Computation of statistical models for stroke detection and their validation Lead Investigator: Stian Johnsen Institution : BrainMiner Ltd E-Mail : stian.johnsen@brainminer.co.uk Proposal ID : 991 Proposal Description: Hypothesis: We can automatically detect stroke lesions in brain CT and DWI scans and identify subjects who have suffered an ischemic or haemorrhagic stroke with high confidence Methods: Computation of statistical models of normalcy that allow us to determine the visible changes in stroke patient scans. Validation of said models on a mixed stroke patient/control population by standard statistical means (e.g., receiver operator curves). The detection algorithm is based on research conducted jointly by the Translational Imaging Group of UCL and the department of neurology of UCLH (both London, UK).