Using MR imaging to predict invasion of the recurrent laryngeal nerve by thyroid carcinoma.

Abstract

We evaluated the accuracy of MR imaging findings for predicting invasion of the recurrent laryngeal nerve by thyroid carcinoma and established an optimal criterion on which to base this prediction. We reviewed MR imaging findings (lesion size and posterior extension of tumor; encirclement of and invasion into the laryngeal cartilage, trachea, and esophagus by the tumor; and the amount of effaced fatty tissue in the tracheoesophageal groove or between the laryngeal cartilage and hypopharyngeal wall) in 66 patients with thyroid carcinoma. The amount of effaced fatty tissue was classified as one of five grades: 1, normal amount of fatty tissue; 2, partly effaced; 3, completely effaced in one MR imaging slice; 4, completely effaced in two contiguous MR slices; and 5, completely effaced in three or more contiguous MR slices. Thirty-two (48%) of the 66 patients had surgically or pathologically verified recurrent laryngeal nerve invasion. Logistic modeling revealed that the amount of effaced fatty tissue (p < 0.001) and the lesion size (p = 0.033) were the significant factors. Using the threshold values for the lesion size to predict invasion, we found that a threshold of more than 2.9 cm showed the highest accuracy, 76%, with 78% sensitivity and 74% specificity. For the amount of effaced fatty tissue, a grade of 3 or more had the highest accuracy, 88%, with 94% sensitivity and 82% specificity. Addition of the lesion size to this criterion did not improve the diagnostic accuracy of using the amount of effaced fatty tissue alone. Invasion of the recurrent laryngeal nerve by thyroid carcinoma was accurately predicted by the finding of effaced fatty tissue on MR imaging.