Dynamic connectivity predicts acute motor impairment and recovery post-stroke

Scaling behaviours of deep learning and linear algorithms for the prediction of stroke severity

Deep learning has allowed for remarkable progress in many medical scenarios. Deep learning prediction models often require 105-107 examples. It is currently unknown whether deep learning can also enhance predictions of symptoms post-stroke in real-world samples of stroke patients that are often several magnitudes smaller. Such stroke outcome predictions however could be particularly instrumental in guiding acute clinical and rehabilitation care decisions. We here compared the capacities of classically used linear and novel deep learning algorithms in their prediction of stroke severity. Our analyses relied on a total of 1430 patients assembled from the MRI-Genetics Interface Exploration collaboration and a Massachusetts General Hospital-based study. The outcome of interest was National Institutes of Health Stroke Scale-based stroke severity in the acute phase after ischaemic stroke onset, which we predict by means of MRI-derived lesion location. We automatically derived lesion segmentations from diffusion-weighted clinical MRI scans, performed spatial normalization and included a principal component analysis step, retaining 95% of the variance of the original data. We then repeatedly separated a train, validation and test set to investigate the effects of sample size; we subsampled the train set to 100, 300 and 900 and trained the algorithms to predict the stroke severity score for each sample size with regularized linear regression and an eight-layered neural network. We selected hyperparameters on the validation set. We evaluated model performance based on the explained variance (R2) in the test set. While linear regression performed significantly better for a sample size of 100 patients, deep learning started to significantly outperform linear regression when trained on 900 patients. Average prediction performance improved by ∼20% when increasing the sample size 9× [maximum for 100 patients: 0.279 ± 0.005 (R2, 95% confidence interval), 900 patients: 0.337 ± 0.006]. In summary, for sample sizes of 900 patients, deep learning showed a higher prediction performance than typically employed linear methods. These findings suggest the existence of non-linear relationships between lesion location and stroke severity that can be utilized for an improved prediction performance for larger sample sizes.

Fronto-striatal dynamic connectivity is linked to dopaminergic motor response in Parkinson's disease

INTRODUCTION

Differences in dopaminergic motor response in Parkinson's disease (PD) patients can be related to PD subtypes, and previous fMRI studies associated dopaminergic motor response with corticostriatal functional connectivity. While traditional fMRI analyses have assessed the mean connectivity between regions of interest, an important aspect driving dopaminergic response might lie in the temporal dynamics in corticostriatal connections.

METHODS

This study aims to determine if altered resting-state dynamic functional network connectivity (DFC) is associated with dopaminergic motor response. To test this, static and DFC were assessed in 32 PD patients and 18 healthy controls (HC). Patients were grouped as low and high responders using a median split of their dopaminergic motor response.

RESULTS

Patients featuring a high dopaminergic motor response were observed to spend more time in a regionally integrated state compared to HC. Furthermore, DFC between the anterior midcingulate cortex/dorsal anterior cingulate cortex (aMCC/dACC) and putamen was lower in low responders during a more segregated state and correlated with dopaminergic motor response.

CONCLUSION

The findings of this study revealed that temporal dynamics of fronto-striatal connectivity are associated with clinically relevant information, which may be considered when assessing functional connectivity between regions involved in motor initiation.

Altered dynamic functional network connectivity states in patients with acute basal ganglia ischemic stroke

BACKGROUND

Dynamic functional network connectivity (dFNC) patterns are successfully able to capture the time-varying features of intrinsic fluctuations throughout a scan. We explored dFNC alterations across the entire brain in patients with acute ischemic stroke (AIS) of the basal ganglia (BG).

METHOD

Resting-state functional magnetic resonance imaging data were acquired from 26 patients with first-ever AIS in the BG and 26 healthy controls (HCs). Independent component analysis, the sliding window method, and the K-means clustering method were used to obtain reoccurring dynamic network connectivity patterns. Moreover, temporal features across diverse dFNC states were compared between the two groups, and the local and global efficiencies across states were analyzed to explore the characteristics of the topological networks among states.

RESULTS

Four dFNC states were characterized for comparison of dynamic brain network connectivity patterns. In contrast to the HC group, the AIS group spent a significantly higher fraction of time in State 1, which is characterized by a relatively weaker brain network connectome. Conversely, compared with HC, patients with AIS showed a lower mean dwell time in State 2, which was characterized by a relatively stronger brain network connectome. Additionally, functional networks exhibited variable efficiency of information transfer across 4 states.

CONCLUSIONS

AIS not only altered the interaction between the different dynamic networks but also promoted characteristic alterations in the temporal and topological features of large-scale dynamic network connectivity.

The relevance of rich club regions for functional outcome post-stroke is enhanced in women

This study aimed to investigate the influence of stroke lesions in predefined highly interconnected (rich-club) brain regions on functional outcome post-stroke, determine their spatial specificity and explore the effects of biological sex on their relevance. We analyzed MRI data recorded at index stroke and ~3-months modified Rankin Scale (mRS) data from patients with acute ischemic stroke enrolled in the multisite MRI-GENIE study. Spatially normalized structural stroke lesions were parcellated into 108 atlas-defined bilateral (sub)cortical brain regions. Unfavorable outcome (mRS > 2) was modeled in a Bayesian logistic regression framework. Effects of individual brain regions were captured as two compound effects for (i) six bilateral rich club and (ii) all further non-rich club regions. In spatial specificity analyses, we randomized the split into "rich club" and "non-rich club" regions and compared the effect of the actual rich club regions to the distribution of effects from 1000 combinations of six random regions. In sex-specific analyses, we introduced an additional hierarchical level in our model structure to compare male and female-specific rich club effects. A total of 822 patients (age: 64.7[15.0], 39% women) were analyzed. Rich club regions had substantial relevance in explaining unfavorable functional outcome (mean of posterior distribution: 0.08, area under the curve: 0.8). In particular, the rich club-combination had a higher relevance than 98.4% of random constellations. Rich club regions were substantially more important in explaining long-term outcome in women than in men. All in all, lesions in rich club regions were associated with increased odds of unfavorable outcome. These effects were spatially specific and more pronounced in women.

Dynamic brain states in spatial neglect after stroke

Previous studies indicated that spatial neglect is characterized by widespread alteration of resting-state functional connectivity and changes in the functional topology of large-scale brain systems. However, whether such network modulations exhibit temporal fluctuations related to spatial neglect is still largely unknown. This study investigated the association between brain states and spatial neglect after the onset of focal brain lesions. A cohort of right-hemisphere stroke patients (n = 20) underwent neuropsychological assessment of neglect as well as structural and resting-state functional MRI sessions within 2 weeks from stroke onset. Brain states were identified using dynamic functional connectivity as estimated by the sliding window approach followed by clustering of seven resting state networks. The networks included visual, dorsal attention, sensorimotor, cingulo-opercular, language, fronto-parietal, and default mode networks. The analyses on the whole cohort of patients, i.e., with and without neglect, identified two distinct brain states characterized by different degrees of brain modularity and system segregation. Compared to non-neglect patients, neglect subjects spent more time in less modular and segregated state characterized by weak intra-network coupling and sparse inter-network interactions. By contrast, patients without neglect dwelt mainly in more modular and segregated states, which displayed robust intra-network connectivity and anti-correlations among task-positive and task-negative systems. Notably, correlational analyses indicated that patients exhibiting more severe neglect spent more time and dwelt more often in the state featuring low brain modularity and system segregation and vice versa. Furthermore, separate analyses on neglect vs. non-neglect patients yielded two distinct brain states for each sub-cohort. A state featuring widespread strong connections within and between networks and low modularity and system segregation was detected only in the neglect group. Such a connectivity profile blurred the distinction among functional systems. Finally, a state exhibiting a clear separation among modules with strong positive intra-network and negative inter-network connectivity was found only in the non-neglect group. Overall, our results indicate that stroke yielding spatial attention deficits affects the time-varying properties of functional interactions among large-scale networks. These findings provide further insights into the pathophysiology of spatial neglect and its treatment.

Prediction of post-stroke motor recovery benefits from measures of sub-acute widespread network damages

Following a stroke in regions of the brain responsible for motor activity, patients can lose their ability to control parts of their body. Over time, some patients recover almost completely, while others barely recover at all. It is known that lesion volume, initial motor impairment and cortico-spinal tract asymmetry significantly impact motor changes over time. Recent work suggested that disabilities arise not only from focal structural changes but also from widespread alterations in inter-regional connectivity. Models that consider damage to the entire network instead of only local structural alterations lead to a more accurate prediction of patients' recovery. However, assessing white matter connections in stroke patients is challenging and time-consuming. Here, we evaluated in a data set of 37 patients whether we could predict upper extremity motor recovery from brain connectivity measures obtained by using the patient's lesion mask to introduce virtual lesions in 60 healthy streamline tractography connectomes. This indirect estimation of the stroke impact on the whole brain connectome is more readily available than direct measures of structural connectivity obtained with magnetic resonance imaging. We added these measures to benchmark structural features, and we used a ridge regression regularization to predict motor recovery at 3 months post-injury. As hypothesized, accuracy in prediction significantly increased (R ² = 0.68) as compared to benchmark features (R ² = 0.38). This improved prediction of recovery could be beneficial to clinical care and might allow for a better choice of intervention.

Motor network reorganization after motor imagery training in stroke patients with moderate to severe upper limb impairment

BACKGROUND

Motor imagery training (MIT) has been widely used to improve hemiplegic upper limb function in stroke rehabilitation. The effectiveness of MIT is associated with the functional neuroplasticity of the motor network. Currently, brain activation and connectivity changes related to the motor recovery process after MIT are not well understood.

AIM

We aimed to investigate the neural mechanisms of MIT in stroke rehabilitation through a longitudinal intervention study design with task-based functional magnetic resonance imaging (fMRI) analysis.

METHODS

We recruited 39 stroke patients with moderate to severe upper limb motor impairment and randomly assigned them to either the MIT or control groups. Patients in the MIT group received 4 weeks of MIT therapy plus conventional rehabilitation, while the control group only received conventional rehabilitation. The assessment of Fugl-Meyer Upper Limb Scale (FM-UL) and Barthel Index (BI), and fMRI scanning using a passive hand movement task were conducted on all patients before and after treatment. The changes in brain activation and functional connectivity (FC) were analyzed. Pearson's correlation analysis was conducted to evaluate the association between neural functional changes and motor improvement.

RESULTS

The MIT group achieved higher improvements in FM-UL and BI relative to the control group after the treatment. Passive movement of the affected hand evoked an abnormal bilateral activation pattern in both groups before intervention. A significant Group × Time interaction was found in the contralesional S1 and ipsilesional M1, showing a decrease of activation after intervention specifically in the MIT group, which was negatively correlated with the FM-UL improvement. FC analysis of the ipsilesional M1 displayed the motor network reorganization within the ipsilesional hemisphere, which correlated with the motor score changes.

CONCLUSIONS

MIT could help decrease the compensatory activation at both hemispheres and reshape the FC within the ipsilesional hemisphere along with functional recovery in stroke patients.

Advanced Diffusion MRI of Stroke Recovery

There is an urgent need for ways to improve our understanding of poststroke recovery to inform the development of novel rehabilitative interventions, and improve the clinical management of stroke patients. Supported by the notion that predictive information on poststroke recovery is embedded not only in the individual brain regions, but also the connections throughout the brain, majority of previous investigations have focused on the relationship between brain functional connections and post-stroke deficit and recovery. However, considering the fact that it is the static anatomical brain connections that constrain and facilitate the dynamic functional brain connections, the microstructures and structural connections of the brain may potentially be better alternatives to the functional MRI-based biomarkers of stroke recovery. This review, therefore, seeks to provide an overview of the basic concept and applications of two recently proposed advanced diffusion MRI techniques, namely lesion network mapping and fixel-based morphometry, that may be useful for the investigation of stroke recovery at the local and global levels of the brain. This review will also highlight the application of some of other emerging advanced diffusion MRI techniques that warrant further investigation in the context of stroke recovery research.

Discovery of anti-stroke active substances in Guhong injection based on multi-phenotypic screening of zebrafish

Ischemic stroke is a leading cause of death worldwide, and it remains an urgent task to develop novel and alternative therapeutic strategies for the disease. We previously reported the positive effects of Guhong injection (GHI), composed of safflower extract and aceglutamide, in promoting functional recovery in ischemic stroke mice. However, the active substances and pharmacological mechanism of GHI is still elusive. Aiming to identify the active anti-stroke components in GHI, here we conducted a multi-phenotypic screening in zebrafish models of phenylhydrazine-induced thrombosis and ponatinib-induced cerebral ischemia. Peripheral and cerebral blood flow was quantified endogenously in erythrocytes fluorescence-labeled thrombosis fish, and baicalein and rutin were identified as major anti-thrombotic substances in GHI. Moreover, using a high-throughput video-tracking system, the effects of locomotion promotion of GHI and its main compounds were analyzed in cerebral ischemia model. Chlorogenic acid and gallic acid showed significant effects in preventing locomotor dyfunctions. Finally, GHI treatment greatly decreased the expression levels of coagulation factors F7 and F2, NF-κB and its mediated proinflammatory cytokines in the fish models. Molecular docking suggested strong affinities between baicalein and F7, and between active substances (baicalein, chlorogenic acid, gallic acid, and rutin) and NF-κB p65. In summary, our findings established a novel drug discovery method based on multi-phenotypic screening of zebrafish, provided endogenous evidences of GHI in preventing thrombus formation and promoting behavioral recovery after cerebral ischemia, and identified baicalein, rutin, chlorogenic acid, and gallic acid as active compounds in the management of ischemic stroke.

Regional homogeneity alterations in multifrequency bands in patients with basal ganglia stroke: A resting-state functional magnetic resonance imaging study

Objective

The aim of this study was to investigate the spontaneous regional neural activity abnormalities in patients with acute basal ganglia ischemic stroke (BGIS) using a multifrequency bands regional homogeneity (ReHo) method and to explore whether the alteration of ReHo values was associated with clinical characteristics.

Methods

In this study, 34 patients with acute BGIS and 44 healthy controls (HCs) were recruited. All participants were examined by resting-state functional magnetic resonance imaging (rs-fMRI). The ReHo method was used to detect the alterations of spontaneous neural activities in patients with acute BGIS. A two-sample t-test comparison was performed to compare the ReHo value between the two groups, and a Pearson correlation analysis was conducted to assess the relationship between the regional neural activity abnormalities and clinical characteristics.

Results

Compared with the HCs, the patients with acute BGIS showed increased ReHo in the left caudate and subregions such as the right caudate and left putamen in conventional frequency bands. In the slow-5 frequency band, patients with BGIS showed decreased ReHo in the left medial cingulum of BGIS compared to the HCs and other subregions such as bilateral caudate and left putamen. No brain regions with ReHo alterations were found in the slow-4 frequency band. Moreover, we found that the ReHo value of left caudate was positively correlated with the NIHSS score.

Conclusion

Our findings revealed the alterations of ReHo in patients with acute BGIS in a specific frequency band and provided a new insight into the pathogenesis mechanism of BGIS. This study demonstrated the frequency-specific characteristics of ReHo in patients with acute BGIS, which may have a positive effect on the future neuroimaging studies.

Suboptimal states and frontoparietal network-centered incomplete compensation revealed by dynamic functional network connectivity in patients with post-stroke cognitive impairment

Background

Neural reorganization occurs after a stroke, and dynamic functional network connectivity (dFNC) pattern is associated with cognition. We hypothesized that dFNC alterations resulted from neural reorganization in post-stroke cognitive impairment (PSCI) patients, and specific dFNC patterns characterized different pathological types of PSCI.

Methods

Resting-state fMRI data were collected from 16 PSCI patients with hemorrhagic stroke (hPSCI group), 21 PSCI patients with ischemic stroke (iPSCI group), and 21 healthy controls (HC). We performed the dFNC analysis for the dynamic connectivity states, together with their topological and temporal features.

Results

We identified 10 resting-state networks (RSNs), and the dFNCs could be clustered into four reoccurring states (modular, regional, sparse, and strong). Compared with HC, the hPSCI and iPSCI patients showed lower standard deviation (SD) and coefficient of variation (CV) in the regional and modular states, respectively (p < 0.05). Reduced connectivities within the primary network (visual, auditory, and sensorimotor networks) and between the primary and high-order cognitive control domains were observed (p < 0.01).

Conclusion

The transition trend to suboptimal states may play a compensatory role in patients with PSCI through redundancy networks. The reduced exploratory capacity (SD and CV) in different suboptimal states characterized cognitive impairment and pathological types of PSCI. The functional disconnection between the primary and high-order cognitive control network and the frontoparietal network centered (FPN-centered) incomplete compensation may be the pathological mechanism of PSCI. These results emphasize the flexibility of neural reorganization during self-repair.

Neuroimaging Techniques as Potential Tools for Assessment of Angiogenesis and Neuroplasticity Processes after Stroke and Their Clinical Implications for Rehabilitation and Stroke Recovery Prognosis

Stroke as the most frequent cause of disability is a challenge for the healthcare system as well as an important socio-economic issue. Therefore, there are currently a lot of studies dedicated to stroke recovery. Stroke recovery processes include angiogenesis and neuroplasticity and advances in neuroimaging techniques may provide indirect description of this action and become quantifiable indicators of these processes as well as responses to the therapeutical interventions. This means that neuroimaging and neurophysiological methods can be used as biomarkers—to make a prognosis of the course of stroke recovery and define patients with great potential of improvement after treatment. This approach is most likely to lead to novel rehabilitation strategies based on categorizing individuals for personalized treatment. In this review article, we introduce neuroimaging techniques dedicated to stroke recovery analysis with reference to angiogenesis and neuroplasticity processes. The most beneficial for personalized rehabilitation are multimodal panels of stroke recovery biomarkers, including neuroimaging and neurophysiological, genetic-molecular and clinical scales.

Post-stroke reorganization of transient brain activity characterizes deficits and recovery of cognitive functions

Functional magnetic resonance imaging (fMRI) has been widely employed to study stroke pathophysiology. In particular, analyses of fMRI signals at rest were directed at quantifying the impact of stroke on spatial features of brain networks. However, brain networks have intrinsic time features that were, so far, disregarded in these analyses. In consequence, standard fMRI analysis failed to capture temporal imbalance resulting from stroke lesions, hence restricting their ability to reveal the interdependent pathological changes in structural and temporal network features following stroke. Here, we longitudinally analyzed hemodynamic-informed transient activity in a large cohort of stroke patients (n = 103) to assess spatial and temporal changes of brain networks after stroke. Metrics extracted from the hemodynamic-informed transient activity were replicable within- and between-individuals in healthy participants, hence supporting their robustness and their clinical applicability. While large-scale spatial patterns of brain networks were preserved after stroke, their durations were altered, with stroke subjects exhibiting a varied pattern of longer and shorter network activations compared to healthy individuals. Specifically, patients showed a longer duration in the lateral precentral gyrus and anterior cingulum, and a shorter duration in the occipital lobe and in the cerebellum. These temporal alterations were associated with white matter damage in projection and association pathways. Furthermore, they were tied to deficits in specific behavioral domains as restoration of healthy brain dynamics paralleled recovery of cognitive functions (attention, language and spatial memory), but was not significantly correlated to motor recovery. These findings underscore the critical importance of network temporal properties in dissecting the pathophysiology of brain changes after stroke, thus shedding new light on the clinical potential of time-resolved methods for fMRI analysis.