Cortical change after a 2-week novel robotic rehabilitation program in children with prior hemispherectomy: pilot imaging study

Partial preservation of sensory and motor functions in the contralateral extremities after hemispherectomy is likely secondary to cortical reorganization of the remaining hemisphere and can be improved by rehabilitation. This study aims to investigate behavioral and structural cerebral cortical changes that may occur after a 2-week novel robotic rehabilitation program in children with prior anatomic hemispherectomy. Five patients with prior anatomic hemispherectomy (average age 10.8 years; all female) participated in a 2-week novel robotic rehabilitation program. Pre- and post-treatment (2 time points) high-resolution structural 3D FSPGR (fast spoiled gradient echo) magnetic resonance images were analyzed to measure cortical thickness and gray matter volume using a locally designed image processing pipeline. Four of the five patients showed improvement in the Fugl-Meyer score (average increase 2.5 + 2.1 SD. Individual analyses identified small increases in gray matter volume near the hand knob area of the primary cortex in three of the five patients. Group analyses identified an increase in cortical thickness near the hand knob area of the primary motor cortex, in addition to other sensorimotor regions. This small pilot study demonstrates that potentially rehabilitation-associated cortical changes can be identified with MRI in hemispherectomy patients.


Introduction
Hemispherectomy surgery is generally performed in patients with medically refractory epilepsy secondary to a large unilateral hemispheric epileptogenic area secondary to perinatal stroke, hemimegalencephaly, multilobar cortical dysplasia, Rasmussen encephalitis, and Sturge-Weber syndrome. While this procedure often results in improved seizure control with a reported long-term seizure freedom rate of around 66-80%, there can be adverse sensorimotor, visual, language, and behavioral impairments [1][2][3][4][5][6].
Studies have found residual motor function in the contralateral extremities after hemispherectomy [7,8] and have shown cortical reorganization of the sensorimotor cortex in the remaining hemisphere post-surgery to allow for this residual motor function [9][10][11][12][13][14]. In a positron emission tomography (PET) study, post-surgery subjects performing a motor task showed activation foci in the remaining hemisphere in multiple regions: premotor area, supplementary motor area (SMA), secondary sensory area, and lingual gyrus [15]. Subcortical changes have also been proposed to account for the residual contralateral motor function. Results of diffusion tensor imaging (DTI) studies have been variable regarding reinforcement of the corticospinal tract in the remaining hemisphere postsurgery [7,[16][17][18]. In our previously published book chapter, Jacokes et al. reported statistically greater cortical gray matter and white matter volume in the unexcised hemisphere in six post-hemispherectomy patients (age range 10-14) compared to a cohort of 437 age-matched healthy control volunteers suggesting cortical and subcortical changes accounting for the residual contralateral motor function [19].
The current study aimed to pilot the behavioral efficacy of a 2-week high intensity rehabilitation program in five children, who had previously undergone hemispherectomy, and to investigate accompanying sensorimotor reorganization using magnetic resonance imaging (MRI) by examining cortical changes before and after the 2-week training period.

Methods
This study was approved by the local IRB committee. Experimental protocols were approved by local IRB and ethics committees, and all participants and their parents gave informed and signed written consent.
This study included five female patients with rightsided anatomical hemispherectomy with an average age of 10.8 years (range: 10-12 years) at the time of therapy. Eligibility criteria included prior hemispherectomy, > 1 year from last brain surgery, age > 10 years at intervention, wellcontrolled seizures, ability to walk at least 32 feet without another person's assistance, and ability to focus for at least 30 min at a time. Four of the five patients were right-handed at the time of surgery, with one patient being too young to determine handedness. The average age at first hemispherectomy surgery was 5.05 years (range: 0.25-9 years). The list of diagnoses that led to hemispherectomy include Rasmussen's Encephalitis, Sturge-Weber Syndrome, brain trauma secondary to prior neurosurgery, and Taylor Type 2 Cortical Dysplasia. Demographic information can be found in Table 1.
All subjects were enrolled in a high-intensity task-oriented robotic rehabilitation program for 2 weeks. Four consecutive days each week, the subjects received 3 h per day of robot-assisted therapy, consisting of 1 h of "Hocoma Lokomat", 1 h of "inMotion ARM" (Bionik), and 1 h of "inMotion ANKLE" (Bionik) training. "Hocoma Lokomat" is a robot-assisted walking therapy to increase muscle strength and range of motion. "inMotion ARM" and "inMotion ANKLE" are also robot-assisted machines that aid in improving movement of their respective names.
In addition to the robot-assisted therapy, the subjects also participated in value-added activities for an additional 3 h per day consisting of a mixture of 1. recreational therapy, 2. adaptive yoga, 3. virtual reality occupational therapy, 4. occupational therapy targeting activities of daily living, and 5. miscellaneous fun activities such as arts and crafts.
Patients 1 and 2 also received an additional 1 h of training with the wrist robot-IMT Wrist-focusing on wrist flexion, wrist extension, ulnar deviation and radial deviation. The wrist training occurred on the same days as the other robotic training; however, Patient 2 participated in the wrist training for 7 days and Patient 1 participated in all 8 days.
The Fugl-Meyer assessment of physical performance has been used in previous neuroimaging studies regarding hemispherectomy, including the functional MRI plasticity study [8,16,20] and was applied here to assess behavioral changes in motor function before and after the therapy. This study used only the upper extremity motor function subscale of the Fugl-Meyer assessment.

Image acquisition
Subjects 1-5 underwent 3D FSPGR (fast spoiled gradient echo) magnetic resonance imaging of each subject 1 day before and 1 day after the 2-week training period. Imaging was performed on a General Electric (GE) Signa HDx 1.5 Tesla MRI scanner at Rancho Los Amigos National Rehabilitation Center, located in Downey, CA, USA. The following parameters were used: TR: 9.3 ms; TE: 3.7 ms; Flip Angle: 13°; NEX: 1; Field of view: 24 × 26.8 cm; Matrix: 256 × 256 mm; Slice thickness: 1 mm.

Image preprocessing
Patients 1-5 who underwent MRI had varying amounts of the right hemisphere removed during surgery. To standardize the variation among subjects, only the completely

Within subject analyses
Individual changes in gray matter volume were assessed using FSL's SIENA program [21,22]. The pre-and posttreatment MRI scans for each subject were processed to remove non-brain tissues from the image. The images were spatially aligned to each other and resampled into the space halfway between the two [23,24]. Tissue-type segmentation was used to identify the brain/non-brain edge points, and the perpendicular edge displacement between the two timepoints was estimated at each edge point [25]. The mean edge displacement was converted into a global estimate of the percentage of brain volume change between the two timepoints. For visualization purposes, these maps were thresholded at 2% and a clustering criteria of 20 voxels was applied.

Group analyses
Maps of the group differences in cortical thickness were calculated using FreeSurfer [26]. All images were aligned to a common surface template, individual cortical thickness values were resampled onto the surface, and spatial smoothing was performed with a 5 mm full width at half maximum (FWHM) surface based Gaussian kernel. A paired t-test was conducted to identify the common regions where all subjects experienced cortical thickness changes. A vertex wise threshold of p < 0.05 was applied to the statistical maps, and multiple comparison correction was applied using precached simulations at a significance level of α < 0.05.

Behavioral
Most patients had an objective increase in the adapted Fugl-Meyer score after the rehabilitation program (Table 2). One patient exhibited no change and no patient displayed a net negative change. Average change in FM score was calculated to be + 2.6.

Individual gray matter volume changes
For patients 1-5, 3D FSPGR MRI sequencing was conducted to elicit changes in cortical thickness for cluster numbers 16, 26, 33, 35, and 42. There is inter-variability between the subjects. Three of the five patients (Patients 1, 2, and 3) showed an increase in gray matter volume of 2-4% in clusters of 20 voxels or larger near the putative hand knob area of the primary motor cortex (Fig. 1). Two of these patients,  . 1 Three of the five patients (Patients 1-3, panels A-C, respectively) showed an increase in gray matter volume near the hand knob area of the primary motor cortex (indicated by a box on the sagittal slice and crosshairs on the axial slice) following a 2-week intensive robotic intervention Patients 1 and 2, had received additional training with the wrist robot. While these maps are necessarily qualitative, changes in gray matter volume due to this intervention were robust enough to be seen in the primary motor cortex of three of the five patients.

Group maps of cortical thickness changes
The group maps identified widespread changes in cortical thickness in the frontal, temporal and parietal cortices. Of particular interest are clusters located in the premotor area, primary motor area near the putative hand knob area, sensory association cortices, and two distinct clusters in the SMA. The larger cluster in the SMA corresponded to a decrease in cortical thickness while another smaller cluster corresponded to an increase (Fig. 2). The premotor area also decreased in cortical thickness, while the primary motor and sensory association cortex increased. Post hoc analyses of the hand knob region of primary motor cortex showed an average change of 0.21 mm (Cohen's d = 0.77).

Discussion
Here, we examined behavioral changes and cortical gray matter changes following a 2-week training period. Along with improvements in patients behaviorally, the results presented in this paper show changes in both cortical volume and thickness in several regions of the remaining hemisphere in each patient. We performed whole brain analyses rather than region of interest analyses on previously identified regions, so as not to exclude cortical regions that may be involved in sensorimotor reorganization. These findings suggest that while there may not be observable behavioral changes during a future training program, cortical and gray matter changes can be measured to assess any potential efficacy. In the future, imaging can be an indicator for personalized medicine-if gray volume matter is increasing, then training therapy may be working and could potentially signal future functional improvement that is clinically meaningful.
To date, there have been no structural neuroimaging studies examining cortical reorganization with locomotor training in hemispherectomy subjects. Task-oriented training has been shown to lead to structural brain plasticity in the healthy human brain [27]. Changes in gray matter have also been observed, for example, after training in juggling [28], golf [29], computer games [30], spatial navigation [31], and mirror reading [32] along with medical students studying for an exam [33]. Additionally, in clinical populations, treatment effects on gray matter have been observed in chronic fatigue syndrome following cognitive behavioral therapy [34], phonological training in dyslexia [35], and pharmacological intervention in schizophrenia [36]. These changes have been observed with as little as 1 week of training [37], and cover a wide range of ages extending to the late 60s [31,38]. Importantly, motor skill learning is associated with structural brain plasticity, and it has been shown that practice time and performance modulate the extent of structural brain changes evoked by long-term training [39]. Together, this suggests that examining changes in gray matter following locomotor training in hemispherectomy patients allows for investigating the plasticity of the remaining hemisphere, as well as the neuro-efficacy of the training protocol. Further study is required to support this relationship.
Brain plasticity is highly variable among hemispherectomy patients after surgery and may be dependent on factors including, but not limited to, the timing of cortical injury or specific cause of the lesion [15]. Factors that did not affect ambulatory outcome status include age at surgery, age at epilepsy onset, etiology of epilepsy, epilepsy duration (onset to surgery), bilateral abnormalities and pre-op ambulatory status [2]. Other studies state that motor function declines with older age at surgery, specifically within the Rasmussen's group [16]. It may also be possible that changes occur in areas besides the cortex including white matter tracts. There are very few studies on this topic, but Wakamoto and colleagues revealed no evidence of significant reinforcement of the contralateral (to the removed hemisphere) corticospinal tracts in patients with hemispherectomy using DTI measurements [17]. They also suggest that Wallerian degeneration most likely occurs in the ipsilateral motor pathway after surgery [17].
Generally, the results of this study show an objective increase in gray matter thickness that coincides with behavioral improvements of movement. In addition to the intervention being well-tolerated by the patients, the intensive burst of intervention may aid in the chronic phase of recovery. This supports the view that repetitive training may induce change in function as demonstrated by imaging. The group results show cortical thickness changes in regions most closely associated with sensorimotor reorganization including the supplementary motor area, primary motor area, and sensory associative zones in previous studies [11,15,20]. However, previous studies have shown changes in only functional reorganization using fMRI or PET imaging [8,10,11,15,20]. The results of this study add to the view that higher order cortical areas have an increased capacity for sensorimotor reorganization when compared to primary motor areas, by providing evidence of a physical change in cortical thickness.
Interestingly, both the individual maps of gray matter volume changes and the group map of cortical thickness changes identified a cluster near the putative hand knob region with an increase in volume shown in three of the five patients and an overall group increase in the cortical thickness. This region is normally associated with motor control of the contralateral hand. This increase in gray matter after training of the ipsilateral hand, suggests that there may have been some sensorimotor reorganization in the primary motor cortex as well. This finding provides evidence that this form of rehabilitation could be successful in both improving behavior and influencing the reorganization of motor function at the cortical level. Our results are congruent with prior studies that have reported similar findings with changes in cortical thickness on structural imaging and functional connectivity on resting state fMRI imaging demonstrated months to years after a precipitating incident like traumatic brain injury [40][41][42][43][44].
While the results of this pilot study are encouraging, it is important to consider the small sample size when interpreting the results. Expanding this training to a larger sample size will increase statistical power and allow for investigating potential correlation between behavioral improvements and cortical changes. Nonetheless, the results are encouraging in that even in this pilot study of five subjects, both behavioral improvements and changes in cortical thickness can be observed at the group level. The goal of future studies is to expand this training protocol to a larger group of hemispherectomy patients and to continue to investigate behavioral and brain-based changes associated with the rehabilitation.