Application of multiparametric diagnostic model based on the 18F-FDG PET/MR in the differential diagnosis of benign and malignant pancreatic tumors / 中华核医学与分子影像杂志

ABSTRACT

Objective:To investigate the value of multiple parameters and the combined model based on 18F-fluorodeoxyglucose (FDG) PET/MR in the differential diagnosis of benign and malignant pancreatic tumors. Methods:A total of 76 patients (50 males, 26 females, age (45.2±18.0) years) with clinically suspected pancreatic tumor who underwent PET/MR between December 2012 and September 2020 in First Medical Center of Chinese PLA General Hospital were retrospectively selected. All patients had definitive diagnoses. PET/MR sequences included T 1 weighted imaging (WI; convention and contrast enhancement), T 2WI, diffusion weighted imaging (DWI) and 18F-FDG PET sequences. Morphological characteristics and scores of lesions in MRI were evaluated. Parameters including parameters of PET and DWI, as well as parameters derived from histogram analysis of apparent diffusion coefficient (ADC) and standardized uptake value (SUV) (maximum (max), minimum (min), mean, median, standard deviation (SD), skewness, kurtosis, entropy) were measured. Independent-sample t test or Mann-Whitney U test were used for data analysis. PET/MR combined model was measured based on logistic regression analysis. Receiver operating characteristic (ROC) curve analysis was used to calculate the differential diagnosis efficacy of PET/MR multiparameters and combined model for benign and malignant lesions. Results:Among 76 patients, 55 were malignant and 21 were benign. (1) Visual evaluation. The main manifestations of pancreatic cancer were unclear margin, abnormal internal MR signal and enhancement, decreased ADC signal and increased radioactive uptake. The main manifestations of benign tumor lesions of pancreas were clear margin, even signal and enhancement, no reduction of ADC, decreased radioactive uptake. (2) Diagnostic efficacy. Multiparameter model established based on logistic regression analysis included SUV max, SUV SD, ADC entropy and ADC skewness. The efficiency of differential diagnosis for benign and malignant pancreatic tumors were shown as follows multiparametric diagnostic model>ADC entropy>MR score>SUV max>SUV SD>ADC skewness. The multiparametric diagnostic model had the best diagnosis efficiency with the area under curve (AUC) of 0.86, the sensitivity of 69.1%(38/55), and the specificity of 100%(21/21) ( z=-8.73, P<0.001). Conclusions:MR score and multiple quantitative parameters obtained from 18F-FDG PET/MR can be used to differentiate benign and malignant pancreatic tumors. Compared with independent parameter of PET/MR, multiparametric model can further improve the diagnostic efficiency.