Optimal size criteria of malignant lymph nodes in the treatment planning of radiotherapy for esophageal cancer: Evaluation by computed tomography and magnetic resonance imaging

Purpose: To determine the optimal size criterion for detection of lymph node metastases from esophageal cancer for radiotherapy by computed tomography (CT) and magnetic resonance (MR) imaging. Methods and Materials: In 58 patients with esophageal cancer treated with subtotal esophagectomy and radical lymph node dissection the preoperative MR (n = 58) images and CT scans (n = 41) were reviewed. The relationship of the CT and MR findings for the neck and mediastinum to the surgical and histopathological results was examined. Five size criteria on malignant lymph nodes were used to construct receiver operating characteristic (ROC) curves for CT and MR, and their detectabilities were evaluated. Results: The specificities of both modalities at the cutoff of 3 mm short-axis diameter were lower than those at the cutoff of 5 mm or more. In contrast, the sensitivities apparently decreased at the cutoff of 10 mm or more. The analysis of the ROC curves showed that the optimal size criterion for malignant lymph nodes was 5 mm for both CT and MR. When the criterion of 5 mm was used, the sensitivity, specificity, and accuracy for CT was 68, 92, and 87%, respectively, and the respective values for MR were 70, 93, and 89%. Although there was no significant difference between the two ROC curves, MR was useful in distinguishing lymph nodes from vascular structures because of the flow void. Conclusions: There was no significant difference between CT and MR in the detection of malignant lymph nodes from esophageal cancer. The optimal size criterion for both CT and MR in the detection of cervical and mediastinal lymph node metastases is 5 mm for short-axis diameter. These results suggest that all regional lymph nodes of 5 mm or more on CT or MR should be regarded as part of the gross tumor volume in the treatment planning of radiotherapy for esophageal cancer. Copyright (C) 1996 Elsevier Science Inc.