Early warning method for radiation detection and related equipment

Abstract

The invention provides an early warning method for radiation detection and related equipment, and is applied to the technical field of data processing. According to the method, target element extraction is carried out on detection service scene information to generate risk influence factors; performing data analysis processing on the historical detection records to generate a standardized detection feature matrix; performing feature dimension screening on the standardized detection feature matrix based on the identity information of the current detection personnel to generate a target risk feature vector; performing safety characteristic analysis on the equipment information used for detection to generate a detection equipment safety characteristic vector; carrying out transfer learning on a preset risk assessment model based on the historical risk data sample to generate an adaptive risk assessment model; and inputting the risk influence factor, the target risk feature vector and the detection equipment security feature vector into an adaptive risk assessment model, carrying out feature fusion through an attention mechanism, and generating radiation equipment security leakage risk early warning result information with confidence.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an early warning method for radiation detection and related equipment. Background Art

[0002] In the field of radiation detection, especially in the safety detection of radiation medical detection equipment, the existing detection and early warning technology has many shortcomings. When assessing the risk of radiation equipment safety leakage, traditional methods often rely on only a single or a few factors for judgment, which makes the assessment of equipment risks not comprehensive and accurate, and it is difficult to detect potential safety hazards in advance. When processing historical detection records, the existing data analysis methods are relatively simple. It is impossible to deeply mine and effectively utilize a large amount of historical data, and it is difficult to extract valuable information from complex data. For example, when analyzing the trend of changes in the radiation value of the equipment, it is impossible to accurately associate factors such as detection frequency, equipment maintenance, and detection operation type with the radiation value, resulting in the inability to accurately grasp the changing laws of the equipment's radiation performance, affecting the judgment of the equipment's safety.

[0003] For the equipment used for testing, the existing security feature analysis is often limited to certain aspects, such as focusing only on the authentication status of the equipment, but lacking in-depth analysis of the equipment's network security status, abnormal operation history, etc. In today's context where medical equipment is becoming increasingly intelligent and networked, equipment network security vulnerabilities may be exploited by criminals, causing equipment to operate abnormally and increase the risk of radiation leakage. However, the existing detection and early warning technology fails to fully cover these potential risk factors and cannot accurately assess the overall security status of the equipment.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0005] The purpose of this application is to provide an early warning method and related equipment for radiation detection, which at least overcomes the problems existing in the prior art to a certain extent, by collecting historical equipment detection records, personnel identity information, business scenario information, equipment information, preset models and historical risk samples. Then, risk influencing factors are extracted from the business scenario, and the historical detection records are analyzed to generate a standardized detection feature matrix. The target risk feature vector is generated by screening according to the identity of the personnel, and the equipment information is analyzed to generate the detection equipment safety feature vector. Subsequently, the preset model is trained based on the historical risk samples to obtain an adaptive risk assessment model. Finally, the above feature vector is input into the adaptive model, fused through the attention mechanism, and the risk warning result with confidence is output. First, each key element is deeply analyzed and processed to generate the corresponding feature vector and model, and then the feature vector is fused through the model, so as to achieve accurate radiation equipment risk warning and ensure the safe use of radiation medical detection equipment.

[0006] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present invention.

[0007] According to one aspect of the present application, a radiation detection early warning method is provided, including: obtaining historical detection records of radiation medical detection equipment, identity information of current detection personnel, detection business scenario information, information of equipment used for detection, a preset risk assessment model and historical risk data samples; extracting target elements from the detection business scenario information to generate risk impact factors, wherein the risk impact factors include detection urgency weights, equipment radiation safety sensitivity parameters and detection personnel qualification authority threshold indicators; performing data analysis and processing on historical detection records to generate a standardized detection feature matrix; performing feature dimension screening on the standardized detection feature matrix based on the identity information of the current detection personnel to generate a target risk feature vector, wherein The target risk feature vector includes the characteristics of the historical detection results of the equipment, the operating habits of the detection personnel, and the abnormal behavior characteristics of the detection process; the security characteristics of the equipment information used in the detection are analyzed to generate a detection equipment security feature vector, where the detection equipment security feature vector includes the detection equipment authentication status coefficient, the detection equipment network security level parameter and the detection equipment abnormal behavior baseline value; based on the historical risk data samples, the preset risk assessment model is transferred and learned to generate an adaptive risk assessment model; the risk influencing factors, target risk feature vectors and detection equipment security feature vectors are input into the adaptive risk assessment model, and the features are fused through the attention mechanism to generate radiation equipment safety leakage risk warning result information with confidence.

[0008] Another aspect of the present application provides an early warning device for radiation detection, including: an acquisition module for acquiring historical detection records of radiation medical detection equipment, identity information of current detection personnel, detection business scenario information, information of equipment used for detection, a preset risk assessment model and historical risk data samples; a processing module for extracting target elements from the detection business scenario information and generating risk impact factors, wherein the risk impact factors include detection urgency weights, equipment radiation safety sensitivity parameters and detection personnel qualification authority threshold indicators; performing data analysis and processing on historical detection records to generate a standardized detection feature matrix; and performing feature dimension screening on the standardized detection feature matrix based on the identity information of the current detection personnel to generate target risk features. The target risk feature vector includes the characteristics of the historical detection results of the equipment, the operating habits of the detection personnel and the abnormal behavior characteristics of the detection process; the security characteristics of the equipment information used in the detection are analyzed to generate a detection equipment security feature vector, wherein the detection equipment security feature vector includes the detection equipment authentication status coefficient, the detection equipment network security level parameter and the detection equipment abnormal behavior baseline value; based on the historical risk data samples, the preset risk assessment model is transferred and learned to generate an adaptive risk assessment model; the risk influencing factors, the target risk feature vector and the detection equipment security feature vector are input into the adaptive risk assessment model, and the features are fused through the attention mechanism to generate the radiation equipment safety leakage risk warning result information with confidence.

[0009] According to another aspect of the present application, an electronic device is provided, comprising: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to execute the above-mentioned radiation detection early warning method by executing the executable instructions.

[0010] According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a second processor, the above-mentioned early warning method for radiation detection is implemented.

[0011] The present application provides a radiation detection early warning method and related equipment, in which the server collects equipment historical detection records, personnel identity information, business scenario information, equipment information, preset models and historical risk samples. Then, the risk influencing factors are extracted from the business scenario, the historical detection records are analyzed to generate a standardized detection feature matrix, the target risk feature vector is generated according to the personnel identity screening, and the equipment information is analyzed to generate the detection equipment safety feature vector. Subsequently, the preset model is trained based on the historical risk samples to obtain an adaptive risk assessment model. Finally, the above feature vectors are input into the adaptive model, fused through the attention mechanism, and the risk warning results with confidence are output. First, each key element is deeply analyzed and processed to generate the corresponding feature vectors and models, and then the feature vectors are fused through the model to achieve accurate radiation equipment risk warnings and ensure the safe use of radiation medical detection equipment.

[0012] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A flowchart showing a radiation detection early warning method provided by an embodiment of the present application; Figure 2 A schematic structural diagram of a radiation detection warning device provided in one embodiment of the present application is shown. DETAILED DESCRIPTION

[0014] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0015] Combine the following Figure 1 To describe the early warning method of radiation detection according to the exemplary embodiment of the present application. It should be noted that the following application scenarios are only shown to facilitate understanding of the spirit and principle of the present application, and the embodiments of the present application are not limited in this regard. On the contrary, the embodiments of the present application are applied to any applicable scenario.

[0016] In one implementation, the present application also proposes a radiation detection early warning method and related equipment. Figure 1 The following schematically shows a flow chart of a radiation detection early warning method according to an embodiment of the present application. Figure 1 As shown, the method is applied to a server, comprising: S101, obtaining historical detection records of radiation medical detection equipment, identity information of current detection personnel, detection business scenario information, detection equipment information, preset risk assessment model and historical risk data samples.

[0017] In one implementation, the hospital's radiation detection management system stores the detection records of all radiation medical detection equipment in the hospital over the years. Taking a CT machine with a model of "RX-2000" as an example, the historical detection records record in detail the time of each detection, such as the first detection on May 10, 2018 after it was first put into use, to the time point of the subsequent annual regular detection; the work number of the detection personnel, such as the "00123" detection personnel who was responsible for the detection work in 2019; the radiation value data at each detection, such as the radiation value of a specific part in the 2020 detection was 2.5μSv / h; and the equipment maintenance situation. In 2021, due to abnormal fluctuations in radiation values ​​due to aging of the equipment, key components were replaced and maintained, and these maintenance records were also recorded in detail. By obtaining these historical detection records, we can clearly understand the radiation performance change trend, maintenance history and detection results of the equipment in different periods, and provide basic data for subsequent analysis.

[0018] When conducting a new inspection of the "RX-2000" CT machine, the inspector logs in through the hospital's internal identity recognition system. The system automatically obtains his identity information, including his name "Li Si", work number "00567", department "Equipment Inspection Department", and inspector qualification level "Intermediate Inspector". Among them, the inspector's qualification level is assessed based on his professional skills training, assessment results and actual work experience. Intermediate inspectors have the ability and qualifications to conduct comprehensive inspections of this model of CT machine. Obtaining this identity information will help to conduct subsequent risk assessments based on the professional capabilities and operating habits of the inspectors. For example, inspectors with different qualifications may have differences in the standardization of operating procedures and the accuracy of data reading, which in turn affects the test results and equipment risk judgments.

[0019] The detection business scenario information covers many aspects. In terms of time, the "RX-2000" CT machine inspection is scheduled for June 2024, which is an annual regular inspection. In terms of the detection environment, the temperature in the hospital's imaging room was 22°C and the humidity was 50% during the inspection, and the environment was relatively stable. In terms of the purpose of the inspection, it is to evaluate the radiation safety of the equipment after use in the past year, to ensure that the equipment meets the relevant radiation safety standards, and to protect the health and safety of patients and medical staff. In addition, before the inspection, the hospital had just introduced a batch of new clinical testing projects, and the frequency of use of the CT machine is expected to increase significantly, which also increases the importance of this inspection. These detection business scenario information is of great significance for determining the urgency of the inspection, the sensitivity of the equipment to radiation safety, and whether the qualifications and permissions of the inspection personnel match, and is the key basis for generating risk influencing factors.

[0020] For the "RX-2000" CT machine being tested, the testers obtained the equipment information through the equipment's own information collection module and the hospital's equipment management system. The basic parameters of the equipment include model "RX-2000", production date "January 2018", and use time of about 6 years. In terms of the equipment's certification status, the CT machine has passed the relevant national radiation safety certification before it was put into use, and the hospital calibrates and reviews the equipment every year in accordance with regulations. The current certification status is valid, and the certification certificate number is "[specific number]". In terms of network security, the equipment is connected to the hospital's internal network, and the hospital's network security department regularly conducts network security assessments on it. The latest assessment report shows that there are some security vulnerabilities in the network connected to the equipment, but temporary protective measures have been taken, and the network security level has been rated as "medium". At the same time, the historical operation log information of the equipment was obtained, and it was found that the equipment had 3 abnormal operation records in the past six months. All of them were when the radiation value temporarily exceeded the normal range when the equipment was started, but soon returned to normal. The frequency and type of these abnormal operations will be used to generate the baseline value of abnormal behavior of the equipment in the future.

[0021] The hospital uses a deep learning model based on a combination of Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) to conduct radiation equipment risk assessment. The overall model is divided into a front-end convolutional layer, an intermediate LSTM layer, and a back-end fully connected layer. The front-end convolutional layer consists of three convolutional blocks, each of which contains a 2D convolutional layer, a batch normalization layer, and a ReLU activation function. The convolutional layer is responsible for extracting local features in the equipment detection data, such as the changing trend of radiation values, the characteristics of equipment operating parameters, etc. The middle LSTM layer contains two layers of LSTM units, which are used to process detection data with time series characteristics and capture the long-term dependencies of the data, such as the changing patterns of the equipment's historical radiation values ​​over time. The back-end fully connected layer consists of two fully connected layers, which integrate and map the features output by the LSTM layer and finally output the risk assessment results. In the convolution layer, the 2D convolution kernel size is 3x3, the step size is 1, and the padding is 1; in the LSTM layer, the number of hidden units is 128; in the fully connected layer, the number of neurons in the first fully connected layer is 64, and the number of neurons in the second fully connected layer is determined according to the risk assessment category. For example, when the risk is divided into three categories: low, medium, and high risk, the number of neurons is 3. When training the model, the Adam optimizer is used, the learning rate is set to 0.001, and the loss function uses the cross entropy loss function.

[0022] Over the years, the hospital has accumulated a wealth of radiation equipment risk cases as historical risk data samples. For example, there have been cases where the same type of equipment has not been maintained for a long time, resulting in radiation leakage, causing patients to be exposed to excessive radiation; there have also been cases where the equipment has been operated improperly by the inspectors, resulting in deviations in the test data that affect risk judgment. These cases record in detail information such as equipment model, inspector information, business scenarios, risk occurrence time, risk type and consequences, providing reference standards and training data for this "RX-2000" CT machine risk assessment, which helps to improve the accuracy and reliability of risk assessment.

[0023] S102, extracting target elements from the detection business scenario information and generating risk impact factors.

[0024] In one implementation, target elements are extracted from the detection business scenario information to generate detection business scenario information features described in natural language, wherein the detection business scenario information features described in natural language are used to characterize the basic information of the detection business scenario. On a certain day in June 2024, Li Si, an intermediate inspector of the hospital equipment inspection department, used professional inspection equipment to conduct annual regular inspections on the CT machine of model "RX-2000" in the computer room of the hospital imaging department. The room temperature was 22°C and the humidity was 50% during the inspection. In addition, the hospital recently introduced new clinical inspection projects, and the frequency of use of the CT machine is expected to increase significantly. Thus, the detection business scenario information features described in natural language are generated: "In June 2024, Li Si, an intermediate inspector, conducted annual regular inspections on the 'RX-2000' CT machine in the computer room of the hospital imaging department (temperature 22°C, humidity 50%). The hospital has introduced new clinical inspection projects, and the frequency of use of the CT machine is expected to increase significantly." This description covers key information of the inspection, such as time, personnel, equipment, environment, and business background, and fully presents the basic situation of the detection business scenario.

[0025] Construct a radiation detection business knowledge ontology library, and map the information features of the detection business scenarios described in natural language based on the radiation detection business knowledge ontology library to generate structured features. The radiation detection business knowledge ontology library constructed by the hospital contains rich information. In terms of equipment, the technical parameters, radiation safety standards, maintenance cycles, etc. of different models of CT machines are clarified; in terms of personnel, the qualification requirements and scope of responsibilities of detection personnel at all levels are stipulated; in terms of business processes, the full process specifications from detection plan formulation, equipment preparation, detection implementation to result evaluation and report generation are covered; in terms of environmental factors, the standard range of temperature, humidity, electromagnetic environment, etc. suitable for detection is defined; and relevant laws and policies are also included, such as "Radiation Protection Requirements for Medical X-ray Diagnosis". By integrating this information, a comprehensive knowledge system is formed to provide a basis for subsequent analysis. Based on the radiation detection business knowledge ontology library, the above natural language description features are mapped. The business type is determined to be "Annual regular inspection of CT machines", and the business processes involved include "Inspection plan execution process" and "Equipment inspection operation process"; the data objects involved are "'RX-2000' CT machine inspection data" and "machine room environment data"; the data belongs to the field of "medical equipment radiation detection"; the business operation subject is "Intermediate inspector (Li Si)"; the structured information related to the inspection environment is "temperature 22℃ (suitable inspection temperature range 18-25℃)" and "humidity 50% (suitable inspection humidity range 30%-70%)"; the equipment related information is "'RX-2000' CT machine (has been in use for 6 years, radiation safety standard [specific standard value])", etc. These structured features convert natural language descriptions into data forms that are easy to analyze and process.

[0026] The structured features are processed to generate business adaptation index, equipment radiation safety importance parameters and inspection personnel qualification compliance indicators. According to the business process specifications in the ontology library, Li Si operates according to the standard inspection process, and all steps from equipment preheating, parameter calibration to radiation value measurement meet the requirements. At the same time, the temperature and humidity of the inspection environment are within the appropriate range, and this inspection is consistent with the development of new business in the hospital. Through the set evaluation algorithm, the business adaptation index is calculated to be 0.9 (out of 1 point), indicating that the inspection business operation is highly adapted to the standard process and current business needs.

[0027] According to the definition of equipment radiation safety in the ontology library, the "RX-2000" CT machine is used for clinical diagnosis, and its radiation safety is directly related to the health of patients and medical staff. In addition, the equipment has been in use for 6 years. As the service life increases, the radiation safety risk may increase. After comprehensive evaluation, the importance parameter of equipment radiation safety was determined to be 0.85 (out of 1 point), indicating that equipment radiation safety is of vital importance in this inspection. Referring to the personnel qualification requirements in the ontology library, intermediate inspectors are qualified to conduct comprehensive inspections of the "RX-2000" CT machine. Li Si, as an intermediate inspector, performed this inspection task and fully met the qualification requirements. Upon inspection, his operations were all within the scope allowed by the qualifications, and the calculated compliance index of the inspection personnel qualifications was 1 (out of 1 point), indicating that the qualification compliance was good.

[0028] The business adaptation index, equipment radiation safety importance parameter and inspection personnel qualification compliance index are converted and calculated to generate the detection urgency weight, equipment radiation safety sensitivity parameter and inspection personnel qualification authority threshold index, which together constitute the risk influencing factor. Calculated by the pre-set conversion formula, the formula is assumed to be: detection urgency weight = business adaptation index × 0.3 + equipment radiation safety importance parameter × 0.7. Substituting the previously calculated value, the detection urgency weight = 0.9 × 0.3 + 0.85 × 0.7 = 0.865. This shows that although the CT machine inspection business process is standardized, due to the high importance of equipment radiation safety, the overall detection urgency is at a high level.

[0029] The conversion is performed based on the equipment radiation safety importance parameter and the established sensitivity standard. The equipment radiation safety importance parameter is set between 0.8-1, the equipment radiation safety sensitivity is high, and the corresponding equipment radiation safety sensitivity parameter is 0.9. Therefore, the equipment radiation safety sensitivity parameter of the "RX-2000" CT machine is 0.9, which means that the equipment radiation safety sensitivity is extremely high. Adjustment is made in combination with the inspection personnel qualification compliance index. Since the qualification compliance index is 1, it indicates that the inspection personnel are fully compliant. When the qualification compliance level is 1, the inspection personnel qualification authority threshold index is set to 1; if there is a qualification violation, the index value is reduced accordingly according to the degree of violation. Therefore, the inspection personnel qualification authority threshold index is 1 this time. Finally, the inspection urgency weight 0.865, the equipment radiation safety sensitivity parameter 0.9 and the inspection personnel qualification authority threshold index 1 together constitute the risk impact factor of this CT machine inspection business scenario, providing a key basis for subsequent risk assessment.

[0030] S103, performing data analysis and processing on historical detection records to generate a standardized detection feature matrix.

[0031] In one implementation, the historical detection records are time sliced ​​to generate a number of detection period records, wherein the detection period records are used to characterize the record information of the radiation equipment detection in each time period divided in the time dimension. The hospital's radiation detection management system stores the historical detection records of the "RX-2000" CT machine since it was put into use in 2018. When performing time slicing, it is divided into months. For example, in 2023, the hospital conducted relatively frequent inspections and maintenance on the CT machine from January to March, and this time period can be divided into multiple detection period records.

[0032] From January 10 to January 15, 2023, one routine test and one additional test due to a minor abnormality in the equipment were conducted during this period. From 9:00 to 10:00 a.m. on January 10, the inspectors conducted routine tests and recorded relevant data such as radiation values; from 3:00 to 4:00 p.m. on January 12, due to a brief fluctuation in the radiation value of the equipment during use, the inspectors conducted another test. This constitutes a test period record, which records in detail the time interval, number of tests and reasons for the tests during this period. From February 5 to February 8, 2023, only one monthly regular test was conducted, which was completed from 10:00 to 11:00 a.m. on February 6. This test period record clearly presents the specific time of the only test during this period. Through this time slicing method, the historical test records of the CT machine are divided into multiple test period records in chronological order, fully displaying the test details in each period.

[0033] Target data extraction and normalization processing are performed on several detection period records to generate detection feature sequences, where the detection feature sequence includes information such as detection frequency, detection radiation value, equipment maintenance status and detection operation type, which is used to characterize the behavioral characteristics of radiation equipment detection. For each detection period record, key information is extracted. Take the detection period record from January 10 to January 15, 2023 as an example: Detection frequency: During these 6 days, a total of 3 tests were conducted, so the detection frequency of this period is 3 times / 6 days=0.5 times / day. Detection radiation value: During the routine detection on January 10, the radiation value of a specific part was 2.3μSv / h; during the additional detection on January 12, the radiation value was 2.5μSv / h (value after fluctuation). Equipment maintenance status: On January 13, the inspector found that a sensor of the equipment had an accuracy deviation and performed calibration maintenance on it. Detection operation type: including routine detection operations, detection operations for abnormal conditions, and sensor calibration and maintenance operations.

[0034] In order to make the test data of different periods comparable, normalization is required. Assuming that the CT machine performs a maximum of 10 tests per month during the busy period (using months as the unit to facilitate comparison of frequencies in different periods), the normalized value of the test frequency from January 10 to January 15 is 0.5÷(10÷30)=1.5 (here the maximum number of tests per month is converted to the average maximum number of tests per day for comparison). For the radiation value, if the maximum radiation value of the CT machine during normal operation is 5μSv / h, the normalized value of the radiation value detected on January 10 is 2.3÷5=0.46, and the normalized value of the radiation value detected on January 12 is 2.5÷5=0.5. By integrating these normalized detection frequencies, detection radiation values, equipment maintenance status (the maintenance status can be encoded, such as no maintenance as 0, calibration maintenance as 1, etc.) and detection operation types (different operation types are encoded, such as routine detection as 1, abnormal detection as 2, calibration maintenance as 3, etc.), we can generate a detection feature sequence, such as [(1.5,0.46,1,1),(1.5,0.5,1,2),(1.5,0.5,1,3)], which is used to characterize the behavioral characteristics of the "RX-2000" CT machine detection during this period.

[0035] The detection feature sequence is dimensionally aligned and standardized to generate a standardized detection feature matrix. Since the detection feature sequence contains different types of information (detection frequency, detection radiation value, equipment maintenance status code, detection operation type code), and the operation type code may lead to inconsistent dimensions (assuming that there are multiple operation types, different numbers of codes will result in different feature vector dimensions), dimension alignment is required. The dimension can be determined based on the maximum number of possible operation types. Assuming that the maximum number of operation types is determined to be 5, the insufficient dimensions are filled with 0. In this way, the above detection feature sequence becomes a form with consistent dimensions, such as [(1.5, 0.46, 1, 1, 0, 0), (1.5, 0.5, 1, 2, 0, 0), (1.5, 0.5, 1, 3, 0, 0)].

[0036] The dimensionally aligned feature vectors are normalized and scaled to eliminate the impact of different dimensions between features. For example, for the two numerical features of detection frequency and detection radiation value, the Z-score normalization method is used. Assume that the mean of detection frequency is , the standard deviation is ; The average value of the detected radiation is , the standard deviation is For the detection frequency of 1.5 in the first eigenvector, the normalized value is ; The normalized value of the detected radiation value is 0.46 After all eigenvalues ​​in each eigenvector are standardized in this way, a standardized eigenvector is obtained. These standardized eigenvectors are arranged in rows to generate a standardized detection feature matrix. Each row represents a eigenvector of a detection period, and each column represents a different feature dimension, so that the detection features have a unified format in time and feature dimensions, which is convenient for subsequent operations such as feature screening and risk assessment based on the identity information of the current detection personnel.

[0037] S104, screening the characteristic dimensions of the standardized detection feature matrix based on the identity information of the current detection personnel to generate a target risk feature vector.

[0038] In one implementation, the standardized detection feature matrix is ​​screened for feature dimensions based on the identity information of the current detection personnel to generate original features of the equipment's historical detection results, original features of the detection personnel's operating habits, and original features of abnormal behaviors during the detection process. The hospital uses the "RX-2000" CT machine for detection, and the identity information of the current detection personnel, Li Si, shows that he is an intermediate detection technician with a work number of "00567". The generated standardized detection feature matrix contains detection feature data of the CT machine by different detection personnel and at different times over the years. Based on Li Si's identity information, the feature vectors corresponding to the detection records related to him are screened out.

[0039] Filter out the radiation value, equipment operation status and other relevant data of the "RX-2000" CT machine that Li Si participated in testing. For example, in the 5 tests that Li Si participated in, the radiation value of each test and whether the equipment was operating normally were recorded. Extract data such as the operation time and operation steps of each test by Li Si. For example, information such as the start-up time of Li Si in each test, the operation sequence of each test link, and the operation interval time. Find out the operation records that Li Si appeared in the test process that did not conform to the standard operating procedures. Assume that in a test, Li Si did not calibrate the equipment in the prescribed order, and this abnormal operation was recorded as the original feature of abnormal behavior in the test process.

[0040] The original features of the historical detection results of the equipment are processed to generate the average radiation detection value and the fluctuation range of the radiation value. The average radiation detection value and the fluctuation range of the radiation value are used to characterize the historical level and stability of the radiation situation of the equipment. The radiation value of the "RX-2000" CT machine in which Li Si participated in the detection is calculated. Assuming that the radiation values ​​in the 5 tests in which Li Si participated were 2.3μSv / h, 2.5μSv / h, 2.4μSv / h, 2.6μSv / h, and 2.2μSv / h, respectively, then the average radiation detection value = (2.3+2.5+2.4+2.6+2.2) ÷ 5 = 2.4μSv / h, which is used to characterize the historical level of the radiation situation of the equipment. Calculate the difference between the maximum and minimum radiation values. In the above five tests, the maximum radiation value was 2.6μSv / h, the minimum was 2.2μSv / h, and the radiation value fluctuation range was = 2.6-2.2 = 0.4μSv / h. This range reflects the stability of the equipment's radiation value.

[0041] The original features of the operating habits of the inspectors are processed to generate the average inspection time and the compliance rate of the inspection steps. The average inspection time and the compliance rate of the inspection steps are used to characterize the time regularity and compliance degree of the inspection personnel's operations. The time spent by Li Si on each inspection is counted. Assuming that the duration of Li Si's 5 inspections is 60 minutes, 55 minutes, 62 minutes, 58 minutes, and 65 minutes respectively, the average inspection time = (60+55+62+58+65) ÷ 5 = 60 minutes, which reflects the time regularity of Li Si's inspection operation. Compare Li Si's inspection steps with the standard inspection steps. The standard inspection steps have a total of 10 steps. In the 5 inspections of Li Si, there was an error in the order of one step in the first inspection, all were correct in the second inspection, two steps in the third inspection did not fully meet the standard, all were correct in the fourth inspection, and one step was omitted in the fifth inspection. Then Li Si's inspection step compliance rate = (9+10+8+10+9) ÷ (10×5) × 100% = 92%, which is used to characterize the compliance degree of the inspection personnel's operations.

[0042] The original features of abnormal behavior in the detection process are processed to generate the frequency of abnormal detection situations. The frequency of abnormal detection situations is used to characterize the frequency of deviations from normal detection operations during the detection process. The number of abnormal operations performed by Li Si during the detection process is counted. In the 50 tests in which Li Si participated, there were 3 abnormal operations (such as the aforementioned failure to calibrate the equipment in the prescribed order). The frequency of abnormal detection situations = 3 ÷ 50 = 6%. This frequency is used to characterize the frequency of deviations from normal detection operations during the detection process.

[0043] The average radiation detection value, radiation value fluctuation range, average detection time, detection step compliance rate, and abnormal detection frequency are processed to generate a target risk feature vector. Set weights for the average radiation detection value, radiation value fluctuation range, average detection time, detection step compliance rate, and abnormal detection frequency. Assume that the weight of the average radiation detection value is 0.2, the weight of the radiation value fluctuation range is 0.2, the weight of the average detection time is 0.1, the weight of the detection step compliance rate is 0.3, and the weight of the abnormal detection frequency is 0.2. Standardize these indicators (for example, assume that the standardized value of the average radiation detection value is 0.8, the standardized value of the radiation value fluctuation range is 0.6, the standardized value of the average detection time is 0.7, the standardized value of the detection step compliance rate is 0.9, and the standardized value of the abnormal detection frequency is 0.3).

[0044] Calculate the target risk feature vector: (0.8×0.2)+(0.6×0.2)+(0.7×0.1)+(0.9×0.3)+(0.3×0.2)=0.16+0.12+0.07+0.27+0.06=0.68. The calculation result 0.68 is a key value in the target risk feature vector. Combined with other similar calculation results (such as considering more test data or other related indicators), a complete target risk feature vector can be generated, providing an important basis for evaluating the risk of the "RX-2000" CT machine in the detection process of Li Si.

[0045] S105, performing security characteristic analysis on the device information used for detection, and generating a detection device security feature vector.

[0046] In one implementation, the authentication status analysis and processing of the equipment information used for testing is performed to generate the authentication status coefficient of the testing equipment. When testing the "RX-2000" CT machine, the hospital adopted a strict equipment authentication system. The system includes the initial registration authentication of the equipment, regular calibration authentication, and real-time monitoring authentication during operation. When the tester is ready to test the "RX-2000" CT machine, the system will automatically conduct a comprehensive check on the authentication status of the equipment.

[0047] First, confirm whether the initial registration information of the equipment is complete and accurate, including whether the basic information such as equipment model, manufacturer, and production date is consistent with the filing information. After verification, the initial registration information of the "RX-2000" CT machine is completely correct. Next, check the regular calibration of the equipment. According to regulations, the "RX-2000" CT machine needs to be calibrated once every six months. The last calibration time was March 2024, and the calibration results showed that all parameters of the equipment were within the standard range. Finally, check the real-time monitoring data during the operation of the equipment to ensure that there are no abnormal conditions that may affect the certification status of the equipment during daily use. No abnormalities were found through real-time analysis of the equipment operation log. Based on the above inspection results, according to the certification status scoring rules set by the hospital, the full score is set to 1 point. If all certification links of the equipment are normal, full marks will be given. Therefore, the certification status coefficient of the detection equipment of the "RX-2000" CT machine is rated as 1.

[0048] Obtain the network security assessment report of the testing equipment and the historical operation log information of the equipment. The hospital's network security department regularly conducts security assessments on all devices connected to the hospital's internal network, and the "RX-2000" CT machine is no exception. The testers obtained the latest network security assessment report of the device from the network security department. The report shows that in the most recent comprehensive assessment, it was found that the network connected to the "RX-2000" CT machine had some potential risks. For example, some network ports have the risk of weak passwords. Although they have not yet suffered substantial attacks, there are certain security risks. At the same time, some abnormal network traffic was detected in the network, but after further analysis, it was initially determined that these traffic were not malicious attacks, and may be interference caused by network failures of other equipment within the hospital.

[0049] In addition, the inspectors also obtained the historical operation log information of the "RX-2000" CT machine from the local storage of the equipment and the hospital's equipment management system. The log records in detail the various operating behaviors of the equipment, including the time of each startup and shutdown, the use of various detection functions, and the adjustment records of equipment parameters. When reviewing the historical operation log, the inspectors found that in the past month, the equipment had performed a total of 20 detection operations, of which 3 operations were abnormal. These 3 abnormal operations all occurred during the equipment startup phase, manifested as radiation values ​​temporarily exceeding the normal range, but quickly returned to normal.

[0050] The network security assessment report is processed to generate the network security level parameters of the detection equipment, where the network security level parameters are used to characterize the security level of the network environment in which the detection equipment is located. The hospital processes the network security assessment report of the "RX-2000" CT machine according to a set of established network security level assessment standards to determine its network security level parameters. The assessment standard comprehensively considers multiple factors, including the severity of network vulnerabilities, the possibility of intrusion risks, the impact of abnormal network traffic, etc., and divides the network security level into three levels: high, medium, and low, corresponding to different numerical ranges, with high level corresponding to 0.8-1, medium level corresponding to 0.4-0.7, and low level corresponding to 0-0.3. For the "RX-2000" CT machine, although there is a risk of weak passwords on some network ports, no security incidents have occurred, and the impact of abnormal network traffic is relatively small. After a comprehensive evaluation by network security experts, it is believed that the network security status of the equipment is at a medium level. Therefore, the network security level parameters of the detection equipment of the "RX-2000" CT machine are determined to be 0.6.

[0051] The historical operation log information of the equipment is processed, the frequency and type of abnormal operations are counted, and the baseline value of abnormal behavior of the detection equipment is generated. The baseline value of abnormal behavior of the detection equipment includes the average frequency of abnormal operations of the detection equipment and the proportion of abnormal operation types. The detection personnel conducted an in-depth analysis of the historical operation log information of the "RX-2000" CT machine, counted the frequency and type of abnormal operations, and generated the baseline value of abnormal behavior of the detection equipment. In the 20 detection operations in the past month, 3 abnormal operations occurred, so the average frequency of abnormal operations is 3÷20=0.15 times / time (here it means the average probability of abnormal operations in each detection operation). From the perspective of abnormal operation types, these 3 abnormal operations are all when the radiation value temporarily exceeds the normal range when the equipment is started, which belongs to the abnormal type of equipment startup. Therefore, among all operations, the proportion of abnormal operation types is 3÷20×100%=15%. These data (the average frequency of abnormal operations is 0.15 times / time and the proportion of abnormal operation types is 15%) together constitute the baseline value of abnormal behavior of the detection equipment of the "RX-2000" CT machine.

[0052] The detection equipment authentication status coefficient, detection equipment network security level parameter and detection equipment abnormal behavior baseline value are processed to generate the detection equipment security feature vector. In order to comprehensively evaluate the safety status of the "RX-2000" CT machine, it is necessary to comprehensively process the detection equipment authentication status coefficient, detection equipment network security level parameter and detection equipment abnormal behavior baseline value to generate the detection equipment security feature vector. The hospital sets corresponding weights for each indicator based on the degree of impact of these indicators on equipment safety. Assume that the weight of the detection equipment authentication status coefficient is 0.4, the weight of the detection equipment network security level parameter is 0.4, and the weight of the detection equipment abnormal behavior baseline value is 0.2.

[0053] Since the abnormal behavior baseline value of the detection equipment includes two sub-indicators: the average frequency of abnormal operations and the proportion of abnormal operation types, the average of these two sub-indicators is first taken, that is, (0.15+0.15)÷2=0.15 (the average value here is to combine the two sub-indicators into a comprehensive indicator for subsequent calculations). Then calculate: (detection equipment certification status coefficient×0.4)+(detection equipment network security level parameter×0.4)+(detection equipment abnormal behavior baseline value×0.2)=(1×0.4)+(0.6×0.4)+(0.15×0.2)=0.4+0.24+0.03=0.67. This 0.67, as a key value in the detection equipment security feature vector, comprehensively reflects the security status of the "RX-2000" CT machine.

[0054] S106, performing transfer learning on the preset risk assessment model based on historical risk data samples to generate an adaptive risk assessment model.

[0055] In one implementation, the number of risk features in the historical risk data samples is counted, and a sampling ratio is generated based on these numbers. The hospital's radiation detection management system stores a large number of historical risk data samples of radiation equipment, covering various types of equipment, operation records of different detection personnel, and detection data in various business scenarios. These samples contain a wealth of risk features, such as radiation values ​​exceeding the standard, untimely equipment maintenance, detection personnel operating errors, and risks caused by network security vulnerabilities. According to statistics, in 1,000 historical risk data samples, risk features related to radiation values ​​exceeding the standard appeared 300 times, risk features due to untimely equipment maintenance appeared 200 times, risk features due to detection personnel operating errors appeared 250 times, risk features due to network security vulnerabilities appeared 150 times, and other risk features (such as environmental factors) appeared 100 times. Assuming that the preset number of samples is 200, the sampling ratio is generated according to the proportion of the number of each risk feature to the total number of samples. The sampling ratio of the risk characteristic of excessive radiation value is 300÷1000=0.3; the sampling ratio of the risk characteristic of untimely equipment maintenance is 200÷1000=0.2; the sampling ratio of the risk characteristic of operating errors by inspection personnel is 250÷1000=0.25; the sampling ratio of the risk characteristic of network security vulnerabilities is 150÷1000=0.15; and the sampling ratio of other risk characteristics is 100÷1000=0.1.

[0056] Based on the sampling ratio, the historical risk data samples are sampled and processed to generate a preset number of sampling risk features. Based on any risk feature and each sampling risk feature, the historical risk data samples are divided into a high radiation risk group and a low radiation risk group, wherein each group contains a preset number of data samples, and at least one data sample carries radiation risk level identification information. According to the above sampling ratio, sampling is performed from the historical risk data samples. The number of samples from the samples related to excessive radiation values ​​is 200×0.3=60; the number of samples from the samples of untimely equipment maintenance is 200×0.2=40; the number of samples from the samples of operating errors of the inspection personnel is 200×0.25=50; the number of samples from the samples of network security vulnerabilities is 200×0.15=30; the number of samples from the samples of other risk features is 200×0.1=20.

[0057] In this way, 200 sampling risk features were finally obtained. These sampling risk features cover different types of risk situations, and the proportion of each type of risk features is similar to the proportion of risk features in the original sample, ensuring the representativeness of the sampling. The risk feature of excessive radiation value was selected as the basis for comparison, and the radiation value exceeding the normal standard by 1.5 times was set as the high radiation risk standard. Among the 200 sampling risk features, the radiation value of each sample was checked one by one.

[0058] It was found that the radiation values ​​of 80 samples exceeded 1.5 times the normal standard, and these 80 samples were divided into the high radiation risk group. For example, in some samples, the "RX-2000" CT machine was not maintained in time during the detection process due to aging of the equipment, resulting in a radiation value of 2 times the normal standard. These samples were classified into the high radiation risk group. The radiation values ​​of the other 120 samples did not exceed 1.5 times the normal standard, and they were divided into the low radiation risk group. For example, in some samples, although there were some minor problems with the equipment, the radiation values ​​were still within the normal range. These samples belonged to the low radiation risk group. Make sure that each group contains a preset number of data samples (each group here contains at least 1 sample), and mark the radiation risk level identification information for each sample. The samples in the high radiation risk group are marked as "high risk" and the samples in the low radiation risk group are marked as "low risk".

[0059] Based on the high radiation risk group and the low radiation risk group, the preset risk assessment model is trained through transfer learning to generate the trained risk assessment model and training results. The deep learning model based on the combination of convolutional neural network (CNN) and long short-term memory network (LSTM) pre-established by the hospital is used as the preset risk assessment model. The data samples of the high radiation risk group and the low radiation risk group are input into the model for transfer learning training. During the training process, the model will learn the characteristic differences between the samples of the high radiation risk group and the low radiation risk group. In the high radiation risk group, the model will focus on the characteristic combination patterns such as the radiation value change trend, equipment status, and detection personnel operation of the samples with excessive radiation values; in the low radiation risk group, the model will learn the relevant characteristic patterns of the samples with normal radiation values. For example, in the high radiation risk group, there may be a situation where the equipment maintenance is not timely and the radiation value exceeds the standard at the same time, and the model will learn this association pattern; while in the low radiation risk group, the equipment operation is relatively stable and the radiation value is normal, and the model will also capture these features. By continuously adjusting the parameters of the model (such as the convolution kernel weight of the convolution layer, the hidden unit parameters of the LSTM layer, etc.), the model can better identify the data characteristics of different risk groups. After the training is completed, the trained risk assessment model and training results are obtained. The training results include the model's risk assessment output for each sample and the comparison with the actual risk label.

[0060] If the data samples containing identification information in the training results are factors that characterize high radiation detection risks, the trained risk assessment model is used as an adaptive risk assessment model. In the training results, check the model's assessment of each sample. If the model accurately identifies data samples in the high radiation risk group that have characteristics such as high radiation values, equipment maintenance problems, and operator errors as high risk, and these characteristics match the actual high radiation detection risk factors, the trained risk assessment model is considered to be able to effectively identify high-risk situations.

[0061] For example, in the training results, some samples of the "RX-2000" CT machine were accurately assessed as high risk by the model because the radiation value exceeded the standard and the equipment was not maintained in time, which was in line with the actual situation. At this time, the trained risk assessment model is used as an adaptive risk assessment model for subsequent risk assessment of new "RX-2000" CT machine detection data. If the training results are not ideal and high-risk factors are not accurately identified, such as the model misjudging some high-radiation risk samples as low-risk, then it is necessary to adjust the sampling strategy (such as re-determining the sampling ratio, increasing the number of samples, etc.), optimize the model parameters (such as adjusting the learning rate, changing the network structure, etc.) or retrain until the model can accurately identify high-radiation detection risk factors.

[0062] S107, input the risk influencing factor, the target risk feature vector and the detection equipment safety feature vector into the adaptive risk assessment model, perform feature fusion through the attention mechanism, and generate radiation equipment safety leakage risk warning result information with confidence.

[0063] In one implementation, the risk influencing factors, target risk feature vectors, and detection equipment safety feature vectors are processed based on an adaptive risk assessment model to generate a risk feature template, an attention weight for each feature dimension, and an attention weight related to key safety indicators of radiation equipment. The hospital uses a trained adaptive risk assessment model (a deep learning model based on a combination of a convolutional neural network (CNN) and a long short-term memory (LSTM) network) to process the risk influencing factors, target risk feature vectors, and detection equipment safety feature vectors obtained from the "RX-2000" CT machine test.

[0064] Assume that among the risk influencing factors obtained previously, the detection urgency weight is 0.865, the equipment radiation safety sensitivity parameter is 0.9, and the detection personnel qualification authority threshold index is 1; the target risk feature vector key value is 0.68; the detection equipment safety feature vector key value is 0.67. The model analyzes these input vectors and generates a risk feature template. This template is an abstract representation of the risk characteristics of the "RX-2000" CT machine detection this time. It integrates multiple factors such as equipment radiation value, detection personnel operation, and equipment safety status, presenting a combination of various factors that may lead to the risk of radiation equipment safety leakage. At the same time, the model assigns attention weights to each feature dimension. For the detection urgency weight, equipment radiation safety sensitivity parameter, and detection personnel qualification authority threshold index in the risk influencing factors, the model assigns different weights according to their importance in risk assessment. For example, given that equipment radiation safety is directly related to the health of patients and medical staff, the model assigns an attention weight of 0.4 to the equipment radiation safety sensitivity parameter; the detection urgency weight also has an important impact on risk judgment, and is assigned an attention weight of 0.3; the detection personnel qualification and authority threshold indicator has a relatively small impact in this detection, and is assigned an attention weight of 0.3.

[0065] The model also assigns corresponding attention weights to the components of the target risk feature vector, such as the average radiation detection value, radiation value fluctuation range, average detection time, detection step compliance rate, and abnormal detection frequency, as well as the detection equipment certification status coefficient, detection equipment network security level parameter, and detection equipment abnormal behavior baseline value in the detection equipment safety feature vector. For example, the average radiation detection value and the radiation value fluctuation range are crucial to the evaluation of equipment radiation safety, and they are assigned an attention weight of 0.25 respectively; the detection step compliance rate reflects the standardization of the detection personnel's operation, and is assigned an attention weight of 0.2; the detection equipment network security level parameter is more critical in terms of equipment safety, and is assigned an attention weight of 0.3, etc. For key indicators of radiation equipment safety, such as radiation value exceeding the standard and the stability of key components of the equipment, the model generates special attention weights. Assuming that the radiation value exceeding the standard has a great impact on risk assessment, it is assigned an attention weight of 0.4; the attention weight of the stability of key components of the equipment is 0.3, etc., to highlight the role of these indicators in risk assessment.

[0066] The risk feature template, the attention weight of each feature dimension, and the attention weight related to the key safety indicators of radiation equipment are processed to generate several radiation equipment safety leakage risk warning results and the confidence value attached to each warning result. The model uses the risk feature template as input for calculation and reasoning. When evaluating the radiation equipment safety leakage risk of the "RX-2000" CT machine, a comprehensive consideration will be made based on the association between the various factors in the risk feature template. For example, the equipment radiation safety sensitivity parameter in the template is high (0.9), combined with the average radiation detection value close to the upper limit of the normal range (standardized value 0.7) and the large fluctuation range of the radiation value (standardized value 0.6), the model will believe that there is a certain risk in radiation. At the same time, the threshold index of the qualification authority of the detection personnel is 1, and the standardized value of the compliance rate of the detection steps is 0.9, indicating that the personnel operate in a standardized manner and are qualified, which has a positive effect on reducing risks, but due to the prominent radiation-related risk factors, the overall risk still needs to be focused on. The certification status coefficient of the detection equipment is 1, indicating that the equipment has passed the certification, but the network security level parameter of 0.6 shows that there is a certain risk in the network, which will also be included in the scope of risk assessment. The model conducts a comprehensive analysis of these factors in the risk feature template and combines their respective attention weights to ultimately determine that the "RX-2000" CT machine presents a medium risk in this detection.

[0067] After calculation, the model may generate multiple warning results for the risk of radiation equipment safety leakage. For example, the first warning result is "medium risk". This is because although the equipment certification status is good (the certification status coefficient is 1), the equipment radiation safety sensitivity is high (the parameter is 0.9), and the average radiation detection value is within the normal range but the fluctuation range is large. These factors combined make the risk at a medium level. The model believes that this judgment has a 75% credibility, that is, the confidence value is 0.75.

[0068] The second warning result is "high risk" because there are certain irregularities in the operating habits of the inspectors (the compliance rate of the inspection steps is 92%, which is close to the full score but still has room for improvement), and there are some loopholes in the network security of the equipment (the network security level parameter is 0.6, which is at a medium security level). Combined with other factors, it is judged to be high risk. However, due to the uncertainty of some data, the model's confidence value for this judgment is 0.6. The third warning result is "low risk", which is based on the fact that the qualifications and permissions of the inspectors are fully compliant (the threshold indicator is 1), the recent equipment maintenance is good, and the radiation value fluctuation range is also within an acceptable range. The model's confidence value for this judgment is 0.85.

[0069] Screen and integrate several radiation equipment safety leakage risk warning results and the confidence values ​​attached to each warning result to generate radiation equipment safety leakage risk warning result information with confidence. From the multiple radiation equipment safety leakage risk warning results generated, it is necessary to screen out the most representative and reliable results and integrate them. The hospital sets the screening rule to select the warning result with the highest confidence as the final result. In the above example, the confidence of "low risk" is 0.85, the confidence of "medium risk" is 0.75, and the confidence of "high risk" is 0.6, so "low risk" is selected as the risk warning result of the "RX-2000" CT machine for this detection.

[0070] The screened warning results and their corresponding confidence values ​​are integrated to form the warning result information of radiation equipment safety leakage risk with confidence. For example, the warning result information finally generated is "The 'RX-2000' CT machine has a low risk in this test, and the confidence is 0.85". According to this warning result, the relevant hospital managers can make further arrangements for the use and maintenance of the "RX-2000" CT machine. If it is a high-risk warning, it may be necessary to stop using the equipment immediately and conduct a comprehensive overhaul; medium-risk requires enhanced monitoring and regular inspections; although low-risk is relatively safe, it is still necessary to continue to pay attention to the status of the equipment to ensure radiation safety.

[0071] In this application, the server collects historical equipment detection records, personnel identity information, business scenario information, equipment information, preset models and historical risk samples. Then, extract risk influencing factors from business scenarios, analyze historical detection records to generate standardized detection feature matrices, generate target risk feature vectors based on personnel identity screening, and analyze equipment information to generate detection equipment safety feature vectors. Subsequently, train the preset model based on historical risk samples to obtain an adaptive risk assessment model. Finally, input the above feature vectors into the adaptive model, fuse them through the attention mechanism, and output risk warning results with confidence. First, conduct in-depth analysis and processing of each key element to generate corresponding feature vectors and models, and then fuse the feature vectors through the model to achieve accurate radiation equipment risk warnings and ensure the safe use of radiation medical detection equipment.

[0072] In one embodiment, if Figure 2 As shown, the present application also provides an early warning device for radiation detection, comprising: Acquisition module 201, used to acquire historical detection records of radiation medical detection equipment, identity information of current detection personnel, detection business scenario information, detection equipment information, preset risk assessment model and historical risk data samples; Processing module 202 is used to extract target elements from detection business scenario information and generate risk impact factors, wherein the risk impact factors include detection urgency weights, equipment radiation safety sensitivity parameters, and detection personnel qualification and authority threshold indicators; perform data analysis and processing on historical detection records to generate a standardized detection feature matrix; perform feature dimension screening on the standardized detection feature matrix based on the identity information of the current detection personnel to generate a target risk feature vector, wherein the target risk feature vector includes equipment historical detection result features, detection personnel operation habit features, and detection process abnormal behavior features; perform security feature analysis on the equipment information used for detection to generate a detection equipment security feature vector, wherein the detection equipment security feature vector includes a detection equipment authentication status coefficient, a detection equipment network security level parameter, and a detection equipment abnormal behavior baseline value; perform transfer learning on a preset risk assessment model based on historical risk data samples to generate an adaptive risk assessment model; input the risk impact factor, target risk feature vector, and detection equipment security feature vector into the adaptive risk assessment model, perform feature fusion through the attention mechanism, and generate radiation equipment safety leakage risk warning result information with confidence.

[0073] The computer-readable storage medium provided in the above-mentioned embodiments of the present application and the radiation detection warning method provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0074] Each embodiment in the present application is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the early warning method, electronic device, electronic device, and readable storage medium embodiment for evaluating radiation detection, since they are basically similar to the embodiment of the early warning method for radiation detection described above, the description is relatively simple, and the relevant parts can be referred to the partial description of the embodiment of the early warning method for radiation detection described above.

Claims

1. A radiation detection early warning method, characterized in that: include: Obtain historical test records of radiation medical testing equipment, identity information of current test personnel, test business scenario information, test equipment information, preset risk assessment models and historical risk data samples; Extract target elements from the detection business scenario information and generate risk impact factors, where the risk impact factors include detection urgency weight, equipment radiation safety sensitivity parameters, and detection personnel qualification authority threshold indicators; Perform data analysis and processing on historical test records to generate a standardized test feature matrix; Based on the identity information of the current inspector, the standardized inspection feature matrix is ​​screened for feature dimensions to generate a target risk feature vector, where the target risk feature vector includes the features of the historical inspection results of the equipment, the operating habits of the inspector, and the abnormal behavior features of the inspection process; Perform security feature analysis on the device information used for detection to generate a detection device security feature vector, wherein the detection device security feature vector includes a detection device authentication status coefficient, a detection device network security level parameter, and a detection device abnormal behavior baseline value; Perform transfer learning on the preset risk assessment model based on historical risk data samples to generate an adaptive risk assessment model; The risk influencing factors, target risk feature vectors and detection equipment safety feature vectors are input into the adaptive risk assessment model, and feature fusion is performed through the attention mechanism to generate radiation equipment safety leakage risk warning result information with confidence.

2. The method according to claim 1, characterized in that Extract target elements from the detection business scenario information and generate risk impact factors, including: Extract target elements from the detection business scenario information to generate detection business scenario information features described in natural language, wherein the detection business scenario information features described in natural language are used to characterize basic information of the detection business scenario; Construct a knowledge ontology database for radiation detection business; Based on the radiation detection business knowledge ontology library, the information features of the detection business scenario described in natural language are mapped and processed to generate structured features; Process the structured features to generate business adaptation index, equipment radiation safety importance parameters and inspection personnel qualification compliance indicators; The business adaptation index, equipment radiation safety importance parameters and testing personnel qualification compliance indicators are converted and calculated to generate the detection urgency weight, equipment radiation safety sensitivity parameters and testing personnel qualification authority threshold indicators, which together constitute the risk influencing factors.

3. The method according to claim 1, characterized in that Perform data analysis on historical test records to generate a standardized test feature matrix, including: Performing time slicing processing on the historical detection records to generate a number of detection period records, wherein the detection period records are used to represent the record information of the radiation equipment detection in each period divided in the time dimension; Extract target data and normalize the records of several detection periods to generate a detection feature sequence, where the detection feature sequence includes information such as detection frequency, detection radiation value, equipment maintenance status, and detection operation type, which is used to characterize the behavioral characteristics of radiation equipment detection; The detection feature sequence is dimensionally aligned and standardized to generate a standardized detection feature matrix.

4. The method according to claim 3, characterized in that Based on the identity information of the current inspector, the standardized inspection feature matrix is ​​screened for feature dimensions to generate a target risk feature vector, including: Based on the identity information of the current inspector, the standardized inspection feature matrix is ​​screened in feature dimensions to generate the original features of the equipment's historical inspection results, the original features of the inspector's operating habits, and the original features of abnormal behaviors during the inspection process; The original features of the historical detection results of the equipment are processed to generate the average radiation detection value and the radiation value fluctuation range, wherein the average radiation detection value and the radiation value fluctuation range are used to characterize the historical level and stability of the radiation situation of the equipment; The original features of the inspectors' operating habits are processed to generate the average inspection time and the compliance rate of the inspection steps. The average inspection time and the compliance rate of the inspection steps are used to characterize the time regularity and compliance degree of the inspectors' operations. Processing the original features of abnormal behavior in the detection process to generate abnormal detection frequency, where the abnormal detection frequency is used to characterize the frequency of deviations from normal detection operation in the detection process; The average radiation detection value, radiation value fluctuation range, average detection time, detection step compliance rate and abnormal detection frequency are processed to generate a target risk feature vector.

5. The method according to claim 1, characterized in that Perform security feature analysis on the device information used for detection to generate a detection device security feature vector, where the detection device security feature vector includes the detection device authentication status coefficient, the detection device network security level parameter and the detection device abnormal behavior baseline value, including: Perform authentication status analysis on the equipment information used in the test to generate the authentication status coefficient of the test equipment; Obtain network security assessment reports of detection equipment and historical operation log information of equipment; Processing the network security assessment report to generate network security level parameters for the detection equipment, wherein the network security level parameters are used to characterize the security level of the network environment in which the detection equipment is located; Process the historical operation log information of the device, count the frequency and type of abnormal operations, and generate a baseline value for detecting abnormal behavior of the device, wherein the baseline value for detecting abnormal behavior of the device includes the average frequency of abnormal operations of the device and the proportion of abnormal operation types; The detection equipment authentication status coefficient, the detection equipment network security level parameter and the detection equipment abnormal behavior baseline value are processed to generate a detection equipment security feature vector.

6. The method according to claim 1, characterized in that Based on historical risk data samples, transfer learning is performed on the preset risk assessment model to generate an adaptive risk assessment model, including: Count the number of risk features in the historical risk data sample and generate sampling ratios based on these numbers; Sampling and processing historical risk data samples based on the sampling ratio to generate a preset number of sampling risk features; Based on a comparative analysis between any risk feature and each sampled risk feature, the historical risk data samples are divided into a high radiation risk group and a low radiation risk group, wherein each group contains a preset number of data samples, and at least one data sample carries radiation risk level identification information; Perform transfer learning training on the preset risk assessment model based on the high radiation risk group and the low radiation risk group to generate a trained risk assessment model and training results; If the data samples containing identification information in the training results are factors that characterize high radiation detection risks, the trained risk assessment model is used as an adaptive risk assessment model.

7. The method according to claim 1, characterized in that The risk influencing factors, target risk feature vectors and detection equipment safety feature vectors are input into the adaptive risk assessment model, and feature fusion is performed through the attention mechanism to generate radiation equipment safety leakage risk warning result information with confidence, including: Based on the adaptive risk assessment model, the risk influencing factors, target risk feature vectors and detection equipment safety feature vectors are processed to generate risk feature templates, attention weights for each feature dimension, and attention weights related to key safety indicators of radiation equipment; Process the risk feature template, the attention weight of each feature dimension, and the attention weight related to the key safety indicators of the radiation equipment to generate several radiation equipment safety leakage risk warning results and the confidence value attached to each warning result; The safety leakage risk warning results of several radiation equipments and the confidence values ​​attached to each warning result are screened and integrated to generate the safety leakage risk warning result information of radiation equipments with confidence.

8. A radiation detection early warning device, characterized in that: The device comprises: The acquisition module is used to obtain historical test records of radiation medical testing equipment, the identity information of the current test personnel, the test business scenario information, the equipment information used for the test, the preset risk assessment model and historical risk data samples; The processing module is used to extract target elements from the detection business scenario information and generate risk impact factors, wherein the risk impact factors include the detection urgency weight, the equipment radiation safety sensitivity parameter and the detection personnel qualification and authority threshold index; perform data analysis and processing on historical detection records to generate a standardized detection feature matrix; perform feature dimension screening on the standardized detection feature matrix based on the identity information of the current detection personnel to generate a target risk feature vector, wherein the target risk feature vector includes the equipment historical detection result characteristics, the detection personnel operation habit characteristics and the detection process abnormal behavior characteristics; perform security feature analysis on the equipment information used for detection to generate a detection equipment security feature vector, wherein the detection equipment security feature vector includes the detection equipment authentication status coefficient, the detection equipment network security level parameter and the detection equipment abnormal behavior baseline value; perform transfer learning on the preset risk assessment model based on the historical risk data sample to generate an adaptive risk assessment model; input the risk impact factor, the target risk feature vector and the detection equipment security feature vector into the adaptive risk assessment model, perform feature fusion through the attention mechanism, and generate radiation equipment security leakage risk warning result information with confidence.

9. An electronic device, characterized in that: include: a first processor; and a memory for storing executable instructions of the first processor; Wherein, the first processor is configured to execute the radiation detection early warning method according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the second processor, the radiation detection early warning method according to any one of claims 1 to 7 is implemented.

Images