By Victoria Higgins, PhD, FCACB
On Tuesday morning, an eager audience gathered for the much-anticipated session, “CCJ hot topics: designing and implementing artificial intelligence in laboratory medicine.” Moderated by Jason Park, MD, PhD, DABCC, the session provided a comprehensive overview of both the promise and practical challenges of bringing artificial intelligence (AI) into the clinical lab.
AI has long been a subject of research, but with the rapid rise of tools like ChatGPT, there has been a renewed and urgent focus on how it can impact every facet of life — including laboratory medicine. “We have already seen AI-based disruption in education and many types of jobs,” Park said. “I am certain AI will change the practice of laboratory medicine over the next 10 years.”
Despite a growing body of studies over the past five years, few AI-based algorithms have been implemented in the clinical lab. Major obstacles remain. For example, many promising findings can’t be replicated across different populations or institutions, and the results often remain confined to single sites. The session tackled these issues head-on by summarizing the current guidelines for AI in laboratory medicine and emphasizing how to design AI studies that are both reproducible and generalizable.
Park opened the session by welcoming attendees and introducing the speakers. Shannon Haymond, PhD, DABCC, FADLM, presented first, covering the risks of implementing AI in laboratory medicine and clinical guidelines for using it, focusing on the need for both technical robustness and human involvement when integrating AI.
Then, He Sarina Yang, PhD, DABCC, FADLM, talked about the key elements required for AI research to achieve broader clinical impact, ensuring reproducibility and generalizability. To demonstrate these essential aspects, she highlighted recent Clinical Chemistry publications, including the use of machine learning to interpret serum protein electrophoresis and immunofixation electrophoresis.
A lively Q&A followed, with discussions focused on the challenges of validating and quality-controlling AI models, including adaptive ones, and how approaches may differ depending on whether models are used for diagnosis or operational workflows.
For Park, the speed of AI’s adoption drove home the urgency of establishing standards. “The reality of AI’s impact hit me when I saw the speed of adoption of ChatGPT,” he explained. “Middle school students with zero interest in computer science [were] exchange[ing] tips on using large language models … ,” while at the same time there was “a rise in manuscript submissions to Clinical Chemistry … on the use of machine learning to create new diagnostic or predictive algorithms,” Park said. “I felt the need to create standard approaches on evaluating the utility and generalizability of these technologies.”
While radiology continues to dominate the AI-enabled devices approved by the U.S. Food and Drug Administration, laboratory medicine is beginning to make inroads. AI applications have been developed for analyzing images, interpreting test results, and generating reports — often tailored for use within individual hospital systems.
The session offered value for both laboratory professionals and AI researchers. While the former gained insight into AI’s potential and hurdles to implementation, the latter came away with a clearer understanding of the quality standards needed for clinical implementation. The speakers hope attendees left with an understanding of the latest AI applications in lab medicine and how to evaluate the quality of AI research and applications.
Looking ahead, Park emphasized the profound impact AI will likely have. “I think within 10 years, AI-based applications will be ubiquitous in laboratory medicine and all of society,” he said. “Having lived through the internet evolution, I will only predict that AI will impact our profession in ways that we cannot currently imagine.”
As AI continues to evolve, sessions like this one will be essential in guiding its thoughtful and responsible integration into healthcare.
