Pattern Recognition of Functional Neuroimage Data of the Human Sensorimotor System After Stroke

This thesis contributes to the study of the regulatory mechanisms of the cerebral circulation and its dysfunction in disease. Novel analytical methods are studied for the characterisation of the neurovascular mechanisms measured by functional magnetic resonance imaging of humans having suffered ischemic stroke.;The results from the trials presented here show that the method is capable of differentiating between the response characteristics from the healthy and stroke patient groups. In the months following ischemic stroke, an abnormal increase in the number of distinct response signals occurs, exhibiting a dispersed spatial pattern and diminished temporal relationship to the motor task that is either extensively delayed, or abnormally early and anti-correlated. The most likely explanation is that these widely dispersed response signals appearing throughout the cerebral cortex of the stroke patients are no longer coupled to colocalised neural activity. According to this premise, our results suggest that the regulatory mechanisms of the cerebrovascular system can remain compromised in the months following ischemic stroke.;In summary, direct evidence is provided here supporting the application of exploratory pattern recognition and Bayesian hierarchical analysis of neuroimage data acquired during the study of ischemic stroke. Specifically, the aforementioned analytical methods are shown to be suitable for monitoring the spatiotemporal characteristics of the neurovascular response to sensorimotor stimuli. Its implementation in longitudinal trials would provide key information for the evaluation of stroke rehabilitation programmes and for the assessment of individual patient recovery.;Functional neuroimage data are acquired from consenting healthy volunteers and ischemic stroke patients with motor deficits who participated in an event-related visual feedback controlled hand motor task. The neuroscience and stroke literature provides the basis for this research on analytical methods for i) exploratory pattern recognition of the cerebrovascular responses to sensorimotor tasks, ii) the characterisation of their space-time structure, and iii) a Bayesian hierarchical model for evaluating their statistical significance for intersubject comparison. Ultimately, these methods aim to identify and distinguish between distinct response signals, thereby making them amenable to expert visual inspection and interpretation.