Table 2.

State-of-the-art studies proposing surgical skill and subtask classification models

Author	Year	Population	Setting	Tasks	Data	Classes	Model*	Accuracy
Wang Y. et al. [35]	2021	18	RAS, laboratory setting	suturing	video recordings	skill level: novice, intermediate, expert	DL	83%
Soangra et al. [36]	2022	26	laparoscopic simulator and RAS, laboratory setting	peg transfer, knot tying	kinematic data and electromyogram	skill level: novice, intermediate, expert	ML	58%
Law et al. [37]	2017	29	RAS, operating room	robotic prostatectomy	video recordings	skill level: binary (good vs. poor)	DL, ML	0.92
Natheir et al. [38]	2023	21	three simulated brain tumor resection procedures on the neuroVR™ platform, laboratory setting	brain tumor resection procedures	EEG	skill level: binary (skilled vs. less skilled)	ML	85%
Zappella et al. [39]	2013	8	RAS, laboratory setting	suturing, needle passing, knot tying	video and kinematic data	task detection: suturing, needle passing, knot tying	DL, ML	80%–94%
Wang et al. [40]	2018	8	RAS, laboratory setting	suturing, needle passing, knot tying	video and kinematic data	skill level: novice, intermediate, expert	DL	91%–95%
Current study	2024	11	RAS, operating room	blunt, cold sharp, and thermal dissection subtasks throughout cystectomy, hysterectomy, and nephrectomy operations	EEG and eye-tracking	skill level (inexperienced, competent, experienced) and subtask type (blunt, cold sharp, and thermal dissection); 9 classes	ML	83%–88%

*ML Machine Learning, DL Deep Learning