Neuroarchitectural Recovery Model of Post-stroke Patients

Overview

The societal burden of stroke patients with persistent neurological deficit is high. It is therefore imperative that the mechanisms of rehabilitation-induced motor recovery be better understood in the hopes of developing more efficacious rehabilitative therapy. The treatment outcomes of people with stroke after rehabilitation vary, with up to 60% of people having residual impairment of the upper limb function. The high variability in rehabilitation-induced recovery prompted researchers and clinicians to develop more efficacious rehabilitative interventions for functional regain in post-stroke patients. However, the mechanisms underlying post-stroke functional regain have not been well articulated. The majority of studies in this area placed a limited scope on associating improvement in functions with changes in the activation of the motor cortices, ranging from normalization to the overactivity of the motor regions. A wider scope, however, needs to include structural changes in the motor cortices, as well as functional and structural changes in other neural substrates, as other non-motor cortices underpin stroke recovery. The results of our pilot study on acute post-stroke patients indicate that both functional and structural brain connections are significantly associated with motor recovery after conventional post-stroke rehabilitation. In addition to sensorimotor cortices, the investigators also found other non-motor areas, such as the superior frontal gyrus and the precuneus, that play important roles in post-stroke rehabilitation-induced recovery. Given the gap in elusive neural processes and in the mechanisms underlying rehabilitation-induced recovery, this proposed study is aimed at providing a better understanding of the functional regain of post-stroke patients by constructing a brain recovery model. As a first attempt, the investigators propose building a basic recovery model based on patients who will undergo constraint-induced movement therapy, a popular evidence-based post-stroke intervention, for capturing training-induced neuroplastic changes. Two groups of chronic stroke patients will be recruited, respectively, for the treatment and control groups. Magnetic resonance imaging (used to map functional and structural brain connections), the clinical assessments of motor impairments, and activities of daily living will be conducted at four time points-namely at the baseline, one week, four weeks, and three months after the treatment commences. The two objectives set for the proposed study are: (1) to characterize the longitudinal changes in functional and structural brain networks, which would differentiate the rate of changes in these networks; and (2) to define the functional and the structural brain network coupling, as well as their contributions to the daily function regain.

A.1. Introduction Between 1990 and 2016, the mean lifetime risk of stroke increased from 23% to 25%.1 Stroke was one disease that resulted in long-term neurological deficits and disabilities1,2, which burden both the family and society. Post-stroke rehabilitation has been designed to promote the regain of daily functions.3 A previous study revealed that the heterogeneity of the neurotrauma due to a stroke, as well as the constraints of neuroplastic responses to clinical interventions, influenced the patient's rehabilitation outcomes.4 For example, substantial variability existed in the gains in motor functions after constraint-induced movement therapy (CIMT). This occurred despite the fact that treatment efficacy had been well established for the intervention.5 High variability in rehabilitation-induced recovery has prompted clinical researchers to develop more efficacious post-stroke rehabilitative interventions.6,7 Other researchers have attempted to derive neurological and clinical biomarkers8-11 for explaining the observed effects of these interventions. A.2. Neuroplasticity underlying stroke recovery Chan (Co-I) and Pang (Co-I) conducted a few randomized clinical trials to test the efficacy of clinical interventions for augmenting the functional regains of post-stroke patients. Chan and his colleagues designed functional training programs based on the motor relearning theory, which showed significantly better motor functions and a higher level of performance in activities of daily living compared with regular practices.12 The second trial was a six-week intervention involving self-regulatory learning training. Post-stroke patients learned the chunking, mental rehearsal, and self-regulation practices of common daily activities, such as laundry and cooking.13,14 Compared with those who received the usual practices, patients in the self-regulatory learning group showed significantly better performance in both trained and untrained daily tasks. More importantly, the results of the untrained tasks suggested that patients in the experimental group managed to generalize what was learned from the training to the new task context and procedures.15 It is noteworthy that the better daily task performance did not associate with changes in the motor functions among the patients. A recent brain imaging pilot study that Chan (Co-I) and colleagues conducted on the self-regulatory training indicated that post-stroke patients showed significant increases in blood-oxygen-level-dependent (BOLD) signals in the right superior and middle temporal gyri in the untrained task trials, as well as in the left orbital frontal, anterior insular, and anterior temporal regions in the untrained-minus-trained task trials. The results of this pilot study suggested that better task performance after self-regulatory training were associated with decision-making, access to semantic memory, and introspective awareness. Post-stroke functional regain, according to these studies, would go far beyond motor functions and self-regulation, which belongs to the executive function. Pang (Co-I) conducted a single-blinded randomized controlled study and showed that an intervention for a dual task, which required the performance of mobility and cognitive tasks simultaneously, could improve the outcomes of a patient with chronic stroke, such as dual-task performance and fall incidence.16 A.3. Neuroarchitectural correlates of motor recovery after post-stroke rehabilitation The results of a pilot study that Hui (PI) conducted on 16 post-stroke patients indicate that both functional and structural brain networks are associated with motor recovery. The changes were found to occur during and after the patients underwent a conventional rehabilitation program in a hospital for up to three months. They included one week as well as one, three, and six months after stroke onset. Longitudinal structural changes in connectivity, derived from diffusion tensor imaging, were observed in the ipsilesional supramarginal gyrus, contralesional anterior cingulate gyrus, superior parietal gyrus, and cuneus (Figure 1). Longitudinal functional changes in connectivity, derived from rest-state functional magnetic resonance imaging (MRI), were also observed in the ipsilesional superior temporal gyrus, pallidum, and thalamus, as well as the contralesional superior, middle, and triangularis inferior frontal gyri; mid-cingulate area; putamen; and pallidum (Figure 2). To better comprehend the structural results, the neuroarchitectural method was used to build network models at the whole brain and regional levels17, as well as their rich-club organizations.18 An increase in the density ratio of the structural connections of rich-club regions occurred (Table 1). These connections were meant to reflect increases in neural communication efficiency among the neural substrates. This was followed by the testing of how structural and functional connections might subserve post-stroke functional recovery. At one week, the investigators found the properties significantly associated with patients' functional regain in the local structural brain network of the ipsilesional postcentral gyrus; paracentral lobule; median and anterior cingulate gyri and putamen; and contralesional superior parietal gyrus, precuneus, dorsolateral superior frontal gyrus, and hippocampus (Figure 3). At one month, it was the local functional brain network of the ipsilesional posterior cingulate gyrus, contralesional precuneus, and superior frontal gyrus that is significantly associated with patients' functional regain (Figure 4). The post-treatment functional gains were also correlated with the rich-club organization of the structural brain connections (Table 2). The results of this pilot study inform the setting of hypotheses in this study. The advantage of using the neuroarchitectural method, such as rich-club organization, rather than conventional fractional anisotropy from diffusion tensor imaging to define longitudinal structural and functional changes is that the brain network that ischemic infarct disrupts is attributable to neurological sequelae.19-21 The relationship between brain changes and recovery should therefore be studied from the perspective of the brain as a network, rather than local regional changes.22 In this study, the investigators plan to apply the neuroachitectural method to build a post-stroke brain recovery model. Taken A.2 and A.3 together, our research work offers insights into the notion that neural substrates beyond the sensorimotor cortices should play important roles in post-stroke recovery. However, they have not been emphasized in previous post-stroke recovery model studies. The investigators also demonstrate the advantages of using the neuroachitectural method, which accounts for longitudinal changes in the brain in terms of structural and functional features, as well as their relationships, at various levels of the brain. The joint characterization of the spatiotemporal changes in structural and functional brain networks would fill a critical knowledge gap regarding the mechanisms underlying post-stroke recovery in response to clinical intervention. This is particularly true for the role that non-sensorimotor cortices would play a role in the recovery process. Furthermore, the knowledge would help to provide a better understanding of the relationships by means of the network coupling strength23 between the dynamic functional and static structural connections in the plastic brain. This claim is supported by our pilot results showing that both types of brain networks were correlated with the rehabilitation-induced recovery and hence their underpinning of responses to post-stroke rehabilitation. The reason for employing network coupling strength in this proposed study is because previous studies showed that its value increased with age24 and decreased with schizophrenia,25 idiopathic generalized epilepsy,23 and small vessel disease with a high white matter lesional load.26 More importantly, the network coupling strength was also found to be correlated with the clinical symptoms,25 the duration23 of the disease, executive function, and memory performance.26 A.4. Neurophysiological correlates of rehabilitation-induced recovery CIMT involves the forced use of the paretic upper limb followed by bilateral upper limb functional training.27 A large number of studies have reported the efficacy of CIMT on improving the motor abilities and daily functions of post-stroke patients.28-30 For instance, a comprehensive review published in 2015 concluded that the clinical effects of CIMT can be supported by the motor learning principles of task- and context-specific and repetitive training.30 However, despite the clinical benefits of CIMT, the neural mechanisms underlying CIMT-induced recovery have not yet been clarified.30 For instance, it is still unclear why improvement in the quality of hand functions involved the recruitment of the ipsilesional superior temporal gyrus after CIMT,31 a neural substrate not directly related to motor functions. The majority of brain imaging studies on CIMT have revealed that post-treatment functional regains were associated with changes in the motor cortices. CIMT-indicated functional regain was correlated with an increase in the size of the ipsilesional motor cortex32,33, as well as activations in the primary sensorimotor5,34-37, supplementary motor5, premotor cortex31, and secondary somatosensory31 cortices. The structural changes associated with the functional regains included an increase in gray matter density in the bilateral sensorimotor cortices and hippocampi,38 as well as bilateral cerebral peduncles and pons.39 Another study revealed structural integrity of the posterior limb of the internal capsule of the cortico-spinal track (CST) relating to CIMT treatment effects.40 One study indicated that the extent of ischemic injury to the CST was a better predictor of CIMT effects compared with infarct volume.41 In this study, CIMT will be used as the intervention for inducing functional improvements in a group of chronic post-stroke patients based on which the recovery model is to be constructed. A.5. Hypothesis and significance The investigators propose constructing a basic brain recovery model for a detailed account of the functional and structural brain changes in a group of chronic stroke patients during the course of CIMT. Two objectives are set for achieving this goal: OBJECTIVE 1: To characterize the longitudinal changes in functional and structural brain networks that would differentiate the rate of changes in these networks. Rationale. Considering that the brain is in continuous interactive states,42 a comprehensive account of the neuroplastic changes that occur over the course of CIMT-induced recovery would be meaningful only when both the structural and functional brain networks are to be jointly modeled in the same stereotactic "framework" (i.e., the same brain template). To achieve this, changes in functional brain networks will be measured using the resting state rather than task-based functional MRI. The main reason is that rest-state data would be less biased by the participants' specific task-taking behaviors, such as motivation, level of attention, compliance, and infarct location.43-46 Resting-state data also require a shorter scanning time and are easier to acquired than task-based data are.47 These advantages are helpful for lowering the potential attrition rate due to the repeated scanning required for the participants. Previous studies on robotic therapy reported significant functional connectivities between the ipsilesional and contralesional motor cortices associated with better therapy-induced functional gains among a group of subacute or chronic post-stroke patients.48,49 Hypothesis 1. Longitudinal changes will take place in the functional and structural connections in the sensorimotor and non-sensorimotor cortices across the baseline as well as one week, four weeks, and three months after CIMT. Sensorimotor cortices would include the primary and secondary sensorimotor cortices. The non-sensorimotor neural substrates would include the dorsolateral prefrontal cortex, cingulate cortices, superior parietal lobule, precuneus, and putamen. Impact. The joint characterization of the longitudinal changes in structural and functional brain networks would fill a critical knowledge gap regarding the mechanisms underlying post-stroke recovery due to CIMT. The results would inform the rates of change among various neural substrates, which could potentially account for individual patient differences in the treatment outcome. In addition, the results would shed light on the potential differences in the temporal courses of functional versus structural brain changes in response to CIMT or other post-stroke rehabilitative interventions. OBJECTIVE 2: To define the functional and structural brain network coupling, and their contributions to the daily function regain from the treatment Rationale. As functional and structural networks are within the same physical brain, both should contribute to the post-intervention regain in function. It is imperative that they are to be measured at the same time points and to be characterized in the same brain template. The neuroarchitectural methods employed in this study will include rich-club analysis for building brain networks at the regional and whole-brain levels, as well as coupling between the two types of networks (see Section B.8. for details) for modeling the functional and structural brain network relationships. Coupling is made possible by computing the correlations between functional and structural brain connections.23 Hypothesis 2. Significant associations will exist between the combined sensorimotor- and non-sensorimotor functional-structural coupling, as well as the CIMT-related functional regain measured using the Fugl-Meyer Motor Scale and Barthel Index. Hypothesis 3. Structural connections within the functional-structural coupling will influence the activation of the functional connections. The influence can be facilitative or inhibitory. Impact. The coupling between functional and structural brain connections will provide an alternative measure of the neuroplasticity underlying rehabilitation-induced recovery. Given the fact that the treatment response could potentially be boosted through the addition of transcranial direct current stimulation to CIMT,50,51 knowledge regarding the coupling between structural and functional connections will inform new targets beyond the sensorimotor systems for brain stimulations to further improve and understand the treatment response. This is particularly important considering the fact that the attention control network (the neural substrates include the cingulate gyrus, inferior frontal gyrus, and temporoparietal cortex52) has been implicated in the improvement in the performance of the motor task when the patient was presented with multisensory stimuli.53 The majority of prior studies focused on the non-invasive stimulation of the motor system and did not consider non-motor neural substrates, such as those in the attention control network.54 B. RESEARCH PLAN AND METHODOLOGY B.1 Experimental design. The proposed study is a longitudinal study of two groups of chronic post-stroke patients, namely a treatment group that will receive CIMT, and a control match-pair group that will receive the usual practices. MRI scans and clinical assessments of motor impairment and daily activity functions will be conducted at four time points: baseline as well as one week, four weeks, and three months after the beginning of CIMT. B.2. Subjects. Based on the effect sizes of the structural changes across time points in the anterior cingulate cortex, superior parietal lobule, precuneus, and putamen gathered in the pilot study (Section A.3.), with the power of the repeated measure and correlations set at 80% and α=0.05, the estimated sample size is 60. Considering that the patient attrition rate for four scans and assessment occasions is 20%, the total sample size for the CIMT group is n=72, and that for the control group is n=36. B. 3. Inclusion/exclusion criteria. Chronic post-stroke patients between 50 and 80 years old, and with the first-time ischemic stroke occurring three months or more after onset will be recruited to participate in the study. The reason for choosing chronic post-stroke patients is to minimize the effect of spontaneous recovery,55 which may confound the treatment-induced recovery. The recruited patients should present with motor deficits in the upper and/or lower extremities due to the stroke as measured using the Fugl-Meyer Motor Scale. The exclusion criteria include patients with voluntary extension ≤ 10° in the metacarpophalangeal or interphalangeal joints, or ≤ 20° in the wrist; severe balance or walking disorders as indicated by the need for assistance in any activities of daily living; significant cognitive decline (score \<16) measured with the The Hong Kong version of Montreal Cognitive Assessment (HK-MoCA);56 or a history of prior stroke, brain neoplasm, intracranial hemorrhages, transient ischemic attacks, diffusion abnormalities due to nonvascular etiology (for example, posterior reversible encephalopathy syndrome or global hypoxic ischemic encephalopathy), or other neurological/psychiatric or medical conditions that preclude active participation in research and/or may alter the interpretation of the behavioral/imaging studies (e.g., dementia, schizophrenia). Chronic stroke patients in the control group will be matched to those in the CIMT group by age, infarct location, and motor and functional status at the baseline. B.4. MRI. All brain scans on the patients will be performed using a 3 Tesla MRI scanner (Prisma, Siemens, Erlangen, Germany) housed in the University Research Facility in Behavioral and Systems Neuroscience (UBSN) at The Hong Kong Polytechnic University. Four scans will be conducted at the baseline and then one week, four weeks, and three months after the treatment commences. Brain scan protocols will make references to those adopted in the Harms et al. human connectome project on aging57. Structural scans use the diffusion-weighted spin-echo echo-planar imaging sequence with b-values of 1500 and 3000 s/mm2 along 92 diffusion-encoding directions. Other imaging parameters are: multiband factor = 4, repetition time/ echo time = 3230/89 ms, field of view = 210 mm, 1.5 mm isotropic (no gap), and acquisition time = 10.7 minutes (with two acquisitions along positive and negative phase-encoding directions). Functional scans (resting state) use a gradient-echo echo-planar imaging sequence while the patient is instructed to rest with the eyes open. Imaging parameters are: repetition/echo time = 800/37 ms, flip angle = 52o, field of view = 208 mm, 2 mm isotropic (no gap), number of dynamics = 488, and acquisition time = 12.8 minutes (with two acquisitions along positive and negative phase-encoding directions). Regional cerebral blood flow (rCBF) will be measured using pseudo-continuous arterial spin labeling. Imaging parameters are post-labeling delay (for the bottom slice)/ label duration = 1642/1500 ms, repetition/echo time = 3580/19 ms, flip angle = 90o, field of view = 215 mm, 2.5 mm isotropic (no gap), 90 pairs of labeling and control images, and acquisition time = 10.6 minutes (with two acquisitions along positive and negative phase-encoding directions). Structural MRI includes a T1-weighted magnetization-prepared rapid gradient echo and fluid-attenuated inversion recovery. In addition, susceptibility-, T2-, and T2\*-weighted imaging will be acquired. B.5. Interventions. Patients will receive CIMT and control intervention at The Hong Kong Polytechnic University. The treatment protocol makes reference to that described in Wolf et al.28,58 and Winstein et al.,59 which consists of a two-week daily rehabilitation program with monitored behavioral shaping and the repetitive task practice of the paretic limb for up to six hours a day during the weekdays, as well as a padded safety mitt worn on the non-paretic limb during 90% of waking hours. Patients assigned to the control group will receive the usual practice program described in Pang (Co-I)16, consisting of flexibility exercises of all limbs, as well as strengthening exercises of the upper limbs without additional cognitive tasks. The research assistant will carry out the implementation of CIMT and the usual practice interventions under the supervision of Chan and Pang (co-I) in this study. B.6. Clinical instruments. The language, cognitive, motor, and functional deficits of patients will be assessed using four clinical instruments. The Hong Kong version of the Oxford Cognitive Screen (HK-OCS)60 is to assess the patient's stroke-specific cognitive impairments, such as visual neglect and apraxia. HK-MOCA56 is to assess the patient's overall cognitive functions. Both HK-OCS and HK-MOCA will be used to screen patients for recruitment. The Fugl-Meyer Motor Scale (FMMS, full score = 100), which includes the assessment of the upper (UE-FM, full score = 66) and lower extremities (LE-FM, full score = 44),61 is to assess the patient's motor recovery in both the upper and lower limbs. The Barthel Index (BI)62 is to assess the patient's regain of self-care and maintenance functions. The FMMS and BI will be used as the outcome measures. Occupational therapy/physiotherapy trainees and a research assistant will conduct all assessments under the supervision of Chan and Pang (co-I). B.7. Data postprocessing and statistical analyses for Hypothesis 1. All postprocessing steps, including whole-brain parcellation, the construction of structural and functional connections, and analysis using the neuroarchitectural method will make reference to those used in the pilot study and reported in previous studies.63-66 Functional and structural brain networks will be constructed using graph theoretical analysis67 and rich-club analysis17 at the regional-brain, whole-brain, and rich-club scales. For testing Hypothesis 1, a linear mixed model will be used to test (1) the effect of time (baseline, one week, four weeks, and three months) and the interaction effect of time on the scores of the clinical assessments and the measures of neural substrates within the identified brain networks. In this analysis, the neural substrate measures within the brain networks and the scores of clinical assessments are responses; occasions (four time points) and group (CIMT vs. control) are the fixed variables; patient subjects constitute the random variable; and age, gender, time after stroke onset, cognitive status, handedness, and those described in Section B.9. are the covariates. The linear model will also be used to test (2) the association between the scores of the clinical assessments and the measures of neural substrates within the identified brain networks. In this analysis, the scores of the clinical assessments are the responses; the network neural substrates measures and group are the fixed variables; patient subjects constitute the random variable; and the covariates are the same as those in (1). Furthermore, the linear model will be used to (3) predict treatment gains by the measures of neural substrates within the identified networks. In this analysis, changes in the scores of the clinical assessments from baseline are the responses; patient subjects constitute the random variable; the network neural substrate measures at baseline are the fixed variables; and the covariates are the same as those in (1) and (2). B.8. Structural-functional connectivity coupling and statistical analyses for Hypotheses 2 and 3. Network coupling strength23 is defined as the correlations between the structural and functional brain connections, which is also known as connectome41. In this study, three types of network coupling strengths will be highlighted: (1) within the sensorimotor cortices, (2) within the non-sensorimotor brain regions, and (3) between the sensorimotor and non-sensorimotor cortices/regions. Besides these three strengths, network coupling strengths will also be determined for the whole brain. These four types of network coupling strengths will be derived for the baseline, one week, four weeks, and three months. For testing Hypothesis 2, the linear mixed model will be used to test (1) the effect of time and the interaction effect of time on the network coupling strengths. In this analysis, the network coupling strengths (positive functional connections) are the responses; occasion and group effects are the fixed variables; subject patients constitute the random variable; and the covariates are the same as those included for Hypothesis 1. The model will also be used to test (2) the association between the scores of the clinical assessments and the four types of network coupling strengths. In this analysis, the scores of the clinical assessments are the responses; the network coupling strengths (positive functional connections) and group are the fixed variables; patient subjects constitute the random variable; and the covariates are the same as those for Hypothesis 1. For testing Hypothesis 3, it involves the differentiation of positive versus negative functional connectivities, as well as testing their effects on the four types of network coupling strengths. The statistical tests will be the same as those for Hypothesis 2, but adding functional connection type (positive versus negative) as a fixed variable. The theoretical basis for adding positive versus negative to the model is to consider the influence of the neural substrates within a task-negative default mode network (DMN) to those that are task positive68. Task-positive neural substrates have been reported as subserving higher-level cognitive processing, such as decision-making and working69. In contrast, neural substrates within the DMN, which formed negative functional connections with those within other functional networks, were found to be associated with a lapse of attention or other cognitive functions70. The differentiation of the network coupling strengths between positive and negative functional connections would be useful for building a more robust post-stroke recovery model. B.9. Covariates. With the goal of accounting for the variability in the load and location of ischemic subcortical infarct, covariates will be included in all data analyses in this study. Alsttot et al. showed that infarct in the brain regions with high centrality has a larger effect on the corresponding brain network,71 and it will be included as a covariate. Another covariate is the total small vessel disease (SVD) score,72 which accounts for the impact of the SVD lesions on the whole brain. Other covariates to be included are comorbidity (such as atrial fibrillation, diabetes, hypertension, and hyperlipidemia) and drug effect (such as antiplatelet, anticoagulant, antihypertensive, and statin). Specific to the analyses of functional connectivity, an additional covariate to be included is the rCBF, which accounts for the potential changes in regional variations in perfusion after a stroke73. B.10. Potential problems and alternative strategies. (1) Although some studies reported the limited recovery potential of chronic post-stroke patients,46 other studies revealed significant motor and functional gains among these patients.74 The main reason for this study to recruit chronic post-stroke patients is to avoid potential biases due to spontaneous recovery in the acute and sub-acute rehabilitative stages9, which may interfere with treatment-induced recovery. (2) The functional connectivity to be modeled in this study is based on the temporal coherence of brain-wide neuronal activities and the integrity of the neural substrates7 across the four time points. They can potentially be confounded by the hemodynamic lag resulting from a stroke75. For the purpose of controlling this, possible regional temporal delays in the resting-state signals will be estimated, and the functional connectome to be derived will be corrected based on the approach that Siegel et al.73 proposed. The hemodynamic lag corrections will be performed prior to the statistical analyses to be conducted on the functional connectivity results as well as their corresponding network metrics.