLF (ps<.05), and OI group had significantly greater MD than the cmTBI group in the bilateral SLF I (ps<.05).
Our analyses evaluating the association between FA or MD and activation in the FPN during the SWM task showed significant group differences. We found significant differences in FA among and within groups in the left dorsal CB and right SLF I covaried by brain activation in the FPN (ps<.05), with post-hoc analysis showing the cmTBI group had greater FA covaried by FPN activation compared to msTBI in these tracts (ps<.01). There were considerably more significant group differences in MD compared to FA, including in some parts of the CC, right ATR, bilateral CB, ILF, MLF, SLF I-II, UF, and left SLF III (ps<.05). Surprisingly, the OI group showed greater MD covaried by FPN activation than the cmTBI group in both right ATR and UF (ps<.05).
Discussion
This study aimed to assess performance differences in functional activation during a SWM task in children with TBI and OI, confirm differences in WM density among and within groups at cross-sections of frontal and posterior tracts, and explore the structure-activation association differences at tract cross-sections in TBI. We found no differences in task performance by group, only for reaction time, unlike most of the current literature. However, our results align with other studies that evaluate brain activation in working memory tasks where the performance did not differ in TBI (Newsome et al., 2008; Newsome et al., 2007).
Interhemispheric tracts such as the CC and CBs are prominent to long-term degradation after TBI, and our results contribute to these findings by showing differences in FA and MD in parts of the CC and the left dorsal CB (Hulkower et al., 2013; Ryan et al., 2018 Adamson et al., 2013; Wang et al., 2021; Beauchamp et al., 2009; Adams et al., 1989; Rutgers et al., 2008; Wilde et al., 2010). Differences in the MLF, a tract where there is not a considerable amount of literature on its integrity after TBI, were unexpected but not surprising given that the MLF has been linked with visual information and working memory, future studies are needed to address this finding in depth (Kalyvas et al., 2020; Latini et al., 2021). We also found differences in MD in the SLF and thalamic radiations, tracts corresponding with literature (Ware et al., 2022; Xiong et al., 2014). Additionally, we expected and found the msTBI group to have significantly lower FA, indicating compromised WM microstructure. We expected a negative relationship between injury severity and MD, however, we found OI had greater MD than cmTBI. While predicted that those with cmTBI would not show WM degradation to the same extent as msTBI, increased diffusivity in the OI group was unanticipated. Corroboration with additional control samples is necessary.
Our main results demonstrate the relationship between FA as well as MD and FPN activation differs among the groups, especially in parts of the left dorsal CB and the right SLF I in FA, and all the tracts of interest except the left ATR and right SLF III in MD. The FA results were expected, with cmTBI having greater FA covaried by activation than msTBI and demonstrating WM degradation. However, MD findings were also unexpected as we found the OI group had a greater MD covaried by activation than the cmTBI group. Next steps include examining group-level simple slopes to fully interpret the nuance of these interactions.
Impact Statement
TBI accounts for millions of emergency department hospitalizations, visits, and fatalities in the United States. Fortunately, timely medical care and current intervention produce a higher chance of survival, leading to more TBI survivors who suffer from numerous neurocognitive morbidities after acute recovery. This project takes a step towards improving our understanding of how TBI-related brain changes subserve functional behavior in children and adolescents, especially in aspects of working memory, a crucial ability throughout their cognitive development. We identified changes in FA in the chronic phase following pediatric TBI, reflecting a persistent identifiable deviation in brain structure after more severe injury. Our findings also highlight a possible deficit in SWM due to WM disruption after TBI. While this was not evident in overall performance level, we did identify differences in ties between WM and the way their brain processes this information. Our next steps are to examine group-level simple slopes of these brain-function interactions to fully characterize how these aspects underlie neurocognition after pediatric TBI.
FABBS Comments on the NIH Minority Health and Health Disparities Strategic Plan for 2026-2030