Identification of high-risk plaque features in intracranial atherosclerosis: initial experience using a radiomic approach