Dynamic connectivity predicts acute motor impairment and recovery post-stroke

Figures

• 

  Figure 1. Lesion overlap of all patients. Most lesions were located in the middle cerebral artery territory. The highest lesion load was found subcortically, affecting white matter pathways and the grey matter of basal ganglia. Of note, our sample included five patients with pontine brainstem infarcts and one patient with small bilateral cerebellar lesions in addition to a primary left-hemispheric subcortical stroke. Right-hemispheric lesions of 19 patients were flipped to the left hemisphere to decrease the spatial heterogeneity of lesions.

• 

  Figure 2. Spatial maps of 14 included intrinsic connectivity networks. Networks were organized in three motor-related functional domains: Sensorimotor (SMN, 8 components, dark blue), subcortical (SC, 3 components, light blue) and cerebellar (CB, 3 components, yellow). Back-reconstruction of networks was based on components extracted in Allen et al.9 L, left; R, right; SMA, supplementary motor area.

• 

  Figure 3. (A) Static and (B) Dynamic functional network connectivity. (A) Darker red colour implies stronger positive, darker blue stronger negative connectivity. Static functional connectivity was, therefore, characterized by strong positive intra-domain connectivity, neutral connectivity between cortical and subcortical motor networks as well as neutral to negative connectivity between either cortical and subcortical motor networks and cerebellar networks. (B) Stated percentages above each state correspond to state-specific fraction times across all subjects. State 1, the most seldom state (median dwell time: 10 windows), featured highly positive intra-domain connectivity and highly negative connectivity between the sensorimotor and subcortical as well as cerebellar domains. In contrast, State 2, which emerged most often (median dwell time: 22 windows), was characterized by particularly weak intra-domain connectivity and mostly neutral inter-domain connectivity. Lastly, State 3 presented highly positive intra-domain connectivity, neutral connectivity between the sensorimotor and subcortical, and slightly negative connectivity between sensorimotor and cerebellar domains (median dwell…

• 

  Figure 4. Dynamic functional connectivity strength variability in relation to upper limb motor impairments and recovery post-stroke. (A) Mean variability of dynamic connectivity throughout the entire scan session. The left plot visualizes the variability averaged over all 54 patients initially recruited. The right plot considers all 30 patients that were followed up after six months. Darker red colour represents higher variability values. Asterisks indicate significant group effects between patients with no, moderate, or severe upper limb impairments or more, less or no recovery, respectively (three-level one-way ANOVA: P < 0.05). Most of the differences in variability were found for connections between the bilateral putamen and cortical sensorimotor networks. (B) Subtraction maps of mean variability values between each subgroup constellation of initial impairment. Red colour implies higher variability values, blue colour lower variability values. Post hoc t-tests revealed some overlapping (‘posterior putamen–left sensorimotor cortex’), yet mostly distinct significantly different connectivity pairs between mildly affected patients and non-affected patients, as well as mildly and…

• 

  Figure 5. Variability of dynamic connectivity and feature importances. Brain renderings visualize the most predictive dynamic connectivity variability pairs for the prediction of mild versus moderate-severe acute motor impairments (left) and the prediction of more or less pronounced recovery in the patients with initial impairment in the first months after stroke (right). The variabilities in dynamic connectivity between SMA and bilateral posterior putamen, as well as lateral paracentral cortex and bilateral posterior putamen were the most relevant for predicting recovery. Correlation analyses revealed that patients with more pronounced recovery featured less variability in these connections.