Related Experiment Video
Updated: Oct 12, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Adaptive Localizing Region-Based Level Set for Segmentation of Maxillary Sinus Based on Convolutional Neural
Xianglong Qi1, Jie Zhong2, Shengjia Cui3
1Liaoning Huading Technology Co., Ltd., Shenyang, Liaoning 110167, China.
Computational Intelligence and Neuroscience
|November 22, 2021
Summary
This study introduces a novel convolutional neural network (CNN) method for accurate maxillary sinus segmentation, improving lesion detection in CT images. The new approach enhances segmentation accuracy, outperforming existing methods.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Maxillary sinus segmentation is crucial for diagnosis, but lesions complicate traditional methods.
- Heterogeneous lesions within the maxillary sinus reduce segmentation accuracy in current algorithms.
Purpose of the Study:
- To develop an adaptive localizing region-based level set method using a convolutional neural network (CNN) for improved maxillary sinus segmentation.
- To enhance segmentation accuracy, particularly in cases with heterogeneous lesions.
Main Methods:
- A novel method combining a convolutional neural network (CNN) with an adaptive localizing region-based level set.
- CNN feature extraction to identify lesion characteristics and guide the level set evolution.
- Interactive contour refinement with speed compensation to prevent trapping in non-target areas.
Main Results:
- The proposed method demonstrated superior performance on a dataset of 200 CT images with lesions.
- Achieved a significant average Dice similarity coefficient improvement of 0.25 over FLS and 0.12 over CRF-FCN.
- Effectively avoided active contour trapping in non-target areas, improving segmentation robustness.
Conclusions:
- The adaptive CNN-based level set method significantly improves maxillary sinus segmentation accuracy, especially in the presence of lesions.
- This novel approach offers a robust solution for challenging medical image segmentation tasks.
- The method shows promise for clinical applications requiring precise maxillary sinus delineation.
