I have just read what is, in my opinion, the best and most comprehensive treatment to date of the potential legal risks for use of artificial intelligence (AI) technology in radiology in an article now available online by the American Journal of Roentgenology. The author, Jonathan Mezrich, MD, JD, MBA, LLM - Assistant Professor in the Department of Radiology and Biomedical Imaging at Yale University School of Medicine - has brought clarity and understanding to his summary and perspective of how the legal system may respond when patients suffer adverse consequences allegedly caused by AI.
Dr. Mezrich, who is both a radiologist and a lawyer, explores the potential legal considerations of AI in radiology in the context of the U.S. legal system. When inevitable errors and patient injury are alleged to be the result of AI, how will the legal system respond? Traditionally, the legal system aims to make that injured patient (or their family) whole through civil litigation, often resulting in monetary damages awarded by the court or through a settlement between the litigants.
The article explores the types of tort law that could be implicated with AI in the health care setting, including medical malpractice, vicarious liability, and products liability. Dr. Mezrich observes, correctly in my view, that which tort remedy emerges to ultimately govern liability and compensate injured patients may meaningfully impact the advancement of AI and its potential utility in radiology.
The article frames the legal consequences by observing the potential ways AI will be used in patient care, and in how reparations of patient injuries through the legal system could evolve.
After describing the necessary elements that plaintiffs must prove in medical malpractice litigation, Dr. Mezrich explores whether such concepts will be relevant if AI evolves to become an autonomous diagnostician. Is there a "physician-patient relationship when the 'physician' is an algorithm? How is an AI algorithm held to the “reasonably prudent radiologist” (or perhaps “reasonably prudent algorithm”) standard, and who could serve as an expert witness to determine this standard? Is there a different standard of care or expectation for an algorithm, and does the expectation change if the algorithm is performing tasks that go beyond the capabilities of the typical human radiologist (e.g., predicting optimum therapy options or responses based on imaging or genomic lesion characterization)?"
Will products liability become the vehicle for legal reparations with legal claims brought against manufacturers and distributors of AI technology in lieu of claims against radiologists? What will be the impact of FDA regulation of AI software? These fascinating and important considerations are considered by the author.
It is clear that Dr. Mezrich is a proponent of the potential value AI can bring to radiologists and their patients. He hopes that policymakers and legislators will adopt reasonable guardrails that will compensate those who suffer injuries but otherwise not hamper the development of AI in providing better patient care.
AI is dependent on the accumulation of data through regular usage for more accurate machine learning. One hopes that groundbreaking technologies like AI will not stall because of the fear of legal exposure when using those very same technologies have so much potential to deliver better health care outcomes.
For anyone interested in the future of AI in radiology, I suggest you add this AJR article to your reading list.