System and method for identifying isotopes

Abstract

The invention relates to a method for identifying isotopes. In particular, the invention relates to a computer-implemented method for determining the presence of a radioactive source and a corresponding system.

Description

Technical Field

[0001] The present invention relates to a method for identifying isotopes. In particular, the present invention relates to a computer-implemented method for determining the presence of a radioactive source and a corresponding system. Background Art

[0002] Over the past decade, there has been an increasing need to screen for nuclear or radioactive sources and to prevent threats to public safety. The deployment of inspection systems and portable monitoring services using passive gamma and neutron radiation detectors provides a useful layer of defense for public safety. However, the goal of improving such detectors in terms of their detection efficiency and accuracy requires meeting the conflicting requirements of minimizing the system's susceptibility to false alarms and having a high sensitivity for the detection of weak radioactive sources, which can be time consuming and costly.

[0003] Developers in the radiation detector industry face several problems. First, environmental noise and signal attenuation can hamper detection accuracy. For example, when a detector is operating on a city street, the walls of buildings and the materials present within the buildings greatly attenuate the radiation and thus affect the detector's ability to identify radioactive sources. Moreover, this problem must be compensated for in real time during field detection. Other factors, such as calibration offsets, the presence of masking sources including naturally occurring radioactive materials (NORMs) and medical isotopes, and widely varying source intensities complicate detection and can even further hinder detection.

[0004] Various methods have been used to detect and identify radioactive sources. The simplest way is to ignore spectral information, consider only the total number of gamma rays from the radioactive source, and alert the user when the total number exceeds the counting threshold. However, there are more complex schemes, such as some machine learning algorithms, which can integrate and use complete radiation spectrum information. The latest machine learning algorithm schemes specifically use the spectral data simulated in the "Automatic Isotope Identification and Quantification Algorithm for Isotope Mixtures in Low Resolution Gamma-ray Spectra" on pages 281-286 of Volume 155 of Radiation Physics and Chemistry, published by M. Kamuda and CJ Sullivan in 2019, as input to the detection algorithm. However, these spectra cannot fully reproduce the complex features that appear in real-world situations, and this may result in low detection accuracy for deployed systems using these schemes.

[0005] The present invention is directed to reducing or overcoming at least some of the problems set forth above. Summary of the invention

[0006] According to a first aspect of the present invention, there is provided an isotope identification method performed by a computing device, comprising the following steps: collecting a plurality of spectral data from a spectral device; generating a plurality of data sets based on the plurality of spectral data by a processor of the computing device; determining a plurality of isotope probabilities by the processor, each of the plurality of isotope probabilities corresponding to a respective data set in the plurality of data sets; wherein the isotope probability represents the probability that a corresponding data set in the plurality of data sets corresponds to a specific isotope; and identifying the isotope based on the plurality of isotope probabilities by the processor.

[0007] The spectroscopic device may be any device suitable for collecting real-world isotope spectra. For example, the spectroscopic device may be a scintillation counter configured to detect and measure the intensity and / or energy of ionizing radiation emitted by the isotope and incident on the scintillation counter. Optionally, the spectral data may be filtered to match the characteristic resolution of the spectroscopic device, thereby reducing artifacts generated by statistical noise.

[0008] Those skilled in the art will appreciate that the term "isotope probability" refers to the probability that a data set is associated with a particular isotope (e.g., 137 Cs) corresponding probability.

[0009] In the process of nuclear decay, gamma rays with different energies emitted by isotopes are detected by a spectral device. The spectral device provides a plurality of spectral data to a computing device, wherein the spectral data may be the energy of the gamma rays detected by the spectral device. A machine learning-based scheme is used to classify or identify the isotopes associated with the gamma rays. Therefore, the present invention provides a method for identifying and classifying a single isotope. The method may also be used in a multi-label situation, wherein an isotope mixture comprising a plurality of isotopes is classified. Advantageously, the method is robust to noise and classifies isotopes with higher accuracy.

[0010] Preferably, each of the plurality of data sets includes binned data representing a functional relationship between a first bin of the binned data and a second bin of the binned data. Specifically, the spectral data collected from the spectroscopic device is grouped according to a plurality of intervals (i.e., bins), and a value is assigned to each interval. Each bin may be, for example, a particular energy range of gamma rays received at the spectroscopic device. An exemplary bin interval is 10 3 -10 4 eV. A bin count is associated with each bin, wherein the bin count represents the number of gamma rays that the spectroscopic apparatus has captured within the bin interval.

[0011] In a preferred embodiment, each of the plurality of data sets comprises a respective bin-ratio vector. It will be appreciated by those skilled in the art that the term "bin-ratio vector" is a vector in which each entry in the vector is a direct ratio between a pair of bin counts of different bins. Advantageously, the use of bin-ratio vectors increases classification accuracy, specifically, for example, where gamma rays of a wide energy range are detected. Specifically, different spectra collected for the same isotope may have bins with significantly different values. Differences between values ​​may occur due to changes in radiation source intensity, position, or background rate. Measures of similarity between spectra may be affected by this partial match. The bin-ratio vector can capture the relative size of the peaks in the spectrum, which in turn can improve the robustness of the method.

[0012] In some embodiments, generating the plurality of data sets includes: binning, by a processor, the plurality of spectral data into n spectral bins; generating, by the processor, a plurality of bin-ratio vectors; and storing, by the processor, a respective bin-ratio vector in each of the plurality of data sets.

[0013] Preferably, generating a plurality of bin-ratio vectors comprises: generating a histogram based on n spectral bins by a processor; generating a histogram ratio matrix M based on the histogram by a processor; wherein the histogram ratio matrix M is a square matrix of length n; and extracting a plurality of bin-ratio vectors from the histogram ratio matrix by a processor; wherein the bin-ratio vectors are k and k=n-1. Specifically, the histogram ratio matrix M is a square matrix of length n, wherein each element is a ratio between a pair of bin counts. Those skilled in the art understand the histogram ratio matrix M as a bin-ratio matrix, wherein each element in the bin-ratio matrix is ​​a direct ratio between a pair of bin counts.

[0014] Preferably, each entry M in the histogram ratio matrix M i,j is the i-th bin divided by the j-th bin. Accordingly, the ratios between bin counts are represented by the histogram ratio matrix M.

[0015] Preferably, the entries of the y-th bin-ratio vector are bin-ratios of a first bin to a second bin, wherein the second bin is separated from the first bin by y bins.

[0016] Accordingly, each bin-ratio vector includes entries with different ratios of the bin counts.

[0017] In an alternative embodiment, the bin data comprises binning differences. That is, the bin data comprises the difference between a first bin count and a second bin count.

[0018] Preferably, the isotope probability is determined by using a kit; the kit includes a plurality of categories generated by a multi-class classifier, each corresponding to a respective data set. Those skilled in the art understand the term "multi-class classifier" as a classification model, and the classification model is configured to classify a data set into one of a plurality of categories. The category represents an isotope. Those skilled in the art understand the term "kit" as a collection of classification models. The multi-class classifier may be selected from, but not limited to, one or more of the following: k-nearest neighbor (k-NN) classification method, logistic regression (LR) classification method, linear discriminant analysis (LDA) classification method, support vector machine (SVM) classification method, artificial neural network (ANN) classification method, and convolutional neural network (CNN) classification method. Those skilled in the art will recognize that any combination of classification methods can be used. Further, those skilled in the art will understand that because the functions of k-NN, LR, LDA, and SVM only support binary classification, additional binary correlation methods are applied to multi-label situations to convert multi-label situations into multiple independent binary problems.

[0019] In some embodiments, the kit is an artificial neural network (ANN) kit; the ANN kit includes a plurality of ANN classes, each ANN class is generated by an ANN, and each ANN class corresponds to a respective data set. It will be understood by those skilled in the art that the term "ANN class" refers to an isotope class or identity determined by an ANN. Accordingly, multiple ANNs, rather than a single ANN, are trained using an ANN kit. Advantageously, the present invention provides improved generalization capabilities. Further advantageously, the variance of the prediction is reduced using the present invention.

[0020] In some embodiments, isotope probabilities are determined based on multiple ANN classes.

[0021] Preferably, each ANN comprises one or more parameters, the parameters being one or more selected from the following ranges: a hidden layer comprising a plurality of neurons; a learning rate; and a regularization strength.

[0022] Preferably, the parameters are selected by applying a random search method.

[0023] Preferably, the multi-class classifier is optimized.

[0024] In some embodiments, the multi-class classifier is optimized by error back propagation, and the multi-class classifier is configured to minimize a cost function with respect to network weights.

[0025] Preferably, the cost function is a cross entropy loss function for monoisotopic classification.

[0026] In some embodiments, the cost function is a mean square error cost function for isotope mixture classification.

[0027] In some embodiments, the cost function includes a penalty term, which is an L2 regularization element.

[0028] In an alternative embodiment, the multi-class classifier is optimized according to the scaled conjugate gradient method.

[0029] Preferably, the multi-class classifier is trained until a threshold difference between the predicted output and the correct output reaches a threshold.

[0030] Preferably, the isotopes are identified by kit prediction.

[0031] In some embodiments, the suite forecast includes: determining an isotope with a highest probability for each of a plurality of data sets; and identifying the isotope with the highest probability having the greatest number.

[0032] It will be appreciated that any feature described herein as suitable for incorporation into one or more aspects or embodiments of the present disclosure is intended to be generalized to any and all aspects and embodiments of the present disclosure. Other aspects of the present disclosure will be understood by those skilled in the art based on the specific embodiments, claims, and drawings of the present disclosure. The above general description and the following detailed description are exemplary and illustrative only and do not limit the claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] One or more embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings, in which:

[0034] Figure 1 It is a system used to identify isotopes;

[0035] Figure 2 Is used for Figure 1 Methods for systematic identification of isotopes;

[0036] Figure 3 is a histogram ratio matrix showing the extraction of bin-ratio vectors;

[0037] Figure 4 is a schematic diagram of an artificial neural network architecture; and

[0038] Figure 5 It is a method used to train artificial neural network models. DETAILED DESCRIPTION

[0039] Figure 1 A system 100 for identifying isotopes is shown. The system 100 includes a computing device 102 in communication with a spectrometer 104.

[0040] The spectrometer 104 is configured to receive gamma radiation from an isotope source and measure the spectral components (i.e., spectral data) of the isotope source. In the present embodiment, the spectrometer 104 is a scintillation counter 104, which is configured to detect and measure the intensity and / or energy of ionizing radiation emitted by the isotope and incident on the scintillation counter. For example, the scintillation counter 104 can measure, for example, 57 Co. 60 Co. 67 Ga, and 131 The energy of any one or more isotopes in I. Those skilled in the art will appreciate that this list of isotopes is for illustrative purposes only and that the scintillation counter 104 can measure the energy representing any known isotope.

[0041] The computing device 102 is configured to receive spectral data from the spectrometer 104 and execute the artificial neural network (ANN) suite model generation method 200 and the isotope identification method 55 to identify the isotopes.

[0042] Before executing the isotope identification method 500, use the isotope identification system 100 according to Figure 2 The training method 200 described in generates an artificial neural network (ANN) suite model or pre-trains an artificial neural network (ANN) suite model.

[0043] In a first step 202, the spectrometer detects a spectrum of gamma rays emitted by a plurality of known isotopes over a period of time to enable the determination of spectral counts. It will be understood by those skilled in the art that the term "known isotope" refers to an isotope known to the user. There may be any number of known isotopes used by the training process, such as 18 isotopes.

[0044] At step 204, computing device 102 collects or receives training spectral data for each of a plurality of known isotopes from spectrometer 104. The training spectral data corresponds to a detected spectrum of gamma rays received from the known isotopes. The spectral data represents energy associated with gamma rays received and measured by spectrometer 104. Accordingly, a gamma-ray spectrum is generated.

[0045] Alternatively, the spectral data can be filtered to match the characteristic resolution of the spectrometer 104, thereby reducing the artificial features generated by statistical noise. Further optionally, randomly generated gain shift noise can be added to the spectral data to simulate a real setting. Gain shift noise can be added by introducing a small random change in the linear term of the quadratic calibration curve of the spectrometer 104. Further optionally, a random background noise pattern can be added to the spectral data. A plurality of different randomly generated background noise patterns can be mixed in random proportions to generate a plurality of background spectra.

[0046] In step 206, for each known isotope in the plurality of known isotopes, the training spectral data is binned into n spectral bins. That is, each training spectral data is grouped according to its energy. For example, for n=1024, 1024 bins may be selected, each bin corresponding to a specific energy range. 7 The measurement energy range is about 10 eV, and each bin can span 3 More preferably, for 0-3x 10 6 The measurement energy range is 2.93 x 10 3 Accordingly, each spectral data is assigned to one of the 1024 bins corresponding to the measured energy.

[0047] In step 208, for each known isotope in the plurality of known isotopes, a training histogram is generated based on n spectral bins. The training histogram includes n items and provides an indication of how many training spectral data fall into each bin. In other words, the training histogram provides bin counts associated with each bin. In this embodiment, the training histogram includes 1024 items.

[0048] In step 210, a training histogram ratio matrix M is generated for each known isotope in the plurality of known isotopes. The training histogram ratio matrix M is a square matrix of length n. In this embodiment, the training histogram ratio matrix M is a square matrix of length 1024. The items M in the training histogram ratio matrix M are i,j is the bin count of the i-th bin divided by the bin count of the j-th bin.

[0049]

[0050] Those skilled in the art will recognize that the ratios are reciprocals of each other, so the upper and lower triangles of the matrix are identical and the matrix diagonal is constant.

[0051] At step 212, a plurality of training bin-ratio vectors are extracted from each of the training histogram ratio matrices M. Figure 3 It is easier to understand the extraction step. Figure 3 A training histogram ratio matrix 300 is depicted for 1024 bin histograms.

[0052] For a training histogram ratio matrix M of n bins, the number of bin-ratio vectors is n-1. Accordingly, a training histogram ratio matrix M with 1024 bins in the present embodiment has 1023 bin-ratio vectors. The entries of the yth training bin-ratio vector are bin-ratios of the first bin to the second bin, where the second bin is separated from the first bin by y bins. Specifically, the first training bin-ratio vector 302 (having a length of 1023) contains the ratios between bin counts of adjacent bins (i.e., separated by 1 bin). The second training bin-ratio vector 304 (having a length of 1022) contains the ratios between bin counts of the second adjacent bins. Due to the separation of 1023 bins, the final (i.e., 1023th) training bin-ratio vector 306 contains the ratio between the last bin count (i.e., the 1024th bin count) and the first bin count. Advantageously, the ratios measured by multiple distance scales between bins introduce feature diversity, thereby improving the robustness of the present method.

[0053] Each bin-ratio vector may be stored in its own dataset.

[0054] At step 214, a first set of isotope predictions is determined for a first bin-ratio vector 302 of each of the training histogram ratio matrices M. The predictions are achieved by inputting the first bin-ratio vector 302 into a first artificial neural network (ANN) model, such as, Figure 4 The ANN model described in .

[0055] like Figure 4 As shown in , ANN architecture 400 includes an input layer 402, a hidden layer 404, and an output layer 406. Each layer in layers 402, 404, 406 includes a plurality of artificial neurons (referred to herein as "nodes") connected via edges. Nodes and edges have associated weights configured to adjust the contribution of specific input data to other input data. The output of each layer in layers 402, 404, 406 is propagated to the next layer, and a function is applied to each output. An exemplary function is a linear rectification function.

[0056] ANN also includes multiple parameters. The multiple parameters include the number of neurons in the hidden layer, the learning rate, and the regularization strength.

[0057] The input layer 402 is configured to receive a bin-ratio vector. Specifically, the input layer 402 includes a plurality of input nodes 402A, 402B, 402C, each of which is configured to receive a respective item in the bin-ratio vector. Accordingly, there are N input nodes, where N is the length of the input bin-ratio vector. For example, in step 512, the input node 402A is configured to receive the first item of the first bin-ratio vector 302 of the histogram ratio matrix 300. The input node 402B is configured to receive the second item of the first bin-ratio vector 302 of the histogram ratio matrix 300. Finally, the input node 402C is configured to receive the 1023rd (i.e., last) item of the first bin-ratio vector 302 of the histogram ratio matrix 300.

[0058] The hidden layer 404 is configured to apply nonlinear transformation to the input received via the input nodes 402A, 402B, 402C. The hidden layer 404 includes a plurality of hidden layer nodes 404A, 404B, 404C, 404D, each of which is an ANN model. The hidden layer 404 may also include a plurality of layers.

[0059] In some embodiments, the hidden layer 404 may include a sigmoid activation layer (not shown), which is configured to apply a sigmoid activation function: 1 / (1+e ―x ). The sigmoid activation layer can be used when a single isotope needs to be identified. The sigmoid activation layer can also be used when multiple isotopes from a single radiation source need to be identified.

[0060] After applying the hidden layers, the output layer 406 is configured to provide an isotope prediction associated with each of the bin-ratio vector entries. Specifically, the output layer 406 includes a plurality of output nodes 406A, 406B, 406C, each of which is configured to output a respective isotope prediction. For example, the output node 406A may output a first isotope prediction associated with a first entry of the first bin-ratio vector 302 of the histogram ratio matrix 300, such as 241 The output node 406B may output a second isotope prediction associated with a second entry of the first bin-ratio vector 304 of the histogram ratio matrix 300, such as 137 The output node 406B may output a third isotope prediction associated with the 1023rd (ie, last) entry of the first bin-ratio vector 306 of the histogram ratio matrix 300, such as 152 Eu.

[0061] In the case of monoisotopic classification, a softmax function is applied to the output node.

[0062] In the case of a multi-label problem, the output layer 406 can be configured to use a sigmoid activation function: 1 / (1+e ―x ), where x is the weighted sum of the inputs through the sigmoid activation function, and then the weighted sum of x is used as the input to the next layer. Accordingly, by applying the sigmoid activation function, the output layer 406 can limit the output to values ​​between 0 and 1. In this case, a threshold can be implemented, where if the output layer provides a value greater than the threshold, a specific isotope can be identified. For example, if the threshold is 0.5, an output layer value greater than 0.5 can indicate a category that is about to associate the output with the output layer value.

[0063] Returning to method 200 , at step 216 , the cost function is evaluated for the first set of isotope predictions.

[0064] In the case of monoisotope classification, the cost function is the cross entropy loss function:

[0065]

[0066] Where K is the total number of output nodes 406, t i is the correct isotopic type, and is the predicted class provided by the respective output node 406. In this case, K is 1023 for the first bin-ratio vector.

[0067] In the case of multiple isotope classification, the cost function is a mean squared error function with L2 regularization to penalize higher values ​​of network weights and prevent overfitting:

[0068]

[0069] Among them, w j is the weight and λ is the regularization strength, t i is the correct isotopic type, and is the predicted category provided by the respective output node 406 .

[0070] In step 218, a second set of isotope predictions is determined for the second bin-ratio vector 304 of each training histogram ratio matrix M in the training histogram ratio matrix M. The prediction is implemented in the same manner as the prediction of step 512, except that a second ANN model is used. In this example, because the length of the second bin-ratio vector 304 is 1022, the second ANN model includes 1022 input nodes and output nodes. The parameters of the second ANN model are configured to minimize the cost function determined in step 514. Accordingly, the new ANN model is added to the ANN suite, and thus the ANN suite to which the new ANN model is added contains a second bin-ratio vector associated with each known isotope in the plurality of known isotopes.

[0071] Steps 216 and 218 are repeated, with the output of step 218 being used to evaluate the cost function in a subsequent iteration of step 216. In addition, each subsequent iteration of step 516 utilizes a subsequent bin-ratio vector for each known isotope in the plurality of known isotopes.

[0072] Steps 216 and 218 may be repeated until a threshold is reached. The threshold may be an output difference threshold. For example, when the predicted category When the difference between the correct isotope class and the output difference threshold is less than or equal to the output difference threshold, the method 500 can end. An exemplary output difference threshold can be 5×10 -2 The threshold value may provide an indication as to whether the ANN model satisfies the performance metric.

[0073] Accordingly, each subsequent ANN is trained by error back-propagation to minimize the cost function with respect to the node weights. The scaled conjugate gradient (SCG) method can be used to speed up convergence.

[0074] Now turn Figure 5 , describes the Figure 1 The system 100 is provided with a method 200 for identifying isotopes.

[0075] In a first step 502, the computing device 102 acquires or receives spectral data from the spectrometer 104. The spectral data represents energy associated with gamma rays received and measured by the spectrometer 104. Accordingly, a gamma-ray spectrum may be generated.

[0076] Optionally, the spectral data may be filtered to match the characteristic resolution of the spectrometer 104 to reduce artifacts resulting from statistical noise.

[0077] In step 504, the spectral data is binned into n spectral bins. That is, each spectral data is grouped according to its energy. For example, for n=1024, 1024 bins may be selected, each bin corresponding to a specific energy range.4 -10 7 The measurement energy range is about 10 eV, and each bin can span 3 Accordingly, each spectral data is assigned to one of the 1024 bins corresponding to the measured energy.

[0078] At step 506, a histogram is generated based on the n spectral bins. The histogram includes n entries and provides an indication of how much spectral data falls into each bin. In other words, the histogram provides a bin count associated with each bin. In this embodiment, the histogram includes 1024 entries.

[0079] In step 508, a histogram ratio matrix M is generated. The histogram ratio matrix M is a square matrix of length n. In this embodiment, the histogram ratio matrix M is a square matrix of length 1024. The items M in the histogram ratio matrix M are i,j is the bin count of the i-th bin divided by the bin count of the j-th bin.

[0080]

[0081] Those skilled in the art will recognize that the ratios are reciprocals of each other, so the upper and lower triangles of the matrix are identical and the matrix diagonal is constant.

[0082] At step 510, a plurality of bin-ratio vectors are extracted from the histogram ratio matrix M. Figure 3 It is easier to understand the extraction step. Figure 3 A histogram ratio matrix 300 is depicted for 1024 bin histograms.

[0083] For a histogram ratio matrix M of n bins, the number of bin-ratio vectors is n-1. Accordingly, a histogram ratio matrix M with 1024 bins in the present embodiment has 1023 bin-ratio vectors. The entries of the yth bin-ratio vector are bin-ratios of the first bin to the second bin, where the second bin is separated from the first bin by y bins. Specifically, the first bin-ratio vector 302 (having a length of 1023) contains the ratios between bin counts of adjacent bins (i.e., separated by 1 bin). The second bin-ratio vector 304 (having a length of 1022) contains the ratios between bin counts of the second adjacent bin. Due to the separation of 1023 bins, the final (i.e., 1023th) bin-ratio vector 306 contains the ratios between the last (i.e., 1024th bin count) and the first bin count. Advantageously, the ratios measured by multiple distance scales between bins introduce feature diversity, thereby improving the robustness of the present method.

[0084] Each bin-ratio vector may be stored in its own dataset.

[0085] At step 512, an isotope prediction is determined for each bin-ratio vector in the bin-ratio vectors. Figure 4 As described in , for each bin-ratio vector in the bin-ratio vector, an artificial neural network (ANN) is used. The isotope prediction can be in the form of an isotope probability, so that the ANN model provides the probability that the radiation source is a specific isotope. Accordingly, for this embodiment, the isotope prediction associated with each bin-ratio vector of the histogram ratio matrix 300 is generated by the method 200. Reference Figure 4 and Figure 5 Discuss further details about ANN models and architectures.

[0086] In the last step 514 of method 200, the final isotope prediction is determined. The final isotope prediction is determined by the suite prediction step. Specifically, the isotope predictions associated with each bin-ratio vector in the bin-ratio vector are combined as determined in step 512 of method 500. In this embodiment, the output of step 514 is the output of the majority isotope prediction associated with each bin-ratio vector in the bin-ratio vector. That is, the isotope with the most isotope predictions or the highest cumulative probability will be output by step 514.

[0087] In this embodiment, the isotope is outputted through 879 output nodes of the output layer 406. 57 Co. Accordingly, the final isotope prediction is 57 Co.

[0088] The descriptions provided herein may be directed to specific implementations.It should be understood that the discussion provided herein is provided for the purpose of enabling one of ordinary skill in the art to make and use any subject matter defined herein by the claimed subject matter.

[0089] It should be noted that the subject matter of the claims is not limited to the implementations and illustrations provided herein, but includes variations of these implementations that include portions of the implementations and combinations of elements of different implementations according to the claims. It should be recognized that, as in any engineering or design project, when developing any such implementation, multiple specific implementation decisions should be made to achieve the developer's specific goals, such as complying with system and business-related restrictions, which may vary from implementation to implementation. Moreover, it should be recognized that the development activities may be complex and time-consuming, but for those of ordinary skill in the art who benefit from the present invention, design, preparation, and production remain routine tasks.

[0090] Various implementations of the present invention have been described in detail, and the accompanying drawings and schematic diagrams are examples of these embodiments. In the detailed description, a plurality of specific details are listed to provide a thorough understanding of this invention. However, the invention provided herein may also be implemented without these specific details. In some other examples, well-known methods, processes, components, circuits, and networks are not described in detail to avoid unnecessarily covering up the details of the embodiments.

[0091] It should also be understood that, although the terms first, second, etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element. The first element and the second element are both elements, respectively, but the two are not considered to be the same element.

[0092] The terms used in the description of the invention provided herein are used for the purpose of describing specific implementations and are not intended to limit the invention provided herein. As used in the description of the invention provided herein and the appended claims, the singular forms "a", "an", and "the" also include plural forms, unless the context clearly indicates otherwise. The term "and / or" used herein refers to and encompasses any and all possible combinations of one or more related listed items. When used in this specification, the terms "includes", "including", "comprises", and / or "comprising" refer to the presence of stated features, integers, steps, operations, elements, and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, parts, and / or combinations thereof.

Claims

1. An isotope identification method performed by a computing device, comprising the following steps: Collecting a plurality of spectral data from a spectral device; generating, by a processor of the computing device, a plurality of data sets based on the plurality of spectral data; determining, by the processor, a plurality of isotope probabilities, each isotope probability in the plurality of isotope probabilities corresponding to a respective data set in the plurality of data sets; wherein the isotope probability represents the probability that a corresponding data set among the multiple data sets corresponds to a specific isotope; and An isotope is identified, by the processor, based on the plurality of isotope probabilities.

2. The method according to claim 1, wherein: Each data set of the plurality of data sets includes binned data, the data set representing a functional relationship between a first bin of the binned data and a second bin of the binned data.

3. The method according to claim 2, wherein: Each of the plurality of data sets includes a respective bin-ratio vector.

4. The method according to claim 3, wherein: Generating the plurality of data sets comprises: Binning the plurality of spectral data into n spectral bins by the processor; generating, by the processor, a plurality of bin-ratio vectors; and A respective bin-ratio vector is stored, by the processor, in each of the plurality of data sets.

5. The method according to claim 4, wherein: Generating the plurality of bin-ratio vectors comprises: generating, by the processor, a histogram based on the n spectral bins; generating, by the processor, a histogram ratio matrix M based on the histogram; Wherein, the histogram ratio matrix M is a square matrix of length n; and extracting, by the processor, the plurality of bin-ratio vectors from the histogram ratio matrix; There are k bin-ratio vectors, and k=n-1.

6. The method according to claim 5, wherein: Each entry M in the histogram ratio matrix M i,j is the i-th bin divided by the j-th bin.

7. The method according to claim 6, wherein: The entries of the y-th bin-ratio vector are the bin-ratios of a first bin to a second bin that is separated from the first bin by y bins.

8. The method according to claim 2, wherein: The bin data includes bin differences.

9. A method according to any preceding claim, wherein: The isotopic probabilities are determined using a set comprising a plurality of classes generated by a multi-class classifier, each class corresponding to a respective data set.

10. The method according to claim 9, wherein: The kit is an artificial neural network ANN kit; the ANN kit includes multiple ANN categories, each ANN category is generated by an ANN, and each ANN category corresponds to a respective data set.

11. The method according to claim 10, wherein: The isotope probability is determined based on the plurality of ANN categories.

12. The method according to any one of claims 9 to 11, wherein: Each ANN includes one or more parameters, wherein the parameters are one or more selected from the following ranges: Hidden layers, including multiple neurons; learning rate; and Regularization strength.

13. The method according to claim 12, wherein: The parameters are selected by applying a random search method.

14. The method according to claim 9, wherein: The multi-class classifier is optimized.

15. The method according to claim 14, wherein: The multi-class classifier is optimized by error back propagation, and the multi-class classifier is configured to minimize a cost function with respect to network weights.

16. The method according to claim 15, wherein: The cost function is a cross entropy loss function for monoisotope classification.

17. The method according to claim 16, wherein: The cost function is a mean square error cost function for isotope mixture classification.

18. The method according to claim 17, wherein: The cost function includes a penalty term, which is an L2 regularization element.

19. The method according to claim 14, wherein: The multi-class classifier is optimized according to the scaled conjugate gradient method.

20. The method according to claim 14, wherein: The multi-class classifier is trained until a threshold difference between the predicted output and the correct output reaches a threshold.

21. A method according to any preceding claim, wherein: The isotopes are identified by kit prediction.

22. The method according to claim 21, wherein: The kit forecast includes: determining, for each of the plurality of data sets, an isotope having a highest probability; and Identify the isotope with the highest probability that has the greatest number.

Images