Annotation Practices in Computational Pathology: A European Society of Digital and Integrative Pathology (ESDIP) Survey Study
Diana Montezuma, Sara P. Oliveira, Yuri Tolkach, Peter Boor, Alex Haragan, Rita Carvalho, Vincenzo Della Mea, Tim‐Rasmus Kiehl, Sabine Leh, Mustafa Yousif, David Ameisen, Mircea-Sebastian Şerbănescu, Norman Zerbe, Vincenzo L’Imperio
Abstract
Integrating digital pathology and artificial intelligence (AI) algorithms can potentially improve diagnostic practice and precision medicine. Developing reliable, generalizable, and comparable AI algorithms depends on access to meticulously annotated data. However, achieving this requires robust collaboration among pathologists, computer scientists, and other researchers to ensure data quality and consistency. The lack of standardization and scalability is a significant challenge when generating annotations and annotated data sets. Recognizing these limitations, the Scientific Committee of the European Society of Digital and Integrative Pathology (ESDIP) performed a comprehensive international survey to understand the current state of annotation practices and identify actionable areas to address critical needs in the annotation process. The analysis and summary of the survey results provide several insights for all stakeholders involved in data preparation and ground truthing, ultimately contributing to the advancement of AI in computational pathology.