Efficient Region-based Classification for Whole Slide Images.

For the past decade, new hardware able to generate very high spatial resolution digital images called Whole Slide Images (WSIs) have been challenging traditional microscopy. But the potential for automation is hindered by the large size of the files, possibly tens of billions of pixels. We propose a fast segmentation method coupled with an intuitive multiclass supervised classification that captures expert knowledge presented as morphological annotations to establish a cartography of a WSI and highlight biological regions of interest. While our primary focus has been the development of a proof of concept for the analysis of breast cancer WSIs acquired after chromogenic immunohistochemistry, this method could also be applied to more general texture-based problems.