Related Experiment Video
Updated: Jan 10, 2026
08:17
Dynamic Lung Tumor Tracking for Stereotactic Ablative Body Radiation Therapy
Published on: June 7, 2015
16.1K
Marker-Less Lung Tumor Tracking from Real-Time Color X-Ray Fluoroscopic Images Using Cross-Patient Deep Learning
Yongxuan Yan1, Fumitake Fujii1, Takehiro Shiinoki2
1Graduate School of Sciences and Technology for Innovation, Yamaguchi University, Ube 755-8611, Japan.
Bioengineering (Basel, Switzerland)
|November 27, 2025
Summary
This study introduces a marker-less tumor tracking framework for radiotherapy, eliminating invasive procedures. The novel deep learning approach accurately tracks tumors using simulated data, showing promise for improved patient care.
Area of Science:
- Medical Physics
- Radiotherapy
- Artificial Intelligence
Background:
- Fiducial marker implantation for radiotherapy tumor localization is invasive and poses risks.
- Current methods often require per-patient retraining, increasing complexity and cost.
Purpose of the Study:
- To develop and assess a marker-less tumor tracking framework using a cross-patient deep learning model.
- To eliminate the need for per-patient retraining in radiotherapy tumor localization.
Main Methods:
- A novel degradation model generated simulated data from digitally reconstructed radiographs (DRRs).
- A Restormer network transformed clinical fluoroscopic images into DRR-like images.
- A DUCK-Net model performed tumor segmentation on DRR-trained data.
Main Results:
- The framework achieved a median 3D Euclidean tumor center tracking error of 1.53 mm.
- Directional errors were low: 0.98±0.70 mm (LR), 1.09±0.74 mm (SI), and 1.34±0.94 mm (AP).
- Average processing time was 179.8 ms per image.
Conclusions:
- The marker-less tumor tracking framework is a feasible proof-of-concept.
- This approach demonstrates potential for a cross-patient, non-invasive solution in radiotherapy.
- Further large-scale validation is needed for broad clinical applicability.
