Explore AI’s evolution, challenges, and breakthroughs at RSNA23. Leaders, including Connie Lehman, MD, PhD, Etta D. Pisano, MD, and Fredrik Strand, MSc, MD, PhD, delve into clinical implementation challenges, cutting-edge research outcomes, and regulatory landscapes. Lehman’s historical review traces AI’s journey from the 1950s to the present concerns. Strand’s studies showcase AI’s potential for efficient breast cancer detection. Pisano dissects U.S. regulatory pathways, urging real-world evidence exploration. The Q&A session provides valuable insights, emphasizing the transformative impact on breast imaging paradigms.
RSNA 2023 emerged as a pivotal arena for the convergence of top leaders in the field of artificial intelligence (AI) for breast imaging. The first day of the 109th Radiological Society of North America Scientific Sessions and Annual Meeting (RSNA23) commenced with a session titled “AI in Breast Imaging,” featuring esteemed speakers Connie Lehman, MD, PhD, FACR, FSBI, from Massachusetts General Hospital; Etta D. Pisano, MD, FACR, from the University of Pennsylvania; and Fredrik Strand, MSc, MD, PhD, from the Karolinska Institute, Stockholm. This session laid the foundation for comprehensive discussions on the history, opportunities, challenges, and the latest updates in AI for breast imaging.
Advanced Technology Evolution in Breast Imaging Unveiled at RSNA 2023
Connie Lehman, a Professor of Radiology at Harvard Medical School, initiated the discourse by delving into the challenges and opportunities inherent in the clinical implementation of AI in breast imaging. She highlighted the early stage of implementation and raised critical questions regarding the translation of reader studies or simulations to community practice and the global applicability of European study results. Lehman emphasized the diverse clinical settings worldwide, including variations in reading protocols, specialist roles, and recall rates. Additionally, she underscored the financial uncertainties surrounding AI adoption, contemplating the value derived from time savings and potential reimbursement models. Offering a historical perspective, Lehman traced the evolution of AI from its inception in the 1950s to the contemporary concerns and regulatory trends observed in the 2020s. She concluded by emphasizing the transformative potential of AI in reshaping screening paradigms through global collaborations and rigorous clinical science.
Fredrik Strand, from the Karolinska Institute, presented findings from three significant research studies, published in esteemed journals such as The Lancet Oncology, The Lancet Digital Health, and Nature Medicine. These studies focused on AI applications for breast cancer detection in screening mammography, employing platforms from vendors like Screenpoint, Lunit, and Kheiron. Strand’s insights highlighted the potential for AI to enhance efficiency in breast cancer detection, particularly in comparison to traditional double-reading methods. He emphasized the varying workloads and detection outcomes associated with AI standalone, AI combined with radiologists, and AI triaging and informing. Additionally, Strand provided key considerations for successful AI implementation, including safety validation, calibration of AI thresholds, and understanding score distributions.
Etta D. Pisano, representing the American College of Radiology (ACR), offered a comprehensive overview of the U.S. regulatory landscape for breast AI. She delineated the intended uses of AI and machine learning software in areas such as triage, detection/localization, diagnosis/characterization, and acquisition/optimization. Pisano addressed the limitations of standalone performance testing and reader studies, urging a critical examination of data representativeness and reader diversity. She also questioned the adequacy of clinical trials, emphasizing the time and expense involved. Pisano proposed the exploration of real-world evidence as a pathway for autonomous AI in breast screening and suggested installing AI algorithms in clinics before FDA authorization for practical insights. Moreover, she highlighted the regulatory advancements in the EU and the UK, with the UK National Health Service planning a Real World Evidence AI trial.
The Q&A session and panel discussion further enriched the discourse, with Lehman offering advice to early career radiologists, encouraging them to embrace the impact they can make in this dynamic subspecialty. The RSNA23 session on AI in breast imaging set the stage for continued exploration and collaboration, emphasizing the need for thoughtful clinical science, regulatory oversight, and global partnerships to harness the full potential of AI in transforming breast imaging practices. As the field advances, these insights from RSNA23 are poised to guide future developments and innovations in AI for breast imaging.
