Development of intelligent multimodal imaging analysis platform to predict stroke motor outcomes

Stroke is a leading cause of adult disability. Being able to predict motor recovery and outcomes soon after stroke could support clinicians to set appropriate goals for treatment and rehabilitation.  This project will derive and train an automated artificial intelligence platform by using machine learning methods to identify lesions and features of the sensorimotor network and whole brain to classify patients according to expected stroke outcomes. This project has the potential to improve the quality and efficiency of rehabilitation. The clinical characteristics will be combined with acute imaging to make predictions for upper limb function and walking outcomes at 3 months post-stroke. We will derive the model using our large retrospective imaging and clinical dataset. The model will then be used to create and train a prediction tool using prospectively collected routine medical imaging data.