Comparison of Deep Learning-Based Auto-Segmentation Results on Daily Kilovoltage, Megavoltage, and Cone Beam CT
Zhixing Wang1, Chengyu Shi1, Carson Wong1
1Department of Radiation Oncology, City of Hope, Duarte, CA, USA.
Technology in Cancer Research & Treatment
|May 21, 2025
Summary
Deep learning auto-segmentation for radiotherapy shows best results with kilovoltage CT (kVCT) imaging compared to kV cone-beam CT (kV-CBCT) or megavoltage CT (MVCT). Manual contouring adjustments are still essential for all modalities.
Area of Science:
- Radiotherapy and Medical Imaging
- Artificial Intelligence in Healthcare
- Computational Anatomy
Background:
- Accurate auto-segmentation of organs-at-risk is crucial for image-guided radiotherapy (IGRT).
- Deep learning models offer potential for automating contouring, but performance varies across imaging modalities.
- Evaluating auto-segmentation accuracy on different CT imaging techniques is essential for clinical implementation.
Purpose of the Study:
- To assess the performance of deep learning auto-segmentation models across various CT imaging modalities.
- To compare the accuracy of auto-segmented contours against manual delineations on kilovoltage CT (kVCT), kV cone-beam CT (kV-CBCT), and megavoltage CT (MVCT).
- To identify the optimal imaging modality for deep learning-based auto-segmentation in IGRT.
Main Methods:
- Phantom studies were conducted to benchmark image quality.
- Retrospective analysis of daily CT images from 60 patients across kVCT, kV-CBCT, and MVCT modalities.
- Deep learning models (convolutional neural networks) were employed for auto-segmentation of organs-at-risk.
- Quantitative metrics including Dice Similarity Coefficient (DSC) and Hausdorff distance were used for comparison with manual contours.
Main Results:
- Auto-segmentation on kVCT demonstrated statistically significant superior agreement with manual contours compared to kV-CBCT and MVCT for most major organs.
- In pelvic cases, kVCT achieved a mean DSC of 0.84±0.05 for bowel, significantly higher than kV-CBCT (0.35±0.23) and MVCT (0.48±0.27).
- In thoracic cases, kVCT yielded a mean DSC of 0.63±0.16 for the esophagus, outperforming kV-CBCT (0.18±0.13) and MVCT (0.22±0.08).
Conclusions:
- Deep learning auto-segmentation models perform best with kVCT images, showing greater concordance with manual delineations than with kV-CBCT or MVCT.
- Despite advancements, manual contour correction remains necessary across all evaluated imaging modalities, especially for organs with low contrast.
- These findings highlight the current capabilities and limitations of deep learning auto-segmentation for adaptive radiotherapy applications.
Related Concept Videos
01:10Computed Tomography
Tomography refers to imaging by sections. Computed tomography (CT) is a non-invasive imaging technique that uses computers to analyze several cross-sectional X-rays to reveal minute details about structures in the body.
The technique was invented in the 1970s and is based on the principle that as X-rays pass through the body, they are absorbed or reflected at different levels. In the technique, a patient lies on a motorized platform while a computerized axial tomography (CAT) scanner rotates...
The technique was invented in the 1970s and is based on the principle that as X-rays pass through the body, they are absorbed or reflected at different levels. In the technique, a patient lies on a motorized platform while a computerized axial tomography (CAT) scanner rotates...
01:28Imaging Studies for Cardiovascular System V: CT
Cardiac computed tomography (CT) scanning is an advanced cardiac imaging technique that utilizes CT technology, with or without intravenous (IV) contrast, to produce accurate cross-sectional virtual slices of specific areas of the heart, coronary circulation, and major blood vessels such as the aorta, pulmonary veins, and arteries. The computer processes these slices to generate three-dimensional images. Multidetector CT (MDCT) is a rapid form of CT scanning that captures multiple slices...