A method and system for automatically identifying abnormal monitoring data of radiation field

Abstract

The invention discloses an automatic identification method and system for abnormal monitoring data of a radiation field, which relates to the field of radiation monitoring technology and comprises the following steps: collecting radiation dose rate data of each monitoring node in a radiation field generated by a historical medical field; extracting features from the collected data to obtain feature values; using the obtained feature values ​​as samples and using the corresponding state marks as labels to train an initial neural network to obtain a radiation dose rate state determination model; collecting real-time radiation dose rate data of each monitoring node, extracting feature values ​​of the real-time data as inputs of the trained radiation dose rate state determination model, obtaining the probability of the existence of an abnormal situation at each monitoring node, generating a plurality of local decisions according to the generated abnormal situation probabilities, fusing the local decisions with the missed judgment rate and the wrong judgment rate of each monitoring node, generating a fusion center decision, and finally determining whether there is an abnormal situation in the radiation field according to the fusion center decision.

Description

Technical Field

[0001] The present invention relates to the field of radiation monitoring technology, and in particular to a method and system for automatically identifying abnormal monitoring data of a radiation field. Background Art

[0002] Modern medical equipment is widely used in the diagnosis and treatment process, some of which will produce ionizing radiation, such as X-ray machines, CT scanners, radiotherapy equipment, etc. In order to ensure the safety of medical staff and patients, the radiation levels of these devices must be strictly monitored. Therefore, it is particularly important to establish an effective radiation data monitoring system.

[0003] Ionizing radiation has high energy and can cause damage to cells and DNA. Long-term exposure may lead to serious health problems such as cancer and genetic damage. Radioactive equipment is frequently used in hospitals, and medical staff and patients may be exposed to radiation. In order to effectively manage and control radiation doses, radiation levels must be monitored and managed in real time: With the rapid development of information construction, radiation monitoring nodes can continuously upload real-time monitoring information of acquired radiation monitoring data through the radiation monitoring network, realize real-time data collection and dynamic monitoring, and provide a data basis for radiation hazard information fusion. However, the current measurement data processing level is low, and the processing method is single.

[0004] Therefore, accurately judging whether radiation data anomalies occur in the area is of great significance for radiation reconnaissance and protection actions. However, due to technical limitations and the complex environment and weather of the hospital, a single radiation monitoring device will inevitably miss or misjudge. On-site reconnaissance can further determine whether radiation conditions occur, but on-site reconnaissance requires highly professional and technical personnel to enter the accident area, and there is a greater risk if the degree of radiation hazard is not understood. Therefore, it is necessary to develop an automatic identification method for abnormal monitoring data of radiation fields, so that it can accurately and efficiently automatically identify abnormal data, so as to better meet the use requirements of medical equipment and improve the level of radiation protection management.

[0005] In the prior art, the publication number CN114386477A discloses a method for automatically identifying abnormal monitoring data of radiation fields. By using the measurement data of multiple points of the radiation field as a data set, the abnormal nodes of the radiation field data stream are identified according to the potential laws of the radiation field data in space and time scales. It can be applied to radiation fields with multiple working conditions. Based on the isolation forest technology, the radiation field monitoring data set is trained, the radiation field data stream is scored and detected in real time, and the new data is added to the training set for iterative calculation and analysis. This method improves the work efficiency of operation and maintenance personnel, better meets the requirements of on-site use of facilities, improves the level of radiation protection management, and has high efficiency in identifying abnormal data. However, in actual use, due to interference in the actual environment, the monitoring nodes may be missed or misjudged, resulting in reduced credibility and practicality of the detection results.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The purpose of the present invention is to provide a method and system for automatically identifying abnormal radiation field monitoring data to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for automatically identifying abnormal radiation field monitoring data, the specific steps comprising:

[0010] Acquire abnormal state radiation dose rate data and normal state radiation dose rate data of each monitoring node in the radiation field generated by the medical venues in several equal-length time periods in history, wherein the abnormal state and the normal state are two states that may affect the human body and will not affect the human body, respectively;

[0011] Extracting data time domain characteristic values ​​according to the collected radiation dose rate data; the time domain characteristic values ​​include maximum value, minimum value, mean value and standard deviation;

[0012] Using the obtained data time domain feature values ​​as samples and the corresponding state marks as labels, the initial neural network is trained to obtain a radiation dose rate state determination model;

[0013] Acquire real-time radiation dose rate data of each monitoring node in the radiation field generated by the medical setting; extract real-time time domain characteristic values ​​from the real-time radiation dose rate data, input the real-time time domain characteristic values ​​into the obtained radiation dose rate state determination model, and obtain the probability that the radiation dose rate of each monitoring node belongs to an abnormal state;

[0014] According to the comparison between the probability that the radiation dose rate of each monitoring node belongs to an abnormal state and the preset abnormal state probability threshold, the local judgment result of each monitoring node is obtained;

[0015] The local judgment results of each monitoring node are combined with the missed judgment rate and wrong judgment rate of each monitoring node to perform information fusion and generate the fusion center judgment. According to the different judgment results of the fusion center, it is judged whether there is abnormal monitoring data in the radiation field.

[0016] Furthermore, the abnormal state radiation dose rate data and normal state radiation dose rate data of each monitoring node in the radiation field generated by the medical venues within several equal-length time periods in history are obtained, and data for T time periods are collected in total, of which the abnormal state radiation dose rate data accounts for 50%, that is, a total of T*0.5 groups of abnormal state radiation dose rate data are collected, and the remaining 50% are normal state radiation dose rate data. The setting method of each monitoring node is to take the radiation generating device as the center of the circle, and arrange them in an equidistant manner in a direction away from the radiation generating device, and the number of monitoring nodes is not less than 1.

[0017] Furthermore, the time domain characteristic value of the radiation dose rate data of each monitoring node is extracted, wherein the formula for extracting the time domain characteristic value is as follows:

[0018] The calculation formula for the maximum and minimum radiation dose rate is:

[0019] F max,j =max(F 1,j ,F 2,j ,…,F n,j )

[0020] F min,j =min(F 1,j ,F 2,j ,…,F n,j )

[0021] In the formula, F max,j and F min,j They represent the maximum and minimum values ​​of the radiation dose rate of the jth monitoring node respectively; n is the number of radiation dose rate data collected in each group;

[0022] The formula for calculating the mean and standard deviation is:

[0023]

[0024] In the formula, F μj and F σj They represent the mean and standard deviation of the radiation dose rate monitored by the jth monitoring node, F i,jIt represents the radiation dose rate of the jth monitoring node at the i-th collection time, where j = 1, 2, ..., m, m represents the number of monitoring nodes, and i = 1, 2, ..., n, where n is the number of radiation dose rate data collected in each group.

[0025] Furthermore, the initial neural network is trained based on the obtained data time domain feature values ​​as samples and the corresponding state marks as labels to obtain a radiation dose rate state determination model, wherein the initial neural network uses a SVM model.

[0026] Furthermore, according to the trained SVM radiation dose rate state judgment model, its output value is converted into a probability value using the probability estimation method, which is achieved through Platt scaling, that is, a sigmoid function is applied to the output of the SVM decision function to map it to a probability value between 0 and 1, where the formula based on Platt scaling is:

[0027]

[0028] In the formula, svm outj It represents the abnormal state probability value of the j-th monitoring node converted by the decision function output of SVM, f(x) is the decision function output of SVM, A is the slope parameter of the sigmoid function, and B is the offset parameter of the sigmoid function.

[0029] Furthermore, according to the comparison between the converted abnormal state probability value and the preset abnormal state probability threshold value yz, a corresponding local decision is output according to different comparison results, wherein the logic of outputting the local decision is:

[0030] When svm outj When ≥0.5*yz, the output local decision is 1;

[0031] When svm outj When <0.5*yz, the output local decision is 0;

[0032] The output local judgment of 1 indicates that the radiation dose rate of the jth monitoring node is abnormal, that is, the radiation dose rate is greater than the radiation dose rate threshold K. The output local judgment of 0 indicates that the radiation dose rate of the jth monitoring node is within the normal range, that is, the radiation dose rate is less than the radiation dose rate threshold K.

[0033] Furthermore, the local decision of each monitoring node is combined with the missed judgment rate and wrong judgment rate of each monitoring node to perform information fusion and generate the fusion center decision. The formula for information fusion is:

[0034]

[0035] Where ZH is the fusion center decision, u j is the local decision output of each monitoring node, P Mj and P Lj are the missed detection rate and wrong detection rate of the jth monitoring node respectively, and τ is the decision threshold.

[0036] Furthermore, according to the generated fusion center decision ZH, the logic for judging whether an abnormal situation occurs is as follows:

[0037] When ZH≥0, it indicates that an abnormal situation has occurred;

[0038] When ZH<0, it means that there is a misjudgment in the local judgment and there is no abnormality in the radiation field.

[0039] The present invention also provides a system for automatically identifying radiation field abnormality monitoring data, and the system for automatically identifying radiation field abnormality monitoring data is used to execute the above-mentioned method for automatically identifying radiation field abnormality monitoring data, including:

[0040] A historical data acquisition module is used to obtain abnormal state radiation dose rate data and normal state radiation dose rate data of each monitoring node in the radiation field generated by the medical venue in a number of equal long time periods in history, wherein the abnormal state and the normal state are two states that will affect the human body and will not affect the human body respectively;

[0041] A feature extraction module, used to extract the time domain feature values ​​of the data according to the collected radiation dose rate data; the time domain feature values ​​include maximum value, minimum value, mean value and standard deviation;

[0042] A state determination model generation module is used to train the initial neural network based on the obtained data time domain characteristic values ​​as samples and the corresponding state marks as labels to obtain a radiation dose rate state determination model;

[0043] A real-time data detection module is used to obtain real-time radiation dose rate data of each monitoring node in the radiation field generated by the medical venue; extract real-time time domain characteristic values ​​from the real-time radiation dose rate data, input the real-time time domain characteristic values ​​into the obtained radiation dose rate state determination model, and obtain the probability that the radiation dose rate of each monitoring node belongs to an abnormal state;

[0044] A local decision module is used to obtain a local decision result of each monitoring node according to the comparison between the probability that the radiation dose rate of each monitoring node belongs to an abnormal state and a preset abnormal state probability threshold;

[0045] The central decision generation module is used to combine the local decision results of each monitoring node with the missed judgment rate and wrong judgment rate of each monitoring node, perform information fusion, generate a fusion center decision, and judge whether there is abnormal monitoring data in the radiation field based on the different decision results of the fusion center.

[0046] Compared with the prior art, the beneficial effects of the present invention are as follows: by collecting radiation dose rate data of each monitoring node in the radiation field generated by historical medical places; extracting features from the collected data to obtain time domain feature values; using the obtained time domain feature values ​​as samples, and using the corresponding state marks as labels, training the initial neural network to obtain a radiation dose rate state determination model; collecting real-time radiation dose rate data of each monitoring node, extracting feature values ​​of the real-time data as inputs of the trained radiation dose rate state determination model, obtaining the probability of abnormal conditions in each monitoring node, generating a number of local decisions based on the generated abnormal condition probabilities, fusing the local decisions with the missed judgment rate and the wrong judgment rate of each monitoring node, generating a fusion center decision, and finally determining whether there is an abnormal condition in the radiation field based on the fusion center decision; the method determines whether a local abnormal condition occurs through a neural network, and then through the detection fusion technology, the local decisions made by each monitoring device are correlated and optimized under certain criteria to generate a global decision result, so that the credibility of the detection result is higher, and the risk of missed judgment and false judgment is also greatly reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0048] Figure 2 It is a schematic diagram of the overall system structure of the present invention. DETAILED DESCRIPTION

[0049] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0050] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0051] Example:

[0052] See also Figure 1, the present invention provides a technical solution:

[0053] A method for automatically identifying abnormal radiation field monitoring data, the specific steps comprising:

[0054] Step 1: Acquire the abnormal state radiation dose rate data and normal state radiation dose rate data of each monitoring node in the radiation field generated by the medical venues in several equal time periods in history, wherein the abnormal state and the normal state are two states that will affect the human body and will not affect the human body respectively;

[0055] The abnormal state radiation dose rate data and normal state radiation dose rate data of each monitoring node in the radiation field generated by the medical venues in several equal-length time periods in history are obtained, and data for T time periods are collected in total, of which the abnormal state radiation dose rate data accounts for 50%, that is, a total of T*0.5 groups of abnormal state radiation dose rate data are collected, and the remaining 50% are normal state radiation dose rate data. The setting method of each monitoring node is to take the radiation-generating device as the center of the circle, and arrange them in an equidistant manner in a direction away from the radiation-generating device, and the number of monitoring nodes is not less than 1.

[0056] Step 2: extracting the time domain characteristic values ​​of the data according to the collected radiation dose rate data; the time domain characteristic values ​​include maximum value, minimum value, mean value and standard deviation;

[0057] The time domain characteristic value of the radiation dose rate data of each monitoring node is extracted, and the formula for extracting the time domain characteristic value is:

[0058] The calculation formula for the maximum and minimum radiation dose rate is:

[0059] F max,j =max(F 1,j ,F 2,j ,…,F n,j )

[0060] F min,j =min(F 1,j ,F 2,j ,…,F n,j )

[0061] In the formula, F max,j and F min,j They represent the maximum and minimum values ​​of the radiation dose rate of the jth monitoring node respectively; n is the number of radiation dose rate data collected in each group;

[0062] The formula for calculating the mean and standard deviation is:

[0063]

[0064] In the formula, F μj and Fσj They represent the mean and standard deviation of the radiation dose rate monitored by the jth monitoring node, F i,j represents the radiation dose rate of the jth monitoring node at the ith collection time, where j = 1, 2, ..., m, m represents the number of monitoring nodes, i = 1, 2, ..., n, where n is the number of radiation dose rate data collected in each group; based on the data collected in T time periods, that is, the radiation dose rate data of each monitoring node of T groups are collected, and a total of T groups of characteristic values ​​are extracted.

[0065] The advantage of feature extraction is that it does not require uploading all the original information. Data can be compressed before fusion, and the number of transmissions is relatively reduced, which is conducive to real-time processing of information. The time domain features of the data are extracted as the input of the neural network. Its main function is to provide key information of the data and help the neural network better understand and process the data. The time domain features can highlight the key patterns and trends in time series data, such as periodicity, trend, peak, valley, etc. These features help the neural network capture important changes in the data and improve the prediction accuracy and generalization ability of the model. Therefore, feature extraction is a very important and effective way of data processing.

[0066] Step 3: Using the obtained data time domain feature values ​​as samples and the corresponding state marks as labels, the initial neural network is trained to obtain a radiation dose rate state determination model;

[0067] The initial neural network uses the SVM model. The Support Vector Machine (SVM) is a supervised learning model used for classification and regression analysis. The SVM classification model performs well in processing high-dimensional data and small sample data sets. The SVM model can be used to distinguish radiation dose rate data into two categories: normal data and abnormal data.

[0068] The kernel function of the SVM classification model is selected as the Gaussian kernel function, and the particle swarm algorithm is used to determine the optimal penalty coefficient C and kernel function parameter g. At the same time, the kernel function K(x s ,x l ); According to the determined optimal penalty coefficient C, kernel function parameter g and kernel function K(x s ,x l ), constructs an optimization problem to solve, where the Gaussian kernel is selected as the kernel function, which is the core of the radial basis function network, and the kernel function K(c s ,l) is based on the formula:

[0069] K(x s ,x l ) = exp(-g‖x s -x l ‖ 2 )

[0070] Where g is the kernel function parameter, x s is the sth sample vector in the input space, x l is the lth sample vector in the input space; the sample vector data is the eigenvalue corresponding to each group of sampling data.

[0071] Step 4: obtaining real-time radiation dose rate data of each monitoring node in the radiation field generated by the medical setting; extracting real-time time domain feature values ​​from the real-time radiation dose rate data, inputting the real-time time domain feature values ​​into the obtained radiation dose rate state determination model, and obtaining the probability that the radiation dose rate of each monitoring node belongs to an abnormal state;

[0072] According to the trained SVM radiation dose rate state judgment model, its output value is converted into a probability value using the probability estimation method, which is achieved through Platt scaling, that is, a sigmoid function is applied to the output of the SVM decision function to map it to a probability value between 0 and 1. The formula for Platt scaling is:

[0073]

[0074] In the formula, svm outj It represents the abnormal state probability value of the j-th monitoring node converted by the decision function output of SVM, f(x) is the decision function output of SVM, A is the slope parameter of the sigmoid function, and B is the offset parameter of the sigmoid function. In f(x), x is the characteristic value of the SVM model input, where A and B are both real numbers and both A and B are not 0. When A is 0, the sigmoid function becomes a constant function. When B is 0, the flexibility of the sigmoid function is limited, which may lead to inaccurate probability output, thereby affecting the classification performance of the model.

[0075] Step 5: According to the probability that the radiation dose rate of each monitoring node belongs to an abnormal state, the local judgment result of each monitoring node is obtained by comparing it with the preset abnormal state probability threshold;

[0076] According to the comparison between the converted abnormal state probability value and the preset abnormal state probability threshold yz, the corresponding local decision is output according to different comparison results, wherein the logic of outputting the local decision is:

[0077] When svm outj When ≥0.5*yz, the output local decision is 1;

[0078] When svm outj When <0.5*yz, the output local decision is 0;

[0079] The output local decision of 1 indicates that the radiation dose rate of the j-th monitoring node is abnormal, and the output local decision of 0 indicates that the radiation dose rate of the j-th monitoring node is within the normal range.

[0080] The local judgment is the preliminary judgment of the system, by setting the abnormal state probability threshold yz, where the setting range of the abnormal state probability threshold yz is: 0.9*svm outj ≤yz≤1.1*svm outj , the minimum abnormal state probability threshold yz is set to 0.9*svm outj To avoid misjudgment due to the abnormal state probability threshold yz being too small, the maximum value of the abnormal state probability threshold yz is set to 1.1*svm outj , to avoid missed judgments; compare the probability that the real-time data output by the neural network model is abnormal data with it. When the probability value of the real-time data of a monitoring node being in an abnormal state exceeds 0.5*yz, it is preliminarily judged that the monitoring node has an abnormal situation, and the local judgment output is 1. When the probability value of the real-time data of a monitoring node being in an abnormal state is less than 0.5*yz, it is preliminarily judged that the monitoring node has no abnormal situation, and the local judgment output is 0.

[0081] Step 6: Combine the local judgment results of each monitoring node with the missed judgment rate and wrong judgment rate of each monitoring node, perform information fusion, generate the fusion center judgment, and judge whether there is abnormal monitoring data in the radiation field according to the different judgment results of the fusion center.

[0082] The local judgment of each monitoring node is combined with the missed judgment rate and wrong judgment rate of each monitoring node to perform information fusion and generate the fusion center judgment. The formula for information fusion is:

[0083]

[0084] Where ZH is the fusion center decision, u j is the local decision output of each monitoring node, P Mj and P Lj are the missed detection rate and wrong detection rate of the jth monitoring node respectively, and τ is the decision threshold.

[0085] Where P Mj and P Lj are the missed detection rate and false positive rate of the jth monitoring node, respectively. The missed detection rate of monitoring equipment is a key indicator for evaluating its performance and reliability, especially in critical radiation facility environments. The missed detection rate is usually defined as the proportion of events that the monitoring equipment fails to detect in actual events; the false positive rate is one of the important indicators for evaluating the performance of radiation monitoring equipment, indicating the frequency of the equipment erroneously issuing an alarm when no actual event occurs. In this embodiment, the missed detection rate of each monitoring device is set to PMj = 8% generates random numbers, and the error rate is P Lj =5% generates random numbers. In practical applications, the missed detection rate and wrong detection rate can be calculated based on the actual monitoring data of each monitoring device; the calculation is based on the formula:

[0086]

[0087] Where τ is the decision threshold, and the formula for the decision threshold τ is:

[0088]

[0089] In the formula, P0 and P1 represent the prior probability of the center decision respectively. Assuming P0=P1=0.5, C 00 It means the cost when the local judgment output is 0 and the fusion center judges that there is no abnormal situation, C 10 It means the cost of the fusion center’s judgment that there is an abnormal situation when the local judgment output is 0, C 01 It means that when the local judgment output is 1, the fusion center judges that there is no abnormal situation, and the cost is C 11 It means that when the local judgment output is 1, the fusion center judges that there is an abnormal situation. The cost of making an incorrect judgment is greater than the cost of making a correct judgment, so C 10 >C 00 , C 10 >C 11 In this embodiment, C is set 10 =C 01 =1, C 00 =C 11 =0, that is, when the fusion center makes a correct judgment on the radiation event, the cost is set to 0; when it makes an incorrect judgment on the radiation event, the cost is set to 1.

[0090] According to the generated fusion center judgment ZH, the logic for judging whether an abnormal situation occurs is:

[0091] When ZH≥0, it indicates that an abnormal situation has occurred;

[0092] When ZH<0, it means that there is a misjudgment in the local judgment and there is no abnormality in the radiation field.

[0093] When the calculation result of ZH is greater than or equal to 0, the judgment confirms that an abnormal situation has occurred, that is, the judgment result of the processing center is that the local judgment is 1, confirming that an abnormal situation has occurred;

[0094] When the calculation result of ZH is less than 0, the judgment confirms that no abnormal situation has occurred, that is, the judgment result of the processing center is that the local judgment is 0, confirming that no abnormal situation has occurred.

[0095] See also Figure 2 The present invention also provides a system for automatically identifying radiation field abnormality monitoring data, and the system for automatically identifying radiation field abnormality monitoring data is used to execute the above-mentioned method for automatically identifying radiation field abnormality monitoring data, including:

[0096] A historical data acquisition module is used to obtain abnormal state radiation dose rate data and normal state radiation dose rate data of each monitoring node in the radiation field generated by the medical venue in a number of equal long time periods in history, wherein the abnormal state and the normal state are two states that will affect the human body and will not affect the human body respectively;

[0097] A feature extraction module, used to extract the time domain feature values ​​of the data according to the collected radiation dose rate data; the time domain feature values ​​include maximum value, minimum value, mean value and standard deviation;

[0098] A state determination model generation module is used to train the initial neural network based on the obtained data time domain characteristic values ​​as samples and the corresponding state marks as labels to obtain a radiation dose rate state determination model;

[0099] A real-time data detection module is used to obtain real-time radiation dose rate data of each monitoring node in the radiation field generated by the medical venue; extract real-time time domain characteristic values ​​from the real-time radiation dose rate data, input the real-time time domain characteristic values ​​into the obtained radiation dose rate state determination model, and obtain the probability that the radiation dose rate of each monitoring node belongs to an abnormal state;

[0100] A local decision module is used to obtain a local decision result of each monitoring node according to the comparison between the probability that the radiation dose rate of each monitoring node belongs to an abnormal state and a preset abnormal state probability threshold;

[0101] The central decision generation module is used to combine the local decision results of each monitoring node with the missed judgment rate and wrong judgment rate of each monitoring node, perform information fusion, generate a fusion center decision, and judge whether there is abnormal monitoring data in the radiation field based on the different decision results of the fusion center.

[0102] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0103] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0104] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0105] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for automatically identifying abnormal radiation field monitoring data, characterized in that: The specific steps include: Acquire abnormal state radiation dose rate data and normal state radiation dose rate data of each monitoring node in the radiation field generated by the medical venues in several equal-length time periods in history, wherein the abnormal state and the normal state are two states that may affect the human body and will not affect the human body, respectively; Data from T time periods are collected, of which 50% are in abnormal state. That is, T*0.5 groups of data from abnormal state are collected, and the remaining 50% are in normal state. The setting method of each monitoring node is to take the device generating radiation as the center of the circle and arrange them in an equidistant manner in the direction away from the device generating radiation. The number of monitoring nodes is not less than 1. Extracting data time domain characteristic values ​​according to the collected radiation dose rate data, wherein the time domain characteristic values ​​include maximum value, minimum value, mean value and standard deviation; The time domain characteristic value of the radiation dose rate data of each monitoring node is extracted, and the formula for extracting the time domain characteristic value is: The calculation formula for the maximum and minimum radiation dose rate is: F max，j =max(F 1，j ,F 2，j ,…,F n，j ) F min，j =min(F 1，j ,F 2，j ,…,F n，j ) In the formula, F max，j and F min，j They represent the maximum and minimum values ​​of the radiation dose rate of the jth monitoring node respectively; n is the number of radiation dose rate data collected in each group; The formulas for calculating the mean and standard deviation are: In the formula, F μj and F σj They represent the mean and standard deviation of the radiation dose rate monitored by the jth monitoring node, F i，j represents the radiation dose rate of the jth monitoring node at the i-th collection time, where j = 1, 2, ..., m, m represents the number of monitoring nodes, and i = 1, 2, ..., n, where n is the number of radiation dose rate data collected in each group; The initial neural network is trained using the obtained data time domain feature values ​​as samples and the corresponding state marks as labels to obtain a radiation dose rate state determination model; The initial neural network uses the SVM model; Acquire real-time radiation dose rate data of each monitoring node in the radiation field generated by the medical setting; extract real-time time domain characteristic values ​​from the real-time radiation dose rate data, input the real-time time domain characteristic values ​​into the obtained radiation dose rate state determination model, and obtain the probability that the radiation dose rate of each monitoring node belongs to an abnormal state; According to the trained SVM radiation dose rate state judgment model, its output value is converted into a probability value using the probability estimation method, which is achieved through Platt scaling, that is, a sigmoid function is applied to the output of the SVM decision function to map it to a probability value between 0 and 1. The formula based on Platt scaling is: In the formula, svm outj represents the abnormal state probability value of the j-th monitoring node converted by the decision function output of SVM, f(x) is the decision function output of SVM, A is the slope parameter of the sigmoid function, and B is the offset parameter of the sigmoid function; According to the comparison between the probability that the radiation dose rate of each monitoring node belongs to an abnormal state and the preset abnormal state probability threshold, the local judgment result of each monitoring node is obtained; According to the comparison between the converted abnormal state probability value and the preset abnormal state probability threshold yz, the corresponding local decision is output according to different comparison results, wherein the logic of outputting the local decision is: When svm outj When ≥0.5*yz, the output local decision is 1; When svm outj When <0.5*yz, the output local decision is 0; The output local judgment of 1 indicates that the radiation dose rate of the jth monitoring node is abnormal, and the output local judgment of 0 indicates that the radiation dose rate of the jth monitoring node is within the normal range; Combine the local judgment results of each monitoring node with the missed judgment rate and wrong judgment rate of each monitoring node to perform information fusion and generate the fusion center judgment. According to the different judgment results of the fusion center, judge whether there is abnormal monitoring data in the radiation field; The formula for information fusion is: Where ZH is the fusion center decision, u j is the local decision output of each monitoring node, P Mj and P Lj are the missed judgment rate and wrong judgment rate of the jth monitoring node respectively, and τ is the decision threshold; According to the generated fusion center judgment ZH, the logic for judging whether an abnormal situation occurs is: When ZH≥0, it indicates that an abnormal situation has occurred; When ZH<0, it means that there is a misjudgment in the local judgment and there is no abnormality in the radiation field.

2. A radiation field abnormality monitoring data automatic identification system, characterized by: The radiation field anomaly monitoring data automatic identification system is used to execute the radiation field anomaly monitoring data automatic identification method according to claim 1, comprising: A historical data acquisition module is used to obtain abnormal state radiation dose rate data and normal state radiation dose rate data of each monitoring node in the radiation field generated by the medical venue in a number of equal long time periods in history, wherein the abnormal state and the normal state are two states that will affect the human body and will not affect the human body respectively; A feature extraction module, used to extract the time domain feature values ​​of the data according to the collected radiation dose rate data; the time domain feature values ​​include maximum value, minimum value, mean value and standard deviation; A state determination model generation module is used to train the initial neural network based on the obtained data time domain characteristic values ​​as samples and the corresponding state marks as labels to obtain a radiation dose rate state determination model; A real-time data detection module is used to obtain real-time radiation dose rate data of each monitoring node in the radiation field generated by the medical venue; extract real-time time domain characteristic values ​​from the real-time radiation dose rate data, input the real-time time domain characteristic values ​​into the obtained radiation dose rate state determination model, and obtain the probability that the radiation dose rate of each monitoring node belongs to an abnormal state; A local decision module is used to obtain a local decision result of each monitoring node according to the comparison between the probability that the radiation dose rate of each monitoring node belongs to an abnormal state and a preset abnormal state probability threshold; The central decision generation module is used to combine the local decision results of each monitoring node with the missed judgment rate and wrong judgment rate of each monitoring node, perform information fusion, generate a fusion center decision, and judge whether there is abnormal monitoring data in the radiation field based on the different decision results of the fusion center.

Images