Paper ID: 2204.14226

Recommendations on test datasets for evaluating AI solutions in pathology

André Homeyer, Christian Geißler, Lars Ole Schwen, Falk Zakrzewski, Theodore Evans, Klaus Strohmenger, Max Westphal, Roman David Bülow, Michaela Kargl, Aray Karjauv, Isidre Munné-Bertran, Carl Orge Retzlaff, Adrià Romero-López, Tomasz Sołtysiński, Markus Plass, Rita Carvalho, Peter Steinbach, Yu-Chia Lan, Nassim Bouteldja, David Haber, Mateo Rojas-Carulla, Alireza Vafaei Sadr, Matthias Kraft, Daniel Krüger, Rutger Fick, Tobias Lang, Peter Boor, Heimo Müller, Peter Hufnagl, Norman Zerbe

Artificial intelligence (AI) solutions that automatically extract information from digital histology images have shown great promise for improving pathological diagnosis. Prior to routine use, it is important to evaluate their predictive performance and obtain regulatory approval. This assessment requires appropriate test datasets. However, compiling such datasets is challenging and specific recommendations are missing. A committee of various stakeholders, including commercial AI developers, pathologists, and researchers, discussed key aspects and conducted extensive literature reviews on test datasets in pathology. Here, we summarize the results and derive general recommendations for the collection of test datasets. We address several questions: Which and how many images are needed? How to deal with low-prevalence subsets? How can potential bias be detected? How should datasets be reported? What are the regulatory requirements in different countries? The recommendations are intended to help AI developers demonstrate the utility of their products and to help regulatory agencies and end users verify reported performance measures. Further research is needed to formulate criteria for sufficiently representative test datasets so that AI solutions can operate with less user intervention and better support diagnostic workflows in the future.

Submitted: Apr 21, 2022