Related Experiment Video
Updated: Nov 28, 2025
07:53
Author Spotlight: Advancing 3D Modeling for Enhanced Diagnosis and Treatment of Pulmonary Nodules in Early-Stage Lung Cancer
Published on: October 13, 2023
1.8K
Enhancing digital tomosynthesis (DTS) for lung radiotherapy guidance using patient-specific deep learning model.
Zhuoran Jiang1,2, Fang-Fang Yin2,3,4, Yun Ge1
1School of Electronic Science and Engineering, Nanjing University, 163 Xianlin Road, Nanjing, Jiangsu, 210046, People's Republic of China.
Physics in Medicine and Biology
|November 25, 2020
Summary
A new patient-specific deep learning model enhances digital tomosynthesis (DTS) for image-guided radiation therapy (IGRT). This method improves image quality and accuracy, making DTS a valuable tool for IGRT applications.
Area of Science:
- Medical Imaging
- Radiation Oncology
- Artificial Intelligence
Background:
- Digital tomosynthesis (DTS) is a low-dose imaging technique for image-guided radiation therapy (IGRT).
- Conventional DTS reconstruction methods (e.g., FDK) produce distortions and poor resolution due to limited scanning angles.
- Existing deep learning methods fail to account for inter-patient variability, leading to blurred edges in restored images.
Purpose of the Study:
- To develop a patient-specific deep learning model for enhancing DTS images.
- To recover volumetric information in DTS by learning patient-specific correlations with ground truth CT images.
- To improve the accuracy and quality of DTS for IGRT applications.
Main Methods:
- A patient-specific deep learning model was trained using DTS and ground truth volumetric images.
- The model leveraged patient-specific prior knowledge to learn image correlations.
- Validation was performed using simulated and real on-board projections from lung cancer patient data.
Main Results:
- Enhanced DTS exhibited CT-like image quality with clear edges.
- Quantitative analysis showed low-intensity errors and high structural similarity to ground truth CT.
- Tumor localization errors were ≤0.7 mm (simulated) and ≤1.6 mm (real projections).
- DTS enhancement was achieved in near real-time.
Conclusions:
- The proposed patient-specific deep learning method effectively enhances DTS image quality.
- The enhanced DTS provides accurate volumetric information for IGRT.
- This method offers a valuable and efficient tool for improving IGRT accuracy and patient outcomes.
