We propose a novel multiple instance learning algorithm for cancer detection in histopathology images. With images labelled at image-level, we rst search a set of region-level prototypes by solving a submodular set cover problem. Regularised regression trees are then constructed and combined on the set of prototypes using a multiple instance boosting framework. The method compared favourably with competing methods in experiments on breast cancer tissue microarray images and optical tomographic images of colorectal polyps.
展開▼