Clinical Uncertainty In Large Vessel Occlusion Ischemic Stroke (CULVO): An Intrarater And Interrater Agreement Study

Abstract Background Limited research exists regarding the impact of neuroimaging modality on endovascular thrombectomy (EVT) decisions for late window large vessel occlusion (LVO) stroke cases. Purpose This study assesses whether perfusion CT imaging: 1) alters the proportion of recommendations for EVT, and 2) enhances the reliability of EVT decision-making compared to non-contrast CT and CT angiography. Materials and Methods We conducted an online survey using 30 patients drawn from an institutional database of 3144 acute stroke cranial CT scans. These cases were presented to 29 stroke or neurointerventional physicians from Canada across two sessions. Physicians evaluated each patient both with and without perfusion imaging and gave EVT recommendations. We used non-overlapping 95% confidence intervals and difference in agreement classification as criteria to suggest a difference between the Gwet AC1 statistics (κ G ). Our outcomes were: 1) the proportion of EVT recommendations, and 2) interrater and intrarater agreement, with or without perfusion imaging. Results In the first round, 29 raters completed the assessment, with 28 finishing the second round. The percentage of EVT recommendations differed by 1.1% with or without perfusion imaging. However, individual decisions changed in 21.4% of cases, with 11.3% against EVT and 10.1% in favor. Interrater agreement (κG) among the 29 raters was similar between non-perfusion CT neuroimaging and perfusion CT neuroimaging (κG = 0.487; 95% CI 0.327, 0.647 and κG = 0.552; 95% CI 0.430, 0.675). The 95% CIs overlapped with moderate agreement in both. Intrarater agreement exhibited overlapping 95% CIs for all 28 raters. κG was either substantial or excellent (0.81-1) for 71.4% (20/28) of raters in both groups. Conclusion The difference in EVT recommendations is minimal with either neuroimaing protocol. Regarding agreement we found that use of automated CT perfusion images does not significantly impact the reliability of EVT decisions for late window LVO patients.