Radiation Protection Dosimetry Advance Access published August 21, 2014 Radiation Protection Dosimetry (2014), pp. 1–10
doi:10.1093/rpd/ncu256
A MONTE CARLO CALIBRATION OF AWHOLE BODY COUNTER USING THE ICRP COMPUTATIONAL PHANTOMS Jenny Nilsson* and Mats Isaksson Department of Radiation Physics, Institute of Clinical Sciences, The Sahlgrenska Academy, University of Gothenburg, Go¨teborg, Sweden *Corresponding author: [email protected] Received 13 January 2014; revised 9 July 2014; accepted 11 July 2014
This article presents a Monte Carlo calibration method for a whole body counter (WBC) system. The reason for undertaking this work stirs from an explicit request from the Swedish Radiation Protection Agency and the Swedish Civil Contingencies Agency. The agencies asked for a whole body measurement method to be used after accidents or other incidents involving ionising radiation. A WBC can be used for whole body measurements in scenarios involving internal exposures of gamma-emitting radionuclides given that it has been calibrated accordingly. The outcome of an emergency is hard to predict, and the calibration method needs to be fast and versatile to accommodate a wide range of possible internal contamination situations. The WBC can be calibrated for a wide range of measurement geometries by using a Monte Carlo model of the WBC together with a computational phantom. The University of Gothenburg, Sweden, has a low activity laboratory equipped with two WBCs that has been used for metabolic studies(1 – 8) and radiation protection measurements. The WBC system presented in this article consists of four large plastic scintillators capable of detecting high-energy gamma photons. When an ionising particle deposits energy in one of the plastic scintillators, electrons in the molecular structure of the scintillator are excited. The electrons de-excite within nanoseconds through an emission of optical photons (scintillation light). Two photomultiplier tubes (PMTs) are mounted on each scintillator. The optical photons detected by the photocathode of the PMT are converted into an electric signal, which is amplified in the PMT dynode structure through
electron multiplication. The output from each PMT is further amplified, summed detectorwise and fed into a multichannel analyser (MCA) that outputs an energy spectrum. To enable a Monte Caro calibration, the Monte Carlo model of the WBC must be capable of rendering an energy spectrum equivalent to an experimental energy spectrum from a WBC measurement. It has been shown that the optical transport in a plastic scintillator must be included in the Monte Carlo simulations of the WBC response(9) and the Monte Carlo model of the WBC is based on the results and conclusions from three previous works(10, 11). The energy deposition by an ionising particle in the plastic scintillators and the generation, transport and detection of the optical photons at the photocathode surface of the PMT were simulated with the Monte Carlo code GATE v6.2 (Geant4 Application for Tomographic Emission, using the Geant4 version 9.5 patch 01)(12). The PMT response and the creation of an energy spectrum were calculated using MATLAB(13). The Monte Carlo model of the WBC is the output from a GATE simulation coupled to the MATLAB model. The aims of this article are to verify the Monte Carlo model of the WBC and to implement the ICRP computational phantoms of the Reference Male and Reference Female(14) into the Monte Carlo model. The implementation of the ICRP computational phantoms should enable a calibration for realistic measurement geometries with respect to nuclear emergencies. The Monte Carlo model was verified by comparing the simulated energy spectrum and simulated counting efficiency with the experimental
# The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected]
Downloaded from http://rpd.oxfordjournals.org/ at University of Ulster on February 22, 2015
A fast and versatile calibration of a whole body counter (WBC) is presented. The WBC, consisting of four large plastic scintillators, is to be used for measurements after accident or other incident involving ionising radiation. The WBC was calibrated using Monte Carlo modelling and the ICRP computational phantoms. The Monte Carlo model of the WBC was made in GATE, v6.2 (Geant4 Application for Tomographic Emission) and MATLAB. The Monte Carlo model was verified by comparing simulated energy spectrum and simulated counting efficiency with experimental energy spectrum and experimental counting efficiency for high-energy monoenergetic gamma-emitting point sources. The simulated results were in good agreement with experimental results except when compared with experimental results from high dead-time (DT) measurements. The Monte Carlo calibration was made for a heterogeneous source distribution of 137Cs and 40K, respectively, inside the ICRP computational phantoms. The source distribution was based on the biokinetic model for 137Cs.
J. NILSSON AND M. ISAKSSON
are given in Equations (3) and (4). 2
E½X ¼ emþs
=2
ð3Þ 2
2
Var½X ¼ e2mþs ðes 1Þ 2
¼ ðE½X Þ2 ðes 1Þ:
ð4Þ
If E[X ] and Var[X ] are known, then m and s can be written as in Equations 5 and 6, respectively. ! 1 Var½X m ¼ ln(E½X Þ ln 1 þ ; ð5Þ 2 ðE½X Þ2 vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !ffi u u Var½X s ¼ tln 1 þ ðE½X Þ2
ð6Þ
The ICRP computational phantoms THEORY The PMT and the lognormal distribution This section gives a statistical description of the multiplication in a PMT and explains the lognormal distribution, which was used in the MATLAB model of the PMT multiplication. In a PMT, photoelectrons are released by optical photons at a photocathode surface. The photoelectrons undergo multiplication in a dynode chain, and the final gain, G, and its relative standard deviation, sG/G, are given in Equations 1 and 2(16). G ¼ Adn
ð1Þ
sG 1 ¼ ð1 dÞ G
ð2Þ
A is the fraction of the photoelectrons collected by the multiplier structure/dynode chain, n is the number of dynodes and d is the number of released electrons at each dynode per incoming electron. In spectrometry, a PMT will cause a deterioration of spectrum resolution since the released number of electrons in a dynode per incoming electron varies. Avariable that is the product of positive and independent random variables can be described by a lognormal distribution(17). Therefore, G was treated as a lognormally distributed variable in the MATLAB model of the PMT. The lognormal distribution is a continuous probability distribution of a random variable whose logarithm is normally distributed. If X is lognormally distributed, and m and s are the mean and standard deviation for the normal distribution of ln(X ), then the expectance value E[X ] and variance Var[X ] for X
The two ICRP computational phantoms are made from whole body computed tomography image sets of a male and a female in a supine position. The image sets have been segmented and scaled to fit the reference values for the reference male (1.76 m and 73.0 kg) and the reference female (1.63 m and 60.0 kg) defined by ICRP(18, 19). Segmentation means that a pixel in an image is given an organ identification number. A tissue is coupled to each organ identification number, and the elemental composition ( percentage by mass) is defined for each tissue. Since the image slice refers to an anatomical thickness given by the slice thickness, each pixel also defines a volume element, i.e. a voxel. The computational phantom of the reference male is a 254` 127` 222 array, in total 1 946 375 voxels and the reference female is a 299` 137` 348 array, in total 3 886 020 voxels. Both phantoms are in the shape of a box where the voxels not associated with an organ are filled with air. ICRP uses the term computational phantom; another commonly used term is voxel phantom. METHOD The verification of the Monte Carlo model The WBC in the Monte Carlo model The WBC consists of four equivalent detectors where two detectors are placed above a patient bed and two below. Figure 1a shows a schematic of the WBC and the placement of the ICRP computational phantoms. In the verification of the Monte Carlo model, the simulated results were compared with experimental results for four point sources; the sources are further presented in the section ‘The sources and their
Page 2 of 10
Downloaded from http://rpd.oxfordjournals.org/ at University of Ulster on February 22, 2015
energy spectrum and experimental counting efficiency for four point sources (in the energy range 661.657– 1115.546 keV) placed at different positions relative to the WBC. After the ICRP computational phantoms had been implemented, a realistic measurement geometry was achieved by using the content and distribution of potassium, K, in the phantoms. The alkali metal potassium is a micronutrient and is found in human tissue, and the alkali metal caesium, Cs, accumulates in human tissue in a similar way as potassium once it has entered the blood system(15). The WBC response was simulated for the naturally occurring radioisotope 40K distributed in the ICRP computational phantom with respect to the phantoms’ potassium content. The WBC response was also simulated for 137Cs, which is often released into the environment after nuclear accidents, using the same distribution as for 40K.
CALIBRATION OF AWHOLE BODY COUNTER
placement for the verification of the Monte Carlo model’. Each source was placed on the patient bed at three positions and for each position an energy spectrum was recorded; the source positions are shown in Figure 1b. The detector housing has a 0.397-mm copper window facing the patient bed and inside each detector, surrounded by an air gap, is a large plastic scintillator measuring 91.5` 76.0` 25.4 cm3. The scintillator is made of NE-102A, which is equivalent to BC400(16). Each scintillator has two 1200 PMTs (EMI 9545A with photocathode surface material Cs3SbO)(20), mounted through a 16.5-cm-long perspex light guide at one short end and at the opposite short end is a reflecting aluminium foil. The signal from each PMT is amplified (ORTEC - 855), summed detectorwise and fed into a MCA. The energy spectrum from the MCA can be shown per detector or as a sum from two or more detectors. A more detailed description of the WBC and the low activity laboratory can be found in Ref. (11). The Monte Carlo model of the WBC included all four detectors and the patient bed. There was one main difference between the Monte Carlo model of the WBC and the real WBC; in the Monte Carlo model, the light guides were omitted and a photocathode surface was placed directly on the short end of the plastic scintillator. To transport optical photons in GATE, the user needs to define the surface properties between two volumes. The surface controls the physical processes an optical photon undergoes when reaching the boundary. Three surfaces were defined: the plastic scintillator surface facing the air, the reflecting aluminium foil surface facing the plastic scintillator and the photocathode surface facing the plastic scintillator. The surface parameters, summarised in Table 1, were chosen based on the results from the extensive work presented in Ref. (9). The composition of the plastic scintillators, C:H ¼
Table 1. The surface parameters used for the optical transport in GATE. Plastic scintillator Surface type Dielectric– dielectric Surface Ground finish sa 68 Reflectivity 1.0 Reflection Specular lobe type Efficiency N/A
Aluminium foil
Photocathode
Dielectric– metal Polished
Dielectric– metal Polished
08 3MTM Specular spike 0
08 0 Specular spike Cs3Sb-O
Reflectivity 3MTM stands for the wavelength-dependent reflectivity for the 3MTM Aluminium Foil Tape 425(21). Efficiency Cs3Sb-O stands for the wavelength-dependent relative efficiency for the photocathode material Cs3Sb-O(20).
10:11 and density 1.032 g cm23, and the following scintillation properties were taken from the data sheet for BC-400(22): scintillation yield ¼ 10 000 optical photons/MeV, a light emission spectrum consisting of seven discrete wavelengths and their respective probabilities, fast time constant ¼ 2.4 ns, absorption length ¼ 250 cm and refractive index ¼ 1.58. The remaining properties, resolution scale, slow time constant, yield ratio and Rayleigh scattering length, were not found in the data sheet. In GATE, the Gaussian distribution of the emitted optical photons per deposited MeV is given by resolution scale ` (absorbed energy` scintillation yield)1/2. In this work, the resolution scale was set to 1. Only one time constant was given in the BC-400 data sheet, and this was used as the fast time constant. The Rayleigh scattering length was determined using the method described by Riggi et al.(21) The refractive index for air was set to
Page 3 of 10
Downloaded from http://rpd.oxfordjournals.org/ at University of Ulster on February 22, 2015
Figure 1. (a) The schematic of the WBC and placement of the voxel phantom. Two detectors are placed above and two below a patient bed, giving a nearly 3p geometry. (b) A schematic of detector 1 from above. For the verification of the Monte Carlo model, point sources were placed on the patient bed (not shown), Ò marks the source positions, and 70, 50 and 30 cm are the distances from the source position to the boundary between detector 1 and detector 4, marked by
doi:10.1093/rpd/ncu256
A MONTE CARLO CALIBRATION OF AWHOLE BODY COUNTER USING THE ICRP COMPUTATIONAL PHANTOMS Jenny Nilsson* and Mats Isaksson Department of Radiation Physics, Institute of Clinical Sciences, The Sahlgrenska Academy, University of Gothenburg, Go¨teborg, Sweden *Corresponding author: [email protected] Received 13 January 2014; revised 9 July 2014; accepted 11 July 2014
This article presents a Monte Carlo calibration method for a whole body counter (WBC) system. The reason for undertaking this work stirs from an explicit request from the Swedish Radiation Protection Agency and the Swedish Civil Contingencies Agency. The agencies asked for a whole body measurement method to be used after accidents or other incidents involving ionising radiation. A WBC can be used for whole body measurements in scenarios involving internal exposures of gamma-emitting radionuclides given that it has been calibrated accordingly. The outcome of an emergency is hard to predict, and the calibration method needs to be fast and versatile to accommodate a wide range of possible internal contamination situations. The WBC can be calibrated for a wide range of measurement geometries by using a Monte Carlo model of the WBC together with a computational phantom. The University of Gothenburg, Sweden, has a low activity laboratory equipped with two WBCs that has been used for metabolic studies(1 – 8) and radiation protection measurements. The WBC system presented in this article consists of four large plastic scintillators capable of detecting high-energy gamma photons. When an ionising particle deposits energy in one of the plastic scintillators, electrons in the molecular structure of the scintillator are excited. The electrons de-excite within nanoseconds through an emission of optical photons (scintillation light). Two photomultiplier tubes (PMTs) are mounted on each scintillator. The optical photons detected by the photocathode of the PMT are converted into an electric signal, which is amplified in the PMT dynode structure through
electron multiplication. The output from each PMT is further amplified, summed detectorwise and fed into a multichannel analyser (MCA) that outputs an energy spectrum. To enable a Monte Caro calibration, the Monte Carlo model of the WBC must be capable of rendering an energy spectrum equivalent to an experimental energy spectrum from a WBC measurement. It has been shown that the optical transport in a plastic scintillator must be included in the Monte Carlo simulations of the WBC response(9) and the Monte Carlo model of the WBC is based on the results and conclusions from three previous works(10, 11). The energy deposition by an ionising particle in the plastic scintillators and the generation, transport and detection of the optical photons at the photocathode surface of the PMT were simulated with the Monte Carlo code GATE v6.2 (Geant4 Application for Tomographic Emission, using the Geant4 version 9.5 patch 01)(12). The PMT response and the creation of an energy spectrum were calculated using MATLAB(13). The Monte Carlo model of the WBC is the output from a GATE simulation coupled to the MATLAB model. The aims of this article are to verify the Monte Carlo model of the WBC and to implement the ICRP computational phantoms of the Reference Male and Reference Female(14) into the Monte Carlo model. The implementation of the ICRP computational phantoms should enable a calibration for realistic measurement geometries with respect to nuclear emergencies. The Monte Carlo model was verified by comparing the simulated energy spectrum and simulated counting efficiency with the experimental
# The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected]
Downloaded from http://rpd.oxfordjournals.org/ at University of Ulster on February 22, 2015
A fast and versatile calibration of a whole body counter (WBC) is presented. The WBC, consisting of four large plastic scintillators, is to be used for measurements after accident or other incident involving ionising radiation. The WBC was calibrated using Monte Carlo modelling and the ICRP computational phantoms. The Monte Carlo model of the WBC was made in GATE, v6.2 (Geant4 Application for Tomographic Emission) and MATLAB. The Monte Carlo model was verified by comparing simulated energy spectrum and simulated counting efficiency with experimental energy spectrum and experimental counting efficiency for high-energy monoenergetic gamma-emitting point sources. The simulated results were in good agreement with experimental results except when compared with experimental results from high dead-time (DT) measurements. The Monte Carlo calibration was made for a heterogeneous source distribution of 137Cs and 40K, respectively, inside the ICRP computational phantoms. The source distribution was based on the biokinetic model for 137Cs.
J. NILSSON AND M. ISAKSSON
are given in Equations (3) and (4). 2
E½X ¼ emþs
=2
ð3Þ 2
2
Var½X ¼ e2mþs ðes 1Þ 2
¼ ðE½X Þ2 ðes 1Þ:
ð4Þ
If E[X ] and Var[X ] are known, then m and s can be written as in Equations 5 and 6, respectively. ! 1 Var½X m ¼ ln(E½X Þ ln 1 þ ; ð5Þ 2 ðE½X Þ2 vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi !ffi u u Var½X s ¼ tln 1 þ ðE½X Þ2
ð6Þ
The ICRP computational phantoms THEORY The PMT and the lognormal distribution This section gives a statistical description of the multiplication in a PMT and explains the lognormal distribution, which was used in the MATLAB model of the PMT multiplication. In a PMT, photoelectrons are released by optical photons at a photocathode surface. The photoelectrons undergo multiplication in a dynode chain, and the final gain, G, and its relative standard deviation, sG/G, are given in Equations 1 and 2(16). G ¼ Adn
ð1Þ
sG 1 ¼ ð1 dÞ G
ð2Þ
A is the fraction of the photoelectrons collected by the multiplier structure/dynode chain, n is the number of dynodes and d is the number of released electrons at each dynode per incoming electron. In spectrometry, a PMT will cause a deterioration of spectrum resolution since the released number of electrons in a dynode per incoming electron varies. Avariable that is the product of positive and independent random variables can be described by a lognormal distribution(17). Therefore, G was treated as a lognormally distributed variable in the MATLAB model of the PMT. The lognormal distribution is a continuous probability distribution of a random variable whose logarithm is normally distributed. If X is lognormally distributed, and m and s are the mean and standard deviation for the normal distribution of ln(X ), then the expectance value E[X ] and variance Var[X ] for X
The two ICRP computational phantoms are made from whole body computed tomography image sets of a male and a female in a supine position. The image sets have been segmented and scaled to fit the reference values for the reference male (1.76 m and 73.0 kg) and the reference female (1.63 m and 60.0 kg) defined by ICRP(18, 19). Segmentation means that a pixel in an image is given an organ identification number. A tissue is coupled to each organ identification number, and the elemental composition ( percentage by mass) is defined for each tissue. Since the image slice refers to an anatomical thickness given by the slice thickness, each pixel also defines a volume element, i.e. a voxel. The computational phantom of the reference male is a 254` 127` 222 array, in total 1 946 375 voxels and the reference female is a 299` 137` 348 array, in total 3 886 020 voxels. Both phantoms are in the shape of a box where the voxels not associated with an organ are filled with air. ICRP uses the term computational phantom; another commonly used term is voxel phantom. METHOD The verification of the Monte Carlo model The WBC in the Monte Carlo model The WBC consists of four equivalent detectors where two detectors are placed above a patient bed and two below. Figure 1a shows a schematic of the WBC and the placement of the ICRP computational phantoms. In the verification of the Monte Carlo model, the simulated results were compared with experimental results for four point sources; the sources are further presented in the section ‘The sources and their
Page 2 of 10
Downloaded from http://rpd.oxfordjournals.org/ at University of Ulster on February 22, 2015
energy spectrum and experimental counting efficiency for four point sources (in the energy range 661.657– 1115.546 keV) placed at different positions relative to the WBC. After the ICRP computational phantoms had been implemented, a realistic measurement geometry was achieved by using the content and distribution of potassium, K, in the phantoms. The alkali metal potassium is a micronutrient and is found in human tissue, and the alkali metal caesium, Cs, accumulates in human tissue in a similar way as potassium once it has entered the blood system(15). The WBC response was simulated for the naturally occurring radioisotope 40K distributed in the ICRP computational phantom with respect to the phantoms’ potassium content. The WBC response was also simulated for 137Cs, which is often released into the environment after nuclear accidents, using the same distribution as for 40K.
CALIBRATION OF AWHOLE BODY COUNTER
placement for the verification of the Monte Carlo model’. Each source was placed on the patient bed at three positions and for each position an energy spectrum was recorded; the source positions are shown in Figure 1b. The detector housing has a 0.397-mm copper window facing the patient bed and inside each detector, surrounded by an air gap, is a large plastic scintillator measuring 91.5` 76.0` 25.4 cm3. The scintillator is made of NE-102A, which is equivalent to BC400(16). Each scintillator has two 1200 PMTs (EMI 9545A with photocathode surface material Cs3SbO)(20), mounted through a 16.5-cm-long perspex light guide at one short end and at the opposite short end is a reflecting aluminium foil. The signal from each PMT is amplified (ORTEC - 855), summed detectorwise and fed into a MCA. The energy spectrum from the MCA can be shown per detector or as a sum from two or more detectors. A more detailed description of the WBC and the low activity laboratory can be found in Ref. (11). The Monte Carlo model of the WBC included all four detectors and the patient bed. There was one main difference between the Monte Carlo model of the WBC and the real WBC; in the Monte Carlo model, the light guides were omitted and a photocathode surface was placed directly on the short end of the plastic scintillator. To transport optical photons in GATE, the user needs to define the surface properties between two volumes. The surface controls the physical processes an optical photon undergoes when reaching the boundary. Three surfaces were defined: the plastic scintillator surface facing the air, the reflecting aluminium foil surface facing the plastic scintillator and the photocathode surface facing the plastic scintillator. The surface parameters, summarised in Table 1, were chosen based on the results from the extensive work presented in Ref. (9). The composition of the plastic scintillators, C:H ¼
Table 1. The surface parameters used for the optical transport in GATE. Plastic scintillator Surface type Dielectric– dielectric Surface Ground finish sa 68 Reflectivity 1.0 Reflection Specular lobe type Efficiency N/A
Aluminium foil
Photocathode
Dielectric– metal Polished
Dielectric– metal Polished
08 3MTM Specular spike 0
08 0 Specular spike Cs3Sb-O
Reflectivity 3MTM stands for the wavelength-dependent reflectivity for the 3MTM Aluminium Foil Tape 425(21). Efficiency Cs3Sb-O stands for the wavelength-dependent relative efficiency for the photocathode material Cs3Sb-O(20).
10:11 and density 1.032 g cm23, and the following scintillation properties were taken from the data sheet for BC-400(22): scintillation yield ¼ 10 000 optical photons/MeV, a light emission spectrum consisting of seven discrete wavelengths and their respective probabilities, fast time constant ¼ 2.4 ns, absorption length ¼ 250 cm and refractive index ¼ 1.58. The remaining properties, resolution scale, slow time constant, yield ratio and Rayleigh scattering length, were not found in the data sheet. In GATE, the Gaussian distribution of the emitted optical photons per deposited MeV is given by resolution scale ` (absorbed energy` scintillation yield)1/2. In this work, the resolution scale was set to 1. Only one time constant was given in the BC-400 data sheet, and this was used as the fast time constant. The Rayleigh scattering length was determined using the method described by Riggi et al.(21) The refractive index for air was set to
Page 3 of 10
Downloaded from http://rpd.oxfordjournals.org/ at University of Ulster on February 22, 2015
Figure 1. (a) The schematic of the WBC and placement of the voxel phantom. Two detectors are placed above and two below a patient bed, giving a nearly 3p geometry. (b) A schematic of detector 1 from above. For the verification of the Monte Carlo model, point sources were placed on the patient bed (not shown), Ò marks the source positions, and 70, 50 and 30 cm are the distances from the source position to the boundary between detector 1 and detector 4, marked by
