Abstract: Objective:To investigate the application value of artificial neural network in laparoscopic surgery training.
Methods:The prospective cohort study was conducted. A total of 158 trainees from the First Hospital Affiliated to Army Medical University between Semptember and November, 2019 who had no experience in laparoscopic technology were selected for laparoscopic surgery training, including 52 graduate students of surgery from grade 2019, 2018 and 2017, 58 surgeons receiving standardized residency training, 12 interns and 36 refresher physicians. The 158 trainees were divided into two groups using the random number table. Trainees trained by artificial neural network laparoscopic simulator were allocated into artificial neural network group, and trainees trained by box laparoscopic simulator were allocated into general laparoscopic simulator group. Trainees in both groups were trained using the laparoscopic simulator for 10 hours (5-day continuous training, 2 hours per day) on fundamentals of laparoscopic surgery. Observation indicators: (1) comparison of operation grades on laparoscopic simulator before and after training in the two groups; (2) comparison of improvement of the operation grades on laparoscopic simulator after training between the two groups. Measurement data with normal distribution were represented as Mean±SD, comparison within groups was analyzed using the paired t test and comparison between groups was analyzed using the independent sample t test. Measurement data with skewed distribution were represented as M (range).
Results:A total of 158 trainees were selected for eligibility, including 140 males and 18 females, aged from 23 to 34 years, with a median age of 27 years. Of the 158 trainees, 79 were in the artificial neural network group and 79 were in the general laparoscopic simulator group. (1) Comparison of operation grades on laparoscopic simulator before and after training in the two groups: operation grades of the nails transferring, pattern cutting, ligation, sewing knots in vivo and sewing knots in vitro for the artificial neural network group before training were 51.2±4.9, 45.6±3.7, 43.0±3.6, 42.1±3.1, and 39.6±3.1, respectively. The above indicators for the artificial neural network group after training were 78.6±3.0, 76.4±3.9, 79.9±2.5, 78.3±3.5, and 84.1±3.8, respectively. There were significant differences in the above indicators for the artificial neural network group before and after training (t=-42.490, -56.256, -80.373, -70.802, -79.742, P<0.05). The above indicators for the general laparoscopic simulator group before training were 50.1±2.9, 45.4±3.9, 42.7±3.0, 42.3±3.4, and 39.2±4.7, respectively. The above indicators for the general laparoscopic simulator group after training were 70.4±5.0, 69.8±4.0, 72.3±3.3, 72.3±3.5, and 72.8±3.2, respectively. There were significant differences in the above indicators for the general laparoscopic simulator group before and after training (t=-28.942, -42.436, -58.357, -52.322, -53.098, P<0.05). (2) Comparison of improvement of the operation grades on laparoscopic simulator after training between the two groups: improvement of the operation grades in the nails transferring, pattern cutting, ligation, sewing knots in vivo and sewing knots in vitro for the artificial neural network group after training were 27.4±5.7, 30.8±5.0, 36.9±4.1, 36.2±4.5 and 39.5±5.4, respectively. The above indicators for the general laparoscopic simulator group after training were 20.3±6.2, 24.4±5.1, 29.6±4.5, 29.9±5.1 and 33.5±5.6, respectively. There were significant differences in the above indicators between the two groups (t=7.597, 7.946, 10.638, 8.200, 6.969, P<0.05).
Conclusion:The introduction of artificial neural network in laparoscopic surgery training can improve the training effects.