Preliminary study of using diffusion kurtosis imaging for characterizing parotid gland tumors

Background Diffusion kurtosis imaging (DKI) is a technique based on the theory that the diffusion of water follows non-Gaussian distribution; until now, few studies have explored the usefulness of DKI in differentiating parotid gland tumors. Purpose To evaluate the potential of DKI in the characterization of parotid gland tumors. Material and Methods DKI was performed in 40 patients with parotid gland tumors (33 benign, 7 malignant). DKI parameters including apparent diffusion coefficient (D-app) and apparent kurtosis coefficient (K-app) were obtained using the DK model and compared. Intraclass correlation coefficient (ICC) was used to evaluate inter-observer agreement of DKI derived parameters. Results Excellent inter-observer agreements (D-app, ICC = 0.988; K-app, ICC = 0.985) were achieved during quantitative measurements. There were no significant differences in D-app (P = 0.055) and K-app (P = 0.915) between the benign and malignant groups. In the subgroup analyses, pleomorphic adenomas showed higher D-app (P < 0.001) and lower K-app (P = 0.037) than malignant tumors. Combination of D-app and K-app showed optimal discriminative value (area under curve [AUC] = 0.913; sensitivity = 88.89%; specificity = 85.71%). Pleomorphic adenomas also showed higher D-app (P < 0.001) and lower K-app (P < 0.001) than Warthin's tumors. A combination of D-app and K-app showed optimal discriminative value (AUC = 1.000; sensitivity = 100%; specificity = 100%). Warthin's tumors demonstrated higher K-app than malignant tumors (P < 0.001), while no significant difference was found on D-app (P = 0.448). The AUC, sensitivity, and specificity of K-app for discriminating Warthin's tumors from malignant tumors at an optimal threshold of 0.735 were 0.905, 88.89%, and 85.71%, respectively. Conclusion DKI may be a promising imaging technique for characterizing parotid gland tumors.