Related Experiment Video
Updated: Jul 26, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
A spatio-temporal network for video semantic segmentation in surgical videos
Maria Grammatikopoulou1, Ricardo Sanchez-Matilla2, Felix Bragman2
1Medtronic plc, London, UK. maria.grammatikopoulou@medtronic.com.
Summary
This study introduces a novel temporal segmentation model for surgical videos, enhancing anatomical identification accuracy and patient safety. The model improves temporal consistency, crucial for reliable intra-operative guidance and surgical education.
Area of Science:
- Computer Vision
- Medical Imaging
- Surgical Technology
Background:
- Semantic segmentation in surgical videos is vital for intra-operative guidance, post-operative analytics, and surgical education.
- Temporal inconsistency in anatomical identification can compromise patient safety.
- Accurate and temporally consistent segmentation models are needed for clinical applications.
Purpose of the Study:
- To propose a novel architecture for modeling temporal relationships in surgical videos.
- To improve the accuracy and temporal consistency of semantic segmentation in surgical videos.
- To address the limitations of static models in capturing temporal dynamics.
Main Methods:
- Developed a temporal segmentation model with a static encoder and a spatio-temporal decoder.
- The encoder processes individual frames, while the decoder learns from frame sequences.
- The proposed decoder is designed to be compatible with existing encoders.
Main Results:
- Evaluated on CholecSeg8k and a private Partial Nephrectomy dataset.
- Mean Intersection over Union (IoU) improved by 1.30% and 4.27% with the temporal decoder.
- Demonstrated improvements in temporal consistency of up to 7.23%.
Conclusions:
- The proposed model advances semantic segmentation in surgical scenes.
- The temporal decoder enhances state-of-the-art static models.
- Improved segmentation accuracy and temporal consistency can lead to better patient outcomes.
