Artificial intelligence to better measure anesthesia doses


In an interesting new intersection between medicine and artificial intelligence, a research team based at MIT, in collaboration with the General Hospital of Massachusetts, United States, was presented a machine learning mechanism that collaborates in the dosage of anesthesia for surgical interventions.

Through a system defined as "simple" by its creators themselves at the time of being made known, it is possible to evaluate the level of unconsciousness of a patient based on somebody's indicators, after the application of a drug.

Precise anesthesia dosage thanks to the help of AI

Anesthesia drugs are a complex universe. There are various formulas, presentations, and effects that apply to an infinity of specific cases.

Part of the success of a surgical intervention depends on the correct administration of these substances since this influences the degree of unconsciousness that is achieved in patients and in the relaxation of certain indicators such as heart or respiratory rate.

The mechanism presented by MIT, following the type of anesthetic used, can produce algorithms that evaluate with high accuracy and reliability, the effect that its application has on patients, depending on their brain activity.

With this new resource, those engaged in anesthesiology would be relieved of significant pressure, as the indicators often used to assess these effects are often superficial and indirect.

Thanks to the AI presented, the new algorithms give anesthesiologists the possibility of keeping the effects of a drug within an expected level, facilitating a timely and efficient administration and reducing the possibility of generating episodes such as crises or postoperative delusions.

The first trials of this study began in 2013, with 10 healthy volunteers, in their twenties, who underwent anesthesia with the commonly used propofol drug. The administration of the drug was controlled by computer, methodically increasing the dose depending on the response of the volunteers to simple requests, about which they had to respond until they could no longer. Then, when they regained consciousness from dose reduction, they regained the ability to respond again. Subsequently, the test was extended to 27 patients of real surgery, who received the same anesthetic drug, and then, expanded the study with sevoflurane, another substance of the same category, with a sample of 17 real patients.

The probes associated with these hearings were measured using electroencephalogram electrodes, used to record the nervous rhythms that reflect the brain activity of the patients, allowing to monitor the brain activity and unconsciousness in real-time.

See the source image

Among the conclusions highlighted by the authors, it is commented that there was a case in which the algorithms were able to detect the decreasing level of unconsciousness of a patient several minutes before the actual assistant anesthesiologist did, which means that if it had been in use during a real surgical intervention, could have provided a timely alert.

And although the bases of this initiative have been discussed in previous research, what MIT presented is the first example of a mechanism of this kind that can distinguish and adapt to different types of drugs and in the future, also to the different ages of patients.

The algorithms resulting from this work are not computationally demanding. The authors noted that for about 2 given seconds of analysis data, the algorithms could make an accurate prediction of a person's level of consciousness in less than a tenth of a second, running only on a standard MacBook Pro computer.

Post a Comment