Table 5.

The application of AI in drug administration control

Study	Aim	Population	AI method	N	Accuracy results	Conclusions
Mendez 2018 [28]	Automatic drug delivery	ASA I and II undergoing general anesthesia for ambulatory surgery	FL closed loop system	85	MDAPE 9,67% (AI) vs. 13,5% (control); MDPE − 3,13% vs. -9,83%;	The system showed better performance parameter values than the manual standard procedure and PID algorithms.
Zaouter 2017 [29]	Closed loop sedation without manual override	Patients undergoing elective Transcatheter Aortic Valve Implantation	Hybrid sedation system	20	MDAPE 23,8%; MDPE − 2,4%;	The System maintained sedation in 95% of cases with no manual override.
Xu 2022 [30]	Computer aided diagnosis system on anesthesia quality control	ASA I to III undergoing sedative EGD and/or colonoscopy	CNN	154	Emergence and recovery time were significantly shorter than that in the control group (p < 0.01)	Satisfaction scores were significantly higher and the emergence and recovery time was significantly shorter.
Syed 2021 [31]	Prediction if a colonoscopy can be successfully completed with moderate sedation	Patients undergoing colonoscopy	Machine learning	10.025	AUC 0,762	The model achieved reliable accuracy in predicting procedures.
Wei 2021 [58]	Assessing intrathecal hyperbaric bupivacaine dose during cesarean section	Term parturient (ASA II to III) presented for elective cesarean section under spinal anesthesia	Machine learning	684	MSE 0,0087; Coefficient of determination 0.807	The model created showed a coefficient of determination of 0.807 to predict intrathecal hyperbaric bupivacaine dose
Marrero 2017 [59]	Predict the hypnotic effect in general anesthesia	Patients undergoing general anesthesia using intravenous hypnotic agent (propofol)	FL	20	RMSE for AFM 36,13 vs. 50,29 for the FM vs. 45,69 for the compartmental model.	The model had significantly better results than fuzzy model and compartment model.
Shieh 2000 [60]	Controller for neuromuscular block with rocuronium	ASA I and II undergoing general anesthesia	FL	10	Standard deviation of the T1%² error 1,82% and the mean T1% error − 0,18%.	The performance is better in intermediate duration of action neuromuscular blocking agents.
Lin 2002 [61]	Predict the hypnotic effect of propofol based on clinical parameters	Patients undergoing elective surgery under total intravenous anesthesia	ANN	270	Sn 82,35%; Sp 64,38%; AUC 0.7552	The model had the ability to predict the hypnotic effect of propofol bolus induction superiorly than the clinician.