MIDL Vision 2022

Metrics Reloaded - A new recommendation framework for biomedical image analysis validation Annika Reinke is a PhD student at DKFZ in Heidelberg, Germany. Her paper is a position paper supported by many scholars in our field. She told us about her work ahead of her presentation today - Poster Session 2.2 - Onsite 11:00-12:00, Virtual 15:20-16:20 There is increasing evidence that flaws related to the validation of AI algorithms are an underestimated global problem . Specifically in the field of automatic biomedical image analysis, the chosen performance metrics often do not reflect the domain interest and thus hinder translation of AI techniques into practice. To overcome this roadblock, Annika and team formed an international consortium of over 60 image analysis expert with the mission to develop a comprehensive framework that guides researchers towards choosing performance metrics in a problem-aware manner. Specifically, they focus on biomedical image analysis problems that can be interpreted as a classification task at image, object or pixel level, namely image-level classification, semantic segmentation, instance segmentation, and object detection tasks. 16 Poster Presentation VISION MIDL