Dosimetric validation study of a flattening filter free SABR treatment by use of a normoxic PAG gel dosimeter with MRI readout

This paper describes clinical studies of the normoxic PAG gel dosimeter with MRI readout for validation of complex flattening filter free (FFF) treatment delivery techniques. Recently our department implemented the Elekta FFF technology for stereotactic ablative radiotherapy (SABR) treatments using dynamic conformal arc (DCAT) and volumetric arc therapy (VMAT). In this study 6MV FFF DCAT and VMAT plans were analysed using 3D gel dosimetry techniques and compared against standard clinical dosimetry QA tools. Some challenges with this validation study included high dose per fraction for SABR, MRI access time and R2-dose calibration errors. A study of the dependency of the dose response on dose rate revealed insignificant dose rate dependence of the normoxic PAG dosimeter up to 20 Gy/min for both photon energies (6 MV and 10 MV).


Introduction
Three-dimensional dosimeters are of great interest in a progressive radiotherapy facility utilising complex and advanced treatments techniques such as high dose rate dynamic SABR. In our Centre, we recently introduced three-dimensional polymer gel dosimeters for encapsulating and validating 6MV FFF VMAT and DCAT treatments. Treatments of this nature, however, are difficult and complex to validate and mainstream quality assurance tools are often limited in their ability to satisfy clinical demands [1][2][3][4][5][6][7]. Three-dimensional polymer gel dosimeters offer significant advantages for validating steep 3D dose distributions of this complexity. This paper discusses some of the challenges encountered while using the normoxic PAG gel dosimeter [8,9] in the clinical setting such as high dose short fractionation regimes, dose rate response and geometric and dosimetric accuracy. Clinical examples are presented for 6MV FFF VMAT and DCAT treatments planned using Monaco™ v 5.10.02 and delivered using the Elekta Agility™ head.

Phantom simulation and treatment planning
The gel container was initially filled with water and imaged using a Siemens Somatom™ Helical CT scanner. Our department uses the Monaco TM Monte Carlo treatment planning system. The 6MV FFF DCAT and VMAT patient plans were copied onto the phantom data set retaining segment shapes and monitor units using a 2 mm calculation grid with 1% statistical uncertainty in dose per plan. The resulting transverse dose distributions were exported with 2 mm centre slice spacing separation to compare with the measured gel dose distributions.

Phantom & calibrations vials
In each study the liquid dosimeter was poured into eleven 15 ml Pyrex Tm glass calibration vials and a 3 litre container of similar type that has been used elsewhere [10] (see Figure 2). The dosimeter was left on the bench top to cool for an hour and then subsequently placed in a refrigerator to set overnight.

Phantom and calibration vial irradiation
Temperature control measures were implemented before irradiation to ensure the dosimeters were in equilibrium with the ambient air temperature. The gel phantom was placed on the treatment couch at the linac iso-centre and irradiated according to the patient's treatment plan. The calibration vials were irradiated to different known doses in water according to the TRS-398 protocol [11].

MRI readout
After equilibrating with the ambient air temperature, the dosimeters were imaged using a Siemens Avanto 1.5T clinical MRI scanner with a head coil. An interleaving multi-echo-pulse sequence consisting of 16 echoes with TE of 40 ms, TR of 3240 ms, FOV 128 x 128 mm 2 , 2 mm slice thickness, pixel bandwidth of 130, 1 acquisition with 21 slices separated by 2 mm centres was used resulting in an imaging time of 42 minutes [12]. Due to MRI access constraints only the minimal imaging parameters were used. It was determined that a TE of 40ms with 16 echoes was sufficient to cover the range of R2dose response within the dosimeter.

Processing and analysis
MRI images were processed using in-house software created in Matlab TM . R2 maps of each axial slice in the phantom were calculated and converted to dose maps using the R2-dose response calibration curve generated from the calibration vials.  Figure 1 shows the R2-dose response curves for the two batches of gels made in this clinical study. The R2-dose sensitivities were 0.31029 s -1 Gy -1 for the DCAT study and 0.29175 s -1 Gy -1 for the VMAT study both with linear regions that extended up to 5 Gy. Figure 2 shows the irradiated DCAT gel phantom along with the MRI dose map. Only central axis transverse dose maps are shown in figures 3 and 4 for each delivery technique compared against the TPS with the associated gamma values. Adjacent to each dose map comparison is a profile taken through the central axis of the dose distribution and the associatred gamma values for the above criteria. In the transverse plane 21 slices at 2 mm separation can be analysed and by interpolation, 128 sagital and coronal dose planes are available to analyse at 1 mm centre separations, which allows the dose distribution in each plane to be assessed. The identified errors in R2 in the phantom are potentially related to reasons discussed elsewhere, such as, over response of R2 in a large phantom volume compared with relatively small calibration vials [13]. Normalisation was performed on the central axis with a small shift applied to locate a better overall fit of the dose distribution. Off-axis dose distributions were also assessed, however due to the limited scope within this abstract these will not be covered in this article.

DCAT plan.
The overall pass rate for the central axis slice was 93.6% with 776 points failing the criteria and 11388 passing. No gamma score was above 2 for this dose plane.

Clinical QA measurement using SNC ArcCheck and CC04 ionisation chamber
The VMAT plan passed the specified gamma criteria at 97.6% with 9 points failing and 360 passing. The highest gamma score was 2.45. The dose at the isocentre measured -2.2% below the TPS value. The DCAT plan passed the specified gamma criteria at 95.5% with 21 points failing and 441 passing. The highest gamma score was 2.7. The point dose measured 2.4% above the TPS value. Figure 5 shows the R2response as a function of dose rate for a dose of 5Gy. The plot shows a weak dependence with increasing dose rate with a linear fit having a gradient of -6x10 -5 s -1 cGy -1 min for 6MV and -9x10 -5 s -1 cGy -1 min for 10MV.

Conclusion
There are many challenges involved in setting up a gel dosimetry program [14][15][16] using MRI [17][18][19][20][21][22][23]. Expertise, time and access constraints are limiting factors in imaging optimization. Clinical requirements dictate that for short fractionation high dose regimes that the geometric and dosimetric parameters of a plan are well known and within tight tolerances. In this study, due to an R2 over response in the dosimeter only the normalised dose distribution was studied. The errors in R2 are potentially related to reasons discussed elsewhere [13]. While the normoxic PAG dosimeter has highly favourable characteristics for radiotherapy applications, it has a limited useful dose range [24] which is problematic for SABR doses. A typical fractionation regime would be 10Gy/fx, however when calculated on the gel phantom doses can be in excess of 20Gy. Ideally a gel will receive doses within the range of its optimal dose resolution [25]. The normoxic PAG dosimeter manufactured in this study exhibited a weak dose rate dependence at FFF dose rates and further work should be conducted to fully characterise this observation. Further development work will continue on our dosimetric and calibration techniques for high dose high dose rate SABR fractionation regimes.