Abstract
Purpose: This study evaluates the accuracy of ultrasound (US) tracking for intra-fraction pancreatic tumor motion during radiotherapy using a phantom-based model. Methods: Respiratory cycle data from 12 pancreatic cancer patients (treated with SBRT) were acquired via 4DCT and transferred to a motion platform. An abdominal phantom was fixed to the platform, and its ground truth motion was recorded using an infrared (IR) optical tracker. An ultrasound system tracked a target tumor within the phantom, and results were compared against the IR data using root-mean-square error (RMSE). Results: 82.2% of US tracking measurements were within 0.5 mm of the IR ground truth, with only 0.7% failing (error > 2.5 mm). Tracking accuracy showed no correlation with respiratory displacement, velocity, or acceleration. Conclusions: The high accuracy demonstrated suggests that US tracking is a viable method for real-time target monitoring, potentially improving radiation dose delivery precision.
Introduction
Pancreatic cancer treatment often involves high-dose Stereotactic Body Radiation Therapy (SBRT). However, the pancreas is subject to significant intra-fraction motion due to respiration, moving up to 24 mm. To ensure target coverage, clinicians typically use large Planning Target Volumes (PTVs) of 15–30 mm. While effective, these large margins increase radiation exposure to adjacent critical organs like the stomach and duodenum, leading to toxicity.
Non-invasive, real-time tracking is essential to reduce these margins. While MRI-guided systems are costly and X-ray-based systems involve additional ionizing radiation, ultrasound offers a cost-effective, non-ionizing alternative with high soft-tissue contrast. This study validates a US-based tracking system using a realistic phantom setup.
Methods and Materials
The researchers developed a sophisticated phantom-based experimental setup to simulate clinical conditions accurately.
Motion Simulation and Phantom
Instead of idealized waveforms, the study used actual respiratory displacement data from 4DCT scans of 12 pancreatic cancer patients. This data drove a motion platform carrying an ABDFAN ultrasound abdominal phantom (Kyoto Kagaku Co, Japan), which contains simulated organ structures and tumors.
Tracking and Validation
A 4D convex ultrasound probe (3-7 MHz) was held against the phantom using a customized "arm-bridge" mechanical system. The tracking algorithm employed intensity-based image-to-image registration, comparing real-time US frames to a manually contoured reference image. Ground truth was established using an external infrared (IR) camera system tracking an optical marker on the phantom.
Key Results
The system demonstrated exceptional precision and reliability across the simulated patient profiles:
- Sub-millimeter Accuracy: 82.2% of all tracking points were within 0.5 mm of the ground truth.
- Clinical Reliability: Only 0.7% of points exceeded the 2.5 mm clinical failure threshold.
- Strong Correlation: Linear regression analysis showed a near-perfect correlation ($r \approx 1.0$) between US-tracked and IR-monitored motion for all patients.
- Dynamic Stability: Tracking errors did not correlate with respiratory velocity or acceleration, indicating the algorithm remains stable even during irregular breathing patterns.
One outlier, "Patient 4," showed a slightly higher RMSE (1.121 mm), likely due to a very short respiratory cycle (1.37s), which may test the temporal resolution limits of the current system.
Discussion and Limitations
While these results are promising, several factors must be addressed before full clinical implementation:
- Deformable Anatomy: The rigid phantom used does not account for the abdominal deformation and organ compression that occurs in live patients under the pressure of a US probe.
- Motion Dimensions: This study focused on Superior-Inferior (SI) motion. Future work must evaluate tracking in the Left-Right (LR) and Anterior-Posterior (AP) directions.
- Acoustic Windows: Patient-specific factors like BMI and bowel gas can degrade US image quality, a challenge less prevalent in phantom studies.
Conclusion
This phantom-based study confirms that ultrasound tracking can monitor pancreatic tumor motion with high accuracy. By providing real-time, non-ionizing feedback, this technology could enable clinicians to reduce safety margins, thereby sparing healthy tissue and allowing for more aggressive dose escalation in pancreatic cancer treatment.
Summary updated