Munenori Uemura, PhD1, Morimasa Tomikawa, PhD, MD, FACS1, Tiejun Miao, PhD2, Tomohiko Akahoshi, MD, PhD1, Satoshi Ieiri, MD, PhD1, Makoto Hashizume, MD, PhD, FACS1. 1Department of Advanced Medical Initiatives, Graduate School of Medical Sciences, Kyushu University, 2TAOS Institute, Tokyo, Japan
Purpose
Our previous work focused on the hand motions of expert and novice surgeons during an operation. Kinematic data describing the motions of a surgeon’s forceps during a skill assessment task were analyzed mathematically, revealing new evidence on hand motions during laparoscopic surgery. This method enabled the surgical motions of expert and novice surgeons to be assigned objective, numerical values. It was not clear, however, whether these values accurately reflect surgeons’ skills during actual laparoscopic surgery. This study therefore investigated whether these newly described factors were accurate assessments of skill in performing laparoscopic surgery.
Methods
Using factors previously identified, we constructed a chaos neural network system to distinguish experienced and inexperienced surgeons. The network consisted of three layers: an Input layer, a Hidden layer and an Output layer. The input layer consisted of eight previously identified input factors. The hidden layer consisted of 30 neurons and the output layer of two neurons as identifiers, 1 (expert) and 0 (novice). The skills of 38 surgeons were analyzed, 11 expert and 27 novice surgeons. The data for each were those obtained in our previous study. Data were optimized until the two groups of surgeons were correctly distinguished.
The examinees in this study consisted of 12 expert surgeons, each of whom had performed more than 500 laparoscopic operations and who had completed the skill assessment task (expert group); and 17 young surgeons, each of whom had performed fewer than 15 laparoscopic operations and had not completed the skill assessment task (novice group). None of these 29 examinees had been included in our previous study.
Objective datasets for each examinee included the paths of the tips of needle holders. To detect differences between the expert and novice groups, we computed the paths of the centers of gravity and relative paths of these tips.
A blinded test was performed to evaluate the newfound factors using the chaos neural network system. All randomized data were inputted into the system. Based on the factors previously identified, the system was analyzed for its ability to distinguish expert from novice surgeons, with the primary outcome being system success rate.
Results
Optimization of the chaos neural network system during mechanical learning was completed at trial 7th. The correlation between inputted supervised learning data and output data was 0.99. The chaos neural network system had a success rate of 79.3% in distinguishing between expert and novice surgeons.
New skill evaluation software was developed that based on these results and is able to indicate skill score rank.
Conclusion
Using the factors identified during our previous study, the chaos neural network system was able to distinguish between expert and novice surgeons and both from surgeons with an unknown skill level. Our proposed method may be useful in the future training and assessment of laparoscopic surgeons.