The proliferative activity of breast tumors, which is routinely estimated bycounting of mitotic figures in hematoxylin and eosin stained histologysections, is considered to be one of the most important prognostic markers.However, mitosis counting is laborious, subjective and may suffer from lowinter-observer agreement. With the wider acceptance of whole slide images inpathology labs, automatic image analysis has been proposed as a potentialsolution for these issues. In this paper, the results from the Assessment ofMitosis Detection Algorithms 2013 (AMIDA13) challenge are described. Thechallenge was based on a data set consisting of 12 training and 11 testingsubjects, with more than one thousand annotated mitotic figures by multipleobservers. Short descriptions and results from the evaluation of eleven methodsare presented. The top performing method has an error rate that is comparableto the inter-observer agreement among pathologists.
translated by 谷歌翻译