Related Experiment Video
Updated: Mar 21, 2026
08:17
Dynamic Lung Tumor Tracking for Stereotactic Ablative Body Radiation Therapy
Published on: June 7, 2015
16.3K
Automatic prompt-guided incremental fine-tuning for offset detection in radiotherapy patient positioning.
Jing Zhang1, Yang Liu1, Yuchi Jiang1
1School of Computer Science and Artificial Intelligence, Liaoning Normal University, Dalian, People's Republic of China.
Physics in Medicine and Biology
|March 19, 2026
Summary
This study introduces an automated, low-cost method for radiotherapy patient positioning using prompt-guided AI. The system enhances accuracy and efficiency, reducing clinician workload and patient discomfort during treatment.
Area of Science:
- Medical Physics
- Artificial Intelligence in Healthcare
- Radiotherapy Technology
Background:
- Accurate patient positioning is critical in radiotherapy (RT) for precise radiation delivery and minimizing off-target exposure.
- Conventional RT positioning relies on manual, often CT-based methods, which are inefficient, uncomfortable for patients, and prone to inconsistencies.
- Existing methods lack automation and adaptability, leading to potential alignment errors and increased treatment times.
Purpose of the Study:
- To develop an automatic, robust, and low-cost method for detecting posture offsets in radiotherapy patients.
- To overcome the limitations of conventional radiotherapy positioning workflows, including inefficiency and potential alignment inconsistencies.
- To improve the precision and consistency of patient positioning during radiation therapy treatments.
Main Methods:
- A prompt-guided incremental fine-tuning model using a large-scale image segmentation backbone was developed.
- The system processes real-time 2D images from a single RGB camera, generating adaptive prompts for robust segmentation.
- A multi-level offset analysis framework (contour, keypoint, pixel levels) quantifies posture deviations, with clinical deployment for data collection and validation.
Main Results:
- The proposed method demonstrated accurate, fast, and stable posture offset detection on real clinical radiotherapy data.
- Positioning consistency and efficiency were substantially improved compared to conventional workflows.
- Ablation studies confirmed the effectiveness and necessity of individual components within the developed framework.
Conclusions:
- The study presents a practical, low-cost solution for radiotherapy patient positioning, reducing workload and improving treatment accuracy.
- Prompt-guided incremental adaptation and multi-level offset analysis show significant potential in real-world radiotherapy environments.
- The findings pave the way for more intelligent and automated radiotherapy positioning systems.
