Related Experiment Video
Updated: Jan 2, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
Liver tumor segmentation in CT volumes using an adversarial densely connected network
Lei Chen1, Hong Song2, Chi Wang1
1School of Computer Science & Technology, Beijing Institute of Technology, 5 South Zhongguancun Street, Beijing, 100081, China.
BMC Bioinformatics
|December 3, 2019
Summary
This study introduces an advanced algorithm for segmenting liver tumors in CT scans, significantly improving accuracy and robustness. The method enhances diagnostic capabilities for physicians, aiding in personalized cancer treatment strategies.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Oncology
Background:
- Malignant liver tumors are a leading cause of mortality.
- Accurate segmentation of liver tumors from CT scans is crucial for diagnosis and personalized treatment.
- Challenges include image noise, similar tissue intensities, and tumor variability, hindering automated segmentation.
Purpose of the Study:
- To develop an automatic and reliable method for liver tumor segmentation from abdominal CT images.
- To improve diagnostic accuracy and treatment planning for liver cancer patients.
Main Methods:
- A novel deep learning network designed for liver tumor segmentation.
- Implementation of an adversarial training strategy to enhance segmentation performance.
Main Results:
- The proposed network achieved an average Dice score of 68.4% for tumor segmentation.
- Adversarial training improved segmentation metrics: ASD decreased from 27.8 to 21, MSD from 147 to 124, VOE from 0.52 to 0.46, and RVD from 0.69 to 0.73.
Conclusions:
- The developed method demonstrates improved performance in liver tumor segmentation.
- Adversarial training significantly enhances the accuracy and robustness of the segmentation results.
