According to MIT and Massachusetts General Hospital (MGH) researchers, artificial intelligence systems may be able to help Anesthesiologists in the operating room. A group of neuroscientists, engineers, and physicians demonstrated how a machine-learning system could continually automate the dosage of the anaesthetic drug propofol in a special edition of Artificial Intelligence in Medicine.
In patient simulations, the programme outperformed traditional methods. Using recorded data from nine actual procedures, it also nearly mirrored the performance of real Anesthesiologists in terms of sustaining unconsciousness. The advances in the algorithm increase the possibility of using computers to maintain patient unconsciousness, freeing up Anesthesiologists to focus on other tasks in the operating room.
Like ensuring that patients remain immobile, pain-free, physiologically stable, and receive adequate anaesthesia. The researchers devised a machine-learning method for learning not only how to dose propofol but also how to maximise the amount of medicine supplied. According to the press release, the research team achieved this by incorporating two related neural networks into the software: a “actor” that determines how much of the drug to dose at any given time, and a “critic” that assists the actor in behaving in a way that maximises the “rewards” specified by the programmer.
For example, the researchers tested with three different rewards while training the algorithm: one that penalised only overdose, one that questioned supplying any dose, and one that had no penalties. In each case, the researchers used complex models of pharmacokinetics, or how quickly propofol doses reach the important brain areas after administration, and pharmacodynamics, or how the medication modifies consciousness once it reaches its target, to train the algorithm.