Abstract
The detection of cervical nodal metastases is important for the prognosis and treatment of head and neck tumors. The purposes of present study were first to assess the ability of apparent diffusion coefficient (ADC) values at 3.0T to distinguish malignant from benign lymph nodes and second to analyze ADC from partitions through a fuzzy C-means (FCM) technique for distinguishing nodal metastasis in head and neck cancer. From July 2009 to June 2010, 22 patients (21 males and 1 female; mean age, 49.8 ± 9.5 years; age range, 28–66 years) scheduled for surgical treatment of biopsy-proven head and neck cancer were prospectively and consecutively enrolled in this study. All patients were scanned on a 3.0T imaging unit (Verio; Siemens, Germany) using a 12-channel head coil combined with a 4-channel neck coil. All lymph nodes seen on DWI images were proceeded ADC calculation from ADC maps. All lymph nodes also were analyzed using in-house software developed using MATLAB. A radiologist manually contoured the lesions, and ADC values for each lesion were divided into 2 (low and high) and 3 (low, intermediate, and high) partitions by using the FCM clustering algorithm. Histologic findings were the reference standard for the diagnosis of lymph nodes metastasis. The ADC values derived from the signal intensity averaged across images obtained with b values of 0 and 800 s/mm2 was 1.086 ± 0.222 × 10-3 mm2/s for benign lymph nodes and 0.705 ± 0.118 × 10-3 mm2/s for malignant lymph nodes (P < .0001). When an ADC value of 0.851 × 10-3 mm2/s was used as a threshold value for differentiating benign from malignant lymph nodes, the best results were obtained with an accuracy of 91.0%, sensitivity of 91.3%, and specificity of 91.1%. From FCM technique, the results showed that the low value ADC clusters were more sensitive (95.7%) in distinguishing malignant from benign lesions than the whole-lesion mean ADC values (78.3%), while retaining a high specificity (approximately 90%). Moreover, receiver operating characteristic curves demonstrated that the low value ADC clusters used as a predictor of malignancy for lymph nodes could achieve a higher area under the curve (0.949 and 0.944 for 2 and 3 partitions, respectively). In conclusion, ADC value is a sensitive and specific parameter that can help to differentiate malignant from benign lymph nodes. But, the ADC cutoff value to distinguish malignant lymph nodes was that metastatic nodes with necrotic areas might have higher ADC values because of necrosis and might be misidentified as benign. the FCM clustering technique as a computed aid to prevent this bias.