Developmental Factors That Predict Head Movement During Resting-State fMRI in Preschool-Age Children

Background: Early childhood marks a period of dynamic neurocognitive development. The preschool age coincides with the onset of many childhood disorders and is a developmental period that is frequently studied to determine markers of neurodevelopmental disorders. Magnetic resonance imaging (MRI) is often used to explore typical brain development and the neural bases of neurodevelopmental disorders. One advantage of MRI is that it provides comprehensive imaging data with high spatial resolution that is not possible with “child-friendly” techniques such as electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS). However, acquiring high-quality MRI data in young children is challenging. The enclosed space and loud sounds can trigger unease and cause excessive head movement. A better understanding of potential factors that predict successful MRI acquisition would increase chances of collecting useable data in children with and without neurodevelopmental disorders. Methods: We investigated whether age, sex, stuttering status, and childhood temperament as measured using the Child Behavioral Questionnaire, could predict movement extent during resting-state functional MRI (rs-fMRI) in 76 children aged 3 to 7 years, including 42 children who stutter (CWS). Results: The CWS did not differ signicantly from controls in temperament or head movement during scanning. Boys exhibited signicantly more movement than girls, and younger age was a consistent predictor of increased movement, unless temperament was considered. Effortful control and negative affectivity signicantly predicted movement especially in young boys. Conclusions: Considering temperament, age, and sex may help predict the success of acquiring useable rs-fMRI data in young children and inform future studies developing better desensitization techniques for pediatric MR neuroimaging.


Background
Though generally considered to be a technique that is di cult to apply in younger populations, MRI has been used at an increasing rate over the last few decades to study brain development in typically developing children and children with neurodevelopmental disorders (1). The onset of many neurodevelopmental disorders, such as dyslexia (2)(3)(4), attention-de cit/ hyperactivity disorder ([ADHD]; 5), speci c language impairment ([SLI]; 6), autism spectrum disorders ([ASD]; 7), intellectual disability (8), and developmental stuttering (9,10) occurs at preschool-age or earlier. The ability to detect subtle neural differences near symptom onset can be greatly enhanced by the acquisition of whole-brain measures using functional magnetic resonance imaging (fMRI), which is crucial to extract both localized and network-level information to expand our understanding of the nature of these neurodevelopmental disorders. Early detection of brain network-level changes associated with childhood neurodevelopmental disorders might pave the way for developing assessments, interventions, and better prognostic markers for these disorders; however, this requires acquiring MRI data from very young children.
Collecting high-quality MRI data from preschool-age children is challenging. Notably, age is an important factor that in uences data quality across various data collection methods in terms of compliance, active participation, and reliability of measurement (11,12). For MRI in particular, it is di cult for young children to remain still for the 30 to 60 minute scanning period required for most research scans (11,12). In addition, young children may exhibit fear or anxiety associated with the scanning environment, which can lead to increased movement during scanning. Previous research has shown that for neuroimaging studies in children 4-6 years of age, researchers may need to increase recruitment for data collection by 20-40% to obtain su cient useable data without movement artifacts (12). Malisza and colleagues (11) indicated that children 2-3 years of age have a signi cantly lower rate of successful MRI completion than children 6-7 years of age. These authors also suggest that researchers should anticipate a failure rate of at least 50% if a study includes children 2-7 years of age (11). Age is consistently found to in uence the "success" of MRI data collection in children. Success of MRI data collection in this study is de ned as data that are relatively free of movement artifacts and are thus of su cient quality to be included in data analysis. In preschool-age children, success rates are especially low (11,12). Given the challenges researchers face in acquiring MRI data in young children, it is vital to understand factors that may interact with the age of the child that could lead to excessive movement during scanning.
Previous research has suggested that scan duration, movement restrictions, being placed within a con ned space, and the excessive noise produced by MRI gradient coils may contribute to a child having di culty completing a scan (13)(14)(15). Potential modulatory factors that have been investigated in clinical disorders and typically developing children include age and sex; however, the contribution of these variables is understudied and thus poorly understood (11,16). Some studies have found that age did not predict willingness to participate in MRI but did predict scanning success (11). The sample sizes in this study differed across age groups however, potentially in uencing their results. Additionally, 2-5 year-olds had a failure rate of 50%, while 6-7 year-olds only had a failure rate of 35%. There is no strong evidence for an effect of sex (11,16); however, not all studies explicitly examined sex as a factor (12). These studies importantly did not examine the differences between sexes within clinical populations, despite a large body of evidence suggesting that different behaviors and developmental patterns are present among boys and girls diagnosed with neurodevelopmental disorders (e.g., [17][18][19][20]. Still, some young children seem to be able to tolerate scanning better than much older children. This observation suggests that interactions with variables other than age may predispose some children to tolerate scanning better than others. One such factor that may in uence pediatric scanning success is temperament, or one's innate behavioral dispositions. For example, temperament affects a child's tolerance to many clinical environments such as doctor and dental visits (e.g., [21][22][23]. To date, few studies have directly investigated temperament as it relates to the successful completion of MRI scanning in children (16,24). From these studies, the best predictors of movement during scanning were poor attention and adaptability skills (i.e., ability to handle novel experiences) among preschool-and school-age children (16,24). Children with neurodevelopmental disorders may display decreased attention regulation and adaptability, which might further compromise their ability to tolerate scanning (25)(26)(27)(28)(29)(30)(31)(32)(33). While some studies have shown that children with neurodevelopmental disorders such as ADHD, and ASD, or developmental delays tend to have fewer successful scans than controls, the reasons for this have not been clari ed, especially in preschool-age children (12,16).
Moreover, temperament factors such as adaptability and attentional control develop differently between the sexes, even in typically developing children. Preschool-age girls tend to have more advanced skills related to effortful control than their age-matched male peers (34). If temperamental factors associated with poor attention and adaptability skills in uence scanning tolerance in children, those with neurodevelopmental disorders may already be at a greater disadvantage and have a lower probability of success, leading to spurious group differences that may be erroneously attributed to the core clinical condition. If temperament factors that in uence MRI scanning develop more slowly in boys than girls, boys with neurodevelopmental disorders may have an even more challenging time tolerating MRI scanning.
Several studies to date have reported that children who stutter (CWS) and children who do not stutter differ in some temperament dimensions. For example, CWS have been shown to demonstrate decreases in inhibition (35)(36)(37)(38), attentional shifting (35,36), attentional focusing (36, 37,39), and attentional regulation (36, 40). They also exhibit increased di culty adapting to change (41)(42), present with greater negative affect (35,42,43), display increased emotional reactivity (40), show higher activity levels (37,42), demonstrate more anger and frustration (35,39), and exhibit more impulsivity (37). Because these factors may contribute to tolerance of the scanning procedures in clinical populations, information that further elucidates relationships between temperament, presence of a clinical condition (in this case, stuttering), and head movement during scanning, could help develop strategies to increase chances of collecting usable MRI data that is comparable across the clinical and control groups.
The current study investigated whether clinical condition (stuttering compared to children who do not stutter, hereafter referred to as "controls"), age, sex, and temperament as assessed by scores on the Child Behavior Questionnaire ([CBQ]; 44), could predict excessive head movement during resting-state fMRI scanning in a cohort of preschool-age children. We focused our temperament analyses on the three CBQ composite scores: effortful control, negative affectivity, and extraversion/surgency. Effortful control re ects a child's ability to control attentional processes and regulate behaviors, such as their ability to maintain focus on a task. Negative affectivity measures children's negative emotional responses, such as a negative response to an adverse, unique, or high-intensity event. Extraversion/surgency describes how extroverted or outgoing a child may be, for example, how willing a child may be open to new experiences (45). Guided by previous research reporting temperament differences between CWS and controls as well as the relationship between temperament and MRI tolerance in children, we tested the following hypotheses. First, we expected that CWS would exhibit signi cantly more movement during scanning than controls. Second, we expected that the groups would differ in temperament indices that may affect tolerance of MRI procedures, including effortful control and negative affectivity, because of their relationship to poor attention and adaptability skills (i.e., ability to handle novel experiences) that have been shown to be associated with increased tolerance of MRI scanning (16, 24). These factors are consistent with the effortful control scale on the CBQ that measures self-regulation of behavior and emotions. Our third hypothesis tested whether the extent of movement during MRI scanning is in uenced by temperamental characteristics associated with lower adaptability skills and attentional control, such as low effortful control and negative affectivity scores. We predicted that these temperamental differences would be associated with the most movement artifacts in their scans and that this relationship would be modulated by age and sex.

Participants
This current study included 76 participants (42 CWS, 34 controls) from an ongoing longitudinal project examining the neural bases of developmental stuttering (46). MRI data collected from the rst visits were used in this study from the larger longitudinal study. Details on recruitment, testing, and in/exclusion criteria for this longitudinal study can be found in Chang & Zhu (47). Brie y, all children were monolingual English speakers and scored within two standard deviations of the mean on all standardized speechlanguage assessments and intelligence tests. Participants included in this study were those with complete CBQ scores and resting-state fMRI (rs-fMRI) datasets collected. The age range of participants was 37-86 months (3-7 years) at the time of scanning (M = 57.99, SD = 14.03). See Table 1 for participant demographics.
All study procedures were approved by the Institutional Review Board at Michigan State University's in accordance with the Declaration of Helsinki. Informed consent/assent was obtained from all children and their parents before participation. Children were given small prizes and stickers for participation and received nominal remuneration.

Behavioral Testing
Speech and language testing All children completed a battery of cognitive and speech-language assessments, as detailed in Chang & Zhu (47 (51,52). Regardless of group, all children scored within the typical range for their age on these standardized assessments. Consistent with previous studies, a child was considered to be stuttering if they met the following criteria: 1) three or more stuttering-like dis uencies per 100 syllables, 2) a diagnosis of 'very mild' or greater based on the index score from the Stuttering Severity Instrument ([SSI-4]; 53) during a narrative story-telling task and a conversation sample collected with a certi ed speech-language pathologist (54), and 3) expressed concern from parent and clinician impression consistent with stuttering diagnosis.

Measures of temperament
Parents completed the 94-item CBQ (44), which assesses the child's temperament within the previous 6 months. This questionnaire uses a 6-point Likert scale to evaluate children between the ages of 3-6 years. Responses to each of the 94 questions are then grouped to form 15 subscales, including activity level, anger/frustration, approach, attentional focusing, discomfort, falling reactivity and soothability, fear, high-intensity pleasure, impulsivity, inhibitory control, low-intensity pleasure, perceptual sensitivity, sadness, shyness, and smiling and laughter (for a description of each variable, see 44). These aforementioned subscales are used to calculate three CBQ composite scores were calculated: effortful control, negative affectivity, and extraversion/surgency. These measures were used in the statistical analyses described below.

MRI data acquisition and head motion measures
Before the MRI visit, all children underwent an MRI training protocol that involved desensitization to the sights and sounds of the scanner environment using games and by visiting a mock MRI scanner. An MRI training protocol that was developed and tested for young children as reported in Theys et al. (55) was implemented to ensure that all children were ready and willing to participate in the subsequent MRI scanning session. On the day of the MRI scan, a trained clinician or graduate student sat next to the child at all times during the scanning session to monitor the child for excessive movement or distress, at which point scanning was stopped and restarted as tolerated. All scans were acquired using a 3T GE Signa HDx MR scanner (GE Healthcare) with an 8-channel head coil in the Department of Radiology at Michigan State University. Each scan session included a 10-minute rs-fMRI scan, which generally occurred after the structural MRI (3D IRFSPGR) and diffusion tensor imaging (DTI) scans (45 minutes total). The rs-fMRI images were collected with echo-planar imaging using the following parameters: 38 contiguous 3-mm axial slices in an interleaved order, echo time = 27.7 ms, repetition time = 2500 ms, ip angle = 80°, eld of view = 22 cm, matrix size = 64 × 64, ramp sampling, with the rst four data points discarded, and each volume of slices acquired 164 times (47). Each rs-fMRI scan was co-registered to the rst volume using rigid body rotation. The movement parameters obtained in the co-registration step were used to calculate the frame-wise displacement (FD), an estimate of volume to volume head movement (56) using AFNI's 1d_tool.py and 1deval. For each participant, the percentage of volumes with FD > 0.5mm was used to determinate the degree of excessive head motion.

Data analysis
Between-group differences All statistical analyses were run using IBM SPSS Statistics, version 26. We rst used separate independent samples t-tests to compare differences between CWS and controls in (a) movement during rs-fMRI scanning and the three CBQ composite scores: (b) effortful control, (c) negative affectivity, and (d) surgency. The effortful control variable did not meet the assumption of homogeneity of variances, as tested by Levene's Test for Equality of Variances (F = 10.77, p = .001), so a Mann-Whitney U test was used to compare CWS and controls for this variable.

Regression analyses
A linear regression analysis was used to further examine group differences in movement during rs-fMRI scans with age and sex entered as covariates. Due to stuttering status being a nonsigni cant predictor in this model (Table 2) and the lack of difference in movement in children who stutter and controls (Table 1), the two groups were collapsed for the subsequent analyses. First, a linear regression assessed the effect of age and sex on head movement during rs-fMRI (Table 3). Additionally, three separate linear regression analyses were conducted to assess the effects of each of the three temperament variables as well as age and sex on head movement during rs-fMRI scanning (Tables 4-6). Lastly, regression models that examined the effects of temperament were conducted separately in girls and boys (Table 7). Individual movement data were transformed using a square root transformation due to not meeting normality of residuals for all regression analyses. All movement analyses were performed using the transformed data.

Results
Between-group differences Children who stutter and controls did not differ signi cantly in age, socioeconomic status, movement during rs-fMRI scanning, negative affectivity, surgency, or effortful control (Table 1). Multiple linear regression was used to examine the effects of age, sex, stuttering status, and their interactions on head movements. Age and sex were both signi cant predictors of head movement (Table 2). Age was negatively associated with head movement; not surprisingly, older children tended to move less than their younger peers. Boys tended to have more volumes with signi cant movement than girls in this model. However, stuttering status (i.e., group) was not signi cantly associated with head movement, even when accounting for age and sex. Therefore, data from the two groups of children were combined for subsequent analyses.   (Table 3). Age was negatively associated with head movement in our model. Boys had more volumes with signi cant movement than girls in the study. Relationship between temperament and head movement Three separate multiple linear regression analyses were conducted to examine whether the three CBQ composite variables (surgency, effortful control, negative affectivity) predicted head movement. Each regression model included ve predictors: main effect of the CBQ variable, sex, age, the interaction between age and the CBQ variable, and the interaction between sex and the CBQ measure (Tables 4-6, Figs. 1-2). The three-way interaction between age, sex, and CBQ variable was not included because the interaction between age and sex was not signi cant in the analysis presented in Table 2. The rst model examined the CBQ variable surgency. The overall model was signi cant (F(5) = 3.250, p = .011, R 2 =. 188); however, none of the predictors were signi cant ( Table 4). The second model examining the effect of effortful control was signi cant (F(5) = 4.691, p = .001, R 2 = .251). Effortful control, sex, and the interaction between sex and effortful control were the signi cant predictors of movement in the model ( Table 5). The signi cant interaction term between effortful control and sex indicates that the relationship between effortful control and movement extent varies by the sex of the child; therefore, two additional regressions models were run, separately in boys and girls. These analyses showed that there was a signi cant negative correlation between effortful control and head movement in boys, but not in girls ( Table 7; Fig. 1). The third model predicting movement based on negative affectivity was also signi cant (F(5) = 4.459, p = .001, R 2 = .242). While sex alone was not a good predictor of movement, negative affectivity and its interaction with sex were signi cant predictors of movement (Table 6). Separate analyses for boys and girls showed that negative affectivity did not predict more head movement in either sex (Table 7; Fig. 2).

Discussion
The present study examined the effects of clinical group (stuttering status), age, sex, and temperament on the extent of head movement during resting-state functional magnetic resonance imaging (rs-fMRI) scanning in preschool-age children. We predicted that children who stutter (CWS) would exhibit more movement during scanning than the control group, as measured by percent of total MRI volumes that contained head movement exceeding 0.5mm. Instead, results indicated no signi cant difference between the two groups of children (Table 1). Furthermore, CWS did not differ signi cantly from the control group on any of the temperament composite measures indexing negative affectivity, surgency, or effortful control.
Based on previous ndings from studies that examined movement during MRI in children (e.g., [11][12], we predicted that effortful control and negative affectivity would be associated with movement during rs-fMRI. CWS did not differ from controls in any temperament score measured by the CBQ, and stuttering status was not a signi cant predictor of movement (Tables 1, 2). Hence, we collapsed the groups for our subsequent analyses examining the predictive effect of temperament on movement. We found that two of the CBQ variables -effortful control and negative affectivity -signi cantly predicted head movement from these analyses.
Both effortful control and negative affectivity had signi cant sex*temperament interactions, suggesting that the effect of both effortful control and negative affectivity on movement is different in boys and girls (Tables 5-6). For example, while examining temperament in separate models for boys and girls, a signi cant negative correlation between effortful control and movement was observed for boys but not in girls (Table 7, Fig. 1). Boys with higher effortful control scores, re ecting better regulation of attentional resources during scanning, moved less during rs-fMRI procedures. Alternatively, boys who showed higher negative affectivity showed a trend of greater movement.
One possible explanation for the disparate relationship between effortful control and head movement among the two sexes in this study is that effortful control matures at different rates in boys compared to girls (for a discussion, see 64). Boys tend to have a more protracted developmental trajectory for effortful control than girls (e.g., 65-67). The difference of maturity rates in some effortful control skills between boys and girls in this study may have in uenced the different relationship between effortful control and head movement. Additionally, another possible explanation for the pattern of difference between sexes is that effortful control mediates the development of externalizing behavior in boys but not girls (e.g., 68).
These previous ndings may explain the observed signi cant relationship between effortful control and movement in boys but not girls in this study. For example, girls may be able to comply with rs-fMRI procedures regardless of their effortful control scores.
While negative affectivity was a signi cant predictor of movement and sex modulated this relationship (Table 6), it was not a signi cant predictor of movement for either boys or girls when they were examined separately ( Table 7). The data show a trend for more movement in boys with higher negative affectivity scores (Table 7; Fig. 2). This result suggests that boys who generally experience higher levels of negative affectivity were less able to tolerate scanning. Additionally, the signi cant interaction between sex and negative affectivity suggests that the relationship between negative affectivity and movement in girls is different from that of boys (Table 6). Negative affectivity is related to feelings of distress during unique or high-intensity experiences. Higher CBQ negative affectivity scores in children in this study could indicate a higher negative response to the novelty of the scanning environment, which may have, in turn, translated into increased movement. Previous studies have found that higher levels of negative affectivity in childhood are associated with more externalizing problems, such as behaviors that are seen as disruptive or problematic, in childhood and later in development as teenagers (69, 70).
Our nding that CWS do not differ signi cantly from controls on temperament scores as measured by the CBQ was somewhat surprising. While previous research reported differences between CWS and their control peers on various tasks or instruments that measure temperament (for a review, see 57), other studies have failed to nd differences between these groups (58-61). It is important to note that the studies of temperament in CWS vary greatly in their methods of sampling (e.g., the children recruited through a clinic vs. from a community sample), age range, and as pointed out by Alm (62), sample sizes as well as the speci c test to measure temperamental differences. Many studies recruit CWS from speech and language clinics (e.g., 35,38,40,41) and would have included children currently undergoing therapy, who may be more likely to present with comorbid clinical disorders such as attention de cit hyperactivity disorder, language delays, and affective disorders. In addition, preschool age CWS who seek therapy may exhibit higher negative reactions to stuttering than those who do not seek treatment for their stuttering.
Thus, CWS who are recruited primarily from clinics have a higher likelihood of differing in temperament measures relative to those that do not seek treatment. Currently, it is di cult to compare the results of this study to others that collected data from older and/or wider age ranges (35,38,39) or studies using different temperament measurements than the CBQ used in this study (38,40,41,43,63). More studies are needed to con rm whether temperament factors differ between CWS and controls, taking into account varied methods of sampling procedures.
Although ndings from the current study were based on rs-fMRI data quality, they can be generalized to brain MRI in general.

Limitations
The data from this study were collected from a longitudinal study of developmental stuttering. These results may be less generalizable to children scanned for purely clinical purposes. For example, our study participants may have certain temperamental characteristics associated with willingness to volunteer in a research study, and in an MRI study in particular. Therefore, due to self-selection, our study may be most relevant to research environments where children volunteer to participate. Future studies involving larger sample sizes, as well as diversity in the pediatric neurodevelopmental populations explored in the study, are necessary to replicate and expand the ndings reported in this study and understand how different conditions may predispose children to tolerate rs-fMRI scanning or not.
Our interpretation of how temperament affects movement during scanning is discussed within the context of previous research that explored disruptive externalizing behaviors, such as those that would get a child in trouble at school or warrant psychological evaluation and intervention. These conditions are quite different from the strict procedures involved in rs-fMRI scanning, which can be challenging even for adults. In this study, excessive movement was de ned by volumes that contained movement greater than 0.5mm. This value was chosen as a stringent movement parameter; however, research teams may be able to accept more or less movement depending on their analyses. This stringent threshold leaves a very small margin for error for these participants. Therefore, an inability to comply with these procedures should not be compared to behaviors of clinical signi cance or any typical behaviors that may be disruptive to everyday life.

Conclusions
We found that in a sample of young children (3-7 years) who underwent rs-fMRI scanning for research purposes, the best potential predictors of head movement during scanning were age, sex, and the temperament factors effortful control and negative affectivity. If temperament is used as a predictor of rs-fMRI tolerance, or MRI in general, in young children, researchers should also consider each temperamental factor differently depending on the sex of the child in their study. Results from this study suggest that assessing temperament may help identify children who could bene t from additional time to desensitize to the MRI environment and scanning procedures. Ultimately, this may lead to improved quality and increased quantity of useable brain scans acquired from young children, which is crucial to furthering our understanding of brain development in children with and without neurodevelopmental disorders.

List Of Abbreviations
ADHD  Figure 1 Sex differences in the relationship between effortful control and movement. For boys, effortful control was negatively correlated with movement. This relationship was not signi cant in girls.