fMRI Study

The two sub-groups of UCP patients and TD children underwent a session of fMRI in which they were instructed to imagine reaching-grasping actions similar to those they performed during the behavioral task. Participants were presented with videoclips showing a central object (sphere, cube or cylinder) and a box (6 × 6 cm), placed on right or left with respect to the object (Figure 1D). Instructions were to observe the presented context, then a cue (a little arrow) would appear in the central part of the screen, instructing children to imagine themselves performing the action with the right non-preferred hand (to imagine grasping the object and placing it into the box located in the position cued by the arrow). A total of 32 experimental video stimuli were presented in blocks lasting 16 s each. Eight task blocks were presented in a single functional run, with 12–16 s of baseline (fixation of a central white cross) after each block (see Figure 1E). Imaging sessions lasted ~10 min.

Both UCP and TD children performed a training phase before the fMRI session aimed at familiarizing them with the experimental procedure. Visual stimuli were presented by means of a digital video system (60 Hz refresh rate) with a resolution of 800 horizontal pixels x 600 vertical pixels with horizontal eye field of 30° (Resonance Technology, Northridge, CA). Sound-attenuating headphones were used to muffle scanner noise and give instructions to participants. Digital transmission of signal to scanner was via optic fiber. Software E-Prime 2 Professional was used for stimulus presentation. Before the beginning of MRI acquisition, children received precise instructions not to make any voluntary movement during the MI task. An MR-compatible camera (acquisition frequency 60 Hz; MRC Systems) was used to video record actual hand movement or mirror movements during all experimental sessions. The absence of movements during motor imagery performance was investigated by two independent observers at the end of the scanning session.

Anatomical T₁-weighted, anatomical T₂-weighted FLAIR FS-ARC, and functional T₂*-weighted MR images were acquired with a 3-T General Electric scanner (MR750 Discovery) equipped with an 8-channel receiver head-coil. A three-dimensional (3D) high-resolution T₁-weighted IR-prepared fast SPGR (Bravo) image covering the entire brain was acquired and used for anatomical reference. Its acquisition parameters were as follows: 196 slices, 280 × 280 matrix with a spatial resolution of 1 × 1 × 1 mm, TR = 9,700 ms, TE = 4 ms, FOV = 252 × 252 mm; flip angle = 9°. Functional volumes were acquired while participants performed the motor imagery task with the following parameters: 37 axial slices of functional images covering the whole brain acquired using a gradient-echo echo-planar imaging (EPI) pulse sequence, slice thickness = 3 mm plus interslice gap = 0.5 mm, 64 × 64 × 37 matrix with a spatial resolution of 3.5 × 3.5 × 3.5 mm, TR = 2,000 ms, TE = 30 ms, FOV = 205 × 205 mm², flip angle = 90°, in plane resolution = 3.2 x 3.2 mm².

Data analysis was performed with SPM12 (Wellcome Department of Imaging Neuroscience, University College, London, UK; http://www.fil.ion.ucl.ac.uk/spm) running on MATLAB 2018a (The Mathworks, Inc.). Structural images were manually centered and reoriented with functional images to the anterior-posterior commissure axis. The first four EPI volumes were discarded to allow for T₁ equilibration effects. For each subject, all volumes were slice timing corrected, spatially realigned to the first volume and un-warped to correct for between-scan motion. Motion parameters were used as regressors of no-interest in the model to account for translation and rotation along the 3 possible dimensions as determined during the realignment procedure. Individual datasets were excluded if excessive head motion was observed (translation > 3 mm or rotation > 3°). T₁-weighted images were segmented into gray, white matter, and cerebrospinal fluid and spatially normalized to a standard Montreal Neurological Institute (MNI) template for pediatric data from the 4.5 to 18.5 year age (50). The spatial transformation derived from this segmentation was then applied to realigned EPIs for normalization and re-sampled in 2 × 2 × 2 mm³ voxels using trilinear interpolation in space. All functional volumes were then spatially smoothed with an 8 mm full-width half-maximum isotropic Gaussian kernel.

The pre-processed functional data for each participant were entered in single-subject whole-brain analysis (51). Blood oxygen level dependent (BOLD) signal was modeled in a General Linear Model (GLM) by a design matrix comprising onset and duration of each event, according to experimental task. This analysis employed event-related convolution models using the hemodynamic response function (HRF) provided by the software SPM12. We used one predictor of interest that was a boxcar function with duration of motor imagery blocks, containing four trials each. Single subject activation maps were produced using a fixed-effect analysis (FFX) at a statistical threshold of p < 0.001 [with cluster level family-wise error (FWE) rate correction for multiple comparisons]. Anatomical description was performed on the basis of probabilistic cytoarchitectonic maps as implemented in Anatomy toolbox for SPM12 (52).

Two multiple regression analyses were performed at multi-subject level, one for UCP and one for TD group, separately, to look for a linear relationship between brain activity (BOLD signal change) during explicit MI task and MIA score (Pearson correlation coefficient) obtained in the mental chronometry paradigm task with the non-preferred hand. These regression analyses were designed to study responses across the entire brain at a threshold of p < 0.001, after application of FWE correction for multiple comparisons at cluster level.

Lesions were manually delineated on the T₂-weighted FLAIR images, using the MRIcron software (http://www.cabi.gatech.edu/mricro/mricro). Lesions were mapped by two expert neuroradiologists (FB and GC) delineating the boundary of the lesion directly on the image for every single transverse slice, using MRIcron. Both MRI scan and lesion shape were then mapped into stereotaxic space using the normalization algorithm provided by SPM12. After normalization, all lesions were carefully reviewed to ensure that lesion maps accurately reflected the extent of lesions in MNI space. Manual adjustments were made if necessary to better match the MNI template.