Evaluation of skin dose calculation factors in interventional fluoroscopy

Abstract Purpose The purpose of this study was to measure fluoroscopic dose calculation factors for modern fluoroscopy‐guided interventional (FGI) systems, and to fit to analytical functions for peak skin dose (PSD) calculation. Methods Table transmission factor (TTF), backscatter factor (BSF), and a newly termed kerma correction factor (KCF) were measured for two interventional fluoroscopy systems. For each setup, air kerma rates were measured using a small ionization chamber in fluoroscopic service mode while selecting kVp, copper (Cu) filter thickness, incident angle, and x‐ray field size at the assumed patient skin locations. Angle dependency on KCF was measured on the GE system at isocenter for angles of 0, 15, 30, and 40 degrees, using a range of kVp, Cu filters, and one field size. An analytical equation was created to fit the data to facilitate PSD calculation. Results For the GE system, oblique incidence measurements show KCF decreased by about 2%, 8%, and 13% for incident angles of 15, 30, and 40°, respectively, relative to KCF at 0 degree. The GE and Siemens systems' KCFs ranged from 0.89 to 1.45, and 0.64 to 1.44, respectively. The KCFs increased with a power of field size, and generally increased with kVp and Cu filter. The average percentage difference between TTF × BSF × f and KCF was 16% at normal incidence. The KCF data were successfully fitted to function of angle, field size, kVp, and Cu filter thickness using seven parameters, with an average R‐squared value of 0.98 and maximum percentage difference of 6.0%. Conclusions This study evaluated scatter factors for two fluoroscopy systems, and dependencies on angle, kVp, Cu filter, and field size, with emphasis on under table beam orientations. Analytical fitting of the data with exposure parameters may facilitate PSD calculations, and more accurately determine the potential for radiation‐induced skin injury.

have investigated real-time and postprocedure radiation skin dose mapping. [7][8][9] The National Council on Radiation Protection and Measurements (NCRP) suggests using the calculated peak skin dose (PSD) as a metric for managing prompt and latent skin reactions to radiation exposure. 10 An equation to calculate patient skin dose is suggested by NCRP Report 168 10 and Jones AK 11 is given by: where K a,r is the air kerma reported at the interventional reference point, 6 d r is the source to reference point distance, SSD is the source-to-skin distance, TTF is the table transmission factor, BSF is backscatter factor of soft tissue, 12 and f is the dose conversion factor from air to soft tissue. 13 This approach has been used by research groups 7-9 studying skin dose mapping and has been shown to be accurate with using the fluoroscope's reported reference air kerma, and measurements of BSF and TTF.
Backscatter factors have been calculated using Monte Carlo simulations for a variety of x-ray beam qualities, 14 and for filtered x-ray beams intended to reduce patient dose. 15 More recently, Wunderle et al measured BSFs for a modern fluoroscope that employs copper (Cu) filtration, using polymethyl methacrylate (PMMA) blocks and an over-table x-ray tube setup. 16 The authors found that BSF depended on kVp, added Cu filter thickness, and x-ray field size, and ranged from 1.18 (using 60 kVp, 0.0 mm Cu, 11 cm field of view) to 1.58 (using 80 kVp, 0.9 mm Cu, 42 cm field of view). Polymethyl methacrylate is a common phantom material and surrogate for soft tissue, 17 but it has been demonstrated that the scatter properties are not identical, 15,18 and a correction is necessary to account for these differences. Corrected BSF measurements may be readily applied to laterally oriented x-ray beams, but may not necessarily apply to under-table exposures due to the attenuation and forward scatter from the table and pad.
The patient support has been shown to affect image quality for automatic brightness control fluoroscopy units. 19 The energy spectrum incident on the patient skin for under-table exposures would therefore differ from those oriented in the lateral direction. DeLorenzo et al. showed that the patient support attenuates the x-ray beam and becomes a source of forward scatter, and some combinations of table and pad can reduce the transmission of air kerma to 60% of its original value. 20 With the presence of the patient support in mind, we hypothesize that PSD calculations would benefit from measuring BSF and TTF simultaneously to produce a single value for x-ray beam orientations that intercept the patient support. We call this value the kerma correction factor (KCF), which intends to convert the inverse square corrected K a,r to D skin by taking into account the forward scatter and potential spectral changes due to the patient support.   Fig. 1(b), setup 2]. The chamber was centered in the x-ray beam, in the "torso" region of the patient support. The KCF was computed by multiplying the air to tissue dose conversion factor (f-factor) by the ratio of exposure rates from setup 2 by inverse square law corrected exposure rates from setup 1 (see Fig. 1), where f is the f-factor, B st PMMA is the ratio of backscatter factors of soft tissue to PMMA, K setup 1 and K setup 2 are kerma rates measured for a given exposure parameter configuration, and SCD 1 (83 cm) and SCD 2 (81.5 cm) are source-to-chamber distances for setup 1 and setup 2, respectively. The f-factor is tabulated with half value layer (HVL) by the International Council on Radiation Protection (ICRP), 12 and was

2.B | Dependency of KCF on incident angle
The GE Innova 2100 system was used to measure the dependence of incident angle on KCF. The system's radiation isocenter was found using a fiducial marker and multiple views, and was measured to be 70 cm from the x-ray source. The RadCal 10x6-60 ion chamber was placed on the patient support pad at the system's isocenter and secured with tape. Next, slabs totaling 20-cm PMMA were placed tightly above the chamber and support pad at the assumed patient skin location. Air kerma rates were measured using all configurations of kVp and Cu filter in Section II.A.1. The lateral gantry angles of 0°(corresponding to the PA view), 15°, 30°, and 40°were used with a nominal field size of 11.8 × 11.8 cm 2 at the chamber location. An angle factor, F θ was calculated as, to correct the KCF measurements at 0°to account for obliquely incident beams.

2.C | Backscatter factor
Backscatter factor was measured using both fluoroscopes to estimate skin doses for laterally oriented beams. The patient support pad was removed and the fluoroscope was oriented in the lateral direction at 90°, without the x-ray field intercepting the patient support. A ceiling-mounted protective face shield was positioned above the x-ray field and was used as a platform to suspend the RadCal 10x6-0.6 ion chamber. The ion chamber was carefully centered in the x-ray field, 90 cm from the x-ray focal spot, to achieve a rotated version of setup 1. Air kerma rates were measured for all previous exposure parameter configurations of kVp, added Cu filtration, and field size, using a 20-cm thick PMMA phantom positioned upright directly behind the suspended ion chamber. Backscatter factor for each exposure parameter configuration without the influence of the patient support was found by dividing exposure rates with and without PMMA blocks behind the chamber, and correcting for scatter property differences between ICRU soft tissue and PMMA.

2.E | Analytical data fitting
The measured KCF and BSF were fitted to an analytical function of kVp, Cu filter, field size, and incident angle. The effect of incident angle was employed as a multiplicative factor to correct data measured for x-ray beams at normal incidence. The final expression will take the form, where B 1 represents the dependency on incident angle, and B 2 describes the dependency on kVp, Cu filter, and x-ray field size. Matlab (Mathworks Inc., Natick, MA, USA) least squares nonlinear curve fitting was used to fit the measured KCF and BSF data to a model, and a representative equation with the highest coefficient of determination (R 2 ) and lowest maximum percentage difference between predicted and measured values was reported.

2.F | Phantom simulation
An interventional fluoroscopic procedure was simulated using an anthropomorphic (adult male), tissue equivalent 21  (ATOM phantom, CIRS, Norfolk, VA, USA) to assess the accuracy of Eq. 6, and to compare the approach of using KCF vs TTF × BSF × f.
Using the Siemens Artis Zee system, the phantom was positioned on the patient support supine, and the 10x 6-0.6 Radcal ion chamber was placed beneath the phantom's thoracic spine region, seen in  to the skin for each beam was estimated using KCF (eq 2) and calculated from measured BSF, TTF, and f-factor estimated using HVL (eq 1), and compared to the ion chamber measurements The GE system's HVLs for all kVp/Cu filter combination were higher than the Siemens system, seen in Table 3, particularly for low values of Cu filter thickness. The trend for both fluoroscopes is that KCF increases with HVL. This would result more penetration through the table and pad, and a higher scatter fraction off the F I G . 2. Anthropomorphic phantom experiment setup using 12 radiation events with different kVp, Cu filter, field size, and angle. Dose to air was measured under the torso of the phantom with a Radcal 10x6-0.6 ion chamber, and corrected using B st PMMA : The measured surface dose was compared to results predicted by Eqs 7-9 and Eq 1. Results are shown in Table 6. PMMA, polymethyl methacrylate.
F I G . 3. KCF as a function of x-ray field size for the Siemens Artis Zee system. KCF increased with field size due to the increased scatter from the edges of the PMMA phantom. The field size dependence was more pronounced as the added Cu filter thickness increased. Field sizes were measured at the ion chamber location. KCF, kerma correction factor; PMMA, polymethyl methacrylate.
PMMA toward the detector. Kerma correction factor for the GE system was consistently larger than for the Siemens system.
F θ increases with kVp, added Cu filter, and decreases with incident angle. F θ is used as a multiplicative factor to correct KCF for obliquely incident gantry angles, θ, where 0°indicates normal incidence of the x-ray beam's central ray into the patient.

3.C | Backscatter factor
The BSF of the GE system and 13.5 × 13.5 cm 2 field size is shown in [Fig. 4(b)]. Full BSF values for the GE system and Siemens system are shown in Tables 1 and 2, respectively. BSF generally increased with Cu filter thickness. For smaller values of Cu filter thickness, BSF increased with kVp, and then decreased for larger values of filter thickness. BSF for the GE system was largest at 60 kVp and 0.9 mm F I G . 4. KCF (a) and BSF (b) with 20-cm thick PMMA phantom and GE Innova 2100, using RadCal 10x6-0.6 ion chamber and 13.5 × 13.5 cm2 field size. KCF generally increases with kVp and added Cu filter for all but the highest energy x-ray beams. BSF has a different relationship with kVp and added Cu filter, reaching a maximum value of 1.41 at 60 kVp and 0.9 mm Cu. BSF, backscatter factor; KCF, kerma correction factor; PMMA, polymethyl methacrylate.
T A B L E 1 GE Innova 2100 KCF and BSF for three field sizes at ion chamber location, using 20-cm PMMA phantom. Cu, and BSF for the Siemens system was largest at 90 kVp and 0.9 mm Cu filter. This difference in shape and maximum value highlights the differences in energy spectrum properties between the two systems.

3.E | Modeling of the broad beam transmission
Several equation forms were investigated to accurately fit the measured data, and our selection criteria focused on maximizing the total R 2 while minimizing the largest percentage difference between measured and fitted values. The transmission data were successfully fitted to the following equations, where θ was the incident angle of the central ray (degrees), x was the x-ray field size at the point of measurement (m 2 ), k was the kVp, and c was the added Cu filter thickness (mm). The coefficients for each patient support, percentage differences from measured values, and R 2 values are shown in Table 5.

3.F | Phantom surface dose
The total accumulated dose and dose per radiation event measured with the ion chamber were compared to the values predicted by eq 8 (KCF fitting method), and using eq 1 (measured TTF × BSF × f method). A summary of the phantom results is shown in Table 6.
The total dose to soft tissue recorded by the ion chamber for 12 beams was 909 mGy, while the dose predicted by the KCF method and fitting eq 8 was 836 mGy, and the total dose calculated using separately measured BSF and TTF with eq 1 was 720 mGy. Dose calculated using separately measured TTF × BSF × f was consistently lower than the measured dose for all 12 beams.

| DISCUSSION
For both the GE and Siemens systems, the table and pad had a substantial effect on the measured air kerma rate. The tables and relatively thick support pad attenuated about 25% of the beam for moderate values of kVp and added Cu filter (Table 4). For unfiltered beams using low kVp, the patient support transmitted as little as 52% of the primary beam. For the GE system, using the 13.5 × 13.5 cm 2 field size and 20-cm PMMA phantom, the average percentage difference between TTF × BSF × f and KCF was 15.8% at normal incidence. For the Siemens system, average percentage differences were 13%-16% across the ranges of kVp, Cu filter, and field size . The setup used for measuring KCF more closely matches   the clinical scenario for under-table x-ray  The angle factor, F θ , for the KCF showed the effect of incident angle on the kerma experienced at the assumed patient skin location, as a fraction of what was measured at normal incidence. F θ is consistent with the increased path length through the patient support with incident angle, and thus, a simple path length correction is a good approximation. This finding was also reported by Rana et al. 7 For angles of 30°and 40°, the entire x-ray beam went through the patient support, but part of the beam impacted the side surface of the 20-cm thick PMMA phantom. This decrease in F θ for larger angles resulting from sideways incidence onto the phantom would presumably become more pronounced with SSD for the same phantom, and is unclear how F θ would change with a rounded phantom that more closely approximates the body habitus of the patient. In practice, one might use eqs 7-9 together with the radiation event information available in the RDSR. For example, KCF and BSF can be estimated easily for each radiation event from the RDSR, which contains kVp and Cu filter data, and effective angle of the beams' central rays. The HVL can be estimated using The fit equation enables a more automated and beam-specific consideration of PSD correction factors when used alongside the T A B L E 6 Twelve beams from the Siemens fluoroscope were used to simulate an interventional procedure on an anthropomorphic phantom torso, and predicted skin dose from eqn 8 and from separately measured BSF and TTF. F-factors were calculated from measured HVL, and ion chamber doses to air were corrected using the same f-factors. Dose program, radiation event data sent from the fluoroscope. For every beam in the anthropomorphic phantom experiment, the ion chamber remained in the center of the field of view, consistent with the setups with PMMA used to measure KCF. This study assumes a uniform intensity across the x-ray field, and one may expect a decrease in primary beam intensity and scatter for off-axis ion chamber measurements.
A limitation of this study is the use of a rectangular PMMA phantom for acquiring the majority of the data. Although this was corrected using B st PMMA , human patients are comprised of soft tissue, bone, and air, and a more thorough investigation of the scattering properties of these materials would be helpful to estimate skin dose more accurately. Another limitation was the use of a single support pad for both systems; previous work has shown a relationship between TTF and support pad thickness. 20 Vendors typically quote the equivalent aluminum thickness (mm Al Eq.) of the patient support, but this concept does not account for the effect of kVp, added Cu filtration, and the materials' scatter properties.

| CONCLUSION
Fluoroscopic skin dose may be calculated using the presented KCFs for under-table x-ray beam geometries, and the BSFs for lateral beam geometries. Kerma resulting from primary beam transmission, patient support forward scatter, and patient material backscatter underestimate the dose by up to 18% at the surface of 20-cm thick PMMA phantoms. Exam-specific radiation event data from the RDSR are available for many fluoroscopic systems, and the analytical equations herein may be used for more automated implementations of skin dose calculations. With the increase in awareness of elevated air kerma rates during complex FGI procedures, this work may aid the clinical physicist in performing peak skin dose calculations, or enable the opportunity to automate the calculation and improve accuracy.

CONFLI CTS OF INTEREST
The authors have no relevant conflict of interest to disclose. F I G . 6. Twelve beam anthropomorphic phantom experiment simulated in custom peak skin dose calculation software. RDSR was converted by Radimetrics into spreadsheet format and imported into modified Volume Viewer. The resulting PSD calculation was performed using KCF for each radiation event. KCF, kerma correction factor; PSD, peak skin dose; RDSR, radiation dose structured reports.