Spectral attenuation curves of photon-counting detector CT scans for the differentiation between mucinous and non-mucinous cyst fluids of pancreatic lesions – a proof-of-concept study

Background Spectral imaging via photon-counting detector CT (PCD-CT) scanners allows the generation of spectral attenuation curves. Our study aimed to test the feasibility of spectral attenuation curve analysis for differentiating mucinous from non-mucinous cyst fluids of pancreatic cystic lesions (PCL). Methods Our study included 74 patients with PCLs. The classic morphological features were evaluated by an expert radiologist. The spectral attenuation curves were generated from the pancreatic-phase scans. The average CT values were measured on the 70 keV (HU 70keV ) and 40 keV (HU 40keV ) virtual monoenergetic images (VMIs), and the differences were calculated (HU Δ40−70keV ). The cases were divided into training (53 mucinous vs. 19 non-mucinous PCLs) and test (10 mucinous vs. 10 non-mucinous PCLs) datasets. The receiver operating characteristic (ROC) curve analysis assessed the discrimination performance. The intra- and interobserver reproducibilities were evaluated by the intraclass correlation coefficient (ICC). Results On 70 keV VMIs, no significant differences were found between the average CT values of mucinous and non-mucinous PCLs, however, the difference was significant in HU Δ40−70keV values ( p < 0.0001). The diagnostic performance of HU Δ40−70keV in differentiating between mucinous vs. non-mucinous PCL had AUCs of 0.92 and 0.92 on the training and test datasets, respectively, with a good interobserver (ICC = 0.82) and excellent intraobserver reproducibility (ICC = 0.94). The conventional clinical-radiological model achieved AUCs of 0.85 and 0.86, and with the addition of HU Δ40−70keV values its performance increased significantly to 0.98 and 0.94 on the training and test datasets, respectively. Conclusions Spectral attenuation curve assessment of cyst fluids could be a useful additional measurement to facilitate the noninvasive differential diagnosis of PCLs.