IG-Net: An Instrument-guided real-time semantic segmentation framework for prostate dissection during surgery for low rectal cancer

Bo Sun ¹, Zhen Sun ², Kexuan Li ²

¹Research Unit of Virtual Body and Virtual Surgery Technologies, Division of Colorectal Surgery, Department of General Surgery, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, No. 1 Shuai Fu Yuan, DongCheng District, Beijing, 100730, China; State Key Laboratory of Virtual Reality Technology and Systems, Beihang University, No. 37 Xueyuan Road, Haidian District, Beijing, 100191, China.
²Division of Colorectal Surgery, Department of General Surgery, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, No. 1 Shuai Fu Yuan, Dongcheng District, Beijing, 100730, China.
³State Key Laboratory of Virtual Reality Technology and Systems, Beihang University, No. 37 Xueyuan Road, Haidian District, Beijing, 100191, China.
⁴Computer school, Beijing Information Science and Technology University, No. 12 Xiaoying East Road, Haidian District, Beijing, 100192, China.
⁵State Key Laboratory of Virtual Reality Technology and Systems, Beihang University, No. 37 Xueyuan Road, Haidian District, Beijing, 100191, China; Zhongguancun Laboratory, No. 51 Kunming Lake South Road, Haidian District, Beijing, 100195, China.
⁶State Key Laboratory of Virtual Reality Technology and Systems, Beihang University, No. 37 Xueyuan Road, Haidian District, Beijing, 100191, China; Peng Cheng Laboratory, No. 2 Xingke 1st Street, Nanshan District, Shenzhen, 518055, China.

⁰Research Unit of Virtual Body and Virtual Surgery Technologies, Division of Colorectal Surgery, Department of General Surgery, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, No. 1 Shuai Fu Yuan, DongCheng District, Beijing, 100730, China; State Key Laboratory of Virtual Reality Technology and Systems, Beihang University, No. 37 Xueyuan Road, Haidian District, Beijing, 100191, China.

Computer Methods and Programs in Biomedicine +

|

October 5, 2024

View abstract on PubMed

Summary

A new IG-Net framework precisely segments the prostate during transanal surgery using instrument features for improved accuracy and speed. This AI tool enhances surgical precision and patient outcomes in low rectal cancer treatment.

Area Of Science

Medical Imaging and Computer-Assisted Surgery
Artificial Intelligence in Healthcare

Background

Accurate prostate dissection is vital in transanal surgery for low rectal cancer to prevent complications like urethral injury.
Challenges include unclear prostate boundaries, irregular anatomy, and surgical smoke, hindering precise dissection.

Purpose Of The Study

To introduce a novel video semantic segmentation framework, IG-Net, for real-time and precise prostate segmentation.
To improve surgical accuracy and patient outcomes in transanal low rectal cancer surgery.

Main Methods

Developed IG-Net, a framework incorporating surgical instrument features for prostate segmentation.
Implemented an instrument-guided module for attention-based local and global segmentation.
Introduced a keyframe selection module using instrument features to optimize segmentation speed and reduce noise.

Main Results

Evaluated on an extensive dataset of 106 video clips (6153 images).
Achieved 72.70% IoU, 82.02% Dice, and 35 FPS.
Demonstrated superior performance compared to previous methods.

Conclusions

IG-Net surpasses existing methods for prostate segmentation in surgical videos.
The framework effectively balances segmentation accuracy and speed.
IG-Net shows robustness against surgical challenges like smoke and unclear boundaries.

Keywords:

Computer vision Deep learning Rectal cancer Surgical video Video semantic segmentation