Related Experiment Video
Updated: Jun 10, 2025
05:18
Author Spotlight: Improving Radiation Therapy Access with Radiation Planning Assistant
Published on: October 6, 2023
1.3K
Clinical target volume (CTV) automatic delineation using deep learning network for cervical cancer radiotherapy: A
Zhe Wu1,2, Dong Wang3, Cheng Xu4
1Department of Digital Medicine, School of Biomedical Engineering and Medical Imaging, Army Medical University (Third Military Medical University), Chongqing, China.
Journal of Applied Clinical Medical Physics
|October 14, 2024
Summary
A novel deep learning algorithm, ResCANet, accurately delineates clinical target volumes for cervical cancer radiotherapy, significantly reducing manual delineation time. This AI tool demonstrates strong generalizability, performing well even in endometrial cancer cases.
Area of Science:
- Radiotherapy and Medical Imaging
- Artificial Intelligence in Oncology
- Cancer Treatment Planning
Background:
- Accurate clinical target volume (CTV) delineation is crucial for effective radiotherapy in cervical cancer.
- Manual delineation is time-consuming and subject to inter-observer variability.
- Deep learning (DL) offers potential for automating and standardizing this process.
Purpose of the Study:
- To evaluate the accuracy and feasibility of a novel deep learning (DL) algorithm, ResCANet, for CTV delineation in cervical cancer radiotherapy.
- To assess the generalization capability of ResCANet in external cervical and endometrial cancer datasets.
Main Methods:
- ResCANet, a DL model based on ResNet-UNet with cascade multi-scale convolution and atrous spatial pyramid pooling, was developed.
- The model was trained and validated on 236 cervical cancer cases (5-fold cross-validation).
- External validation was performed on 54 cervical and 42 endometrial cancer cases, comparing DL delineation against manual contours using Dice Similarity Coefficient (DSC), Sensitivity (SEN), Positive Predictive Value (PPV), and 95% Hausdorff Distance (95HD).
Main Results:
- Internal validation showed mean DSC of 74.8%, SEN 81.5%, PPV 73.5%, and 95HD of 10.5 mm.
- External validation yielded comparable results: cervical cancer (DSC 73.4%, SEN 72.9%, PPV 75.3%, 95HD 12.5 mm) and endometrial cancer (DSC 77.1%, SEN 81.1%, PPV 75.5%, 95HD 10.3 mm).
- Approximately 85% of cases using DL-aided delineation required minor or no revisions, with delineation time reduced to under 30 minutes.
Conclusions:
- The proposed ResCANet algorithm demonstrates high accuracy and feasibility for automatic CTV delineation in cervical cancer radiotherapy.
- The model exhibits excellent generalizability, performing effectively in endometrial cancer cases as well.
- DL-based automatic delineation, as implemented by ResCANet, can improve efficiency and consistency in radiotherapy planning without compromising quality.
