Emulating Clinical Workflow of Scintillator Array Dosimetry for FLASH PBS Proton Therapy

Abstract

Purpose

The emergence of ultra-high dose rate (UHDR) pencil-beam scanning (PBS) proton therapy and associated clinical trials has outpaced the capabilities of in vivo treatment validation systems. Current dosimetry tools are limited in spatiotemporal resolution and are insufficient for clinical FLASH treatment monitoring. This work aims to demonstrate the ability to monitor UHDR proton therapy in an emulated clinical in vivo setting using a novel scintillation imaging system. We demonstrate the system’s ability and evaluate the accuracy of measuring 2D dose and dose rate maps with verification against treatment plan projection.

Methods

A novel optical dosimetry imaging system composed of a 1kHz intensified CMOS camera, stereo-vision system, and a deformable scintillator array was deployed at a clinical proton therapy center for end-to-end validation. A conventional lung PBS plan was modified to simulate FLASH treatment at 99nA and 250MeV. Validation was performed using an anthropomorphic chest phantom through cumulative dose comparison with gafchromic film, spot position analysis, and PBS dose rate area verification. Clinical workflow was evaluated for impact on treatment time and compatibility to prepare the system for large animal and clinical studies.

Results

Scintillation imaging showed high gamma passing rates comparing the cumulative dose distribution to film (92.5% at 1%/1mm, >99.9% at 2%/2mm). Spot monitoring achieved sub-millimeter accuracy (0.32±0.19mm deviation). Planned and measured surface dose rates showed excellent agreement, with a 0.71% difference in coverage at 40Gy/s. Time added to treatment workflow was limited to array placement, contributing ~1 minute. Post-beam array activation showed negligible additional exposure (671nSv/hr).

Conclusion

This study presents the first practical implementation of scintillator array dosimetry for UHDR PBS proton therapy, enabling real-time dose and dose rate monitoring over complex geometries. The system offers a novel approach to in vivo treatment validation with clinical compatibility, providing unique metrics for UHDR proton beam dynamics.

1. Introduction

Pencil beam scanning (PBS) proton therapy has experienced widespread implementation over the last 15 years because of its ability to administer highly modulated, conformal radiation doses with superior organ-at-risk (OAR) sparing as compared to photon therapy¹. The scanning nature of PBS delivery typically leads to treatment plans with highly complex spatiotemporal fields and several energy layers. This approach of Bragg peak dose modulation offers numerous advantages in organ at risk (OAR) sparing, target dose conformity, and local dose rate control. On the other hand, it also introduces stringent delivery constraints since small deviations from the treatment plan may result in significant variation in dose delivery² through range uncertainty, patient motion, spot size, and particle travel characteristics^3,4.

The development of ultra-high dose rate (UHDR) therapy has introduced the possibility for enhanced normal tissue sparing while retaining tumor control, as demonstrated with UHDR electron^5–7, proton^8–10 and X-ray beams^11,12, leading to the emergence of both animal studies and clinical trials^9,13–16. Electron UHDR is currently most accessible for research and trials, with access to proton UHDR being less widespread and X-ray UHDR more limited still. Despite accessibility constraints, PBS UHDR proton therapy is experiencing increased interest and represents a promising FLASH implementation due to its superior dose conformity in deep-seated targets and the capacity to achieve UHDR parameters using existing clinical infrastructure. On the other hand, the lack of understanding of the exact physiological processes responsible for FLASH¹⁷ combined with ongoing studies of biological parameters such as oxygen content^18,19 and different dose rate definitions²⁰ led to reports of varying critical dose rate thresholds²¹ and treatment outcomes. It has been observed that the level of tissue sparing depends on a specific dose rate²² which is spatially heterogeneous in a PBS treatment field²³. Currently existing dosimetry tools are unable to operate at the specifications required for spatial and temporal measurement in UHDR radiation deliveries. Many detectors fail to respond linearly in UHDR regimes²⁴ or inherently do not provide spatial information necessary to monitor dose rate throughout the treatment field. Point dosimetry offers only extremely limited measurement due to aforementioned spatial heterogeneity and the need for submillimeter localization that is impractical in clinical setting²⁵.

This need for UHDR-capable dose and dose rate monitoring techniques has driven research toward spatially distributed dosimetry. Yang et al proposed a 2D ionization chamber array placed in the beamline at the scanning nozzle snout,²⁶ achieving high spatial, temporal, and dosimetric accuracy. Likewise, multi-layer strip ionization chamber 2D monitoring was proposed by Harrison et al., showing a high gamma passing rate for spots delivered to a flat surface²⁷. Rather than being fully two-dimensional, the strip ion chamber approach relies on reconstructing 2D dose/dose rate maps from cumulative charge projections to two orthogonal directions precluding accurate use with non-symmetric, non-Gaussian proton spot shapes. A 2D scintillator sheet imaging study recently showed promising capability of dose/dose rate monitoring for UHDR applications, achieving high agreement without forfeiting signal between measurement positions^23,28. The challenge of PBS proton pre-treatment QA is therefore met with several potential solutions, but translation to in vivo monitoring presents additional complexity. Patient surface geometry is usually non-regular and deforms with unpredictable motion before and during the treatment, which in the case of PBS may result in significant degradation of the treatment quality²⁹. In addition to the lack of 2D sensing, existing in vivo dosimetry solutions also fail to adapt to the motion and local surface deformation.

Localized methods for in vivo dosimetry have been proposed such as film, OSLDs/TLDs, diodes, or scintillating fibers³⁰ with varying levels of success²¹. The main drawbacks of such point dosimetry include the inability to present dose rate information (film and OSLD/TLDs) and their limited spatial resolution, which prevents comprehensive dose and dose rate monitoring across the full treatment. 2D printed scintillators have been proposed³¹ but so far have been limited to photon beams and would likewise present issues with calibration, angular correction, and direct plan comparison. Imaging 2D Cherenkov and/or luminescence emission from tissue has been investigated^32,33, but its potential for clinical adoption is not yet clear due to the lack of a robust optical absorption and scattering correction and low light yield.

Consequently, the lack of viable in vivo PBS proton therapy monitoring tools has significantly limited the availability of assessment metrics, resulting in few approaches for validating such deliveries. This work delivers the first comprehensive end-to-end application of a high-speed scintillation imaging system using a semi-rigid deformable mesh for validation of time-resolved 2D dose and dose rate maps during UHDR PBS proton treatment delivery on an anthropomorphic phantom. This validation is key to confirming that the resulting metrics are adequate for effective treatment monitoring, while ensuring the technique presents minimal disruption to the clinical workflow – both essential for facilitating integration into large animal and clinical UHDR PBS trials.

2. Methods

2.1. Setup

The dosimetry system (BeamSite Ultra, DoseOptics LLC) consisted of an intensified 2-megapixel ultra-fast complementary metal-oxide semiconductor (CMOS) camera operating at a 1 kHz frame rate and 99.8% duty cycle, accompanied by two 2-megapixel stereo-vision cameras on either side. Both cameras were time-synchronized using software timestamps. The system used 50 mm focal length lenses, configured for a 40×30 cm² field of view at 2 m distance from isocenter. The separation of the stereo-vision cameras was set to 15 cm (based on ~1/30 of the working distance) to ensure reliable capture of the mesh geometry with <0.5 mm accuracy.³⁴

The array was constructed of 485 plastic scintillating elements (Penn-Jersey X-Ray, Blue-800 Scintillator) connected via a polyester mesh forming an array with 7.5 mm inter-element spacing and 0.5 mm gaps. The procedure for application is likewise outlined in Figure 1. We used an 11×21 cm² scintillator array consisting of 6.5 mm inscribed diameter hexagonal elements (17 staggered rows of 29 elements, omitting the edge element every 2 rows). The array was adhered to one side of a medically approved adhesive (3M Medical Tape 9917, Double Sided Spunlaced Nonwoven Fabric, 62# Liner) as part of the pre-treatment preparation. Another approach, which may be preferred for patients with sensitive skin or early onset of erythema, is to attach the array to a plastic wrapping using the same procedure and lay over the target area, thereby still retaining contact while preventing topical irritation. The array was then laid over the target surface where it deformed to geometric variations. The removal was likewise simple, as the array is internally rigid which allows for the adhesive to be peeled off without damaging the general structure. We followed the workflow where the array is applied right after patient’s (here phantom) positioning using CBCT and removed immediately after treatment.

Figure 1:

The proposed treatment monitoring setup for conventional or UHDR PBS proton therapy is shown. A camera suite, consisting of an intensified CMOS ultra-fast camera and two white light USB cameras, is located outside of the couch or gantry operating field. The procedure for the scintillator array setup, application, and removal is likewise outlined.

To simulate a dose/dose rate-optimized plan in UHDR conditions, a 3-field conventional dose rate lung SBRT PBS plan was converted to a simulated FLASH plan. First the spot positions from the multi-energy plan were collapsed into one 250 MeV energy layer. Neighboring spots were then combined through spatial averaging, weighting the position in proportion to the MU/spot, until all the minimum spot dwell time was at least 2.8 ms, assuming delivery at 99 nA. The spot delivery pattern was then reordered to increase dose rate by minimizing the travel time between neighboring spots. While we acknowledge that this technique did not lead to a fully optimized FLASH beam pattern such as those seen in prior studies,^35,36 it yielded a good representation of such a plan both with respect to dose rate gradients and pencil beam trajectory. The plan was delivered using a Varian ProBeam system operating at 250 MeV and 99 nA. The medial chest region was chosen as the delivery area due to the high spatial variation of the superficial anatomy to validate the conformity of the scintillator mesh and its practical application. Additional application sites are shown as examples in Figure 1, highlighting applicability to a variety of treatment areas and orientations. System performance is also not expected to vary with placement location or plan complexity, provided that the array is entirely visible to the iCMOS camera, and the planned fields are entirely encompassed within the array bounds with sufficient temporal delay (>5 seconds) between sequential deliveries. Finally, to support the feasibility of clinical translation, the impact of the proposed procedure on treatment setup time, and the level of material activation was investigated.

2.2. Acquisition and Processing Workflow

The proposed monitoring and validation technique is outlined in Figure 2. Prior to treatment, the additional water equivalent thickness (WET) presented by the array may generally be incorporated either through passive correction by adding a bolus to the target planning surface in the treatment planning system, or by attaching the mesh to the patient during the CT simulation. During treatment, the array attachment procedure closely resembled that of bolus placement. Owing to the large mesh area coverage, the exact array positioning is not important as long as the full treatment field is covered. Prior to and following the application, the array may be sterilized using sanitizing wipes, making it reusable for future applications

Figure 2:

Proposed workflow for applying the scintillator mesh monitoring technique. The complete data processing pipeline is shown, including image darkfield and flat field correction, geometric correction based on stereo-vision reference, angular correction, spot profile interpolation, and intensity to dose translation. The array correction section of the processing workflow is highlighted in green.

During delivery, the scintillation images were acquired and saved as 8-bit raw files at 1 kHz frame rate, while the stereo-vision images were acquired at 30 Hz and saved as 16-bit raw files. The complete processing workflow is outlined as a subset of Figure 2. Frames containing signal above the noise floor were isolated by identifying the range between the first and last instances of detected scintillation signal. The remaining dataset served as background images for background correction. After the delivery in a pre-processing stage, each signal frame underwent dark field and flat field correction to eliminate constant offset and light response inhomogeneity of the intensifier and CMOS sensor. Dark field acquisition is done by placing a lens can over the iCMOS camera and acquiring images. Flat field acquisition was done following the QA procedure described by Clark et al²³, delivering a large uniform field to a flat calibration board covering the full camera FOV. The background frames were averaged and subtracted from the signal frames to isolate the scintillation signal. The pre-processed image stack was then subjected to geometric correction using centroid positioning information from a known geometric reference, angular correction, and signal profile interpolation as detailed in Section 2.3.

2.3. Scintillator Array Signal Corrections

2D scintillation imaging has previously been proposed for dose verification²³, but in vivo clinical translation has proven to be difficult due to the number of non-trivial corrections necessary to ensure that the intensity-to-dose relationship is maintained throughout the delivery. These corrections account for geometric, angular, and continuity effects since the scintillation map must provide data comparable to the plan in beams-eye view, retain its dose linearity, and provide a continuously sampled dose profile. The proposed semi-rigid mesh was specifically designed to address these limitations, which are evaluated here.

The scintillation and stereo-vision processing workflow³⁷ is depicted in Figure 3. Image pair data from the stereo-vision camera pair was used to relate the position of each scintillating element in both stereo images to its 3D location and construct a 3D point cloud of the scintillating elements at each time frame. The point cloud was then used to transform the intensified CMOS image to the same perspective as the gantry, or beam’s eye view. The angular dependence of scintillator emission was likewise corrected using the 3D information. For each scintillator centroid, the corresponding surface normal vector was identified and by referencing the gantry and camera position in the point cloud coordinates. The observation and irradiation angles with respect to each mesh element were derived and used to individually correct the output of each element. Previously derived constraints on acceptable observation angles limit delivery monitoring to 75 degrees, but would be addressed in a multi-camera setup.³⁷ Lastly, the lack of signal between neighboring array elements at each spot was accounted for by using the scintillation intensity distribution as a reference and fitting an a-priori spot kernel obtained from a separate film measurement (EBT XD, Ashland) of a single spot delivered prior to treatment, aligned along the couch axis. The final corrected scintillation map was translated into a dose map using an intensity-to-dose linearity curve obtained from pre-treatment scintillator characterization.

Figure 3:

Data processing and intensity correction workflow showing the use of stereo-vision imaging to geometrically correct the intensified CMOS images to beams-eye-view, use the individual-element normal vectors to correct for the angular dependence of emission, and account for inter-element intensity loss by fitting the single spot profile to a known single-spot kernel delivered to film.

2.4. Delivery Evaluation and Metrics

Following the treatment, dose and dose rate metrics were used to validate the delivery. Cumulative dose profiles were compared against the repeated delivery onto flat gafchromic film using gamma analysis with increasingly strict criteria. Taking advantage of the temporal information, centroid analysis was performed to analyze spatial deviation between the dwell points specified in the treatment plan and those identified through scintillation imaging. In this case, a dwell point was defined as the average centroid position of any successive spots that were separated by less than 0.5 mm to exclude frames where the beam was scanning between spots. PBS dose rate was also evaluated as defined by Folkerts et al.²⁰ and is represented in equation 1.

$$\dot{D_{\mathit{\text{PBS}}}}\left(\overrightarrow{x}\right)=\frac{\left(D\left(\overrightarrow{x}\right)-D_{†}\right)-D_{†}}{t\left(\overrightarrow{x}\right)}$$ #(1)

Where the dose rate $\dot{{(D}_{PBS})}$ at a point in the field $\left(\overrightarrow{x}\right)$ is governed by the time (t) that a dose between the threshold $\left(D_{†}\right)$and $\left(D\left(\overrightarrow{x}\right)-D_{†}\right)$ is delivered. To achieve meaningful comparison, a ground truth model is necessary but currently existing dose monitoring methods are not able to provide single spot dose information with sufficiently high temporal resolution. A previously developed and tested³² independent MATLAB simulation algorithm was therefore used to create a comparable ground truth dose rate map. This algorithm uses a map of the planned spot dwell positions with associated dwell times and spot velocities in X and Y axes within the beams-eye-view frame of reference. This information is then used along with a single spot profile obtained from film to simulate the delivery at any desired frequency (1kHz was used in this work to match the imaging frame rate), adjusting the spot profile accordingly to achieve the correct cumulative dose map while providing single spot dose information with single frame temporal resolution. A $D_{†}$ threshold of 0.5 Gy (~3% of maximum observed dose) was chosen for the simulated and imaged PBS dose rate calculations based on work by Clark et. al showing PBS dependence on $D_{†}$ selection²³.

3. Results

The scintillator array application procedure was carried out using the anthropomorphic phantom and timed to gauge the additional time added to the treatment workflow. Attachment onto the treatment area added ~1 min to the total time and initialization of the imaging software was done while the room was being secured, presenting no waiting time. The additional radiation activity of the mesh after 250 MeV proton beam irradiation was estimated to be 671 nSv/hr by sampling the activity at the surface using a survey meter (Thermo Scientific FH40) immediately following the delivery. The plan delivered a total of 24,987 MU leading to an induced activity of 2.68E-2 nSv/hr/MU following the Varian ProBeam definition of an MU.

The resulting intensity map from the sample UHDR PBS lung plan delivery was directly comparable to the ground truth profile. Such cumulative dose map validation is most representative of typical QA methods and is therefore the first metric considered for in vivo delivery verification. Figure 4 presents a comparison of the cumulative dose map derived from imaging the scintillation response on the curved chest phantom to a dose map from a separate delivery of the same plan onto flat film. Given that these were two distinct deliveries, we assumed that machine stability was maintained across successive irradiations based on an observed inter-delivery variation of about 2–3%, consistent with other Varian FLASH systems.³⁸ Gamma analysis was used to quantitatively analyze the monitoring quality of the scintillation imaging method. Decreasing percent and distance deviation thresholds were used to test the robustness of the imaging technique for dose monitoring and are also shown in Figure 4. A greater dependence on distance variation is observed, but generally high agreement is seen across all applied threshold values. Despite the high global agreement, local deviations are still present as is highlighted in the progression from 3%/3mm to 1%/1mm, particularly at the centers of high-intensity hot spots and in the lower regions between major spots.

Figure 4:

Cumulative dose map comparison of the imaged scintillation profile and a flat film delivery evaluated using gamma analyses ranging from 3% to 1% and from 3mm to 1mm. 3%/3mm, 2%/2mm, and 1%/1mm gamma maps are highlighted to showcase the regions of greatest disagreement.

Scintillation array surface imaging at 1 kHz also facilitates single-spot monitoring, offering insight into the accuracy of spot delivery and dwell locations compared to the treatment plan. Figure 5 displays the cumulative imaged dose map with an overlay showing the spot centroids for both planned and imaged datasets. Each spot stack is categorized into dwell positions (depicted as large circles) and travel positions (depicted as small circles), based on the proximity of centroids. Dwell positions, defined as consecutive spots with centroids within 0.5 mm of each other, remained consistent at 80 spots between the plan and imaging session. To assess centroid localization accuracy, the deviation between imaged and planned positions was quantified for each dwell point using Euclidean distance. The histogram of distance errors indicates a mean error of 0.32 mm and a standard deviation of 0.19 mm. Further analysis separated centroid distance errors into X and Y axes, indicating positive (right/up) and negative (left/down) deviations for each of the 80 spots.

Figure 5:

Temporal information used for delivery analysis, showing overlay of single spot distribution and travel path, along with a histogram of Euclidean centroid and X/Y single axis distance errors. PBS dose rate comparison between the simulated plan and the imaged scintillator emission is also shown with a dose rate area histogram providing quantitative coverage comparison showing high agreement of 40 Gy/s coverage with a percent error of 0.71%.

Taking advantage of the temporal resolution afforded by single-spot monitoring, dose and temporal information can be leveraged to determine the Pencil Beam Scanning (PBS) dose rate. Figure 5 also presents a comparison of the PBS dose rate map derived from simulation with the map obtained from scintillation array imaging. A high global agreement is seen in both maps with accentuated dose rates at the sides of the field compared to the center. For a more direct comparison, a PBS dose rate area histogram is provided, showing the fractional area of the total field affected by variation in local PBS dose rate. Consistently, both the planned and imaged dose rate maps exhibit a steep decrease in fractional area covered by increasing dose rate, which is expected given the low dose rate regions around the center of the delivery. Such analysis was extended to assess the fraction of the field achieving the commonly cited FLASH threshold of 40 Gy/s, allowing for direct comparison between the imaged field and the planned delivery. Specifically, the planned PBS dose rate distribution attained a dose rate of 40 Gy/s in 84.3% of the field, while the imaged map shows a coverage of 83.7%, resulting in a percent error of 0.71% between the two. The mean dose rate for the field was likewise calculated to be 191.3 Gy/s.

4. Discussion

This work presents the practical implementation of scintillator array imaging as a novel dosimetry technique for FLASH radiotherapy, demonstrating its application using a spatially and temporally complex UHDR treatment delivery onto an anthropomorphic phantom. We present the first extensive evaluation of treatment monitoring and verification metrics at the patient surface, providing a pathway for clinical implementation while validating both dosimetric accuracy and workflow integration.

Clinical translation of this system requires minimal adaptation of the existing treatment procedures. A phantom workflow test demonstrated minimal disruption to the typical operating procedure, with the additional time limited to array placement and contributing approximately one minute. The geometric calibration for stereo-vision matching follows procedures similar to surface-guided radiation therapy³⁹ and allows the array to achieve sub-millimeter localization accuracy in absolute coordinate space. As a result of the real-time stereo-vision imaging, the system is also set up to offer surface geometry guidance at up to 30 frames per second, providing potential to enhance treatment setup verification. Regarding patient safety considerations, irradiation of the scintillator array did not induce significant radiation activity, with post-irradiation surveys of the array reading a 671 nSv/hr. The array has a WET of 1.1 mm, and this could be accounted for during the treatment planning process. For the application in this study, treatment plans were created using transmission beams, so added range uncertainty is not clinically significant, given the total water-equivalent thickness of the patient in the beams-eye-view is less than 34 cm.

The proposed technique is the first to yield a full surface dose map from a delivery over a geometrically complex target, so no guidelines are currently in place for validating the delivery. Gamma analysis was selected given its prevalence for pre-treatment validation, and a method of threshold variation was proposed to identify regions of greatest deviation between the monitored delivery and assumed ground truth. Although the presented comparison only showed the analysis of the cumulative dose profile, the proposed method can also be applied at any arbitrary timepoint during the radiation delivery. In this manner, each spot can be independently verified if any gross dosimetry errors were identified, providing additional validation capability even in plans with UHDR delivery speeds. The imaging technique demonstrates a viable approach to monitoring cumulative dose delivery as evident by the high passing rate under strict gamma parameters. Furthermore, by increasing the gamma index, any present discrepancies are highlighted which may help guide clinical adjustments if these are identified at significantly low gamma criteria.

Together with the centroid positioning information, temporal dose data was used to determine the PBS dose rate. The planned and imaged dose rate distributions showed high agreement in the cumulative dose rate area histogram analysis at the 40 Gy/s FLASH threshold, with 0.71% error. During this set of experiments, it was noted that the measured delivery time differed slightly from the predicted delivery time using the previously validated simulation²³. This offset, on the order of ~7% (65 ms), is likely due to the variability in scanning time when moving relatively large distances between spots. There is also some inherent uncertainty in scanning nozzle current, which other groups have noted to be up to 5%³⁶, and could lead to slight variability in delivery times at each spot position. Given that the camera response was previously characterized with flat deliveries²³ and showed no deviation with continuous fields, this suggests an additional benefit of in vivo scintillation imaging. Additionally, no work has been done to evaluate the consistency of the PBS proton beam hold time when traveling between spots with sufficient spatial separation to cause the beam to temporarily turn off (for spot transitions ≥ 1 cm). If the delivery is imaged with a fixed and consistent frame rate, these beam parameters may be monitored directly, leading to more accurate dose rate evaluation.

The work presented here demonstrates a proof-of-concept implementation of a scintillation imaging system and workflow for in vivo surface dose monitoring. The system has shown to successfully verify spot positions and dose dynamics at the surface, but these measurements alone cannot guarantee accurate dose delivery at depth. Even with correct surface dosimetry, factors such as patient positioning uncertainties and anatomical changes can significantly impact the delivered dose distribution throughout the target volume. To address this limitation, future research will explore Monte Carlo depth projection methods to estimate dose volume histograms and dose-rate volume histograms based on measured surface dose dynamics. Furthermore, the current system uses a single camera to collect the scintillation signal which precludes it from monitoring delivery to optically blocked anatomy or regions with a surface-to-camera angle of >80 degrees.³⁷ Future iterations will introduce a multi-camera system to cover a wider range of viable imaging angles and increase the utility of the proposed system. Finally, the current validation was performed using transmission beams, where scintillator interactions occurred primarily in the plateau region without observable LET-dependent quenching. The LET of a 250 MeV pencil beam has been shown to remain at ~1 keV/um until ~25mm away from the Bragg peak deposition⁴⁰ which suggests that for the energy used in this work, the scintillator array is expected to behave linearly at the point of measurement, and is observed to do so. For applications involving Bragg-peak flash delivery or conventional pencil beam scanning with varying energies, the impact of ionization quenching effect should be considered.^24,41,42

5. Conclusion

This study proposes a novel implementation of scintillation imaging for in vivo monitoring of ultra-high dose rate deliveries of PBS proton therapy. Imaging and correcting the emission from a deformable scintillator array throughout the treatment duration allows for the collection of cumulative dose and temporally distributed spot and dose rate profiles. Compared to other proposed dosimetry methods, this technique is the first to offer sub-0.5mm spatial and 1ms temporal resolution, while reporting on the complete delivery field instead of pre-determined locations with point readouts. There is a growing interest in UHDR therapy within the radiation therapy community, which is only increasing the importance and need for devices that can provide in vivo monitoring and delivery validation with such a vast range of analysis metrics. This method has the potential to offer a highly efficient way of recovering surface dosimetry with minimal impact on clinical time and proves to be particularly beneficial for treatments with complex spatiotemporal profiles and strict delivery requirements.

Data Availability Statement:

Research data will be shared upon reasonable request to the corresponding author.