All the studies were conducted with the approval of the animal care committee at Seoul National University Hospital. The experiments were performed using six New Zealand White rabbits (Central Biological Laboratory, Seoul, Korea) weighing 2.5-3.0 kg each. The animals were allowed food and water ad libitum between the various procedures. They were anesthetized by an intramuscular injection of 50 mg/kg of ketamine hydrochloride (Ketalar®, Yuhan Yanghang, Seoul, Korea) and 20 mg/kg xylazine hydrochloride (Rompun® Bayer Korea, Seoul, Korea).

To induce head and neck cancer, 0.5 ml of a cell suspension of VX2 carcinoma was inoculated subcutaneously into the cranial section of the lower third of both auricles. The experimental auricular VX2 rabbit carcinoma was prepared in a manner described previously (12). The MR imaging was performed approximately three weeks after VX2 inoculation. Solid auricular masses were palpable in all animals.

The anesthetic regimen used for tumor implantation was also used for MR imaging. The ear vein of a rabbit was cannulated with a 22-gauge infusion set that was connected to a 1 mL syringe loaded with contrast agent. All the animals were imaged in the prone position. All examinations were performed using a 1.5-T superconducting magnet (Signa Horizon, GE Medical Systems, Milwaukee, WI) and with employing a standard transmit-receive quadrature head coil. After obtaining the routine localization images, the T1-weighted spin-echo (SE) sequence was performed in the coronal plane with using the following parameters: a TR of 400 ms, a TE of 12 ms and two excitations. The slice thickness used was 2 mm without a gap, the matrix number was 256×160 and the field of view (FOV) was 16 cm. The T2-weighted fast SE sequence (TR/TE: 3500/84 ms, 2 excitations) was also obtained. The coronal three-dimensional fast spoiled gradient-recalled echo (3DSPGR) T1-weighted images (TR/TE, 19.7/1.8 ms; FOV, 20×20 cm) were obtained before and 2 minutes after 0.1 mmol/kg of body weight gadopentetate dimeglumine (Magnevist® Schering, Germany) had been administered by manual injection via an ear vein.

All MR images were assessed with an emphasis placed on the parotid and caudal mandibular lymph nodes. The images were retrospectively evaluated by consensus between two experienced radiologists (who unaware of the study details) for the lymph node size, the signal intensity ratios and the enhancement pattern. No attention was paid to the stage of the auricle carcinoma. The locations of the suspicious lymph nodes were designated as parotid or mandibular. All lymph nodes on the right or left side were interpreted and the results were compared with the pathological findings.

The craniocaudal longitudinal diameters of the lymph nodes (maximum diameter) were measured on the selected T1-weighted and T2-weighted MR images. On the contrast-enhanced T1-weighted MR images, circular regions of interest (1.0-2.0 mm in diameter) were positioned on the hypointense area to determine signal intensities of the lymph nodes by consensus between the two radiologists. The signal intensity ratio was defined as the postcontrast signal intensity of an affected node relative to the adjacent muscle, and this was calculated for the metastatic and hyperplastic lymph nodes.

After the MR images were obtained, the animals were sacrificed by administering a lethal dose (90 mg/kg) of intravenous sodium pentobarbital (Pentothal® , Choong Wae Pharmacy, Seoul, Korea). The parotid and mandibular lymph nodes were exposed to determine their positions and then they were carefully removed. The pathologic specimens were fixed in 10% phosphate-buffered formaldehyde for subsequent embedding in paraffin and the pathologic analysis. The histologic sections were obtained in planes paralleling the MR imaging orientations. Serial sections of the entire lymph nodes were obtained and subjected to hematoxylin-eosin (HE) staining, which was used to assess the lymph node morphology.

Two observers simultaneously reviewed all the slides with using a double-headed light microscope. The total number of nodes, the number of metastatic and hyperplastic lymph nodes, the number of necrotic metastatic nodes and the maximum diameters of the nodes were documented. Discrepant interpretations were resolved by consensus.

The mean sizes of metastatic and hyperplastic lymph nodes were compared by using the independent t-test in the SigmaStat 2.0 and SAS program for Windows version 11.0 (SPSS Science, Chicago, and the SAS Institute, Cary, NC). We investigated the use of node size and the signal intensity ratio as the criteria for the differentiation of metastatic and hyperplastic lymph nodes, and for calculating the accuracies, sensitivities, specificities and the positive and negative predictive values. Receiver operating characteristic (ROC) curve analyses were also performed by calculating the areas under the appropriate ROC curves (Az) (14). Binormal ROC curves were fitted using ROCKIT 0.9 software (15).

The size distributions of all the lymph nodes and the relations between the size measurements and the signal intensity ratios are shown in Figure 3. The average maximum diameters of the metastatic and hyperplastic lymph nodes were 10.5±3.2 mm and 8.0±3.6 mm, respectively, which was not statistically significant (p > 0.05). The mean signal intensity ratios of the metastatic and hyperplastic lymph nodes were 0.82±0.16 and 2.02±0.86, respectively, which was statistically significant (p < 0.05). The correlation coefficient between the node size and signal intensity was 0.06, which was not statistically significant (p > 0.05)

The contrast-enhanced T1-weighted MR images showed peripheral high and central low signal intensity (n = 13) or homogenous high signal intensity (n = 23) (Figs. 1, 2). Of the 13 lymph nodes with peripheral high and central low signal intensity, nine nodes were confirmed to be metastatic (Fig. 1) and were four hyperplastic (Fig. 2) by microscopic examination. Of the 23 nodes with homogenous high signal intensity, three nodes were metastatic and 20 nodes were hyperplastic.

Using a cutoff value of a maximal diameter of 10 mm in the coronal plane, the sensitivity and specificity were 58% (7/12) and 79% (19/24), respectively, and they were 83% (10/12) and 50% (12/24), respectively, at a cutoff value of 7 mm. Using a signal intensity ratio less than one as a diagnostic criterion for a metastatic lymph node, the MR imaging had an accuracy of 81% (29/36), a sensitivity of 75% (9/12), a specificity of 83% (20/24), a positive predictive value of 69% (9/13) and a negative predictive value of 87% (20/23) for the detection of lymph node metastasis.

By ROC analysis, the Az value with a 95% confidence interval (CI) for a signal intensity ratio less than 1 on the contrast-enhanced T1-weighted MR images was 0.81 (95% CI: 0.63, 0.92), whereas that for a size measurement criterion of 10 mm was 0.72 (95% CI: 0.54, 0.86) (Fig. 4). The difference was significantly significant (p < 0.05).