Multiphase CTA-derived tissue maps aid in detection of medium vessel occlusions

Medium vessel occlusions (MeVOs) can be challenging to detect on imaging. Multiphase computed tomography angiography (mCTA) has been shown to improve large vessel occlusion (LVO) detection and endovascular treatment (EVT) selection. The aims of this study were to determine if mCTA-derived tissue maps can (1) accurately detect MeVOs and (2) predict infarction on 24-h follow-up imaging with comparable accuracy to CT perfusion (CTP). Two readers assessed mCTA tissue maps of 116 ischemic stroke patients (58 MeVOs, 58 non-MeVOs) and determined by consensus: (1) MeVO (yes/no) and (2) occlusion site, blinded to clinical or imaging data. Sensitivity, specificity, and area under the curve (AUC) for MeVO detection were estimated in comparison to reference standards of (1) expert readings of baseline mCTA and (2) CTP maps. Volumetric and spatial agreement between mCTA- and CTP-predicted infarcts was assessed using concordance/intraclass correlation and Dice coefficients. Interrater agreement for MeVO detection on mCTA tissue maps was estimated with Cohen’s kappa. MeVO detection from mCTA-derived tissue maps had a sensitivity of 91% (95% CI: 80–97), specificity of 82% (95% CI: 70–90), and AUC of 0.87 (95% CI: 0.80–0.93) compared to expert reads of baseline mCTA. Interrater reliability was good (0.72, 95% CI: 0.60–0.85). Compared to CTP maps, sensitivity was 87% (95% CI: 75–95), specificity was 78% (95%CI: 65–88), and AUC was 0.83 (95% CI: 0.76–0.90). The mean difference between mCTA- and CTP-predicted final infarct volume was 4.8 mL (limits of agreement: − 58.5 to 68.1) with a Dice coefficient of 33.5%. mCTA tissue maps can be used to reliably detect MeVO stroke and predict tissue fate.