33rd Journal Symposium: AI’s full potential (with human oversight)
Pinpointing machine learning’s benefits in anesthesiology.

SPE15 – 33rd Journal Symposium: Artificial Intelligence, Machine Learning, and Data Analytics in Anesthesiology
1:30-4:30 p.m. | Sunday, October 20
Room 119AB
Artificial intelligence (AI) has the potential to transform perioperative medicine, but it should be viewed as another tool in the anesthesiologist's toolkit, rather than a panacea for perioperative solutions.
Sachin Kheterpal, MD, MBA, the Kevin K. Tremper Professor of Anesthesiology at the University of Michigan Medical School in Ann Arbor, said AI has many potential applications in the data-rich clinical environment of perioperative medicine, with new ones emerging all the time. These include the mature image analysis, classification, and identification that have already been implemented in radiology and ophthalmology as well as large language models that build conversational bots. But these advances, he cautions, should not be taken lightly in the field of anesthesiology.
“There are thousands of data points, waveforms, and pieces of information collected just during the intraoperative period, not to mention the preoperative and postoperative times,” he said. “These massive, time-series data are perfectly suited for modern AI tools that can handle multimodal data streams and millions of features. However, the second-to-second decision-making in the operating room requires timely AI guidance and strong safeguards due to the risk of many of our pharmacologic and procedural interventions. The issue of data shift, where the algorithm’s performance degrades over time, is also an issue for all fields.”
Dr. Kheterpal will be joined by a panel of experts to discuss how AI might benefit anesthesiology in Sunday’s “33rd Journal Symposium: Artificial Intelligence, Machine Learning, and Data Analytics in Anesthesiology.”
Some of the biggest challenges facing AI usage, Dr. Kheterpal said, are not found in the technology itself, but rather in the ethical and regulatory hurdles that must first be cleared before it can be properly integrated into any medical setting.
“The rapidly changing landscape of regulations and ethics makes it very difficult for AI developers to understand the bounds of compliance and ethics,” he said. “Even foundational aspects of AI such as dataset creation, where patient data are aggregated across centers for secondary use, is coming under ethical scrutiny. And the regulatory landscape across countries and jurisdictions is so varied that industry has to prepare for completely different levels of scrutiny, potentially impeding investments.”
Despite these hurdles, Dr. Kheterpal said he sees enormous potential for AI and machine learning, mostly in terms of augmenting what anesthesiologists do on a day-to-day basis.
“Given the fast pace of clinical decision-making in anesthesiology and the high-risk nature of our interventions, AI will continue to augment and advance clinician decision-making rather than replace it,” he said. “The concept of ‘clinician in the loop’ is particularly necessary for anesthesiology. There may be some administrative, back-office functions, such as billing and coding, which more fully embrace AI with occasional human validation and checks, but without a human involved in every interaction.”