A Preprocessing Pipeline for Histopathological Whole Slide Images

Digital pathology (DP) refers to the use of a virtual microscope for the digital scanning of tissue slides, producing Whole-slide Images (WSI). WSIs are very high-resolution and information-rich. Different laboratories and scanners can have slightly different protocols and color variations that can be present even within the same lab and scanner. A WSI can contain a variety of artifacts. These artifacts may result from the tissue being cauterized or from the preparation of the slide. We propose a preprocessing pipeline to cover necessary steps and filters to detect regions of the WSI that should be discarded before further processing with machine learning models.