SBI Webinar: Navigating Imaging AI Bias
Bias in artificial intelligence (AI) is a complex, context-dependent topic. Understanding and identifying sources of bias in imaging AI can mitigate potential adverse impact on patients. Commercially available breast imaging AI solutions may propagate statistical bias inherent in model develop, such as selection bias, and demonstrate data shift, adversely impacting model performance without physician awareness. “Out-of-the-box” breast imaging models may not be generalizable to all patients and can perpetuate health disparities.
Understanding sources of imaging AI bias enables creation of specific strategies to mitigate its impact. Furthermore, awareness of imaging AI bias can translate to new and unfamiliar technologies, including natural language processing/large language models, equipping users to prospectively address potentially harmful sources of AI bias.
In this webinar, Dr. Ali Tejani provides a foundation for imaging AI bias in the context of breast imaging. He discusses context-dependent definitions of bias in imaging AI and enumerates common sources of bias, proposing specific strategies to detect imaging AI bias and mitigate its impact. Accordingly, this webinar will provide tools for attendees to assess the current state of AI initiatives at their own institution and develop strategies to address potential sources of bias. Additionally, this webinar provides actionable steps for attendees to actively participate in pre-deployment evaluation and continued monitoring of AI solutions.
CME Information
Earn up to 1 AMA PRA Category 1 Credit(s)TM
CME Released: 08/31/2023 CME Expires: 8/31/2026
Statement of Accreditation- This activity has been planned and implemented in accordance with the accreditation requirements and policies of the Accreditation Council for Continuing Medical Education (ACCME) through the joint providership of Michigan State Medical Society and Society of Breast Imaging. The Michigan State Medical Society is accredited by the ACCME to provide continuing medical education for physicians.
AMA Credit Designation Statement- The Michigan State Medical Society designates this live activity for a maximum of 1 AMA PRA Category 1 Credit(s)TM. Physicians should claim only the credit commensurate with the extent of their participation in the activity.
- Understand different types of imaging AI bias.
- Implement strategies to prospectively mitigate the impact of bias prior to deploying AI algorithms.
- Identify sources of bias influencing the performance of breast imaging AI solutions.
- Understand the role of breast radiologists as domain experts in guiding AI model creation.
