EurAI Advanced Course on Artificial Intelligence 2022 + TAILOR SUMMER SCHOOL
Framing Trust in Medical AI
Lecturer: Jose M. Juarez, University of Murcia
Overview
In this tutorial we will introduce the medical Artificial Intelligence field and how to handle the major concerns of doctors to adopt AI-based technologies in daily clinical practice.
The development of trustworthy AI system is multidisciplinary requiring ethical, legal, and technological measures.
During this tutorial, the students will have a general vision of existing initiatives made by the European Union on the ethical and legal framework related to AI in healthcare (ethics guidelines, GDPR, AI , medical devices). The tutorial will overview most popular explainable methods on AI (e.g. LIME, SHAP, saliency maps) highlighting their advantages and drawbacks from the clinician’s perspective. To illustrate the different challenges of medical AI, this tutorial is driven by several examples obtained from the recent literature on the AI and medical fields.
Objectives
- To gain a better understanding of applying AI in healthcare settings.
- To identify principal interdisciplinary factors for a trustworthy AI project according to EU guidelines.
- To be aware of advantages and limitations of current eXplainable AI.
Materials, references and links
Slides \& Materials
Part I: Humans, Healthcare and Ethics
- Algorithm Watch: link
- Building Trust in AI: link
- EU Ethics Guidelines for Trustworthy AI: link
- Ethics and AI, prof Guido Boella. University of Turin - PhD online course organized by SIpEIA.
Part II: AI Understanding in Healthcare
- SHAP [Lundberg-Lee2017]Scott Lundberg and Su-In Lee. A unified approach to interpreting model predictions. CoRR, abs/1705.07874, 2017. arXiv:1705.07874. GITHUB LIBRARY
- LIME [Ribeiro et al 2016] Ribeiro, Marco Tulio, Sameer Singh, and Carlos Guestrin. Why should I trust you?: Explaining the predictions of any classifier. Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining. ACM (2016). LIBRARY WEBSITE
- [Amann 2020] Amann, J., Blasimme, A., Vayena, E. et al. Explainability for artificial intelligence in healthcare: a multidisciplinary perspective. BMC Med Inform Decis Mak 20, 310 (2020). https://doi.org/10.1186/s12911-020-01332-6
- [Gu et al 2019] Gu J, Tresp V. Saliency methods for explaining adversarial attacks. arXiv 2019; published online Aug 22. http://arxiv.org/ abs/1908.08413 (preprint).
Part III: Regulations for the Medical Industry
- GDPR: LINK
- EU Artificinal Intelligence Act (proposal) LINK
- EU Medical Device Regulation LINK
Other links: Conferences and workshops on XAI
- XAI-Healthcare: International Workshops on eXplainable AI in Healthcare LINK
- X-KDD: International Workshops on eXplainable Knowledge Discovery in Data Mining LINK
- EXTRAAMAS: International Workshops on EXplainable and TRAnsparent AI and Multi-Agent Systems LINK
