"Sweet spot" for endoleak detection: optimizing contrast to noise using low keV reconstructions from fast-switch kVp dual-energy CT.

Objective: To assess endoleak detection and conspicuity using low-kiloelectron volt (keV) monochromatic reconstructions of single-source (fast-switch kilovolt [peak]) dual-energy data sets.

Methods: With approval of the institutional review board, multiphasic dual-energy computed tomographic (CT) scans for aortic endograft surveillance were retrospectively reviewed for 39 patients. Two abdominal radiologists each performed 2 separate reading sessions, at 55-keV and standard 75-keV reconstruction, respectively. The readers tabulated endoleak presence, conspicuity on 1-to-5 scale, and type overall and in arterial and venous phases. Originally, dictated reports in medical records were used as criterion standard.

Results: Original dictations identified 19 endoleaks (9 abdominal and 10 thoracic), 13 of which were type II. The blinded readers (R1 and R2) exhibited good to very good intraobserver and interobserver agreement. Endoleak detection was higher at 55 keV than at 75 keV (sensitivity, 100% (95% confidence interval [CI], 82.4%-100.0%) and 84.2% (95% CI, 60.4-96.6%) at 55 keV vs 79% (95% CI, 54.4-94.0%) and 68.4% (95% CI, 43.5%-87.4%) at 75 keV in venous phase). Further, endoleak conspicuity ratings (where original dictation showed positive leak) were higher at 55 keV than at 75 keV, which was a significant difference for R2 in the overall ratings (P = 0.03) and for both readers in the venous phase ratings (R1, P = 0.01; R2, P = 0.004). There was no difference in endoleak type characterization between the kiloelectron volt levels.

Conclusion: Sensitivity for endoleak detection and overall endoleak conspicuity ratings were both higher at 55 keV than 75 keV, favoring the inclusion of a lower-energy monochromatic reconstruction for endoleak surveillance protocols with dual-energy computed tomography.