The lifespan of the automated anesthesiologist was short yet tempestuous; obtained US Food and Drug Administration (FDA)’s approval in 2013, reported to be used by four hospitals in 2015, and left the market in March 2016. During which many an anesthesiologist and nurse anesthetist got offended by its promised to replace them and the American Society of Anesthesiologists had to campaign against it.
The Sedasys machine was not as revolutionary as its earlier counterpart – McSleepy, which successfully performed the first ever robotic surgery with DaVinci on a prostatectomy patient in 2010, but it was the first to go commercial. That’s why medical professionals are eager to question its safety and reliability in the long run, although its main role was to sedate patients undergoing colonoscopy or gastrointestinal (GI) endoscopy.
Lessons learnt from the automated anesthesiologist
Intestinal experts said the use of Propofol in Sedasys probably marked its downfall since it’s a kind of drug with wide variability which makes it challenging to achieve consistent mild to moderate sedation. Besides, the standard Sedasys protocol requires six minutes while most diagnostic upper GI endoscopy procedures can be accomplished within 5-10 minutes, this seriously undermines the efficiency of Sedasys.
Coupled with the discrepancy between the increase demand for deep Propofol sedation or general anesthesia but FDA does not approve computer-assisted personalized sedation system (CAPS) that administer beyond moderate sedation. Unless drastic changes occur along the way, we may never witness Sedasys and other related products’ return.
What else can AI do to help a human anesthesiologists?
Dr. James Philip, director of clinical bioengineering at Brigham & Women’s hospital’s department of anesthesiology and perioperative and pain medicine professor of Harvard medical school said in the World medical innovation forum took place this April, there is a need for more data and variables. Presently, anesthesiology related AI is in need of more data.
Pharmacologic variables like the effect and secondary effect of each drug on patients, time taken for these effects to show, their flow profiles, kinetics and dynamic are all missing yet they are helpful for anesthesiologists to provide better care. There is also a need to present these data thoughtfully and display them clearly for clinicians to access what actually marks a good or bad outcome in a patient, so that the overwhelming data will not go to waste.
The more realistic pathway for AI and machine learning, will probably be creating heuristics for anesthesiologists and have experts stepping into related research. At the same forum, other anesthesiologists and technology experts had showcased their latest research including a programme which aims to recognize patients who may experience difficult intubation via face ratio and use of AI to determine patients’ consciousness. After all, if it’s a dexterity challenge for AI to make patients “sleep”, perhaps they could help anesthesiologists to help us “sleep better”.