Related Experiment Video
Updated: Jul 9, 2026
05:40
Interrogating Cell-Cell Interactions in the Salivary Gland via Ex Vivo Live Cell Imaging
Published on: November 17, 2023
Intelligent surgical workflow recognition-based skill assessment for minimally invasive submandibular gland resection
Zhongkai Ma1, Yufei Hua1, Lin Que1
1State Key Laboratory of Oral Diseases & National Center for Stomatology & National Clinical Research Center for Oral Diseases & Department of Head and Neck Oncology West China Hospital of Stomatology, Sichuan University, Chengdu, China.
NPJ Digital Medicine
|July 7, 2026
Summary
Artificial Intelligence (AI) enhances minimally invasive submandibular gland (SMG) resection by recognizing surgical workflows. This intelligent model improves surgical phase recognition and reduces annotation time.
Area of Science:
- Oral Medicine
- Surgical Workflow Analysis
- Artificial Intelligence in Surgery
Background:
- AI applications in dentistry primarily focus on static images, neglecting dynamic surgical video data.
- Analyzing dynamic video data in oral medicine surgery is underexplored.
- Minimally invasive submandibular gland (SMG) resection generates rich, dynamic surgical information.
Purpose of the Study:
- To introduce AI-miSMG, an intelligent model for recognizing surgical workflows in minimally invasive SMG resection.
- To establish a standardized annotation protocol for endoscopic SMG resection procedures.
- To evaluate the model's performance in surgical phase recognition and workflow analysis.
Main Methods:
- Developed AI-miSMG, a deep learning model trained on 73 endoscopic SMG resection videos (386,122 frames).
- Utilized a standardized annotation protocol dividing procedures into Creation, Position, Separation, Inspection, and Idle phases.
- Validated the model on a multicenter dataset (85,913 images) from four centers.
Main Results:
- AI-miSMG achieved an overall accuracy of 0.87 on the external validation dataset.
- The model demonstrated feasibility for surgical phase recognition and workflow analysis.
- Annotation time was reduced by approximately 47% (from 94.00 min to 49.90 min).
Conclusions:
- A deep learning-based workflow recognition model (AI-miSMG) was developed for minimally invasive SMG resection.
- The model shows feasibility for surgical phase recognition, workflow analysis, and model-assisted annotation.
- AI-miSMG offers potential for enhancing surgical training and efficiency in oral medicine.