Background and PurposeIn acute ischemic stroke (AIS) management, CT-based thrombus density has been associ-ated with treatment success. However, currently used thrombus measurements are proneto inter-observer variability and oversimplify the heterogeneous thrombus composition. Ouraim was first to introduce an automated method to assess the entire thrombus density andthen to compare the measured entire thrombus density with respect to current standardmanual measurements.Materials and MethodIn 135 AIS patients, the density distribution of the entire thrombus was determined. Density dis-tributions were described using medians, interquartile ranges (IQR), kurtosis, and skewedness.Differences between the median of entire thrombus measurements and commonly appliedmanual measurements using 3 regions of interest were determined using linear regression.ResultsDensity distributions varied considerably with medians ranging from 20.0 to 62.8 HU andIQRs ranging from 9.3 to 55.8 HU. The average median of the thrombus density distribu-tions (43.5±10.2 HU) was lower than the manual assessment (49.6±8.0 HU) (p<0.05).The difference between manual measurements and median density of entire thrombusdecreased with increasing density (r = 0.64; p<0.05), revealing relatively higher manualmeasurements for low density thrombi such that manual density measurement tend overes-timates the real thrombus density.@en

Introduction
Thrombus density may be a predictor for acute ischemic stroke treatment success. However, only limited data on observer variability for thrombus density measurements exist. This study assesses the variability and bias of four common thrombus density measurement methods by expert and non-expert observers.

Methods
For 132 consecutive patients with acute ischemic stroke, three experts and two trained observers determined thrombus density by placing three standardized regions of interest (ROIs) in the thrombus and corresponding contralateral arterial segment. Subsequently, absolute and relative thrombus densities were determined using either one or three ROIs. Intraclass correlation coefficient (ICC) was determined, and Bland–Altman analysis was performed to evaluate interobserver and intermethod agreement. Accuracy of the trained observer was evaluated with a reference expert observer using the same statistical analysis.

Results
The highest interobserver agreement was obtained for absolute thrombus measurements using three ROIs (ICCs ranging from 0.54 to 0.91). In general, interobserver agreement was lower for relative measurements, and for using one instead of three ROIs. Interobserver agreement of trained non-experts and experts was similar. Accuracy of the trained observer measurements was comparable to the expert interobserver agreement and was better for absolute measurements and with three ROIs. The agreement between the one ROI and three ROI methods was good.

Conclusion
Absolute thrombus density measurement has superior interobserver agreement compared to relative density measurement. Interobserver variation is smaller when multiple ROIs are used. Trained non-expert observers can accurately and reproducibly assess absolute thrombus densities using three ROIs.@en