Abstract TP87: Gaze Deviation on Initial CT as Predictor of Large Vessel Occlusion

Introduction: Treatment of large vessel occlusion (LVO) stroke is traditionally preceded by imaging evaluation of vessels using computed tomographic angiography (CTA). However, CTA may not be feasible due to renal dysfunction or contrast allergy; in smaller hospitals, CTAs may not be easily obtainable. Given this, early LVO signs on initial non-contrast computed tomography (CT) could prove practical, such as the hyperdense vessel sign (HVS). We aimed to investigate the sensitivity of another potential indicator of LVO on CT, gaze deviation, and to compare it to HVS. Methods: We conducted a retrospective analysis of consecutive patients with anterior circulation LVO stroke admitted to a comprehensive stroke center from July 2015 to July 2018, who underwent endovascular therapy (ET). All patients had confirmed LVO on CTA prior to ET. Initial head CTs were reviewed for HVS and gaze deviation by two neurology-trained examiners and compared to official diagnostic radiology reports. Two analyses were performed: detection of HVS and gaze deviation by each examiner, diagnostic radiologist with sensitivity calculation; and interrater agreement for identification of HVS and gaze deviation using a Kappa Coefficient between the two primary examiners and between diagnostic radiologist. Results: A total of 121 patients were included. Sensitivity of HVS by examiner 1 was 46% (56/121) and 2 was 60% (72/121). Sensitivity of HVS by in-house radiology was 36% (42/116) and OSH radiology was 27% (24/90). Sensitivity of gaze deviation by examiner 1 was 65% (79/121) and 2 was 79% (95/121). Inter-rater reliability of HVS for examiners 1 and 2 had substantial agreement (K=0.64), and for in-house and OSH radiology had fair agreement (K=0.27). Inter-rater reliability of gaze deviation for examiners 1 and 2 had substantial agreement (K=0.69). No radiology report noted gaze deviation, despite 2 reports of unrelated ocular findings. Conclusions: Gaze deviation may be a more sensitive and more reliable indicator of LVO on initial non-contrast head CT than the hyperdense vessel sign. This is the largest sample size to examine and directly compare signs of LVO. More research should be invested in this and other indicators of LVO that can assist in the efficient triage and treatment of these severe strokes.