Impact of detector selections on inter‐institutional variability of flattening filter‐free beam data for TrueBeam™ linear accelerators

Abstract This study evaluates the type of detector influencing the inter‐institutional variability in flattening filter‐free (FFF) beam‐specific parameters for TrueBeam™ linear accelerators (Varian Medical Systems,Palo Alto, CA, USA). Twenty‐four beam data sets, including the percent depth dose (PDD), off‐center ratio (OCR), and output factor (OPF) for modeling within the Eclipse (Varian Medical Systems) treatment planning system, were collected from 19 institutions. Although many institutions collected the data using CC13 (IBA Dosimetry, Schwarzenbruck, Germany) or PTW31010 semiflex (PTW Freiburg, Freiburg, Germany) ionization chambers, some institutions used diode detectors, diamond detectors, and ionization chambers with smaller cavities. The OCR data included penumbra width, full width at half maximum (FWHM), and FFF beam‐specific parameters, including unflatness and slope. The data measured by CC13/PTW31010 ionization chambers were compared with those measured by all other detectors. PDD data demonstrated the variations within ±1% at the dose fall‐off region deeper than peak depth. The penumbra widths of the OCR measured with the CC13/PTW31010 detectors were significantly larger than those measured with all other detectors (P < 0.05). Especially the EDGE detector (Sun Nuclear Corp., Melbourne, FL, USA) and the microDiamond detectors (model 60019; PTW Freiburg) demonstrated much smaller penumbra values compared to those of the CC13/PTW31010 detectors for the 30 × 30 mm2 field. There was no difference in the FWHM, unflatness, and slope parameters between the values for the CC13/PTW31010 detectors and all other detectors. OPF curves demonstrated small variations, and the relative difference from the mean value of each data point was almost within 1% for all field sizes. Although the penumbra region exhibited detector‐dependent variations, all other parameters showed tiny interunit variations regardless of the detector type.


| INTRODUCTION
Stereotactic radiotherapy has demonstrated excellent local control of intracranial and extracranial localized tumors. [1][2][3] Recently, linear accelerators (linacs) equipped with flattening filter-free (FFF) beams have become commonly used around the world. For conventional treatments, photon beams with high intensities at the center of the beam become flattened by a flattening filter. For stereotactic treatments with small fields, however, a flattened beam is not essential.
By eliminating the filter, the beams can deliver a very high dose rate, which decreases the treatment time. [4][5][6] It has been indicated in several studies that the interunit variability in modern linacs is small, likely owing to the improved manufacturing. 7  (PTW Freiburg, Freiburg, Germany) from multiple institutions, and they reported that the interunit variability was very small. 9 While these ionization chambers are often used for beam data collection, various other detectors with small sensitive volumes, such as ionization chambers with smaller cavities, diode detectors, diamond detectors, and plastic scintillators, are also employed, especially for measuring small-field beams. It has been demonstrated in many studies that the detector type has an influence on the penumbra of offcenter ratio (OCR) profiles and output factors (OPFs). 10 It has been shown in many studies that the characteristics of FFF beams differ from those of flattened beams in terms of the cone-shaped OCR, lower effective beam energy affecting the percent depth dose (PDD), photon energy spectrum affecting the water-air stopping power ratio, [15][16][17] and high dose per pulse affecting the ion recombination coefficient. 18,19 Moreover, the type of the detector may affect the collected beam data. In addition, there are some dosimetric parameters specialized for FFF beams because of their unique profile shape. 20 However, no one has reported how detector selection affects these parameters. Here we investigate the impact of detector selection on the FFF beam-specific parameters for beam data collected from multiple institutions.

2.A | Data collection
According to institutional agreement, 24 sets of TrueBeam™ data were collected from 19 institutions. All data were measured for modeling within the Eclipse TPS, and treatment fields were collimated with jaws. Data were submitted in the format of the three-dimensional scanning water phantoms or W2CAD format, a format for data registration of the Eclipse TPS. The field sizes (FSs) of the collected data were 30, 100, and 200 mm square fields. PDD and crossline OCR data were measured with a source-to-surface distance of 100 cm. The OPF data were collected in a Microsoft Excel (Microsoft Corp., Redmond, WA, USA) spreadsheet. All collected data were imported in Akilles RT software (RADLabInc., Osaka, Japan) to create a database. The details of the detectors evaluated in this study are listed in Table S1. The number of detectors used for data collection is listed in Tables 1 and 2. Most of the institutions used either the CC13 or PTW31010 semiflex ionization chamber. Only one institution used the PTW30013 Farmer-type ionization chamber to measure OPF and compared their data with the RBD.

2.B | Data analysis
All PDD data were resampled to the data with a 1 mm interval and normalized at 100 mm depth, as data normalized at the peak depth will be affected by noise around the peak. For each data point, the mean value and standard deviation (SD) of 24 machines were calculated. In order to evaluate the variability, the maximum SD (SD max ) was calculated for the dose fall-off region deeper than the peak depth.
where Dose centralÀaxis and Dose XÀoffÀaxis represent the central-axis dose level and the dose level at a certain off-axis position, respectively. Slope was calculated as where y represents the dose at the coordinate x, whereas x 1 and x 2 represent the positions located at one-third and two-thirds of the half of FWHM from the central axis, respectively.
The parameters of CC13/PTW31010 and all other detectors were compared using Wilcoxon signed-rank test using Microsoft Excel. Statistical significance was set at a P < 0.05.         Abbreviations: d 10 , dose at a 10 cm depth; FFF, flattening filter-free; FS, field size; FWHM, full width at half maximum; OCR, off-center ratio; SD, standard deviation; min-max, minimum-maximum. *P < 0.05.

| RESULTS
significantly larger penumbra width than other detectors including diode and diamond detectors. The inner diameters of cavity of the CC13 and PTW31010 chambers are 5.5 and 6 mm, respectively, and they are much larger than the sensitive area of diode and diamond detectors (Table S1). Such differences of the sensitive volume greatly affected the measured data at steep regions. Similar data have previously been reported in many studies. 10,11 In contrast, the type of detector showed modest impacts on the shape of the curves, such as the dose fall-off region of PDD and field region of OCR defined as 80% FWHM, probably because the dose variations are not steep.
In addition, the FFF-specific parameters, including unflatness and slope, showed very small variations among detectors. Although a few data showed statistically significant differences, the variations were within 2%. For flattened beams, the photon energy spectrum changes in off-center region because of the thickness of flattening filter, whereas the spectrum of the FFF beams is not changed. 23 Therefore, the energy spectrum will be stable in the field region where the FFF-specific parameters are evaluated. However, it has been reported that large FS results in the increase of scattered photons with low energy, leading to an overresponse of diode detectors. 10 Abbreviations: d 10 ,dose at a 10 cm depth; FFF, flattening filter-free; FS, field size; FWHM, full width at half maximum; min-max, minimum-maximum; OCR, off-center ratio; SD, standard deviation. *P < 0.05.
F I G . 4. Output factor (OPF) curves of 6 and 10 MV flattening filter-free beams are shown in the insets, with the relative differences between each curve and the mean values with field sizes ranging from 30 × 30 to 400 × 400 mm 2 Tx™ and reported that the OPF values varied in a detector-dependent manner. 14 In addition, the authors also showed that variations were significantly reduced when applying output correction factors,

| CONCLUSIONS
In this study, we investigated the interunit variability in TrueBeam™ linacs among multiple institutions, focusing on FFF-specific parameters and detector selection. Although the penumbra region demonstrated detector-dependent variations, all other parameters, including the slope and unflatness, exhibited very small interunit variations, regardless of the detector type.

ACKNOWLEDGMENTS
We are deeply grateful to the institutions that provided their beam data for this study. This study was supported by JSPS KAKENHI Grant Number 17K15802 and Katarou-kai Golden Beam Data Working Group.

CONF LICT OF I NTEREST
The second author, Y. Akino, is a developer of the commercial software Akilles RT, which was used for analysis in this study.