Related Experiment Video
Updated: Aug 25, 2025
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
FBCU-Net: A fine-grained context modeling network using boundary semantic features for medical image segmentation
Mei Yu1, Kaijie Pei2, Xuewei Li1
1College of Intelligence and Computing, Tianjin University, Tianjin, 300350, China; Tianjin Key Laboratory of Cognitive Computing and Application, Tianjin, 300350, China; Tianjin Key Laboratory of Advanced Networking, Tianjin, 300350, China.
This study introduces FBCU-Net, a novel deep learning model for medical image segmentation that improves accuracy by focusing on boundary semantic features. The network effectively distinguishes boundary pixels, enhancing overall segmentation performance for various medical imaging applications.
Area of Science:
- Medical Image Analysis
- Deep Learning
- Computer Vision
Background:
- Accurate medical image segmentation is crucial for diagnosis and treatment planning.
- Deep learning methods struggle with segmenting tissue boundaries due to ambiguous pixel features.
- Boundary pixels often contain mixed features, complicating precise segmentation.
Purpose of the Study:
- To develop a novel deep learning network, FBCU-Net, for improved medical image segmentation.
- To enhance the accuracy of tissue boundary segmentation by leveraging boundary semantic features.
- To reduce the impact of irrelevant features on boundary pixels for more discriminative representations.
Main Methods:
- Proposed FBCU-Net incorporating fine-grained contextual modeling based on boundary semantic features.
- Introduced new supervision information to identify and classify boundary pixels.
- Developed a boundary enhancement strategy to improve network focus on critical boundary regions.
- Generated boundary region representations to capture semantic features of boundaries.
Main Results:
- FBCU-Net demonstrated superior boundary segmentation performance compared to state-of-the-art methods.
- The model achieved better overall segmentation accuracy across diverse medical image datasets (TUI, ISIC-2018, 2018 Data Science Bowl, Glas, BUSI).
- Evaluations confirmed FBCU-Net's effectiveness in reducing the influence of irrelevant features on boundary pixels.
Conclusions:
- FBCU-Net offers a promising approach for accurate medical image segmentation, particularly at tissue boundaries.
- The proposed boundary-focused strategy significantly enhances segmentation quality.
- The model shows substantial potential for clinical applications requiring precise medical image analysis.
