Related Experiment Video
Updated: Sep 3, 2025
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
HPM-Net: Hierarchical progressive multiscale network for liver vessel segmentation in CT images
Wen Hao1, Jing Zhang2, Jun Su1
1School of Computer Science and Communication Engineering, Jiangsu University, Zhenjiang 212013, China.
Insights
This study introduces a novel hierarchical progressive multiscale learning network (HPM-Net) for accurate liver vessel segmentation in 3D CT images, improving upon existing methods.
Area of Science:
- Medical Imaging
- Radiology
- Computer-Aided Diagnosis
Background:
- Accurate segmentation of liver vessels in 3D CT images is crucial for diagnosing and planning treatments for liver diseases.
- Challenges include irregular vessel structures, image noise, and information loss during down-sampling, especially for small vessels.
Purpose of the Study:
- To develop an advanced deep learning framework for precise liver vessel segmentation.
- To address the limitations of existing methods in capturing fine vessel details and preventing information loss.
Main Methods:
- Proposed a hierarchical progressive multiscale learning network (HPM-Net) incorporating internal and external progressive learning.
- Introduced a dual-branch progressive 3D Unet with a dual-branch progressive (DBP) down-sampling strategy to preserve detailed information.
- Implemented a deep supervision mechanism to accelerate convergence and enhance network training.
Main Results:
- The HPM-Net achieved an average Dice coefficient of 75.18% and a sensitivity of 78.84% on the 3Dircadb dataset.
- These results indicate superior performance compared to the original network in liver vessel segmentation.
Conclusions:
- The proposed HPM-Net demonstrates significant accuracy in segmenting liver vessels from CT images.
- The method effectively overcomes challenges associated with vessel irregularity and information loss during segmentation.
Background And Objective:
The segmentation and visualization of liver vessels in 3D CT images are essential for computer-aided diagnosis and preoperative planning of liver diseases. Due to the irregular structure of liver vessels and image noise, accurate extraction of liver vessels is difficult. In particular, accurate segmentation of small vessels is always a challenge, as multiple single down-sampling usually results in a loss of information.
Methods:
In this paper, we propose a hierarchical progressive multiscale learning network (HPM-Net) framework for liver vessel segmentation. Firstly, the hierarchical progressive multiscale learning network combines internal and external progressive learning methods to learn semantic information about liver vessels at different scales by acquiring receptive fields of different sizes. Secondly, to better capture vessel features, we propose a dual-branch progressive 3D Unet, which uses a dual-branch progressive (DBP) down-sampling strategy to reduce the loss of detailed information in the process of network down-sampling. Finally, a deep supervision mechanism is introduced into the framework and backbone network to speed up the network convergence and achieve better training of the network.
Results:
We conducted experiments on the public dataset 3Dircadb for liver vessel segmentation. The average dice coefficient and sensitivity of the proposed method reached 75.18% and 78.84%, respectively, both higher than the original network.
Conclusion:
Experimental results show that the proposed hierarchical progressive multiscale network can accurately segment the labeled liver vessels from the CT images.
