Abstract
We present γ-ray radiation detection in a neutron radiation environment using a monolithic active pixel sensor (MAPS) camera without conversion or shielding layers. The measured output signal is the sum of the pedestal value, noise, and real radiation response signal. The sensor response shows that the MAPS camera is sensitive to neutrons and can capture a single photon. The number of pixels with a signal exceeding 100 exhibits a strong dependence on the dose rate and is the best indicator of this value. Therefore, a MAPS camera can be efficiently used as a radiation detection sensor in a robotic system, further limiting human errors in performing radiation detection in complex nuclear radiation environments.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Robots are frequently used in radiation environments, such as the aftermath of nuclear accidents and nuclear facility patrols. For example, after the Fukushima nuclear accident, robots were used to investigate the area. One of the most important missions for robots is nuclear radiation detection. However, the radiation environment of a nuclear accident is complex, including mixed radiation fields of neutrons and gamma rays. Traditional γ-ray detectors are vulnerable to nuclear accidents and are susceptible to a mixed radiation field. They are usually characterized by their large volume, high costs, and poor radiation resistance. Cameras are an essential part of robotic systems. With the rapid development of camera technology, increasing attention has been paid to the utilization of a camera as a radiation detector. Monolithic active pixel sensors (MAPSs) form camera image sensors. They are sensitive to visible light and respond to ionizing radiation. Research on direct radiation detection is progressing with the development in MAPS technology. In particular, commercial off-the-shelf MAPSs have been used for radiation detection. MAPSs exhibit a high radiation resistance, stable and reliable operation, high cost performance, and a wide detection range. The radiation effect of charged particles is evident due to MAPSs being semiconductor-based sensors [1]. In recent years, the detection of uncharged particles using MAPSs has attracted significant attention. Wei et al. proposed an effective algorithm to distinguish radioactive events in video images collected by an unshielded MAPS camera [2]. They also evaluated the response of smart phone MAPS cameras to γ-rays and demonstrated the use of smartphones for photon detection [3]. Tith et al. also reported research on the use of mobile phones with MAPS cameras for radiation detection, which confirmed that smart phones can be used as γ-ray measuring devices once calibrated, and also for radiation safety control of high-level radioactive sources, such as industrial radiography, γ-ray irradiation facilities, and medical treatment [4]. Pérez et al. studied the radiation response characteristics of a MAPS to α, β, and γ rays, and confirmed that this type of sensor can be a cheaper alternative to personal dosimeters, based on other detection techniques [5]. It was also reported that a MAPS can be used to classify particles [6], and is sensitive to soft X-rays [7]. Its radiation response characteristics can be utilized for radiometric imaging [8]. Servoli and Magalotti et al. also reported a study on real-time detection using a MAPS as a personal dosimeter [9–11]. Ma et al. used Advanced RISC Machines (ARM) microcontrollers and ZigBee modules in combination with MAPS to detect low-energy radiation [12,13]. Galimberti et al. and Wang et al. reported successful radiation detection using commercial off-the-shelf MAPS [14,15]. Velthuis et al. and Page et al. demonstrated that a MAPS can be successfully used to measure radiation profiles in radiotherapy [16,17]. Arbor et al. studied a real-time detection method for fast neutrons and thermal neutrons, using a 14-µm thick neutron sensitive layer and complementary metal oxide semiconductor (CMOS) MAPS [18]. While these previous studies have confirmed the use of MAPSs as uncharged particle detectors, research on the relationship between integration time, radiation dose rate, and radiation response signal has rarely been reported, especially under low dose rate irradiation conditions. Meanwhile, the number of response events in the pixel array is linearly dependent on the neutron radiation fluence from the same neutron source, and Xue et al. explained that the dark signal in a MAPS is affected by neutron radiation from a nuclear reactor [19], which must also affect γ-ray detection. Research on the difference in energy responses, and the method for processing radiation data under a low dose rate and a mixed neutron radiation field, is required for further research.
In this paper, the radiation response of MAPSs to γ-rays and neutrons is reported. The differences in the radiation response of MAPSs were studied using characteristic response events and histograms of the irradiated frames. It has been built on a method for data extraction and response signal processing. A linear relationship between radiation response and different γ-ray energies was obtained for small values of dose rate. The results of this paper provide theoretical and experimental support for exploring the performance of MAPSs in direct radiation detection of γ-rays in a neutron radiation environment.
2. Experiments
2.1 Sensors
The experiments presented in this paper were undertaken using the Sony IMX 222LQJ-C MAPS, which is a 2.43M pixel sensor. The pixels were designed in a 0.18 µm CMOS process with four transistors and a pinned photodiode. The pixel pitch was 2.8 µm. The chips were operated at an analog voltage of 2.7 V, a digital voltage of 1.2 V, and an interface voltage of 1.8 V. The sensor was integrated on the sensor board and provided 8-bit data. The integration time of the sensor was adjustable from 1/25 s to 1/10000 s. An Ambarella system on chip (A5S ARM) was used to generate readout signals and for digital signal processing. The glass covering the front of the sensor was removed, and the sensors were covered using a layer of opaque plastic material to help insulate the sensor from contamination due to the surrounding visible light.
2.2 Experimental setup
The experimental setup is shown in Fig. 1. The experiment was performed in a laboratory with gamma and neutron sources. The sensor module was placed in a container and placed on a slider in front of an 8-cm collimating hole of the γ-rays’ radiation facility. A γ-ray radiation facility with four collimated γ sources was used. One 60Co source emitting gamma photons with energies of 1.17 MeV and 1.33 MeV and an activity of 188 mCi, and three 137Cs sources with photon energies of 0.662 MeV and activities of 11.3 mCi, 225 mCi, and 2.17 Ci were used. A spherical isotropic neutron source, equipped with 252Cf and stored in an incompletely-shielded well located below the laboratory, was also used. The activity of the neutron source was 1×105 Bq. In the γ-ray radiation experiments, the neutron source was kept in a repository located below the laboratory, and the radiation dose rate of the γ-ray was controlled by changing the source and the distance between the source and the sensor. The dose rate was calibrated using a thermo γ-ray detection system with a range of 0.01 µGy/h to 100 mGy/h. In the neutron radiation experiments, a MAPS was placed 50 cm away from the radiation source. The entire array of the test sensor was exposed to radiation in both the neutron and γ-ray experiments. All experiments were conducted at the China Institute of Atomic Energy (CIAE), and all data were recorded online. The experimental temperature was maintained at 21 °C. The signal data were transmitted using 4800 LX cables, and the maximum video bitrate was 13 Mbps. Video files were recorded using a computer, at 25 fps during all experiments, and the data were imported using MATLAB R2019a (Math Works Inc., Natick, MA, USA) and further split into individual frames. During the experiments, the control systems of the aperture, shutter, gain, and white balance were placed manually, and noise reduction functions and exposure compensation functions were turned off.
3. Results and analyses
3.1 MAPS signal
The output signal indicates the radiation response signal of the source to be assessed. The composition of the output response signal of the MAPS in the radiation environment was analyzed. The measured output signal from a pixel (Vpixel) is the sum of the pedestal value (Vped), noise, and a real radiation response signal (Vrad). As an equation, the signal can be expressed as
where VBG_rad denotes the radiation response signals from the background radiation environment and Vnoise denotes the sum of random variation noise, common mode noise, and fixed pattern noise (FPN), which are caused by radiation damage.To extract the real signal from the ionizing particles to be measured, corrections need to be made for the background radiation, noise, and pedestal. The first and last terms on the right side of the equation can be evaluated using a dedicated run, without any radiation source (inside the shield). The latter term must be measured after each irradiation experiment to determine the value of FPN caused by radiation damage. The third term of the equation is also a radiation response signal, but it needs to be separated from the background experimental environment. This value of a MAPS can be measured without a radiation source, but still placed in the radiation experimental environment.
3.2 Neutron radiation
When neutrons pass through the MAPS, the non-ionization energy loss deposited in the space charge region of the MAPS and the main interactions are inelastic and elastic interactions. Some bulk defects are understood to be present, which can become centers for the generation and recombination of electron-hole pairs, carrier trapping, the compensation of donors or acceptors, and the tunneling of carriers. This can increase the mean dark signal and dark signal non-uniformity.
The background noise of the sensor was measured before and after the experiment, to test the FPN caused by radiation damage. The experiment was carried out with a shield and without a radioactive source. Figure 2 shows the histogram of the FPN in a dark frame of the sensor before and after neutron irradiation. As shown in the figure, the FPN with grayscale values between 1 and 5 are the most common before irradiation, and the number of pixels with grayscale values greater than 10 account for less than 0.2% of the total. Figure 2(b) shows the FPN after 5 min of irradiation with the 252Cf neutron source. By comparing the distribution of FPN before and after irradiation, the damage of the radiation on the FPN can be ignored for this dose.
The 252Cf source was used to perform neutron radiation experiments in the laboratory, and Fig. 3 shows the scatter plot of the neutron radiation response signal. As shown in the figure, during irradiation by the neutron source, most pixel values fall in the range of 10 to 20. There are also a few pixels whose values are close to 50. Figure 3 shows the spatial distribution characteristics of the radiation response pixels. Small high-value pixels appeared at the edge of the pixel array, but most of the high-value pixels are distributed evenly in the center. This indicates that data processing efficiency can be improved by selecting pixels in the central part of the array.
The data in Fig. 3 were processed using a contour, as shown in Fig. 4. The figure depicts the number of response pixels in each range of pixel values in the sensor array. This result is crucial for delimiting the threshold of the background noise. The level with a pixel value of 15 contained the largest number of pixels. The distribution characteristics of a scatter plot of the neutron radiation experiment are similar to those of the background radiation test; however, the overall pixel value of the pixel array increased, and a few pixels with values ranging from 0 to 5 were also obtained. The distribution of background radiation noise was similar to that of neutron radiation; however, when the neutron source was used for the radiation experiment, the value of each pixel in an entire pixel array increased.
Figure 5 shows the histogram of the background radiation noise and the 252Cf neutron source radiation signal of the sensor. The neutron radiation peak is shifted to the right compared to the background radiation. Approximately all pixel values were distributed in the range of 0 to 40. The peak position for the 252Cf neutron source irradiation curve is to the right of one of the backgrounds.
Figure 6 shows the bar chart of a typical neutron radiation event in 10 × 10 pixel regions. As can be observed, neutrons’ radiation response events in the captured frames show minute, but sharp peaks due to interactions. It was deduced from Eq. (1) that the final output signal of each pixel in the neutron radiation event included noise and signal from the background and neutron radiation. This event showed that larger values are in the middle, while the peak value is similar to the background radiation event. However, neutron radiation mainly induced an increase in the pedestal of events, which implies that the pixel value of each pixel in pixel array increased as a whole.
The neutron radiation experimental results show that a MAPS is sensitive to neutrons. The response is mainly characterized by an increase in the pixel value of the pixel array, and the peak value of events being close to 50. Therefore, it was concluded that the background radiation in the laboratory was mainly caused by neutron radiation. There were scattered neutrons in the laboratory, and the radiation source of these scattered neutrons was a 252Cf neutron source that was used for the experiment, and stored in the source library below the laboratory. Neutrons were scattered in the laboratory, and the background radiation was provided only by the stored neutron source. This explains the similarity of the pixel value distribution characteristics between the neutron and background radiation.
3.3. Photon response under neutron radiation
Figure 7 shows the scatter plot of the γ-ray radiation in a frame of the sensor with the 60Co and 137Cs source, under a 252Cf neutron source radiation environment. It can be observed that there are noticeable peaks in the frames, and the pixel values of most pixels are below 50. The two scatter plot figures have the same division.
Figures 8 and 9 show the bar chart and heatmap of the 60Co and 137Cs γ source radiation events, respectively, in a frame of the sensor. As shown in the figures, events caused due to γ-rays are different from neutron events. Neutron events exhibit a relatively symmetrical peak; however, γ-rays show a series of pixels in being hit, and a great value of radiation response pixels. The insignificant effect on the surrounding pixels is mainly caused by the f secondary particles generated by Compton scattering. Due to the high ionizing capacity of γ photons, the electrons formed from the energy deposited in the pixel filled the space charge region almost instantaneously during the integral time, and the pixel value was large and close to saturation. However, not all rays are incident on the pixel completely perpendicularly due to scattering, while the size of a pixel is on the micron scale. Different γ-ray incident directions created different peak patterns of response events. The γ photon events were caused by the superposition of multiple incident photons. As shown in Figs. 8(b) and 9(b), the pixel in an event with a pixel value greater than 250, can be considered as a direct deposition of energy by γ-rays in the space charge region of the pixel; however, there is a pixel with a pixel value of only 12, in the response event center area of Fig. 8, which is similar to the superposition of noise and background radiation. This proves the non-existence of deposition of γ-ray energy in the pixel, and hence a response event can be a superposition of multiple photon radiation events. Pixels with a large value in the event in Fig. 9(b) were relatively concentrated. This could have been an event produced by a single photon or its secondary ionizing particle, and generated by multiple photons simultaneously incident to adjacent pixels. It is shown that a MAPS is equipped with the ability to capture a single photon, due to the space charge area of the pixel being miniscule with a low response probability. As can be seen from Figs. 8 and 9, pixels had already reached their saturation value mid-event. These data may result in a large statistical error when the radiation dose rate is represented by the pixel value. Therefore, data close to saturation should be excluded. However, these saturated data are also valuable, and will not affect the results when the radiation dose rate is represented by the count of radiation response pixels.
4. Discussion and application
Few radiation response events were generated in one frame due to the low activity of the γ sources utilized in the experiments. Hence, we could not obtain a statistically accurate result by counting the events in a frame. Thus, events in 300 consecutive frames were counted together to increase the statistical significance. Figure 10 shows the pixel value distribution in 300 consecutive frames captured with 137Cs and 60Co γ-ray sources, at different dose rates. The results of the neutron radiation study show that the noise and background neutron radiation were mainly concentrated in the range of 1 to 50, while part of the peak value was close to 60. Pixels with a value larger than 60 were mainly caused by direct or indirect γ-ray ionization radiation. The values of the pixels directly affected by γ-rays were either large, or nearly saturated. Although a MAPS is also sensitive to neutron radiation, the characteristics of the response events are very different. Neutron radiation in the environment does not affect γ-ray detection. As shown in Fig. 10, there is an obvious difference in the pixel value distribution for values above 50.
Counting the number of pixels with signals exceeding 50 could be utilized to extract the dose rate. However, the proportion of pixels exceeding 50 was very small. To obtain a more accurate dose-rate value, data from various subsequent frames were used in this study. Furthermore, the distribution of pixel values above 100 was reasonably flat, and therefore, the number of pixels with signals exceeding 100 were utilized to extract a reliable value for dose rate detection. This is explained by the fact that the data in the range of pixel values greater than 100 are more conducive to the relationship between the response signal and dose rate. For a MAPS, a pixel value of 100 is a suitable threshold. This threshold setting method is suitable for all MAPSs in γ-ray radiation detection, at a low photon fluence.
Figures 11 and 12 show the pixel value distribution in a frame taken with a 137Cs and a 60Co γ-ray source for different integration times, ranging from 1/25 s to 1/8000 s. The number of pixels with a signal exceeding 50 decreased significantly with shorter integration, at the same dose rate. Especially when the integration time was less than 1/1000 s, the number of pixels with a signal between 20 to 50 also started to decrease. However, even if the integration time was configured as 1/8000 s, there were still very few pixels with a value greater than 150. This can be contributed to the ionized charge captured by the charge area of the pixel being very slight when the integration time is short; however, when some ionized charge is generated in or near the space, the charge area still collects charge. At low dose rates, the longer the integration time, the clearer the response signals.
As mentioned prior, the most accurate value for the dose rate can be obtained by calculating the number of pixels with a threshold value of 100 to 250. The number of pixels with a signal threshold exceeding 100 best reflects the dose rate. The data of each signal threshold were fitted using a third-order cubic fitting curve. The fitting formula is shown in Eq. (1), and the coefficients are listed in Table 1 for both 137Cs and 60Co γ-ray source experiments. The linearity (R2) of all fitting curves was larger than 0.99.
Figures 13 and 14 show the number of pixels with a signal threshold of 100 to 250 as a function of the dose rate for 137Cs and 60Co γ-ray sources. An obvious growth trend can be observed from the curves in the figures. However, we noticed that the lower the signal threshold and the faster the count, the larger the dose rate.We used a binarization method to process the data in the γ-ray irradiation experiments. Figures 15 and 16 show the relationship between the irradiation response and integration time with 60Co and 137Cs γ-ray sources. It can be observed that the number of pixels increased with a longer integration time. The solid lines denote the results of a linear fit, where the linearity of the fitting curve was greater than 0.99. This shows that there is a linear relationship between integration time and the number of pixels. The slope of the fitting curves decreased with an increase in the pixel threshold value. The slope of the 60Co fitting line was higher than that of 137Cs at the same pixel threshold value, owing to the greater energy of the 60Co γ-rays that generated more pixels with a signal exceeding the threshold value.
Verification experiments were conducted to compare the deviation between the measured and calculated dose rates, by fitting formulas. Three measured dose rates for the 137Cs γ-ray sources, 0.59, 1.77, and 5.92 mGy/h, and three measured dose rates with the 60Co γ-ray source: 0.59, 1.03, 3.31 mGy/h were taken into consideration. Table 2 lists the verification results and deviations. As given in the table, for the first test point at 0.59 mGy/h of the 60Co γ-ray source experiment, the deviation between the calculated result and the measured value is the minimum when the signal threshold value was 60. Additionally, for all other experiments, 100 was a desirable threshold. Overall, we noticed that a larger threshold value increased the deviation.
In conclusion, the number of pixels with a signal exceeding 100, best reflects the dose rate. This signal is further linearly dependent on the integration time. The key to low-dose-rate γ-ray radiation detection using a MAPS is to select an exceptional statistical method to determine the dose rate, such that it can be determined effectively, accurately, and in real time by calculating the number of pixels with a signal exceeding 100. Therefore, a MAPS can be used as a detector in robotic systems at irradiation dose rates above 500 µGy/h. A robot with a MAPS radiation detector can replace humans in performing radiation detection in complex nuclear radiation environments, such as nuclear accidents.
5. Conclusions
In this paper, it has been explained that a MAPS is sensitive to neutron and gamma radiation and can be utilized for direct detection, without a conversion layer. The response is mainly characterized by an increase in the pixel value of the pixel array, the peak value of events being close to 50. The MAPS was given the opportunity to capture a single photon only because the space charge area of the pixel was miniscule and the response probability was low. The number of pixels with a signal exceeding 100 is a favorable estimator of the dose rate. This threshold isolates the effect of neutron radiation on γ-ray detection. A binarization method was used to process the data in the γ-ray irradiation experiments, and the relationship between the count and the dose rate was proved to be linear. Hence, MAPSs can be used to accurately measure the radiation dose in real time and thus, can be used as a radiation detection sensor for robot systems in nuclear environments.
Funding
Education Department of Hunan Province (18B268); Natural Science Foundation of Hunan Province (2020JJ5499); National Natural Science Foundation of China (11905102).
Acknowledgments
The authors would like to express their sincere gratitude to the China Institute of Atomic Energy for providing the 60Co γ source and nuclear radiation detector.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. L. Servoli and P. Tucceri, “Use of CMOS imagers to measure high fluxes of charged particles,” J. Instrum. 11(03), P03014 (2016). [CrossRef]
2. Q. Wei, Z. Wang, T. Dai, and G. Yu, “Nuclear Radiation Detection Based on Un-covered CMOS Camera under Static Scene,” Atomic Energy Science and Technology 51(1), 175–179 (2017). [CrossRef]
3. Q. Wei, R. Bai, Z. Wang, R. Yao, Y. Gu, and T. Dai, “Surveying ionizing radiations in real time using a smartphone,” Nucl. Sci. Tech. 28(5), 70 (2017). [CrossRef]
4. S. Tith and N. Chankow, “Measurement of γ-Rays Using Smartphones,” Open J. Appl. Sci. 06(01), 31–37 (2016). [CrossRef]
5. M. Pérez, M. Haro, I. Sidelnik, L. Tozzi, D. Brito, C. Mora, J. Blostein, M. Berisso, and J. Lipovetzky, “Commercial CMOS pixel array for beta and gamma radiation particle counting,” in Proceedings of the 2015 Argentine School of Micro-Nanoelectronics, Technology and Applications (EAMTA). 11–16 (2015).
6. M. Pérez, J. Lipovetzky, M. S. Haro, I. Sidelnik, J. J. Blostein, F. A. Bessia, and M. G. Berisso, “Particle detection and classification using commercial off the shelf CMOS image sensors,” Nucl. Instrum. Methods Phys. Res., Sect. A 827, 171–180 (2016). [CrossRef]
7. M. S. Haro, F. A. Bessia, M. Pérez, J. J. Blostein, and J. Lipovetzky, “Soft X-rays spectroscopy with a commercial CMOS image sensor at room temperature,” Radiat. Phys. Chem. 167, 108354 (2020). [CrossRef]
8. F. A. Bessia, M. Pérez, J. Lipovetzky, N. A. Piunno, H. Mateos, I. Sidelnik, J. J. Blostein, M. S. Haro, and M. G. Berisso, “X-ray Micrographic Imaging System Based on Cots CMOS Sensors,” Int. J. Circ. Theor. App. 46(10), 1848–1857 (2018). [CrossRef]
9. L. Servoli, M. Paolucci, A. Pentiricci, P. Placidi, D. Magalotti, S. Fabiani, A. Calandra, L. Bissi, A. Maselli, M. Scarpignato, R. Di Lorenzo, R. Cicioni, A. Scorzoni, L. Colleluori, S. Chiocchini, N. Forini, A. C. Di Pilato, and F. P. Cappotto, “Personnel Real Time Dosimetry in Interventional Radiology,” Phys. Medica 32(12), 1724–1730 (2016). [CrossRef]
10. D. Magalotti, P. Placidi, M. Dionigi, A. Scorzoni, and L. Servoli, “Experimental Characterization of a Personal Wireless Sensor Network for the Medical X-ray Dosimetry,” IEEE Trans. Instrum. Meas. 65(9), 2002–2011 (2016). [CrossRef]
11. D. Magalotti, P. Placidi, S. Fabiani, L. Bissi, M. Paolucci, A. Scorzoni, A. Calandra, G. Verzellesi, and L. Servoli, “A Wireless Personal Sensor Node for Real Time Dosimetry of Interventional Radiology Operators,” in Proceedings of the International Conference on Applications in Electronics Pervading Industry, Environment and Society1–7 (2016).
12. X. Wang, S. Zhang, G. Song, G. Song, D. Guo, C. Ma, and F. Wang, “Remote measurement of low-energy radiation based on ARM board and ZigBee wireless communication,” Nucl. Sci. Tech. 29(1), 4 (2018). [CrossRef]
13. Q. Chen, Y. Yuan, C. Ma, and F. Wang, “Gamma Measurement Based on CMOS Sensor and Arm Microcontroller,” Nucl. Sci. Tech. 28(9), 122 (2017). [CrossRef]
14. C. Wang, S. Hu, C. Gao, and F. Chang, “Nuclear radiation degradation study on HD camera based on CMOS image sensor at different dose rates,” Sensors 18(2), 514 (2018). [CrossRef]
15. C. L. Galimberti, F. A. Bessia, M. Pérez, M. G. Berisso, and J. Lipovetzky, “A Low Cost Environmental Ionizing Radiation Detector Based on COTS CMOS Image Sensors,” in Proceedings of the 2018 IEEE Biennial Congress of Argentina (ARGENCON) (2018).
16. J. J. Velthuis, R. P. Hugtenburg, D. Cussans, M. Perry, C. Hall, P. Stevens, H. Lawrence, and A. Mckenzie, “The VANILLA sensor as a beam monitoring device for X-ray radiation therapy,” Appl. Radiat. Isot. 83, 8–11 (2014). [CrossRef]
17. R. F. Page, N. L. Abbott, J. Davies, E. X. Dyke, H. J. Randies, J. J. Velthuis, S. Fletcher, S. D. Gregory, C. Hall, and A. John, “Towards using a Monolithic Active Pixel Sensor for in vivo beam monitoring of Intensity Modulated Radiotherapy,” Nucl. Instrum. Methods Phys. Res., Sect. A 731, 295–298 (2013). [CrossRef]
18. N. Arbor, S. Higueret, H. Elazhar, R. Combe, P. Meyer, N. Dehaynin, F. Taupin, and D. Husson, “Real-time Detection of Fast and Thermal Neutrons in Radiotherapy with CMOS Sensors,” Phys. Med. Biol. 62(5), 1920–1934 (2017). [CrossRef]
19. Y. Xue, Z. Wang, B. He, Z. Yao, and J. Jin, “Theoretical and experimental study of the dark signal in CMOS image sensors affected by neutron radiation from a nuclear reactor,” AIP Adv. 7(12), 125222 (2017). [CrossRef]