Optimal window setting for 40 keV virtual monoenergetic images derived from dual-layer spectral detector CT pancreas dynamic enhanced scanning

Objective: To explore the image quality of 40 keV virtual monoenergetic images (VMI) derived from dual-layer spectral detector CT (DLCT) pancreas dynamic enhanced scanning and its optimal window setting. Methods: From January to July 2019, 28 patients who underwent pancreas enhanced DLCT scan within one week before surgery and pathologically confirmed of pancreatic neuroendocrine tumors (pNETs) were retrospectively enrolled. Conventional polyenergetic images (PI) and 40 keV virtual monoenergetic images (VMI_{40 keV}) were generated after scanning. CT value of normal pancreatic parenchyma, lesion, abdominal subcutaneous fat, abdominal aorta and portal vein were measured in PI and VMI_{40 keV}. The contrast-to-noise ratio (CNR) of the pNETs lesion was calculated. All these objective results were compared between VMI_{40 keV} and PI using paired t test. Individual window settings (W-Ind, including window width and window level) of VMI_{40 keV} were recorded. Calculated window settings (W-Calc) were mathematically calculated via regression analysis and optimized window settings (W-Opt) were obtained. Subjective image quality was assessed with a 5-point scale and compared among VMI_{40 keV} with different window settings (W-Std, W-Ind, W-Calc and W-Opt) using Friedman test, and compared PI with standard abdominal window setting (W-Std) and VMI_40keV with W-Opt settings using Wilcoxon test. The maximum diameter of lesion was measured and compared with one-way ANOVA analysis among PI and VMI_{40 keV} with different windows settings. Results: For VMI_{40 keV} in both arterial phase and portal vein phase, the CT attenuation [(464.0 ± 136.7), (375.4 ± 79.2) HU] of pNETs lesion were statistically significantly higher than those in PI [(168.8 ± 38.0), (140.5 ± 23.5) HU] (t = -16.107, -22.225, P < 0.001), CNR (16.5 ± 11.1, 10.9 ± 6.1) were also statistically significantly higher than those in PI (4.5 ± 2.9, 3.0 ± 1.9) (t = -7.838, -9.781, P < 0.001), while with lower image noise in VMI_{40 keV} [(11.8 ± 1.5), (11.8 ± 1.4) HU] than PI (13.1 ± 1.5, 12.9 ± 1.3 HU) (t = 6.356, 3.891, P < 0.001). The subjective score for PI with W-Std and VMI_{40 keV} with W-Std, W-Ind, W-Calc, W-Opt in arterial phase were 4 (1), 1 (0), 5 (0), 5 (0.75), 5 (1), and which in portal vein phase were 3.5 (1), 1 (0), 5 (0), 5 (0), 5 (1). The subjective score of VMI_{40 keV} with different window settings had statistical differences (X² = 76.143, 76.000, P < 0.001). Compared to the image quality of PI with W-Std settings, VMI_{40 keV} with W-Opt settings have higher objective score (Z = 4.685, 4.235, P < 0.001). The maximum diameter of lesion has no statistical difference among PI and VMI_{40 keV} with different window settings (F = 0.008, 0.004, P > 0.999) in both arterial phase and portal vein phase. Conclusions: The VMI_{40 keV} in pancreas dynamic enhancement scanning derived from dual-layer spectral detector CT have higher image quality than PI. Due to changes of CT value of tissue in VMI_{40 keV}, it is recommended to optimize window settings (window width / window level, 880/230 HU for arterial phase and 840/260 HU for portal vein phase) to obtain the best image quality. (authors)