2.1 Participants
A total of 83 children (41 stuttering, 26 boys; 42 controls, 21 boys) between 3 and 11 years of age participated. All were monolingual native North American English speakers without concomitant developmental disorders (e.g., dyslexia, attention-deficit/hyperactivity disorder, learning delay, psychiatric conditions). All children underwent careful screening to ensure typical speech and language developmental history except for the presence of stuttering in the experimental groups. The children who stutter and controls were matched in age, handedness (Oldfield, 1971), and socioeconomic status (Hollingshead, 1975). While most participants were strongly right-handed, 6 children were left-handed (2 persistent, 1 recovered, 3 control) and 7 ambidextrous (4 persistent, 1 recovered, 2 control). All participants were tested on a battery of standardized speech, language, and cognitive tests, audiometric hearing screening, oral-motor screening, and cognitive evaluations. The tests included the Peabody Picture Vocabulary Test (PPVT-3; Dunn, Dunn, & Lenhard, 2007), Expressive Vocabulary Test (EVT-2; Williams & Williams, 2007), Goldman-Fristoe Test of Articulation (GFTA-2; Goldman, 2000), Wechsler Preschool and Primary Scale of Intelligence (WPPSI-III; for children 2: 6–7:3; Wechsler, 2002), and Wechsler Abbreviated Scale of Intelligence (WASI; for children aged 7 and up; Wechsler, 1999). Children were excluded if scores fell below two standard deviations (SD) of the mean on any standardized assessments. The average test scores for each group are listed in Table 1.
Stuttering severity was assessed by collecting samples of spontaneous speech elicited through storytelling and conversational tasks with a parent and a certified speech-language pathologist. These samples were video recorded for further offline analyses. We calculated percent stuttered utterances per number of syllables based on narrative samples containing a conversation with the clinician and a monologue elicited with storytelling with a pictures-only book (‘Frog, where are you?’; Mayer, 1969). In addition, the Stuttering Severity Instrument (SSI; Riley, 2009) was used to examine the frequency and duration of disfluencies occurring in the speech sample, as well as any physical concomitants associated with stuttering; all of these measures were incorporated into a composite stuttering severity rating. To determine the measurement reliability of the SSI scores, an intraclass correlation coefficient was calculated based on the ratings from two independent judges on children’s speech samples.
While all children who stutter were diagnosed with stuttering at the initial study visit, they were categorized as recovered or persistent based on measurements acquired in subsequent early visits that occurred up to 4 times for each child. A child was categorized as persistent with an SSI-4 score >10 at two consecutive follow-up visits, and the onset of stuttering had been at least 36 months prior to his most recent visit. A child was considered recovered with an SSI-4 score ≤10 (corresponding to “very mild”) at two consecutive follow-up visits. Such determination also required the consideration of percent occurrence of stuttering-like disfluencies (%SLD) in the speech sample (≥3 for persistent) as well as clinician and parental reports. Similar criteria were used to determine recovery or persistency in previous studies (Yairi & Ambrose, 1999). Using these criteria, we identified 13 recovered children (7 boys), hereafter “recovered” and 28 persistent children (18 boys), hereafter “persistent”. For controls, the inclusion criteria included never having been diagnosed with stuttering, no family history of stuttering, lack of parental concern for their child’s speech fluency, with %SLD <3%.
Table 1 Demographic information and behavioral test scores of children who stutter and control participants included in this study
|
Controls
n = 42 (21 boys)
|
|
Persistent
n = 28 (18 boys)
|
|
Recovered
n = 13 (8 boys)
|
|
Mean (SD)
|
Range
|
Mean (SD)
|
Range
|
Mean (SD)
|
Range
|
Age
|
6.52 (2.03)
|
3.25-10.75
|
6.52 (1.95)
|
3.08-10.33
|
5.77 (2.31)
|
3.66-11.00
|
SES (Maternal Education)
|
6.36(0.61)
|
5–7
|
6.21 (0.83)
|
4-7
|
6.30 (0.63)
|
5-7
|
Full-Scale IQ a
|
114.68 (14.33)
|
84-144
|
106.78(14.30)
|
81-138
|
104.00 (15.47)
|
88-130
|
Performance IQ
|
111.58 (15.92)
|
77-145
|
107.85 (13.24)
|
79-135
|
101.53 (14.91)
|
86- 134
|
Verbal IQ a
|
117.04 (14.93)
|
87-153
|
105.32 (13.83)
|
77 - 137
|
106.00 (15.54)
|
88-136
|
PPVT c
|
118.26 (13.61)
|
95-151
|
109.79 (12.33)
|
87-147
|
110.69 (15.44)
|
86-134
|
EVT c
|
115.65 (14.33)
|
90 - 149
|
106.57 (11.66)
|
89-134
|
106.84 13.94)
|
87-137
|
GFTA
|
105.31 (7.95)
|
81-123
|
101.86 (8.58)
|
77-118
|
107.38 (8.32)
|
96-121
|
%SLD b
|
1.08 (0.87)
|
0.0-3.23
|
6.88 (6.55)
|
1.10-30.20
|
4.62 (3.33)
|
0.20-12.0
|
%OD
|
4.93 (2.67)
|
0.0- 13.73
|
5.61 (2.91)
|
1.00-12.70
|
4.60 (1.96)
|
2.20- 9.0
|
SSI-4 at initial visit
|
N/A
|
N/A
|
21.18 (7.96)
|
12-48
|
16.00 (6.00)
|
6-28
|
a Controls exhibited significantly higher scores than the persistent and recovered CWS groups.
b Both persistent and recovered groups exhibited significantly higher scores than in controls.
c Controls exhibited significantly higher scores than the persistent CWS group.
SD, standard deviation; SES, socioeconomic status; IQ, intelligence quotient; PPVT, Peabody Picture Vocabulary Test; EVT, Expressive Vocabulary Test; GFTA, Goldman-Fristoe Test of Articulation; SSI-4, Stuttering Severity Instrument Edition 4; %SLD, stuttering-like disfluencies (e.g., sound-syllable repetitions, word repetitions, sound prolongations) occurring per 100 words during conversational speech; %OD, other disfluencies (e.g., interjections, phrase repetitions) occurring per 100 words during conversational speech.
In addition to the speech-language and cognitive tests, all children went through mock scanner training during a separate visit to familiarize them to the scanner environment and procedures and to practice keeping still while lying inside the bore for stretches of time. Recordings of MRI scanner noise were played during this session so that children were aware that they would hear loud sounds during scanning. This session was repeated in some children as needed. All children were paid a nominal remuneration and were given small prizes (e.g., stickers) for their participation. All procedures used in this study were approved by the Michigan State University Institutional Review Board.
2.2 MRI acquisition
MRI scans for this study were acquired on a GE 3T SignaVR HDx MR scanner (GE Healthcare) with an 8-channel head coil. During each scanning session, 180 T1-weighted 1-mm3 isotropic volumetric inversion recovery fast spoiled gradient-recalled images, with CSF suppressed, were collected to cover the entire brain with the following parameters: time of echo 3.8 ms, time of repetition of acquisition 8.6 ms, time of inversion 831 ms, repetition time of inversion 2,332 ms, flip angle 8 degrees, and receiver bandwidth +/-20.8 kHz. After collecting T1 data, high-order shimming procedures were performed to improve magnetic field homogeneity. The dMRI data were obtained with a dual spin-echo echo-planar imaging sequence for 12 minutes and 6 seconds with the following parameters: 48 contiguous 2.4-mm axial slices in an interleaved order, field of view 22 x 22 cm, matrix size 128 x 128, number of excitations =2, echo time 77.5 ms, repetition time 13.7 s, 25 diffusion-weighted volumes (one per gradient direction) with b=1,000 s/mm2, one volume with b=0 and parallel imaging acceleration factor=2. A member of the research staff sat inside the scanner room next to the child being scanned for the duration of the scans to monitor participant comfort and to ensure the child was able to cooperate with scanning protocols. Children watched a movie to help them stay still during the acquisition of volumetric T1-weighted scans and dMRI scans.
2.3 Diffusion MRI data preprocessing
Diffusion MRI data were preprocessed individually using the open-source software mrDiffusion (https://github.com/vistalab/vistasoft/tree/master/mrDiffusion) implemented in MATLAB R2017b. Eddy current distortions and subject motion in the diffusion-weighted images were removed by a 14-parameter constrained non-linear co-registration algorithm based on the expected pattern of eddy current distortions (Rohde et al., 2004). Each diffusion-weighted image was registered to the non-diffusion (b0) image; the b0 image was registered automatically to the T1 image, which had been aligned to the canonical MNI template. The combined transformation, incorporating both eddy-current correction and anatomical alignment, was applied to the raw diffusion data, and the transformed images were resampled at 2 × 2 × 2 mm isotropic voxels. Diffusion tensors were then fit using a robust least-squares algorithm. Eigenvectors and eigenvalues of the tensor were extracted. FA was calculated as the normalized standard deviation of the eigenvalues of the diffusion tensor, and MD was calculated as the average of all three eigenvalues (Pierpaoli & Basser, 1996). Head motion was quantified in each participant by calculating the degree of motion correction in each volume relative to prior volume (Bruckert et al., 2019). Average displacement across participants 0.046+/- 0.030, and there was no group difference in the displacement (p>0.05).
2.4. Tract identification and segmentation
Five cerebellar peduncles (CPs) (bilateral ICP, bilateral SCP, and MCP) were identified and quantified using the open-source software Automated Fiber Quantification (AFQ) package (Yeatman et al., 2012). AFQ consists of three main processing steps: (1) whole-brain fiber tractography, (2) automatic tract segmentation based on template region of interest (ROIs) warped to native space, and (3) automatic tract quantification and cleaning. A whole-brain fiber group was tracked using a deterministic streamlines tracking tractography (STT) algorithm (Mori et al. 1999). Based on previous dMRI studies in children (Bruckert et al., 2019), the tracking algorithm was seeded with a white matter mask defined as all the voxels with FA value greater than 0.15. Tracking proceeded in both directions along the principal diffusion axes and stopped when FA estimated at the current position dropped below 0.10 or when the angle between the last path segment and next step direction was greater than 30°. Fiber tract segmentation was based on waypoint ROIs, which were first defined on the JHU MNI T1 template and then back-transformed to each participants’ native space (Bruckert et al., 2019; Jossinger et al., 2021). The core of the tract was calculated by defining 30 sample points along the tract and computing the robust mean position of the corresponding sample points. After tract segmentation, an automated cleaning algorithm was used to remove fiber longer than 1 standard deviation from the mean fiber length and spatially deviated more than 4 standard deviations from the core of the tract.
A small number of children who stutter were excluded from each analysis because a tract could not be segmented or did not conform to anatomical norms. One child who stutters was excluded from the MCP, 1 from the right SCP, and 2 from the left SCP analyses. Diffusion properties (FA, MD) were calculated at 30 equidistant nodes along the central portion of each fiber tract bounded by the same two waypoint ROIs used for tract segmentation (Bruckert et al., 2019; Jossinger et al., 2021; Yeatman et al., 2012). Figure S1 shows the tracts of interest identified in one representative participant and visualization of FA profiles in each group and each CP.
2.5. Statistical analysis
Analyses were conducted on the core of each CP tract between the two waypoint-ROIs used for tractography.
Group differences. Mean tract-diffusion indices (FA and MD) were calculated by averaging values of all 30 nodes for each participant and in each peduncle (hereafter referred to as “tract-FA and tract-MD”). Independent t-tests were used to compare tract-FA and tract-MD between groups (CWS, controls) in the MCP, bilateral ICP, and SCP. In addition, we evaluated group differences of the local diffusivity measures along the 30 nodes in each tract (hereafter, node_FA, node_MD), which may provide better sensitivity compared to the mean FA and MD for the whole tract. These along-tract group comparison statistics were corrected using a non-parametric permutation test (p<0.05) and controlling the family-wise error (FWE) corrected alpha at 0.05 (Nichols & Holmes, 2002).
Linear regression and correlations. To account for the additional factors that could contribute to the group differences of diffusion properties in the CPs, a linear regression model was built to test group differences when age, sex, and IQ were included as covariates. Meanwhile, Pearson correlation coefficients were used to examine age correlations with diffusion properties of CPs and to test whether diffusion properties of CPs showing significant group differences were associated with stuttering severity.