Background: Clinical and radiological outcomes of endovascular thrombectomy (EVT) are related to etiologies of large vessel occlusion (LVO) in acute stroke. However, preprocedural computed tomography angiography (CTA) or CT perfusion imaging can hardly distinguish embolic occlusion from atherosclerotic occlusion. We hypothesized that quantitative multiphase CTA (mCTA) of LVO may predict occlusion types and thrombectomy outcome. Methods: We retrospectively evaluated the consecutive stroke patients who had undergone mCTA and EVT <6 hours of onset at two independent medical centers. The intra-arterial radiodensities of Hounsfield unit (HU) were measured to examine the HUdistal/proximal ratio using receiver operating characteristic curve analysis. The derived cut-off value was re-examined in an independent cohort. Results: In the derivation cohort (n = 102), 81 patients (79.4%) were embolic occlusion without severe residual intracranial atherosclerotic stenosis (ICAS[−]) and 21 patients were atherosclerosis-related occlusion (ICAS[+]) based on digital subtraction angiography (DSA). The optimal cut-off to predict embolic occlusion was HU ratio <0.6 measured at 2 mm from the occlusion site (maximum area under the curve = 0.87; sensitivity 96%; specificity 81%). This cut-off also independently predicted successful recanalization using stent-retrievers and/or contact aspiration (modified Treatment in Cerebral Ischemia score ≥2b; p = 0.002) after adjusting for age, atrial fibrillation, and collateral circulation score, but not predicted favorable outcome at 3 months post stroke. Importantly, in the validation cohort (n = 95, 80% embolic occlusion), this HU ratio cut-off similarly predicted occlusion types and recanalization outcome, respectively. Conclusion: The mCTA-based quantitative radiodensities of acute LVO provides preprocedural predictive values of DSA-determined occlusion types and thrombectomy outcomes.