4.1 Participants
All participants were recruited between November 2015 and August 2017. The participants signed an informed consent and the study was in accordance with the Declaration of Helsinki. The Institutional Review Board (IRB) of the University of Minnesota approved the study.
Adults with stroke were included if they (i) were between 18-85 years old, (ii) were at least 6 months after a stroke that occurred in adulthood, (ii) had a stable ischemic infarct with a lesion in the right middle cerebral artery area, with resulting left hemiplegia/paresis, (iv) did not receive ongoing rehabilitation at the time of testing, (v) were able to read, hear and comprehend instructions in English. Exclusion criteria were: (i) contra-indications for Magnetic Resonance Imaging (MRI) scanning; (ii) other medical conditions precluding full participation, including (other) brain injuries/illnesses, cognitive impairment, severe unilateral spatial neglect or severe proprioceptive loss, hindering them to feel finger movements on the affected side, severe apraxia, severe aphasia, or contractures that restricts them from keeping the outstretched arm in a relaxed position.
Healthy adults of similar age than adults with stroke were recruited through fliers and through the University website. Healthy participants had to be medically stable; able to read, hear and comprehend instructions in English. We excluded healthy adults who had brain injuries, who had complete proprioceptive loss, contractures in the arm, who were medically unstable, illiterate in English, having cognitive impairments, or for whom MRI scanning was contra-indicated.
4.2 Research design
This is an exploratory pilot study with single arm pre-post experimental design in which participants with stroke were tested with behavioural tests, structural and resting-state functional MRI at baseline; behavioural test, structural and resting-state functional MRI after 6 weeks of therapy (CMR, 3 times a week for 45min/session), and behavioural tests at follow-up, 1 year after the therapy has ended. Healthy volunteers were recruited as a control group and were tested once on behavioural tests, structural and resting-state functional MRI.
4.3 Clinical assessments
All participants (i.e. healthy adults as well as adults with stroke) underwent screening tests, i.e., Bell’s test for neglect (cut-off <29/35)[50], Mini-Mental State Examination-short version MMSE®-2-BV™ (cut-off <13/16)[51], Apraxia Screen test (cut-off <5/12 for severe apraxia),[52] Aphasia Rapid Test [cut-off >21/26 (lower score is better outcome)],[53] Edinburgh Handedness Inventory[54], the National Institutes of Health Stroke Scale (NIHSS)[55], Raven’s Progressive Matrices to measure abstract reasoning,[56] and the Corsi Block Tapping Task[57] to test visuo-spatial short-term memory, exteroceptive and proprioceptive sensibility (tested clinically for position and motion sense), two-point discrimination, in stereognosis, in which participants need to correctly identify objects by feeling the object in the hand with eyes closed,[58] and Numeric Rating scale[59] for pain in upper limb. Participants also completed medical and general health questionnaires.
Motor function of the affected arm was evaluated with the “Motor Evaluation for Upper Extremity in Stroke Patients” (MESUPES) in which scores are given for the perceived muscle tone during passive arm movements; for assisted and active arm movements, and for active hand and finger movements, including dexterity tasks (e.g., rotating a dice with thumb and index finger)[14,60]. Secondary clinical measures were the Jebsen-Taylor Hand Function Test[61], which tests hand dexterity (such as putting beans in a contained with a spoon); Fugl-Meyer upper extremity test[62]; ABILHAND in which participants reflect on how easily daily bimanual tasks can performed[63]; Stroke Impact Scale[64]; EuroQol-5D-5L[65]; Global rating scale of change[66]; Frenchay test[67], which enquires how often participants are engaged in daily activities; and the Warwick-Edinburgh Mental Well-Being Scale.[68] The global rating of change scale (GROC) evaluates the patient’s grading of stroke recovery in terms of upper limb motor function after therapy, ranging from -7 (i.e., a very great deal worse) to 7 (i.e., a very great deal better)[66].
4.4 Cognitive multisensory rehabilitation (CMR)
CMR, whose original Italian name is “riabilitazione neurocognitiva”, is also translated in other publications as cognitive therapeutic exercises[69], Cognitive Sensory Motor Training[19], neurocognitive therapeutic exercise[70], cognitive exercise therapy[18], or (neuro)cognitive approach[71,72].
CMR incorporates conscious perception of body positions and movements during (multi)sensory discrimination exercises[15–17]. The treating therapists, with years of clinical practice and specialized in this therapy, gave 35 minutes of discrimination exercises embedded in functional movements followed by 10 minutes of applying the learned strategies during activities of daily living. The difficulty of the exercises was set at +/-75% success, monitored by the number of correct responses given to an exercise, to promote learning, engage motivation and avoid frustration. CMR uses several types of discrimination exercises: Participants discriminated shapes, length, weight, distance, resistance, textures or compare kinaesthetic information with visual information for integration of multisensory information[9,16]. Solving the discrimination task is combined with reflection and a learning process, prompted by the therapist on how the limb (was) moved or was positioned. Focusing attention on movements helped participants control muscle tone and relax their hand,[73,74] thereby creating the potential for movements to re-emerge. Exercises focused on activating only those muscles relevant for the movement, adapting a correct relaxed sensation of the movement in proper form; relate and compare movement with the affected and unaffected arm; understanding the different movement components and how they relate to each other; and recognizing spatial and temporal cues.
Furthermore, kinaesthetic motor imagery was included to increase correct sense perception and, if present, to decrease pain[17,70]. Depending on the severity of the upper limb motor impairment, exercises transitioned from passive, assisted, active, and to functional movements progressively throughout the therapy. All activities were directly related to functional tasks. Several body joints were included in the exercise to integrate speed and dexterity in functional movements.
4.5 Structural and resting-state functional MRI (fMRI) acquisition and pre-processing
Structural MRI acquisition was acquired using a T1 weighted MPRAGE image [TR=2.5s, TE=3.65ms, 1mm3 voxels], as well as T2 weighted SPACE image [TR=3s, TE=565ms, 1mm3 voxels], and SPACE based FLAIR [TR=5.0s, TE=394ms, 1mm3 voxels] images to quantify lesion extent. The 3T Siemens auto-align longitudinal repositioning system was used to ensure comparable head positioning across the scan sessions. Structural Preprocessing was performed using the containerized FMRIPREP[75] version 1.2.2 with standardized BIDS formatted NIFTI data.[76] Each T1-weighted (T1w) volume was corrected for intensity non-uniformity (INU) using N4BiasFieldCorrection[77] and skull-stripped using antsBrainExtraction.sh using the OASIS template. Brain surfaces were reconstructed using recon-all from FreeSurfer,[78] and the brain mask estimates were refined with a custom variation of the method to reconcile ANTs-derived and FreeSurfer-derived segmentations of the cortical gray-matter of Mindboggle.[79] Spatial normalization to the ICBM 152 Nonlinear Asymmetrical template[80] was performed through nonlinear registration with the antsRegistration tool of ANTs,[81] using brain-extracted versions of both T1w volume and template. Additionally, for participants with focal brain lesions, a lesion mask was added during spatial normalization to standard space.[82] Brain tissue segmentation of cerebrospinal fluid (CSF), white-matter (WM) and gray-matter (GM) was performed on the brain-extracted T1w using FAST.[83]
Resting-state functional connectivity (FC) is a measure of the temporal correlation of activation between spatially separate brain regions.[84] fMRI acquisition for resting-state was obtained with a T2*-weighted multiband echo planar acquisition tipped 30 degrees relative to the AC-PC plane. This sequence was designed to measure whole-head BOLD-contrast with optimal temporal and spatial resolution, and to reduce signal dropout [TR=0.871s; TE=34.20ms; Flip Angle=65deg; 72 slices; multiband factor 6; 2.20mm isotropic resolution]. fMRI Preprocessing was also performed with the containerized FMRIPREP[85] 1.2.2 version with standardized BIDS formatted NIFTI data.[76] First, a reference volume and its skull-stripped version were generated using a custom methodology of FMRIPREP. Functional data were slice time corrected using 3dTshift from AFNI[86] and motion corrected using mcflirt.[87] Distortion correction was performed using an implementation of the TOPUP technique[88] using 3dQwarp.[86] This was followed by co-registration to the corresponding T1w using boundary-based registration[89] with 9 degrees of freedom, using bbregister (FreeSurfer). Motion correcting transformations, field distortion correcting warp, BOLD-to-T1w transformation and T1w-to-template (MNI) warp were concatenated and applied in a single step using antsApplyTransforms (ANTs v2.1.0) using Lanczos interpolation. Physiological noise regressors were extracted applying CompCor.[90] Principal components were estimated for the two CompCor variants: temporal (tCompCor) and anatomical (aCompCor). A mask to exclude signal with cortical origin was obtained by eroding the brain mask, ensuring it only contained subcortical structures. Six tCompCor components were then calculated, including only the top 5% variable voxels within that subcortical mask. For aCompCor, six components were calculated within the intersection of the subcortical mask and the union of CSF and WM masks calculated in T1w space, after their projection to the native space of each functional run. Framewise displacement and DVARS[91] were calculated for each functional run using the implementation of Nipype. ICA-based Automatic Removal Of Motion Artifacts (AROMA) were used to generate aggressive noise regressors and to create a variant of data that was non-aggressively denoised.[92]Resting-state fMRI: set-up and statistical analysis
Participants underwent a resting-state fMRI scan of 14.50 minutes with eyes open, maintaining fixation with a restful mind and they were asked to not fall asleep. The Harvard-Oxford cortical atlas was split into regions corresponding to each of the left and right hemisphere, resulting in a total of 96 regions of interest (ROI). The resulting resting-state BOLD-contrast time series were extracted from each ROI from each participant by averaging the voxel-level time courses within each ROI. For each participant, Pearson correlation coefficients were computed to construct functional connectivity metrics between the OP1/OP4 in the affected (i.e., right) hemisphere, thereby serving as the seed region in the analysis and the other ROIs. Two-sample t-tests were used to determine differences between the participants with stroke and healthy controls in the FC between the seed region and each ROI. The Benjamini-Hochberg procedure for controlling the false discovery rate (FDR) was used to adjust the p-values to account for multiple comparisons.
4.6 Behavioural data: Statistical analysis
Behavioural data were calculated with JMP®, Version 13, SAS Institute Inc., Cary, NC, 1989-2007. The results of the Shapiro-Wilk test to verify normal distribution of ratio data informed the decision to conduct two-sample t-tests or Mann-Whitney U tests to compare both groups at baseline. Nominal data were calculated with Chi square (test of independence) between groups (https://www.socscistatistics.com/tests/chisquare/Default2.aspx). Repeated measures ANOVA was used to compare the clinical data in adults with stroke at baseline, after CMR, and at 1-year follow-up. Subgroup analyses for sex, race or ethnicity were not feasible due to the small sample size.