Pattern interpretation by cellular automata (PICA)--evaluation of tumour cell adhesion in human melanomas.

In routine pathology, the evaluation of the pattern of a tumour at scanning magnification often reveals diagnostic and prognostic features indicating that the biological properties of the tumour cells are related to the morphological pattern. For further evaluation of the relationship between functional properties of the cells on the one hand and the pattern on the other, we propose the pattern interpretation by cellular automata (PICA) procedure. The PICA system consists of an import module transferring real histological images into the data structure of a cellular automaton, a measurement module generating a comparable quantitative description of real and simulated images, a cellular automation designed to simulate tumour growth and invasion at the histological level, a database consisting of the morphological results obtained in simulated patterns, an interpretation module linking real histological images to the knowledge stored in the database and an image synthesis and display module. By comparing real images to computer-simulated patterns, PICA facilitates an estimation of functional properties based on the static histological pattern of a given tumour. Using the example of tumour cell adhesion, it is demonstrated that the degrees of tumour-tumour and tumour-stroma adhesion significantly affect the resulting simulated patterns, that, in turn, the morphological evaluation of the patterns enables a reproducible estimation of adhesion and that estimates of adhesion in real images of malignant melanoma of the skin are of prognostic significance. PICA may serve as an additional in situ evaluation technique linking morphological features to functional properties.