Abstract
Background: In acute ischemic stroke, non-invasive perfusion imaging can guide the decision to perform endovascular intervention. A subset of these patients who are ineligible for IV TPA or for whom it has failed may benefit by going directly to endovascular intervention without delays imposed by non-invasive imaging. In these patients, an angiographic biomarker of viable brain tissue such as the capillary blush described by the Capillary Index Score (CIS) will be important for decision making in the angiography suite. Indeed, a favorable CIS score is associated with good outcomes when recanalization is achieved. However, any ordinal angiographic scale is observer dependent and limited by scale properties. Methods: Here we used our novel perfusion angiography software (PerfAngio) to extract cerebral blood volume (CBV) maps from conventional angiograms acquired during endovascular intervention at either UPMC or UCLA. Areas of angiographic blush were selected manually from a subset of the angiograms. These blush areas trained a machine learning model to identify features of angiographic blush from CBV maps to produce blush maps. Results: In the figure, we show PerfAngio’s blush map in a patient with acute proximal MCA occlusion prior to endovascular recanalization. At each pixel cool colors represent low likelihood of capillary blush and hotter colors represent higher likelihood of blush. These color maps allow for spatial characterization of the blush and quantifies it as a continuous variable rather than according to an ordinal scale. Conclusions: PerfAngio blush maps allow for automation and quantification of the blush seen during conventional angiography. These maps render data that does not depend on observer interpretation and provide spatial information about the capillary blush that is not captured by the CIS. Since PerfAngio blush maps can be acquired in real time, they are amenable for use in the angiography suite to inform the decision to recanalize or not.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have