DeyeCOM(++) had 100% specificity for large vessel occlusion, whereas DeyeCOM(--) had a 100% specificity for absence of large vessel occlusion. Sustained DeyeCOM, whether positive or negative, is a strong predictor of ultimate diagnosis that could lead to quicker endovascular treatment times.

Introduction: Rapid imaging in acute stroke is critical and often occurs before full examination. Early, reliable exam findings clarify diagnosis and improve treatment times. Suspected stroke patients typically undergo both acute non-contrast CT and CT Angiogram (CTA). The DeyeCOM sign (conjugate gaze deviation on CT) has been described as a predictor of ischemic stroke with good specificity. In this study, we further evaluate a sustained DeyeCOM sign on both CT and CTA in early prediction of anterior large vessel occlusion (LVO). Methods: We retrospectively reviewed 46 acute stroke code patients between April and June 2017 from the prospective, IRB approved UCSD Stroke Registry. Patients had both CT and CTA as part of their acute stroke work-up as standard of care. DeyeCOM(+) sign was defined as a conjugate gaze deviation on CT or CTA of at least 15 degrees measured by caliper technique on Agfa IMPAX software (v.6.6.1.3525). DeyeCOM(++) was defined as sustained gaze deviation on both CT and CTA. Results: Three groups of patients were observed: 1) gaze deviation on both CT and CTA (DeyeCOM(++)), 2) non-sustained gaze deviation (+ on CT or CTA but not both), and 3) no gaze deviation (DeyeCOM(--)). All patients with DeyeCOM(++) (8/8, 100%) had anterior LVO. Of those with non-sustained gaze deviation, 2/7 (29%) had LVO. No DeyeCOM(--) patients (0/31, 100%) had LVO. Of the DeyeCOM(--) patients, 16/31 (52%) were stroke mimics and 15/31 (48%) had strokes that were not LVO. The specificity and sensitivity of DeyeCOM(++) for LVO was 100% (CI .90-1.0) and 80% (CI .44-.97). The specificity and sensitivity of DeyeCOM(--) for absence of LVO was 100% (CI .69-.10) and 86% (CI .70-.95). Conclusion: In this study, DeyeCOM(++) had 100% specificity for anterior LVO while DeyeCOM(--) had a 100% specificity for absence of LVO. DeyeCOM(++) false positive rates were low. Sustained DeyeCOM, whether positive or negative, is a strong predictor of ultimate diagnosis. These results should support practitioners’ confidence that DeyeCOM(++) reflects LVO stroke and is unlikely to be a mimic and DeyeCOM(--) patients likely do not have LVO. This data point can potentially lead to quicker endovascular treatment times. Further comparisons in this dataset are ongoing and work in a larger data set is needed.

Background: Identifying a last known well (LKW) time surrogate for acute stroke is vital to increase stroke treatment. Previous research have used an MRI DWI vs. FLAIR mismatch to estimate LKW time to within 6 hours. DWI signal intensity initially increases from onset of stroke, but mapping a reliable time course to the signal intensity has not been demonstrated. Methods: We retrospectively reviewed stroke code patients between 1/2016 and 6/2017 from the prospective, IRB approved UCSD Stroke Registry. Patients who had MRI brain from onset, with or without intervention, are included. All ischemic strokes were confirmed with ADC correlation and timing from onset to MRI was calculated. Raw DWI intensity was measured in grays (Gy) using IMPAX software and compared to contralateral side for control, for a relative DWI intensity (rDWI). LKW and MRI time were collected by chart review. Correlation was assessed using Pearson correlation coefficient between DWI intensity, rDWI, and time to MRI imaging. 1.5T, 3T, and combined modalities were examined. Results: 97 patients were included in this analysis. Overall, there is a good correlation (0.39, p<0.001) for minutes from onset to MRI and DWI intensity. There was good correlation with the 1.5T group (0.39, p=0.001) and very good correlation with the 3T group (0.60, p=0.001). There were no significant differences in demographic or time interval between the two MRI types. Conclusions: There is good correlation between DWI intensity and minutes from onset to MRI. This suggests a time-dependent DWI intensity response and supports the potential use of DWI intensity measurements to extrapolate a LKW time in unknown cases. Further studies in a larger dataset within the hyperacute period are being pursued to increase both experience and generalizability.