Calibration of medical gamma cameras for estimation of internal contamination from 137Cs

Calibration of 22 gamma camera units was performed at 15 hospitals in southern and western Sweden to estimate 137Cs contamination in humans in a supine static geometry, with a new developed calibration protocol and phantom. The minimum detectable activities (MDAs) and the estimated committed effective doses (CEDs) were calculated for each calibration. Generic calibration factors were calculated for five predetermined groups based on the detector type and manufacturer. Group 1 and 2 included NaI-based gamma cameras from General Electrics (GEs) with a crystal thickness of 5/8′′ and 3/8′′ respectively. Group 3 and 4 included NaI-based gamma cameras from Siemens Healthineers with a crystal thickness of 3/8′′, with a similar energy window as the GE NaI-based cameras and a dual window respectively. Group 5 included semiconductor-based gamma cameras from GE with a CdZnTe (CZT) detector. The generic calibration factors were 60.0 cps kBq−1, 52.3 cps kBq−1, 50.3 cps kBq−1, 53.2 cps kBq−1 and 48.4 cps kBq−1 for group 1, 2, 3, 4, and 5 respectively. The MDAs ranged between 169 and 1130 Bq for all groups, with measurement times of 1–10 min, corresponding to a CED of 4.77–77.6 μSv. A dead time analysis was performed for group 1 and suggested a dead time of 3.17 μs for 137Cs measurements. The dead time analysis showed that a maximum count rate of 232 kcps could be measured in the calibration geometry, corresponding to a CED of 108–263 mSv. It has been shown that semiconductor-based gamma cameras with CZT detectors are feasible for estimating 137Cs contamination. The generic calibration factors derived in this study can be used for gamma cameras of the same models in other hospitals, for measurements in the same measurement geometry. This will increase the measurement capability for estimating internal 137Cs contamination in the recovery phase following radiological or nuclear events.


Introduction
In case of an accident at a nuclear power plant, radioactive elements may be released into the environment and cause contamination of workers as well as members of the public.Referring to the Chernobyl accident in 1986, the release was dominated by the noble gases 85 Kr and 133 Xe, followed by 131 I and 132 Te with half-lives on the order of days, and 137 Cs and 134 Cs with half-lives on the order of years.Other gamma-emitting radionuclides that were released have half-lives on the order of days or months, except for 106 Ru with a half-life of about one year [1].Although the risk of contamination may be lowered by mitigating actions such as sheltering and relocation, methods to determine possible contamination are still needed [2].The need for extensive monitoring of contamination has been clearly demonstrated following the two large nuclear power plant accidents at Chernobyl and Fukushima [3][4][5].
Measurements to estimate the body burden of radioactive elements, notably caesium isotopes, are often performed as whole-body measurements using detector systems dedicated to these types of measurements.These whole-body counters (WBCs) are equipped with detectors (e.g.NaI, HPGe, or plastic scintillators) that are shielded from the surrounding background radiation, either partially or totally enclosed in steel rooms.The detectors can operate in stationary mode or scan the subject in various patterns.Owing to shielding, WBCs are generally characterised by a very low background and, hence, a low minimum detectable activity (MDA).However, the number of WBCs is limited, and a lack of measurement resources can be expected following a radiological or nuclear (RN) event, as demonstrated after the Fukushima accident 11th of March 2011 when more than 50 new WBCs was purchased and installed in the Fukushima prefecture and during the period June 2011 to end of September 2013, over 156 000 subjects were scanned for possible internal contamination [3].
Gamma cameras for medical imaging, also called single-photon emission computed tomography, have proven to be a useful supplement to WBCs in the case of radiological accidents that require the assessment of the internal radiation dose to affected subjects (e.g.[6][7][8]).These gamma cameras can be found in most hospitals, and several units can be found in larger hospitals.Medical gamma cameras are, therefore, both abundant and well-distributed over a country compared to the available WBCs [6].Most gamma cameras are equipped with two detectors of NaI; however, in recent years, gamma cameras equipped with detector elements of CdZnTe (CZT) have emerged and have increased in popularity.However, methodological difficulties exist regarding both detector types when they are applied to the measurement of internal contamination in the case of nuclear accidents.The possibility of adjusting the energy window used for spectrum acquisition is often limited, with a high-energy threshold typically below 600 keV for modern gamma cameras.Thus, no photo peaks can be detected for 137 Cs.Some commercial gamma cameras also have limitations regarding data output, which in some cases require auxiliary software to extract full uncorrected spectrum data.
In a previous study, the medical gamma camera model GE Discovery NM/CT 670 Pro was investigated by considering the calibration factors and MDA for 152 Eu, 137 Cs, and 60 Co [6].The measurements were performed using an IRINA whole-body phantom (set UPh-08T) [9], and the results suggested that this gamma camera could be used to estimate the whole-body activity of the above radionuclides.However, in the analysis, we used special software to extract the spectrum data, and this method cannot be expected to be generally available at all hospitals.In addition, the use of IRINA phantom is cumbersome for hospitals; therefore, the calibration process must be simplified.
Following the war in Ukraine and the threat to Ukrainian nuclear installations, the Västra Götalandsregionen (VGR), responsible for medical health care in western Sweden, felt the need to prepare gamma cameras in the region for the assessment of internal contamination in case of contaminated refugees arriving in Sweden.By cooperating with the VGR, we had access to gamma cameras in the region and were able to study how the previously developed calibration method could be implemented in clinical settings.Therefore, the aim of this study was to develop a calibration phantom and a calibration method that could be used by hospital staff for emergency preparedness in the recovery phase after the accident [10], as well as to calculate the dead time losses for different activity levels for the measurement of 137 Cs with gamma cameras.

The phantoms
Two identical phantoms were constructed using water-filled containers.One of the phantoms contained only water and was used for the background measurements, whereas the other contained water with added 137 Cs (figure 1).Each phantom was constructed of twelve 5 l-water containers, and the lower seven containers (legs and pelvis) were filled with 3 l of water each and the upper five (chest and head) with 5 l of water to obtain a realistic distribution of the body mass.The total weight and length of each phantom were approximately 49 kg and 150 cm, respectively.
The aim of constructing the phantoms was to have a simple phantom that could be easily transported between hospitals and could be accurately reconstructed in the proper measurement geometry at each gamma camera facility.To be a reasonable representative for adults and children, the phantoms should also mimic the IRINA P3 phantom, with a weight and length of 41 kg (representing a 50 kg human) and 160 cm, respectively.The activity distribution was slightly different from the IRINA P3, with 11% of the total activity in the head, 43% in the upper body, and 46% in the lower body; whereas, The IRINA P3 activity distribution was 9% in the head, 50% in the upper body and 41% in the lower body.The activity concentration of 137 Cs was 0.1234 Bq ml −1 , and the total activity of the water container phantom with 137 Cs was 5.676 kBq ± 3% on the 26th March 2022, which was below the legal transportation limit, enabling transportation without an additional permit.Each water container was marked with a letter from A-L (AW-LW for the water-only phantom), and the water containers were constructed at the same position for every measurement.Hydrochloric acid (HCl) was added to all containers and diluted to a concentration of 0.1 M HCl to maintain a low pH of the solution to avoid adsorption to the container walls by keeping the 137 Cs in ion form.The location of the hospitals and the number of gamma camera units included in the study at each site, marked with different shades of green and sizes depending on the number of units.Two of the three active nuclear power plants in Sweden are marked with a radioactive symbol.The six regions are colour coded as Västra Götaland (teal), Halland (red), Kronoberg (orange), Kalmar County (pink), Blekinge (purple) and Skåne (yellow).The map was retrieved from the software Google Earth Pro (Google, USA) [11].

The hospitals
The 15 hospitals included in this study were situated in the south and west of Sweden in six regions: Västra Götaland, Halland, Kronoberg, Kalmar County, Blekinge, and Skåne (figure 2).
Two of the three active nuclear power plants in Sweden are marked with a radioactive symbol in the figure.The third active nuclear power plant in Sweden was not covered by the map.The gamma cameras were operated at three sites at Sahlgrenska University Hospital in Gothenburg: Sahlgrenska Hospital, Östra Hospital, and Queen Silvia Children´s Hospital.Not all gamma camera units in operation in the six regions and at the 15 hospitals were included in the study.

The gamma camera systems
Gamma camera models from two manufacturers were used in this study: General Electrics Healthcare (GE) and Siemens Healthineers (Siemens).Table 1 lists the gamma camera models, total number of units, crystal thickness (for NaI-based detectors), and location, with the number of units in brackets.A total of 22 gamma camera units from 15 hospitals were used in this study.All gamma camera models had two detector heads, and all models, except GE Discovery NM 630, had a Computed Tomography (CT) unit attached to the gamma camera; however, CT imaging was not used in this study.All measurements at each gamma camera unit were performed by the authors.
Measurements were performed without attached collimators using only a plexiglass plate to protect the detector (i.e.intrinsic mode) and enhance the detection efficiency [7,12,13].The Siemens Symbia Intevo did not have a Plexiglass plate for detector protection and was operated without any substitute for the collimator.
A whole-spectrum analysis was used, where all detected counts in the chosen energy window, which was set to cover the entire spectrum, were used as the signal.The signal was collected as the total number of detected counts for both detector heads in the gamma camera software, which was easily accessible to the user.These data were referred to as processed data.The raw spectra data retrieved with special software, that is, List mode data, were only available for GE gamma cameras.These spectra were retrieved and analysed to enable a comparison of the phantom constructed in this study with the IRINA P3 phantom used by Hjellström and Isaksson [6], where raw data were used exclusively.Raw data were also used for the dead time analysis.
The energy interval for the NaI-based detectors was set to 30-510 keV, and an additional energy interval of 511-588 keV was used for the Siemens cameras because this option was available for these units.Semiconductor-based gamma cameras can only measure energies up to 250 keV, according to the manufacturer, and an energy interval of 40-250 keV was chosen [14].The field of view (FOV) of the detectors and the actual energy window set in the gamma camera software, with the energy interval representing the energy window, are listed in table 2. Five gamma camera groups were identified based on the different characteristics of the models, including manufacturer, detector crystal thickness, and detector type.The groups are 1.GE gamma camera units with 5/8 ′′ NaI crystal thickness, 2. GE gamma camera units with 3/8 ′′ NaI crystal thickness, 3. Siemens gamma camera units with 3/8 ′′ NaI crystal thickness and a similar energy window as the GE NaI-based cameras, 4. Siemens gamma camera units with 3/8 ′′ NaI crystal thickness with a Dual Window (DW) and 5. GE gamma camera units with CZT semiconductor crystals.All gamma camera units studied were included in either of these groups, and the Siemens units were included in both group 3 and 4.

The calibration geometry
Calibration measurements were performed in the supine position, with the gamma camera detector heads stationary above and below the phantom, respectively, at an angle of 180 • (figure 3).The gamma camera setting used was longitudinal setting '40' for the GE cameras (centred over the abdominal region) and '40' for the Siemens cameras, resulting in the same position of detector FOV in relation to the phantom.Energy, uniformity, and linearity maps for 99m Tc were used for all gamma camera models, as 99m Tc is a commonly  used isotope in nuclear medicine and the maps are expected to be available and updated at all units, and minimum adjustment for the personnel is needed.The distance between the detectors was fixed at 69 cm (64 cm between the plexiglass plates) for all gamma cameras.This is the standard distance for GE gamma cameras and can be easily reproduced at the Siemens units.The measurement geometry and gamma camera settings replicated the setup used by Hjellström and Isaksson [6] for the IRINA P3 phantom in the supine position.
All calibration measurements followed the same scheme: two background measurements without a phantom, three background measurements with the water-filled phantom, ten measurements with 137 Cs phantom, three background measurements with the water-filled phantom and, finally, two background measurements without a phantom.Each measurement lasted 10 min (real time).All potential radioactive sources were removed far away from the gamma camera room to prevent the detection of unwanted signals.However, some measurements had to be excluded because of different interferences from the vicinity, as described in the measurement environment section below.
Background measurements were performed without a water-filled phantom present to study the impact of different background radiation levels that each gamma camera room might have on the calibration factor.This is important if measurements are made without a background phantom available.In the case of an RN event, the background radiation level could be altered and water-filled background phantoms might not be available at each gamma camera unit for background measurements.Therefore, additional calibration factors were calculated based on the background measurements without the water-filled phantom.

Construction uncertainty
To study the uncertainty in phantom construction, five additional, repeated calibrations were conducted separately on the two GE NaI 5/8 ′′ gamma camera units, to study the deviation in the resulting calibration factors.The phantoms were deconstructed and reconstructed between calibration measurements.The five additional calibration measurements followed a shorter measurement scheme than the calibration: first two background measurements with the water-filled phantom, then three with the 137 Cs phantom, followed by two additional background measurements with the water-filled phantom.All measurements were performed in the calibration geometry and lasted for 10 min each (real time), and the deviation was calculated in comparison to the original calibration.

Generic calibration factors
Generic calibration factors based on the detector crystal size and manufacturer were calculated by computing the arithmetic mean value of the calculated calibration factors for all gamma camera units in each predetermined group (see table 2).

Dead time measurements
For higher contamination levels, that is, higher activities, it is important to consider dead time losses.Higher activities expose the detectors to more radiation, meaning that the detector must handle more pulses per second.Since the detector can only handle a specific amount of input per time frame, higher activities will result in pulses not being detected because the detector is occupied with processing the previous pulse.This is often referred to as the dead time, τ , when the detector is unable to register new pulses.The dead time is specific to each detector system and photon energy, and this has previously been studied for the Anger camera (NaI-based gamma camera), for example, 99m Tc (e.g.[15,16]).However, to the best of our knowledge, dead time analysis has not been previously performed for 137 Cs measurements.
As described by Evans and Knoll [17,18], two different models are commonly used to define dead time behaviour: paralyzable and nonparalyzable responses.Modern NaI-based gamma cameras have been observed to behave as a paralyzable system up to a certain count rate [19], and the relationship between the dead time (τ ), measured count rate (m), and true count rate (n) for a paralyzable detector response can be described by equation ( 1) [18] m = ne −nτ . ( The losses due to dead time increase with the fluence rate to which the detector is exposed, and thus, with the count rate.This means that the higher the contamination level, the higher the dead time losses.Modern GE NaI-based gamma camera systems can neither display the live time nor the dead time, only the real time.This means that the operator is unaware of dead time losses when measuring, which could result in a large underestimation of the actual activity measured at higher contamination levels, demonstrating the need for dead time analysis. A specific setup, made from a composition board, was built for dead time measurements and placed on a tripod.The FOV of the detector, 40 × 54 cm 2 , was marked on the composition board, and five source holders were positioned in each corner of the FOV and at the centre, as shown in figure 4(a).These positions, marked 1-5 in the figure, were chosen to provide as much homogeneous exposure of the detector as possible.Air levels were used in both the XY and ZY directions to ensure a perpendicular angle between the detector and the setup.The measurements were performed with one detector head to enable larger distances to the detector, and therefore, a wider range of exposure levels.The measurement geometry is illustrated in figure 4(b), with a 2 m distance between the detector and the source.Five solid 137 Cs sources were used in small bottles, each containing approximately 7 MBq.The sources are denoted A-E, with the corresponding activities at the measurement date: A = 6.85 MBq, B = 6.57MBq, C = 7.35 MBq, D = 6.41 MBq and E = 6.65 MBq.
Two different methods for dead time estimation were used: the Dual Source (DS) method and a Variant of the Decay Source (VDS) method [17,18].The count rate using an open window containing the entire spectrum (table 2) was used as the signal, and dead time analysis was performed on both the processed and raw data.Background measurements were performed using the measurement geometry displayed in figure 4, and the number of background counts was subtracted for all the measurements.The dead time was only studied for the GE gamma camera system with 5/8 ′′ NaI crystal thickness (group 1).
The DS method is based on three different measurements from two sources, S 1 and S 2 : a measurement using only S 1 , a measurement using both S 1 and S 2 , and a measurement using only S 2 .The sources should be at the same distance from the detector and should not affect the detection probability when both are measured simultaneously.Equation ( 2) was then used to calculate the dead time [15] where m 1 is the measured count rate for S 1 , m 2 is the measured count rate for S 2 and m 12 is the count rate measured for S 1 and S 2 simultaneously.To achieve as equal activities for S 1 and S 2 as possible, sources A and D (13.27 MBq) were used as S 1 and B and E (13.22 MBq) as S 2 .The measurements were performed, and dead times were calculated for three different distances between the sources and the detector (1, 2, and 2.75 m) to study the impact of different irradiation levels.The four sources were placed in the middle position of the setup (position 3), i.e. centred at the FOV of the detector.
The second method was the VDS method.The decay source method is used for radioactive elements with a relatively short half-life, enabling measurements of different activity levels in the same setup within a short time frame.Because 137 Cs has a half-life of approximately 30 years, this method is not practically feasible; instead, the distance between the detector and the sources was varied to observe the response to different irradiation levels, which was called the VDS method.The sources were placed in five holders in the setup to achieve homogenous irradiation of the detector and measured at 14 different distances: 30, 40, 50, 60, 75, 90, 100, 125, 150, 175, 200, 225, 250, and 275 cm.The sources were placed in the following positions: A at position 5, B at position 4, C at position 1, D at position 3, and E at position 2, see figure 4(a).The irradiation level of the detector at each distance was calculated by a point kernel approach, using MATLAB (The MathWorks Inc., USA) by analytically calculating the dose rate from all five sources separately at each point of a 400 × 540 matrix representing the FOV of the detector, using the dose rate of 137 Cs at 1 m and the inverse-square law.The mean value of the calculated dose rate experienced by the detector at each distance, in this study referred to as 'Theoretical dose rate' , was used in the analysis as a replacement of the activity, which is normally used for the decay source method.The MATLAB script also enabled theoretical visualisation of the homogeneity of the irradiation at each distance by studying the standard deviation in the matrix.The result was assumed to follow the relation between the measured and true count rates described by equation (1), and the dead time was simply derived from the derivative of equation ( 1) as a function of n, where the largest measured count rate was assumed to be the global maximum.
Four different dead time values, τ , were calculated: three for different distances using the DS method and one using the VDS method.The curves describing the paralyzable detector response were calculated for each τ using equation ( 1): The count rates measured using the VDS method reflect the detector response for different exposure levels; however, the true count rate, n, for each measured count rate is unknown.To estimate the accuracy of the different calculated τ , the true count rate for each measured count rate was calculated by iteratively solving equation ( 1) for each τ .
The fractional loss of the signal due to dead time effects, ε, can be calculated using equation ( 3) [20].The measured count rates for different levels of dead time losses were calculated using the dead time τ from the VDS method (3)

Calculations
The calibration factor, CF (cps kBq −1 ), was calculated using equation ( 4), where C Cs137 is the mean number of counts from the measurements with the phantom containing 137 Cs, with a measurement time of t Cs137 ; C water is the mean number of counts from the measurements with the water-filled phantom, with a measurement time of t water , and A Cs137 is the decay-corrected 137 Cs activity in the phantom.The same energy window and measurement time were used for C Cs137 and C water .This equation was also used for the calibration factors based on an empty background by replacing C water and t water with C BKG and MDA was calculated using equation ( 5) with a coverage factor of k = 1.645 [21] The uncertainty of the calibration measurements was calculated as the combined uncertainty of the activity, background measurements and 137 Cs measurements.The uncertainties of the background measurement and 137 Cs measurements were calculated as the standard deviation of the total number of counts, assuming Poisson statistics.
The calculations of the committed effective dose (CED), which can be estimated from the corresponding MDAs for the different gamma camera groups, and the CED that can be estimated from the maximum count rates for different dead time losses, were based on dose coefficients from ICRP Publication 119 for members of the public [22].Dose coefficients for inhalation of Type M aerosols, with an activity median aerodynamic diameter (AMAD) of 1 µm, for 137 Cs and 15 year olds were used, as the 49 kg calibration phantom is representative for the 15-year-olds adolescent reference values (56 and 53 kg for male and female respectively) from ICRP Publication 89 [23].The CED was calculated for measurements at 1 d, 3 d and 1 week after intake, by using the bioassay functions from ICRP Publication 137 for abovementioned specifications [24].

The measurement environment
Measurements without a collimator in combination with whole-spectrum analysis render the detectors highly sensitive.This is an advantage when measuring in a steady/unchanging environment but can be a great disadvantage when unwanted sources of ionising radiation are present, as discussed by Hjellström and Isaksson [6].Examples include injected patients moving in corridors, vials of short-lived radionuclides in surrounding rooms, trash cans and laundry bags, residual technegas from patient inhalation in adjacent rooms and 99m Tc generators being replaced.Close-by CTs being operated can result in showers of low-energy photons hitting the detectors.This results in a transient increase in the background, which can be difficult to notice unless the spectrum is continuously monitored by personnel.It is important to pay close attention to the abovementioned contributors because the detector crystal thickness of the medical gamma camera is optimised to measure low-energy photons (in comparison with e.g. 137Cs) and high activity levels are used in nuclear medicine procedures, meaning that a small drop of 99m Tc on the floor can significantly contribute to the measurement.This contribution varies between measurements owing to the short half-life.The whole-spectrum analysis includes all detected photons in the spectrum, indicating that the method is sensitive to these unintended contributions.

Calibration factors and MDA
Table 3 shows the gamma camera ID code (GC ID), location, gamma camera model, detector type, calibration factor with and without the water-filled phantom, and MDA for 10 min and 1 min measurement times based on the calibration with a water-filled phantom for each gamma camera unit.The combined uncertainties of the calibration factors are indicated in parentheses.GC IDs 1 and 2 were included in group 1, GC IDs 3-14 in group 2, GC IDs 15-19 in group 3 and 4, and GC IDs 20-22 in group 5.
The calibration factors for the dual energy window for the Siemens units, group 4, are not listed in table 3.If a dual energy window is used the calibration factor for GC ID 15 is 67.1 (3.6%) instead of 64.0 Table 3.The gamma camera ID code (GC ID), the location, the gamma camera model, the detector type, calibration factor with and without the water-filled phantom, the MDA for 10 min and 1 min measurement time for each gamma camera unit.The combined uncertainties for the calibration factors are displayed in the brackets.Figure 5 shows the calibration factors for calibration with the water-filled phantom for the processed data, where each gamma camera unit is noted with the GC ID in table 3. The different gamma camera groups are marked with different colours: orange for group 1, blue for group 2, pink for group 3, yellow for group 4, and green for group 5. Dotted lines indicate the mean value of each group and are marked with the corresponding colours.Note that the two calibration factors of GC ID 15, one obtained with a single energy window and the other with a dual energy window, were not included in the mean value for group 3 and 4. The mean value including this GC ID 15 is 53.0 cps kBq −1 for group 3 and 56.0 cps kBq −1 for group 4.
Figure 6 shows the calibration factors resulting from the analysis of the raw data, that is, the spectrums retrieved from the GE NaI gamma cameras (group 1 and 2).

Construction uncertainty
The repeated calibrations for the two GE NaI 5/8 ′′ gamma camera units, GC ID 1 and 2, resulted in a maximum deviation below 4% between all repeated calibrations and the original calibration factors, displayed in table 3.
The generic calibration factor based on the raw data was 62.3 cps kBq −1 for group 1 and 54.7 cps kBq −1 for group 2.
Table 4.The mean value for the different gamma camera groups, which is proposed as generic calibration factors for the specific characteristics of the group.The outlier for Siemens GC ID 15 is not included in the generic calibration factor.The generic calibration factor is based on the processed data.

Group
Gamma camera Generic calibration factor (cps kBq

Dead time measurements
Table 5 lists the calculated dead time, τ , for the different methods used.Dead time measurements have been carried out using GC ID 1 (group 1).The homogeneity of the theoretical dose rates for the VDS method is displayed as standard deviations, along with the mean value of the theoretical dose rate and the measured count rate (processed data), in table 6.
Figure 7 shows the measured count rate for different theoretical dose rates, i.e. distances between the sources and the detector, of GC ID 1.These are the results of the VDS method, which are presented in table 6. Green dots, 'R-Measured count rate' , represent the raw data and the yellow dots, 'P-Measured count rate' , represent the processed data.The results appear to follow the behaviour of the paralyzable model described by equation (1).
Figure 8 shows the relation between the measured count rate and the calculated true count rate based on the different τ , with a linear curve corresponding to no dead time losses (m = n).Two curves, one for processed data ('R' , solid line) and one for raw data ('P' , dotted line), were derived for each τ with equation ( 1).The true count rate for each measured count rate (table 6, figure 7) was calculated by iteratively solving equation ( 1) for each τ and is noted as 'Fm' in the figure.1) for each dead time value, are noted as dots with different shapes and colours, depending on the dead time value.'DS' represents the dual source method and 'VDS' the variant of the decay source method.

Committed effective dose
Table 7 shows the CED, for inhalation of Type M aerosols (AMAD = 1 µm) for 137 Cs and 15-year-olds, that can be estimated based on the MDA for measurement times of 1, 2, 5, and 10 min for different detector groups, for measurement at 1 d, 3 d and 1 week after intake.The MDA and CED are displayed in intervals, with the highest and lowest MDA for each group and the CED to which these values correspond.The MDA is dependent on the background radiation level and therefore varies between different units.
Table 8 lists the calculated count rate expected to be measured at different levels of signal loss due to dead time and the activity that the count rate corresponds to in the measurement geometry used in this study.The CED that the activities correspond to, for inhalation of Type M aerosols (AMAD = 1 µm) for 137 Cs and 15-year-olds, for measurement at 1 d, 3 d and 1 week after intake, is also shown.This is based on the generic calibration factor for gamma camera group 1 (GE, NaI, 5/8 ′′ ) for the processed data, meaning that it is specific for the measurement geometry used in this study.The count rate is for two detector heads, and the dead time used is 3.17 µs (VDS method, processed data).Note that 63% dead time loss is the maximum The count rates at different levels of fractional dead time losses and the estimated activity levels in the calibration geometry for gamma camera group 1, based on the generic calibration factor for the processed data.The estimated CED that the activity corresponds to, for inhalation of Type M aerosols (AMAD = 1 µm), for 137 Cs and 15-year-olds, for 1 d, 3 d and 1 week after intake, is also displayed.

Calibration factors and MDA
The generic calibration factors for the five predetermined groups (table 4) reflected the individual calibration factors for the gamma camera units with relatively low standard deviations, and there were no apparent discrepancies between the different gamma camera models included in the same group.Therefore, the generic calibration factors are considered applicable to other gamma camera units that match the characteristics of the groups and are not included in this study.However, this can only be observed for the models included in this study.The applicability of the CZT semiconductor-based gamma camera, which, to the best of our knowledge, has not been previously investigated for 137 Cs estimation, was shown to be feasible.The generic calibration factor for the crystal thickness 3/8 ′′ did not differ substantially between the GE and the Siemens models and calibration of additional Siemens units would reveal possible discrepancies.However, the Siemens units have somewhat lower calibration factors, which could be due to the lower limit of 35 keV in comparison to 30 keV for the GE units or to the slightly smaller FOV.The Plexiglass plate, used only for GE models, could have acted as a scatterer to some degree and altered the detection efficiency owing to lower photon energies exposing the detector; however, Monte Carlo simulations need to be performed to verify this.The Siemens units could measure a higher energy range (511-588 keV), and using this dual window, a more sensitive calibration factor was obtained.It was possible to choose an even larger energy window (588 keV ± 13%, see table 2), which would partially include the photo peak of 137 Cs, but the generic calibration factor including this extended energy window (53.24 cps kBq −1 ) indicates that this is not the case, and the stated upper limit of 588 keV by the manufacturer seems more accurate [25].This was also observed for the CZT units, where the upper limit of the energy window could be set higher than 250 keV; however, there was no indication of an increased count rate.
There is mainly one outlier in the calibration measurements, the Siemens Symbia Intevo (GC ID 15), with a calibration factor of 64.0 ± 3.7% cps kBq −1 for group 3 and 67.1 ± 3.6% cps kBq −1 for group 4.This calibration factor was significantly higher and was therefore excluded from the generic calibration factors of groups 3 and 4. Further investigation is required to explain the origin of this discrepancy.GC ID 13 could be interpreted as an outlier as well, owing to its low calibration factor in comparison, but it was decided to be included in the generic calibration factor.Removing this data point will increase the generic calibration factor by 0.8%.The MDAs were low; hence, activities resulting in a low CED could be estimated even for short measurement times.This is due to the large detectors, the use of whole-spectrum analysis, and a stable background.This is important to keep in mind because an elevated background, which might exist following an RN event, increases MDA.MDAs correspond to an estimated CED below 77.6 µSv, even for a 1 min measurement, performed 1 week after the intake.
The choice of phantom size was based on a desire to keep the size and weight manageable for transport and handling at the hospital since the variation in calibration factor between phantom sizes relatively low [6].The calibration factor from Hjellström and Isaksson [6], based on raw data from measurements of the same gamma cameras as GC ID 1 and 2, was 74.5 cps kBq −1 for the IRINA P3 phantom.The generic calibration factor based on measurements of the water phantom in this study, which were constructed to mimic the IRINA P3 phantom, was 62.3 cps kBq −1 for the raw data (group 1).These calibrations were determined using the same measurement geometry and calculated from the raw data retrieved from the GE software.The IRINA factor was approximately 19.6% higher, which was a significant difference, most probably owing to the differences between the phantoms.However, this discrepancy between phantoms was expected and highlighted the difficulty in calibrating whole-body measurements.For example, the European Commission Technical Recommendations for Monitoring Individuals for Occupational Intakes of Radionuclides recommends assuming a scattering factor of 1.2 for lognormal uncertainty distribution, which corresponds to a relative uncertainty of 20% [26].
The repeated calibration measurements performed on the two GE Discovery NM/CT 670 Pro (group 1), with the purpose of investigating the construction uncertainty of the phantoms at each gamma camera unit, resulted in a maximum deviation below 4% from the original calibration, indicating that there was no need to add an additional uncertainty due to construction of the phantom at each unit.

The measurement environment
Some measurements could not be used due to a disturbed background caused by e.g. a CT being operated, administration of tecnegas in an adjacent area, or a patient treated with a radiopharmaceutical walking by.These measurements were repeated if possible, but this was not the case for all of them.This resulted in fewer measurements being analysed for some gamma camera units.However, this should not significantly affect the calibration factors to a greater extent.In some cases, the gamma camera room was contaminated with 99m Tc during calibration; in these cases, the background radiation level was decay-corrected.
As previously mentioned, the resulting MDA is relatively low; however, this requires a stable background because we are dealing with the entire spectra and quite a few net pulses.The nuclear medicine department, where gamma cameras are situated, is also an environment that is prone to fluctuating backgrounds because of its daily use.
Comparing the calibration factors based on a phantom background with those based on an empty background, all calibration factors (except GC ID 22) were higher for the latter.This shows that the phantom background is higher than the empty background, indicating that the phantom alters the background radiation.The phantom could also shield the detector from background radiation in some cases, as observed in GC ID 22 and to some extent in GC ID 18.For GC ID 18 there was a well-defined 99m Tc contamination, which the phantom shielded.
Comparing the calibration factors for an empty background, ranging from 70.6-210 for group 2, with the calibration factors for a phantom background, ranging from 47.8-55.5, it is obvious that it is important to use a background phantom for the subtraction of background radiation.In practical use of the proposed method, it is important that a measurement of a human subject is used as background, otherwise the presence of 40 K in the subject will give a false contribution to the measured activity.Note that the 40 K contribution was not included in the derived calibration factors in this study.However, this only affects low-activity levels because the efficiency of detecting 40 K has been observed to be low for gamma cameras [6].It is important to ensure that the personnel used for this background measurement are not contaminated; otherwise, the activity may be underestimated for the subject.

Dead time analysis
The calculated dead time, τ , based on the DS method varies for different distances from the detector, and thus, different fluence rates.This was also observed by Heemskerk and Defrise [16], Silosky et al [19], and Adams et al [15] at different activity levels.The longest estimated dead time was obtained using the VDS method, which is regarded as the most accurate method because it is based on the largest actual measurable count rate.Figure 8 shows the relation between the measured count rate and the calculated true count rate for different dead time values for the paralyzable detector model (equation ( 1)).It can be observed that the dead time values from the DS method suggest that a larger count rate could be measured than we observed with the VDS method, displayed in figure 7. The measured count rates shown in figure 7 indicate that this was not the case.
The DS method is based only on three different dose rates related to the activity level and does not consider the largest measurable count rate.Silovsky et al [19] stated that the combined source S 12 in the DS method should preferably be as close to the maximum measurable count rate as possible.In this study, we measured a count rate of 110 kcps when the combined source from the DS method was placed at its closest position to the detectors (100 cm), which was 90% of the observed maximum count rate from the VDS source method.This registered count rate resulted in a 9% lower estimation of the dead time compared with the VDS method.For 200 cm and 275 cm distances, the estimated dead times were 27% lower and 35% lower, respectively.This indicates that a closer distance or higher activities of the sources, resulting in a measured count rate equal to the maximum, would yield results similar to the VDS method.In conclusion, the DS method should be used with caution because it does not necessarily reflect the correct dead time of the detector system for the measured radionuclide.Heemskerk and Defrise [16] discussed the uncertainties in using the DS method and recommended a variation of the graphical method discussed by Knoll [18].However, the DS method is more accessible and reflects dead time losses relatively well for lower count rates.For a measured count rate of 91.3 kcps, processed data, there was only a deviation of 6.8% between the calculated true count rate between the VDS method and the DS method at 100 cm, which gave the S 12 closest to the maximum measurable count rate.For the DS method with a 275 cm distance, the deviation was 24.1%.
Unlike Heemskerk and Defrise [16] and Silosky et al [19], who observed that modern gamma cameras only follow the paralyzable detector model up to the maximum observed count rate, our results for the observed count rate appear to follow the paralyzable detector model even after the maximum.This could be energy dependent, and thus count rate dependent, since we only observed count rates up to approximately 120 kcps, whereas Heemskerk and Defrise [16] observed a maximum of approximately 200 kcps, and Silosky et al [19] observed a maximum of 300-400 kcps, both for 99m Tc, with uncollimated detectors and open energy windows.However, both these studies were performed on a 3/8 ′′ crystal thickness, for broader energy windows, and other gamma camera models (Phillips, Siemens and an older GE model).This makes comparison difficult; however, Desy et al [27] studied the dead time for a GE Discovery 670 Pro 3/8 ′′ , using collimators and 99m Tc, and obtained a paralyzable detector response after the maximum observed count rate similar to this study, with a maximum of 200 kcps using an open energy window.
The dead time analysis was only performed for the NaI detector with the crystal thickness of 5/8 ′′ , and no measurements were performed on the CZT nor the 3/8 ′′ (GE or Siemens).However, the calibration factors suggest that the NaI 5/8 ′′ gamma camera will experience more counts per second per activity content and will likely experience a higher dead time for a specific activity than the other models, if identical signal processing system is assumed.Therefore, the dead time analysis done in this study may deviate for the 3/8 ′′ gamma cameras, and the dead time has also been shown to differ between manufacturers [15,19].Dead time measurements were performed for only one of the detector heads, detector 2, and there might have been a slight variation between the detector heads, which could affect the estimated activity and CED for the calibrated geometry.It is also important to consider that the dead time, and thus the dead time losses, might change for measurements with a phantom or a subject because these measurements change the characteristics of the spectrum owing to more scattered radiation.
The homogeneity of the theoretical dose rate experienced by the detector, calculated using MATLAB, is shown as a standard deviation in table 6.The standard deviation was largest for the closest distance, where it was 4.9% and lowest for the greatest distance.The deviation of the homogeneity is considered small but could affect the results to a smaller degree.
The hypothetical maximum detectable activity of 137 Cs, calculated by Hjellström and Isaksson [6] based on a maximum count rate of 230 kcps for 99m Tc and one detector, as stated by the manufacturer, was 35-49.2MBq.However, this was only calculated for a distant sitting geometry, but this maximum count rate can be compared with the maximum count rate (one detector) calculated for 137 Cs in this study, which was approximately 124 kcps (raw data).This is a significant difference and indicates an increase in the dead time with increasing energy.A plausible explanation for this difference is that photons from 137 Cs produce a larger number of secondary charges in the detector and that this compensates for the slightly lower probability for photoelectric interaction.

Recommendations following an RN event
Following an RN event, there may be a great need for measurement capabilities, and medical gamma cameras have proven to be useful tools for the estimation of 137 Cs.Generic calibration factors are recommended for use in the gamma camera models included in this study, and potentially for other gamma camera models that suit the characteristics of different generic factor groups.The results from the processed data are intended to enable quick analysis of the estimated activity in the subject, since the usage of the raw data is more time-demanding, requires certain software, and is only applicable to GE NaI gamma cameras.
Whole-spectrum analysis should be used with caution because multiple contaminants in the subject will affect the results, and this can be the case if measurements are performed shortly after a release.However, experiences from Japan after the Fukushima accident show that a need for assessment of internal contamination is not restricted to the first months after an accident but can remain for several years [3].Simulations of the effective dose rate from Chernobyl fallout in Sweden show that the contribution from internal exposure due to the two Cs-isotopes will exceed the contribution from external exposure due to short-lived radioisotopes about 4 months after the fallout, and after about one year 137 Cs will give the largest contribution [28].It can thus be expected that the internal contamination measured during the recovery phase will be dominated by the two Cs-isotopes.
The CED coefficient from these two radioisotopes is the same, except for insoluble forms, e.g.fuel fragments [24].Thus, the measurements could still be used to approximately estimate the CED from ingestion.Although the detector response for 134 Cs will not be equal to that for 137 Cs, the deviation between the calibration factors is expected to be about a factor of 2, based on the weighted mean photon energy for 134 Cs at about 700 keV and calibration data from previous measurements [6].During the time period where whole-body measurements are needed (recovery phase and later follow-up), the abundance of other radionuclides in food is expected to be known from other kinds of measurements and used to estimate the contribution to the gamma camera measurements.It is also recommended to use non-contaminated personnel for background measurements to include the 40 K contribution and to exclude the need for a background phantom.
When dead time losses occur, they must be corrected or considered.If a subject is being measured, it could be deceiving to just observe the measured count rate since higher activity levels could give the same count rate as lower, see e.g. the theoretical dose rate of 10 µSv h −1 and 1 µSv h −1 in figure 7, both resulting in approximately the same measured count rate.To control whether the activity level is higher or lower, the subject can move slightly further away from the detectors between measurements.If the count rate increase, it indicates a higher activity level; if the count rate decrease, it indicates that the subject has a lower activity level.However, this can be difficult to achieve with the measurement geometry used in this study, in which the subject lies in a supine position between the detectors.For higher activities, it could be easier to use the distant sitting geometry used by Hjellström and Isaksson [6], where the chair could be moved backwards away from the detector to check whether it had a higher or lower contamination level.
An alternative to measuring without a collimator, when a higher intake is suspected, could be to calibrate 137 Cs using a High Energy General Purpose (HEGP) collimator to decrease the sensitivity.This could enable the estimation of higher CED than 263 mSv (measurement at 1 week after intake), which is the highest CED that could be estimated in the geometry used in this study, which could be requested in a large-scale accident.However, all gamma camera units might not have a HEGP collimator available, whereas all gamma cameras can be measured without a collimator, and utilisation of the distant sitting geometry introduced by Hjellström and Isaksson [6] could thus be a better alternative.

Conclusions
22 gamma camera units at 15 hospitals in southern and western Sweden were calibrated to estimate 137 Cs in a supine, static geometry.MDAs for 1, 2, 5 and 10 min measurement time has been calculated and corresponds to an estimated CED below 77.6 µSv, for inhalation of Type M aerosols (AMAD = 1 µm) for 137 Cs and 15 year olds and measurements performed 1 d, 3 d and 1 week following an intake.The semiconductor-based gamma camera, CZT, was shown to be feasible for estimating 137 Cs contamination.The calibration should be used for 137 Cs contamination only, as the activity, and thus the CED, otherwise may be incorrectly estimated.The method is therefore intended to be used in the recovery phase.
Generic calibration factors for five different groups based on certain gamma camera characteristics were proposed.Group 1 is NaI-based gamma cameras from GE with a crystal thickness of 5/8 ′′ and had a generic calibration factor of 60.0 cps kBq −1 .Group 2 is NaI-based gamma cameras from GE with a crystal thickness of 3/8 ′′ and had a generic calibration factor of 52.4 cps kBq −1 .Group 3 is NaI-based gamma cameras from Siemens with a crystal thickness of 3/8 ′′ and a similar energy window as the GE NaI-based cameras and had a generic calibration factor of 50.3 cps kBq −1 .Group 4 is NaI-based gamma cameras from Siemens with a crystal thickness of 3/8 ′′ and a dual energy window and had a generic calibration factor of 53.2 cps kBq −1 .Group 5 consisted of semiconductor-based gamma cameras from GE with a CZT detector and had a generic calibration factor of 48.4 cps kBq −1 .
A dead time analysis has been performed on the GE Discovery NM/CT 670 Pro, with a crystal thickness of 5/8 ′′ , and the results suggest a dead time of 3.17 µs for processed data.The measurable count rate related to the activity and the CED that can be estimated were calculated for different levels of dead time losses with a maximum detectable count rate of 232 kcps, corresponding to a signal loss due to dead time of 63%, an activity of 3.82 MBq and a CED of 108, 210 and 263 mSv for measurement at 1 d, 3 d and 1 week after intake correspondingly.

Figure 1 .
Figure 1.One of the two identical water-filled container phantoms constructed for this study.This phantom is filled with only water and marked with AW-LW for identification of each water container.

Figure 2 .
Figure 2. The location of the hospitals and the number of gamma camera units included in the study at each site, marked with different shades of green and sizes depending on the number of units.Two of the three active nuclear power plants in Sweden are marked with a radioactive symbol.The six regions are colour coded as Västra Götaland (teal), Halland (red), Kronoberg (orange), Kalmar County (pink), Blekinge (purple) and Skåne (yellow).The map was retrieved from the software Google Earth Pro (Google, USA)[11].

Figure 3 .
Figure 3.The calibration geometry, with the two detector heads stationary above and below the phantom, at an angle of 180 • .The water container phantom with water only is positioned in the calibration geometry.

Figure 4 .
Figure 4.The set-up for the dead time measurements.Figure (a) shows the composition board setup for the sources, with the five position holders in place and five sources present.Figure (b) shows the measurement geometry, with the setup placed at a 2 m distance from the detector.

Figure 5 .
Figure 5. Calibration for all gamma camera units, noted with respective GC ID.Mean values for each generic calibration factor group, marked as a dotted line in corresponding colour.The mean for group 3 and 4 (pink and yellow) does not include the outlier GC ID 15.DW denotes the Dual window for Siemens 3/8 ′′ .

Figure 6 .
Figure 6.Calibration factors for the raw data from the GE gamma camera model, noted with respective ID code.Mean values for each generic calibration factor group, marked as a dotted line in corresponding colour.

Figure 7 .
Figure 7.The measured count rate, m, of the GC ID 1 for different theoretical dose rates.

Figure 8 .
Figure 8. Curves based on equation (1), with the four different calculated dead time values.The raw data (List mode) is represented by dotted, darker lines and the processed data by solid, lighter lines.The measured values, fitted into the curve by iteratively solving equation (1) for each dead time value, are noted as dots with different shapes and colours, depending on the dead time value.'DS' represents the dual source method and 'VDS' the variant of the decay source method.

Table 1 .
Gamma camera models included in the study, the number of units, the crystal thickness (or detector type) and the location of the models, with the number of gamma cameras at each location in brackets.

Table 2 .
The gamma camera model groups, the FOV of the detectors, the energy window used for the measurements and the actual energy interval that these energy windows represent.

Table 5 .
The dead time, τ , calculated for the raw data and the processed data for the different methods used.

Table 6 .
The measured count rate, m, (processed data), the mean value of the theoretical dose rate and the standard deviation, 1σ, (%) of the dose rate that the detector is theoretically exposed to.

Table 7 .
The MDA and CED for the different detector groups and measurement times.The CED is for inhalation of Type M aerosols (AMAD = 1 µm) for 137 Cs and 15-year-olds, for measurement at 1 d, 3 d and 1 week after intake.