Image analysis of mesothelioma. I. differentiation of mesothelioma from adenocarcinoma of the lung.

OBJECTIVE To explore the usefulness of nuclear micromorphometric analysis for the differentiation between epithelial mesothelioma and metastatic adenocarcinoma in the chest wall. STUDY DESIGN High-resolution images of 2,100 nuclei from 27 cases of epithelial mesothelioma and 15 cases of adenocarcinoma of the lung were recorded. Stepwise discriminant analysis and a nonparametric classifier were applied to derive estimates for a case diagnosis correct classification rate. RESULTS Nuclei from epithelial mesothelioma and adenocarcinoma of the lung showed statistically significantly different properties, but there was a region of overlap in feature space such that approximately 15-20% of cases could not be correctly classified. The lesion signatures derived from the mesothelioma cases with discriminant function scores that might result in case misclassification and the cases of adenocarcinoma of the lung spanned a similar range of degree of nuclear abnormality. However, the distribution of nuclear abnormality values for the mesothelioma cases has a mode at 0.87 SD from normal, whereas the distribution seen in lung adenocarcinoma cases had a mode at about 3.7 SD. CONCLUSION Cases of epithelial mesothelioma and adenocarcinoma of the lung have nuclei with a wide range of deviation from normal in the spatial and statistical distribution of their nuclear chromatin. For approximately 80% of cases, correct case classification can be provided by nuclear micromorphometric analysis. Cases of epithelial mesothelioma with highly abnormal nuclei overlap in feature space with nuclei from adenocarcinoma of the lung. However, it is possible that characterization by a lesion signature may allow correct assignment for those cases.