Diagnostic Accuracy and Reading Times of Subtraction CT for Detecting Extravasation in Pelvic Fractures

Background Accurate, timely detection of extravasation is essential for managing pelvic fractures, but distinguishing extravasation from bone fragments at CT is challenging. Although several studies demonstrate the benefits of subtraction CT, research on its use for detecting extravasation in pelvic fractures is lacking. Purpose To compare the diagnostic accuracy and reading times of dynamic contrast-enhanced CT with and without subtraction images for detecting extravasation in pelvic fractures. Materials and Methods This retrospective study, conducted at a level 1 trauma center, included patients with pelvic fractures admitted from April 2015 to July 2023. Subtraction CT images were generated by subtracting the arterial phase from the venous phase of dynamic contrast-enhanced CT. Three radiologists interpreted the dataset, which included plain, arterial phase, and venous phase images. One month later, they reevaluated the same dataset with the addition of subtraction CT images. For each dataset, they documented the presence and location of extravasation, along with their confidence scores. The reference standard for extravasation was established by consensus between two radiologists independent of the readers. Performance assessments included figure of merit in alternative free-response receiver operating characteristic analysis, sensitivity, positive predictive value, and reading time. Diagnostic measures were compared using generalized estimating equations, and reading time was analyzed with a linear model using a generalized estimating equation. Results Sixty-six patients (median age, 65 years [IQR, 36.3-72.0 years]; 34 male patients) were identified. Mean figure of merit values were higher for CT subtraction than without (0.82 vs 0.73; mean difference, 0.09 [95% CI: 0.05, 0.12]; P < .001). Mean sensitivity was higher with subtraction (81%) than without (63%; P < .001), with no compromise in positive predictive value. Reading time was reduced with subtraction (mean difference, -8.7 seconds [95% CI: -14.7, -2.8]; P = .004). Conclusion Subtraction CT improved diagnostic accuracy for detecting extravasation and reduced reading time in patients with pelvic fractures. © RSNA, 2026 Supplemental material is available for this article. See also the editorial by Dreizin in this issue.