Background: Volumetry of lymph nodes potentially better reflect asymmetric size alterations independently of lymph node orientation in comparison to metric parameters (e.g. long-axis diameter).
Purpose: To distinguish between benign and malignant lymph nodes by comparing 2D and semi-automatic 3D measurements in MSCT.
Material and methods: FDG-18 PET-CT was performed in 33 patients prior to therapy for malignant melanoma at stage III/IV. One hundred and eighty-six cervico-axillary, abdominal and inguinal lymph nodes were evaluated independently by two radiologists, both manually and with the use of semi-automatic segmentation software. Long axis (LAD), short axis (SAD), maximal 3D diameter, volume and elongation were obtained. PET-CT, PET-CT follow-up and/or histology served as a combined reference standard. Statistics encompassed intra-class correlation coefficients and ROC curves.
Results: Compared to manual assessment, semi-automatic inter-observer variability was found to be lower, e.g. at 2.4% (95% CI 0.05-4.8) for LAD. The standard of reference revealed metastases in 90 (48%) of 186 lymph nodes. Semi-automatic prediction of lymph node metastases revealed highest areas under the ROC curves for volume (reader 1 0.77, 95%CI 0.64-0.90; reader 2 0.76, 95%CI 0.59-0.86) and SAD (reader 1 0.76, 95%CI 0.64-0.88; reader 2 0.75, 95%CI 0.62-0.89). The findings for LAD (reader 1 0.73, 95%CI 0.60-0.86; reader 2 0.71, 95%CI 0.71, 95%CI 0.57-0.85) and maximal 3D diameter (reader 1 0.70, 95%CI 0.53-0.86; reader 2 0.76, 95%CI 0.50-0.80) were found substantially lower and for elongation (reader 1 0.65, 95%CI 0.50-0.79; reader 2 0.66, 95%CI 0.52-0.81) significantly lower (p < 0.05).
Conclusion: Semi-automatic analysis of lymph nodes in malignant melanoma is supported by high segmentation quality and reproducibility. As compared to established SAD, semi-automatic lymph node volumetry does not have an additive role for categorizing lymph nodes as normal or metastatic in malignant melanoma.