A supervised method for calculation of perfusion/diffusion mismatch volume in acute stroke

We present our initial experience with a supervised technique for calculating the mismatch between lesion volumes on Perfusion and Diffusion weighted images of acute stroke. Our method was designed to be reliable and easy to use. It requires the intervention of a trained user at two points during the segmentation process: first to identify normal and lesion tissue, and second, to match a filter to artifact size. The time to analyze a single patient is 15 minutes or less. Statistical test showed us effect of rater or day. This performance suggests that personnel at multiple sites could use this method in future studies to evaluate the meaning and utility of the mismatch volume. Most of the variability in our measurements appears to be due to variability in the perfusion (MTT) images. We expect that this variability would be minimized with higher doses of contrast agent.