Purpose: To investigate the diagnostic performance of dual-layer dual-energy CT (dlDECT) in the evaluation of adrenal nodules.
Methods: In this retrospective study, 66 patients with triphasic dlDECT (unenhanced, venous phase (VP), delayed phase (DP)) for suspected adrenal lesions were included. Virtual unenhanced images (VUE) were derived from VP acquisitions. Reference diagnoses were established with true unenhanced (TUE) attenuation, absolute washout, follow-up imaging and pathological data. Attenuation for adrenal lesions and abdominal tissues was acquired on TUE, VUE, VP and DP images. VUE and TUE attenuation were compared in all included tissues. Characterization of adrenal nodules based on TUE and VUE attenuation was investigated. ROC analysis was used to determine an adjusted threshold for diagnosing lipid-rich adenomas.
Results: Seventy-three adrenal nodules (mean size: 18.9 ± 8.9 mm) were identified in 66 patients (38 females, 28 males; age: 61 ± 13 years) including adenoma (n = 65), metastases (n = 2), pheochromocytoma (n = 3), adrenocortical carcinoma (n = 1) and myelolipoma (n = 2). Mean attenuation of all included tissues except for the abdominal aorta (p = 0.11) was significantly higher in VUE compared to TUE images, including the attenuation of adrenal nodules (20.0 ± 17.2 vs. 7.1 ± 19.8; p < 0.05). Classification of adrenal adenomas as lipid-rich based on VUE attenuation ≤ 10 HU yielded a sensitivity/specificity of 0.2/1.0, while an adjusted threshold of ≤ 22 HU yielded a sensitivity/specificity of 0.82/0.85.
Conclusion: dlDECT-derived VUE images overestimated attenuation in adrenal nodules, resulting in low sensitivity for diagnosis of lipid-rich adenomas using the established 10 HU threshold. Based on an adjusted threshold (≤ 22 HU) a higher sensitivity was attained, yet at the expense of a lower specificity, warranting further validation.
Keywords: Adrenal glands; Adrenal nodules; Dual-energy CT; Lesion characterization; Virtual unenhanced images.
© 2021. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.