Related Experiment Video
Updated: Sep 20, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Deep Learning Based Real-Time Semantic Segmentation of Cerebral Vessels and Cranial Nerves in Microvascular
Ruifeng Bai1,2, Xinrui Liu2,3, Shan Jiang1
1Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Sciences, Changchun 130033, China.
Cells
|June 10, 2022
Summary
This study introduces MVDNet, a fast AI model for segmenting cerebral vessels and cranial nerves in microvascular decompression (MVD) images. It offers a balance of accuracy and speed for treating conditions like trigeminal neuralgia.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neurosurgery
Background:
- Accurate segmentation of cerebral vessels and cranial nerves is crucial for treating trigeminal neuralgia (TGN) and hemifacial spasm (HFS).
- The visual similarity between these structures in microvascular decompression (MVD) images presents a significant challenge for real-time automated segmentation.
Purpose of the Study:
- To develop a lightweight and fast semantic segmentation network, MVDNet, specifically designed for MVD scenarios.
- To achieve a favorable balance between segmentation accuracy and processing speed for clinical applications.
Main Methods:
- Proposed a novel Microvascular Decompression Network (MVDNet) featuring a Light Asymmetric Bottleneck (LAB) module for context encoding and a Feature Fusion Module (FFM) for combining feature levels.
- The network was designed without reliance on pretrained models, emphasizing efficiency and reduced parameter count.
Main Results:
- MVDNet achieved a mean Intersection over Union (mIoU) of 76.59% on the MVD test dataset.
- The network boasts a small parameter count (0.72 million) and a fast inference speed of 137 FPS on a single GTX 2080Ti GPU.
Conclusions:
- MVDNet demonstrates high efficiency and speed for semantic segmentation in MVD images, outperforming previous methods.
- The developed network offers a practical solution for real-time image analysis in neurosurgical procedures, potentially improving treatment outcomes for TGN and HFS.
