Objective

The angiographic spot sign (AS) on CT angiography (CTA) is known to be useful for predicting expansion in intracranial hemorrhage, but its use is limited due to its relatively low sensitivity. Recently, dual-energy computed tomography (DECT) has been shown to be superior in distinguishing between hemorrhage and iodine. This study aimed to evaluate the diagnostic performance of hematoma expansion (HE) using DECT AS in traumatic intracranial hemorrhage.

Methods

We recruited participants with intracranial hemorrhage confirmed via CTA for suspected traumatic cerebrovascular injuries. We evaluated AS on both conventional-like and fusion images of DECT. AS is grouped into three categories: intralesional enhancement without change, delayed enhancement (DE), and growing contrast leakage (GL). HE was evaluated by measuring hematoma size on DECT and follow-up CT. Logistic regression analysis was used to evaluate whether AS on fusion images was a significant risk factor for HE. Diagnostic accuracy was calculated, and the results from conventional-like and fusion images were compared.

Results

Thirty-nine hematomas in 24 patients were included in this study. Of these, 18 hematomas in 13 patients showed expansion on follow-up CT. Among the expanders, AS and GL on fusion images were noted in 13 and 5 hematomas, respectively. In non-expanders, 10 and 1 hematoma showed AS and GL, respectively. In the logistic regression model, GL on the fusion image was a significant independent risk factor for predicting HE. However, when AS was used on conventional-like images, no factors significantly predicted HE. In the receiver operating characteristic curve analysis, the area under the curve of AS on the fusion images was 0.71, with a sensitivity and specificity of 66.7% and 76.2%, respectively.

Conclusions

GL on fusion images of DECT in traumatic intracranial hemorrhage is a significant independent radiologic risk factor for predicting HE. The AS of DECT fusion images has improved sensitivity compared to that of conventional-like images.

PMID 37498879