Combined morphometrical and syntactic structure analysis as tools for histomorphological insight into human lung carcinoma growth.

Histological sections of formalin-fixed, paraffin-embedded tissue comprising 60 surgical specimens of human lung carcinoma were Feulgen stained. The histomorphological images were transferred to an automated image analysing system (VISIAC) and analysed as follows. The geometrical centers of tumor cell nuclei were defined as vertices, and the minimum spanning tree (MST) was calculated based on the two-dimensional distance between the vertices. Segmentation of the images was performed semiautomatically by interactive definition of nuclei of interest and automated detection of nuclear boundaries. Several morphometric features of tumor cell nuclei were measured including size, DNA-content (extinction), and form factor, and were set in relation to parameters of the MST. The following results were obtained: DNA-content and tumor cell nucleus size ('center cell') of different microscopic tumor growth patterns are related to the number of nearest neighboring cells. No relation was found in the neighboring (surrounding) cells. The different cell types of lung carcinoma, i.e., the different microscopic tumor textures expressed the relation of center cell features to the parameters of MST. A high amount of DNA content in branching points of the MST for epidermoid carcinoma may be interpreted as carcinoma growing in epidermoid textures tend to proliferate from tumor cell nuclei related to at least one neighboring cell. The opposite was found for large cell anaplastic carcinoma (no perceptible microscopic textures of the tumors) which showed the highest DNA content in tumor cell nuclei but which was not related to any neighboring cells. This technique allows analysis of growth centers and microenvironment conditions in human lung cancer in relation to tumor texture at the light microscopy level.