Warthin's tumor is the second most frequent benign tumor of the parotid gland, with no risk of malignant evolution. That is why surgery should be avoided if the preoperative diagnosis is certain. The aim of the study was to assess the added value of a decisional algorithm for the preoperative diagnosis of Warthin's tumor.

This retrospective IRB-approved study included 75 patients who underwent standardised MRI with conventional sequences (T1- and T2-weighted images, and T1 post-contrast sequences with fat saturation) and functional sequences: diffusion (b0, b1000) and perfusion MR. Two independent readers reviewed the images using the decisional algorithm. The conclusion of each reader was: the lesion is or is not a Warthin's tumor. The MRI conclusion was compared with histology or with cytology and follow-up. We calculated the Cohen's kappa coefficient between the two observers and the sensitivity and specificity of the algorithm-helped-reading for the diagnosis of Warthin's tumor.

Seventy-five patients; histology (n=61) or cytology and follow-up (n=14) results revealed 20 Warthin's tumors and 55 other tumors. Using the algorithm, sensitivity and specificity were 80–96%, and 85–100%, respectively for readers 1 and 2. The Cohen's kappa coefficient between the two observers was 0.79 (P<0.05) for the diagnosis of Warthin's tumor.

Our decisional algorithm helps the preoperative diagnosis of Warthin's tumor. The specificity of the technique is sufficient to avoid surgery if a parotid gland tumor presents all the MRI characteristics of a Warthin's tumor.