Objective
This study aims to develop an automated surgical skill assessment tool using deep learning technology for surgical training.
Background
In robot-assisted surgery (RAS), precise surgical operations are crucial for prognosis. Currently, the training of novice surgeons relies on the on-site guidance and skill assessment by experienced physicians, and artificial intelligence (AI) technology is expected to enhance the objectivity of the existing assessment system.
Methods
A dual-network architecture consisting of convolutional neural networks and long short-term memory networks (LSTM) was constructed to extract spatiotemporal features from surgical videos, enabling surgical action recognition and skill assessment.
Results
Twenty-one participants (16 novice surgeons and 5 experienced surgeons) completed 16 different laparoscopic robot-assisted surgeries on a pig model.The action recognition network achieved an accuracy of 96.0% in recognizing surgical operations, and the GradCAM filter was applied to enhance model interpretability; the skill assessment network achieved a classification accuracy of 81.3% for both novice and experienced surgeons, generating a visual skill assessment map.
Conclusion
The study indicates that AI can be used for automated surgical action recognition and skill assessment. The pig model provides a systematic collection of data across different skill levels, which is difficult to obtain in clinical settings.Future research needs to validate the model’s effectiveness in identifying operational errors, providing real-time feedback, and offering actionable skill assessments in real clinical environments.
👆Long press to save and view anytime
Journal Introduction
Surgical Endoscopy is a well-known journal focusing on gastrointestinal, endoscopic techniques, and innovative interventional surgical technologies.
-
Published in collaboration with the American Society of Gastrointestinal and Endoscopic Surgeons (SAGES) and the European Association for Endoscopic Surgery (EAES) since 1987.
-
Provides an academic platform for discussions on surgical practice and innovative technologies.
-
Offers early professional guidance to researchers from a renowned editorial team.
-
Provides high exposure for published articles through societies, libraries, and social media.
👆Long press to save and view anytime
Click “Read the original text” to view the English original
