Result interpretation method and device of intelligent medical diagnosis system, storage medium and computer program product

Abstract

The application discloses a result interpretation method and device of an intelligent medical diagnosis system, a storage medium and a computer program product, relates to the technical field of intelligent medical treatment, and comprises the following steps: acquiring patient clinical data; inputting the patient clinical data into a pre-constructed cost-sensitive decision tree regularization neural network model; performing prediction processing on the patient clinical data through a neural network part in the cost-sensitive decision tree regularization neural network model to obtain a prediction result; and performing interpretation processing on the patient clinical data and the prediction result through a cost-sensitive decision tree regularization part to obtain an interpretation result. By introducing the concept of cost sensitivity, constructing a cost-sensitive decision tree, adding the cost-sensitive decision tree in the regularization neural network model, improving the generalization ability of the model, and making the intelligent medical diagnosis system maintain stable interpretation performance, the application compensates for the disadvantage that current intelligent medical diagnosis systems lack interpretation capability.

Description

Technical Field

[0001] This application relates to the field of intelligent medical technology, and in particular to the result interpretation methods, devices, storage media, and computer program products of intelligent medical diagnostic systems. Background Technology

[0002] With the development of computer hardware, computer technologies, led by machine learning, have demonstrated powerful data processing capabilities. Artificial intelligence methods, represented by deep learning, are bringing new changes to the medical industry. Intelligent medical diagnostic systems combining deep learning and other AI methods can assist doctors in diagnosis. For example, intelligent medical diagnostic systems with powerful data processing capabilities can assist in predicting preeclampsia (PE), a condition with an unclear pathogenesis and a high mortality rate. However, because current intelligent medical diagnostic systems generally lack the ability to interpret diagnostic results, they either do not interpret the results from the perspective of human expert decision-making or simply lack the ability to provide professional interpretations. Consequently, the diagnostic predictions assisted by intelligent medical diagnostic systems are not accepted or trusted.

[0003] Therefore, how to overcome the shortcomings of intelligent medical diagnostic systems in lacking interpretability has become an urgent technical problem to be solved.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this application is to provide a method, device, storage medium, and computer program product for interpreting the results of an intelligent medical diagnostic system, aiming to address the technical problem of compensating for the lack of interpretability in intelligent medical diagnostic systems.

[0006] To achieve the above objectives, this application proposes a method for interpreting the results of an intelligent medical diagnostic system, the method comprising:

[0007] Obtain patient clinical data;

[0008] The patient's clinical data is input into a pre-constructed cost-sensitive decision tree regularized neural network model;

[0009] The patient's clinical data is predicted by the neural network part of the cost-sensitive decision tree regularized neural network model, and the prediction result is obtained by the cost-sensitive decision tree regularized part. The interpretation result is obtained by the interpretation of the patient's clinical data and the prediction result.

[0010] In one embodiment, before the step of inputting the patient's clinical data into a pre-constructed cost-sensitive decision tree regularized neural network model, performing prediction processing through the neural network part of the cost-sensitive decision tree regularized neural network model to obtain a prediction result, and the step of the cost-sensitive decision tree regularization part interpreting the patient's clinical data and prediction result to obtain an interpretation result, the method further includes:

[0011] Construct a tree-regularized neural network model based on the pre-collected training dataset;

[0012] A cost-sensitive decision tree is constructed based on the training dataset and the tree-regularized neural network model;

[0013] The cost-sensitive decision tree is added to the tree-regularized neural network model to obtain the cost-sensitive decision tree-regularized neural network model.

[0014] In one embodiment, the step of constructing a tree-regularized neural network model based on a pre-collected training dataset includes:

[0015] Obtain the pre-collected training dataset and deep learning model framework;

[0016] The training dataset is input into the deep learning model framework to obtain the neural network prediction results;

[0017] Construct a binary classification decision tree based on the neural network prediction results and the training dataset;

[0018] The regularization term is calculated based on the binary classification decision tree and the deep learning model framework.

[0019] The regularization term is added to the deep learning model for regularization processing to obtain a tree-regularized neural network model.

[0020] In one embodiment, after the step of adding the regularization term to the deep learning model for regularization processing to obtain a tree-regularized neural network model, the method further includes:

[0021] When the system has not reached the iteration limit, the training data is input into the tree regularized neural network model to obtain a new neural network prediction result, and the execution step is returned: construct a binary classification decision tree based on the neural network prediction result and the training dataset.

[0022] In one embodiment, the step of constructing a cost-sensitive decision tree based on the training dataset and the tree-regularized neural network model includes:

[0023] The feature contribution is obtained based on the training dataset and the tree regularized neural network model;

[0024] A cost-sensitive decision tree is constructed based on the contribution of the aforementioned features.

[0025] In one embodiment, the step of adding the cost-sensitive decision tree to the tree-regularized neural network model to obtain the cost-sensitive decision tree-regularized neural network model includes:

[0026] The tree regularization term is calculated based on the cost-sensitive decision tree;

[0027] Adding the tree regularization term to the tree regularization neural network model yields a cost-sensitive decision tree regularization neural network model.

[0028] In one embodiment, after the step of adding the tree regularization term to the tree regularized neural network model to obtain a cost-sensitive decision tree regularized neural network model, the method further includes:

[0029] If the system does not reach the iteration limit, return to the execution step: obtain the feature contribution based on the training dataset and the tree regularized neural network model.

[0030] Furthermore, to achieve the above objectives, this application also proposes a result interpretation device for an intelligent medical diagnostic system, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the result interpretation method for the intelligent medical diagnostic system as described above.

[0031] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the result interpretation method of the intelligent medical diagnostic system as described above.

[0032] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the result interpretation method of the intelligent medical diagnostic system as described above.

[0033] One or more technical solutions proposed in this application have at least the following technical effects:

[0034] By inputting patients' clinical data into a pre-constructed cost-sensitive decision tree regularized neural network model for prediction and interpretation, not only are predictions obtained, but also interpretations from the perspective of medical experts. By configuring a cost-sensitive decision tree regularized neural network model into an intelligent medical diagnostic system, the system can not only generate predictions but also provide detailed explanations, revealing key factors influencing diagnostic results and their interactions. Introducing the concept of cost sensitivity aligns the interpretations with expert experience, further enhancing the interpretability of the intelligent medical diagnostic system. The application of regularization techniques in the cost-sensitive decision tree regularized neural network model reduces overfitting on training data and improves generalization ability on unseen data, maintaining stable interpretability and compensating for the lack of interpretability inherent in intelligent medical diagnostic systems. Attached Figure Description

[0035] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart illustrating the first embodiment of the result interpretation method for the intelligent medical diagnostic system of this application.

[0038] Figure 2 This is a flowchart illustrating the second embodiment of the result interpretation method for the intelligent medical diagnostic system of this application.

[0039] Figure 3 This is a flowchart illustrating the third embodiment of the result interpretation method for the intelligent medical diagnostic system of this application.

[0040] Figure 4 A flowchart illustrating the overall training process for constructing a tree-regularized neural network model;

[0041] Figure 5 A flowchart illustrating the overall training process for constructing a cost-sensitive decision tree regularized neural network model;

[0042] Figure 6 This is a schematic diagram of the module structure of the result interpretation device of the intelligent medical diagnostic system according to an embodiment of this application;

[0043] Figure 7 This is a schematic diagram of the hardware operating environment involved in the result interpretation method of the intelligent medical diagnostic system in this application embodiment.

[0044] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0045] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0046] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0047] The main solution of this application is to pre-construct a cost-sensitive decision tree regularized neural network model in the intelligent medical diagnostic system, and then use this model to interpret the prediction results of the intelligent medical diagnostic system, obtaining interpretations that align with the experience of medical experts and are trustworthy. Furthermore, by adding a cost-sensitive decision tree to the tree regularized neural network model, a more concise tree-structured interpretation graph that better reflects expert knowledge can be generated, overcoming the current shortcomings of intelligent medical diagnostic systems in lacking interpretability.

[0048] This embodiment considers that relying solely on clinical features to diagnose certain diseases with complex definitions and ambiguous pathogenesis can increase the likelihood of misdiagnosis. For example, preeclampsia (PE) is a pregnancy syndrome, one of the most common medical complications during pregnancy, and a leading cause of maternal mortality. However, PE often presents with subtle clinical features, and there are no effective treatments other than induced labor. Intelligent medical diagnostic systems, with their powerful data processing capabilities, can assist doctors in diagnosing and predicting diseases including PE. However, most current intelligent medical diagnostic systems lack interpretability for their predictions. In systems that do offer interpretability, the models used only explain the prediction results by calculating the contribution of model features. For instance, intelligent medical diagnostic systems using the SHAP method to interpret prediction results, while the feature contribution of the SHAP method can represent the importance of different features in the prediction to some extent, thus explaining the focus of the model's prediction, do not explain the prediction process itself; their interpretation of the model remains data-driven. Interpretable methods for calculating the feature contribution of computational models, such as the SHAP method, do not offer explanations from the perspective of human expert decision-making. This leads to the predictions of intelligent medical diagnostic systems being unacceptable and untrusted by medical experts. This phenomenon demonstrates that current medical diagnostic systems lack interpretability and struggle to provide robust explanations for their predictions, making them difficult for the public to trust.

[0049] Therefore, this application provides a solution by adding a cost-sensitive decision tree to a tree-regularized neural network model, thus constructing a cost-sensitive decision tree regularized neural network model. This model interprets the prediction results of the intelligent medical diagnostic system from the perspective of human expert decision-making, making the predictions acceptable and trustworthy to medical experts and overcoming the current shortcomings of intelligent medical diagnostic systems in lacking interpretability.

[0050] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or intelligent medical diagnostic system capable of performing the above functions. The following description uses an intelligent medical diagnostic system as an example to illustrate this embodiment and the subsequent embodiments.

[0051] Based on this, embodiments of this application provide a method for interpreting the results of an intelligent medical diagnostic system, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the result interpretation method of the intelligent medical diagnostic system of this application.

[0052] In this embodiment, the result interpretation method of the intelligent medical diagnostic system includes steps S10 to S20:

[0053] Step S10: Obtain patient clinical data;

[0054] Intelligent medical diagnostic systems are medical aids that utilize artificial intelligence technology to achieve rapid and accurate diagnosis of patients' diseases and provide treatment suggestions. These systems typically rely on technologies such as big data analysis, machine learning, and deep learning to extract correlations and patterns from massive amounts of medical clinical data, establishing connections between diseases and symptoms / signs.

[0055] Acquiring patient clinical data is a prerequisite for intelligent medical diagnostic systems to generate predictive results and corresponding interpretations. Sensors can monitor patients' physiological parameters and disease indicators in real time, and the collected data can be wirelessly transmitted to the cloud for further analysis and prediction by the intelligent medical diagnostic system.

[0056] Step S20: Input the patient's clinical data into a pre-constructed cost-sensitive decision tree regularized neural network model;

[0057] Cost-sensitive decision trees are decision tree algorithms that consider the costs of misclassification for different categories. By introducing a cost matrix, the tree construction process is adjusted to account for the misclassification costs of different categories during classification. By incorporating physician expert knowledge as a cost, features deemed important by physicians are prioritized for input into the model while ensuring accurate predictions, generating interpretations that align with physician experience.

[0058] Tree-regularized neural network models combine decision trees and regularized neural networks. Regularized neural networks introduce regularization terms into neural networks to prevent overfitting. Regularization terms typically include L1 and L2 regularization, which penalize model complexity by adding the sum of the absolute values ​​of the weights or the sum of the squares of the weights to the loss function, encouraging the model to learn simpler, more generalizable features. Tree-regularized neural network models not only combine the intuitive interpretability of decision trees with the powerful learning capabilities of neural networks, but also utilize regularization to prevent overfitting.

[0059] The cost-sensitive decision tree regularized neural network model is a further development of the tree regularized neural network model by introducing the concept of cost-sensitive decision trees. This means that the model, during its construction, not only considers regularization to prevent overfitting, but also takes into account the cost of misclassification for different categories in order to optimize classification performance.

[0060] Understandably, while the tree regularized neural network model combines decision trees and regularized neural networks, the decision trees in this model are just ordinary decision tree algorithms and do not consider the cost of misclassification of different categories. It is necessary to further add cost-sensitive decision tree algorithms to this model, and introduce medical expert knowledge as a cost. Only then can the final model interpret the prediction results from the perspective of medical experts, so that the prediction results can be accepted and trusted, and the shortcomings of the current intelligent medical diagnosis system lacking interpretability can be made up for.

[0061] Step S30: The patient's clinical data is predicted using the neural network component of the cost-sensitive decision tree regularized neural network model to obtain a prediction result; the cost-sensitive decision tree regularized component interprets the patient's clinical data and the prediction result to obtain an interpretation result.

[0062] Predictive results refer to the automated output of intelligent medical diagnostic systems based on input medical data, after a series of technical analyses, providing information about a patient's disease status, risk predictions, or treatment recommendations. Predictive results include the type of disease, the degree of lesion, its development trend, and treatment suggestions, aiming to assist doctors in making more accurate diagnostic and treatment decisions.

[0063] Patient clinical data is used as input and fed into a cost-sensitive decision tree regularized neural network model. The neural network part of the model analyzes the patient clinical data to obtain prediction results. The cost-sensitive decision tree regularization part of the model interprets the prediction results to provide an interpretation that aligns with the decision-making perspective of medical experts.

[0064] Specifically, as a feasible implementation, step S30 may include: First, patient clinical data is passed as input data to the tree-regularized neural network model part of the cost-sensitive decision tree regularized neural network model to obtain prediction results. Further, the cost-sensitive decision tree regularized neural network model analyzes the prediction results using a cost-sensitive decision tree. Based on a predefined cost matrix, considering the misclassification costs of different categories, the most suitable classification path is selected, and an explanation result is generated. The explanation result is a tree-structured explanation diagram, typically involving an explanation of the model's decision-making process, such as which features have a significant impact on the classification results, and the prediction probabilities of different categories. These explanation results can help doctors better understand the model's decision-making basis, improving the reliability and interpretability of the prediction results.

[0065] The above is only one feasible implementation of step S30 provided in this embodiment. This embodiment does not specifically limit the specific implementation of step S30.

[0066] This embodiment provides a method for interpreting the results of an intelligent medical diagnostic system. By inputting patient clinical data into a pre-constructed cost-sensitive decision tree regularized neural network model for prediction and interpretation, it not only obtains the predicted results but also interprets these results from the perspective of medical experts. By configuring the cost-sensitive decision tree regularized neural network model into the intelligent medical diagnostic system, the system can not only generate predicted results but also provide detailed interpretations, revealing the key factors influencing the diagnostic results and the interactions between these factors. By introducing the concept of cost sensitivity, the interpretation results are aligned with expert experience, thereby further improving the interpretive capability of the intelligent medical diagnostic system. The application of regularization techniques in the cost-sensitive decision tree regularized neural network model reduces overfitting on training data and improves the model's generalization ability on unseen data, enabling the intelligent medical diagnostic system to maintain stable interpretive performance and compensating for the lack of interpretive capability in intelligent medical diagnostic systems.

[0067] Based on the first embodiment of this application, a second embodiment of this application is proposed. In the second embodiment of this application, content that is the same as or similar to that in the first embodiment described above can be referred to the above description and will not be repeated hereafter.

[0068] Based on this, please refer to Figure 2 , Figure 2 This is a flowchart illustrating the second embodiment of the result interpretation method of the intelligent medical diagnostic system of this application.

[0069] In this embodiment, steps S121 to S123 are included before step S20, which involves inputting the patient's clinical data into a pre-constructed cost-sensitive decision tree regularized neural network model:

[0070] Step S121: Construct a tree-regularized neural network model based on the pre-collected training dataset;

[0071] Pre-collected training datasets typically contain a wealth of historical clinical data with rich information and features. By constructing tree-regularized neural network models, the complex structures and patterns within this data can be effectively utilized. Tree-regularized neural network models can learn deep-seated relationships within the data and continuously optimize their structure during training to better adapt to the characteristics of the data.

[0072] Tree-regularized neural network models refer to the use of decision tree structures as a regularization method to improve the generalization ability of neural network models and prevent overfitting.

[0073] Building tree-regularized neural network models based on pre-collected training datasets allows for full utilization of the data's characteristics and advantages, improving the model's generalization ability and performance. This is of great significance for handling complex medical problems, assisting doctors in making more accurate and reliable decisions.

[0074] Specifically, in one feasible implementation, step S121 may include:

[0075] First, prepare the data needed to construct the tree-regularized neural network model.

[0076] Preprocess the pre-collected training dataset, including data cleaning, missing value handling, feature selection or extraction, and data normalization, to ensure that the quality and format of the dataset are suitable for model training.

[0077] Furthermore, a basic neural network structure is designed to build a tree-regularized neural network model.

[0078] Designing a basic neural network architecture involves determining the number of layers, the number of neurons in each layer, and the type of activation function. Choosing a network architecture suitable for the task, such as a fully connected network, convolutional neural network, or recurrent neural network, depends on the type of data and the nature of the problem.

[0079] Furthermore, a tree regularization term is introduced during the training process of the neural network.

[0080] Tree regularization is typically achieved by adding a regularization term related to the neural network weights to the loss function. This regularization term can be defined based on certain characteristics of the decision tree, such as its splitting criterion or its complexity. For example, the splitting criterion of a decision tree can be transformed into a constraint on the neural network weights. By defining a regularization term, sparsity in the network weights near the decision boundary can be encouraged, thus introducing tree regularization into the neural network architecture. Another example is the introduction of tree regularization by combining the output of the neural network with the structure of the decision tree. The neural network learns complex feature representations, and the decision tree uses these representations for final classification or regression.

[0081] Furthermore, the trained tree regularized neural network model is validated and evaluated.

[0082] The trained model is evaluated using a validation dataset to check if its performance meets expectations. Based on the validation results, the model's structure, parameters, or regularization terms can be adjusted to optimize its performance.

[0083] Finally, evaluate the model's performance on the test dataset to ensure it has good generalization ability. Once the model reaches a satisfactory performance level, it can be applied to real-world scenarios.

[0084] It should be noted that the specific implementation details and parameter settings need to be adjusted and optimized according to the specific task and dataset. Furthermore, due to the complexity of the model and the potentially high computational cost of the training process, it is necessary to utilize high-performance computing resources or distributed computing frameworks to accelerate the model training and evaluation process.

[0085] In this embodiment, a suitable neural network model framework is constructed based on the pre-collected dataset, the neural network model framework is trained using the dataset, and the trained model is validated and evaluated to ensure that the trained model has good generalization ability, prevent overfitting, and enable interpretation of prediction results, thus giving the intelligent medical diagnostic system the ability to interpret prediction results.

[0086] The above is only one feasible implementation of step S121 provided in this embodiment. This embodiment does not specifically limit the specific implementation of step S121.

[0087] Step S122: Construct a cost-sensitive decision tree based on the training dataset and the tree-regularized neural network model;

[0088] Although tree-regularized neural network models have the ability to interpret prediction results, their interpretation results are not consistent with expert experience. However, by adding cost-sensitive operations to the original tree-regularized neural network to construct a cost-sensitive decision tree, it is possible to interpret the prediction results from the perspective of medical experts.

[0089] The construction of a cost-sensitive decision tree requires defining a cost matrix based on the training dataset, incorporating the experience and knowledge of medical experts as costs into the cost matrix, and then constructing the cost-sensitive decision tree based on the cost matrix and the feature representations extracted from the tree-regularized neural network model.

[0090] First, a regularized neural network model is used to extract feature representations from the training dataset. The training data is fed into the model, and the outputs of the hidden layers or the last layer are obtained as feature representations. These feature representations capture the deep structure and patterns of the data, which are crucial for subsequently building cost-sensitive decision trees.

[0091] Furthermore, a cost matrix is ​​defined to quantify the cost of misclassification for different classes. The cost matrix is ​​a two-dimensional array where each element represents the cost of misclassifying a sample from one class to another. This matrix reflects the relative importance of misclassification for different classes and can be set according to the specific application scenario.

[0092] Furthermore, a cost-sensitive decision tree is constructed based on the feature representations and the cost matrix. This process is similar to constructing a standard decision tree, but it requires consideration of the misclassification costs in the cost matrix. At each node of the tree, the impact of different feature splits on the cost is evaluated, and the split that minimizes the total cost is selected. The optimal split is chosen by calculating the cost-weighted purity improvement for each split.

[0093] Furthermore, to prevent overfitting and improve the generalization ability of the cost-sensitive decision tree, pruning is necessary. Pruning involves deleting some subtrees or leaf nodes to simplify the tree structure and reduce the risk of overfitting. In addition, other optimization operations can be performed on the tree, including adjusting parameters such as tree depth and the minimum number of samples per leaf node.

[0094] Finally, the constructed cost-sensitive decision tree is evaluated using a validation dataset. By calculating performance metrics (such as accuracy and cost loss) on the validation dataset, the effectiveness and generalization ability of the model can be assessed. Based on the evaluation results, the model can be further adjusted and optimized.

[0095] Based on the training dataset and the tree-regularized neural network model, a cost-sensitive decision tree that considers misclassification is constructed. By introducing expert knowledge, the prediction results are interpreted from the perspective of expert decision-making, generating a more concise tree structure interpretation diagram that is more in line with expert knowledge, so that the prediction results are recognized and trusted.

[0096] Step S123: Add the cost-sensitive decision tree to the tree-regularized neural network model to obtain the cost-sensitive decision tree-regularized neural network model.

[0097] Adding cost-sensitive decision trees to a tree-regularized neural network model combines the advantages of decision tree and neural network algorithms to construct a cost-sensitive decision tree-regularized neural network model. The cost-sensitive decision tree-regularized neural network model is a high-level conceptual description that can interpret prediction results from an expert decision-making perspective, compensating for the lack of interpretability in intelligent medical diagnostic systems.

[0098] Adding cost-sensitive decision trees to a tree-regularized neural network model requires modifying the model's loss function. Inspired by cost-sensitive decision trees, the structure of the tree-regularized neural network is simplified through pruning strategies. Furthermore, the neural network is trained using a cost-sensitive loss function while simultaneously applying tree regularization. Finally, the optimal model configuration is found by adjusting several hyperparameters, such as the learning rate, regularization strength, and cost-sensitive weights.

[0099] It should be noted that the cost-sensitive decision tree regularized neural network model combines the advantages of two different model frameworks, which may make training the cost-sensitive decision tree regularized neural network model more complex and difficult. Therefore, in practical applications, sufficient experiments and validations are needed to ensure the effectiveness and practicality of the proposed model.

[0100] In this embodiment, a tree-regularized neural network model is constructed using a pre-collected training dataset, enabling the intelligent medical diagnostic system to interpret prediction results. To further address the current shortcomings of lacking interpretability in medical diagnostic systems by allowing the system to interpret predictions from the perspective of medical experts, a cost-sensitive decision tree is constructed and added to the tree-regularized neural network model, resulting in the final cost-sensitive decision tree-regularized neural network model. This model analyzes prediction results from an expert decision-making perspective, making the predictions of the intelligent medical diagnostic system trustworthy and acceptable, thus enhancing the usability of the predictions.

[0101] Based on the first and / or second embodiments of this application, a third embodiment of this application is proposed. In this third embodiment, content that is the same as or similar to that in embodiments one and two described above can be referred to the above description and will not be repeated hereafter.

[0102] Based on this, please refer to Figure 3 , Figure 3 This is a flowchart illustrating the third embodiment of the data cleaning method provided in this application.

[0103] In this embodiment, step S121, which constructs a tree-regularized neural network model based on the pre-collected training dataset, includes steps S1 to S5:

[0104] Step S1: Obtain the pre-collected training dataset and deep learning model framework;

[0105] To ensure the learning effectiveness and performance of the final model and improve the efficiency of training and deployment, a deep learning model framework is chosen as the basis for training the model. The pre-collected training datasets are typically pre-processed and labeled, making the model training process more efficient.

[0106] Deep learning model frameworks have multiple hidden layers, enabling them to learn and represent the complexity of data. A Multilayer Perceptron (MLP), on the other hand, has an input layer, one or more hidden layers, and an output layer, making it the simplest model framework. Since deep learning model frameworks have far more network layers than MLPs, the main difference between the two lies in the number of network layers. For ease of understanding, we will use the simplest MLP as an example to explain how to obtain a training dataset. Let a single sample input be x. n The corresponding real label is represented as y. n The training dataset is a combination of sample inputs and corresponding real labels, including the real feature data of the samples, such as clinical data like patients' height, weight, and blood indicators.

[0107] Step S2: Input the training dataset into the deep learning model framework to obtain the neural network prediction results;

[0108] The training dataset is input into the deep learning model framework to obtain the initial neural network prediction results. These neural network results are essentially predicted labels corresponding to samples in the pre-training dataset. The deep learning model framework transforms the sample data in the training dataset into predicted labels, i.e., the neural network prediction results. The true labels corresponding to the sample data in the training dataset can then be used to test and evaluate the completed model.

[0109] Since let the input of a single sample be x n The corresponding real label is represented as y. n The prediction function in the deep learning model framework is represented as in This represents the prediction result, i.e., the neural network prediction result, and the sample data x. n The predicted labels are given, and the vector W represents the model frame parameters.

[0110] Step S3: Construct a binary classification decision tree based on the neural network prediction results and the training dataset;

[0111] Using the neural network predictions as labels and the training dataset as input, a binary classification decision tree is constructed. Essentially, the neural network predictions output by the deep learning model framework are used as the training targets for the binary classification decision tree, and the tree is built based on the features of the training dataset.

[0112] Based on the prediction function in the deep learning model framework Using sample sets and the prediction set of the corresponding neural network net(W) Construct a proxy dataset and input it into a basic decision tree model framework to generate a binary classification decision tree that can accurately reproduce the given sample input and network prediction.

[0113] Step S4: Calculate the regularization term based on the binary classification decision tree and the deep learning model framework;

[0114] The average path length of the generated binary classification decision tree is calculated, and the average path length is used as the prediction target. The parameters in the deep learning model framework are used as inputs and fed into an independent MLP network for fitting. The MLP calculates the estimated value of the average path length based on the parameters of the deep learning model framework, and outputs the estimated value as a regularization term.

[0115] First, calculate the average path length of the generated binary classification decision tree.

[0116] The average path length is defined as the length of a sample input x. n The average number of decision tree nodes that must be traversed during prediction, assuming the function for calculating path length is Path Len(tree, x), then the formula for calculating the average path length is as follows:

[0117]

[0118] In the formula for calculating the average path length, the average path length is used as the output of the tree regularization function Ω(W), where N represents the data capacity of the samples contained in the training dataset, and n = 1.

[0119] Furthermore, the average path length is used as the prediction target, and the parameters in the deep learning model framework are used as inputs, which are then fed into a separate MLP network for fitting. The MLP calculates an estimate of the average path length based on the parameters of the deep learning model framework, and outputs the estimate as a regularization term.

[0120] Based on the above formula for calculating the average path length, the complete algorithm for the tree regularization function Ω(W) is as follows:

[0121]

[0122] It should be noted that Ψ(W) represents the tree regularization term, and λ represents the strength of the regularization term. The input is a dataset containing N samples. And predictions involving neural networks The output is the average path length.

[0123] Additionally, it's important to note that the tree regularization function Ω(W) is not differentiable. The process of building the decision tree in the implementation of the tree regularization function is non-differentiable, which means the tree regularization function Ω(W) is also non-differentiable with respect to the neural network parameters W. This makes it impossible to apply gradient descent for iterative training of the neural network after adding the tree regularization term. While derivative-free optimization techniques can be used, this would significantly increase the training cost.

[0124] Therefore, we choose to introduce proxy functions. Instead of the original tree regularization function Ω(W), a surrogate function is used to compute the tree regularization term. This surrogate function is implemented in a separate MLP and obtains the output of the tree regularization function Ω(W), i.e., the estimated average path length, by mapping the original neural network parameters W. Because of the surrogate function... It is implemented using a separate MLP, which is differentiable relative to the original neural network.

[0125] Let vector θ represent the parameters of the selected independent MLP network approximator, and ∈ be the regularization strength, with a value greater than 0. Then the surrogate function... The optimization can be achieved by minimizing the squared error loss with respect to the tree regularization function Ω(W), as shown in the following formula:

[0126]

[0127] It should be noted that the vector θ represents the parameters of the selected independent MLP network approximator, and ∈ is the regularization strength. Ω(W) is the surrogate function, and Ω(W) is the tree regularization function, j = 1.

[0128] The above formula can achieve the effect of training optimization, so that the estimated value of the average path length reaches the optimal solution, providing a foundation for subsequent model construction.

[0129] Step S5: Add the regularization term to the deep learning model for regularization processing to obtain a tree-regularized neural network model.

[0130] By integrating the calculation logic of the loss function and the tree regularization term into the deep learning model framework, and performing a training loop on the model framework so that the regularization term is calculated and applied in each iteration, a tree regularized neural network model is obtained.

[0131] Training the model framework involves passing input data to the model and obtaining its output. Further, a custom loss function is used to calculate the loss between the output and the true label, including data fitting loss and a tree regularization term. Then, the gradient is calculated based on the loss and propagated back to the model's parameters. Finally, an optimizer is used to update the model's parameters based on the gradient, thus constructing a tree-regularized neural network model.

[0132] In this embodiment, the training dataset is input into a deep learning model framework to obtain neural network prediction results. The neural network predictions are used as labels, and the training data is used as input to construct a binary classification decision tree. The average path length of the generated binary classification decision tree is calculated. The average path length is used as the prediction target, and the parameters of the deep learning model are used as input, fed into an independent MLP network for fitting. The MLP calculates an estimate of the average path length based on the parameters of the deep learning model, and its output is used as a regularization term. The regularization term is added to the deep learning model for regularization, resulting in a tree-regularized neural network model. By configuring the tree-regularized neural network model in the intelligent medical diagnostic system, interpretability is added to the prediction results, enabling the intelligent medical diagnostic system to have a certain interpretability of the prediction results.

[0133] Based on the above embodiments of this application, a fourth embodiment of this application is proposed. In this fourth embodiment, content that is the same as or similar to that in the above embodiments can be referred to the above description, and will not be repeated hereafter.

[0134] In this embodiment, after step S5, which involves adding the regularization term to the deep learning model for regularization processing to obtain a tree-regularized neural network model, step S6 is further included:

[0135] When the system has not reached the iteration limit, the training data is input into the tree regularized neural network model to obtain a new neural network prediction result, and the execution step is returned: construct a binary classification decision tree based on the neural network prediction result and the training dataset.

[0136] The parameters of a tree-regularized neural network model, such as weights and biases, are randomly initialized or set through some pre-training method. Through iterative training, these initial parameters can be adjusted based on the training results of the previous round, enabling the model to better fit the data and improve prediction performance.

[0137] Understandably, during the initial data training, the training dataset is input into the deep learning model framework to obtain the neural network's prediction results. After one round of data training, the deep learning model framework incorporates a tree regularization term, becoming a tree-regularized neural network model with regularization capabilities. To better fit the data and improve prediction performance, the tree-regularized neural network model after the first round of data training undergoes iterative training.

[0138] When the system does not reach the pre-set iteration limit, the training dataset is input into the tree regularized neural network model built in the previous iteration, and a new neural network prediction result is obtained based on the newly built tree regularized neural network model.

[0139] After obtaining the new neural network prediction results, return to the execution steps: construct a binary classification decision tree based on the neural network prediction results and the training dataset, consistent with the steps and calculation formulas of the previous round of constructing the binary classification decision tree. After constructing the new binary classification decision tree, continue to calculate the average path length of the binary classification decision tree, and iteratively train the tree regularization neural network model based on the average path length to improve the generalization ability of the final model and prevent overfitting.

[0140] In this embodiment, the tree-regularized neural network model is iteratively trained. Before the system reaches a preset number of iterations, iterative training is performed based on the model construction results of the previous round. The model parameters are continuously adjusted and the loss function is optimized so that the model can achieve better performance on the given task.

[0141] Based on the above embodiments of this application, a fifth embodiment of this application is proposed. In this fifth embodiment, content that is the same as or similar to that in the above embodiments can be referred to the above description, and will not be repeated hereafter.

[0142] In this embodiment, step S122, which constructs a cost-sensitive decision tree based on the training dataset and the regularized neural network model, may include steps A10 to A20:

[0143] Step A10: Obtain the feature contribution based on the training dataset and the regularized neural network model;

[0144] The training dataset is input into the regularized neural network model, and the SHAP method is used to obtain the feature contribution of each feature in the training dataset, i.e., the Shapley value.

[0145] It should be noted that SHAP (SHapley Additive exPlanations) is an explanatory model based on Shapley values ​​in game theory, used to interpret the predictions of machine learning models. This method calculates the importance value (Shapley value) of each feature variable in each sample, thereby explaining the predictive performance of the model.

[0146] SHAP is used to assign values ​​to the importance of different features. While ensuring the accuracy of the prediction results, features that doctors consider important are prioritized for modeling, and interpretation results that conform to doctors' experience are generated.

[0147] Because the interpretation results align with doctors' knowledge, they are more likely to be accepted by doctors and trusted in the diagnostic model. Taking the medical field as an example, the collected data includes patient height, weight, plasma protein A levels, and plasma urea nitrogen levels. Based on doctors' knowledge, the feature of plasma urea nitrogen level is more conducive to interpreting the disease diagnosis. If the interpretive model built using plasma urea nitrogen level performs similarly to the interpretive models built using the other three features, then the more important plasma urea nitrogen level should be prioritized as the feature for constructing the interpretive model.

[0148] Step A20: Construct a cost-sensitive decision tree based on the feature contribution.

[0149] The cost-sensitive decision tree is constructed using a three-generation iterative binary tree algorithm. Its data partitioning criterion is information gain, which is calculated based on information entropy. The calculation method for information entropy is as follows:

[0150]

[0151] In the formula, D represents the sample set, and p k Let |y| represent the proportion of samples of class k in sample set D, and let |y| represent the number of sample classes in the sample set.

[0152] The formula for calculating information gain is as follows:

[0153]

[0154] Where D v This indicates that feature a in D takes the value a. v The sample set is given by Gain(D,a), which represents the calculated information gain.

[0155] Cost-sensitive operations require incorporating the cost value corresponding to the attribute into the calculation of information gain, transforming the original information gain calculation method into cost information gain. The calculation method is as follows: CostGain(D, a) = Gain(D, a) / Cost(a), where Cost(a) is the cost of attribute a, Gain(D, a) represents the calculated information gain, and CostGain(D, a) represents the cost information gain.

[0156] To address scenarios where expert knowledge cannot be incorporated as a cost, feature contribution is chosen as an additional metric to measure feature importance, achieving the same effect as cost-information gain.

[0157] The cost information gain is obtained through the above calculations. Combined with the feature contribution calculated by the SHAP method, the cost-sensitive decision tree is constructed using the iterative binary tree 3rd generation algorithm.

[0158] It should be noted that the iterative binary tree 3rd generation algorithm consists of three parts: preorder traversal, inorder traversal, and postorder traversal. It mainly uses a top-down greedy search to traverse the possible decision space. Greedy search means that at each step, the best decision is made at the current step, hoping that such locally optimal choices will lead to a globally optimal solution. The cost-sensitive decision tree constructed by this algorithm can find an interpretation path that is more in line with the expert's perspective, thereby generating subsequent interpretation results.

[0159] In this embodiment, expert knowledge is incorporated as a cost to calculate the cost-information gain. Simultaneously, considering the case where expert knowledge cannot be incorporated as a cost, a cost-sensitive decision tree is constructed by combining the feature contribution calculated using the SHAP method. By constructing this cost-sensitive decision tree, the final model can interpret the prediction results of the intelligent medical diagnostic system from the perspective of expert decision-making, thereby overcoming the current shortcomings of intelligent medical diagnostic systems in lacking interpretability.

[0160] Based on the above embodiments of this application, a sixth embodiment of this application is proposed. In this sixth embodiment, content that is the same as or similar to that in the above embodiments can be referred to the above description, and will not be repeated hereafter.

[0161] In this embodiment, step S123, which involves adding the cost-sensitive decision tree to the tree-regularized neural network model to obtain the cost-sensitive decision tree-regularized neural network model, may include steps A30 to A40:

[0162] Step A30: Calculate the tree regularization term based on the cost-sensitive decision tree;

[0163] Because cost factors are considered in the selection of each split point during the construction of a decision tree, unlike traditional decision trees, cost-sensitive decision trees prioritize split points that minimize overall cost.

[0164] The purpose of tree regularization is to control the complexity of the model and prevent overfitting. In cost-sensitive decision trees, the calculation of regularization may incorporate cost factors.

[0165] Therefore, the tree regularization term is calculated by identifying the split points generated during the construction of the cost decision tree. Different weights are assigned to different split points to obtain the specific tree regularization term. Calculating the tree regularization term may also involve adjusting some hyperparameters, such as the regularization strength and the upper limit of the number of leaf nodes. The selection of these hyperparameters affects the balance between model performance and complexity. A weighted loss function is needed to optimize the calculation of the tree regularization term.

[0166] Step A40: Add the tree regularization term to the regularized neural network model to obtain a cost-sensitive decision tree regularized neural network model.

[0167] The loss function typically consists of two parts: a traditional loss term (such as cross-entropy loss or mean squared error loss) that measures the difference between the model's prediction and the true label; and a tree regularization term that controls the model's complexity and takes cost factors into account. By adding the tree regularization term to the loss function of the regularized neural network model, a cost-sensitive decision tree regularized neural network model can be constructed.

[0168] In this embodiment, the tree regularization term of the cost-sensitive decision tree is calculated and added to the tree regularized neural network model to obtain the cost-sensitive decision tree regularized neural network model. This model incorporates expert knowledge as a cost, enabling the intelligent medical diagnostic system to generate a tree-like explanatory graph. This allows the system to interpret prediction results from the perspective of medical experts, thus overcoming the current shortcomings of intelligent medical diagnostic systems in lacking explanatory power.

[0169] Based on the above embodiments of this application, a seventh embodiment of this application is proposed. In this seventh embodiment, content that is the same as or similar to that in the above embodiments can be referred to the above description, and will not be repeated hereafter.

[0170] In this embodiment, after step A40 of adding the tree regularization term to the regularized neural network model to obtain the cost-sensitive decision tree regularized neural network model, step A50 is also included;

[0171] Step A50: If the system has not reached the iteration limit, return to the execution step: obtain the feature contribution based on the training dataset and the regularized neural network model.

[0172] The regularization term in a regularized neural network model is calculated from the parameters of a binary decision tree and the deep learning model framework. However, after adding the tree regularization term of a cost-sensitive decision tree, the regularization term of the regularized neural network model changes, leading to changes in model parameters and feature contributions. To ensure the model's interpretability, it is necessary to re-obtain the feature contributions based on the adjusted cost-sensitive decision tree regularized neural network model and further optimize the model.

[0173] To achieve better optimization results for cost-sensitive decision tree regularized neural network models, the model parameters need to be adjusted. First, an iteration limit needs to be set. When the system does not reach the iteration limit, the model performance is gradually improved to find more accurate and meaningful feature contributions.

[0174] It is important to note that an appropriate iteration limit should be set during the iteration process to avoid overfitting or wasting computational resources. Close monitoring of the model's performance and feature contribution trends is also necessary to adjust optimization strategies promptly.

[0175] Furthermore, based on a pre-set iteration limit, the feature contribution is recalculated using the training dataset and a regularized neural network model.

[0176] If the iteration limit has not been reached, the regularized neural network model is adjusted based on the results of the previous iteration or new requirements. This adjustment is achieved by changing the type of regularization term, adjusting the regularization strength, and modifying the network structure. The feature contributions are then recalculated using the training dataset and the adjusted regularized neural network model.

[0177] In this embodiment, before the system reaches its iteration limit, the cost-sensitive decision tree regularized neural network model is continuously adjusted based on the current feature contribution evaluation results, enabling the model to more accurately capture feature relationships in the data. Each iteration further optimizes the model's parameters, thereby improving interpretation performance.

[0178] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the result interpretation method of the intelligent medical diagnostic system of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0179] Reference Figure 4 As shown, Figure 4 This is a flowchart illustrating the overall training process for constructing a tree-regularized neural network model. The following section combines... Figure 4 This paper describes the overall training process for constructing the tree-regularized neural network model in this application.

[0180] First, the training dataset is input into the deep learning model framework to obtain the neural network prediction results. The deep learning model framework is a subset of neural networks; neural networks are the overarching concept of deep learning models. Figure 4 The neural network shown is a deep learning model framework in its specific operation. The training dataset includes sample data and the corresponding ground truth labels. The neural network prediction result obtained by inputting the training dataset into the deep learning model framework is actually the predicted label of the sample data.

[0181] Furthermore, the neural network predictions are used as labels, and the training dataset is used as input to construct a binary classification decision tree. The neural network predictions are combined with the sample data in the training dataset to form a dataset, which is then fitted to obtain the binary classification decision tree.

[0182] Furthermore, the complexity of the binary classification decision tree is calculated. The complexity is determined by calculating the average path length of the binary classification decision tree. The average path length is the average number of decision tree nodes that must be traversed when predicting sample data.

[0183] Furthermore, the average path length is used as the prediction target, and the parameters of the deep learning model framework are used as input, which are then fed into a separate MLP network for fitting. The average path length and the basic parameters of the deep learning model framework are combined to form a surrogate dataset, which is then fed into the separate MLP network for fitting. The MLP calculates an estimate of the average path length based on the parameters of the deep learning model framework, and outputs the average path length estimate as a regularization term.

[0184] Finally, the calculated regularization term is added to the deep learning model framework to regularize the model, resulting in a tree-regularized neural network model. A deep learning model framework is a type of neural network; acquiring an existing framework can shorten the model training process. The calculated regularization term is actually an estimate of the average path length of the binary classification decision tree; therefore, the deep learning model with the added regularization term is named a tree-regularized neural network model. To ensure model performance and prevent overfitting, iterative training is required.

[0185] Reference Figure 5 As shown, Figure 5 This is a flowchart illustrating the overall training process for constructing a cost-sensitive decision tree regularized neural network model. The following section combines... Figure 5 This paper describes the overall training process of constructing the cost-sensitive decision tree regularized neural network model in this application.

[0186] First, the training dataset is input into the tree-regularized neural network model obtained in the previous model building process. The SAHP method is then applied to the tree-regularized neural network model to obtain the feature contribution value corresponding to each sample in the training dataset. Figure 5 The Shapley value shown.

[0187] Furthermore, feature contribution is used as a cost to construct a cost-sensitive decision tree. The cost-sensitive decision tree also incorporates expert knowledge as a cost, while using feature contribution as a cost addresses scenarios where expert knowledge cannot be included as a cost. This allows the use of Shapley value as an additional metric to measure feature importance, thus expanding the model's applicability.

[0188] Furthermore, the complexity of the cost-sensitive decision tree is calculated to obtain the tree regularization term. This tree regularization term is then added to the regularized neural network model in this round to obtain the cost-sensitive decision tree decision regularized neural network model.

[0189] Finally, to ensure model performance and prevent overfitting, the cost-sensitive decision tree decision regularized neural network model is trained iteratively. When the system has not reached the iteration limit, the model continues training, returning to the steps for calculating feature contributions.

[0190] This application also provides a result interpretation device for an intelligent medical diagnostic system. Please refer to [link / reference]. Figure 6 The result interpretation device of the intelligent medical diagnostic system includes:

[0191] Prediction module 10 is used to acquire patient clinical data, input the patient clinical data into a pre-created cost-sensitive decision tree regularized neural network model, and obtain prediction results through the neural network part of the cost-sensitive decision tree regularized neural network model.

[0192] Acquiring patient clinical data is a prerequisite for intelligent medical diagnostic systems to generate predictive results and corresponding interpretations. Sensors can monitor patients' physiological parameters and disease indicators in real time, and the collected data can be wirelessly transmitted to the cloud for further analysis and prediction by the intelligent medical diagnostic system.

[0193] Patient clinical data is used as input data and passed to a cost-sensitive decision tree regularized neural network model. The neural network model part of the cost-sensitive decision tree regularized neural network model performs predictive processing on the patient clinical data to obtain the prediction results.

[0194] The interpretation module 20 is used to interpret the patient's clinical data and prediction results through the cost-sensitive decision tree regularization part of the cost-sensitive decision tree regularized neural network model to obtain the interpretation result.

[0195] The cost-sensitive decision tree regularization part of the neural network model combines patient clinical data to analyze the prediction results, obtaining an interpretation that aligns with the decision-making perspective of medical experts.

[0196] The result interpretation device for the intelligent medical diagnostic system provided in this application employs the result interpretation method for the intelligent medical diagnostic system in the above embodiments, and can solve the technical problem of result interpretation in intelligent medical diagnostic systems. Compared with the prior art, the beneficial effects of the result interpretation device for the intelligent medical diagnostic system provided in this application are the same as the beneficial effects of the result interpretation method for the intelligent medical diagnostic system provided in the above embodiments, and other technical features in the result interpretation device for the intelligent medical diagnostic system are the same as the features disclosed in the method of the above embodiments, and will not be repeated here.

[0197] This application provides a result interpretation device for an intelligent medical diagnostic system. The result interpretation device for an intelligent medical diagnostic system includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the result interpretation method of the intelligent medical diagnostic system in the first embodiment described above.

[0198] The following is for reference. Figure 7 This document illustrates a schematic diagram of a result interpretation device suitable for implementing the intelligent medical diagnostic system of the embodiments of this application. The result interpretation device of the intelligent medical diagnostic system in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 7 The result interpretation device of the intelligent medical diagnostic system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0199] like Figure 7As shown, the result interpretation device of the intelligent medical diagnostic system may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 1002 or a program loaded from storage device 1003 into random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the result interpretation device of the intelligent medical diagnostic system. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the result interpretation device of the intelligent medical diagnostic system to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows a result interpretation device of an intelligent medical diagnostic system with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0200] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0201] The result interpretation device for the intelligent medical diagnostic system provided in this application, employing the result interpretation method described in the above embodiments, can solve the technical problem of result interpretation in intelligent medical diagnostic systems. Compared with the prior art, the beneficial effects of the result interpretation device for the intelligent medical diagnostic system provided in this application are the same as those of the result interpretation method for the intelligent medical diagnostic system provided in the above embodiments, and other technical features in the result interpretation device for the intelligent medical diagnostic system are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0202] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0203] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0204] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the result interpretation method of the intelligent medical diagnostic system in the above embodiments.

[0205] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0206] The aforementioned computer-readable storage medium may be included in the result interpretation device of the intelligent medical diagnostic system; or it may exist independently and not be assembled into the result interpretation device of the intelligent medical diagnostic system.

[0207] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the result interpretation device of the intelligent medical diagnostic system, the result interpretation device of the intelligent medical diagnostic system: acquires patient clinical data; inputs the patient clinical data into a pre-constructed cost-sensitive decision tree regularized neural network model; performs prediction processing on the patient clinical data through the neural network part of the cost-sensitive decision tree regularized neural network model to obtain a prediction result; and performs interpretation processing on the patient clinical data and the prediction result through the cost-sensitive decision tree regularized part to obtain an interpretation result.

[0208] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0209] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, 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 containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated 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 the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0210] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0211] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the result interpretation method of the above-described intelligent medical diagnostic system, thereby solving the technical problem of result interpretation in intelligent medical diagnostic systems. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the result interpretation method of the intelligent medical diagnostic system provided in the above embodiments, and will not be repeated here.

[0212] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the result interpretation method of the intelligent medical diagnostic system as described above.

[0213] The computer program product provided in this application can solve the technical problem of result interpretation in intelligent medical diagnostic systems. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the result interpretation method for intelligent medical diagnostic systems provided in the above embodiments, and will not be repeated here.

[0214] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for interpreting the results of an intelligent medical diagnostic system, characterized in that, The method includes: Obtain patient clinical data; The patient's clinical data is input into a pre-constructed cost-sensitive decision tree regularized neural network model; The patient's clinical data is predicted by the neural network part of the cost-sensitive decision tree regularized neural network model to obtain the prediction result, and the patient's clinical data and the prediction result are interpreted by the cost-sensitive decision tree regularized part to obtain the interpretation result. The steps preceding the input of the patient clinical data into the pre-built cost-sensitive decision tree regularized neural network model also include: Construct a tree-regularized neural network model based on the pre-collected training dataset; A cost-sensitive decision tree is constructed based on the training dataset and the tree-regularized neural network model; The cost-sensitive decision tree is added to the tree-regularized neural network model to obtain the cost-sensitive decision tree-regularized neural network model. The steps of constructing a tree-regularized neural network model based on the pre-collected training dataset include: Obtain the pre-collected training dataset and deep learning model framework; The training dataset is input into the deep learning model framework to obtain the neural network prediction results; Construct a binary classification decision tree based on the neural network prediction results and the training dataset; The regularization term is calculated based on the binary classification decision tree and the deep learning model framework. The regularization term is added to the deep learning model for regularization processing to obtain a tree-regularized neural network model.

2. The method as described in claim 1, characterized in that, After the step of adding the regularization term to the deep learning model for regularization processing to obtain a tree-regularized neural network model, the method further includes: When the system has not reached the iteration limit, the training data is input into the tree regularized neural network model to obtain a new neural network prediction result, and the execution step is returned: construct a binary classification decision tree based on the neural network prediction result and the training dataset.

3. The method as described in claim 2, characterized in that, The step of constructing a cost-sensitive decision tree based on the training dataset and the tree-regularized neural network model includes: The feature contribution is obtained based on the training dataset and the tree regularized neural network model; A cost-sensitive decision tree is constructed based on the contribution of the aforementioned features.

4. The method as described in claim 3, characterized in that, The step of adding the cost-sensitive decision tree to the tree-regularized neural network model to obtain the cost-sensitive decision tree-regularized neural network model includes: The tree regularization term is calculated based on the cost-sensitive decision tree; Adding the tree regularization term to the tree regularization neural network model yields a cost-sensitive decision tree regularization neural network model.

5. The method as described in claim 4, characterized in that, After the step of adding the tree regularization term to the tree regularization neural network model to obtain the cost-sensitive decision tree regularization neural network model, the method further includes: If the system does not reach the iteration limit, return to the execution step: obtain the feature contribution based on the training dataset and the tree regularized neural network model.

6. A result interpretation device for an intelligent medical diagnostic system, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the result interpretation method of the intelligent medical diagnostic system as described in any one of claims 1 to 5.

7. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the result interpretation method of the intelligent medical diagnostic system as described in any one of claims 1 to 5.

8. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the result interpretation method of the intelligent medical diagnostic system as described in any one of claims 1 to 5.

Images