Medical dispute high-risk crowd identification method and system based on neural network

Abstract

The invention relates to a medical dispute high-risk crowd identification method and system based on a neural network. The method comprises the steps that multi-dimensional information of a patient is acquired and preprocessed; constructing a deep feedforward neural network and inputting the preprocessed multi-dimensional information; the preprocessed multi-dimensional information is divided into a training set and a verification set, the training set is used for training the deep feed-forward neural network, and the verification set is used for verifying and optimizing the deep feed-forward neural network; and early warning information output by the deep feedforward neural network is sent to an intelligent terminal of a medical worker. By adopting the method, through an advanced deep learning technology and multi-dimensional data integration, high-precision identification and real-time dynamic monitoring of high-risk crowds with medical disputes are realized, the management efficiency and patient satisfaction of medical institutions are remarkably improved, the occurrence of medical disputes is effectively reduced, and the method has important practical application value.

Description

Technical Field

[0001] The present application relates to the technical field of risk management and prevention in the medical field, and particularly to a method and system for identifying high-risk groups of medical disputes based on a neural network. Background Art

[0002] With the rapid development of the medical industry, the increasing number of medical disputes has become an important challenge faced by the modern medical and health system, which is significantly related to the accelerating iteration of medical technology and the increasing public health demands. Medical disputes not only harm the physical and mental health of patients, but also cause significant economic burdens and reputational damages to medical institutions. Therefore, accurately identifying high-risk groups of medical disputes and taking preventive measures in advance is of great significance for maintaining harmonious doctor-patient relations and ensuring medical safety.

[0003] Existing methods for identifying high-risk groups of medical disputes mainly rely on the experience and subjective judgment of medical staff, lacking systematicness and scientificity. For example, a study analyzed numerous medical dispute cases and summarized nineteen characteristics of potential high-risk groups, including the psychological and mental conditions of patients, special identities, and whether they had ever had medical disputes with other hospitals. In recent years, academic research has constructed a "Simple Scale for Preventing and Controlling Medical Dispute Risks" by systematically summarizing key risk factors in medical dispute events. This tool aims to achieve quantitative assessment and dynamic monitoring of potential doctor-patient conflicts, thereby effectively reducing the incidence of disputes in high-risk medical treatment links. There are also studies that identify potential high-risk patients by constructing a "portrait of dispute patients" and using big data technology to conduct multi-dimensional analysis of patient information.

[0004] However, these methods have the following limitations: The data types are relatively limited, mainly concentrated on the basic information and medical records of patients, lacking diversity, so it is difficult to comprehensively reveal the risk characteristics of patients. Due to the lack of association between different data types, the accuracy of the identification model is affected, and there may be large errors in the identification results. Summary of the Invention

[0005] 1. Problems to be Solved

[0006] Based on this, it is necessary to provide a method and system for identifying high-risk groups of medical disputes based on a neural network that can comprehensively reveal the risk characteristics of patients and improve the identification effect for the above technical problems.

[0007] 2. Technical Solutions

[0008] In the first aspect, the present application provides a method for identifying high-risk groups of medical disputes based on a neural network. The method includes:

[0009] Obtain the multi-dimensional information of the patient and perform preprocessing. The multi-dimensional information of the patient includes at least basic information, medical records, psychological assessment, and social background;

[0010] Construct a deep feedforward neural network and input the preprocessed multi-dimensional information;

[0011] Divide the preprocessed multi-dimensional information into a training set and a validation set. The training set is used to train the deep feedforward neural network, and the validation set is used to verify and tune the deep feedforward neural network;

[0012] Send the warning information output by the deep feedforward neural network to the intelligent terminal of the medical staff.

[0013] In one embodiment, obtaining the multi-dimensional information of the patient includes:

[0014] Query the internal information of the patient in the preset in-hospital information system. The internal information includes at least age, gender, occupation, education level, diagnosis and treatment records, and medical service experience;

[0015] Query external data in the preset database. The external data includes at least regional disease epidemic trends and social and economic backgrounds;

[0016] Construct the multi-dimensional data of the patient by combining external data and internal data.

[0017] In one embodiment, obtaining the multi-dimensional information of the patient and performing preprocessing includes:

[0018] When the multi-dimensional data is a continuous value, perform standardization based on a preset method;

[0019] When the multi-dimensional data is an enumerated value, construct a mapping relationship through a preset one-hot encoding;

[0020] When the multi-dimensional data is text information, process it based on word embedding technology;

[0021] Integrate the preprocessed multi-dimensional data into an input vector.

[0022] In one embodiment, constructing a deep feedforward neural network includes:

[0023] Construct a 5-layer deep feedforward neural network. The 0th layer is the input layer, the 4th layer is the output layer, and the 1st - 3rd layers are hidden layers;

[0024] For the lth layer (l = 1, 2, 3), assume that this layer has n l nodes, and each node receives the weighted input from the previous layer l - 1 and performs a non-linear transformation through an activation function. Specifically:

[0025] a0 = x;

[0026] z l = W l a l-1 + b l for l = 1, 2, 3;

[0027] a l = f l (z l ) for l = 1, 2, 3;

[0028] y = f L (z L );

[0029] Where: is the weight matrix. is the bias vector, is the unactivated weighted input, a0 is the input vector of the input layer of the deep feedforward neural network, is the output after activation of the l-th layer, f l is the ReLU activation function, which is defined as f L is the sigmoid function, which is defined as

[0030] In one embodiment, iterative training is performed based on the preprocessed multi-dimensional data;

[0031] In each training iteration, a batch of samples is input into the network, and the predicted output is calculated through forward propagation

[0032] The loss function is used to calculate and measure the difference between the predicted output and the true label y. The loss function is defined as:

[0033]

[0034] where N is the number of input vectors of the batch of samples;

[0035] The gradient of the loss with respect to each parameter is calculated using the chain rule, and the weights and biases are updated through the backpropagation algorithm to minimize the loss function;

[0036] First, calculate the errors of the output layer and each layer. The formula is:

[0037]

[0038] where δ L is the error term (also known as the residual) of the output layer, δ l is the error term of the l-th layer, f l ′is the derivative of the activation function f l ; ⊙ represents element-wise multiplication;

[0039] Calculate the gradients of the weights and biases based on the gradient calculation formula:

[0040]

[0041] Use the Stochastic Gradient Descent (SGD) algorithm to reset the weights and biases of each layer, and the update rule is:

[0042]

[0043]

[0044] where η is the learning rate;

[0045] Repeat the above steps until the loss converges.

[0046] In one embodiment, sending the warning information output by the deep feedforward neural network to the smart terminal of the medical staff includes:

[0047] Input the multi-dimensional information into the verified deep feedforward neural network and calculate the corresponding risk probability;

[0048] Compare the risk probability with a preset risk threshold;

[0049] When the risk probability is greater than the preset risk threshold, send the relevant warning information to the smart terminal of the medical staff.

[0050] In a second aspect, the present application also provides a system for identifying high-risk groups of medical disputes based on a neural network. The system includes:

[0051] A multi-dimensional information preprocessing module for obtaining and preprocessing the multi-dimensional information of the patient, where the multi-dimensional information of the patient at least includes basic information, medical records, psychological evaluation, and social background;

[0052] A neural network construction module for constructing a deep feedforward neural network and inputting the preprocessed multi-dimensional information;

[0053] A neural network training module for dividing the preprocessed multi-dimensional information into a training set and a validation set, where the training set is used to train the deep feedforward neural network, and the validation set is used to verify and optimize the deep feedforward neural network;

[0054] A warning information sending module for sending the warning information output by the deep feedforward neural network to the smart terminal of the medical staff.

[0055] In a third aspect, the present application also provides a computer device. The computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the following steps are implemented:

[0056] Obtain multi-dimensional information of a patient and perform preprocessing. The multi-dimensional information of the patient includes at least basic information, medical records, psychological assessment, and social background;

[0057] Construct a deep feedforward neural network and input the preprocessed multi-dimensional information;

[0058] Divide the preprocessed multi-dimensional information into a training set and a validation set. The training set is used to train the deep feedforward neural network, and the validation set is used to verify and optimize the deep feedforward neural network;

[0059] Send the warning information output by the deep feedforward neural network to the intelligent terminal of medical staff.

[0060] In a fourth aspect, the present application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the following steps are implemented:

[0061] Obtain multi-dimensional information of a patient and perform preprocessing. The multi-dimensional information of the patient includes at least basic information, medical records, psychological assessment, and social background;

[0062] Construct a deep feedforward neural network and input the preprocessed multi-dimensional information;

[0063] Divide the preprocessed multi-dimensional information into a training set and a validation set. The training set is used to train the deep feedforward neural network, and the validation set is used to verify and optimize the deep feedforward neural network;

[0064] Send the warning information output by the deep feedforward neural network to the intelligent terminal of medical staff.

[0065] In a fifth aspect, the present application also provides a computer program product. The computer program product includes a computer program. When the computer program is executed by a processor, the following steps are implemented:

[0066] Obtain multi-dimensional information of a patient and perform preprocessing. The multi-dimensional information of the patient includes at least basic information, medical records, psychological assessment, and social background;

[0067] Construct a deep feedforward neural network and input the preprocessed multi-dimensional information;

[0068] The preprocessed multi-dimensional information is divided into a training set and a validation set, wherein the training set is used to train the deep feedforward neural network, and the validation set is used to verify and tune the deep feedforward neural network;

[0069] Send the warning information output by the deep feedforward neural network to the smart terminal of medical staff.

[0070] 3. Beneficial effects

[0071] This application adopts the above method with high accuracy: by integrating multi-dimensional data, the present invention can more comprehensively reflect the risk characteristics of patients and improve the accuracy of identification.

[0072] The deep feedforward neural network, with its powerful multi-level feature extraction capabilities, can autonomously learn and mine the complex nonlinear relationships hidden in the data, thereby significantly improving the prediction accuracy and generalization performance of the model. The test results on the validation set show that the recognition accuracy of the present invention is significantly higher than that of traditional methods, and it can more effectively identify high-risk groups.

[0073] Real-time dynamic monitoring: The present invention can update and analyze the patient's medical records and related data in real time, and realize dynamic monitoring of high-risk groups. According to the analysis of medical dispute data in general hospitals, regular data updates can help medical institutions to timely discover potential risks and issue early warnings, so as to take effective measures to prevent the occurrence of medical disputes.

[0074] Efficient information push and management mechanism: The identification results are immediately delivered to the terminal devices of relevant medical staff, ensuring the timeliness and accuracy of the information. Medical staff can obtain early warning information of high-risk patients in a timely manner and take targeted preventive measures, such as strengthening communication, informing treatment plans in detail, and conducting psychological counseling, etc., to improve management efficiency and patient satisfaction.

[0075] By identifying high-risk groups in advance and taking preventive measures, such as enhancing medical transparency, improving medical quality, strengthening doctor-patient communication, complying with medical standards, fulfilling the obligation to inform, etc., the present invention can significantly reduce the occurrence of medical disputes and maintain a harmonious doctor-patient relationship. Reducing the incidence of medical disputes not only helps to protect the legitimate rights and interests of patients, but also reduces the economic and reputation losses of medical institutions and improves the quality of medical services.

[0076] The method and system of the present invention have good scalability and can be easily integrated into the existing medical information system without large-scale system transformation. They are suitable for medical institutions of different sizes and types and have wide applicability and promotion value.

[0077] In summary, through advanced deep learning technology and multi-dimensional data integration, the present invention realizes high-precision identification and real-time dynamic monitoring of high-risk groups in medical disputes, significantly improves the management efficiency of medical institutions and patient satisfaction, effectively reduces the occurrence of medical disputes, and has important practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 It is the overall architecture diagram of the method for identifying high-risk groups in medical disputes based on neural network in one embodiment;

[0079] Figure 2 It is the detailed structure diagram of data collection in one embodiment;

[0080] Figure 3 It is the detailed structure diagram of the data preprocessing unit in one embodiment;

[0081] Figure 4 It is the architecture diagram of the deep feedforward neural network model in one embodiment;

[0082] Figure 5 It is the flowchart of model training in one embodiment;

[0083] Figure 6 It is the structural block diagram of the system for identifying high-risk groups in medical disputes based on neural network in one embodiment;

[0084] Figure 7 It is the internal structure diagram of a computer device in one embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0085] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0086] In recent years, Deep Feedforward Neural Networks (DFFN) have made remarkable progress in multiple fields. Especially in the medical field, through applications such as disease prediction, diagnosis, and treatment plan formulation, the medical level and treatment effect have been further improved. Deep feedforward neural networks can effectively process complex non-linear relationships through multi-level feature extraction and abstraction, improving the accuracy and generalization ability of the model. In the medical field, deep feedforward neural networks have been successfully applied to medical image recognition, disease diagnosis, and drug research and development, etc.

[0087] The present invention aims to provide a method for identifying high-risk groups of medical disputes based on a deep feedforward neural network. By constructing portraits of high-risk groups, we use preset analysis models and prediction models to conduct multi-dimensional data analysis on the target population, accurately predict the disease types and risk levels of each target object, and push the prediction results to their terminal devices in real time, so as to achieve dynamic monitoring and precise management of high-risk groups.

[0088] In one embodiment, as Figure 1 shown, in this embodiment, the method includes the following steps:

[0089] Step 202, obtain multi-dimensional information of the patient and perform preprocessing.

[0090] Among them, the multi-dimensional information of the patient includes at least basic information, medical records, psychological assessment, and social background.

[0091] Step 204, construct a deep feedforward neural network and input the preprocessed multi-dimensional information.

[0092] Step 206, divide the preprocessed multi-dimensional information into a training set and a validation set.

[0093] Among them, the training set is used to train the deep feedforward neural network, and the validation set is used to verify and optimize the deep feedforward neural network.

[0094] Step 208, send the warning information output by the deep feedforward neural network to the intelligent terminals of medical staff.

[0095] In the above method for identifying high-risk groups of medical disputes based on a neural network, the present invention realizes high-precision identification and real-time dynamic monitoring of high-risk groups of medical disputes through advanced deep learning technology and multi-dimensional data integration, significantly improves the management efficiency of medical institutions and patient satisfaction, effectively reduces the occurrence of medical disputes, and has important practical application value.

[0096] In one embodiment, obtaining the multi-dimensional information of the patient specifically includes: querying the internal information of the patient in a preset in-hospital information system, where the internal information includes at least age, gender, occupation, education level, diagnosis and treatment records, and medical service experience; querying external data in a preset database, where the external data includes at least regional disease prevalence trends and social and economic backgrounds; combining the external data and the internal data to construct the multi-dimensional data of the patient.

[0097] Among them, in the data collection stage, as Figure 2As shown, patient basic information, medical records, and medical service experiences are mainly extracted from the in-hospital information system. At the same time, we need to draw information crucial for identifying high-risk groups of medical disputes from external data sources, covering the patient's economic status, psychological assessment results, and social background information, etc.

[0098] Patient basic information is the basis for identification, including information such as age, gender, occupation, and educational level. Medical records provide detailed medical history information and are an important basis for assessing potential medical disputes. Medical records are the key to preventing medical disputes and should detail the medical history (including past diagnoses and surgical experiences), current diagnosis, medication use (long-term medications and drug allergy history), test reports (such as blood test and imaging results), as well as the doctor's professional opinions and assessment of the disease progression. This information helps to understand the patient's overall health status and treatment compliance.

[0099] Medical service experiences reflect the patient's subjective evaluation and actual interaction with hospital services. This includes satisfaction scores, complaint history, and feedback. The patient's attitude towards the service and any dissatisfaction expressed by them are important clues for identifying potential disputes, while their feedback provides valuable reference for improving service quality and preventing future disputes.

[0100] The patient's insurance coverage, economic situation, payment ability, and family support system are crucial. Insurance coverage and payment ability are directly related to the likelihood of cost-related disputes, while good economic and family support can improve treatment outcomes and reduce the risk of disputes.

[0101] External data sources such as public health statistics and socioeconomic statistics should also be considered. Regional disease prevalence trends assist in judging the prevalence and severity of specific diseases, while socioeconomic statistics help to understand the socioeconomic background of the patient's community, thus predicting possible influencing factors.

[0102] The raw data needs to be cleaned before it can be effectively utilized. A preliminary review of the raw data will be conducted to check its format standardization, missing value status, and the existence of abnormal records. By calculating the basic statistics of each feature (such as mean, median, standard deviation) to understand the data distribution characteristics, and using charts (such as histograms, box plots) to visually display the data features, helping to identify potential problems or patterns.

[0103] For records with missing values, the missing proportion and its impact on the analysis will be evaluated. For records with a small missing proportion and little impact on the overall analysis, rows or columns containing missing values will be selected for deletion. For cases with few missing values, the missing values will be filled according to business logic or using statistical methods (such as mean / median imputation, KNN interpolation). For critical diagnostic or treatment information, models will be constructed using other relevant features to predict and fill in the missing values.

[0104] For duplicate records, the patient's unique identifier (such as the medical record number) will be used to find exactly the same or partially the same records. One record will be selected for retention, and the information of other duplicates will be integrated into this record, or the redundant duplicate records will be directly deleted to ensure the uniqueness of each patient's data.

[0105] Based on medical knowledge or statistical methods (such as Z-score, IQR rule), determine which values are abnormal. For outliers clearly caused by data entry errors, try to correct these errors. If the outliers are truly unreasonable and the number is small, they will be deleted. Conversely, if the outliers may contain important information, they should be retained and given special attention during the modeling process.

[0106] For continuous numerical features such as age and body temperature, use Min-Max Scaling or Z-score normalization to ensure that different features have similar scales. For time series data such as follow-up records and hospitalization logs, ensure that all timestamps are correctly parsed and converted to a unified time format; adjust the sampling frequency as needed, such as converting daily data to weekly summaries.

[0107] Remove unnecessary content such as HTML tags, special characters, and stop words from doctor comments and patient feedback. Decompose sentences into word or phrase units (tokenization), and normalize different forms of words to their basic forms (lemmatization / stemming).

[0108] In one embodiment, obtaining and preprocessing the patient's multi-dimensional information includes: when the multi-dimensional data is continuous, standardizing it based on a preset method; when the multi-dimensional data is an enumerated value, constructing a mapping relationship through a preset one-hot encoding; when the multi-dimensional data is text information, processing it based on word embedding technology; integrating the preprocessed multi-dimensional data into an input vector.

[0109] Among them, the data preprocessing unit is responsible for converting the cleaned input data into an input vector suitable for a deep feedforward neural network. The input vector will comprehensively reflect each patient's personal background information, diagnosis and treatment history, and other relevant information. The data preprocessing unit will convert different information of patients into appropriate vector forms according to the data type to ensure data consistency and the effectiveness of model training.

[0110] For information of different data types, different processing methods will be adopted. For continuous values, such as age or body temperature, etc., their original numerical values can be directly used as part of the input. In specific situations, continuous variables may need to go through a normalization or standardization process so that the model can learn and generalize more effectively. For example, techniques such as Min-Max Scaling or Z-score standardization can be used to make all features have a similar scale.

[0111] For enumerated values, they will be represented by one-hot encoding or an embedding layer. When the number of categories is small, such as gender, occupation, etc., one-hot encoding creates a binary flag for each category, where only the position corresponding to the category is 1 and the rest are 0. For features with a large number of categories, such as ICD-10 diagnostic codes, an embedding layer will be used for encoding.

[0112] For text information, such as unstructured data like doctor's comments, patient feedback, etc., we mainly use the Word2Vec word embedding technique for processing.

[0113] Time series data, such as regular examination results or long-term medication records, contains data points that change over time. Such data will be organized into fixed-length time windows, and the data points within each window are arranged in chronological order, and statistical features such as mean, standard deviation, trend, etc. can be calculated.

[0114] Finally, all preprocessed and encoded features will be integrated into a large input vector, which contains all selected features and is ready for training by a deep feedforward neural network.

[0115] In one embodiment, referring to Figure 5 , constructing a deep feedforward neural network includes:

[0116] Construct a 5-layer deep feedforward neural network, where the 0th layer is the input layer, the 4th layer is the output layer, and the 1st - 3rd layers are hidden layers;

[0117] For the lth layer (l = 1, 2, 3), assume that this layer has n l nodes, and each node receives a weighted input from the previous layer l - 1 and undergoes a non-linear transformation through an activation function. Specifically:

[0118] a0 = x;

[0119] z l = W l a l-1 + b l for l = 1, 2, 3;

[0120] al = f l (z l ) for l = 1, 2, 3;

[0121] y = f L (z L );

[0122] Where: is the weight matrix. is the bias vector, is the unactivated weighted input, a0 is the input vector of the input layer of the deep feedforward neural network, is the output after activation of the l-th layer, f l is the ReLU activation function, which is defined as f L is the sigmoid function, which is defined as

[0123] In one embodiment, iterative training is performed based on the preprocessed multi-dimensional data;

[0124] In each training iteration, a batch of samples is input into the network, and the predicted output is calculated through forward propagation

[0125] The loss function is used to calculate and measure the difference between the predicted output and the true label y. The loss function is defined as:

[0126]

[0127] where N is the number of input vectors of the batch of samples;

[0128] The gradient of the loss with respect to each parameter is calculated using the chain rule, and the weights and biases are updated through the backpropagation algorithm to minimize the loss function;

[0129] First, the errors of the output layer and each layer are calculated. The formula is:

[0130]

[0131] where δ L is the error term (also known as the residual) of the output layer, δ l is the error term of the l-th layer, f l ′ is the derivative of the activation function f l ⊙ represents element-wise multiplication;

[0132] Based on the gradient calculation formula, the gradients of the weights and biases are calculated:

[0133]

[0134] Using the Stochastic Gradient Descent (SGD) algorithm, reset the weights and biases of each layer, and the update rule is as follows:

[0135]

[0136] where η is the learning rate;

[0137] Repeat the above steps until the loss converges.

[0138] Among them, when starting to train the model, the Xavier / Glorot initialization strategy will be used to assign initial values to the weights of each layer of the neural network model to ensure good gradient flow at the start of training.

[0139] In one embodiment, sending the warning information output by the deep feedforward neural network to the intelligent terminal of medical staff includes: inputting multi-dimensional information into the verified deep feedforward neural network and calculating the corresponding risk probability; comparing the risk probability with a preset risk threshold; when the risk probability is greater than the preset risk threshold, sending the relevant warning information to the intelligent terminal of medical staff.

[0140] Among them, the trained deep feedforward neural network model can be used for identifying high-risk groups in medical disputes; first, through the above data collection process, collect and clean the relevant information of the target population to make it an input vector acceptable to the deep feedforward neural network. Input the input vector into the trained DFFN model for prediction, and the probability of each sample belonging to the high-risk group will be output. Set a threshold according to business requirements and actual situations, such as 0.5 or a higher value, to distinguish high-risk and low-risk patients.

[0141] Considering other factors, such as the professional judgment of doctors and the special situations of patients, make a final risk assessment; once the system determines that a certain patient is a high-risk group, a warning message will be automatically generated immediately and relevant medical staff will be notified in a timely manner so that they can quickly take necessary preventive measures.

[0142] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are sequentially shown according to the indications of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of the steps or stages in other steps or other steps.

[0143] Based on the same inventive concept, an embodiment of the present application further provides a neural network-based medical dispute high-risk population identification system for implementing the above-mentioned neural network-based medical dispute high-risk population identification method. The implementation solutions provided by this system to solve problems are similar to the implementation solutions recorded in the above method. Therefore, the specific limitations in one or more embodiments of the neural network-based medical dispute high-risk population identification system provided below can refer to the limitations on the neural network-based medical dispute high-risk population identification method in the above text, and will not be repeated here.

[0144] In one embodiment, as Figure 6 shown, a neural network-based medical dispute high-risk population identification system is provided, including: a multi-dimensional information preprocessing module, a neural network construction module, a neural network training module, and a warning information sending module, where:

[0145] The multi-dimensional information preprocessing module is used to obtain and preprocess the multi-dimensional information of the patient. The multi-dimensional information of the patient at least includes basic information, medical records, psychological evaluation, and social background;

[0146] The neural network construction module is used to construct a deep feedforward neural network and input the preprocessed multi-dimensional information;

[0147] The neural network training module is used to divide the preprocessed multi-dimensional information into a training set and a validation set. The training set is used to train the deep feedforward neural network, and the validation set is used to verify and optimize the deep feedforward neural network;

[0148] The warning information sending module is used to send the warning information output by the deep feedforward neural network to the intelligent terminal of the medical staff.

[0149] In one embodiment, the multi-dimensional information preprocessing module is further configured to: query the internal information of the patient in a preset in-hospital information system, where the internal information at least includes age, gender, occupation, education level, diagnosis and treatment records, and medical service experience; query external data in a preset database, where the external data at least includes regional disease epidemic trends and socioeconomic backgrounds; and construct multi-dimensional data of the patient by combining the external data and the internal data.

[0150] In one embodiment, the multi-dimensional information preprocessing module is further configured to: when the multi-dimensional data is a continuous value, perform standardization based on a preset method; when the multi-dimensional data is an enumerated value, construct a mapping relationship through a preset one-hot encoding; when the multi-dimensional data is text information, perform processing based on a word embedding technique; and integrate the preprocessed multi-dimensional data into an input vector.

[0151] In one embodiment, the neural network construction module is further configured to: construct a 5-layer deep feedforward neural network, where the 0th layer is the input layer, the 4th layer is the output layer, and the 1st - 3rd layers are hidden layers; for the lth layer (l = 1, 2, 3), assume that this layer has n l nodes, and each node receives a weighted input from the previous layer l - 1 and performs a non-linear transformation through an activation function. Specifically: a0 = x; z l = W l a l-1 + b l for l = 1, 2, 3; a l = f l (z l ) for l = 1, 2, 3; y = f L (z L ); where: is the weight matrix. is the bias vector, is the unactivated weighted input, a0 is the input vector of the input layer of the deep feedforward neural network, is the output after activation of the lth layer, f l is the ReLU activation function, and its definition is f L is the sigmoid function, and its definition is

[0152] In one embodiment, the neural network training module is further configured to: perform iterative training based on the preprocessed multi-dimensional data; in each training iteration, input a batch of samples into the network, calculate the predicted output through forward propagation calculate the difference between the predicted output and the true label y using a loss function, and the loss function is defined as: Among them, N is the number of input vectors of the samples in this batch; the gradient of the loss with respect to each parameter is calculated using the chain rule, and the weights and biases are updated through the backpropagation algorithm to minimize the loss function; first, the errors of the output layer and each layer are calculated, and the formula is: Among them, δ L is the error term (also known as the residual) of the output layer, and δ l is the error term of the l-th layer, and f l ' is the derivative of the activation function f l ; ⊙ represents element-wise multiplication; the gradients of the weights and biases are calculated based on the gradient calculation formula: Using the Stochastic Gradient Descent (SGD) algorithm, the weights and biases of each layer are reset, and the update rule is: Among them, η is the learning rate; the above steps are repeated until the loss converges.

[0153] In one embodiment, the early warning information sending module is further configured to: input the multi-dimensional information into the verified deep feedforward neural network and calculate the corresponding risk probability; compare the risk probability with a preset risk threshold; when the risk probability is greater than the preset risk threshold, send the relevant early warning information to the intelligent terminal of the medical staff.

[0154] Each module in the above medical dispute high-risk population identification system based on a neural network can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor in the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.

[0155] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 7 shown. The computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a method for identifying high-risk populations in medical disputes based on a neural network.

[0156] In one embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as Figure 7As shown in the figure. The computer device includes a processor, a memory, a communication interface, a display screen, and an input system connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be implemented through WIFI, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a method for identifying high-risk groups in medical disputes based on neural networks.

[0157] Those skilled in the art can understand that Figure 7 the structure shown in the figure is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0158] In one embodiment, a computer device is provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0159] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0160] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0161] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0162] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0163] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0164] The above-described embodiments only represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for identifying high-risk groups in medical disputes based on neural networks, characterized in that, The method includes: Obtaining multi-dimensional information of a patient and performing preprocessing, where the multi-dimensional information of the patient at least includes basic information, medical records, psychological assessment, and social background; Constructing a deep feedforward neural network and inputting the preprocessed multi-dimensional information; Dividing the preprocessed multi-dimensional information into a training set and a validation set, where the training set is used to train the deep feedforward neural network, and the validation set is used to verify and optimize the deep feedforward neural network; Sending the warning information output by the deep feedforward neural network to the intelligent terminal of medical staff.

2. The method for identifying high-risk groups of medical disputes based on neural network according to claim 1, wherein, The obtaining of the multi-dimensional information of the patient includes: Querying the internal information of the patient in a preset in-hospital information system, where the internal information at least includes age, gender, occupation, education level, diagnosis and treatment records, and medical service experience; Querying external data in a preset database, where the external data at least includes regional disease prevalence trends and social and economic backgrounds; Constructing multi-dimensional data of the patient by combining the external data and the internal data.

3. The method for identifying high-risk groups of medical disputes based on a neural network according to claim 2, wherein The obtaining of the multi-dimensional information of the patient and performing preprocessing includes: When the multi-dimensional data is a continuous value, performing standardization based on a preset method; When the multi-dimensional data is an enumerated value, constructing a mapping relationship through a preset one-hot encoding; When the multi-dimensional data is text information, processing it based on word embedding technology; Integrating the preprocessed multi-dimensional data into an input vector.

4. The method for identifying high-risk groups of medical disputes based on neural network according to claim 3, characterized in that The constructing of the deep feedforward neural network includes: Constructing a 5-layer deep feedforward neural network, where the 0th layer is the input layer, the 4th layer is the output layer, and the 1st - 3rd layers are hidden layers; For the l-th layer (l = 1, 2, 3), assume that there are n l nodes in this layer. Each node receives the weighted input from the previous layer l - 1 and performs a non-linear transformation through an activation function. Specifically: a0 = x; z l = W l a l-1 + b l for l = 1, 2, 3; a l = f l (z l ) for l = 1, 2, 3; y = f L (z L ); Wherein: is the weight matrix. is the bias vector, is the unactivated weighted input, and a0 is the input vector of the input layer of the deep feedforward neural network, is the output after activation of the l-th layer, and f l is the ReLU activation function, which is defined as f L is the sigmoid function, which is defined as 5. The method for identifying high-risk groups of medical disputes based on neural network according to claim 4, wherein The method further includes: Performing iterative training based on the preprocessed multi-dimensional data; In each training iteration, a batch of samples is input into the network, and the predicted output is calculated through forward propagation Calculate the difference between the predicted output and the true label y using a loss function, and the loss function is defined as: and the true label y, and the loss function is defined as: where N is the number of input vectors of the batch of samples; Calculating the gradient of the loss with respect to each parameter using the chain rule and updating the weights and biases through the backpropagation algorithm to minimize the loss function; First, calculating the errors of the output layer and each layer, and the formula is: where δ L is the error term (also known as the residual) of the output layer, and δ l is the error term of the l-th layer, f l ′ is the derivative of the activation function f l , and ⊙ represents element-wise multiplication; Calculating the gradients of the weights and biases based on the gradient calculation formula: Using the stochastic gradient descent (SGD) algorithm, resetting the weights and biases of each layer, and the update rule is: where η is the learning rate; Repeating the above steps until the loss converges.

6. The method for identifying high-risk groups of medical disputes based on neural network according to claim 5, characterized in that, The sending of the warning information output by the deep feedforward neural network to the intelligent terminal of medical staff includes: Inputting the multi-dimensional information into the verified deep feedforward neural network and calculating the corresponding risk probability; Comparing the risk probability with a preset risk threshold; When the risk probability is greater than the preset risk threshold, sending relevant warning information to the intelligent terminal of medical staff.

7. A medical dispute high-risk population identification system based on a neural network, characterized in that, The system includes: A multi-dimensional information preprocessing module, which is used to obtain multi-dimensional information of a patient and perform preprocessing, where the multi-dimensional information of the patient at least includes basic information, medical records, psychological assessment, and social background; A neural network construction module, which is used to construct a deep feedforward neural network and input the preprocessed multi-dimensional information; A neural network training module, which is used to divide the preprocessed multi-dimensional information into a training set and a validation set, where the training set is used to train the deep feedforward neural network, and the validation set is used to verify and optimize the deep feedforward neural network; The early warning information sending module is used to send the early warning information output by the deep feedforward neural network to the intelligent terminal of medical staff.

8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method described in any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, the steps of the method described in any one of claims 1 to 6 are implemented.

Images