Parotid masses: prediction of malignancy using magnetization transfer and MR imaging findings.

Abstract

We determined the most accurate criteria for predicting malignancy of masses in the parotid gland using magnetization transfer ratios. Lesion-to-muscle magnetization transfer ratios obtained with a spoiled gradient-recalled acquisition in a steady state sequence with a 1-kHz off-resonance pulse were measured in 72 parotid masses (52 benign lesions, 20 malignant tumors). Various MR imaging findings and lesion-to-muscle magnetization transfer ratios were simultaneously assessed using a logistic model to determine the useful factors for predicting malignancy. We also studied the clinical usage of magnetization transfer ratios. Of the MR imaging findings, poorly defined margins showed the highest accuracy, 81%, with 60% sensitivity and 88% specificity. Of the lesion-to-muscle magnetization transfer ratios, a ratio of greater than 0.71 was most accurate (85%), with 90% sensitivity and 83% specificity. All four recurrent tumors and 10 (91%) of 11 secondary tumors were correctly diagnosed using the magnetization transfer ratio analysis. The logistic model revealed that the margin characteristics (p = 0.084) and lesion-to-muscle magnetization transfer ratios (p < 0.001) were statistically significant predictors for malignancy. A combined criteria of poorly defined margins and a lesion-to-muscle magnetization transfer ratio of greater than 0.71 raised the accuracy to 86% and specificity to 96%, but the sensitivity decreased to 60%. A combination of MR imaging findings and lesion-to-muscle magnetization transfer ratios was the most accurate predictor of malignancy.