Background: In stroke erythrocyte-rich thrombi are more sensitive to intravenous thrombolysis with recombinant tissue plasminogen activator (IV-rtPA) and have higher density on non-contrast CT (NCCT). We investigated the relationship between thrombus density and recanalization and whether persistent occlusions can be predicted by Hounsfield unit (HU) measurements. Methods: In 88 IV-rtPA-treated patients with intracranial ICA or MCA occluding thrombus and follow-up imaging, thrombus and contralateral vessel attenuation measurements were performed on thin-slice NCCT. Mean absolute and relative HU were compared between patients with persistent occlusion (modified Thrombolysis in Cerebral Infarction system, grade 0/1/2a) and recanalization (grade 2b/3). Univariate and multivariate (adjusted for stroke subtype, clot burden score, occlusion site and time to thrombolysis) odds ratios for persistent occlusion were calculated. Additional prognostic value for persistent occlusion was estimated by adding HU measurements to the area under the curve (AUC) of known determinants and calculating optimal cut-off values. Results: Patients with persistent occlusion (n = 19) had significant lower mean HU (absolute 52.2 ± 9.5, relative 1.29 ± 0.20) compared to recanalization (absolute 63.1 ± 10.7, relative 1.54 ± 0.23, both p < 0.0001). Odds ratios for persistent occlusion were 3.1 (95% confidence interval, CI 1.6-6.0) univariate and 3.1 (95% CI 1.7-5.7) multivariate per 10 absolute HU decrease and 3.2 (95% CI 1.6-6.5) univariate and 4.1 (95% CI 1.8-9.1) multivariate per 0.20 relative HU decrease. Attenuation measurements significantly increased the AUC (0.67) of the known determinants to 0.84 (absolute HU) and 0.86 (relative HU). Cut-off values of <56.5 absolute HU and <1.38 relative HU showed optimal predictive values for persistent occlusion. Conclusions: Thrombus density is related to recanalization rate. Lower absolute and relative HU are independently related to persistent occlusion and HU measurements significantly increase discriminative performances of known recanalization determinants.