# Machine learning combine with nomogram to guide the establishment of endoscopic assistant system for gasless transaxillary endoscopic thyroidectomy

## TL;DR

This paper uses machine learning and a nomogram to develop a system for training endoscopic assistants in gasless transaxillary endoscopic thyroidectomy, aiming to improve surgical efficiency.

## Contribution

A novel endoscopic assistant system combining machine learning and nomogram analysis to optimize surgical processes in thyroidectomy.

## Key findings

- The learning curve coefficient of goodness of fit was R² = 0.807, indicating strong model performance.
- Skilled assistants showed less fluctuation in surgical time during the mastery stage compared to unskilled assistants.
- Multivariate analysis identified CET, LWT, CUSUM, and TRT as statistically significant factors affecting the surgical process.

## Abstract

To explore the influence related factors of endoscopic assistant in gasless transaxillary endoscopic thyroidectomy by using machine learning and nomogram, and construct an endoscopic assistant system.

A skilled endoscopic assistant(Group A, n = 50)and an unskilled endoscopic assistant(Group B, n = 50)were randomly included in participating operation, CUSUM was calculated and learning curve was constructed. Age and other factors were included to study CET, TRT, TST, LWT and CUSUM. Univariate and multivariate analysis were conducted to find out the relevant factors affecting the surgical process. Nomogram and machine learning were constructed to find out the influence of relevant factors on the surgical process.

The learning curve coefficient of goodness of fit R2=0.807. The cases in the learning stage of skilled assistant and unskilled assistant were 20 and 28. In mastery stage, the surgical time of skilled assistant had less fluctuation than that of unskilled assistant. There were statistical significance in CET(p = 0.001), TST(p = 0.001), LWT(p = 0.002), CUSUM(p = 0.019). In multivariate analysis, CET(p = 0.004), LWT(p = 0.013), CUSUM(p = 0.025), TRT(p = 0.018) showed statistical significance. Nomogram was successfully constructed based on the relevant factors explored by the multivariate analysis, and the influence of each relevant factors were explored by machine learning. The system was constructed through preclinical training, preoperative preparation, and share experience from the perspective of the endoscopic assistant according to the procedure process.

It is necessary to train endoscopic assistant to build an endoscopic assistant system, and improve the surgical process by shortening CET, TRT and reduce LWT times. The importance of experience accumulation to improve the efficiency of surgery should be emphasized.