Abstract
Background/Aim: A mobile system allowing hospital medical personnel to prepare for the administration of radiation mitigators prior to receiving casualties is desirable. Materials and Methods: We evaluated a portable spectroscopic personal radiation detector for use as an ambulance-based unit for early detection and identification of gamma radiation. We tested the sensitivity, time-to-identification, and radionuclide identification accuracy rates, change in detector response to vehicle operation, interference from cardiac equipment, and internal versus external radiation source location. Results: We detected radiation sources in each of 119 trials using a humanoid phantom in a moving ambulance with a primary radionuclide identification accuracy of 96%. Typical identification time was around two minutes (149±95 s). Conclusion: Our observations suggest this mobile system is a potential pre-hospital arrival tool allowing for rapid preparation of radiation mitigators.
With increasing numbers of potentially available radiation mitigators (1-5), hospital personnel need to deliver appropriate agents quickly to known or suspected radiation casualties. A radiological emergency (e.g. dirty bomb, fission bomb, nuclear reactor accident, willful disposal of radiation-emitting isotopes) presents unique dangers to the public and challenges the medical response team to act efficiently and accurately in handling casualties. Early detection might effectively limit the number of people exposed after radionuclide detection. Rapid identification of radiation contamination by first responders would also allow for quicker treatment and may help prevent the spread of radioactive material within a hospital.
A challenge to the use of a radiation detection device is accuracy. It is paramount to be able to distinguish between the presence in the body of one or more planned medical gamma-emitting radionuclides used in nuclear medicine and clinical radiation brachytherapy from other pathogenic exposures; the former are not an emergency, and the latter might indicate that a radiological accident or terrorist event has occurred.
The use of portable gamma detectors is part of field surveys for radiation and fallout detection (6-11). Commercial detector systems are available for use by law enforcement bodies and homeland security (6, 7). These systems detect radiation sources outside of an ambulance and have large sensitive detectors for surveying large land areas for contamination or lost sources (12-16). These features limit use in emergency medical care.
We sought to examine the utility of the Emergency Radiation Detection and Notification System (EDNS) for managing potential victims of radiation from a terrorist event or nuclear accident (9, 17). Previous studies have demonstrated the effectiveness of using NaI detectors placed in the ambulance bay, Emergency Department entrance, nursing station, or hospital corridor for detection of the movement of non-medical radiation sources (17).
A handheld spectroscopic personal radiation detector (SPRD) was previously evaluated in a stable non-moving laboratory setting to test initial detector placement parameters and shielding effects of radiation sources placed in a humanoid phantom (9). We now extend this work to the setting of a moving ambulance and in simulation of likely operational conditions.
The SPRD has identification capability while being small and rugged enough to incorporate into ambulance equipment (9). The present trials and measurements were designed to evaluate the use of this device in an ambulance with simulated pickup and transport of a patient (humanoid phantom) that was experimentally modified to contain each of several radioactive materials. We used a radiological phantom (18) and bolus for simulation of a clothed patient. The setting of the ambulance environment and potential interference from other electronic equipment in the ambulance, potential impacts of vehicle motion, shielding and detector distance from radiation source effects resulting from the limited space, and other influences on the operation of the detector were all monitored. Several factors that would impact field use were also evaluated. These included: i) time taken to detect a radiation source on the phantom, ii) time taken to identify the radionuclide(s) present, and iii) the accuracy of the identification in trials involving one compared to two known radionuclides.
Materials and Methods
Setting. The experimental procedure modeled the detection and identification of an unknown radionuclide contamination present on a patient being loaded into an ambulance equipped with an SPRD (9, 17) (Figure 1). We measured the time required to detect and identify the radionuclide(s) present, and the accuracy of the identification as the primary outcomes. Additional measurements included identification trials to check for interference from ambulance equipment including defibrillator discharge, background count rate and dose rate measurements, and a survey of detector response for different source location and detector mounting location combinations.
For all isotope identifications and measurements, we prepared a radiological phantom (18) with one or two sources placed on the chest, sufficiently far away from the radiation detector - the Interceptor (about 10 m), so as not to change the detectable background gamma count rate. For each trial, we loaded the phantom into the ambulance and then a timekeeper measured the time to detection and time to identification, along with the nuclides reported by the device, and confidence levels of its result. We removed the phantom from the ambulance after each trial and prepared for the next trial. Separate sets of trials were carried out with and without up to 3 cm of Superflab bolus (Mick Radio-Nuclear Instruments, Inc., Mt. Vernon, NY) placed on top of the source, which simulated external and internal contamination.
Ambulance and equipment. We used a fully equipped advanced life support (ALS) ambulance. The ambulance was a large Type I with a patient compartment mounted onto a heavy truck chassis (Figure 1b). We discharged the cardiac monitor/defibrillator on an ALS training mannequin up to its maximum charge of 360 J to test for possible interference with the radiation detector placed nearby.
We obtained all data over two consecutive days. The ambulance driver assisted with loading and unloading the gurney and operated the cardiac equipment, radios, and drove the ambulance for the vehicle motion studies. We secured the radiological humanoid phantom torso to the gurney using the attached belts (18). The route used for motion studies was approximately 0.4 km around the perimeter of the lot.
Radiation detectors. The Interceptor™ is a handheld SPRD manufactured by Thermo Scientific (Thermo Fisher Scientific Inc., Pittsburgh, PA) (17). The device measures 68×122×30 mm3 (W×L×D), with a 5.6 cm backlit LCD display in a rugged case (Figure 1c). It uses three cadmium-zinc-telluride (CdZnTe) sensors, two for monitoring the gamma dose rate and one high-resolution detector for nuclide identification. The two dose-rate sensors were each 15×13×5 mm3 with a sensitivity (137Cs) of 1.5 cps per μR h−1 and a dose rate range response of 5 μR h−1 to 100 mR h−1 (±30 %). The high-resolution CdZnTe sensor volume was 7×7×3.5 mm3, attached to a 1024 channel analyzer with an energy resolution (137Cs) 2.4-3.5% (crystal-dependent) and range 30 keV-1.5 MeV.
We mounted the detector at one of each of three locations: ceiling, lower wall near the gurney, or upper wall near the radio and oxygen supply.
While the Interceptor supported radio communication via the bluetooth protocol, no radio equipment existed for these trials. We logged data from the device's LCD screen manually and recovered other data from the accompanying software via USB connection to a PC.
Radionuclides. We used barium-133, cesium-137, and cobalt-60 (Table I). These industrial gamma emitters have been used in previous tests of the hospital EDNS system and were used in the present experiments (9, 17). The isotopes used represent candidates detected after an industrial contamination and also are sources that might be used in a terrorist radiation dispersal device (19). This set of radionuclides yields good energy coverage for testing the detector, with X-rays and gamma ray energies ranging from 80 keV to 1.33 MeV. The activities were 3.40, 7.07, and 3.18 MBq respectively for the sealed barium-133, cesium-137, and cobalt-60 sources used (Table I). These activities are comparable to the National Regulatory Commission (NRC) annual intake limits (19) and are well below International Atomic Energy Agency (IAEA) D-values guidelines for what constitutes hazardous quantities for emergency response situations. Being of relatively low activity, these sources present a good test of the minimum sensitivity required for us to determine whether we had an accurate detector.
We sought to develop conditions to prevent false alarms from the detection of clinical isotopes. While most of the radionuclides found in hospitals and currently used in clinical medicine have short half-lives, measured in minutes or hours, several (e.g. Tl-201, I-131) can remain detectable to sensitive detectors for 1-3 months; these would be found in an ambulance if an exposed individual received a planned treatment prior to the need for any additional therapeutic intervention. These therapeutic clinical isotopes if present in the body currently present a detection challenge for radiation detectors at airports and geographic borders (11) as they may set-off what would be a false alarm. Therefore, we chose to include several of these clinical isotopes as controls allowing us to assess and identify the transport of patients previously treated with nuclear medicine. Activities of administered radionuclides used in nuclear medicine also may vary by procedure but may be more than 100 MBq (15).
As controls for the present studies, we used examples of the clinical background gamma emitters seen in the hospital environment, an NaI-based system installed in the Emergency Department of our hospital was the site for studies in which we detected: 99mTc, I-131, F-18, In-111, Sm-153, Tl-201, Xe-33, I-123, Ga-67, Rb-82, and Sr-89. 99mTc, I-131, and F-18 data was collected during a one year period for our prior study (9, 17). I-125, with a half-life of almost two months, is of concern for detection by detectors deployed in public places. Low-dose rate brachytherapy seeds (Cs-137, Pd-103, and I-125) are also candidates for detection (9, 17).
We chose I-125 in the form of sealed brachytherapy seeds for the control medical radiation source. This isotope was ideal due to its usable lifetime, safety in sealed form and common availability. I-125 also emits at the lower limit of the Interceptor's range, providing a good test of how the identification algorithm would handle low-energy sources that are in its library.
Preparation of humanoid phantom and bolus. Our prior study showed the Interceptor identification software was susceptible to misidentification due to energy-dependent shielding effects (9). To simulate human tissue, we used a Rando® humanoid phantom (The Phantom Laboratory, Salem, NY) and Superflab bolus (Mick Radio-Nuclear Instruments Inc., Mt. Vernon, NY, USA) (Figure 1a). The bolus material was necessary because the sealed sources were too large to place inside the cavity in our phantom. By varying the number of layers of bolus on top of the radioactive source, the attenuation of gamma rays could be measured, simulating internal contamination at different tissue depths.
Isotopes used in experiments with comparison to Nuclear Registry Commission occupational annual limits and International Atomic Energy Agency emergency danger values. Each of the five isotopes listed were placed on the surface of the phantom as described in the Materials and Methods section. The activity, half-life, and characteristics of each isotope are shown.
The distance from the source to detector was between 64 and 196 cm depending upon the relative placement of the source and detector. We surveyed dose rate and count rate with the source in each of three locations, representing the feet, chest and head of a patient on the ambulance gurney. We also mounted the detector on the wall near the gurney's head end, the ceiling above the chest, and on the wall near the radio and gas ports (Figure 1c). The ceiling location had the least obstructed view to the sources and the most consistent distance between source locations. The scenarios of primary concern for internal contamination are ingestion and inhalation, represented by distances between 70 and 120 cm in our setup. Table II details the distance measurements for each of the configurations tested.
Experimental protocol. We performed a total of 119 trials, 54 using a single radionuclide and 65 using two radionuclides, pairing I-125 with each of several industrial sources. Eight measurements used Gd-153, a radionuclide not present in the Interceptor's software library. This allowed us to assess how the device and its identification algorithm responded to an ‘unknown’ source.
Ten of the trials fully simulated the emergency response by driving the vehicle after loading the phantom patient (Figure 1a and b), then driving around a standard hospital parking lot (Figure 1d) at a speed of 24-40 kph (no people present in the parking garage). The detector (Figure 1c) identified the radionuclide(s) present or was shut off after reaching the maximum planned experimental data-acquisition time (30 min). The time to identification reported by the device included processing time and was typically greater than the true acquisition time (live time) by 5-10 sec, found by comparing the log files on the device to our recorded data.
Results
The Interceptor SPRD promptly detected radiation from each of the sources studied in under 5 s of loading the prepared phantom into the ambulance in all of our trials and confirmed they interference from either the defibrillator or other electronic equipment present in the ambulance. The count rate was high enough in all cases to trigger the automatic identification routine, except when using the Superflab bolus material on the I-125 source. While the addition of bolus over the phantom shielded the low-energy gamma emissions, the device did alarm briefly but did not begin the identification routine without user input. For Gd-153, a source not present in its library the detector primarily reported Sm-153 as the primary candidate, with confidence level (CL) between 80-90.
Distance from patient to each detector mounting position. The phantom used for each ambulance trial was placed in the ambulance as described in the Materials and Methods. The isotopes shown in Table I were placed on the phantom at the anatomical positions stated. The distance from the ambulance wall and ceiling from the isotope location to the detector are shown.
Survey measurement results. The dose rate recorded in our survey measurements was fit to a power-law formula as follows:
where D is the dose rate at the detector, l is the distance (Table III) from the source and m is the slope, and b the intercept. The results of individual fits per radionuclide to this model are shown in Figure 2. Cs-137, Ba-133, and Co-60 all had results close to the inverse square as expected. Some deviation is expected as the upper wall position was partially obstructed when the source was placed at the feet. The slope for I-125 data is steeper than those for the other isotopes (p<0.000186) due to greater attenuation in air for its low-energy gamma rays. The detector's lower limit in energy detection is about 25-30 keV compared to the 35 keV gamma peak for I-125 used for identification as listed in the device's manual.
The ambulance, phantom, detector position, and vehicle driving route used in the present experiments. a: Humanoid Rando phantom and positioning of radiation test sources. b: Positioning of phantom on the ambulance gurney and loading into the ambulance. c: Detector placement on the wall of the ambulance. d: Driving route around the Hospital parking lot used for motion experiments (Image Source, Roof of Presbyterian University Hospital, Pittsburgh, PA, USA).
Identification time. The time to identification required was typically less than 5 min for the radionuclides studied, with an average of 149±95 s with the source and detector locations used for the identification trials. As Figure 3 shows, the time to identification increased with increasing thickness of bolus for I-125, which has low-energy gamma spectrum. The shortest identification times were in the two-source trials where the combined activity resulted in the device quickly reaching its statistical threshold. The time to identification for two-radionuclide trials was usually 20-90 s.
Accuracy of measurements. In 54 single-radionuclide trials, the Interceptor correctly identified the source present 96% [95% confidence interval (CI)=86% to 99%] of the time. The reported confidence level for correct trials was 56-97%, with a median value of 74%. However, in 91% (95% CI=80% to 96%) of these trials, the device reported a secondary radionuclide. The CIs reported for the falsely identified second radionuclide were 40-64%, with a median value of 51%.
Interceptor accuracy for confidence level bands for the first and second isotope reported for all trials. The accuracy of measurements is shown for pairs of isotopes in experiments with two different radiation sources. Further data are shown in Figure 3 and presented as described in the Materials and Methods.
In two-radionuclide trials, the first nuclide reported was correct 93% (95% CI=83% to 97%) of the time and the second nuclide reported was correct 41% (95% CI=30% to 54%) of the time. The reported confidence level for the primary radionuclide was 46-94%, with a median value of 74%. The reported confidence level for the second radionuclide was 40-73%, with a median value of 54%. While the confidence level for the secondary radionuclide was usually much lower in these cases, no clear boundary separated the two radionuclide and single nuclide cases solely based on confidence level (Figure 4, Tables IV and V).
To further investigate how the confidence level reported by the proprietary identification algorithm correlated to the empirical accuracy of the identification results, a logistic regression model was fit to the data using the R statistics software environment (21):
where π(x) is interpreted as the probability of the dependent variable equaling a ‘success’ as a function of parameters βi and explanatory variables xi. The parameters are the regression coefficients for fitting the model to the variable number of independent predictor variables, which can be either continuous variables, such as detector confidence level, or binary predictors, such as being the second radionuclide listed. To simplify the model, the primary and secondary nuclides were treated as independent trials with success being if the reported nuclide was one of the sources
present. Thus, if two radionuclides were identified in a single source trial, at least one of them had to be incorrect. The most obvious explanatory variable is the confidence level of
the identification algorithm, considering additional variables only if they improved the goodness-of-fit by a statistical significant degree, assessed by comparing the difference between the model deviance and the null deviance to a Chi-square distribution with degrees of freedom equal to the difference in the number of estimated parameters.
Dose rate versus distance from phantom surface to the ambulance detector for sealed sources of Ba-133 (3.40 MBq), Cs-137 (7.07 MBq), Co-60 (3.18 MBq), and I-125 (16.32 MBq). Note the steeper slope for I-125, due to increased attenuation of low-energy gamma rays. Source: orange circle Ba133; green diamond Co60; blue circle Cs137; and purple plus sign 1125.
Time to identification in seconds using the Thermo Fisher Interceptor at 1.2 m (mounted on the ambulance ceiling) sealed sources for Ba-133 (3.40 MBq), Cs-137 (7.07 MBq), Co-60 (3.18 MBq), and I-125 (16.32 MBq). Lighter-shaded points correspond to increased thickness of Superflab bolus on top of the source.
The histograms of correct and incorrect assignments are shown in Figure 4, along with the resultant prediction of a logistic regression with CL as the only independent explanatory variable. In only three trials with a confidence above 70% was the reported radionuclide misidentified. The fitted curve is the probability of an identification being correct as a function of the confidence level reported by the device. For device confidence levels much above 60%, the probability of being correct for our data is higher than that suggested for the device.
The first and second radionuclides are shown separately in Figure 4b and c. The histogram for the second radionuclide reported shows the unreliability of its assignment. While the resultant fit is similar to that for the combined data, in this case, the fit is more speculative as there are no events in the high confidence region.
The bolus thickness and ordinality of the identification were added to the model as independent variables. The bolus thickness was not a statistically significant predictor of correct identification. The order of the results (whether the radionuclide was the first or second) was a marginally significant coefficient [p(>|z|) of 0.07], but may be biased by the relative number of single- to two-source trials, since when the model is restricted to only the two source trials, the coefficient loses significance [p(>|z|) nearly 1].
The mean and maximum dose rates and count rates were also tested as predictors, with no improvement in fit quality. Attempts to predict the number of sources present and the presence of bolus were unsuccessful. There was insufficient information to predict either of these categories based on the detector results.
We calculated the coefficients, standard errors and statistical significance of the terms of our best logistic regression mode. We found that the confidence levels had a positive correlation with the data collected. The exponential function of the coefficient, the odds ratio (OR=en), gives the likelihood of success, i.e. correct identification, for each unit increase in the predictive variable. For each percentage increase in confidence level of the reported radionuclide, the likelihood of correct identification increases by a factor of 1.16.
Accuracy of measurements: Interceptor identification of the current isotopes when isotope 1 is known (a); and when isotope 2 is known (b). c: Histogram of CL reported for correct (dark gray) and incorrect (light gray) identification of the isotope present. Each curve is the predicted probability from a logistic regression of the data as described in the text of the Materials and Methods, and Results sections.
For the number (first or second) of the radionuclide, the corresponding odds ratio was 0.42, meaning the identification is less than half as likely to be correct if listed second for our data, independent of the confidence level (Figure 3b and c). However, this term had a large standard error and marginal significance (rejected at the 95% level). We concluded that there was a need for further study to determine the true predictive power. If the model was restricted to only the two radionuclide trials carried out, we conclude that the order coefficient may lose significance.
Data interpretation. The survey and time-to-identification results show that an SPRD with small detector volume such as the Interceptor is sensitive enough to detect and identify low to moderately activity gamma emitters in an ambulance. The sources tested had 2-8% of activity considered dangerous in an unsealed form according to the IAEA. Radiation was detected in every trial and triggered automatic identification in all cases, except for the very low gamma-energy source (I-125) with attenuation from the bolus. The identification times were typically 5 min or less, sufficiently quick to give notice to both the first responders and target hospital in a real emergency. For sources with high activity, identification time neared 1 min, which would allow for multiple measurements during a typical ambulance ride (Tables II, III, IV and V).
The accuracy of the identification was excellent for the primary radionuclide reported. Radionuclides with high confidence (>85%) levels were accurate in all trials. However, in only four out of 54 single-radionuclide trials did the device report just one radionuclide. The legitimacy of the secondary nuclide reported by the device is more difficult to assess. In only 14 trials was the confidence of the second radionuclide moderate (65-84%), with an associated identification accuracy of 78%.
Neither the presence of a bolus nor vehicle motion was a factor in predicting the accuracy of identification. Likewise, the presence of a bolus could not be predicted from the data reported by the device, hence differentiating internal and external sources solely from detector data is not possible.
The order of radionuclides was marginally significant [p(>|z|)=0.07] in the logistic regression, with a radionuclide listed second being half as likely to be correctly identified as one listed first. If the number of sources present is not known, as would be the case in responding to an emergency where radiation is not expected, it is difficult to assess whether a second radionuclide listed would be genuine. The logistic regression (Table V) predicted an accuracy of above 80% for a confidence level above 70%, despite few data in this region for a secondary radionuclide.
Discussion
With increased concern for the possibility of a radiation-based terrorist event, emergency response personnel, as well as hospitals, who will care for casualties, have become increasingly aware of the need to manage victims exposed to or contaminated with radiation isotope.
With the availability of novel radiation mitigators that can be given 24 h or more after radiation exposure (1-5), with a high likelihood of improving recovery of critical organs, including the bone marrow and gastrointestinal system, it has become equally important to identify casualties not significantly exposed to radiation and who, therefore, do not require such new agents. In previous studies, we reported the placement of radiation detectors in the hospital environment, including the ambulance bay (arrival site for ambulances), the nurses station in the emergency room, and at critical places within the hospital (9, 17). We concluded that a better system would identify potentially exposed patients in the ambulance prior to arriving at a destination for the management of injuries. In the setting of a fission bomb or dirty bomb, it is likely that management of casualties will aIso require care for injuries that may not be related to ionizing irradiation alone. Victims will likely present with traumatic brain injury, heat, penetrating wounds, and concussions. For this reason, it becomes important to identify not only whether patients have been exposed to radiation, but also whether they are contaminated with radiation-emitting isotopes, whether on their clothing or internally (inhaled or ingested).
We evaluated several factors specific to the Thermo Fisher Interceptor device use in an ambulance, including the device placement for shielding and distance effects, possible interference from other ambulance equipment and potential impacts from vehicle motion. We conclude that an SPRD such as the Interceptor used in our study is well suited for use by first responders in an ambulance (18-21). Our observations were limited by the simulation design and humanoid used, although obviously this type of work in a more real setting has practical barriers.
The unattended mode provides for automatic identification upon detection of gamma radiation above background without user input, allowing the operators to attend to the patient. A device of such small size can easily be incorporated into existing ambulance equipment without much modification and with a removable mount, allowing for portability for triaging outside of the ambulance if needed. The housing is rugged, with the option of simplified display modes, loud alarms, and sufficient battery life for hours of unplugged use when needed.
The accuracy of detection of the primary radionuclide was excellent for the sources studied. Even with the relatively low activity of the sources used, detection was instantaneous and identification was easily completed multiple times over during transit for more active sources. The major concern was the frequent reporting of secondary radionuclide identification when only a single source was present. The data with successful size rarity single radionuclide identification was powerful. However, the experiments with multiple isotopes were less robust. To use the device the data suggest that the device software may require a preset to the top of two possibilities to meet a very minimal requirement for detection of a second isotope. If not corrected, the readings could lead to confusion for a first responder, especially if the primary radionuclide listed were a medical source, and a potentially incorrect secondary industrial source were identified. The second radionuclide might be a genuine reading or simply the second best fit to the energy spectrum detected. The user interface suggested the former but the data show better agreement with the latter possibility. A possible option would be to take repeated measurements and to see if the secondary ‘source’ was consistently identified as the same radionuclide. For widespread use of the SPRD, one would need to consider the expected number of medical source detections for a population served in a geographic area surveyed. In each case, the operator would likely be presented with the identification result of multiple radionuclides, which may be insufficient to allow one to disregard a medical source as being non-threatening.
The present ambulance-based Radiation Detector System allows rapid and accurate identification of radioisotopes likely to be associated with radiation-based terrorist events or a nuclear reactor explosion. This ambulance-based detection system has potential value for use in identifying radiation-contaminated individuals from a radiation-based terrorist event or nuclear reactor accident.
Acknowledgements
The Authors would like to thank Kevin Bohner in his assistance in acquiring the sources and in data collection. They also thank Garrette Riley for operating the ambulance and equipment, and assisting in data collection.
- Received May 14, 2015.
- Revision received May 15, 2015.
- Accepted May 17, 2015.
- Copyright © 2015 The Author(s). Published by the International Institute of Anticancer Research.
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.