Clinical usefulness of spectral images derived from dual-layer spectral detector CT for the detection of pancreatic neuroendocrine tumor lesions / 中华放射学杂志

ABSTRACT

Objective:To investigate the clinical value of iodine density map and low keV virtual monoenergetic images (VMI) derived from dual-layer spectral detector CT (DLCT) for the pancreatic neuroendocrine tumors (pNETs) detection.Methods:From January to June 2019, data of 23 pathologically confirmed patients of pNETs were retrospectively analyzed. All of the patients underwent pancreas enhanced DLCT scanning within 1 week before surgery. The conventional polyenergetic images (PI), iodine density map and 40, 50, 60, 70 keV VMI were generated. One resident radiologist with 3 years’ experience and one senior radiologist with over 10 years’ experience interpreted the images for the lesion detection independently using the following image series PI, VMI (40-70 keV), PI combined with iodine density map. Lesion detection rates were recorded and compared among different image series. The CT value and noise of lesion, normal pancreatic parenchyma, and abdominal subcutaneous fat were measured in PI and VMI in both arterial and portal vein phases. The contrast-to-noise ratio (CNR) of lesion was calculated. The CT value of lesion and normal pancreatic parenchyma, CNR of lesion, and image noise were compared using repeated one-way ANOVA test. Subjective image quality was assessed with a 5-point scale and compared with Friedman test.Results:A total of 26 lesions were confirmed from 23 patients. For resident radiologist and senior radiologist, the detection rates of pNETs lesion using PI were 76.9% (20/26) and 84.6% (22/26) respectively, and both improved to 92.3% (24/26) using image series of 40 and 50 keV VMI. For senior radiologist, the pNETs lesion detection rate was further improved to 96.2% (25/26) using image series of PI with iodine map. The CT value of lesion and normal pancreatic parenchyma, CNR, and image noise had statistical differences among PI and VMI (40-70 keV) in both arterial and portal vein phase ( P<0.001). The mean CT attenuation and CNR of lesion in VMI increased significantly as the energy level decreased.The CNR of lesion in all VMI (40-70 keV) was significantly higher than that in PI. The median of subjective scores of image quality in PI and VMI (40-70 keV) were 3, 3, 4, 4, and 5 respectively, and the difference was statistically significant (χ2=66.393, P<0.001). Conclusions:The low keV VMI derived from DLCT can increase the CT value and CNR of pNETs, and the lesion detection rate can be improved combined with iodine density map. The CNR of pNETs is the highest in 40 keV VMI, and image noise is still lower than that of PI, so 40 keV VMI is recommended for clinical application.