Accurate pathological patch classification with regression

摘要: Histopathological image analysis is important for cancer diagnosis. However, current computer cannot support the end-toend prediction of whole slide images (WSI) with gigapixel resolution. Most works slice high-resolution WSI into patches and classify them as tumor/normal tissues. classification based on only uses category labels according to proportion cells. For example, if a patch contains more than 50% cells, its label “tumor”, otherwise it “normal”. Obviously, although this scheme can achieve good performance, unreasonable because doctors do not simply by whether cells above or not. If ratio information in be fully utilized, rationality accuracy improved. In article, we firstly notice cell propose new model structure, combining regression modules. We introduce branch reference connected layer structure model, then combine branch. When classifying patches, predicts which assist better patchlevel classification. addition, find that baseline regression-assisted have different properties patches. Combining two branches final further improve performance. apply our method large datasets performance state-of-the-art methods. Source code available at https://github.com/ICLAB/ RACN.