MOSFET dosimeter characterization in MR‐guided radiation therapy (MRgRT) Linac

Abstract Purpose With the increasing use of MR‐guided radiation therapy (MRgRT), it becomes important to understand and explore accuracy of medical dosimeters in the presence of magnetic field. The purpose of this work is to characterize metal‐oxide‐semiconductor field‐effect transistors (MOSFETs) in MRgRT systems at 0.345 T magnetic field strength. Methods A MOSFET dosimetry system, developed by Best Medical Canada for in‐vivo patient dosimetry, was used to study various commissioning tests performed on a MRgRT system, MRIdian® Linac. We characterized the MOSFET dosimeter with different cable lengths by determining its calibration factor, monitor unit linearity, angular dependence, field size dependence, percentage depth dose (PDD) variation, output factor change, and intensity modulated radiation therapy quality assurance (IMRT QA) verification for several plans. MOSFET results were analyzed and compared with commissioning data and Monte Carlo calculations. Results MOSFET measurements were not found to be affected by the presence of 0.345 T magnetic field. Calibration factors were similar for different cable length dosimeters either placed at the parallel or perpendicular direction to the magnetic field, with variations of less than 2%. The detector showed good linearity (R2 = 0.999) for 100–600 MUs range. Output factor measurements were consistent with ionization chamber data within 2.2%. MOSFET PDD measurements were found to be within 1% for 1–15 cm depth range in comparison to ionization chamber. MOSFET normalized angular response matched thermoluminescent detector (TLD) response within 5.5%. The IMRT QA verification data for the MRgRT linac showed that the percentage difference between ionization chamber and MOSFET was 0.91%, 2.05%, and 2.63%, respectively for liver, spine, and mediastinum. Conclusion MOSFET dosimeters are not affected by the 0.345 T magnetic field in MRgRT system. They showed physics parameters and performance comparable to TLD and ionization chamber; thus, they constitute an alternative to TLD for real‐time in‐vivo dosimetry in MRgRT procedures.

Recently, preliminary studies were done on MOSFET detectors for use in MR-IGRT 60 Co based modality with a 0.345 T magnetic field. 14,15 The effect of magnetic field was investigated with regard to depth dose, linearity of response, and angular dependence. No significant difference in dose at depth was found with or without the magnetic field. 15 Knutson et al. 14 reported a slight increase in MOSFET response (~5%), attributed to induced currents from the dynamic magnetic field; hence, they advised to perform a special calibration procedure to account for the change in detector response to the dynamic magnetic field. According to Thorpe et al., 16 MOS-FET-based dosimeters, under the influence of 1 Tesla magnetic field, can be used as dosimeters for MRI-guided radiotherapy (MRI Linac) with no effect of the magnetic field.
In this study, we intend to perform an extensive characterization of the commercially available MOSFET dosimeters (Best Medical Canada) in the MRIdian ® Linac system settings for several radiation characteristics and for typical patient MRgRT treatments, and compare them to TLD and IC commissioning results, to determine their suitability as an in-vivo patient dose verification tool. The MOSFET detector has a sensitive area of 0.2 mm 2 × 0.2 mm 2 and 0.5 μm in SiO 2 thickness 5,6 . The silicon chip is packaged on a polyimide (Kapton) flexible circuit, sealed with an organic epoxy of 1.8 g/cm 3 density, resulting in 1.3 mm total thickness and 2.5 mm width dosimeter. The MOSFET physical build-up due to epoxy is approximately 0.8 mm, corresponding to an inherent water equivalent build-up of~1.5 mm, when corrected for density.

| MATERIALS AND METHODS
The MOSFET dosimeter mode of operation is described elsewhere. 6,8 Briefly, the detector is a dual P-type MOSFET composed of two transistors, where hole transport dominates the channel current. The threshold voltage, V th , which is the gate voltage allowing current conduction through the drain to the source, changes with radiation dose; this parameter change, ΔV th in mV, is proportional to the dose and is measured to establish the calibration factor of the dosimeter.
All MOSFET measurements were performed on MRIdian ® Linac system equipped with 6 MV flattening filter free (FFF) inline linac.

2.A | Calibration
The calibration factor is used to convert the measured detector response in mV to dose in cGy. Each MOSFET was calibrated before its use. MOSFETs of different lengths (long and short) were calibrated to evaluate the cable length effect on the calibration factor. Calibration was also performed by aligning the MOSFET parallel and perpendicular ( Fig. 1) to the main magnetic field orientation.

2.B | Monitor unit linearity
To determine the useful dynamic range of the MOSFET, 100-600 MUs were delivered in a step size of 100 MUs at 1.5 cm depth, 90 cm SAD, with 9.96 cm 2 × 9.96 cm 2 field size, and gantry parked at 0°. Three dose measurements for each MU setting were recorded, and the average MOSFET response in mV was determined.

2.C | Output factor
Field size dependence was studied from 0.83 to 20.75 cm 2 field sizes at 5 cm depth, at gantry 0°. For each field size, dose was recorded for 200 MUs delivered at 90 cm SAD. Two dose measurements for each MU setting were recorded, and the average MOSFET response, normalized to the 9.96 cm 2 × 9.96 cm 2 field size, was used to determine the output factor. Results were compared to the IC data.

2.D | Directional dependence
Directional dependence of MOSFET was investigated by placing the detector in a cylindrical water phantom (ViewRay Inc. Cleveland, OH). 200 MUs were delivered from different gantry angles: 0°to 330°range, with 30°increments, and for a field size of 9.96 cm 2 × 9.96 cm 2 . Beams passing through the couch, with inherent attenuation, were included. Figure 2 shows the setup. TLD and IC measurements were performed in the same phantom setup and gantry rotations. MOSFET response was the average of three dose readings for each gantry angle. All measurements were normalized to the average response for all angles and for each detector. The setup used allows only relative comparisons of angular response between F I G . 1. Setup used for metal-oxide-semiconductor field-effect transistors (MOSFET) calibration: MOSFET was placed parallel and perpendicular to the magnetic field. MOSFET aligned perpendicular to the main magnetic field is shown in the figure.
F I G . 2. ViewRay QA cylindrical water phantom setup used to investigate directional dependence using ionization chamber and metal-oxide-semiconductor field-effect transistors (MOSFET).
| 129 detectors and is anisotropic in nature; hence, quantitative angular dependence data cannot be inferred from these measurements. from the surface simulates skin thickness of clinical concern for skin toxicity in radiation therapy.

2.F | IMRT QA
IMRT QA was performed for three clinical sites: mediastinum, liver, and spine. Liver was planned for 50 Gy in 5 factions, spine was planned for 24 Gy in single fraction, and mediastinum was a conventional treatment for 30 Gy in 10 fractions. QA plans were generated on MRIdian ® linac treatment planning system (TPS). MR compatible ArcCHECK ® (Sun Nuclear Corporation, Melbourne, FL, USA) was used for IMRT dose verification. A special holder was built to hold the MOSFET at the point of measurement. The active volume of the MOSFET was facing gantry 0°f or all measurements. IMRT QA dose distribution is shown in Fig. 3.
MOSFET measurements were compared to IC measurements. IMRT QA setup for treatment delivery is shown in Fig. 4.

3.A | Calibration
Calibration factors for MOSFET detectors with different cable lengths are shown in Table 1. Consistency in the MOSFET calibration factor over different cable lengths was observed. Due to an insignificant difference in calibration factors for different cable lengths, MOSFETs with longer cables were used in this study. The long cable MOSFET allowed safer distances between the dosimeters and the reader, placed outside the 5 Gauss line.
Placement of the MOSFET detectors parallel or perpendicular to the magnetic field resulted in similar calibration factors, as shown in Table 2; the maximum percentage difference in calibration factor for the two setups was 1.8%. This number is within the detector reproducibility of 2.4% at two standard deviations (SD) as shown in Table 2.
Chuang et al. 6 reported a reproducibility measurement of 1.5% at 1 SD for 1 Gy at 6 MV beam energy. For all measurements, calibration factors were within the range of 1.11-1.14 mV/cGy for all detectors.
Kinhikar et al. 18 reported calibration factors within 1.10-1.12 mV/cGy and 2% SD for a 6-MV photon beam with a helical tomotherapy. The results obtained in the MRgRT linac setting are within the range of values reported elsewhere for conventional radiotherapy linacs. 6,18 The MOSFET measurements with the MR magnetic field turned On, in comparison to the Off condition, did not influence the MOSFET dose response; the dose response ratio was within ±2% for 100-600 MUs dose range, which is within the dosimeter reproducibility.

3.C | Output factor
The output factor for MOSFET and IC is shown in Fig. 6. The MOS-FET behavior is consistent with the IC results. The absolute discrepancy in output factors for the two detectors was within 0.2%-2.2% for all field sizes investigated. This difference is close to the 2% reproducibility of the dosimeter. The average MOSFET output factor for 0.83 to 20.75 cm field sizes was 0.954; the same as the IC. The estimates of the output factor standard deviation were 10.4% and 9.9% for MOSFET and IC, respectively, for the fields of interest. The consistency of the MOSFET output factor and its small size make it an attractive alternative to the ionization chamber to verify the output factor at different field sizes.   and 53.3% for MOSFET measurement and MC calculation, respectively. In another study by Xiang et al. 19 on MOSFET PDD response compared to MC calculations in the build-up region (0.1 to 2 cm), done at 6-MV energy and with no magnetic field, an agreement within 3%-5% was found, similar to our result (5%). The study also showed that MOSFET PDD measurements were 60.3% and 42% at 2 and 1 mm depths, respectively. 19 It should be noted that the mea-

3.F | IMRT QA
IMRT QA plan for mediastinum, liver, and spine are shown in Fig. 3(a-c).
IMRT QA for these plans was performed using the MOSFET dosimeter and compared to IC results. Results of IMRT QA are summarized in shown that the MOSFET measured dose and the dose calculated by the planning system agreed within 5%. 6,9,20,21 These results make MOSFET attractive for in-vivo dose verification during IMRT QA in MRgRT at 0.345 T magnetic field strength.

| DISCUSSION
In-vivo dosimetry is an important part of patient quality assurance (QA) in radiotherapy. ViewRay's MRIdian ® Linac system is relatively a new treatment modality with fully integrated MR-guided, 6 MV FFF photon beam. There is no standardized report for guidelines on QA devices and procedures for MRgRT, and therefore it is critical to perform careful characterization of dosimeters before use in clinical practice.
MOSFETs were clinically proven dosimeters for quality assurance of conventional IMRT beams with no magnetic fields 5,6,9,20 as they are small in size, energy independent, linear, and isotropic, hence allowing dose measurements in high dose gradient fields at multiple energies and beam orientations.
It was our intent to extend their use for MRgRT modality as an in-vivo dosimeter; to this end, extensive characterization of its radiation characteristics in the presence of the low magnetic field of 0.345 T, used by the ViewRay's MRIdian ® Linac system, was performed.
The magnetic field is known to affect semiconductor devices 10,11 as in some conditions it deviates carriers toward the Si/SiO 2 interface, resulting in change of the magnetoresistance of the conduction channel. In our MOSFET measurements, the effect of the magnetic field on detector response was found to be negligible and within the detector reproducibility; this is likely related to the detector operation in the low drain current region at the low onset V th voltage, which makes the channel magnetoresistance and Hall effect negligible. 10 In the presence of the magnetic field and with no air gaps, the ERE effect 12 is unlikely to affect the MOSFET readouts especially when the dosimeter is placed under full build-up material at Dmax (~1.5 cm), as described in the above calibration test setup. This is consistent with the observed MOSFET behavior during calibration, with MR magnetic field turned On or Off.
In the special case where MOSFETs are used with partial or no build-up, such as in surface dose measurements, the ERE effect was found in this study to have no effect on dose measurements, contrary to what was anticipated for phantom-air interface at higher magnetic fields. 12,13 Indeed, Raajmakers et al. 12   IC, ionization chamber; IMRT QA, intensity modulated radiation therapy quality assurance; MOSFET, metal-oxide-semiconductor field-effect transistors.