CT tube life prediction method, device, server and computer-readable storage medium

Abstract

This application proposes a method, device, server, and computer-readable storage medium for predicting the lifespan of a CT tube, relating to the field of medical devices. The tube lifespan curve is generated based on multiple cumulative exposure times and the predicted failure probability corresponding to each cumulative exposure time. The predicted failure probability is generated by processing the cumulative exposure times into a pre-generated failure prediction model, which is trained based on multiple sample lifespan data and corresponding failure probability labels. The current cumulative exposure time of the CT tube and the target tube lifespan curve are obtained. Based on the target tube lifespan curve, the current failure probability of the CT tube is determined according to the current cumulative exposure time. If the current failure probability exceeds a preset failure threshold corresponding to the CT tube, a failure alarm is issued. This method captures the true degradation characteristics of the CT tube, improves the fitting accuracy of the CT tube failure probability curve, and ensures the normal use of the CT equipment.

Description

Technical Field

[0001] This application relates to the field of medical devices, and more specifically, to a method, apparatus, server, and computer-readable storage medium for predicting the lifespan of a CT tube. Background Technology

[0002] With the rapid development of medical imaging technology, computed tomography (CT) equipment has become an indispensable instrument for disease diagnosis in modern hospitals. In the entire CT system, the X-ray tube (hereinafter referred to as the CT tube) is its core hardware component. It is a high-value, high-precision vacuum electronic device responsible for generating the X-ray beam used for imaging. The performance of the CT tube directly determines not only the clarity and diagnostic accuracy of the CT images, but also whether the entire CT equipment can continuously and stably provide clinical services. If the CT tube suddenly fails, it will not only result in high replacement costs, but also cause unplanned downtime of the CT equipment, leading to a series of chain reactions such as interrupted imaging services, prolonged patient waiting times, and disruption of clinical treatment processes.

[0003] Currently, lifespan analysis and prediction methods in related technologies are still mainly based on parametric statistical models. Among them, the most widely used method is the Weibull distribution-based analysis method. This method assumes that the failure time of all CT tubes follows a specific mathematical distribution, thereby constructing the failure curve of the CT tube. Although this method is simple to calculate and the results are intuitive, it also has certain limitations.

[0004] Specifically, this method is highly dependent on prior assumptions. It must be assumed that the failure time of the CT tube strictly follows the specific function shape of the Weibull distribution. However, in actual applications, the lifespan data of the CT tube often exhibits complex and atypical characteristics. In this case, if the analysis method based on the Weibull distribution with a fixed distribution shape is still used, it is difficult to capture the true degradation characteristics of the CT tube, resulting in low fitting accuracy. The generated failure curve deviates significantly from the actual empirical distribution, making it difficult to detect the degradation of the CT tube in a timely manner during the application of CT equipment, thus affecting the normal use of CT equipment. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a method, device, server and computer-readable storage medium for predicting the lifespan of a CT tube, so as to capture the true degradation characteristics of the CT tube, improve the fitting accuracy of the CT tube failure probability curve, and ensure the normal use of CT equipment.

[0006] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows: In a first aspect, this application provides a CT tube life prediction method applied to a server. The server is communicatively connected to multiple CT devices. The server stores tube life curves corresponding to each CT tube type. The tube life curves characterize the relationship between the cumulative exposure time and the failure probability of the CT tube. The tube life curves are generated based on multiple cumulative exposure times and the predicted failure probabilities corresponding to each cumulative exposure time. The predicted failure probabilities are generated by inputting the cumulative exposure times into a pre-generated failure prediction model. The failure prediction model is trained based on multiple sample life data and the failure probability labels corresponding to the sample life data. The method includes: Obtain the current cumulative exposure time of the CT tube in any of the CT devices and the target tube life curve corresponding to the CT tube; Based on the target X-ray tube life curve, the current failure probability of the CT tube is determined according to the current cumulative exposure time; If the current failure probability exceeds the preset failure threshold corresponding to the CT tube, a failure alarm message is issued.

[0007] In an optional implementation, the X-ray tube life curve is obtained through the following steps: For each of the CT tube types, multiple sample lifetime data corresponding to the tube type are obtained. Multiple failure probability labels are generated based on the multiple sample lifetime data. The failure probability labels and the sample lifetime data are input into a pre-built failure prediction model for training to obtain a trained failure prediction model. Multiple cumulative exposure durations corresponding to the CT tube type are obtained, and each cumulative exposure duration is input into the trained failure prediction probability for processing to obtain the first failure probability corresponding to each cumulative exposure duration. By performing range constraints and order-preserving regression on each of the first failure probabilities, the predicted failure probabilities corresponding to each of the cumulative exposure durations are obtained. Curve fitting is performed based on the cumulative exposure time and the predicted failure probability corresponding to the cumulative exposure time to obtain the tube life curve corresponding to the CT tube type.

[0008] In an optional implementation, generating multiple failure probability labels based on multiple sample lifetime data includes: The lifetime data of each sample are sorted according to their numerical values ​​to obtain the X-ray tube lifetime sequence; For each sample lifetime data in the X-ray tube lifetime sequence, the failure probability label corresponding to the sample lifetime data is calculated based on the rank of the sample lifetime data in the X-ray tube lifetime sequence and the number of sample lifetime data.

[0009] In an optional implementation, the step of performing range constraints and order-preserving regression on each of the first failure probabilities to obtain the predicted failure probability corresponding to each of the cumulative exposure durations includes: The first failure probability is constrained according to a preset value range so that each first failure probability is within the value range, thereby obtaining the second failure probability. The second failure probabilities are monotonically smoothed according to the preset prediction trend so that the change trend of the second failure probabilities conforms to the prediction trend, thereby obtaining the predicted failure probabilities corresponding to each cumulative exposure duration.

[0010] In an optional implementation, the step of inputting the failure probability label and the sample lifetime data into a pre-built failure prediction model for training to obtain a trained failure prediction model includes: Construct feature vectors corresponding to the lifetime data of each sample based on the CT tube type; The feature vector and failure probability label corresponding to each of the sample lifetime data are input into the pre-built failure prediction model for training, and the trained failure prediction model is obtained. The feature vector is ,in, For the feature vector, The sample lifetime data, The CT tube type ID is used; the failure prediction model includes an input layer, multiple fully connected layers, and an output layer, and the number of neurons in each fully connected layer decreases sequentially with the data transmission direction.

[0011] Secondly, this application provides a CT tube life prediction device applied to a server. The server is communicatively connected to multiple CT devices. The server stores tube life curves corresponding to each type of CT tube, and the tube life curves characterize the relationship between the cumulative exposure time and the failure probability of the CT tube. The tube life curves are generated based on multiple cumulative exposure times and the predicted failure probabilities corresponding to each cumulative exposure time. The predicted failure probabilities are generated by inputting the cumulative exposure times into a pre-generated failure prediction model. The failure prediction model is trained based on multiple sample life data and the failure probability labels corresponding to the sample life data. The device includes: The acquisition module is used to acquire the current cumulative exposure time of the CT tube in any of the CT devices and the target tube life curve corresponding to the CT tube; The determination module is used to determine the current failure probability of the CT tube based on the target tube life curve and the current cumulative exposure time. The determining module is further configured to issue a failure alarm message if the current failure probability exceeds the preset failure threshold corresponding to the CT tube.

[0012] In an optional embodiment, the apparatus further includes: The training module is used to acquire multiple sample lifetime data corresponding to each of the CT tube types, generate multiple failure probability labels based on the multiple sample lifetime data, and input the failure probability labels and the sample lifetime data into a pre-built failure prediction model for training to obtain a trained failure prediction model. The generation module is used to obtain multiple cumulative exposure times corresponding to the CT tube type, input each cumulative exposure time into the trained failure prediction probability for processing, and obtain a first failure probability corresponding to each cumulative exposure time; perform range constraints and order-preserving regression on each first failure probability to obtain the failure probability corresponding to each cumulative exposure time; and perform curve fitting based on each cumulative exposure time and the failure probability corresponding to the cumulative exposure time to obtain the tube life curve corresponding to the CT tube type.

[0013] In an optional implementation, the training module is further configured to sort the sample lifetime data according to their numerical values ​​to obtain a tube lifetime sequence; and to calculate the failure probability label corresponding to each sample lifetime data in the tube lifetime sequence based on the rank of the sample lifetime data in the tube lifetime sequence and the number of sample lifetime data.

[0014] Thirdly, this application provides a server including a processor and a memory, wherein the memory stores a computer program that can be executed by the processor, and the processor can execute the computer program to implement the CT tube life prediction method according to any of the foregoing embodiments.

[0015] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the CT tube life prediction method as described in any of the foregoing embodiments.

[0016] The CT tube life prediction method, apparatus, server, and computer-readable storage medium provided in this application embodiment include a server storing tube life curves corresponding one-to-one with each CT tube type. These tube life curves characterize the relationship between the cumulative exposure time and failure probability of the CT tube. The server can generate multiple predicted failure probabilities corresponding to cumulative exposure times by pre-training a failure prediction model based on multiple sample life data and corresponding failure probability labels. Then, the tube life curve is generated based on these multiple cumulative exposure times and their corresponding predicted failure probabilities. In this way, the failure prediction model can be trained based on real operating data, and multiple predicted failure probabilities corresponding to cumulative exposure times can be generated based on this model, thus ensuring that the predicted failure probability conforms to the actual degradation characteristics of the CT tube. Furthermore, a curve is obtained based on the mapping relationship between cumulative exposure time and predicted failure probability. Since this curve is not based on an idealized mathematical distribution assumption but is directly generated based on points corresponding to multiple mapping relationships, the fitting accuracy can be improved. In practical applications, the server can determine the current failure probability of the CT tube based on the target tube life curve and the current cumulative exposure time of the CT tube. When the current failure probability exceeds the preset failure threshold, the server can promptly issue a failure alarm message to prompt staff to intervene in time and ensure the normal use of the CT equipment.

[0017] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A block diagram of a server provided in an embodiment of this application is shown; Figure 2 This paper illustrates a flowchart of a CT tube life prediction method provided in an embodiment of this application. Figure 3 Box plots showing the lifespan distribution of five different CT tube models are presented. Figure 4 The lifetime probability density function estimation curves for five different CT tube models are shown. Figure 5 This diagram shows a comparison of the fitting curves for type A X-ray tubes. Figure 6This diagram shows a comparison of the fitting curves for the B-type X-ray tube. Figure 7 This diagram shows a comparison of the fitting curves for the C-type X-ray tube. Figure 8 This diagram shows a comparison of the fitting curves for the D-type X-ray tube. Figure 9 This diagram shows a comparison of the fitting curves for the E-type X-ray tube. Figure 10 A schematic diagram illustrating the prediction accuracy of the Weibull method is shown. Figure 11 A schematic diagram illustrating the prediction accuracy of the CT tube life prediction method provided in this application embodiment is shown. Figure 12 This paper illustrates a functional block diagram of a CT tube life prediction device provided in an embodiment of this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0023] Figure 1 Please refer to the block diagram of the server provided in the embodiments of this application. Figure 1The server includes a memory, a processor, and a communication module. These components are electrically connected directly or indirectly to enable data transmission or interaction. For example, they can be electrically connected via one or more communication buses or signal lines.

[0024] Memory is used to store computer programs or data that can be executed by a processor. Memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc.

[0025] The processor is used to read / write data or computer programs stored in the memory and execute the computer program to implement the CT tube life prediction method provided in the embodiments of this application.

[0026] The communication module is used to establish communication connections between the server and other communication terminals via the network, and to send and receive data via the network.

[0027] It should be understood that, Figure 1 The structure shown is only a schematic diagram of the server structure; the server may also include components such as... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown. Figure 1 The components shown can be implemented using hardware, software, or a combination thereof.

[0028] In this embodiment, the server can also communicate with multiple CT devices, and the server stores the tube life curves corresponding to each type of CT tube. The tube life curves represent the correspondence between the cumulative exposure time of the CT tube and the failure probability.

[0029] In this embodiment, the cumulative exposure time refers to the total time during which X-rays are actually radiated during the entire scanning process from the time the CT tube is put into use, and the unit can be seconds. This value can be continuously accumulated by the exposure timing module built into the CT equipment.

[0030] The failure probability refers to the probability that a CT tube of a certain model will experience functional failure under a given cumulative exposure time, and its value ranges from [0,1]. Understandably, a higher failure probability indicates that the CT tube is more likely to fail under that cumulative exposure time. In one possible implementation, the tube lifespan curve is generated based on multiple cumulative exposure times and the predicted failure probability corresponding to each cumulative exposure time. The predicted failure probability is generated by inputting the cumulative exposure times into a pre-generated failure prediction model, which is trained based on multiple sample lifespan data and the failure probability labels corresponding to the sample lifespan data.

[0031] It should be noted that the degradation of CT tubes can be affected by complex environmental and usage factors. Simply fitting failure curves according to a specific mathematical distribution, as is often the case in related technologies, is insufficient to fully learn and characterize the failure features of CT tubes. Therefore, in this embodiment, the server first obtains multiple sample lifetime data and corresponding failure probability labels based on CT tubes in actual operating environments. Then, it trains a failure prediction model based on the sample lifetime data and failure probability labels, thereby overcoming the reliance on idealized assumptions in traditional methods and enabling the model to fully learn the failure characteristics of CT tubes.

[0032] Furthermore, considering the significant differences in lifespan between different types of CT tubes, failure prediction models can be trained for different tube types. In one possible implementation, the CT tube type can refer to the manufacturer and / or model of the CT tube.

[0033] In this embodiment, the sample lifetime data refers to the cumulative exposure time of multiple failed CT tubes under the corresponding CT tube type. The value can be obtained from the hospital's maintenance logs or the disassembly and inspection report provided by the manufacturer. For example, if a certain type A CT tube can no longer stably produce qualified X-rays after 178695.8 seconds of exposure, then this value is the sample lifetime data of that CT tube.

[0034] The failure probability label is a probability identifier assigned to the above sample lifespan data, that is, it represents approximately this proportion of CT tubes of this model that will fail at the time point corresponding to the sample lifespan data.

[0035] In this embodiment, after generating failure prediction models for each CT tube type, the server can calculate the predicted failure probability for each cumulative exposure time based on the failure prediction models for each tube type and multiple cumulative exposure times. At this time, multiple sets of mapping relationships between cumulative exposure times and predicted failure probabilities can be obtained. By taking the cumulative exposure time as the horizontal axis and the predicted failure probability as the vertical axis, multiple points can be obtained. In this case, the server can generate a smooth tube life curve based on multiple points, which can characterize the failure features of this CT tube type.

[0036] Understandably, the more cumulative exposure time and predicted failure probability data points there are, the higher the accuracy of the X-ray tube life curve. The specific number can be selected based on the actual application, such as 100,000 data points. The following will use the above... Figure 1 The server in this application serves as the execution entity, and the CT tube life prediction method provided in the embodiments of this application is described executively.

[0037] Specifically, Figure 2 For a flowchart illustrating the CT tube life prediction method provided in this application embodiment, please refer to [link / reference]. Figure 2 The method includes: Step S20: Obtain the current cumulative exposure time of the CT tube in any CT device and the target tube life curve corresponding to the CT tube.

[0038] Step S21: Based on the target X-ray tube life curve, determine the current failure probability of the CT tube according to the current cumulative exposure time.

[0039] Step S22: If the current failure probability exceeds the preset failure threshold corresponding to the CT tube, a failure alarm message is issued.

[0040] In this embodiment, the server can acquire the current cumulative exposure time of the CT tubes in each CT device in real time or at preset intervals, and determine the target tube life curve corresponding to each CT tube based on the tube type of each CT tube.

[0041] In this embodiment, for each CT tube, the server can determine the current failure probability corresponding to the current cumulative exposure seconds on the target tube lifespan curve corresponding to that tube. Understandably, this current failure probability characterizes the statistical probability that the CT tube has failed or is about to fail under the current cumulative exposure duration.

[0042] Furthermore, the server can compare the current failure probability with a preset failure threshold. If the current failure probability exceeds the preset failure threshold, it indicates that the degradation of the CT tube has reached a critical level requiring intervention. At this time, the server can automatically issue a failure alarm message to prompt maintenance personnel to carry out timely inspection, maintenance, or replacement preparations.

[0043] In this embodiment, the preset failure threshold for each CT tube type may be different, and can be determined based on the tube life curve and actual application conditions of the CT tube type.

[0044] In this way, the entire process does not require human experience or deep understanding of the internal mechanism of the equipment. It can achieve objective quantitative assessment and proactive risk warning of the CT tube life status simply by using measurable cumulative exposure time and tube life curve that fully characterizes tube failure.

[0045] The CT tube life prediction method provided in this application involves storing tube life curves in the server that correspond one-to-one with each CT tube type. These tube life curves characterize the relationship between the cumulative exposure time and the failure probability of the CT tube. The server can generate multiple predicted failure probabilities corresponding to the cumulative exposure time by pre-training a failure prediction model based on multiple sample life data and corresponding failure probability labels. Then, the tube life curve is generated based on the multiple cumulative exposure times and their corresponding predicted failure probabilities. In this way, the failure prediction model can be trained based on real operating data, and multiple predicted failure probabilities corresponding to the cumulative exposure time can be generated based on the failure prediction model, thus ensuring that the predicted failure probability conforms to the actual degradation characteristics of the CT tube. In addition, a curve is obtained based on the mapping relationship between the cumulative exposure time and the predicted failure probability. Since this curve is not based on an idealized mathematical distribution assumption, but is directly generated based on the points corresponding to multiple mapping relationships, the fitting accuracy can be improved. In practical applications, the server can determine the current failure probability of the CT tube based on the target tube life curve and the current cumulative exposure time of the CT tube. When the current failure probability exceeds the preset failure threshold, the server can promptly issue a failure alarm message to prompt staff to intervene in time and ensure the normal use of the CT equipment.

[0046] Next, we will provide a possible implementation method for generating the X-ray tube life curve.

[0047] Specifically, the server can first obtain multiple sample lifetime data corresponding to each CT tube type, generate multiple failure probability labels based on the multiple sample lifetime data, and input the failure probability labels and sample lifetime data into the pre-built failure prediction model for training to obtain the trained failure prediction model.

[0048] Next, multiple cumulative exposure times corresponding to the CT tube type are obtained. Each cumulative exposure time is input into the trained failure prediction probability for processing to obtain the first failure probability corresponding to each cumulative exposure time. Range constraints and order-preserving regression are performed on each first failure probability to obtain the predicted failure probability corresponding to each cumulative exposure time. Based on each cumulative exposure time and the predicted failure probability corresponding to the cumulative exposure time, curve fitting is performed to obtain the tube life curve corresponding to the CT tube type.

[0049] In this embodiment, for any CT tube type, the server can first acquire sample lifetime data of multiple CT tubes with known actual lifespans under that type, and generate a corresponding failure probability label for each sample lifetime data based on the relationship between the sample lifetime data. Then, the server can input these sample lifetime data and their corresponding failure probability labels into a pre-built failure prediction model for supervised training, thereby obtaining a trained failure prediction model for that tube type.

[0050] After obtaining the trained failure prediction model corresponding to the CT tube type, the server can generate a set of multiple cumulative exposure durations covering the entire life cycle for the CT tube type.

[0051] In one possible implementation, the effective cumulative exposure time range may differ for different types of CT tubes. The server can expand the value based on the sample lifetime data corresponding to the CT tube type to obtain multiple cumulative exposure times.

[0052] Specifically, the server can first statistically analyze the minimum and maximum values ​​of all sample lifetime data for this type of X-ray tube to determine the effective lifetime range actually observed. Then, it can appropriately extend the range to both ends and perform discrete sampling at a higher density within the extended range to generate a set of cumulative exposure times that cover a wider time span and are far more numerous than the original samples (e.g., expanded from 10 samples to 100,000 points).

[0053] This expanded set of cumulative exposure times not only fully encompasses the lifetime distribution boundaries revealed by existing samples, but also provides a physically reasonable support interval for model extrapolation, ensuring that the final fitted X-ray tube lifetime curve has reliable predictive ability in the initial, middle, and long tail segments, avoiding curve distortion or key indicator calculation deviations due to insufficient sampling.

[0054] After obtaining multiple cumulative exposure durations, the server can input each cumulative exposure duration into the failure prediction model that has been trained above to obtain the corresponding first failure probability. Range constraints and order-preserving regression are then applied to all first failure probabilities in sequence to ensure that the final output sequence of predicted failure probabilities strictly satisfies the mathematical definition and physical meaning of the Cumulative Distribution Function (CDF).

[0055] Based on this, the server can generate a continuous, smooth, and monotonically undecreasing curve based on each cumulative exposure duration and its processed predicted failure probability; this curve represents the X-ray tube lifespan curve for that CT tube type. In this embodiment, the server needs to perform the above steps separately for different CT tube types to generate the X-ray tube lifespan curve for each type.

[0056] In one possible approach, the median rank method can be used to determine the corresponding failure probability label for each sample lifetime data.

[0057] In this embodiment, the server can sort the lifetime data of each sample according to their numerical values ​​to obtain the X-ray tube lifetime sequence; for each sample lifetime data in the X-ray tube lifetime sequence, the server can calculate the failure probability label corresponding to the sample lifetime data based on the rank of the sample lifetime data in the X-ray tube lifetime sequence and the number of sample lifetime data.

[0058] In practical applications, the server can first acquire a set of sample lifetime data corresponding to any type of CT tube. This data represents the cumulative exposure time of that type of tube in a real operating environment, from its initial use to failure, typically measured in seconds. These lifetime data points are discrete time points and do not have pre-defined probabilistic meanings. To model the lifetime prediction task as a supervised learning problem, the server needs to assign a continuous probability value between zero and one to each sample lifetime data point, serving as a failure probability label for subsequent training of the failure prediction model.

[0059] To this end, the server uses the median rank method, a standard method in reliability engineering, to construct the Empirical Cumulative Distribution Function (CDF). This method does not rely on any preset failure distribution form and can generate statistically reasonable probability labels based solely on the sorting position and total number of samples.

[0060] In this embodiment, the server can arrange all sample data corresponding to any CT tube type in ascending order of numerical values, forming an ordered tube lifespan sequence. Based on this, the server can calculate the cumulative failure probability label corresponding to the i-th sample lifespan data in this sequence, according to its rank i (sorting position) in the sequence and the total number n of all valid samples corresponding to that tube type, using the Bernard approximation formula. The formula is:

[0061] in, This represents the lifetime data of the i-th sample after sorting. The corresponding cumulative failure probability takes values ​​within the range [0,1].

[0062] For example, if a certain type of CT tube has 10 valid sample lifetime data, the server can first arrange these 10 data in ascending order to form a tube lifetime sequence, such as 100,000 seconds, 110,000 seconds, ..., 190,000 seconds. Then, for each sample lifetime data in the sequence, the server calculates the cumulative failure probability corresponding to the sample lifetime data according to the above formula, and uses the cumulative failure probability as the failure probability label corresponding to the sample lifetime data.

[0063] It should be noted that the median rank method is a standard statistical method used in reliability engineering to construct an empirical cumulative distribution function from a finite number of discrete failure samples; while the Bernard approximation formula is the most commonly used and mature mathematical implementation of the median rank method in practical engineering applications.

[0064] Understandably, the core of the median rank method lies in not presupposing any specific probability distribution form, but directly generating robust probability estimates based on the ranking position of the observed data. Compared with the simple average rank method, the median rank method has a smaller estimation bias of the median value of the overall distribution in the case of small samples, and thus can provide a more representative and stable continuous probability supervision signal for failure prediction models.

[0065] In this way, the server can transform discrete X-ray tube lifetime observation data into probabilistic targets with clear mathematical meaning and physical interpretability without relying on prior parameterization assumptions such as the Weibull distribution. This satisfies the basic requirements of supervised learning for the continuity and orderliness of labels, and ensures that the generated failure probability labels naturally conform to the definition of the empirical cumulative distribution function, laying the foundation for the subsequent construction of smooth, monotonic and physically consistent X-ray tube lifetime curves.

[0066] Next, the lifetime data of each sample and its corresponding failure probability label can be input into the pre-built failure prediction model for training.

[0067] In this embodiment, the server can first construct feature vectors corresponding to the lifespan data of each sample according to the CT tube type, and then input the feature vectors and failure probability labels corresponding to the lifespan data of each sample into the pre-constructed failure prediction model for training, so as to obtain the trained failure prediction model.

[0068] The feature vector is ,in, For feature vectors, For sample lifetime data, The CT tube type ID is used; the failure prediction model includes an input layer, multiple fully connected layers, and an output layer, with the number of neurons in each fully connected layer decreasing sequentially with the data transmission direction.

[0069] In this embodiment, the server can construct a feature vector corresponding to the sample lifetime data based on the characteristics of the bathtub curve.

[0070] It should be noted that the bathtub curve is a reliability model that describes the failure rate of equipment as high in the early stage, stable in the middle stage, and accelerating in the final stage. In this embodiment, the bathtub curve is used to construct feature vectors not to forcibly fit the tube life distribution into a certain preset parameterized form (such as the Weibull distribution), but to use the general degradation stage law revealed by the curve to enhance the model's ability to learn and express the real degradation process without assuming the failure distribution pattern.

[0071] Understandably, CT tubes do not degrade at a uniform rate, but rather exhibit varying degradation rates across different time intervals. Specifically, the initial stage often sees drastic changes due to break-in and defect exposure, the middle stage tends to stabilize, and the final stage sees accelerated deterioration due to material fatigue and performance degradation. Based on this, the input feature vector can be designed piecewise based on the bathtub curve, i.e., by introducing a logarithmic transformation. Characterizing drastic changes in the early stages, using linear terms and square root terms. Characterizing stable development in the medium term, and employing higher-order polynomials. It indicates an accelerated deterioration trend in the final stage.

[0072] Based on this, the feature vector That is, it is characterized as: ,in, This is the CT tube type ID, used to distinguish different tube types.

[0073] Understandably, this embodiment uses the failure probability label obtained based on the median rank method as the supervision signal, and uses the bathtub curve to define the input feature vector. Therefore, there is no need to make additional assumptions about the specific failure mechanism. Instead, it is designed based on the statistical patterns that the time series may present at different stages, which improves the model's ability to express nonlinear structures and avoids dependence on specific failure distribution forms.

[0074] Considering the potential differences in numerical magnitudes between different features, directly inputting them into the neural network could lead to instability in the gradient update process, thereby affecting the model's convergence speed and training results. Therefore, in this embodiment, the server can also perform Z-score standardization on the aforementioned feature vectors, that is, calculate the mean of each feature on the training set. and standard deviation and according to the formula Transformation is performed to unify the features of each dimension to a standard scale with a mean of 0 and a standard deviation of 1, so as to avoid instability in neural network training due to differences in dimensions.

[0075] After obtaining the processed feature vector, the server can input the feature vector and its corresponding failure probability label into a pre-built failure prediction model for training.

[0076] In one possible implementation, the failure prediction model can be a deep, fully connected neural network, such as a DNN (Deep Neural Network) or an MLP (Multilayer Perceptron). This failure probability model can include an input layer, hidden layers, and an output layer, wherein the input layer can receive the aforementioned six-dimensional feature vector. The hidden layer can include multiple fully connected layers (e.g., four fully connected layers), with the number of neurons decreasing sequentially with the direction of data transmission, namely 400, 300, 200, and 100, respectively. This allows for layer-by-layer compression of the feature space and extraction of higher-order abstract features. In addition, each fully connected layer can use ReLU (Rectified Linear Unit) as the activation function to balance sparsity and gradient propagation stability. The output layer can be a single neuron, outputting the predicted failure probability corresponding to the cumulative exposure time of the input.

[0077] In this embodiment, the goal of model training is to minimize the mean squared error between the predicted distribution and the empirical rank distribution, and an L2 regularization term is introduced to prevent overfitting. Let the training sample set be... ,in, Let i be the feature vector corresponding to the lifetime data of the i-th sample. Let be the failure probability label corresponding to the i-th sample lifetime data, and N be the total number of sample lifetime data. Then, its loss function is defined as:

[0078] in, These are the hyperparameters of the failure prediction model. For the output of the model, This is the regularization coefficient, set according to the actual application, for example, set to 0.001.

[0079] In the above definition of the loss function, the first term The standard mean square error is used to constrain the predicted distribution to approximate the empirical median-rank distribution. The second term... This is an L2 regularization term used to penalize excessively large parameter norms, thereby reducing model complexity and mitigating overfitting.

[0080] Alternatively, an Adaptive Moment Estimation (Adam) optimizer can be used to iteratively update the parameter θ to optimize the failure prediction model. For example, the initial learning rate can be set to 0.001, and the learning rate can be dynamically decayed according to the loss decrease during training, thereby improving the fitting accuracy and result stability in the later stages while ensuring rapid convergence in the early stages.

[0081] In this way, the server can train a failure prediction model with good generalization ability under limited sample conditions.

[0082] In this embodiment, after obtaining the trained failure prediction model corresponding to each CT tube type, the server can input multiple cumulative exposure times corresponding to each CT tube type into the failure prediction model corresponding to that CT tube type for processing, and obtain multiple predicted failure probabilities corresponding to that CT tube type.

[0083] Then, the server can generate tube life curves for each CT tube type based on the cumulative exposure time and predicted failure probability.

[0084] Considering that although neural network models have a strong approximation ability in regression tasks, their original output may still have problems such as output values ​​exceeding the limit and local non-monotonic changes in the curve in practical applications, which are inconsistent with physical laws. Therefore, it is necessary to constrain the predicted failure probability output by the failure prediction model to ensure that the final generated X-ray tube life curve satisfies the mathematical and physical definitions of the cumulative distribution function.

[0085] Specifically, the server can constrain the range of each first failure probability according to a preset value range so that each first failure probability is within the value range to obtain the second failure probability; and perform monotonically smoothing on each second failure probability according to a preset prediction trend so that the change trend of the second failure probability conforms to the prediction trend to obtain the failure probability corresponding to each cumulative exposure duration.

[0086] In this embodiment, the server can perform range constraints and monotonic smoothing on the predicted failure probability to correct the original predicted value output by the failure prediction model to a failure probability value that conforms to both the probability interval [0,1] and the objective law of equipment degradation (monotonically does not decrease with cumulative exposure time).

[0087] In practical applications, the server can first obtain the first failure probabilities output by the failure prediction model, and then impose range constraints on each first failure probability, that is, substitute each first failure probability into the formula. In the interval projection, where This is the first failure probability. This is the second failure probability obtained after range constraints. This operation forces all out-of-bounds values ​​to be mapped to the closed interval [0,1].

[0088] Subsequently, the server can use an order-preserving regression method to monotonically smooth all second failure probabilities. That is, while minimizing the overall fitting error, a monotonically non-decreasing constraint is applied to the second failure probability sequence, such that for any two cumulative exposure durations... The corresponding second failure probability satisfies ,Right now The corresponding second failure probability should be less than or equal to The corresponding second failure probability.

[0089] Understandably, this process does not change the overall upward trend of the prediction curve, but only makes minimal perturbation adjustments in areas where local oscillations or reversals occur, ensuring that the final predicted failure probabilities strictly satisfy the mathematical definition of the cumulative distribution function and the physical consistency of the CT tube degradation process.

[0090] In one example, for a certain type of CT tube, taking five cumulative exposure times (in seconds) as input, the server inputs the five cumulative exposure times of 1000, 2000, 3000, 4000, and 5000 into the trained failure prediction model, and obtains the first failure probabilities corresponding to each cumulative exposure time: -0.01, 0.25, 0.22, 0.68, and 1.02. Among them, -0.01 and 1.02 obviously exceed the probability interval [0,1], while 0.22 does not conform to the monotonically decreasing trend.

[0091] The server can first constrain the range of the above five first failure probabilities, and take the values ​​of -0.01 and 1.02 that are closest to the probability interval [0,1] as their corresponding second failure probabilities, that is, take 0 for -0.01 and 1 for 1.02 to obtain the following second failure probabilities: 0, 0.25, 0.22, 0.68, and 1.

[0092] Next, ordinal regression can be performed on each of the aforementioned second failure probabilities. At this point, the target second failure probability that does not satisfy the monotonically decreasing trend can be taken and its adjacent preceding second failure probability, and their average value can be calculated as the new target second failure probability. In this example, 0.22 and 0.25 can be chosen, and their average value of 0.235 can be used to replace 0.22 as the predicted failure probability corresponding to 3000, so that the final predicted failure probabilities are: 0, 0.25, 0.235, 0.68, and 1.

[0093] The following section further explains the performance of the CT tube life prediction method provided in this application embodiment, based on experimental data.

[0094] In this embodiment, the data used comes from the CT IoT platforms of several large hospitals, specifically the actual operational lifespan data of five typical CT tube models. These CT tube models are designated as Model A, Model B, Model C, Model D, and Model E. The raw data underwent rigorous screening, removing samples with incomplete records or obvious logical errors. The final valid sample sizes were: 72 samples for Model A, 29 samples for Model B, 19 samples for Model C, 17 samples for Model D, and 14 samples for Model E. All samples use the time from the first use of a single CT tube to the occurrence of irreparable failure as their sample lifespan data, in seconds. This sample lifespan data has not been truncated or deleted in any way, representing a complete failure time observation.

[0095] Specifically, Figure 3 For box plots showing the lifespan distribution of five CT tube models, please refer to [link / reference]. Figure 3 Significant differences exist in the lifespan distribution among different CT tube models. Models B and C have wider lifespan ranges and larger overall distribution areas; Model A's lifespan is concentrated in the lower range; Models D and E have relatively moderate distributions. A certain degree of dispersion and long-tail characteristics can be observed in all models, with scatter points outside the box plot reflecting that the lifespans of some individuals deviate from the main distribution range.

[0096] Figure 4 For the probability density function (PDF) estimation curves of the lifetime of five CT tube models, please refer to [link / reference]. Figure 4The density curves of different models differ in peak position, width, and shape. Model A exhibits a narrow, low-lying unimodal distribution; Model B shows a more dispersed density distribution; Model C's density peak is located in the mid-to-high lifetime range; and the density curves of Models D and E show some fluctuations in shape. Overall, the lifetime distribution of each model exhibits asymmetry and cross-model differences, and the PDF curves further reveal the distribution characteristics of the lifetime samples in different ranges.

[0097] The failure prediction model in this embodiment can be trained and validated in a standard deep learning computing environment. The entire process, including feature engineering, neural network modeling, constraint handling, and index calculation, is implemented based on the Python deep learning framework.

[0098] The model input features include six dimensions: The system includes a classification identifier, ModelID, representing the X-ray tube model. ModelID can be an integer from 1 to 5, corresponding to X-ray tube models A, B, C, D, and E, respectively. All continuous features are Z-score standardized, meaning they are normalized using the mean and standard deviation of the training set to ensure that each dimension is comparable in numerical scale.

[0099] The experiment evaluated three metrics: root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination. Among them, RMSE is used to measure the absolute deviation between the predicted failure probability output by the failure prediction model and the failure probability label calculated by the median rank method. The smaller the value, the higher the overall fitting accuracy of the model to the empirical distribution. MAPE reflects the relative deviation of the prediction error at different probability levels, presented as a percentage, which is convenient for comparing prediction stability across intervals. This is used to characterize the model output's ability to explain the variability of the failure probability label. Its value ranges from 0 to 1. The closer it is to 1, the better the model fit and the stronger its ability to restore the true distribution trend.

[0100] Table 1 shows a comparison of the prediction performance of the five types of CT tubes under the Weibull model and the failure prediction model (DNN) provided in the embodiments of this application. Please refer to Table 1.

[0101] Table 1

[0102] As shown in Table 1, the failure prediction model (DNN) in this embodiment significantly outperforms the traditional Weibull distribution model across all five CT tube models. Taking the A-type tube as an example, the RMSE of the DNN is 0.0135, far lower than the 0.0685 of the Weibull model; the MAPE is 3.68%, a decrease of more than twenty percentage points compared to the 24.83% of the Weibull model. It reached 0.9978, which is significantly higher than the 0.9430 of the Weibull model.

[0103] The remaining models also showed a consistent trend: the RMSE of the DNN model for the B-type X-ray tube was 0.0199, lower than the 0.0327 of the Weibull model; the RMSE of the DNN model for the C-type X-ray tube was 0.0292, only about one-third of the 0.0903 of the Weibull model; the DNN models for the D-type and E-type X-ray tubes both maintained lower RMSE and MAPE, and all models of X-ray tubes... All scores were above 0.95. Overall, DNN outperformed DNN in prediction across all five CT tube models, indicating that this method has a stronger ability to adapt to the complexity, asymmetry, and long-tail characteristics exhibited by the lifespan data of different CT tube models.

[0104] Furthermore, from the perspective of curve fitting accuracy, the X-ray tube life curve obtained by the method provided in this application embodiment has significantly higher fitting accuracy than that of the Weibull model.

[0105] Figure 5 A schematic diagram comparing the fitting curves corresponding to model A X-ray tube. Figure 6 A schematic diagram comparing the fitting curves corresponding to model B X-ray tubes. Figure 7 A schematic diagram comparing the fitting curves corresponding to the C-type X-ray tube. Figure 8 A schematic diagram comparing the fitting curves corresponding to model D X-ray tube. Figure 9 This is a comparative diagram of the fitted curves for the E-type X-ray tube. The horizontal axis represents the cumulative exposure time (the diagram shows the cumulative exposure times in seconds), and the vertical axis represents the failure probability (the diagram shows the cumulative failure probability). Please refer to... Figures 5-9 This allows us to visually see the fit between the tube life curve fitted by the predicted failure probability output by the DNN model and the curve formed by the failure probability label calculated by the median rank method.

[0106] For the five CT tube models (A, B, C, D, and E), the tube lifespan curves generated by the DNN model consistently follow the overall trend of the empirical cumulative failure probability point cloud obtained through the median-rank method throughout the entire lifespan. Particularly in the medium and high probability ranges, it accurately reflects the local curvature and slope changes of the empirical curve. In contrast, the tube lifespan curves generated by the Weibull model generally exhibit significant deviations at both the low and high lifespan ends, with varying degrees and timings of deviation for different CT tube models. This indicates that the Weibull model, limited by its fixed parameterized function form, struggles to flexibly adapt to the diverse degradation behaviors exhibited by different CT tube models.

[0107] also, Figure 10 This is a schematic diagram illustrating the prediction accuracy of the Weibull method. Figure 11 For a schematic diagram illustrating the prediction accuracy of the CT tube life prediction method provided in this application embodiment, please refer to [link / reference]. Figures 10-11 As can be seen, the scatter distribution between the failure probability obtained by the Weibull model and the empirical cumulative failure probability is relatively discrete compared to the ideal reference line y=x, with systematic deviations in both low-probability and high-probability regions. In contrast, the scatter distribution between the failure probability obtained by the DNN model and the empirical cumulative failure probability is highly concentrated near the diagonal y=x, exhibiting a narrow band distribution with only minor fluctuations in a few local areas. This indicates that the DNN model maintains good predictive stability and calibration capability across the entire probability range.

[0108] Clearly, the DNN model not only outperforms the Weibull model in terms of numerical error metrics, but also demonstrates a significant advantage in the rationality of the predicted curve's shape and overall reliability, providing a solid foundation for subsequent reliability index calculations based on this distribution.

[0109] Finally, based on the X-ray tube life curve output by the DNN model, multiple reliability indicators can be obtained, including but not limited to mean time between failures, median life, B10 life, B90 life, and life range.

[0110] Mean time between failures (MTBF) reflects the average time from when a CT tube is put into operation until its first failure. In this embodiment, the mean time between failures... According to the formula The calculation yielded that, It is a reliability function, and , The discrete predicted values ​​output by the failure prediction model can be numerically approximated using the trapezoidal integral method in actual calculations. A larger mean time between failures (MTBF) indicates a higher overall lifespan for this type of CT tube.

[0111] Median lifetime refers to the cumulative duration of time when the failure probability output by the failure prediction model is 0.5. Characterized as This means that 50% of CT tubes will fail before this time point. Median lifetime can stably characterize the central trend of the lifetime distribution of this type of CT tube.

[0112] B10 lifetime refers to the cumulative distribution duration when the failure probability output by the failure prediction model is 0.1. Characterized as This means that 10% of CT tubes will fail before this point in time. B10 life is a key indicator with clear decision-making significance in engineering practice, and is often used as the technical lower limit for setting the warranty period of CT tubes, planning the quantity of spare parts inventory, and determining the time for the first preventive maintenance.

[0113] B90 lifetime refers to the cumulative distribution duration when the failure probability output by the failure prediction model is 0.9. Characterized as This means that 90% of CT tubes will fail before this point in time. The B90 lifetime profile focuses on the upper tail of the CT tube lifetime distribution, reflecting the longest service time that the vast majority of CT tubes can achieve. It is an important reference for assessing lifetime dispersion and developing high reliability assurance strategies.

[0114] Lifespan This refers to the difference between B90 lifetime and B10 lifetime, characterized as This metric describes the degree of dispersion in the lifespan performance of individual CT tubes of the same model. A larger lifespan range indicates a more significant difference in lifespan between individual CT tubes of that model. This means that there are both early-failure and long-serving CT tubes of that model, resulting in poor overall lifespan consistency and greater uncertainty in hospital equipment management. Conversely, a smaller lifespan range indicates a more concentrated lifespan distribution for CT tubes of that model, better performance stability, and easier development of a unified maintenance and spare parts strategy.

[0115] In this embodiment, Table 2 shows the reliability indicators corresponding to each CT tube model. Please refer to Table 2.

[0116] Table 2

[0117] Table 2 shows that there are stratified differences in lifespan characteristics among different CT tube models. Specifically, the B-type tube has the highest mean time between failures (MTBF) at 577,335.3 seconds, a median lifespan of 734,195.3 seconds, and a B90 lifespan of 1,179,971 seconds. The C-type tube follows closely behind, with an MTBF of 484,532.1 seconds, a median lifespan of 608,470.2 seconds, and a B90 lifespan of 864,332.7 seconds. In contrast, the A-type tube has an MTBF of 189,269.1 seconds, a median lifespan of 178,695.8 seconds, and a B90 lifespan of 331,225.6 seconds. While the E-type tube has a B90 lifespan of 1,700,084 seconds, its B10 lifespan is only 3,760.447 seconds, resulting in a lifespan range of 1,696,324 seconds, reflecting extreme individual variation in lifespan. Model D X-ray tube has moderate performance across the board, with a mean time between failures (MTBF) of 308,497.6 seconds and a lifespan of 568,317.1 seconds.

[0118] The aforementioned differences in lifespan characteristics directly support the formulation of differentiated equipment management strategies: for model B and model C X-ray tubes, the preventive maintenance cycle can be appropriately extended and the level of regular spare parts inventory can be maintained; for model D X-ray tubes, a conventional periodic operation and maintenance plan can be adopted; while for model A and model E X-ray tubes, it is necessary to strengthen early operation status monitoring, shorten inspection intervals, and increase the proportion of spare parts reserves. In particular, for model E X-ray tubes, an additional safety stock should be set up to cope with the risk of sudden replacement caused by the high uncertainty of lifespan.

[0119] To perform the corresponding steps in the above embodiments and various possible methods, an implementation of a CT tube life prediction device is given below. Optionally, the CT tube life prediction device can adopt the above-described... Figure 1 The server's device structure is shown. For further details, please refer to... Figure 12 , Figure 12 This is a functional block diagram of a CT tube life prediction device provided in an embodiment of this application. It should be noted that the basic principle and technical effects of the CT tube life prediction device provided in this embodiment are the same as those in the above embodiments. For the sake of brevity, any parts not mentioned in this embodiment can be referred to the corresponding content in the above embodiments. The CT tube life prediction device includes: an acquisition module and a determination module.

[0120] This acquisition module is used to obtain the current cumulative exposure time of the CT tube in any CT device and the target tube life curve corresponding to the CT tube.

[0121] Understandably, this acquisition module can be used to perform the above step S20.

[0122] The determination module is used to determine the current failure probability of the CT tube based on the target tube life curve and the current cumulative exposure time.

[0123] The determination module is also used to issue a failure alarm message if the current failure probability exceeds the preset failure threshold corresponding to the CT tube.

[0124] Understandably, this determining module can be used to perform the above steps S21 to S22.

[0125] Optionally, the CT tube life prediction device also includes a training module and a generation module.

[0126] This training module is used to acquire multiple sample lifetime data corresponding to each CT tube type, generate multiple failure probability labels based on the multiple sample lifetime data, and input the failure probability labels and sample lifetime data into a pre-built failure prediction model for training to obtain a trained failure prediction model.

[0127] This generation module is used to obtain multiple cumulative exposure times corresponding to CT tube types, input each cumulative exposure time into the trained failure prediction probability for processing, and obtain the first failure probability corresponding to each cumulative exposure time; perform range constraints and order-preserving regression on each first failure probability to obtain the failure probability corresponding to each cumulative exposure time; and perform curve fitting based on each cumulative exposure time and the failure probability corresponding to the cumulative exposure time to obtain the tube life curve corresponding to the CT tube type.

[0128] Optionally, the training module is also used to sort the lifetime data of each sample according to their numerical values ​​to obtain the X-ray tube lifetime sequence; and to calculate the failure probability label corresponding to each lifetime data of the sample based on the rank of the lifetime data in the X-ray tube lifetime sequence and the number of lifetime data of the sample.

[0129] Optionally, the generation module is further configured to constrain the range of each first failure probability according to a preset value range so that each first failure probability is within the value range to obtain a second failure probability; and to monotonically smooth each second failure probability according to a preset prediction trend so that the change trend of the second failure probability conforms to the prediction trend to obtain the predicted failure probability corresponding to each cumulative exposure duration.

[0130] Optionally, this training module is also used to construct feature vectors corresponding to the lifespan data of each sample based on the CT tube type; input the feature vectors and failure probability labels corresponding to the lifespan data of each sample into a pre-built failure prediction model for training, thereby obtaining a trained failure prediction model; the feature vectors are... ,in, For feature vectors, For sample lifetime data, The CT tube type ID is used; the failure prediction model includes an input layer, multiple fully connected layers, and an output layer, with the number of neurons in each fully connected layer decreasing sequentially with the data transmission direction.

[0131] Optionally, the above modules can be stored in the form of software or firmware. Figure 1 The memory shown is either stored in or embedded in the server's operating system (OS), and can be used by... Figure 1 The processor executes the commands. Meanwhile, the data and program code required to execute these modules can be stored in memory.

[0132] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the CT tube life prediction method provided in this application.

[0133] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0134] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0135] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a computer-readable storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned computer-readable storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0136] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for predicting the lifespan of a CT tube, characterized in that, An application is made to a server that communicates with multiple CT devices. The server stores X-ray tube lifespan curves corresponding to each CT tube type. These lifespan curves characterize the relationship between the cumulative exposure time and the failure probability of the CT tube. The lifespan curves are generated based on multiple cumulative exposure times and the predicted failure probabilities corresponding to each cumulative exposure time. The predicted failure probabilities are generated by inputting the cumulative exposure times into a pre-generated failure prediction model. This failure prediction model is trained based on multiple sample lifespan data and the failure probability labels corresponding to the sample lifespan data. The method includes: Obtain the current cumulative exposure time of the CT tube in any of the CT devices and the target tube life curve corresponding to the CT tube; Based on the target X-ray tube life curve, the current failure probability of the CT tube is determined according to the current cumulative exposure time; If the current failure probability exceeds the preset failure threshold corresponding to the CT tube, a failure alarm message is issued. The X-ray tube life curve is obtained through the following steps: For each of the CT tube types, multiple sample lifetime data corresponding to the tube type are obtained. Multiple failure probability labels are generated based on the multiple sample lifetime data. The failure probability labels and the sample lifetime data are input into a pre-built failure prediction model for training to obtain a trained failure prediction model. Multiple cumulative exposure times corresponding to the CT tube type are obtained, and each cumulative exposure time is input into the trained failure prediction model for processing to obtain the first failure probability corresponding to each cumulative exposure time. By performing range constraints and order-preserving regression on each of the first failure probabilities, the predicted failure probabilities corresponding to each of the cumulative exposure durations are obtained. Based on the cumulative exposure time and the predicted failure probability corresponding to the cumulative exposure time, curve fitting is performed to obtain the X-ray tube life curve corresponding to the CT tube type. The step of inputting the failure probability label and the sample lifetime data into a pre-built failure prediction model for training to obtain a trained failure prediction model includes: Construct feature vectors corresponding to the lifetime data of each sample based on the CT tube type; The feature vector and failure probability label corresponding to each of the sample lifetime data are input into the pre-built failure prediction model for training, and the trained failure prediction model is obtained. The feature vector is ,in, For the feature vector, The sample lifetime data, The CT tube type ID is used; the failure prediction model includes an input layer, multiple fully connected layers, and an output layer, and the number of neurons in each fully connected layer decreases sequentially with the data transmission direction.

2. The method according to claim 1, characterized in that, The step of generating multiple failure probability labels based on multiple sample lifetime data includes: The lifetime data of each sample are sorted according to their numerical values ​​to obtain the X-ray tube lifetime sequence; For each sample lifetime data in the X-ray tube lifetime sequence, the failure probability label corresponding to the sample lifetime data is calculated based on the rank of the sample lifetime data in the X-ray tube lifetime sequence and the number of sample lifetime data.

3. The method according to claim 1, characterized in that, The step of performing range constraints and order-preserving regression on each of the first failure probabilities to obtain the predicted failure probability corresponding to each of the cumulative exposure durations includes: The first failure probability is constrained according to a preset value range so that each first failure probability is within the value range, thereby obtaining the second failure probability. The second failure probabilities are monotonically smoothed according to the preset prediction trend so that the change trend of the second failure probabilities conforms to the prediction trend, thereby obtaining the predicted failure probabilities corresponding to each cumulative exposure duration.

4. A CT tube life prediction device, characterized in that, An application is provided in a server that communicates with multiple CT devices. The server stores X-ray tube lifespan curves corresponding to each CT tube type. These lifespan curves characterize the relationship between the cumulative exposure time and the failure probability of the CT tube. The lifespan curves are generated based on multiple cumulative exposure times and the predicted failure probabilities corresponding to each cumulative exposure time. The predicted failure probabilities are generated by inputting the cumulative exposure times into a pre-generated failure prediction model. This failure prediction model is trained based on multiple sample lifespan data and the failure probability labels corresponding to the sample lifespan data. The device includes: The acquisition module is used to acquire the current cumulative exposure time of the CT tube in any of the CT devices and the target tube life curve corresponding to the CT tube; The determination module is used to determine the current failure probability of the CT tube based on the target tube life curve and the current cumulative exposure time. The determining module is further configured to issue a failure alarm message if the current failure probability exceeds the preset failure threshold corresponding to the CT tube. The device further includes: The training module is used to acquire multiple sample lifetime data corresponding to each of the CT tube types, generate multiple failure probability labels based on the multiple sample lifetime data, and input the failure probability labels and the sample lifetime data into a pre-built failure prediction model for training to obtain a trained failure prediction model. The generation module is used to obtain multiple cumulative exposure times corresponding to the CT tube type, input each cumulative exposure time into a trained failure prediction model for processing, and obtain a first failure probability corresponding to each cumulative exposure time; perform range constraints and order-preserving regression on each first failure probability to obtain a failure probability corresponding to each cumulative exposure time; and perform curve fitting based on each cumulative exposure time and the failure probability corresponding to the cumulative exposure time to obtain the tube life curve corresponding to the CT tube type. The training module is further configured to construct feature vectors corresponding to the lifetime data of each sample based on the CT tube type; input the feature vectors and failure probability labels corresponding to the lifetime data of each sample into a pre-constructed failure prediction model for training, thereby obtaining a trained failure prediction model; the feature vectors are... ,in, For the feature vector, The sample lifetime data, The CT tube type ID is used; the failure prediction model includes an input layer, multiple fully connected layers, and an output layer, and the number of neurons in each fully connected layer decreases sequentially with the data transmission direction.

5. The apparatus according to claim 4, characterized in that, The training module is further configured to sort the lifetime data of each sample according to their numerical values ​​to obtain a tube lifetime sequence; and to calculate the failure probability label corresponding to each lifetime data in the tube lifetime sequence based on the rank of the lifetime data in the tube lifetime sequence and the number of lifetime data.

6. A server, characterized in that, It includes a processor and a memory, the memory storing a computer program that can be executed by the processor to implement the CT tube life prediction method according to any one of claims 1-3.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the CT tube life prediction method as described in any one of claims 1-3.

Images