Form factor-based radioactive signal classification method

Abstract

Method for classifying radioactive signals generated by a scintillator in response to radiation, the method comprising the steps of: Acquiring (501) a digital signal representative of radioactive radiation by means of an acquisition device comprising at least one scintillator and a photodetector, Calculating (502) a form factor from the digital signal, Applying (503) a classification method to the calculated form factor to classify the radioactive radiation according to at least two classes corresponding respectively to different types of radiation. Figure 5

Description

Title of the invention: Method for classifying radioactive signals based on a form factor

[0001] The invention relates to the field of devices for detecting radioactive signals based on organic or inorganic scintillators. More specifically, it relates to a method and a device for classifying radioactive signals detected by such devices.

[0002] The invention applies in particular in fields for which there is an interest in detecting and identifying a type of radiation, in particular the field of nuclear security but also medical imaging by X-rays.

[0003] Organic and inorganic scintillators are materials that are used to detect radiation of different types, for example gamma radiation or neutron radiation.

[0004] Scintillants are generally associated with photodetectors which transform the light generated by scintillation into electrical signals which correspond to a pulse. The shape of the pulse depends on the type of radiation.

[0005] A general problem in this field is to be able to identify precisely to which type of radiation a pulse detected by a photodetector corresponds. In other words, it is a problem of classifying the detected pulses according to the type of radiation which is captured by the scintillator.

[0006] The scientific publication [1] describes several signal processing methods which are all derived from a method called Comparison Charge Method (CCM). This method consists of discriminating the measured pulses according to their shape. More precisely, it is based on an estimation of the decay time of the amplitude of the pulse which differs according to the nature of the radiation. The criterion used to characterize a pulse is the ratio between two integrals of the signal

[0007] ttt _ QlaiI i * 1 ratio Q '

[0008] Qtaii is an integral calculated on the end of the signal defined by an interval [tshort+l, tiong -1] (J 5:—F .A" / '

[0009] Qtotai is an integral calculated over the entire signal.

[0010] ~ total 2 ^i-1 i

[0011] Xi designates the amplitude of the signal at the time abscissa i.

[0012] A disadvantage of the CCM method is that the TTTratio criterion is very dependent the type of scintillator and photodetector used. In other words, the method must be calibrated specifically for each device because the results based on this statistical criterion are not invariant depending on the technology used for radiation detection.

[0013] In particular, the time length of the first integral defined by the parameters tiong and tshort has an influence on the discrimination of the radiations and is strongly dependent on the architecture of the detector and the photo-detector.

[0014] Reference [2] describes another method based on the quadratic mean value of the signal to discriminate neutron radiation from gamma radiation by means of semiconductors and ionization chamber detectors.

[0015] The invention provides a new method for classifying pulses generated by an organic or inorganic scintillator in response to radiation in order to identify the type of radiation detected.

[0016] The proposed method is based on a new statistical indicator, the form factor, which has the interesting property of depending less on the type of detector used than the CCM method and in particular does not require fine optimization of parameters, outside the classification threshold, as in the case of the CCM method.

[0017] The form factor is used as an indicator coupled with a direct classification method or based on an artificial intelligence model.

[0018] The form factor can be coupled with other indicators as input to an automatic classification method which can be trained to discriminate the pulses according to several classes each corresponding to a given type of radiation.

[0019] The subject of the invention is a method for classifying radioactive signals generated by a scintillator in response to radiation, the method comprising the steps of: - Acquire a digital signal representative of radioactive radiation by means of an acquisition device comprising at least one scintillator and one photodetector, - Calculate a form factor from the digital signal, - Apply a classification method to the calculated form factor to classify radioactive radiation into at least two classes corresponding respectively to different types of radiation.

[0020] According to a particular aspect of the invention, the classification method comprises at least one comparison of the form factor to at least one predetermined threshold.

[0021] According to a particular aspect of the invention, the threshold is determined from a statistical distribution of the values ​​of the form factor calculated for all the types of radiation to be classified.

[0022] According to a particular aspect of the invention, the classification method comprises running an automatic classification model trained to classify form factor values ​​according to radiation type.

[0023] According to a particular aspect of the invention, the automatic classification model is configured to receive as input the calculated form factor and at least one other statistical indicator calculated from the digital signal.

[0024] According to a particular aspect of the invention, the at least one other statistical indicator is taken from: a ratio between two integrations of the digital signal respectively on a final part of the acquired signal and on the entirety of the acquired signal, a ratio between the maximum value of the signal on its average.

[0025] According to a particular aspect of the invention, the automatic classification model is a classification model with more than two classes.

[0026] According to a particular aspect of the invention, the form factor is equal to the ratio between the quadratic mean of the digital signal and the average of the digital signal.

[0027] According to a particular aspect of the invention, the types of radiation to be classified are taken from: neutron radiation or gamma radiation.

[0028] The subject of the invention is a device for classifying radioactive signals comprising an organic or inorganic scintillator, a photodetector, an analog-digital converter and a processing unit configured to execute the steps of the method according to the invention.

[0029] Other characteristics and advantages of the present invention will appear more clearly on reading the description which follows in relation to the following appended drawings.

[0030] [Fig-1] represents an example of pulses detected by a detection device radiation comprising a scintillator,

[0031] [Fig.2a] illustrates the application of a pulse discrimination method according to the prior art applied to the example of [Fig.l],

[0032] [Fig.2b] represents a distribution diagram of the values ​​of the statistical indicator used by the prior art method for the example of [Fig.2a],

[0033] [Fig.3a] represents a second example of a distribution diagram of the values ​​of the statistical indicator used by the prior art method,

[0034] [Fig.3b] represents a third example of a distribution diagram of the values ​​of the statistical indicator used by the prior art method,

[0035] [Fig.3c] represents a fourth example of a distribution diagram of the values ​​of the statistical indicator used by the prior art method,

[0036] [Fig.4] represents a diagram of a device for classifying radioactive signals according to one embodiment of the invention,

[0037] [Fig.5] represents a flowchart of a method for classifying ra signals dioactives according to one embodiment of the invention,

[0038] [Fig.6a] represents a first example of a distribution diagram of the values ​​of the statistical indicator used by the method according to the invention,

[0039] [Fig.6b] represents a second example of a distribution diagram of the values ​​of the statistical indicator used by the method according to the invention,

[0040] [Fig.l] represents, on an amplitude-time diagram, an example of two pulses 101, 102 respectively relating to neutron radiation (101) and gamma radiation (102). The two pulses are acquired by means of an acquisition chain comprising an organic or inorganic scintillator, a photodetector and an analog-digital converter.

[0041] As can be seen in [Fig.l], the pulse generated by the photodetector in response to the radiation has a first phase of rapid growth and then a second phase of slower decay. The shape of this second decay phase depends on the nature of the radiation and therefore makes it possible to identify the type of radiation.

[0042] As indicated in the preamble, the prior art method known as the charge comparison method or CCM method aims to calculate a statistical indicator TTTratio which depends on two parameters tshort and tiong which define two time intervals to calculate two integrals of the signal. Figure 2a represents an example illustrating the calculation of the ratio t-t't' _ for values ​​of tshort = 20 ns and tiong = 500 ns. The integral Q 111 ratio ““ ' total is calculated on the entire signal while the integral QtaU is calculated on the end of the signal excluding the ascending part of the pulse.

[0043] The CCM method is essentially used to discriminate between two types of radiation, for example gamma radiation and neutron radiation.

[0044] To achieve this discrimination, the TTTratio indicator is compared to a detection threshold and depending on whether the value of the indicator is below or above the threshold, the presence of one of the two types of radiation is deduced.

[0045] [Fig.2b] represents an example of a distribution function of the values ​​of the TTTratio indicator obtained with the parameters of [Fig.2a]. These results were obtained with a plastic scintillator and a 252Cf radioactivity source which is a mixed source of neutron and gamma radioactivity.

[0046] [Fig.2b] shows the distribution of the values ​​obtained. The shape of the distribution function corresponds to the superposition of two Gaussian functions respectively centered on two different means. These two functions correspond to the respective distributions associated with neutron and gamma radiation. Thus, the detection threshold can be set from this result so as to best separate the two distribution functions.

[0047] A disadvantage of the prior art CCM method is that it is highly dependent to the setting of the tshort parameter which defines the QtaU integral. Depending on the value of this parameter, the distribution function of the indicator values ​​may change, which also results in a modification of the detection threshold.

[0048] Figures 3a, 3b and 3c show other distribution diagrams of the values ​​of the CCM method indicator for different values ​​of the parameters tiong and t short. It can be seen from these figures that the distribution functions change depending on the value of the parameters.

[0049] Another disadvantage of the CCM method is that it is also very dependent on the physical characteristics of the acquisition chain, in particular the type of scintillator and the type of photodetector used. In particular, the CCM method requires adjusting the tshort parameter according to the type of acquisition chain. In other words, the optimal value of this parameter must be determined for each different acquisition chain to maintain optimal classification performance independent of the equipment.

[0050] The invention relates to a new method for classifying pulses generated by a radiation detector which makes it possible to overcome the drawbacks of the CCM method.

[0051] [Fig.4] represents a diagram of a system for detecting and classifying radioactive signals according to the invention.

[0052] The system according to the invention is placed near a source of radioactivity 401, for example a source of neutron, Gamma, Alpha or Beta radiation or any other radiation

[0053] The system comprises a scintillator 402, for example an organic or inorganic scintillator, a plastic or liquid scintillator.

[0054] The scintillator 402 generates light by scintillation in response to the radiation emitted by the radioactivity source 401. The light emitted by the scintillator 402 is captured by a photo-detector 403 which transforms it into an electrical signal. The photo-detector 403 is for example a photomultiplier, for example based on PMT tubes (Photo Multiplier Tubes) or a photo-multiplier on SiPM (Silicon Photo Multiplier).

[0055] The characteristic pulse of the radiation is then isolated from the electrical signal 404 and digitally converted by an analog-to-digital converter 405.

[0056] The digital signal (as shown for example in [Fig. 1]) is then transmitted to a processing unit 406 which performs the pulse classification. The processing unit 406 is configured to execute the steps of the radioactive signal classification method described in the flowchart of [Fig.5].

[0057] The first step 501 of the method corresponds to the acquisition of the signal.

[0058] The second step 502 of the method is executed by a first calculation module

[0059]

[0060]

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071]

[0072]

[0073] 407 and consists of calculating a form factor from the entire digital signal. The form factor FF is equal to the ratio between the root mean square and the mean of the signal. FF = RMS= fl= xi corresponds to the signal samples for i varying from 0 to N-1, N being the number of samples. The number N of samples is for example equal to the total number of digitized samples which corresponds to the duration tiong. The third step 503 of the method consists of a classification step applied to the form factor FF calculated in step 502. According to a first embodiment of the invention, the classification step 503 is implemented by a simple classification module 408 which consists of comparing the form factor to one or more classification thresholds. In the case of a binary classification, i.e. into two types of radiation, the classification threshold can be determined in the same way as for the CCM method, i.e. by analyzing the distribution of the form factor values ​​for test acquisitions. Figures 6a and 6b represent two examples of distributions of the form factor values ​​obtained under the same conditions as those of Figures 3a to 3c presenting the results of the CCM method. [Fig.6a] is obtained for a signal duration equal to 1000ns while [Fig.6b] corresponds to a signal duration equal to 800 ns. It is noted that, unlike the CCM method, the distribution of the form factor values ​​does not depend on the tshort parameter used for the CCM method and more generally does not depend on any parameter to be adjusted other than the length of the signal. It is also noted in Figures 6a and 6b that the two Gaussian curves are more clearly separable than in the case of the CCM method illustrated in Figures 3a to 3c. From the distribution of the form factor values, calculated on a set of test acquisitions, we therefore set a classification threshold to separate the two Gaussian functions. An example of threshold S is shown in [Fig.6a]. In an alternative embodiment, the classification step is performed to classify more than two types of radiation. In this case, a number m of thresholds is set to classify m+1 distinct radiations. Generally, a Gaussian mixture model can be used to model the distribution of the form factor values ​​according to the types of radiation in order to deduce the threshold(s) allowing the separation of the different Gaussian functions.

[0074] The threshold(s) set depend on the length of the signal which depends on the acquisition chain. Thus, the classification thresholds must be defined for each acquisition chain. The module 408 implements a comparator between the value of the form factor and the classification threshold(s) to deduce the type of radiation.

[0075] In another embodiment of the invention, the classification step 503 is carried out by means of an artificial intelligence engine 409 based on an automatic classification method.

[0076] The artificial intelligence model takes as input the form factor values ​​and classifies the values ​​according to their association with two or more distinct radiation types.

[0077] The model is previously trained on training data which can be acquired for different types of radiation or simulated.

[0078] The artificial intelligence model is for example a random forest algorithm or an artificial neural network, for example a multi-layer perceptron or a convolutional neural network.

[0079] In an alternative embodiment, the classification method may also be based on the analysis of histograms as described in reference [3].

[0080] In another embodiment variant, when an artificial intelligence model is implemented, for example an artificial neural network, several other statistical indicators calculated from the signal can be provided as input to the model to improve the learning of the classification of the different types of radiation.

[0081] For example, in addition to the form factor, at least one other statistical indicator can be calculated and provided as input to the model from among: the TTTratio indicator, the mean of the signal, the variance of the signal, the difference between the maximum value and the minimum value of the signal, an asymmetry coefficient of the signal, an impulse indicator equal to the maximum value of the signal divided by its mean, a peak factor of the signal, a maximum value of the absolute value of the signal, a quadratic mean of the signal, a Kurtosis coefficient.

[0082] The statistical indicators which have the most significant influence on the discrimination of radiation types are the form factor, the TTTratio factor and the impulse indicator. Advantageously, these three indicators are calculated as a priority to be provided as input to the model.

[0083] The artificial intelligence model is trained via supervised training using training data obtained for different types of detectors and different types of radiation.

[0084] The processing unit 406 of the detection device according to the invention can be implemented by means of software and / or hardware elements, for example comprising an integrated circuit, a programmable logic circuit of the FPGA type or a microcontroller or even a digital signal processor.

[0085] In an alternative embodiment of the invention, the processing unit 406 can be implemented in a remote calculation server. In this case, the signals measured by the detector are transmitted to this server.

[0086] The invention makes it possible to avoid the necessary parameter adjustments of the methods of the prior art and proposes a solution which does not require any parameter to be adjusted to calculate the statistical indicator making it possible to characterize the radiation. In particular, the length of the signal can be estimated directly from the characteristics of the components of the acquisition chain. However, the proposed method does not require adjusting other parameters as is the case for the CCM method. References

[0087] [1] Gamage, KAA, MJ Joyce, and NP Hawkes. "A comparison of four different digital algorithms for pulse-shape discrimination in fast scintillators." Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 642.1 (2011): 78-83

[0088] [2] Stankovié, Srboljub J., et al. "MSV signal processing System for neutron-gamma discrimination in a mixed field." Nuclear technology and radiation protection 27.2 (2012): 165-170.

[0089] [3] French, Robert M., et al. "A histogram-difference method for neutron / gamma dis crimination using liquid and plastic scintillators." IEEE Transactions on Nuclear Science 64.8 (2017): 2423-2432.

Claims

Claims

1. Method for classifying radioactive signals generated by a scintillator in response to radiation, the method comprising the steps of: - Acquiring (501) a digital signal representative of radioactive radiation by means of an acquisition device comprising at least one scintillator and a photodetector, - Calculating (502) a form factor from the digital signal, - Applying (503) a classification method to the calculated form factor to classify the radioactive radiation according to at least two classes corresponding respectively to different types of radiation.

2. A method of classifying radioactive signals according to claim 1 wherein the classification method (503) comprises at least one comparison of the form factor to at least one predetermined threshold.

3. A method of classifying radioactive signals according to claim 2 wherein the threshold is determined from a statistical distribution of the form factor values ​​calculated for all the types of radiation to be classified.

4. A method of classifying radioactive signals according to claim 1 wherein the classification method comprises running an automatic classification model trained to classify the form factor values ​​according to the type of radiation.

5. A method of classifying radioactive signals according to claim 4 wherein the automatic classification model is configured to receive as input the calculated form factor and at least one other statistical indicator calculated from the digital signal.

6. Method for classifying radioactive signals according to claim 5 in which the at least one other statistical indicator is taken from: a ratio between two integrations of the digital signal respectively on a final part of the acquired signal and on the entire acquired signal, a ratio between the maximum value of the signal on its average.

7. A method of classifying radioactive signals according to any one of claims 4 to 6 wherein the classification model au- automated is a classification model with more than two classes.

8. A method of classifying radioactive signals according to any preceding claim wherein the form factor is equal to the ratio between the root mean square of the digital signal and the average of the digital signal.

9. A method of classifying radioactive signals according to any preceding claim wherein the types of radiation to be classified are taken from: neutron radiation or gamma radiation.

10. Device for classifying radioactive signals comprising an organic or inorganic scintillator (402), a photodetector (403), an analog-digital converter (405) and a processing unit (406) configured to carry out the steps of the method according to any one of the preceding claims.

Images