Abstract
The value of automatic image analysis in the investigation of nucleolus regions (AgNOR) has been examined in tissue sections of 52 malignant and 30 benign breast lesions. Determination of the AgNOR number per cell alone revealed a considerable overlap between benign (range 1.2-3.8) and malignant specimens (range 1.5-16.2). They differed however, highly significantly (P less than 0.001) in their AgNOR sizes. In benign breast disorders the mean AgNOR area per tumour ranged from 0.22 microns2 to 1.07 microns2 (mean 0.39 microns2), whereas in carcinomas AgNOR sites ranged from 0.05 microns2 to 0.22 microns2 (mean 0.09 microns2). AgNOR counts showed a good correlation with histopathological grade (P less than 0.05), aneuploidy (P less than 0.01), proliferation rate as determined by Ki67 immunostaining (P less than 0.01), as well as oestrogen and progesterone receptor content (P less than 0.01). Image analysis proved to be advantageous over AgNOR counting alone as it facilitated the standardization of the AgNOR technique itself and thus, significantly improved its diagnostic specifity.