2017/08/14 by Daniel Bug, Bug, Daniel, Steffen Schneider +11
Computer Science · Medicine · #AI in cancer detection #Computer Vision and Pattern Recognition (cs.CV) #Digital Imaging for Blood Diseases #FOS: Computer and information sciences #Radiomics and Machine Learning in Medical Imaging
paper · pdf · doi:10.48550/arxiv.1708.04099
openalex publication_date 2017/08/14 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/28
While human observers are able to cope with variations in color and\nappearance of histological stains, digital pathology algorithms commonly\nrequire a well-normalized setting to achieve peak performance, especially when\na limited amount of labeled data is available. This work provides a fully\nautomated, end-to-end learning-based setup for normalizing histological stains,\nwhich considers the texture context of the tissue. We introduce Feature Aware\nNormalization, which extends the framework of batch normalization in\ncombination with gating elements from Long Short-Term Memory units for\nnormalization among different spatial regions of interest. By incorporating a\npretrained deep neural network as a feature extractor steering a pixelwise\nprocessing pipeline, we achieve excellent normalization results and ensure a\nconsistent representation of color and texture. The evaluation comprises a\ncomparison of color histogram deviations, structural similarity and measures\nthe color volume obtained by the different methods.\n