A multiparametric analysis based on DCE-MRI to improve the accuracy of parotid tumor discrimination.

Abstract

Recently, semiquantitative time-intensity curve (TIC) analysis based on DCE-MRI and apparent diffusion coefficient (ADC) value-based diffusion-weighted imaging (DWI) were used to improve the diagnostic efficiency when diagnosing parotid tumors (PTs). However, quantitative DCE-MRI biomarkers have not been emphasized previously. To explore the diagnostic efficiency of perfusion parameters alone or in combination based on quantitative DCE-MRI and DWI in the differential diagnosis of PTs. In total, 112 patients with parotid masses were prospectively recruited in our hospital from August 2013 to March 2017. All patients were evaluated with DCE-MRI and DWI before surgery. TIC and quantitative parameters based on DCE MRI and ADCs were analyzed. Receiver operating characteristic analysis and linear discriminant analysis (LDA) was used to determine their diagnostic performance. In total, 87% (27/31) of pleomorphic adenoma (PA) showed type A TIC, 74% (65/88) of Warthin's tumors showed type B TIC, and 95% (19/20) of malignant tumors showed TIC type C. Pearson X2 test showed a significant difference between TIC patterns in benign and malignant tumors (X2 = 38.78, p < 0.001). ROC analysis revealed that ADC achieved the best diagnostic performance for distinguishing PA and Warthin's tumor from others, with area under the curve (AUC) values of 0.945 and 0.925 (p < 0.01), respectively. Furthermore, the TIC type was the only useful biomarker for distinguishing malignant from benign PTs, with an AUC of 0.846 (p < 0.01). Concerning the accuracy of the combined application of multiple parameters of DCE-MRI and ADC values, a combination of TIC pattern and extracellular volume ratio (Ve) provided the best results among five protocols, producing the highest accuracy of 0.75, followed by the combined use of the TIC pattern and ADC (accuracy was 0.70). TIC pattern in combination with the Ve biomarker based on DCE-MRI could achieve optimal diagnostic accuracy in the differential diagnosis of PTs.