Related Experiment Video
Updated: Dec 14, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.2K
Automatic CT image segmentation of maxillary sinus based on VGG network and improved V-Net
Jiangchang Xu1, Shiming Wang1, Zijie Zhou2
1Institute of Biomedical Manufacturing and Life Quality Engineering, State Key Laboratory of Mechanical System and Vibration, School of Mechanical Engineering, Shanghai Jiao Tong University, Dongchuan Road 800, Minhang District, Shanghai, 200240, China.
Summary
This study introduces an automatic method for segmenting the maxillary sinus (MS) using CT images. The approach enhances diagnostic efficiency and accuracy by improving segmentation performance.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Artificial Intelligence in Medicine
Background:
- Accurate maxillary sinus (MS) segmentation is crucial for clinical diagnoses.
- Manual segmentation is time-consuming and prone to inaccuracies.
- Existing automatic methods require parameter tuning and initial seed points, limiting efficiency.
Purpose of the Study:
- To develop an accurate, efficient, and automatic method for maxillary sinus (MS) segmentation from CT images.
- To overcome the limitations of manual and current automatic segmentation techniques.
- To improve the clinical applicability of MS analysis.
Main Methods:
- Proposed an automatic CT image segmentation method combining VGG network for slice classification and an improved V-Net with edge supervision.
- VGG network classifies CT slices to prevent segmentation failures in slices without MS.
- Improved V-Net integrates edge loss to enhance MS region segmentation and reduce misjudgments.
Main Results:
- VGG network achieved 97.04% accuracy in classifying CT slices with/without MS.
- The proposed method yielded high segmentation performance with Dice score of 94.40%, IoU of 90.05%, and precision of 94.72%.
- Significantly reduced region misjudgment compared to U-Net and V-Net, with 3D reconstruction errors within ±1 mm.
Conclusions:
- The developed method enables efficient, accurate, and automatic segmentation of the maxillary sinus (MS).
- The approach improves segmentation accuracy and reduces errors compared to existing methods.
- This technique has the potential to significantly enhance clinical workflow efficiency and diagnostic capabilities.
