Breast tissue microarrays (TMAs) facilitate the study of very large numbers of breast tumours per histological section, but their scoring by pathologists is time consuming, prone to observer variability, and not without error. This paper reports the use of ordinal regression to predict the scores of TMA spots subjected to progesterone receptor immunohistochemistry. We compare the use of global features obtained via two different methods, one involving and the other dispensing with accurate segmentation of epithelial cell nuclei. In addition, we investigate the effect of analysing only regions of interest (ROIs) within each spot, as opposed to analysing the whole spots. The use of the entropy of the posterior probability distribution over category labels for avoiding uncertain decisions is demonstrated.
|Title of host publication
|Medical Image Understanding and Analysis
|British Machine Vision Association and Society for Pattern Recognition
|Number of pages
|Published - 2009