QuA-WSI

With the large throughput of digitized whole-slide images (WSIs) in the NKI, ensuring high-quality image standards for each WSI is important for both pathologists and artificial intelligence (AI) models. Numerous artifacts can be introduced during the (pre-)processing and digitization of WSIs, which can negatively impact pathologists' assessment, as well as AI model predictions. Thus, this project aims to develop a segmentation model for automatic quality assessment and artifact detection in WSIs.

The inclusion of both H&E and IHC stained slides in combination with the diverse artifact classes (such as out-of-focus areas, tissue folding, pen marking, air bubbles, staining artifacts, foreign objects, coverslip and biological containment) will allow for the training of a robust AI model that can segment these noisy areas independent of the used staining method.

Such a tool can be used in the quality control pipeline to improve and ensure the quality of WSIs produced by scanners. Additionally, this research can be used for other digital pathology workflows and (AI) research to reduce the negative effect of WSI quality on model performance. Time and resources will be saved by providing an automated solution to the quality assessment problem of WSIs.