Related Experiment Video
Updated: Jul 12, 2026
06:39
Three-dimensional Location Approach with Silk Thread Guided Laparoscopic Segmentectomy for Liver Tumor
Published on: May 23, 2025
Real-time anatomy recognition in laparoscopic liver resection using video segmentation AI model.
Haisu Tao1,2,3, Kangwei Guo1,2,3, Yijun Yang4
1Department of Hepatobiliary Surgery I, General Surgery Center, Zhujiang Hospital, Southern Medical University, Guangzhou, China.
NPJ Digital Medicine
|July 10, 2026
Summary
Vivim, an AI model using Mamba architecture, enhances surgical navigation for laparoscopic liver resection (LLR) by accurately segmenting critical intrahepatic structures like the Glissonean pedicle and hepatic vein.
Area of Science:
- Medical image analysis
- Computer vision in surgery
- Artificial intelligence in healthcare
Background:
- Laparoscopic liver resection (LLR) faces challenges due to complex intrahepatic vascular anatomy and limited visualization, increasing intraoperative injury risk.
- Accurate identification of anatomical structures is crucial for safe and effective LLR procedures.
Purpose of the Study:
- To develop and evaluate Vivim, a novel video segmentation model based on the Mamba architecture, for improved anatomical structure identification during LLR.
- To assess Vivim's performance against existing state-of-the-art models in segmenting the Glissonean pedicle (GP) and hepatic vein (HV).
Main Methods:
- Vivim, a Mamba-based video segmentation model, was developed integrating long-range dependency modeling with recurrent network efficiency.
- The model was trained and evaluated on a multicenter dataset of 15,865 annotated frames from 45 surgical videos.
- Performance was benchmarked against several image- and video-based models using Dice scores and inference speed.
Main Results:
- Vivim outperformed baseline models, achieving Dice scores of 0.71 (single-target) and 0.66 (multi-target) for GP and HV segmentation.
- The model demonstrated real-time inference at 25 frames per second and strong generalization on external datasets.
- Clinical validation by 13 surgeons confirmed Vivim's ability to improve recognition speed and accuracy in surgical navigation.
Conclusions:
- Vivim, a Mamba-based AI framework, shows significant promise for AI-assisted surgical navigation in LLR, improving intraoperative decision-making.
- The model offers a potential solution to bridge the gap between laboratory precision and clinical reliability in complex surgical procedures.
- Further development is needed to address challenges in differentiating visually similar vascular structures.
