Image analysis for the evaluation of p53 expression in human cancers.

We compared the results of a subjective microscopic evaluation and an interactive image analysis of p53 expression in different human tumours. Expression of p53 was immunocytochemically detected by using the monoclonal antibody PAb 1801, and the evaluation was made blindly by two observers. Image analysis was performed on the same sections using DISCOVERY. The image was segmented into objects and background by interactive thresholding. Several morphological and densitometric features were selected in order to remove artefacts, lymphocytes, stromal cells and overlapping nuclei from the counting. The study was performed on series of 30 cases for different tumour types: breast, colon, oral cavity, ovary and lung cancer. Spearman's correlation coefficient ranged from 0.72 to 0.95 (P = 0.0001). Image analysis can therefore be considered as a quick and alternative approach to microscopic evaluation of nuclear immuno-determinations.