Investigation of Post-stroke Dynamic Functional Connectivity Network and Its Utilization on Motor Function Evaluation

Stroke is the leading cause of long-term disability, and more than half of stroke survivors have motor disabilities of various degrees. However, the brain demonstrates an important capability to modify its activity in reaction to internal or external stimuli through reorganizing its neural connections after injuries such as stroke or traumatic brain injury (TBI). Hence, stroke patients can partially regain or restore impaired body function by engaging in appropriate rehabilitation exercises. Functional connectivity (FC) analysis with resting-state functional magnetic resonance imaging (RS-fMRI) has been applied to evaluate motor behavior through the analysis of the neural network organization process. Nevertheless, FC analysis, particularly dynamic functional connectivity (DFC) analysis, has not been widely used in post-stroke rehabilitation assessment. Hence, this project aims to conduct a long-term follow-up investigation on post-stroke patients to analyze the relationship between FC dynamics and motor function improvement. Unlike conventional static functional connectivity analysis, which assumes that the FC is static during an fMRI scan, DFC analysis evaluates the temporal evolution of the FC in seconds and produces a brain representation based on the connectivity changes of the time-varying network. Based on a comprehensive literature review, a research gap of time-varying functional networks in stroke is identified. Hence a novel DFC analysis pipeline was initially developed in this investigation with the aim to examine the dynamics of the post-stroke functional network in stroke. The effectiveness of the implemented analytical pipeline was also verified on fMRI data from healthy aged people due to the close relationship between age and stroke. The brain functions are fulfilled by a set of functionally specialized modules, which implies that post-stroke function impairment was related to the alteration of the modules in the brain function network. Based on the estimated time-varying network, the dynamic reconfiguration in the post-stroke brain was next investigated. With the presented method of multilayer temporal network, the dynamic changes in the brain function network of stroke patients with different severity levels were studied. Furthermore, the dynamic adaptive behaviours of the post-stroke brain derived from the time-varying network were measured, and the role of those brain behaviors in the motor function recovery following stroke was explored. Specifically, as stroke rehabilitation is a time-dependent process, brain dynamic behaviors recognized from the multilayer temporal network were first exploited to track stroke recovery. Followed by using machine learning methods, the whole-brain dynamic behaviors were then utilized to predict the post-stroke motor recovery status and explain differences in individual motor recovery. Results from the investigation show that the brain will dynamically adjust its functional network to adapt to stroke damage. The post-stroke transient network states become locally weakly connected and exhibit more segregation and higher modularity. Besides, it has been found that the brain will reconfigure its network to regain the balance of segregation and integration. Notably, the network reconfiguration has been demonstrated to be severity-dependent; mild and severe patients exhibit distinct reconfiguration patterns. Further explorations show that the dynamic behaviors of the brain network correlate with the post-stroke motor function recovery stages. The patients who are at higher recovery stages indicate significantly higher levels of network recruitment and flexibility. It has also shown that the brain's ongoing adjustments lead to varying expected outcomes among patients. The ridge regression method inputting the network measurement achieves the highest Area Under the Curve(AUC) of 85.93\%. Post-analysis also illustrates that network recruitment plays a crucial role in post-stroke recovery. In summary, the post-stroke motor function recovery process has been studied longitudinally by using the dynamic functional connectivity approach. The DFC analysis effectively demonstrates how the brain dynamically adapts to stroke attacks and how brain behavior correlates with motor recovery. The improved understanding of post-stroke brain network dynamics could be further used by rehabilitation professionals to make more precise prescriptions, provide reasonable prognosis reports, and offer more personalized and effective interventions to help stroke survivors regain lost function and enhance their quality of life.