Myoepithelial sialadenitis versus low-grade non-Hodgkin's lymphoma of the salivary gland in FNAB: is discrimination by means of an image processing system possible?

The diagnosis of myoepithelial sialadenitis (MESA) in fine needle aspiration biopsy may be difficult. There is a dense lymphocytic infiltration in the gland and discrimination between a hyperimmune reaction and a low grade non-Hodgkin's lymphoma (NHL) of B-cell origin may be impossible. To get additional diagnostically helpful criteria, texture feature analysis on routinely obtained FNAB's of the salivary gland was applied. In the data set 36/36 cases of low grade B-NHL confirmed by histology and 10/13 histologically confirmed cases of MESA could be classified correctly by means of an image processing system. The chromatin structure of each nucleus was classified by texture features (n = 6), which were determined according to the method of Harms et al. For statistical analysis of the cell types a classification tree based on the commercial program CART was applied. The data set of 49 cases was proved by the crossvalidation test 10 fold. The calculated diagnosis for each case suggests that this method may be helpful in the cytologically doubtful cases.