Vendor‐independent skin dose mapping application for interventional radiology and cardiology

Abstract Purpose The purpose of this paper is to present and validate an originally developed application SkinCare used for skin dose mapping in interventional procedures, which are associated with relatively high radiation doses to the patient’s skin and possible skin reactions. Methods SkinCare is an application tool for generating skin dose maps following interventional radiology and cardiology procedures using the realistic 3D patient models. Skin dose is calculated using data from Digital Imaging and Communications in Medicine (DICOM) Radiation Dose Structured Reports (RDSRs). SkinCare validation was performed by using the data from the Siemens Artis Zee Biplane fluoroscopy system and conducting “Acceptance and quality control protocols for skin dose calculating software solutions in interventional cardiology” developed and tested in the frame of the VERIDIC project. XR‐RV3 Gafchromic films were used as dosimeters to compare peak skin doses (PSDs) and dose maps obtained through measurements and calculations. DICOM RDSRs from four fluoroscopy systems of different vendors (Canon, GE, Philips, and Siemens) were used for the development of the SkinCare and for the comparison of skin dose maps generated using SkinCare to skin dose maps generated by different commercial software tools (Dose Tracking System (DTS) from Canon, RadimetricsTM from Bayer and RDM from MEDSQUARE). The same RDSRs generated during a cardiology clinical procedure (percutaneous coronary intervention—PCI) were used for comparison. Results Validation performed using VERIDIC's protocols for skin dose calculation software showed that PSD calculated by SkinCare is within 17% and 16% accuracy compared to measurements using XR‐RV3 Gafchromic films for fundamental irradiation setups and simplified clinical procedures, respectively. Good visual agreement between dose maps generated by SkinCare and DTS, RadimetricsTM and RDM was obtained. Conclusions SkinCare is proved to be very convenient solution that can be used for monitoring delivered dose following interventional procedures.


| INTRODUCTION
Interventional procedures in radiology and cardiology are associated with relatively high radiation doses to the patient's skin which may lead to skin reactions. 1 Even though peak skin dose (PSD) assessment can be accomplished with a wide selection of detectors, 2 using software tools is more convenient and economical. At the present time, most equipment vendors have developed online solutions for skin dose calculations. CareMonitor by Siemens and DoseWise by Philips are solutions which basically provide the accumulated peak air kerma in the current projection. 3 On the other hand, Dose Map from GE is an advanced two-dimensional (2D) solution, [4][5][6] while Dose Tracking System (DTS) from Canon (Toshiba) represents a state of the art three-dimensional (3D) solution for skin dose mapping. [7][8][9] The most important drawback of previously mentioned solutions is that they are vendor specific and cannot be used in fluoroscopy systems from other manufacturers.
Utilizing Digital Imaging and Communications in Medicine (DICOM) Radiation Dose Structured Report (RDSR) generated at the end of the intervention is the only way to make vendor-independent solutions. 10 RDSR was added to the DICOM standard 11 with intention to standardized format of recording all the information related to the exposure parameters used for each irradiation event undergone by the patient. RDSR contains all the necessary technical, geometric, and dosimetric data necessary to assess the patient skin dose. In addition to online solutions mentioned above, there are commercial offline software tools which utilize DICOM headers and/ or RDSRs for skin dose calculations such as em.dose from Esprimed, 12,13 RDM by MEDSQUARE, 3 DOSE from Qaelum, 14 NEXO[DOSE] ® by Bracco, 15,16 Radimetrics TM from Bayer, 17 and Skin Dose Map ® tool integrated in DoseWatch ® by GE Healthcare. 18 Other software solutions can be found in literature. [19][20][21][22][23] The objective of this paper is to present an originally developed skin dose mapping application SkinCare that can be readily used with interventional units from different manufacturers. The application was validated by using the data from the Siemens Artis Zee fluoroscopy system and conducting the "Acceptance and quality control protocols for skin dose calculating software solutions in interventional cardiology" developed and tested in the frame of the Validation and Estimation of Radiation Skin Dose in Interventional Cardiology (VERIDIC) project. 24 VERIDIC project, funded under European Joint Programme for the Integration of Radiation Protection Research, H2020 (Grant agreement No 662287), was focused on the skin dose calculation (SDC) software products in interventional cardiology, with an aim to contribute to the harmonization and the validation of SDC software products in interventional cardiology. 24 Additionally, SkinCare's dose maps generated using different RDSRs were compared visually with different validated commercial software tools for Canon (Toshiba), GE, Philips, and Siemens fluoroscopy systems in order to verify the correctness of applied geometric algorithm.

2.A | SkinCare
SkinCare is an application tool for generating skin dose maps following interventional radiology and interventional cardiology procedures using the realistic 3D patient models. RDSRs from Canon (Toshiba), GE, Philips, and Siemens were used for development of the application, making SkinCare compatible with major fluoroscopy unit vendors. SkinCare is a standalone desktop application that runs in any of the available web browsers. Easy configuration of the system correction factors and patient models provide fast way for generating skin dose maps and visualization of RDSR content.

2.A.1 | Patient modeling
Patient models were created using MB-Lab, 25 an open-source plugin for free and open-source 3D computer graphics software Blender. 26 SkinCare has a library of 42 3D patient models of height ranging from 150 cm to 210 cm, with an increment of 10 cm. Male and female models consider three different body types: thin, standard size, and obese. All models have arms-down pose corresponding to patient supine position. Two additional models of sizes 30 × 30 × 15 cm 3 and 35.56 cm × 43.52 cm (14" × 17") represent water phantom and XR-RV3 Gafchromic film, respectively, for the purpose of quality control (QC) tests.

2.A.2 | Patient positioning
Since different manufacturers of the fluoroscopy systems define 3D position of isocenter in relation to proprietary point in space, it is necessary to determine the offsets for RDSR's attributes Table Lateral Position,

2.A.3 | Skin dose calculation
Skin dose calculation is based on determining the affected points of the 3D patient model by the x-ray beam. Once the 3D positions of x-ray tube focal spot, detector, and patient are found for every irradiation event using data from RDSR, dose is calculated as entrance surface dose, that is, only for 3D points in which x-ray beam enters the patient model (x-ray beam exit points are not relevant).
Dose is calculated by using [Eq. (1)]: were oriented in such a way that the yellow side of the film was fac- ing the x-ray tube. The film pieces were irradiated to 16 air kerma values between 0 and 10 Gy using RQR8 standard beam quality. 29 The conversion of air kerma to absorbed dose to skin is considered to be equal to one. 30 Scanning of the irradiated films was performed after 24 hr of exposure with the Hewlett-Packard (HP) Scanjet 7650 flatbed scanner. VueScan scanning software was used for linear (raw) scanning which enables data acquisition straight from the scanner's sensor without any manipulation from the scanning software. Most software packages that come with consumer scanners do not offer this ability and perform processing on raw data. Film pieces were scanned with a resolution of 300 PPI as 48 Bit RGB TIF files.
Scanned images were analyzed in the Python programming language using only the red channel in order to ensure maximum sensitivity.
To obtain an image value corresponding to a particular air kerma, a square region of interest (ROI) was formed with dimensions of 120 px (ROI was approximately 1 cm 2 ). The formed ROI was shifted throughout the whole image, and pixel values inside the ROI were averaged. ROI with the lowest average pixel value corresponds to the highest dose. Response of the XR-RV3 Gafchromic films as a function of film dose was modeled using a rational function proposed by Lewis et al. 31 : where x represents the lowest average pixel value of ROI at dose D, and a, b, and c are coefficients of the rational function. Calibration curve was obtained using Levenberg-Marquardt nonlinear curve fitting algorithm.
Uncertainties related to the use of XR-RV3 Gafchromic films for patient skin dose assessment in the interventional environment have been estimated in our previous work and published in Medical Physics. 30 It is shown that overall uncertainty of skin dose measure-   were recorded for each exposure. Calibration factor for each exposure was calculated as:

Evaluation of calibration factor
where CFis the calibration factor, K a,i is the incident air kerma measured by R100B, Ais the field area in the plane of the R100B, and KAPis air kerma-area product measured by the KAP meter.

Evaluation of table attenuation factor
TAF was measured using the frontal tube (Plane A) set at the under- A radiopaque ruler was used to measure dimensions of x-ray field for setup 2.
Service-mode projections were acquired for each value of the xray tube voltage ranging from 60 to 120 kVp with 10 kVp increments and varying copper thicknesses of 0, 0.1, 0.2, 0.3, 0.6, and 0.9 mm. Also, three field sizes of 10 × 10 cm 2 , 15 × 15 cm 2 , and 20 × 20 cm 2 in the plane of R100B have been employed. Air kerma rates measured by R100B were recorded for each exposure. Table at-tenuation factor for each exposure was calculated as: where TAFis the table attenuation factor, K a,1 is the air kerma rate from setup 1,K a,2 is the air kerma rate from setup 2, and the ratio d2 d1 2 stands for the inverse-square-law correction for distance.
For a beam incident on the table surface at a nonperpendicular angle, θ, TAF should be multiplied with oblique factor, F θ , which is defined as a relative fraction of transmission between zero and nonzero angles of incidence. Using methodology proposed by Rana et al., 9 oblique factor was calculated as:

3.A | SkinCare
The graphical user interface (GUI) of SkinCare is shown in Figure 3.
It is a single-page long-scrolling standalone application that runs in   shown in Figure 6. Fitting coefficients have the following values: a ¼ À851:1,b ¼ 1:754 * 10 7 ,c ¼ À4828.   Table 4. With the increase in the xray tube voltage and by increasing the field size and added Cu filtration of the primary x-ray beam, TAF value has increased. The recorded TAF values ranged from 0.69 to 0.88.  considerably as it can be seen from Table 7 and Fig. 9. Figure 9 shows comparison between scanned XR-RV3 Gafchromic films and dose maps obtained using SkinCare's XR-RV3 Gafchromic phantom for all three simplified clinical procedures. It should be noted that the scanner used for geometrical comparison in Figure 9 was not used for calibration and PSD assessment due to the presence of horizontal lines on scanned images. On both of the images for the same procedures there are letters A-D used for comparison of the spatial distribution of dose and numbers 1-4 used for dose comparison. In   The goal of future research is the validation considering the clinical use of SkinCare and expanding SkinCare's capabilities to support all interventional radiology and cardiology procedures and setups.

CONF LICT OF I NTEREST
The authors have no relevant conflict of interest to disclose.