Abstract TP72: Localization of Large Vessel Occlusion (LVO) Stroke in the Posterior Circulation Based on Perfusion Imaging
Introduction: The vascular territory of an occluded large intracranial vessel can be reliably identified by CT or MR perfusion imaging. Furthermore, prior studies have shown that perfusion imaging can reliably predict the specific vessel that is occluded in anterior circulation large vessel strokes. We evaluated whether perfusion imaging can predict the specific vessel occlusion (vertebral, basilar, or posterior cerebral) in posterior circulation strokes. Hypothesis: We hypothesized that the occluded vessel could be inferred from the perfusion imaging results in >80% of patients with an acute stroke due to large vessel occlusion in the posterior circulation using the simultaneous CTA or MRA as the gold standard. Furthermore, the inter-rater agreement between a vascular neurologist and a neuroradiologist would be > 90%. Agreement Coefficients (AC1) were determined. Methods: From a multicenter cohort of consecutive patients with posterior circulation stroke, we included patients with documented occlusion of the Basilar Artery (BA) posterior cerebral Artery (PCA) or vertebral artery (VA) who had perfusion imaging (MRI or CT) processed by RAPID software. Perfusion images were evaluated blinded to the angiography or any other brain imaging results. The primary outcome measure was agreement on LVO location based on the CTA/MRA results. Results: 74 patients were eligible: age 63±2, female 32%, median NIHSS 15 (IQR 5-24). The distribution of large vessel occlusions on CTA/MRA was BA (74%), PCA (14%) and VA (12%). Perfusion imaging was able to correctly predict the occluded vessel in 63 (85%), AC1 = 0.82 (95% CI 0.72-0.92), p<0.001. Interrater agreement (n=41) was high [AC1 = 0.94 (95% CI 0.87-1.0), p < 0.001]. Conclusion: Perfusion imaging can predict the site of vessel occlusion (vertebral, basilar, or posterior cerebral) in posterior circulation strokes with good accuracy and high inter-rater agreement.