Related Experiment Video
Updated: Oct 26, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.1K
Automatic segmentation of rectal tumor on diffusion-weighted images by deep learning with U-Net
Hai-Tao Zhu1, Xiao-Yan Zhang1, Yan-Jie Shi1
1Department of Radiology, Key Laboratory of Carcinogenesis and Translational Research (Ministry of Education/Beijing), Peking University Cancer Hospital & Institute, Beijing, China.
Journal of Applied Clinical Medical Physics
|August 3, 2021
Summary
A novel deep learning approach using a volumetric U-Net accurately segments rectal tumors on diffusion-weighted images (DWI). This automated method outperforms semi-automatic techniques, improving efficiency and precision in rectal cancer diagnosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Manual segmentation of rectal tumors on volumetric images is time-consuming and subjective.
- Automatic segmentation on diffusion-weighted imaging (DWI) is challenging due to image quality issues.
Purpose of the Study:
- To propose a volumetric U-shaped neural network (U-Net) for automatic rectal tumor segmentation on DWI.
- To evaluate the performance of the proposed U-Net against a semi-automatic method.
Main Methods:
- A volumetric U-Net was developed using diffusion-weighted images from 300 locally advanced rectal cancer patients.
- The U-Net processed volumetric DWI data for tumor segmentation.
- Performance was evaluated using the Dice Similarity Coefficient (DSC) and compared to a semi-automatic segmentation method.
Main Results:
- The deep learning U-Net achieved a higher mean DSC (0.675 ± 0.144) compared to the semi-automatic method (0.614 ± 0.225).
- Statistical analysis revealed a significant difference (p=0.035) in DSC between the two methods.
- The U-Net demonstrated superior segmentation accuracy for rectal tumors on DWI.
Conclusions:
- Volumetric U-Net effectively automates rectal tumor segmentation on DWI.
- This deep learning approach offers a promising tool for improved rectal cancer assessment.
