Tissue detection is an important first step before applying any of the computational algorithms on the whole slide images.

Most of the digitized pathology tissue sections contain histological artifacts such as out-of-focus areas, tissue folds, black spots, air bubbles and many others.

GrandQC is a fully automatized, powerful deep learning-based tool that effectively segments tissue from whole slide images and detects/segments seven most common types of artifacts.

This allows masking these regions from any further processing by the algorithms and helps to decide if recut / rescan is necessary.

GrandQC is available at:

https://github.com/cpath-ukk/grandqc

In our paper

"Quality control stress test for deep learning-based diagnostic model in digital pathology" in Modern Pathology 34, 2021 (DOI),

we suggest a quality control stress test pipeline using a number of synthetically reproduced histological artifacts.

The instructions of how you can generate the artifacts to stress test you models can be found in https://github.com/cpath-ukk/Artifact