Abstract 50: A Novel Method for Quantification of Brain Edema in Ischemic Stroke

Abstract

Background: Measurement of final infarct volumes are typically done on MRI or CT scans obtained at 30 or 90 days after symptom onset. At these delayed time points, true infarct volume is often underestimated because of atrophy. Moreover, late imaging leads to a high percentage of missing data due to early mortality and poor compliance with late imaging. Obtaining an early assessment of infarct volume avoids these limitations; however, it overestimates the true infarct volume because of vasogenic edema. The aim of this study was to develop a novel approach to quantify brain edema so that final infarct volumes can be approximated on day 5 scans. Method: We analyzed MRI scans from 15 acute stroke patients enrolled in the DEFUSE 2 study. The edema volume was determined by comparing spatially-coregistered ADC maps obtained at baseline and at day 5. Edema displaces CSF from the sulci and the ventricles. We therefore estimated the edema volume as the change in CSF volume between baseline and follow-up. CSF displacement was determined by analyzing ADC maps; a voxel by voxel comparison was performed and voxels with ADC values that shifted from the CSF range (2.0-4.4 x 10 -3 mm 2 /s) to the parenchymal range (0.60-1.05 x 10 -3 mm 2 /s) were considered to represent edema (yellow region in figure ). The analysis was performed on 5 control patients with no or very small lesions (2.4±2.0 mL on FLAIR) and 10 patients with large lesions (240±115 mL on FLAIR). Results: The mean volume of CSF displacement in the control group was 1.6±2.4 mL. CSF displacement (edema volume) in the large infarct group ranged from 33 to 152 mL. The estimated edema volume, as a percentage of the overall day 5 FLAIR lesion volume, ranged from 15% to 52%. Linear regression analysis performed between day 5 FLAIR lesion volumes and the estimated edema volumes produced a correlation coefficient (r)of 0.78 (P< 0.01) with a slope of 0.31. Conclusion: Vasogenic edema can be measured based on assessment of CSF displacement. Using this novel technique we determined that the day 5 FLAIR lesion volume, on average, overestimates the final infarct volume by 31%. These findings suggest that the final infarct volume can be accurately estimated from an MRI scan obtained during the vasogenic edema phase. Validation of this technique, using data sets that have infarct volume determinations at multiple time points, is in progress and results will be presented.