Scholarlink

Purpose <p>Simulation-based training offers considerable benefits in surgical education, especially in mastering minimally invasive techniques such as laparoscopy. However, its widespread adoption faces challenges, particularly in teaching scalability and consistent assessment of trainees’ performance. Artificial Intelligence (AI) has emerged as a promising tool to address these issues, offering possible solutions for optimizing exercise evaluation in simulation programs. This study aims to validate the integration of the YOLO v4 artificial intelligence model into a laparoscopic simulation training program from a video-based feedback-oriented platform.</p> Methods <p>Focusing on object detection and time measurement, the evaluation included a dataset of 7673 videos, extracting 100 random video frames per exercise for object detection with precision, recall, and F1-score calculations based on AI expert tags for each frame, and using 80 videos per exercise for time analysis compared to established times by platform expert teachers.</p> Results <p>With a total of 1100 frames assessed, the YOLO v4 model consistently exhibited high precision with a mean precision of 0.94 and mean recall of 0.89 across object detection tasks, with F1-scores ranging from 0.81 to 0.97. Time measurement accuracy based on 880 assessed videos showed a correlation with mean Pearson correlation coefficient of 0.960 for mean absolute error (MAE) across exercises with the timings of platform expert evaluators.</p> Conclusions <p>The YOLO v4 model proves to be an effective tool for laparoscopic training assessments. The model's high precision and strong correlation with expert assessments demonstrate its usefulness in optimizing the evaluation process in simulated laparoscopic training exercises. This intervention leads the way in surgery education by effectively addressing the need for scalable and standardized assessment methods facilitating the transition towards automated evaluations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A scalable solution: effective AI implementation in laparoscopic simulation training assessments

  • 期刊论文

  • 发表日期：

    2025年4月3日

查看全文
  • Enrique Cruz,
  • Rafael Selman,
  • Úrsula Figueroa,
  • Francisca Belmar,
  • Cristián Jarry,
  • Diego Sanhueza,
  • Gabriel Escalona,
  • Martín Carnier,
  • Julián Varas

摘要

Purpose

Simulation-based training offers considerable benefits in surgical education, especially in mastering minimally invasive techniques such as laparoscopy. However, its widespread adoption faces challenges, particularly in teaching scalability and consistent assessment of trainees’ performance. Artificial Intelligence (AI) has emerged as a promising tool to address these issues, offering possible solutions for optimizing exercise evaluation in simulation programs. This study aims to validate the integration of the YOLO v4 artificial intelligence model into a laparoscopic simulation training program from a video-based feedback-oriented platform.

Methods

Focusing on object detection and time measurement, the evaluation included a dataset of 7673 videos, extracting 100 random video frames per exercise for object detection with precision, recall, and F1-score calculations based on AI expert tags for each frame, and using 80 videos per exercise for time analysis compared to established times by platform expert teachers.

Results

With a total of 1100 frames assessed, the YOLO v4 model consistently exhibited high precision with a mean precision of 0.94 and mean recall of 0.89 across object detection tasks, with F1-scores ranging from 0.81 to 0.97. Time measurement accuracy based on 880 assessed videos showed a correlation with mean Pearson correlation coefficient of 0.960 for mean absolute error (MAE) across exercises with the timings of platform expert evaluators.

Conclusions

The YOLO v4 model proves to be an effective tool for laparoscopic training assessments. The model's high precision and strong correlation with expert assessments demonstrate its usefulness in optimizing the evaluation process in simulated laparoscopic training exercises. This intervention leads the way in surgery education by effectively addressing the need for scalable and standardized assessment methods facilitating the transition towards automated evaluations.