TL;DR
This paper introduces an automated method to detect residual tumor motion during deep-inspiration breath-hold (DIBH) radiotherapy using cone-beam CT (CBCT) projections. By analyzing fiducial markers, the technique provides daily motion assessments, crucial for refining treatment margins and enhancing precision in adaptive radiotherapy.
What Problem Does This Paper Address?
Radiotherapy for pancreatic cancer aims to deliver precise radiation doses to the tumor while sparing surrounding healthy tissues. Respiratory motion, however, can cause the tumor to shift, necessitating larger treatment margins to ensure adequate coverage. While deep-inspiration breath-hold (DIBH) techniques are employed to minimize this motion, residual movements can still occur. Current methods for assessing this residual motion are often manual, time-consuming, or require additional equipment, limiting their daily use. This paper addresses the need for a reliable, automated, and cost-effective method to quantify daily residual tumor motion during DIBH to improve treatment accuracy and patient safety.
Methods
The authors developed an automated algorithm to detect fiducial markers in cone-beam CT (CBCT) projection images. These markers, implanted endoscopically, serve as surrogates for the tumor's position. The core of the method involves:
- Marker Localization: Using predicted fiducial positions based on the planning CT and known imaging geometry (Equations 1 & 2), the algorithm searches for markers within specific image patches.
- Gradient Analysis: To identify marker edges, the method calculates gradients in multiple directions (horizontal, vertical, and oblique) across the image patch. This involves analyzing intensity differences over one or two pixels (Equations 3-14).
- Gradient Combination: The calculated gradients are then combined using square roots of products (Equations 15-22) to create "heights" representing marker features.
- Pattern Matching: The combined gradient information is used to match the detected marker patterns to the expected fiducial positions from the planning CT.
- 3D Position Calculation: Lateral (X) and longitudinal (Y) coordinates of the fiducials are calculated from a series of projection images. Outliers are identified and removed based on deviations from expected positions and curve-fitting results (Equations 23-28).
- Motion Analysis: The algorithm quantifies residual motion by analyzing the differences in fiducial positions within a single breath-hold, between two breath-holds, and the "gap" between the end of one breath-hold and the beginning of the next.
This approach is novel in its automation and its use of CBCT projections for daily assessment, avoiding the need for additional imaging or manual intervention.
Datasets
The study utilized data from nine pancreatic cancer patients treated at their institution. For each patient, multiple CT simulation scans were acquired under DIBH conditions. Additionally, numerous CBCT projection images were collected during treatment. This retrospective dataset was used to validate the developed automated method.
Key Insights and Ideas
- Residual Motion is Pervasive: Even with DIBH, residual tumor motion is consistently observed in all patients, highlighting the limitations of breath-holding alone for achieving perfect immobilization.
- Daily Assessment is Feasible: The automated method demonstrates that daily quantification of residual motion from readily available CBCT projections is achievable.
- CBCT Projections Hold Valuable Information: Unlike reconstructed CBCT volumes which average out motion, the raw projection data contains detailed information about real-time, albeit short-duration, movements.
- Motion Variability: The study reveals significant variability in residual motion, both within a single breath-hold and between consecutive breath-holds, underscoring the need for patient-specific margins.
Key Results
- Intrafraction Motion: During repeated breath-hold simulations, an average intrafraction target motion of 1.9 ± 2.2 mm was observed, with a maximum displacement of 8 mm.
- CBCT Motion: Within a single CBCT scan, residual motion reached up to 7.3 mm. The average drifting range across 94 CBCT scans was 3.8 ± 1.1 mm.
- Standard Deviation of Drift: The average standard deviation of drift was 1.5 ± 0.5 mm.
- Inter-Breath-Hold Motion: Significant drift (1.3 ± 1.2 mm) and inter-breath-hold gaps (2.6 ± 2.0 mm) were detected within the same CBCT scan.
- Margin Implications: The findings suggest that planning margins should be at least twice the standard deviation of fiducial marker positions (approximately 1.5 mm), indicating a need for margins around 3 mm to cover 95% of positions.
Limitations and Future Work
- Limited Scope of Motion Tracking: The current method primarily tracks motion in the longitudinal (Y) direction, as this is typically the largest component of respiratory motion. Motion in orthogonal directions, which can change with gantry rotation, is not continuously tracked.
- Marker Relative Position Changes: The relative positions of fiducial markers might change between simulation CT acquisition and treatment, potentially complicating marker identification.
- Preliminary Study: The study involved a small cohort of nine patients, limiting the ability to draw definitive conclusions or predict daily motion patterns.
- CBCT vs. Treatment Motion: The evaluated motion occurs during CBCT acquisition (approx. 40s) and may differ from motion during actual beam-on time.
- Future Directions: The authors propose further studies with more patients, validation of residual motion consistency throughout a treatment fraction, and correlation of marker motion with other motion monitoring techniques (e.g., ultrasound, Calypso, MR-guided systems). They also aim to investigate day-to-day changes in residual motion patterns and explore real-time monitoring of other physiological signals.
Summary updated