Abstract
Purpose:
This study demonstrates a robust Cherenkov imaging-based solution to MR-Linac daily QA, including mechanical-imaging-radiation isocenter coincidence verification.
Methods:
A fully enclosed acrylic cylindrical phantom was designed to be mountable to the existing jig, indexable to the treatment couch. An ABS plastic conical structure was fixed inside the phantom, held in place with 3D-printed spacers and filled with water allowing for high edge contrast on MR imaging scans. Both a star shot plan and a four-angle sheet beam plan were delivered to the phantom; the former allowed for radiation isocenter localization in the x-z plane (A/P and L/R directions) relative to physical landmarks on the phantom, and the latter allowed for the longitudinal position of the sheet beam to be encoded as a ring of Cherenkov radiation emitted from the phantom, allowing for isocenter localization on the y-axis (S/I directions). A custom software application was developed to perform near-real-time analysis of the data by any clinical user.
Results:
Calibration procedures show that linearity between longitudinal position and optical ring diameter is high (R2 > 0.99), and that RMSE is low (0.184 mm). The star shot analysis showed a minimum circle radius of 0.34 mm. The final isocenter coincidence measurements were in the lateral, longitudinal, and vertical directions were −0.61 mm, 0.55 mm, and −0.14 mm respectively, and the total 3D distance coincidence was 0.83 mm, with each of these being below the 2 mm tolerance.
Conclusion:
This novel system provided an efficient, MR safe, all in one method for acquisition and near-real-time analysis of isocenter coincidence data. This represents a direct measurement of the 3D isocentricity. The combination of this phantom and the custom analysis application makes this solution readily clinically deployable after longitudinal analysis of performance consistency.
1. INTRODUCTION
The installation of magnetic resonance guided radiation therapy systems (MRgRT) in increasing numbers has introduced the benefits of real-time MR imaging to adaptive radiation therapy planning on a MR-Linac platform. These benefits include superior soft tissue contrast compared to x-ray imaging and continuous imaging during treatment, enabling high positional accuracy and robust inter-fraction adjustments. However, this technology has more quality assurance (QA) procedure complications than with conventional accelerators, due to a variety of reasons including the compatibility of traditional detectors with the magnetic environment and the effect of the magnetic field on dosimetric measurements1. Daily QA of a traditional linac can be performed with an onboard imager or detector array capable of measuring both the therapeutic MV beam and the kV imaging beam in order to assess output and isocenter verification2. This convenience does not carry over to MR-Linac systems, as the MR modality is not readily probed by these detectors and onboard x-ray imaging device are not universally available on MR-Linac models.
One particular daily QA measurement that is a challenge on MR-Linac systems is the verification of the mechanical, imaging and radiation isocenter coincidence, described in the AAPM Task Group 142 report with an acceptable tolerance of ≤ 2 mm3. This is a crucial parameter to verify against baseline, and there is responsibility placed on the physicist to determine an adequate method to evaluate the consistency of the alignment4. There have been procedures in the literature describing methods to perform this measurement, based on film, polymer gel, spatial localization of individual ion chambers, or MR-compatible ion chamber arrays5–7. These methods can be somewhat cumbersome and time-consuming either with setup or post-exposure processing or rely on reduced information to infer 3D isocenter coincidence. Currently, for the MRIdian system (ViewRay, Cleveland, OH, USA), the vendor-recommended method involves two pieces of radiochromic film, the first embedded in the provided cylindrical daily QA phantom and the second wrapped around the outer diameter, and subsequently irradiating with a five-field star shot. The readout of these films is then performed in post processing, which yields the radiation isocenter in the x-z plane, as well as along the y-axis since the beam is forward-peaked with the lack of flattening filter. However, in order to relate this measurement to the imaging isocenter, the laser marks must be manually marked on the film and analyzed in the post-processing coordinate system in order to translate back to the treatment coordinate system. While accurate, this process is very time consuming and not suited to be carried out on a daily basis. However, it remains the only method to directly quantify the 3D isocentricity.
It has been demonstrated that Cherenkov emission is proportional to dose deposited by monoenergetic x-ray beams, however the energy dependence of Cherenkov emission complicates the relation to dose for realistic clinical x-ray beams8. However, Cherenkov emission has been demonstrated as a beneficial tool for QA measurements related to beam geometry9, as Cherenkov intensity is monotonically related to dose. Cherenkov and scintillation imaging have been demonstrated as useful methods for various QA procedures in both traditional image guided radiotherapy and MRgRT10–12, and this is exploited in this context for high resolution 3D positional measurement of the radiation beam. In this study, we developed a custom Cherenkov imaging-based solution to measuring mechanical-imaging-radiation isocenter coincidence on the MRIdian system, entitled Visual Isocenter Position Enhanced Review (VIPER). This system consists of a physical phantom and analysis software application. The phantom is fully enclosed with deliberate design considerations for compatibility with the existing daily QA phantom jig, as well as conventional ion chamber output measurement.
2. METHODS
2.1. Phantom Design
The physical component of the VIPER system is a novel isocenter coincidence phantom, which was designed to be similar in shape to the existing MRIdian daily QA phantom, in that it consisted of a cylindrical acrylic body filled with water. Figure 1 shows the design of the VIPER phantom as rendered from SOLIDWORKS (Dassault Systèmes SOLIDWORKS Corp., Waltham MA, USA). The outer housing was composed of four custom-machined pieces of clear acrylic: a front end cap, rear end cap, cylindrical body, and rear hollowed plug for ion chamber insertion. The two circular end caps each with a diameter of 13.85 cm were machined out of 0.5” thick sheets of acrylic. The front and rear caps each had a flattened bottom side with a protruding tab specifically matching the commercial ViewRay daily QA phantom so as to be compatible with that phantom’s jig for couch mounting, as shown in Figure 2. As a stock component used for ease of fabrication, the cylindrical body of the phantom had a smaller outer diameter than the caps at 5.25” and an inner diameter of 4.5”, with a length of about 6.5”. Each cap was attached to the cylindrical body using 12 countersunk M3 PEEK screws, and a water-tight seal was obtained with a 1/16” wide O-ring placed in a machined groove in each end cap.
Figure 1.
3D renders of the VIPER phantom from the front (A) and rear (B), along with a deconstructed view of the assembly highlighting individual components (C).
Figure 2.
(A) Comparison of the manufacturer daily QA phantom (left) and the VIPER phantom designed for this study. (B) Compatibility of the VIPER phantom with the manufacturer provided jig.
While the outer face of the front end cap was untouched for optical imaging, three holes were threaded through the rear end cap. Two symmetric 1/4 NPT holes hosted quick-turn couplings for simplified connection to a water pump for air-bubble removal, while a central threaded port hosted the O-ring sealed hollowed plug. The plug design contained a hollow port for insertion of an MR-compatible A28 ionization chamber for daily output measurements (Figure 1(B)).
The four inner components of the phantom in order from rear to front included a rear spacer, square grid, radioluminescent conical frustum, and front spacer (Figure 1(C)). The inner components were all circular with outer diameters coincident with the inner diameter of the cylindrical body for simplified and secure assembly. The two spacers were 3D printed using black filament and were used to hold the central features of the design in place. Each spacer interfaced with their respective end caps via two additional O-rings, which were concentric with the couplings between the end caps and outer cylindrical body, in order to ensure a tight fit of the internal components. Matte black filament was used in printing to minimize potential optical reflections and leakage from the transparent outer acrylic. The length of the rear spacer was specified such that it would clear the outer length of the ion chamber plug. The grid component, which was a 1 cm spaced square grid inscribed in the circular profile of a 1 cm thick disk, was also 3D printed using black matte filament. This component was included for potential MRI spatial distortion tests and was held between the rear spacer and conical frustrum.
The conical frustum was the primary component of the phantom design and was custom milled out of white ABS plastic. Ionizing radiation passing through plastic yields Cherenkov emission, as does radiation passing through the surrounding water; however, Cherenkov photons in the plastic undergo multiple scattering events, thereby removing most if not all of the original directionality of the emission, when compared to the photon distribution in water. This increases the amount of light emitted from the ABS target that makes it to the camera, providing optical contrast between the plastic and water in the Cherenkov image. The slope of the edge of the cone was 45° to allow for 1:1 encoding of cone diameter to axial position. The front-facing side of this component was etched with a crosshair during machining, as well as a thin ring around the sloped face, which when combined provide the physical center of the component. Black paint was applied to these features to increase contrast in the optical background images. On the rear side of the structure there was a hollow cone-shaped cavity, the tip of which was coincident with the markings on the front side. When submerged this cavity filled with water, and the tip of the cavity was visible on an MR image due to the high contrast between plastic and water.
2.2. Phantom Irradiation and Image Acquisition
All measurements were made on the MRIdian MR-Linac system with MRI scanning and the single 6 MV flattening filter free (FFF) X-ray beam. The phantom was placed in the jig and purposefully aligned with offsets of 2 cm from the lasers in each direction and sent to isocenter after the imaging coils were in place, with caution taken to keep the viewport clear from the camera perspective as shown in Figure 3. First, the VIPER phantom was scanned with a built-in sequence with the highest spatial resolution available (92 s scanning time and a voxel size of 0.15 cm in each dimension). The center point of the conical structure was isolated on the scan, and appropriate shifts were applied to co-align this point with the coordinates (0, 0, 0) on the MR image. After repositioning, a second identical scan was taken and used as the baseline scan for the system. For all subsequent isocenter coincidence measurements, the phantom was always aligned to −2 cm in each direction, and appropriate shifts were obtained from the image registration built in to the ViewRay console, ensuring alignment of the VIPER phantom center with the imaging center.
Figure 3.
Photographs of the treatment bunker (A) as well as close-up views of the VIPER phantom aligned with imaging coils in the bore (B) and the wall-mounted camera aligned along the longitudinal axis.
For the measurement of the radiation isocenter, two custom plans were developed, as shown in Figure 4. The first plan was a star shot consisting of five equally spaced beams at angles of 0, 72, 144, 216, and 288 degrees. Each beam was a 1.5 cm (lateral, X) × 5 cm (longitudinal, Y) centered rectangle, and was optimized to deliver 100 cGy per field to the center of the phantom, resulting in a delivery of 684.6 monitor units (MU). The second plan consisted of four wide sheet beams with dimensions of 27.4 cm (lateral, X) × 0.83 cm (longitudinal, Y) at the four cardinal angles of 0, 90, 180, and 270 degrees, which when intersected with the conical structure generated a ring of Cherenkov emission. This second field size reflects the maximum and minimum field width in the lateral and longitudinal dimensions, respectively, in order to achieve the broadest and thinnest sheet of radiation possible. This plan was optimized for 150 cGy per field to the phantom center, which resulted in a delivery of 858.3 MU.
Figure 4.
Dose distributions overlayed on the MR images of the VIPER phantom for both the sheet irradiation (A) and the star shot (B).
Each plan was delivered and imaged using a custom C-Dose camera (DoseOptics LLC, Lebanon, NH, USA) and the accompanying C-Dose Research software. Similar commercially-available C-Dose cameras have been characterized in depth in a previous study13. Cumulative, background-subtracted Cherenkov images as well as cumulative background images were acquired at 17 Hz. Acquired image dimensions were 1920 by 1200 pixels. The software featured real-time background subtraction, which allowed for the room lights to remain on and therefore features of the VIPER phantom were readily discernable in the background images. The camera was outfitted with a blue-sensitive image intensifier to increase the blue-weighted Cherenkov signal in the phantom. The camera was fixed with a 200 mm lens (Canon Inc, Tokyo, Japan) with an aperture of f/6.8, and mounted to the wall such that the optical axis aligned vertically and laterally with isocenter, at a distance of 5 m longitudinally. The aperture size used provided sufficient depth of field such that the entire conical structure was in focus.
2.3. Custom Software and Image Analysis
The software component of the VIPER system consisted of a custom standalone application written in MATLAB and deployed on a clinical workstation, which was used to analyze the cumulative images acquired (one Cherenkov image and one background image for each plan noted in the previous section). This choice was made so that the data could be analyzed on the spot by any user without the need for post-processing and analysis. The app design consisted of four tabs for the different portions of the analysis: isocenter coincidence in the X-Z plane, isocenter coincidence along the Y axis, three-dimensional results, and calibration.
2.3.1. Isocenter Coincidence in the X-Z Plane
The first portion of the VIPER analysis software was designed to perform a star shot analysis on the first set of images to assess the radiotherapy isocenter coincidence in the lateral (X) and vertical (Z) directions. First, both the Cherenkov and background images are loaded into the application, and a flat-field correction is applied to both images (see calibration description in section 2.3.4.). Next, the center of the crosshair is clicked on the background by the user, and the software automatically detects the true center by assessing a 10 mm radial profile around the point clicked and finding the intersection point of the two lines. This resulting point is taken as the phantom center in the X-Z plane.
Next, the user specifies the number of beams used in the star shot and drags a circular ROI on the Cherenkov image around the edge of the star shot pattern. The radial profile used to find the peaks of each beam is an average of 10 concentric profiles centered around the user-drawn circle which is then median filtered with a 5-pixel moving window. Then, the resulting radial profile is rescaled from -1 to 1, and all zero crossings are determined; the peaks are taken as the midpoints of these zero crossings corresponding to positive values on the profile, and the number of peaks is cross-checked with the user-defined value. Opposing peaks were then joined by line segments, and the median of the intersection points of these segments was taken as the start for a search for the minimum circle, as described in previous work14. Once the minimum circle is found, the center point of that circle is taken as the position of the radiation isocenter in the X-Z plane.
2.3.2. Isocenter Coincidence Along the Y Axis
The second portion of the VIPER analysis software was designed to measure the physical longitudinal distance between the phantom center and radiation field center. First, both the Cherenkov and background images of the second treatment plan are loaded into the software in an identical manner as that described in section 2.3.1. Next, using a circular ROI-drawing tool, the etched ring is outlined by the user on the background image; a short 5 mm radial profile is then acquired at and centered on each point along the outline of drawn ROI, and the position of minimum pixel value along each profile is taken to be the position of the ring outline at that angle. These values are the averaged across all angles to determine the radius of the ring in the background image.
Next, the average radius of the optical ring displayed in the Cherenkov image is calculated automatically by software in three steps: a radial profile starting from the center point (as defined by the crosshair position determined prior) is acquired at each angle; the average position of the 40% maximum values along the rising and falling edge of each profile is taken to be the radius of the optical ring at the corresponding angle; the average radius over all angles is calculated and displayed.
Lastly, the diameters of the etched ring in the background image and the optical ring in the Cherenkov image are subtracted, and the resulting difference is mapped to a longitudinal difference in the Y-positions of the phantom center and the radiation isocenter via a linear fit determined from a pre-defined calibration curve (see section 2.3.4.).
2.3.3. Three Dimensional Isocenter Coincidence
Once the two portions of the analysis described above are complete, the differences between the phantom center and radiation isocenter in each of the three dimensions, dx, dy, and dz, are displayed along with the 3D vector magnitude dr. For each of these quantities, the numerical value is displayed in green if the absolute value falls below the TG-142 threshold of 2 mm; otherwise, it is displayed in red.
2.3.4. Image Analysis Calibration
The software requires three calibration steps in order to perform its function: flat-field image import, pixel size definition, and diameter-to-longitudinal position calibration. Flat-field images are performed with the camera in an identical physical configuration to the normal operating state and acquired with the C-Dose Research software using flood illumination via an 8” × 8” uniform LED board (Advanced Illumination, Rochester, VT, USA). The frame averaged image is then loaded into the software and normalized, and subsequently all loaded Cherenkov and background images are divided by this flat-field image to correct for vignetting and intensifier artifacts. The pixel size is determined by measuring a known feature on the phantom, such as the crosshair width on the conical structure in the phantom. This value is then used for pixel to mm conversion in all analysis steps.
The diameter to-longitudinal position calibration is performed by imaging the phantom under a specific set of pre-determined conditions: first, the phantom is aligned to the center of the MR image, in an identical fashion to daily setup prior to irradiation. Next, the plan containing the four sheet beams of radiation is delivered while Cherenkov images are being acquired. This is repeated five more times at longitudinal shifts (in mm) of −5, −2.5, −1, 1, and 2.5. At each longitudinal position, the radii of the physical and optical rings are calculated on the cumulative Cherenkov and background images, respectively, using the methods described above (section 2.3.2.). A linear fit is performed between the differences in the two diameters at each position and the corresponding longitudinal shifts. The fit parameters are saved and are subsequently applied to the measured differences in the ring radii during analysis.
2.4. Comparison with Existing Methodologies
In order to test the validity of the novel methods described, two comparative measurements were made. First, the Cherenkov starshot image was analyzed using the open-source Python library Pylinac, to assess both the calculated minimum circle radius and the x-z radiation isocenter position thereby validating the efficacy of the x-z analysis protocol presented in this study. Second, the 3D isocenter was measured using the vendor-recommended film-based method described above, as well as vendor-provided analysis software, in order to verify agreement between the accepted and novel methods.
3. RESULTS
3.1. Initial Diameter-to-Longitudinal Position Calibration
Using the pixel measurement of the known diameter of the front face of the conical structure (28 mm), the physical size in the imaging plane represented by one pixel was determined to be 0.24 mm. Figure 5 shows select Cherenkov images and radial profiles resulting from the diameter-to-longitudinal position calibration described previously (section 2.3.4). The diameter of the etched ring in the corresponding background images of the VIPER phantom was found to be 63 mm, matching the dimension measured in SOLIDWORKS. This value was subtracted from each of the measured optical ring diameters (radii marked with vertical red lines in Figure 5) and plotted against the known longitudinal shifts; the resulting plot and linear fit are also shown. These numerical data are displayed in Table 1. The relationship between the shift and optical ring diameter was found to be highly linear with an R2 value of 0.997. Additionally, the root-mean-square error was calculated as 0.184 mm, which gives an idea of the uncertainty in the y-position measurement. The extracted fit coefficients were used subsequently in the longitudinal isocenter coincidence analysis.
Figure 5.
Results from the diameter-to-longitudinal position calibration, including select Cherenkov images (left), corresponding normalized average radial profiles (center), and linear fit to the longitudinal shift vs. ring diameter difference data (right). Positions of the 40% maximum rising and falling edges are shown in dotted blue, while extracted optical radius is shown in red.
Table 1.
Correspondence between observed optical diameter shift sand linear axial shifts.
| Optical Diameter Shift (mm) | Linear axial shift (mm) |
|---|---|
| 7.93 | −5.0 |
| 3.12 | −2.5 |
| 0.21 | −1.0 |
| −1.97 | 0.0 |
| −4.23 | 1.0 |
| −7.08 | 2.5 |
3.2. Isocenter Coincidence Measurement
Figure 6 shows both the lateral-vertical and longitudinal components of the analysis, captured from the custom application. Figure 6(A) shows the background image of the VIPER phantom, including the view of the crosshair. Figure 6(B) shows the Cherenkov image used in the longitudinal analysis, which utilizes the calibration curve calculated previously (Figure 5). The physical diameter of the etched ring was found to be 63.34 mm, and the optical diameter was found to be 62.33 mm, yielding a diameter difference of 1.01 mm. The star shot image is shown in Figure 6(C), and the corresponding radial profile along with the beam peaks is displayed with a remapped x-axis to match the coordinate system used by ViewRay in Figure 6(D); using this data, the radius of the minimum circle was found to be 0.34 mm.
Figure 6.
Results of the isocenter coincidence analysis from the custom software: (A) background image of the VIPER phantom with the crosshair center and etched ring highlighted in orange; (B) optical ring detected (shown in red) overlayed on the Cherenkov image from the sheet irradiation for isocenter detection along the y-axis; (C) Analysis overlayed on the star shot Cherenkov image, along with minimum circle in the zoomed-in subplot for isocenter localization in the x-z plane (minimum radius found to be 0.34 mm); (D) Circular star shot profile acquired from the analysis in (C).
Based on the image analysis, the following differences between the phantom center and radiation isocenter were calculated: dx = −0.61 mm, dy = 0.55 mm, dz = −0.14 mm, and dr = 0.83 mm. These results show that the isocenter coincidence is within the 2 mm tolerance described in TG-142.
3.3. Comparison Measurements
The output from the Pylinac starshot analysis module reported a minimum circle radius of 3.75 mm and an x-z radiation isocenter position relative to the VIPER software result of Δx = 0.29 mm and Δz = 0.94 mm (1.1 and 3.9 pixels respectively). The results of the 3D MR-RT isocenter coincidence from the vendor-recommended film-based method were found to be dxFilm = −0.31 mm, dyFilm = 0.15 mm, dzFilm = 0.79 mm, and drFilm = 0.86 mm.
4. DISCUSSIONS
The VIPER system represents a near-real-time alternative to currently used film-based solutions for imaging-radiation isocenter coincidence verification, making it a feasible method for measuring this parameter on a daily basis. Additionally, the software developed as part of this system makes it useable by all clinical personnel immediately after data acquisition. The inclusion of an ionization chamber housing allows this phantom to be used for output measurements as well, with comparison to baseline of either an off-axis field irradiation, an off-axis chamber measurement, or a known longitudinal couch shift to center the chamber. This combination of developments allows this phantom to perform all TG-142 dosimetric measurements for MR-Linac daily QA. The VIPER analysis results, in 3D and along each axis, were found to be within 1 mm agreement both with the output from the Pylinac starshot analysis of the Cherenkov image as well as the film-based method results.
The analysis procedure including the use of the custom software application designed for this system requires the consistency of three calibration settings: the flat field image, the pixel size, and the diameter-to-longitudinal position calibration. The flat field image is important to correct for relative intensity differences in different regions of the image due to vignetting and intensifier defects, and this feature remains unchanged in the absence of physical camera modifications, such as internal adjustments or lens replacement. The pixel size is sensitive to changes in the imaging plane, such as focus and camera position, and therefore should be cross-checked on a regular basis to ensure there are no physical changes to the system; however, the aperture was set small to ensure that the depth of focus was large enough to keep the entire depth of the conical structure in focus. The 200 mm lens was chosen to optimize the resolution and field of view, as the camera is mounted 5 m from the isocenter (out of the 5 gauss line), but also to achieve an image closer to a true orthographic projection of the VIPER phantom in order to avoid analysis considerations from the perspective view. The diameter-to-longitudinal position calibration is crucial to the efficacy of the y-axis analysis and is similarly sensitive to camera position and focus as the pixel size. However, this is mitigated by the camera mount, which only allows for rotational adjustment in the lateral and vertical directions and is tightly fixed after initial setup. Still, the image field of view should be checked for positional changes on a routine basis to ensure continued viability of the calibration. It is also worth noting that this calibration method provides a relative diameter comparison between the phantom and beam positions along the y-axis; one could perform an absolute axial calibration instead by delivering separate beams at several known y-positions, however this method would be subject to an asymmetrical axial dose distribution due to the lack of a flattening filter, which would likely impact the ability to measure the true center of the field along the y-direction.
One effect that was observed was the systematic offset in apparent optical diameter compared to the physical diameter of the intersection ring as a given longitudinal position. As shown in Figure 5 and Table 1, there is a -2 mm offset in the apparent diameter, which is explained by the fact that Cherenkov light produced deeper in the ABS conical structure compared to the surface is able to diffuse upward along the inner outline of the ring, which biases the diameter measurement downward. This effect is corrected for in the linear calibration step, however. Additionally, in the optical diameter calculation, the decision to use specifically 40% of the maximum intensity as the rising and falling edges on the averaged radial profile, as opposed to 50% for example, was made in order to avoid the subtractive impact of the etched ring on the absolute intensity value; this compromise yielded the most consistent replication of the true location of the radiation isocenter.
While the camera and DC power supply were kept out of the 5 gauss line, there was a small effect on MR image quality when the power supply was plugged into the wall outlet, likely due to interfering radiofrequency noise. This is solely related to powering the camera, as data is transmitted from the camera to the console over fiber. To avoid any possible distortions, the camera was only powered when irradiating the VIPER phantom, and the unit was unplugged during MR imaging.
The use of a 2 mm tolerance for the isocenter coincidence measurement is based off the TG-142 Report guidelines for non-SRS/SBRT treatments, however there is no such guideline specifically for MR-guided external beam radiotherapy treatments. With the prevalence of daily adaptive planning for such treatments using newly acquired imaging, as well as the common use of hypofractionation, it is difficult to determine what the correct tolerance should be for this application. For the purposes of this study, we apply the 2 mm tolerance for agreement with prior studies, however, longitudinal analysis of the performance of this method over time will clarify if it provides sufficient accuracy for compliance with a 1 mm tolerance.
As has been described in prior publications5,7, one inherent limitation to the resolution of this measurement is the spatial resolution of the available scan sequences, which is currently limited to 1.5 mm × 1.5 mm × 1.5 mm. While this is not necessarily the granularity of the image registration algorithm providing the couch shifts, it is a factor that limits measurement resolution; this positioning uncertainty has been estimated previously at 0.5 mm7. This could explain some of the discrepancy between the measurements made by the VIPER system and those made using other programs and methods, although the accuracy and precision of each individual method would need to be understood in order to validate such a claim.
Future work will involve a longitudinal assessment of this method compared to both film and ion chamber profiler-based isocenter coincidence verification methods, as well as tracking dose and dose rate constancy using an embedded ion chamber in the VIPER phantom compared with the manufacturer-provided daily QA phantom. Additionally, development of a time-gated, non-intensified camera would greatly reduce the cost of implementing a system such as this in the clinic in an efficient manner, although a typical off-the-shelf camera with sufficient exposure settings would also work in this application at the expense of convenience. Lastly, although designed for MR-Linac QA, this novel system could be applied to conventional accelerators with kV imaging; this application is also planned for future work.
5. CONCLUSIONS
The VIPER system described in this study allows for a convenient, streamlined, and near-real-time measurement of imaging-radiation isocenter coincidence verification for MR-Linac systems, overcoming the disadvantages of traditional film-based methods. The fully enclosed phantom design and custom analysis software yield a robust solution for MR-Linac daily QA. The source code for the VIPER program as well as data used in this study are available online at https://github.com/daniel-a-alexander/VIPER.
Acknowledgements
This work has been predominantly funded by NIH grant R01 EB023909 with hardware support through R44 CA232879 and with the support of the Norris Cotton Cancer Center shared resources in 5P30 CA023108–41. The authors would like to thank Samuel Streeter and Alberto Ruiz for their input on the physical design of the VIPER phantom, as well as ViewRay for assistance with design specs.
Footnotes
Disclosures and Conflicts of Interest
DA, PB, JA, RZ, and DG have patent 63/157,240 pending which is related to this work. BP, PB, RZ, and DG have patents on the use of Cherenkov imaging in radiotherapy dosimetry licensed to DoseOptics, LLC. BP is the president and PB is the principal scientist at DoseOptics LLC, which supplied cameras and software for this study. DA is a research consultant for DoseOptics LLC outside the context of this work.
Contributor Information
Daniel A. Alexander, Thayer School of Engineering, Dartmouth College, Hanover, NH, USA.
Petr Bruza, Thayer School of Engineering, Dartmouth College, Hanover, NH, USA.
Aris G. Rassias, Thayer School of Engineering, Dartmouth College, Hanover, NH, USA
Jacqueline M. Andreozzi, Moffitt Cancer Center, Tampa, FL, USA
Brian W. Pogue, Thayer School of Engineering and Geisel School of Medicine, Dartmouth College, Hanover, New Hampshire, USA, Norris Cotton Cancer Center, Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire, USA
Rongxiao Zhang, Norris Cotton Cancer Center, Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire, USA.
David J. Gladstone, Geisel School of Medicine and Thayer School of Engineering, Dartmouth College, Hanover, New Hampshire, USA.
References
- 1.Jelen U, Begg J. Dosimetry needs for MRI-linacs. J Phys Conf Ser. 2019;1305:012010. doi: 10.1088/1742-6596/1305/1/012010 [DOI] [Google Scholar]
- 2.Sun B, Goddu SM, Yaddanapudi S, et al. Daily QA of linear accelerators using only EPID and OBI. Med Phys. 2015;42(10):5584–5594. doi: 10.1118/1.4929550 [DOI] [PubMed] [Google Scholar]
- 3.Klein EE, Hanley J, Bayouth J, et al. Task Group 142 report: Quality assurance of medical acceleratorsa). Med Phys. 2009;36(9Part1):4197–4212. doi: 10.1118/1.3190392 [DOI] [PubMed] [Google Scholar]
- 4.Smith K, Balter P, Duhon J, et al. AAPM Medical Physics Practice Guideline 8.a.: Linear accelerator performance tests. J Appl Clin Med Phys. 2017;18(4):23–39. doi: 10.1002/acm2.12080 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Mittauer KE, Dunkerley DAP, Yadav P, Bayouth JE. Characterization and longitudinal assessment of daily quality assurance for an MR-guided radiotherapy (MRgRT) linac. J Appl Clin Med Phys. 2019;20(11):27–36. doi: 10.1002/acm2.12735 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.A Method to Determine the Coincidence of MRI-Guided Linac Radiation and Magnetic Isocenters - Latifi Kujtim, Moros Eduardo G., Zhang Geoffrey, Harrison Louis, Feygelman Vladimir, 2019. Accessed November 16, 2020. 10.1177/1533033819877986 [DOI] [PMC free article] [PubMed]
- 7.Dorsch S, Mann P, Elter A, et al. Measurement of isocenter alignment accuracy and image distortion of an 0.35 T MR-Linac system. Phys Med Biol. 2019;64(20):205011. doi: 10.1088/1361-6560/ab4540 [DOI] [PubMed] [Google Scholar]
- 8.Glaser AK, Zhang R, Gladstone DJ, Pogue BW. Optical dosimetry of radiotherapy beams using Cherenkov radiation: the relationship between light emission and dose. Phys Med Biol. 2014;59(14):3789. doi: 10.1088/0031-9155/59/14/3789 [DOI] [PubMed] [Google Scholar]
- 9.Miao T, Bruza P, Pogue BW, et al. Cherenkov imaging for linac beam shape analysis as a remote electronic quality assessment verification tool. Med Phys. 2019;46(2):811–821. doi: 10.1002/mp.13303 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ashraf MR, Bruza P, Pogue BW, et al. Optical imaging provides rapid verification of static small beams, radiosurgery, and VMAT plans with millimeter resolution. Med Phys. 0(0). doi: 10.1002/mp.13797 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Andreozzi JM, Mooney KE, Brůža P, et al. Remote Cherenkov imaging-based quality assurance of a magnetic resonance image-guided radiotherapy system. Med Phys. 2018;45(6):2647–2659. doi: 10.1002/mp.12919 [DOI] [PubMed] [Google Scholar]
- 12.Alexander DA, Zhang R, Brůža P, Pogue BW, Gladstone DJ. Scintillation imaging as a high-resolution, remote, versatile 2D detection system for MR-linac quality assurance. Med Phys. 2020;47(9):3861–3869. doi: 10.1002/mp.14353 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Alexander DA, Bruza P, Farwell JCM, et al. Detective quantum efficiency of intensified CMOS cameras for Cherenkov imaging in radiotherapy. Phys Med Biol. 2020;65(22):225013. doi: 10.1088/1361-6560/abb0c5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.González A, Castro I, Martınez JA. A procedure to determine the radiation isocenter size in a linear accelerator. Med Phys. 2004;31(6):1489–1493. doi: 10.1118/1.1755491 [DOI] [PubMed] [Google Scholar]