Responsibly integrating machine learning and AI into lab medicine

The infusion of machine learning (ML) and artificial intelligence (AI) into clinical laboratories heralds a new era of analytical prowess, promising unprecedented operational efficiency and diagnostic accuracy. However, poorly selected, developed, or managed ML solutions can harm patients, decrease laboratory efficiency, exacerbate health disparities, and expose institutions to security and privacy breaches. Hence, the importance of meticulous quality management cannot be overstated.

Unfortunately, many laboratorians are unfamiliar with these emerging technologies and may be unprepared to assume responsibility for them. Monday afternoon’s scientific session, titled “Operationalizing AI in Lab Medicine: Approaches for Effective Machine Learning Integration and Deployment,” will address the urgent need for practical guidance. Led by three distinguished informaticists with expertise in AI and ML, it will provide invaluable insights into how to effectively implement, monitor, and manage these tools within laboratory settings. The focus will be on ensuring consistent, high-quality results for both in-house and vendor-based AI/ML solutions.

Thomas Durant, MD, will kick off the session by introducing AI/ML applications in the clinical laboratory. He plans to make his presentation accessible to lab professionals at all levels of expertise and cover current and future use cases, describing how models can be both developed and deployed.

Next, David McClintock, MD, will describe how to critically appraise ML/AI tools, whether developed internally, externally, or collaboratively. His comprehensive presentation will cover accuracy, ease of deployment, ethics, and regulations.

Finally, Nicholas Spies, MD, will present a case study that describes a real-world deployment of an AI tool to reinforce the lessons from the first two presentations. The session will conclude with a panel discussion with all three presenters.

A key point attendees will take away from this session is that developing an accurate AI/ML model is an early step toward successful clinical application. Much of the cost and effort lies in the deployment and maintenance phases. Attendees will acquire the necessary knowledge and skills to responsibly wield the awesome power of AI/ML.