Abstract WMP15: Hyperperfusion on Arterial Spin Labeling: Objective Decision Support Using Pattern Recognition
Background: Hyperperfusion detected on arterial spin labeling (ASL) images acquired after stroke onset has been shown to correlate with subsequent hemorrhagic transformation (HT). Presence of hyperperfusion is typically detected by visual review of arterial spin labeling cerebral blood flow (CBF). Such a review is subjective as it is challenged by inter-reader variability, noise, and lack of standard threshold. We present in this study a quantitative hyperperfusion detection model that can provide an objective decision support for the interpretation of ASL CBF maps and rapidly delineate hyperperfusion regions. Methods: ASL Cerebral blood flow (CBF) maps of acute stroke patients presenting with an occlusion in the MCA territory were coregistered to a standardized atlas space. To achieve reliable detection of ASL hyperperfusion, we formalize the problem as a nonlinear classification that relates regional voxel intensity values to the corresponding binary label (normal or hyperperfused). Our method takes into account the healthy contralateral hemisphere and its CBF intensity values during the determination of hyperperfusion of a voxel. Each input feature vector combines the regional intensity values at the voxel of interest, its contralateral matched region, and the distribution of the difference between them. Each input vector is associated to a label corresponding to the presence of hyperperfusion that was manually established by consensus between experts. The predicted hyperperfusion regions were compared to a groundtruth that manually established by two researchers. Results: A total of 361 ASL scans were collected from 221 patients (age=72±17 years; 45% males). Hyperperfusion was detected in 76 patients that were subsequently used in our analysis. An AUC of 83±5% was reached after a leave-one-out cross-validation, which corresponds to the accuracy in detecting hyperperfusion compared to manual delineation of hyperperfusion on ASL CBF maps. Conclusion: Pattern recognition based on a nonlinear regression can provide an accurate and objective measure of hyperperfusion on ASL CBF images and could therefore improve the detection of hemorrhagic transformation in acute stroke patients.