Abstract
Robotic radiosurgery/Radiotherapy is increasingly adopted in clinics, and quality assurance (QA) of CyberKnife’s variable-aperture Iris™ collimators requires sub-millimeter precision. Conventional film-based QA for the 12 Iris™ cone sizes (ranging from 5 to 60 mm) is both time consuming (120 minutes for all or 30 minutes for 3 cone sizes) and highly experience dependent. To improve the efficiency, a high-resolution 2D diode detector array, sampling every 2.5 mm, was evaluated for Iris™ aperture size QA. This study focused on a spatial frequency analysis, a dose profile reconstruction, and a sensitivity study to beam size variances. Dose profiles of the 12 cones scanned with a high-resolution diode in a water tank were utilized as the gold standard for comparison. Spatial Fourier transform of these profiles were analyzed to explore applicable sampling frequency. Next, the dose profiles were artificially sampled with a 2.5 mm gap, and then interpolated using a Python-based cubic B-spline. Finally, sensitivity of the diode array system to various field sizes was measured by changing source to detector distance. We found, utilizing the diodes system, QA time was reduced to less than 10 minutes. Spatial frequency of the dose profile showed little contribution beyond 0.2 mm⁻¹, so a Nyquist sampling of 0.4 mm⁻¹ is appropriate for dose verification, corresponding to a 2.5 mm gap. Dose profiles were reconstructed using Cubic B-spline with good agreements to nominal for cones 7.5 mm and larger. The measured Iris™ size using the SRS MapCheck had a standard error of ±0.12 mm. Primarily, the 2D Diode array with a spatial resolution 0.4 mm⁻¹ is appropriate for dose verification for these cones above 7.5 mm, and its application would substantially improve Iris™ QA efficiency.
The Challenge of Iris Collimator Quality Assurance
In robotic radiosurgery, precision is paramount. The CyberKnife system utilizes an "Iris" collimator—a sophisticated device featuring two banks of tungsten segments that move to create 12 distinct circular-ish (dodecagonal) beam sizes, ranging from 5 mm to 60 mm. Because these beams are used to treat small tumors near critical structures, their sizes must be verified with sub-millimeter accuracy.
The current "gold standard" for this verification is radiochromic film. While accurate, film is logistically demanding: it requires specific batches, sensitive handling, hours of processing time, and expert interpretation. A full 12-cone QA session can take up to two hours, leading many clinics to sample only 3 of the 12 cones during monthly checks. This study evaluates an electronic detector that can match film’s accuracy while significantly reducing the time investment.
Methodology
The researchers evaluated the SRS MapCheck, a 2D array of 1,013 n-type solid-state diodes spaced 2.5 mm apart. To determine if this 2.5 mm spacing was sufficient for small-field dosimetry, three main strategies were employed:
- Spatial Frequency Analysis: Fast Fourier Transforms (FFT) were performed on "gold standard" dose profiles (originally scanned in a water tank) to determine the high-frequency information content within the beam.
- Profile Reconstruction: A Python-based cubic B-spline interpolation was used to reconstruct continuous dose profiles from the discrete 2.5 mm diode samples.
- Sensitivity Testing: Since the Iris collimator cannot be programmed to arbitrary sizes, the team simulated field size errors by varying the Source-to-Detector Distance (SDD). Adjusting the detector distance changes the projected beam size on the array, allowing for the detection of minute deviations.
Key Insights and Results
The study revealed that the Nyquist-Shannon sampling theorem is central to electronic QA. FFT analysis showed that the majority of the dose profile signal exists at spatial frequencies below 0.2 mm⁻¹. A sampling frequency of 0.4 mm⁻¹ (corresponding to a 2.5 mm gap) is sufficient to reconstruct the signal without significant loss of information for most field sizes.
Performance Metrics
- Accuracy: The system measured Iris aperture sizes with a standard error of ±0.12 mm.
- Linearity: Sensitivity studies showed a high coefficient of determination ($R^2$ between 0.9814 and 0.9988) for field size measurements across different distances.
- Reconstruction Quality: For cones $\ge$ 7.5 mm, the cubic B-spline interpolation matched the water tank "gold standard" with a maximum difference of only 0.6% at off-axis distances.
- Efficiency: Setup took approximately 5 minutes, and beam delivery took 3 minutes, representing a >90% reduction in total QA time compared to film.
Limitations and Future Outlook
The primary limitation identified is the 5 mm cone. The steep gradients and small area of the 5 mm beam push the 2.5 mm diode spacing past its physical limits, resulting in higher errors (~2.5%). For this specific field size, clinics may still need to rely on film or higher-resolution monolithic silicon-diode arrays.
Future work could involve refining interpolation algorithms or integrating diode array data directly into the CyberKnife’s control software for real-time, automated collimator calibration. This transition from film to electronic arrays represents a significant step forward in clinical efficiency and standardization for robotic radiosurgery.
Summary updated