SIIM-ACR Data Science Summit 2024

As more of us integrate imaging AI into our clinical practices, we're recognizing the need to ensure that the AI tools we use will both initially reliably perform as expected and maintain that functionality over time. At this year's SIIM-ACR Data Science Summit, co-hosted with the Society for Imaging Informatics in Medicine (SIIM) meeting in Washington, DC, on Wednesday, June 26, we center our focus on the theme "From data to diagnosis with tested, transparent, and trustworthy AI". Our sessions will examine the attributes of trustworthy AI, including ethics, evidence, explainability, and equity, while also addressing the origins and implications of bias in AI development and use.  Dr. Paul Yi, Director, University of Maryland Medical Intelligent Imaging (UM2ii) Center and recipient of an ACR Innovation Grant for research in AI bias, sets the stage for the day with his keynote presentation titled, "The Road to Biased AI Is Paved With Good Intentions: Lessons Learned in Developing Trustworthy AI in Radiology". This overview promises insights into the root causes for AI bias and offers strategies for end-users to proactively identify and mitigate risks for AI bias before they develop.  

To be effective, radiologists and our patients must trust the healthcare AI applications for patient care. Our next session delves into the characteristics of trustworthy AI encompassing ethical considerations, transparency, explainability, reliability, and safety across the AI landscape. We'll emphasize the critical role of transparency and explainability in ensuring trustworthiness in AI development and provide participants an introduction to some practical strategies for implementing and monitoring their AI products. 

Before integrating AI tools in routine clinical practice, radiologists must have confidence in the robustness and reliability of their products. Our next session further explores the theme of trustworthiness by examining real-world solutions for testing AI products for dependability throughout the AI lifecycle. With the increasing use of language models in healthcare, radiologists will require assurances that the information they receive is accurate. In this session we also explore some best practices and testing strategies for using generative AI in healthcare. Finally, we will take a comprehensive look at AI monitoring, examining methods to achieve the goal of continuous AI monitoring to provide real-time quality assurance. 

An integral aspect of ensuring AI safety for our patients involves guaranteeing that the algorithms we employ are fair, unbiased, and uphold patient privacy and individual rights. Our Summit co-chair, Dr. Howard Chen, will lead a session covering a 2024 overview of AI bias detection, Image IntelliGently, which highlights challenges facing the development and use of safe and effective AI for pediatric imaging AI, and potential biases in large language models. 

The Summit will conclude with a session moderated by Christoph Wald, ACR Informatics Commission Chair and Vice-chair of the ACR Board of Chancellors, that will update attendees on activities of the ACR and specific resources from the College that promote trustworthy AI, including updates to AI Central, the ACR AI Lifecycle Project, and the ACR Recognized Center of Excellence for Healthcare AI (ARCH-AI) and their implications in the regulatory process.  

As in years past, our aim is to ensure the 2024 SIIM-ACR Data Science Summit remains relevant to radiologists, data scientists and industry leaders. We focus on important aspects of the successful implementation of trustworthy imaging AI for end-users and patients. Each session includes a panel discussion where participants can ask questions and provide additional insights. Whether attending virtually or in person in Washington, DC, the 2024 Data Science Summit offers a unique opportunity for attendees to collaborate, learn, and engage with the critical issues shaping the future of AI in medical imaging.