In a blind, prospective trial, 42 patients undergoing 51 cervical lymp
h node dissections for head and neck malignancies were investigated by
a preoperative MRI scan. Histological examination of the dissection s
pecimens recorded the distribution, size, percentage of neoplastic inv
olvement and any extra-capsular spread of the sampled nodes. The MRI c
orrectly diagnosed all 17 positive nodes (sensitivity 100%). However i
t could not differentiate between enlarged reactive and malignant node
s, so there was a significant false positive rate (specificity 53%). R
outine use of MRI scans will allow an unnecessary neck dissection to b
e avoided in a patient with a clinically negative neck who also has a
negative MRI scan.