Ona Wu1, Lawrence L. Latour2, Shlee S. Song3, Karen L. Furie4, Steven Warach2, Lee H. Schwamm4
1Athinoula A Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, MA, United States; 2National Institute of Neurological Diseases and Stroke, National Institutes of Health, Bethesda, MD, United States; 3Department of Neurology, Cedar Sinai Medical Center, Los Angeles, CA, United States; 4Department of Neurology, Massachusetts General Hospital, Boston, MA, United States
MR WITNESS is a multi-center clinical trial of thrombolysis in acute stroke patients with unwitnessed onset. Enrolled subjects must exhibit MRI patterns consistent with early stage stroke: FLAIR negative or exhibit FLAIR signal intensity increase less than 15% compared to normal tissue. We investigated the inter-rater and inter-site reproducibility of this algorithm among 15 readers from two sites. We compared the performance to a simpler qualitative assessment of FLAIR positive or negative. We found that combining FLAIR+signal intensity had Fleiss κ=0.89 compared to κ=0.74 using simple FLAIR assessments. MR WITNESS algorithm is a robust, and reproducible approach.