Early View
RESEARCH ARTICLE
Open Access

Performance characterization of a novel hybrid dosimetry insert for simultaneous spatial, temporal, and motion-included dosimetry for MR-linac

Prescilla Uijtewaal

Corresponding Author

Prescilla Uijtewaal

Department of Radiotherapy, University Medical Center Utrecht, Utrecht, The Netherlands

Correspondence

Prescilla Uijtewaal, Department of Radiotherapy, University Medical Center Utrecht, Heidelberglaan 100, 3584 CX Utrecht, The Netherlands.

Email: [email protected]

Search for more papers by this author
Pim Borman

Pim Borman

Department of Radiotherapy, University Medical Center Utrecht, Utrecht, The Netherlands

Search for more papers by this author
Benjamin Cote

Benjamin Cote

Medscint, Québec, Quebec, Canada

Search for more papers by this author
Yoan LeChasseur

Yoan LeChasseur

Medscint, Québec, Quebec, Canada

Search for more papers by this author
François Therriault-Proulx

François Therriault-Proulx

Medscint, Québec, Quebec, Canada

Search for more papers by this author
Rocco Flores

Rocco Flores

Modus QA, London, Ontario, Canada

Search for more papers by this author
Stephanie Smith

Stephanie Smith

Modus QA, London, Ontario, Canada

Search for more papers by this author
Grant Koenig

Grant Koenig

Modus QA, London, Ontario, Canada

Search for more papers by this author
Bas Raaymakers

Bas Raaymakers

Modus QA, London, Ontario, Canada

Search for more papers by this author
Martin Fast

Martin Fast

Department of Radiotherapy, University Medical Center Utrecht, Utrecht, The Netherlands

Search for more papers by this author
First published: 13 December 2023

Abstract

Background

Several (online) adaptive radiotherapy procedures are available to maximize healthy tissue sparing in the presence of inter/intrafractional motion during stereotactic body radiotherapy (SBRT) on an MR-linac. The increased treatment complexity and the motion-delivery interplay during these treatments require MR-compatible motion phantoms with time-resolved dosimeters to validate end-to-end workflows. This is not possible with currently available phantoms.

Purpose

Here, we demonstrate a new commercial hybrid film-scintillator cassette, combining high spatial resolution radiochromic film with four time-resolved plastic scintillator dosimeters (PSDs) in an MRI-compatible motion phantom.

Methods

First, the PSD's performance for consistency, dose linearity, and pulse repetition frequency (PRF) dependence was evaluated using an RW3 solid water slab phantom. We then demonstrated the MRI4D scintillator cassette's suitability for time-resolved and motion-included quality assurance for adapt-to-shape (ATS), trailing, gating, and multileaf collimator (MLC) tracking adaptations on a 1.5 T MR-linac. To do this, the cassette was inserted into the Quasar MRI4D phantom, which we used statically or programmed with artificial and patient-derived motion. Simultaneously with dose measurements, the beam-gating latency was estimated from the time difference between the target entering/leaving the gating window and the beam-on/off times derived from the time-resolved dose measurements.

Results

Experiments revealed excellent detector consistency (standard deviation ⩽ 0.6%), dose linearity (R2 = 1), and only very low PRF dependence (⩽0.4%). The dosimetry cassette demonstrated a near-perfect agreement during an ATS workflow between the time-resolved PSD and treatment planning system (TPS) dose (0%–2%). The high spatial resolution film measurements confirmed this with a 1%/1-mm local gamma pass-rate of 90%. When trailing patient-derived prostate motion for a prostate SBRT delivery, the time-resolved cassette measurements demonstrated how trailing mitigated the motion-induced dose reductions from 1%–17% to 1%–2% compared to TPS dose. The cassette's simultaneously measured spatial dose distribution highlighted the dosimetric gain of trailing by improving the 3%/3-mm local gamma pass-rates from 80% to 97% compared to the static dose. Similarly, the cassette demonstrated the benefit of real-time adaptations when compensating patient-derived respiratory motion by showing how the TPS dose was restored from 2%–56% to 0%–12% (gating) and 1%–26% to 1%–7% (MLC tracking) differences. Larger differences are explainable by TPS-PSD coregistration uncertainty combined with a steep dose gradient outside the PTV. The cassette also demonstrated how the spatial dose distributions were drastically improved by the real-time adaptations with 1%/1-mm local gamma pass-rates that were increased from 8 to 79% (gating) and from 35 to 89% (MLC tracking). The cassette-determined beam-gating latency agreed within ⩽12 ms with the ground truth latency measurement. Film and PSD dose agreed well for most cases (differences relative to TPS dose <4%), while film-PSD coregistration uncertainty caused relative differences of 5%–8%.

Conclusions

This study demonstrates the excellent suitability of a new commercial hybrid film-scintillator cassette for simultaneous spatial, temporal, and motion-included dosimetry.

1 INTRODUCTION

To maximize healthy tissue sparing in the presence of interfractional or intrafractional motion during stereotactic body radiation therapy (SBRT), several (online) adaptive radiotherapy techniques have been developed. Interfractional anatomical variation can be compensated for by performing a prebeam plan adaptation based on a daily position verification scan.1, 2 Intrafractional motion can be mitigated by online adaptations such as gating,3 tumor tracking,4, 5 and trailing.6 The introduction of the MR-linac, a linear accelerator with integrated MRI scanner, increases the availability of these adaptive radiation treatments.7, 8 The treatment complexity and motion-delivery interplay associated with adaptive strategies raise the demand for MR-compatible moving phantoms with integrated time-resolved dosimeters to validate the online adaptive end-to-end workflows.9, 10

Currently available phantoms used for patient QA have several limitations, which complicate the validation of online plan adaptations on the MR-linac. They typically lack a motion component, MR-compatibility, or their integrated dosimeters often contain metal11, 12 preventing real-time imaging. Most MR-compatible motion phantoms use radiochromic film as dosimeter, because of its high water equivalence in the MV photon range, high spatial resolution, 2D dose mapping, and MR-compatibility.13 However, film dosimetry lacks temporal resolution. These deficiencies drastically limit the patient QA for the current clinical online adaptive treatments (e.g., gating and trailing), and hampers the introduction of newly developed adaptive strategies (e.g., multileaf collimator [MLC] tracking14). A promising alternative dosimeter suitable for use in an MR-linac is a plastic scintillation dosimeter (PSD). A PSD contains a scintillator that emits optical photons proportional to the absorbed energy. An optical fiber guides this photon flux to optical readout equipment for dose tallying.15-18 The biggest challenge for PSDs are dosimetric inaccuracies caused by the background signal or stem effect that is produced during irradiation of the optical fiber. This effect is due to Cerenkov and fluorescence emission,17, 19, 20 and its intensity is affected by strong magnetic fields.21-24 Accounting for both the Cerenkov and fluorescence emissions independently using a hyperspectral approach, effectively removes the stem effect.24 In a recent study, we demonstrated that the HYPERSCINT RP200 PSD provides excellent time-resolved dose readouts in a 1.5 T Unity MR-linac, and that its performance is not affected by motion, the magnetic field, or by MRI scanning.25 However, this characterization was performed using a single PSD, providing insufficient volume coverage to validate adaptive radiotherapy workflows in an end-to-end fashion.

To simultaneously harness the spatial resolution of radiochromic film, the temporal resolution and dosimetric accuracy of PSDs, and include a motion component, we developed the novel MR-compatible MRI4D scintillator cassette together with Modus QA (Modus Medical Devices Inc., London ON) and Medscint (Medscint, Quebec City QC). Both radiochromic film26 and PSDs15, 19 have a high degree of water equivalence in the MV photon range, avoiding dosimetric disruptions or interplay. The new dosimetry cassette seamlessly integrates with the Modus QA Quasar MRI4D motion phantom. In addition, the MR-compatible phantom contains a trackable target that can be programmed to move in cranial-caudal (CC) direction.

The purpose of this study was to demonstrate the performance of the new commercial MRI4D scintillator cassette in a 1.5 T MR-linac, combining radiochromic film and four time-resolved PSDs in a motion phantom. First, we characterized the performance of the individual PSDs in the MRI4D scintillator cassette in terms of repeatability, dose linearity, and pulse repetition frequency (PRF) dependence to demonstrate if a cassette with four PSDs performs as well as the previously validated single PSD.25 We then combined the cassette with the Quasar phantom to demonstrate its suitability for simultaneous spatial, time-resolved, MR-compatible, and motion-included QA in a range of (online) adaptive radiotherapy scenarios including the quantification of beam gating latency.

2 MATERIALS AND METHODS

All experiments were performed on a Unity MR-linac (Elekta AB, Stockholm, Sweden), featuring a 1.5 T MR scanner and a 7-MV flattening filter free (FFF) linac with its beam axis perpendicular to the main magnetic field. To facilitate online adaptive treatments, the Unity system was equipped with Elekta's comprehensive motion management (CMM) solution. CMM continuously monitors the 3D tumor motion using 2D-interleaved sagittal and coronal balanced cine MRI (acquisition voxel size = 3.66 × 3.66 × 5.00 mm3). The image update rate of the cine was 6 Hz, but because they are interleaved, the update rate per plane is 3 Hz. We used two different experimental setups containing either an RW3 solid water slab phantom or the Quasar MRI4D phantom.

2.1 MRI4D scintillator cassette

The newly developed MRI4D scintillator cassette can be inserted in the Quasar phantom's movable cylinder (Figure 1). Both the cassette and the cylinder are made from acrylic, providing a high degree of water equivalence in the MV photon range.27 When the cassette is inserted into the Quasar phantom, it exactly intersects the phantom's ∅3-cm trackable spherical target. The cassette consists of two rectangular plates (thickness = 2 and 4 mm) that are connected with a hinge. A radiochromic film for film dosimetry can be placed in the narrow space between the plates. Small pins in three of the cassette's corners create small indents in the film to retrospectively co-register different films and to accurately pinpoint the target position on the film. Four PSDs (size = 1 × 1 × 1 mm3) are embedded in the thicker 4-mm cassette plate at fixed positions. The thickness of the plates positions one PSD exactly at the center of the phantom's trackable spherical target. The three other PSDs are positioned in the longitudinal direction at 15, 18, and 21 mm from the central PSD exactly on the edge of the gross tumor volume (GTV), the planning target volume (PTV; assuming a 3-mm GTV-to-PTV margin), and 3 mm outside the PTV. The film is positioned 1 mm anteriorly. The PSDs are connected by 20-m long individual optical fibers to Medscint's HYPERSCINT RP200 scintillation dosimetry platform. During all experiments, the time-resolved dose measurements were performed at the platform's maximal sampling frequency (Fs) of 14.8 Hz.

Details are in the caption following the image
(a) HYPERSCINT RP200 scintillation dosimetry platform with MRI4D scintillator cassette connected. (b) Inside of the MRI4D scintillator cassette, showing the positions of the PSDs in the cassette relative to the GTV. (c) Dosimetry cassette integrated into the Quasar MRI4D phantom positioned in the MR-linac bore.

2.2 Calibration procedure

2.2.1 Film dosimetry calibration

During the experiments, we used radiochromic EBT3 (Ashland Inc, Wayne, NJ, USA) films with a dynamic dose range of 0.1–20 Gy. To avoid air pockets around the film, we added a thin layer of water in the MRI4D scintillator cassette under and above the film. To analyze the films, they were digitized with a 0.16 × 0.16 mm2 resolution using an Epson Expression 10000XL flatbed scanner (Seiko Epson Corp, Nagano, Japan) with a resolution of 150 dpi. The digitalized films were processed using in-house developed software performing a triple-channel dosimetry analysis with lateral corrections.28, 29

2.2.2 PSD calibration

To calibrate the MRI4D scintillator cassette, we first used an Orthovolt (Xstrahl 200 X-Ray Therapy System) with a 150-kV beam energy and delivered 105 MU to separate scintillation and fluorescence light sources. The scintillation spectrum was identified by irradiating the PSDs directly in air, while the fluorescence emission was identified by only irradiating the optical fibers in air. To correct the Cerenkov emission, we taped the optical fibers in crossline direction on a 10-cm high RW3 solid water phantom. The optical fibers were irradiated from gantry angles 45° and 315° using a 7-MV beam and 500 MU in the 1.5 T Unity MR-linac. The scintillation signal was then normalized to 500 cGy from gantry angle 0° in the MR-linac using an RW3 solid water slab phantom.

2.3 Linearity, PRF dependence, and repeatability

The dose linearity, repeatability, and PRF dependence of the four individual PSDs in the MRI4D scintillator cassette were evaluated in the 1.5 T MR-linac using an RW3 solid water slab phantom. The MRI4D scintillator cassette was positioned between the solid water slabs with the PSDs at a depth of 5 cm and a source axis distance (SAD) of 143.5 cm (Figure 2a). The phantom was positioned in the machine with detector 1 at the machine isocenter. Because the PSDs are fixated in the MRI4D scintillator cassette, the other three detectors were consequently located more caudally from the machine isocenter. The dose linearity was verified by comparing the detector response for measurements of 10–1000 MU from gantry 0° from a 10× 10 cm2 field using the maximum dose rate (425 MU/min with a PRF of 275 Hz). These measurements were repeated 10 times to evaluate the repeatability. The PRF dependence was verified by delivering 100 MU at varying PRFs of 32–260 Hz. The dose per pulse was not varied.

Details are in the caption following the image
Experimental phantom setups with MRI4D scintillator cassette. (a) RW3 slab phantom containing the MRI4D scintillator cassette at a 5-cm depth. (b) MRI4D scintillator cassette integrated into the Quasar phantom positioned on a cradle. (c) Quasar phantom with integrated MRI4D scintillator cassette on a 20° ramp introducing additional AP motion. Here, the couch is shown retracted relative to the isocenter position used for measurements.

To evaluate the repeatability of measurements in the Quasar phantom, we positioned the MRI4D scintillator cassette in the Quasar phantom. The phantom was positioned feet first-supine in a cradle in the bore such that detector 1 was positioned at the machine isocenter and the off-center detectors were positioned at the caudal target periphery. We used a typical 15-beam IMRT 8 × 7.5-Gy lung SBRT plan with a ∅3 cm target (GTV) and a 3-mm GTV-to-PTV margin. We repeated the measurement five times and between measurements we took the cassette out of the phantom, repositioned the film, and inserted it back into the phantom. This reflects the standard procedure to replace the irradiated radiochromic film. The PSD readings between the different repetitions were compared to evaluate the effect of potential repositioning differences.

2.4 Adaptive radiotherapy validation

To evaluate the spatial, time-resolved, MR-compatible, and motion-included dosimetric performance of the MRI4D scintillator cassette, we used the cassette during several (online) adaptive radiation treatments. During these experiments, the MRI4D scintillator cassette was placed inside the Quasar phantom (Figure 2b,c).

2.4.1 Adapt-to-shape

We first evaluated the spatial and MR-compatibility performance of the MRI4D scintillator cassette during our clinical adapt-to-shape (ATS) procedure. An ATS procedure corrects for interfractional anatomical variation.1 First, a position verification 3D MRI (T2-weighted, acquisition voxel size = 1.3 × 1.3 × 2.0 mm3) was acquired (setup as shown in Figure 2b). The contours of the original treatment plan were propagated onto the 3D MRI, taking into account any positional changes or deformations. Based on these new, visually checked contours and the original treatment plan, an updated treatment plan was calculated by the treatment planning system (TPS) Monaco 6.20 (Elekta AB, Stockholm, Sweden). We used a 15-beam IMRT 8× 7.5-Gy lung SBRT plan with a 3-mm GTV-to-PTV margin. To introduce anatomical variation, we moved the Quasar's target 1 cm caudally.

2.4.2 Intrafractional drift correction/Trailing

Then we tested the simultaneous spatial, time-resolved, MR-compatibility, and motion-included performance of the MRI4D scintillator cassette during a prostate IMRT treatment with constant GTV motion and multiple online baseline motion corrections. To correct for baseline motion during treatment, the Unity MR-linac is equipped with an intrafractional baseline motion correction feature (i.e., trailing6). During treatments, the tumor motion is continuously monitored by the CMM system. When baseline motion is observed, an intrafractional motion correction can be manually triggered to apply segment aperture morphing to the treatment plan based on the detected baseline shift. To validate the comprehensive QA performance of the MRI4D scintillator cassette during these corrections, we used a 7-beam 5 × 7.25-Gy prostate plan with anisotropic GTV-to-PTV margins (left-right and CC = 2 mm, anterior–posterior = 3 mm), following our clinical template. To introduce both CC and anterior–posterior (AP) motion, the Quasar phantom was positioned on a 20° ramp (Figure 2c). We programmed the phantom with artificial linear drift motion (drift = 2 mm/min) and patient-derived prostate motion ( d r i f t ¯ $\overline{drift}$ = 0.4 mm/min). The ramp decomposed the artificial motion into CC = 1.9 mm/min and AP = 0.7 mm/min and the patient-derived prostate motion was decomposed into an average drift of C C ¯ $\overline{CC}$ = 0.4 mm/min and A P ¯ $\overline{AP}$ = 0.1 mm/min. An intrafractional drift correction was applied each time ≥10% of the GTV drifted out of the PTV.

2.4.3 Gating

To further validate the MRI4D scintillator cassette's spatial, time-resolved, MR-compatibility, and motion-included performance, we used the cassette during a lung IMRT treatment with respiratory beam gating. We performed a free-breathing gating delivery and limited the irradiation of the target to end-exhale. The Quasar phantom was programmed with Lujan CC motion (cos4, Apeak-to-peak = 20 mm, T = 4 s) and patient-derived respiratory CC motion ( A peak to peak ¯ $\overline{{A}_{\textit{peak}-\textit{to}-\textit{peak}}}$ = 18 mm, T ¯ $\bar{T}$ = 5 s, drift ¯ $\overline{\textit{drift}}$ = 0.01 mm/min). We defined a gating window using a displacement threshold (Figure 3), such that the beam was automatically gated when the target motion exceeded the threshold. The motion-specific displacement thresholds created a ∼50% duty cycle. To dosimetrically validate the performance of the MRI4D scintillator cassette, we used the 15-beam IMRT 8 × 7.5-Gy lung SBRT plan described above.

Details are in the caption following the image
Beam-on and beam-off latency determination. (a) The blue blocks indicate when dose was measured by the dosimetry cassette, and the red vertical lines indicate when the target crosses the gating threshold. The time differences indicate the beam-on/beam-off latency. (b) Determination of exact beam-on/beam-off time using the summed dose of the first/last data point in the maximum dose plateau and the four previous/next data points. The ratio of the summation to the average maximum dose indicates how much earlier/later the beam was turned on/off.

In addition to a dosimetric evaluation of gating, the MRI4D scintillator cassette's time-resolved measurements could also be used simultaneously to evaluate the beam-on and beam-off system gating latency. To determine the system latency, we used the phantom reported positions (2 kHz) and the PSD time-resolved dose measurements (14.8 Hz). Both the phantom control software and the MRI4D scintillator cassette software ran on the same computer such that phantom positions and the dose measurements were logged and timestamped by the same computer clock. If the phantom is connected directly to the beam-generator, the phantom could simultaneously report the phantom positions and the beam-on/off signal (latency <1 ms). When logging the PSD-measured dose and the beam-on/beam-off phantom signal during the same gating measurement, a one-time cross-correlation could be performed between the two signals, such that the phantom and PSD time axis could be perfectly aligned. The cross-correlation between the beam-on/beam-off signal of the phantom signal and the PSDs resulted in a 40.5-ms shift of the PSD time axis. Based on the phantom reported positions, we retrospectively determined when the target crossed the gating threshold thus entering or leaving the gating window (Figure 3a). The time-resolved dose measurements indicated the beam-on/beam-off times by measuring exactly when dose was delivered per gating interval. Because the MRI4D scintillator cassette measures dose at 14.8 Hz, beam on/off gating is likely to result in partial dose in at least one dose sampling interval. To determine precisely when the beam was turned on/off, we first selected the timestamps where the maximum dose plateau per beam-on interval is reached first and last (Figure 3b). Then the dose measured at the first timestamp and at the four previous timestamps was summed to determine the exact beam-on time and the dose measured at the last timestamp and the following four timestamps were summed to determine the exact beam-off time. The ratio of the summed dose to the mean maximum dose of that beam-on interval allowed us to infer the exact beam on/off time point. Note that this assumption only holds since the Unity linac reaches its maximum output within 10 ms. Because for an IMRT plan, not every segment covers all PSDs at all times, and because a PSD could be moved out of the beam as a result of the respiratory phantom motion, we included all four PSDs in the analysis and selected per gating interval the first timestamp for beam-on time and the last timestamp for beam-off time. The time difference between the crossing of the displacement threshold and the start/end of the dose measurement during a gating interval is the beam-on/beam-off latency. To evaluate the PSD-derived beam-gating latency, we compared the derived latency to the ground-truth measurement of latency directly derived from the beam generator.

2.4.4 MLC tracking

Lastly, we evaluated the spatial, time-resolved, MR-compatible, and motion-included performance of the MRI4D scintillator cassette during a lung IMRT treatment combined with MLC tracking. Here the MLC aperture was continuously reshaped to realign the treatment beam with the target positions derived from cine MRI using research software.5, 14 The Quasar phantom was programmed with Lujan CC motion (cos4, Apeak-to-peak = 20 mm, T = 4 s, drift = 1 mm/min) and patient-derived respiratory CC motion ( A peak to peak ¯ $\overline{{A}_{\textit{peak}-\textit{to}-\textit{peak}}}$ =11 mm, T ¯ $\bar{T}$ = 3 s, drift ¯ $\overline{\textit{drift}}$ = 0.3 mm/min). The system latency was compensated by a ridge linear regression predictor.5 We again used the 8 × 7.5-Gy lung SBRT plan.

2.4.5 Doses evaluation

To evaluate the comprehensive dosimetric performance of the MRI4D scintillator cassette during the various (online) treatment adaptations, the dose measurements were compared to a static reference delivery and to the TPS dose. The TPS dose was calculated with the minimum TPS grid spacing of 1 mm and 0.5% statistical uncertainty per control point to minimize discretization artefacts. In addition, we increased the TPS dose resolution to a 0.5-mm grid spacing by performing a 3D cubic-spline interpolation over the 3D dose map during post-processing. The TPS dose was registered to the corresponding static film-measured dose map using intensity-based registration to validate the similarity of the TPS dose and the film dose. The TPS dose was registered to the PSD's dose using the position verification 3D MRI and the PSD's geometric positions in the Quasar phantom. PSD 1 was positioned exactly at the center of the GTV. This position is traceable in the TPS dose by determining the center of the GTV in the corresponding 3D MRI. Since the distance between the different PSDs is fixed, the TPS dose for the other PSDs can also be determined. The film measurements with and without treatment adaptation were registered to the corresponding static reference film to demonstrate the effect of adaptive radiation treatments. The registration was performed using a semi-automatic registration with a point-matching algorithm to register the films based on the three corner indents created by the MRI4D scintillator cassette. A 1%/1 mm and 3%/3-mm local gamma-analysis evaluated the correspondence between static and adaptive radiation deliveries. Only dose values >10% of the prescribed dose were included in the analysis to reduce calibration-induced uncertainties. The GTV and PTV coverage was quantified using dose area histograms (DAHs). Small coregistration uncertainties (⩽0.6 mm5) between the different dose modalities were expected, which might induce dose deviations when comparing the PSD point doses between the different modalities. For each PSD in both the film and TPS dose maps, we extracted the dose gradient in a 1-mm range around the registered PSD positions to express a dosimetric uncertainty range.

3 RESULTS

3.1 Linearity, PRF dependence, and repeatability

The measured dose as a function of prescribed MU gives excellent dose linearity (R2 = 1.0) for all four detectors. The dose delivered per MU was very similar for the different prescriptions with a relative standard deviation per PSD < 0.5%. The standard deviation of the repeatability measurement was similar for all detectors and ranged between 0.1%-0.4% for the different dose prescriptions. All four PSDs have a PRF dependence ⩽0.4%. The detector consistency was similar for the different PRFs with a relative standard deviation over five repetitions of 0.0%-0.3%. Replacing film between five 15-beam IMRT measurements yielded a relative standard deviation of 0.6% for detector 4 and 0.1%–0.2% for the other PSDs.

3.2 Adaptive radiotherapy validation

Figure 4 demonstrates the spatial performance of the MRI4D scintillator cassette for the different adaptive radiation treatments. The simultaneously measured PSD, film, and TPS dose at the PSD positions are summarized in Table 1, including the corresponding film dose gradients. The cassette's spatial performance is further described by the local gamma pass-rates and by the DAH GTV and PTV dose in Table 2. Figure 5 shows the time-resolved PSD measurements during the different online adaptive treatments.

TABLE 1. PSD and film dose as percentage difference relative to the TPS dose to demonstrate the dosimetric differences measured by the hybrid dosimetry insert with and without adaptations. Dose gradients [%/mm] indicate the dosimetric uncertainty for the film dose. The TPS gradient uncertainty is similar to film uncertainty.
PSD 1 PSD 2 PSD 3 PSD 4
TPS (cGy) PSD (Δ%) Film (Δ%) TPS (cGy) PSD (Δ%) Film (Δ%) TPS (cGy) PSD (Δ%) Film (Δ%) TPS (cGy) PSD (Δ%) Film (Δ%)
ATS
ATS 1054 1 −1 [0%/mm] 943 −1 −2 [4%/mm] 775 0 −1 [14 %/mm] 445 −2 0 [33 %/mm]
No ATS - 1 −4 [0%/mm] - −39 −36 [19%/mm] - −63 −62 [36 %/mm] - −74 −79 [36 %/mm]
Trailing
Static 732 0 −1 [0 %/mm] 713 0 0 [0%/mm] 695 1 0 [2%/mm] 661 −1 −2 [4%/mm]
Drift motion
No trailing - 1 −4 [0 %/mm] - −11 −18 [4 %/mm] - −29 −32 [11 %/mm] - −53 −53 [24 %/mm]
Trailing - 2 1 [0 %/mm] - −3 −3 [1 %/mm] - −8 −7 [4 %/mm] - −22 −20 [12 %/mm]
Prostate motion
No trailing - 1 −1 [0 %/mm] - −3 −6 [1 %/mm] - −7 −10 [4 %/mm] - −17 −18 [7 %/mm]
Trailing - 2 0 [0 %/mm] - 1 −3 [1 %/mm] - 1 −3 [1 %/mm] - −2 −5 [4 %/mm]
Gating
Static 1029 0 0 [0%/mm] 941 0 −1 [3%/mm] 824 1 −1 [6%/mm] 578 0 1 [17%/mm]
Lujan cos4 motion
No gating - −4 −7 [1%/mm] - −38 −39 [5%/mm] - −43 −42 [8%/mm] - −45 −40 [14%/mm]
Gating - 0 −2 [0%/mm] - −1 −2 [3%/mm] - −1 −2 [6%/mm] - −1 2 [15%/mm]
Respiratory motion
No gating - −2 −6 [1%/mm] - −44 −43 [8%/mm] - −51 −50 [10%/mm] - −56 −52 [17%/mm]
Gating - 0 0 [0%/mm] - −5 −5 [4%/mm] - −8 −12 [9%/mm] - −12 −11 [16%/mm]
MLC tracking
Static 1050 1 1 [1%/mm] 950 0 1 [2%/mm] 850 1 1 [6%/mm] 579 0 1 [22%/mm]
Lujan cos4 motion
No tracking - −1 −2 [0%/mm] - −6 −8 [3%/mm] - −7 −8 [4%/mm] - 13 21 [3%/mm]
Tracking - 2 1 [0%/mm] - 1 −3 [3%/mm] - −1 −7 [9%/mm] - −7 −11 [21%/mm]
Respiratory motion
No tracking - 1 −1 [0%/mm] - −13 −16 [3%/mm] - −24 −27 [7%/mm] - −26 −23 [16%/mm]
Tracking - 2 1 [0%/mm] 1 −3 [7%/mm] −1 −7 [10%/mm] 1 −5 [13%/mm]
TABLE 2. Gamma pass-rates and DAH dose comparisons that demonstrate MRI4D scintillator cassette's spatial dose performance with and without adaptations. Dose values are expressed relative to their respective prescription dose.
Gamma pass-rate DAH GTV DAH PTV
(1%/1 mm) (3%/3 mm) D98/% (cGy) D50% (cGy) D2% (cGy) D95% (cGy)
ATS
ATS 118 132 139 108
TPS 90 100 116 132 140 107
No ATS 16 52 93 92 136 76
Trailing
Static 97 100 103 95
TPS 84 98 97 101 104 95
Artificial drift motion
No trailing 12 37 86 95 98 82
Trailing 26 95 97 101 104 95
Prostate motion
No trailing 41 80 93 89 101 90
Trailing 68 97 94 99 105 93
Gating
Static 117 131 137 107
TPS 92 100 115 131 137 105
Lujan (cos4) motion
No gating 9 31 82 122 134 76
Gating 86 100 118 130 136 108
Respiratory motion
No gating 8 26 78 124 133 71
Gating 79 100 120 132 139 107
MLC tracking
Static 122 135 142 111
TPS 92 100 120 134 140 107
Lujan (cos4) motion
No tracking 9 48 74 126 139 67
Tracking 93 100 121 135 142 110
Respiratory motion
No tracking 35 84 112 130 139 98
Tracking 89 100 122 135 142 109
Details are in the caption following the image
Dose profiles and dose gradients for different treatment adaptations extracted from the measured film dose at the line of the registered PSD positions. The dose gradients show the dose difference (%) per millimeter cc-shift in the dose line profile. The lowest 5% of the dose was excluded to reduce film calibration uncertainties. The gray and red areas depict the GTV and PTV positions relative to the profiles. Circular markers represent the dose measured by the PSDs. Partial circles indicate overlapping measurements. Note that the treatment plan for trailing differs (5× 7.25-Gy prostate SBRT plan) from the other panels (8× 7.5-Gy lung SBRT plans).
Details are in the caption following the image
Time-resolved PSD results of the different online motion adaptations during treatments with patient-derived motion. Each column represents a PSD. The arrows in the trailing panel indicate when a correction was applied.

3.2.1 Adapt-to-shape

Figure 4a summarizes the film and PSD performance of the MRI4D scintillator cassette during an ATS procedure. The measured profiles demonstrate that applying an ATS procedure aligns the dose with the TPS dose. Since an ATS procedure is a pre-beam treatment adaptation, no motion compatibility was tested here and the same profile describes the static and ATS profile (Figure 4a). Both the cassette's PSD and film dose agree well with the TPS dose for a static delivery (Table 1). Comparing the PSD-measured dose and the film-measured dose to the TPS dose yielded dose differences of 0.4%–1.8% and 0.4%–1.6% respectively. The MRI4D scintillator cassette reliably detected the dose difference when no ATS was performed, with dose differences of −1, −39, −63, and −74% for, respectively, PSD 1-4. PSD and film dose had a good agreement relative to the TPS dose (1%–5%), whereby the largest relative difference (5%) in PSD 4 might be due to a coregistration error in the steep dose gradient (36%/mm).

The MRI4D scintillator cassette measurements also demonstrate an excellent agreement between the statically measured spatial film dose and the TPS dose with a local 1%/1-mm gamma pass-rate of 90% (Table 2), and the DAH results show similar GTV and PTV coverage with dose differences ⩽1.7% for the D98%, D2%, D2% GTV dose and D95% PTV dose. The cassette also demonstrates that without ATS procedure, the local 1%/1-mm gamma pass-rate was only 16%, and the DAH D98%, D95%, D2% GTV dose and D95% PTV dose were reduced by 21, 31, 2, and 30%, respectively.

3.2.2 Intrafractional drift correction/Trailing

To expand the validation of the MRI4D scintillator cassette, we tested its simultaneous spatial, time-resolved, MR-compatibility, and motion-included performance during a prostate IMRT treatment combined with trailing to compensate for intrafractional drift motion. To ensure that the GTV would not drift more than 10% out of the PTV, we applied five drift corrections per delivery to compensate for linear 2-mm/min drift motion or for patient-derived prostate motion. The dose profiles in Figure 4b show the spatial and PSD results measured by the MRI4D scintillator cassette. The cassette perfectly demonstrates how trailing improves the PSD dose during linear drift motion from dose differences (Table 1) of 1, −11, −29, and −53% compared to the TPS dose to differences of only 2, −3, −8, and −22% for PSD 1-4. More subtle differences during patient-derived prostate motion were also measured by the PSDs, whereby dose differences were reduced from 1, −3, −7, and −17 to 2, 1, 1, and −2%. The film and PSD dose agreed well with dose differences relative to the TPS dose of 0%–3%. Only for PSD 2 when no trailing was applied to reduce artificial drift motion, a larger relative difference was found (7%).

The MRI4D scintillator cassette's spatial dose measurements (Table 2) confirm its PSD results by demonstrating that trailing improved a local 3%/3-mm gamma pass-rate from 37 and 80% to 95 and 97% for, respectively, linear 2-mm/min drift motion and patient-derived prostate motion. In addition, it demonstrated that trailing reduced the D95% PTV dose difference between the static reference from 14 to 0% for the linear motion and from 5 to 2% for the patient-derived motion.

The top panel in Figure 5 shows that time-resolved PSD measurements perfectly measure how trailing restored the delivered dose to the TPS dose, without being disturbed by the repositioning of the beam (segment aperture morphing) applied during trailing. The staircase pattern shows when the drift corrections were applied, as the beam was gated during the recalculation of the treatment plan.

3.2.3 Gating

The MRI4D scintillator cassette's simultaneous spatial, time-resolved, MR-compatibility, and motion-included performance was further evaluated during a lung IMRT treatment combined with respiratory beam gating. Figure 4c shows the spatial performance of the dosimetry cassette during a gating treatment. The cassette demonstrates that gating perfectly restored the dose to the TPS dose. This is also measured by the PSDs (Table 1), which measured delivered dose that was very comparable to the TPS dose (differences ⩽2%) when gating Lujan motion, while it clearly demonstrates that the delivered dose markedly deteriorates to dose differences of 38%–45% in PSDs 2–4 without gating. This was also demonstrated for patient-derived respiratory motion, where PSD-TPS dose differences in PSDs 2–4 were 44%–56% without gating and were only 5%–12% with gating. PSD 1 was dosimetrically hardly affected by the motion since it moved through a relatively flat dose plateau (Table 1). The PSD and film dose agree well with dose differences relative to the TPS dose of 0%–5%. The largest relative difference of 5% was found in detector 4 when no gating was applied during Lujan motion due to the steep dose gradient in the film (14%/mm) at the PSD position.

The spatial dose measurements of the MRI4D scintillator cassette also demonstrate well how much gating improved the dose distribution with 3%/3-mm local gamma pass-rates (Table 2) improved from 31 to 100% for Lujan motion and from 26 to 100% for patient-derived respiratory motion. Additionally, the dosimetric cassette measured the DAH D98%, D50%, D2% GTV dose and the D95% PTV dose and showed that the differences compared to static were improved from −33, −5, −3, and −33% to 4, 1, 1, and 1% for patient-derived respiratory motion. These measurements were similar during Lujan motion.

The time-resolved PSD measurements (Figure 5), could also be used to derive the beam gating latency, shown in Figure 6. The MRI4D scintillator cassette estimated an average beam-on and beam-off latency of 74.6 ± $-74.6\pm$ 77.9 ms and of 80.0±71.2 ms for the Lujan motion. This estimation differed only by 10.0±30.9 ms and 12.8 ± $-12.8\pm$ 22.7 ms from the average ground truth latency. For the patient-derived motion, the average latency was 20.5±129.5 ms and of 82.7±103.3 ms with a difference of 5.2 ± $-5.2\pm$ 35.9 ms and 12.1 ± $-12.1\pm$ 24.7 ms.

Details are in the caption following the image
Beam gating latency based on the MRI4D scintillator cassette's PSDs and based on the beam pulses.

3.2.4 MLC tracking

Lastly, the dosimetry performance of the MRI4D scintillator cassette was evaluated during respiratory MLC tracking. The measured dose profiles in Figure 4d show that the dosimetry cassette clearly demonstrates the difference between treatments with and without MLC tracking. This was also reflected in the PSDs measurements (Table 1) that show how MLC tracking restored the dose to the TPS dose with dose differences of 1%–7% for Lujan motion and it could measure even more subtle differences of 0.5%–2% for patient-derived respiratory motion. Without motion compensation, the PSDs clearly demonstrate how the dose deteriorates with dose differences in PSD 2–4 of 6%–13% and 13%–26%. The time-resolved results in Figure 5 also show how the PSDs can accurately measure the dose during MLC tracking, despite the continuous motion of both the GTV and the treatment beam. Compared to gating, they also illustrate the much more efficient MLC tracking delivery. The film and PSD dose agreed well for PSD 1 (differences relative to TPS dose ⩽2%). For PSD 2, we found differences relative to the TPS dose of 2%–4%, for PSD 3 1%–6%, and for PSD 4, relative differences of 3-8%. These differences could again be due to the steep dose gradients where these PSDs were positioned, whereby small coregistration errors could cause relatively large mutual differences (Figure 4d).

The MRI4D scintillator cassette's high spatial resolution dose maps (Figure 7) clearly show the effect of MLC tracking. The dose maps support the PSD measurements by demonstrating large hot and cold spots that clearly demonstrate under/overdosage without MLC tracking, while MLC tracking restored this to the static dose. This difference is also clearly reflected in the local 1%/1-mm gamma pass-rates (Table 2) of 9 and 35% for Lujan and patient-derived respiratory motion without MLC tracking, and high pass-rates of 93 and 89% with MLC tracking. Moreover, the high spatial resolution dose map enables a DAH D95% PTV dose analysis, which clearly indicates a difference in PTV coverage between MLC tracking (110% of the prescription dose) and without MLC tracking (67% of the prescription dose) for Lujan motion, while it also revealed more subtle differences for the patient-derived respiratory motion of 98%–109% coverage.

Details are in the caption following the image
Film-based dose maps with and without MLC tracking during a treatment delivery with added Lujan + drift motion. Treatments with and without MLC tracking are compared to a static reference by a dose difference map and a gamma map. The static dose map is also compared to the TPS dose.

4 DISCUSSION

In this study, we demonstrated the suitability of a new commercial MRI4D scintillator cassette in a 1.5 T MR-linac, combining radiochromic film and four time-resolved PSDs in a motion phantom. The results show that the PSD readings of the MRI4D scintillator cassette were consistent with a previously validated HYPERSCINT RP200 PSD25 for which the performance was determined to be excellent in terms of consistency and dose linearity and PRF dependence. Furthermore, the results show that the MRI4D scintillator cassette integrates seamlessly with the Quasar motion phantom to perform time-resolved and motion-included QA in a range of (online) adaptive radiation treatment scenarios. The time-resolved measurements can also be used to accurately quantify the beam-gating latency.

The MRI4D scintillator cassette's PSDs had perfect dose linearity (R2 = 1), a very low PRF dependency (⩽0.4%) that was well within tolerance (2.0%),30 and a very high consistency (relative standard deviation <0.4%) that was also well within tolerance (1.0%).31 The results were also perfectly in line with previous single probe measurements on the 1.5 T MR-linac.25 Replacing the radiochromic film between full IMRT plan measurements introduced a relative standard deviation ⩽0.6%, which is also within the 1.0% tolerance31 and is indicative of the ease of use of the cassette. This high accuracy created the opportunity to compare the dose readings of several IMRT deliveries (e.g., with different motion adaptation strategies) with both film and PSD dose.

The MRI4D scintillator cassette's excellent spatial dose performance was able to conclusively demonstrate how an ATS procedure restored the delivered dose to the TPS dose, with a local 1%/1-mm gamma pass-rate of 90%. Additionally, the accurate PSD dose measurements highlighted the clear dosimetric gain of an ATS procedure in key areas of the dose distribution, whereby dose differences compared to the TPS dose were reduced from 1%–74% to only 0%–2% differences. These excellent results indicate that the MRI4D scintillator cassette can also accurately measure dose in very steep dose gradients (26%/mm). The simultaneously measured PSD and film dose also agreed very well (differences relative to the TPS dose 1%–5%). Only for detector 4, a small coregistration error (<0.5 mm) introduced a 5% relative difference in the really steep dose gradient (36%/mm) at detector 4. Additionally, PSD 4 received here <20% of the prescribed dose. Pixels containing <10%–20% of the prescribed dose are often excluded from film analysis because they suffer from the largest relative film calibration uncertainties.32-34 The film and PSDs also have slightly different positions in the dosimetry cassette, whereby the film was positioned 1 mm anteriorly from the PSDs. However, this induces a maximum dose difference of 0.5% for our treatment plans. Since all adaptive procedures started with an ATS procedure, the static dose is actually an ATS dose. We, therefore, expected similar static versus TPS dose comparisons for all measurements, which is exactly what both the film and PSDs measurements confirmed.

During the trailing deliveries, we demonstrated that the MRI4D scintillator cassette is suitable for time-resolved and motion-induced dosimetry. The cassette's PSD measurements showed the benefits of trailing by showing dosimetric improvements in PSD 2–4 of −11% to −3%, −29% to −8%, and −53% to −22% compared to the TPS dose. The consistent dose measurement of PSD 1 between trailing and no trailing (difference relative to TPS ⩽1%) also demonstrated how stable the PSD measurements are in subtle gradients (<0.2%/mm), despite being subjected to drift motion. The MRI4D scintillator cassette's PSDs also measured dose very accurately when being subjected to more fluctuating patient-derived prostate motion, where trailing reduced the dose differences relative to the TPS dose in PSD 2–4 from 3%–17% to 1%–2%. The dose differences between the static PSD reading and the TPS dose were only 0%–0.9%, despite the 20° angulation of PSDs due to the used phantom ramp. This is in line with single probe measurements also showing no angular dependency for the HYPERSCINT PSD in the 1.5 T MR-linac.25 Similar to the ATS procedure, the MRI4D scintillator cassette's excellent spatial dose measurements provide additional insight into the dosimetric benefit of trailing, with local 3%/3-mm gamma pass-rates improved to 95 and 97% for linear drift motion and patient-derived prostate motion. The film dose and PSD dose agreed perfectly for most PSDs (differences relative to TPS dose 0%–3%). Larger relative differences (7%), were most likely due to small coregistration errors (<0.5 mm) in a dose gradient (4%/mm) and to film calibration-related uncertainties.35-37

The measurements during gating demonstrated that the MRI4D scintillator cassette is very suitable for time-resolved and motion-included dosimetry. The PSD measurements showed how gating reduced the dose differences in PSDs 2–4 compared to the TPS dose from 38%–45% to differences <1% for Lujan motion and from 44%–56% to 5%–12% for patient-derived motion. The small dose difference between PSD 4 and TPS when gating Lujan motion (<1%), demonstrates how accurately the cassette can measure in steep dose gradients (15 %/mm) while being subjected to motion. The spatial dose maps provided by the dosimetry cassette offer great insight into the necessity and performance of gating, with 3%/3-mm gamma pass-rates improved from 31 and 26% to 100% for Lujan and patient-derived respiratory motion. Differences between PSD and film dose relative to the TPS dose were small (0%–5%), where the largest relative difference (5%) was found in PSD 4 w/o gating due to small coregistration errors (<0.5 mm) and the steep dose gradient (14%/mm).

In addition to accurate dosimetry measurements, the time-resolved PSDs simultaneously measured the beam-gating latency at high precision (<13 ms), compared to the ground-truth beam generator signal. The high temporal precision distinguishes negative and positive average beam-gating latency, as it found a beam-on latency of Lujan (−74.6±77.9 ms) and for patient-derived motion (20.5±129.5 ms). A negative beam-on latency indicates that the beam was turned on before the GTV entered the gating window, while a positive beam-on latency suggests that the beam was turned on slightly after the GTV entered the gating window. The large spread of latency values might be due to the position estimation uncertainty of the CMM system, as the interleaving of coronal and sagittal cine MRI introduces some jitter on the derived position trace. It is important to note that the gating software uses a prediction filter to mitigate the gating latency, which explains the relatively small latency values.

Lastly, the MRI4D scintillator cassette was validated during MLC tracking. The results demonstrated that the cassette evaluated the time-resolved and motion-included MLC tracking performance very well. The PSD 2–4 measurements here showed how MLC tracking reduced the dose differences compared to the TPS dose from 6%–13% to 0.7%–7% for Lujan motion and from 13%–26% to 0.6%–1% for patient-derived motion. This demonstrates that the MRI4D scintillator cassette highlights the necessity and the effectiveness of tracking, with accurate dose measurements even in steep dose gradients (maximally 22%/mm), while being subjected to motion. The 7% dose difference found in PSD 4 with MLC tracking could be caused by PSD-TPS coregistration errors, or by small inaccuracies in the MR-derived positions that might also have introduced small tracking errors. The high spatial films also excellently showed how MLC tracking mitigated dose inaccuracies and increased the gamma pass-rates for Lujan+drift motion (93%) and for patient-derived respiratory motion (89%). Differences between PSD and film dose were similar to the gating measurements.

In summary, we have established the suitability of the new commercial MRI4D scintillator cassette in a 1.5 T MR-linac, and its excellent patient-specific and motion-included QA performance when combined with the Quasar phantom, indicating its suitability for routine online adaptive QA as well as end-to-end validation of (future) online adaptive radiotherapy approaches. In the future, the addition of more scintillation points could be desirable to provide better volume coverage. The flexibility of the individually positionable PSDs also introduces the possibility for a dosimetry device with deformable target,38 which could provide additional dosimetric insight for the development and validation of deformable MLC tracking.39

5 CONCLUSION

This study demonstrated the suitability of a new commercial MRI4D scintillator cassette in a 1.5 T MR-linac that provides accurate dosimetry measurements by simultaneously using high spatial-resolution radiochromic film and accurate time-resolved PSDs. Combined with the Quasar phantom, the MRI4D scintillator cassette provides meaningful patient-specific and motion-included QA for a range of (online) adaptive radiotherapy scenarios, including the quantification of beam gating latency.

ACKNOWLEDGMENTS

Prescilla Uijtewaal and Martin Fast acknowledge funding by the Dutch Research Council (NWO) through project no. 17515 (BREATHE EASY). We acknowledge research agreements with Medscint (Quebec City QC, Canada), with Modus QA (London ON, Canada) and with Elekta AB (Stockholm, Sweden).

    CONFLICT OF INTEREST STATEMENT

    The authors declare no conflicts of interest.