Automated grading in breast cancer by image analysis of histological sections
Abstract
The aim of the study was to test the solidity of the conventional grading of breast cancer and to divide otherwise non-specified invasive ductal carcinomas into different groups by means of automated microscopic image analysis. Therefore, a pilot study was started in which biopsies were examined from 350 patients who had undergone standardized operations between 1984 and 1987. A wide range of methods was used for characterization of tumor differentiation by image analysis of Feulgen-stained histological sections, including karyometry, chromatin structure analysis, and histometry. For measurements, the Robotron A 6471 image analysis system with AMBA/R software was used. Based on previous studies into preneoplasia of the mammary gland, the prevalence of karyometric features could be demonstrated. Moreover, some new parameters of the nuclear chromatin structure seem to be also correlated with histological grading and, therefore, might be of prognostic relevance, although this will have to be proved by statistical analysis of morphometric and clinical data when the follow-up study will be finished.