An ACR (American College of Radiology) CT accreditation adult abdomen phantom (diameter = 16 cm) was used for this study and is shown in Figure 1(a) and Figure 1(b). The phantom was scanned with a GE (General Electric Company, Milwaukee, Wisconsin, USA) 4-slice LightSpeed CT scanner using a brachytherapy protocol (1.25 mm CT slice thickness, 50 cm field-of-view (FOV)). Images were then transferred to an Oncentra treatment planning system (Elekta Medical System Ltd., Stockholm, Sweden. version 4.3). A treatment plan was developed with a 3.0 cm diameter cylinder simulated. The prescription was 500 cGy per fraction at 5 mm above the cylinder surface. Thus, 2.0 cm away from the center in the radius direction would receive 500 cGy per fraction. The treatment plan was then transferred to the microSelectron HDR unit. A Fluke (Fluke Electronics Corporation, Everett, WA) Model 451 P pressurized ion chamber survey meter was placed with the detector sensitive region touching the phantom surface for dose rate measurement. The University of Wisconsin Accredited Dosimetry Calibration Laboratory calibrated the survey meter with ±6.5% uncertainty. Figure 1 shows the CT scanning and treatment setup for the adult abdomen phantom. The purpose of this phantom study is to get the range of dose rate on the phantom surface and to calculate the calibration factor for the following Monte Carlo simulation to convert fluence to absolute dose per simulated particle.

The EGSnrcMP egs_inprz (version 1.0) program (National Research and Council Canada) was used for the Monte Carlo simulation. EGSnrcMP egs_inprz is a graphic user interface (GUI) for the NRC RZ user-codes

(http://nrc-cnrc.github.io/EGSnrc/doc/pirs801-egsinprz.pdf). DOSRZnrc user code was selected. The adult abdomen phantom was simulated with PMMA materials and the Ir-192 microSelectron spectrum was chosen for the simulation. Uniform isotropically radiating disk of finite size was used and the settings for source simulation were: RMINBM = 0, RBEAM = 0.06, ZSMIN = 5.0 and ZSMAX = 5.35. The Ir-192 Source dimension (diameter and active length) information followed the paper published by Mowlavi et al. in 2008 [18] . In total, 10⁹ particles were simulated with ECUT = 0.521 MeV and PCUT = 0.01 MeV settings. Figure 2(a) depicts the geometry of the simulation. Longitudinal and radius profiles for the simulation were recorded. The 3 mm margin outside of the diameter = 16 cm was assumed to be detector region and used for recording the CT-like device detector signal (here is the deposited dose information) purpose.

In-house Matlab (R2010a, The MathWorks, Natick, MA, USA) code was written to post-process the simulated results. The inverse radon transform (“iradon” function in Matlab) was performed and the 2D CT image was generated. The inverse radon transform will reconstruct the image from the projection data in the 2D array. A de-noising technique with wavelet decomposition at level 5 and soft thresholding (“wavedec2” function in Matlab with N = 5 and wname = “db5”) was performed on original reconstructed 2D image to get better image quality. The wavedec2 function will return the wavelet decomposition of the image at certain level. With the same inverse radon transform, de-noised image showed better reconstructed result.

The measured dose rate using the 451 P survey meter was 8.3 R/h (the apparent source activity was 8.11 Ci at the measurement time), which was converted to air dose using the factor 0.877 cGy/R and a treatment time of 332.3 s. The measured dose was 0.67 cGy. If ±6.5% uncertainty was taken into account, the measured result was then 0.67 ± 0.04 cGy. The Monte Carlo simulation result was then calibrated at 2.0 cm away from the phantom center receiving 500 cGy. The calibration factor was 2 × 10¹³ particles/Gy. The dose was 0.71 ± 0.20 cGy at a radius distance of 16.3 cm, which was 5% higher than the measured result. Factors for the difference between measurements and simulations include the survey meter’s calibration uncertainty and the Monte Carlo simulation’s uncertainty of ±28.0% @ depth 16.3 cm due to the weak signal of the brachytherapy source. However, the measurement was still in the simulation range. Please note that the Monte Carlo uncertainty is independent of the source activity. The source activity was used for deriving the calibration factor for Monte Carlo simulation in order to get absolute dose information. If the source is weaker, the delivery time for the measurement will be longer, but the integral dose will be the same.

Figure 2(b) shows the profile of one normalized signal result at a distance of 16.0 cm to 16.3 cm along the z direction. Figure 2(c) shows the original 2D CT sinogram image and Figure 2(d) shows the re-constructed image. Figure 2(e) shows the de-noised 2D CT sinogram image and Figure 2(f) shows the re-con- structed image of Figure 2(e). The 2D CT was constructed using a dimension of 255 × 255. If 50% of the maximum signal was considered to be source diameter boundary, the source diameter dimension was then 1.3 mm (2 × 160 mm/255 pixel × 1 pixel), which is comparable to the 1.1 mm diameter reported by Mowlavi et al. in 2008 [18] . If we use the de-noised image shown in Figure 2(d) for comparison, the signal to noise ratio would be 19.91 db. Histogram plots for Figure 2(d) and Figure 2(f) show a clear improvement in signal to noise ratios and are presented in Figure 3(a) and Figure 3(b).

A 2^nd Monte Carlo simulation was performed with the source retracted by 5 cm. The results were merged with the first simulation to simulate two source dwell positions. The magnitude of the 2^nd simulation was weighted at 30% of the first simulation in order to differentiate the two signals. Figure 4 shows the dose/particle plot for the simulation of the two differentially weighted dwell positions. The second source center position calculated from the Monte Carlo simulation was 50.2 mm compared to the programmed center position of 50.175 mm. The sub-millimeter accuracy of results clearly illustrates correct source position and relative weight.