External Validation of Atherosclerotic Neuroimaging Biomarkers in Emergent Large‐Vessel Occlusion

Intracranial atherosclerosis related large vessel occlusion (ICAS‐LVO) is the major cause of failed mechanical thrombectomy. ICAS‐LVO causes reocclusion or a fixed focal stenosis, leading to suboptimal revascularization and poor functional outcomes. We aimed to externally validate 4 preidentified imaging biomarkers of ICAS‐LVO: absent hyperdense sign, Hounsfield units (Hu ratio ≤1.1 and Delta Hu <6) and truncal‐type occlusion, observed on admission noncontrast computed tomography and computed tomography angiography in patients presenting with emergent large‐vessel occlusion (ELVO). We conducted a retrospective cohort observational study of consecutive patients presenting with acute M1/terminal internal carotid artery occlusions undergoing mechanical thrombectomy. Inability to locate a hyperdense vessel on noncontrast computed tomography at the corresponding ELVO on computed tomography angiography was labeled absent hyperdense sign. Delta Hu and Hu ratio were defined as the difference and ratio of the Hu of the ELVO on noncontrast computed tomography and its mirror contralateral patent vessel, respectively. ELVO was classified as truncal‐type occlusion if the bifurcation distal to the occlusion was spared on computed tomography angiography. ICAS‐LVO was defined as the presence of fixed focal stenosis or reocclusion after mechanical thrombectomy. Statistical analysis was performed using C statistics, receiver operating characteristic curve analysis, and multivariate logistic regression. Of 161 patients, 30 (18.6%) had suspected ICAS‐LVO. Absent hyperdense sign had a sensitivity of 90% and specificity of 87% (area under the curve [AUC], 0.88), in predicting ICAS‐LVO. Hu ratio ≤1.1 (AUC, 0.89) and Delta Hu <6 (AUC, 0.96) had sensitivity of 100% and 97% and specificity of 79% and 95%, respectively. Truncal‐type occlusion showed a sensitivity of 75% and specificity of 98% (AUC, 0.87). When comparing receiver operating characteristic AUC, Delta Hu <6 was significantly better than absent hyperdense sign ( P =0.006); Hu ratio ≤1.1 ( P =0.006); and truncal‐type occlusion ( P =0.02). Combination of neuroimaging biomarkers using noncontrast computed tomography and computed tomography angiography in ELVO identify ICAS‐LVO with high predictive power. Larger, prospective, multicenter studies are warranted to further evaluate their effectiveness in diagnosing ICAS‐LVO.