Digital image analysis in cytological diagnosis: a morphometric analysis on pleural mesotheliomas

In the course of the last years digital image analysis has become increasingly important in the field of quantitative pathology. This may be due to the technical development of computers which have now reached a high user-friendly standard. In the present paper we present: (1) the development of a simple system for the measurement of digital image analysis; (2) the development of a general data structure applicable to various measurements and classification strategies; and (3) the analysis of the value of invariant moments (IM) as regional descriptors to distinguish mesothelial cells from mesothelioma cells in cytologic smears of pleural fluid. The results show that the invariant moments contain information concerning the geometric distribution of the grey values of the pixels as well as the staining intensity. The invariant moments present combination parameters and are therefore ideal morphometric parameters. With the invariant moments 7 out of 7 patients with mesotheliomas and 8 out of 11 control subjects could be assigned correctly in a nonparametric discriminant analysis procedure.