Rural medical service method and system based on retired doctor resources and artificial intelligence

Abstract

The invention belongs to the technical field of medical data processing, and relates to a rural medical service method and system based on retired doctor resources and artificial intelligence. The method comprises the following steps: collecting historical medical data of left-behind old people; collecting case data of all patients diagnosed and treated by all the silver-age doctors; constructing a neural network model based on deep learning; training to obtain a vector similarity model; acquiring historical medical data of the current left-behind old people and inputting the historical medical data into the vector similarity model to obtain a first vector similarity; obtaining a target doctor of the left-behind old person; obtaining case data of the current left-behind old person and case data of all patients diagnosed and treated by the target doctor, and calculating to obtain a second vector similarity; screening to obtain target case data; and selecting a final diagnosis and treatment scheme of the current patient from the diagnosis and treatment schemes corresponding to the target case data. The diagnosis and treatment doctor with the most matched diagnosis and treatment effect can be obtained, and the diagnosis and treatment efficiency and the treatment satisfaction degree of the patient can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical data processing, and in particular, relates to a rural medical service method and system based on retired doctor resources and artificial intelligence. Background Art

[0002] With the continuous development of Internet technology, telemedicine services have become an important medical service model. Patients can communicate with doctors through Internet medical platforms at home, providing more accurate diagnosis and treatment plans. At the same time, the development of Internet of Things technology and remote payment technology has also provided telemedicine with more convenient mobile payment and drug delivery methods.

[0003] The Internet medical platforms of existing medical institutions are usually aimed at urban residents, because Internet medical registration, medical treatment and payment usually require technical support such as network environment, smartphones and mobile payment. However, most of these left-behind elderly people in rural areas are very old and do not have smartphones or cannot use them proficiently. Therefore, they often cannot enjoy the convenient and efficient Internet medical services of existing medical institutions.

[0004] It is difficult for clinical doctors to find time to provide timely and satisfactory Internet medical services to more patients on Internet medical platforms. Therefore, medical institutions often arrange young doctors with low qualifications to provide Internet medical services, while the left-behind elderly in rural areas go to offline medical institutions that are closer for treatment. These offline hospitals are generally township health centers, which are short of medical equipment and find it difficult to solve the difficult and complicated diseases of the left-behind elderly in rural areas.

[0005] Internet medical doctors in existing medical institutions need to spend a lot of time asking questions in the early stages of seeing a patient due to a lack of prior knowledge of the patient's vital signs and other information. In addition, items that originally required examinations and tests on patients are also difficult to carry out under the Internet medical model. Complex inquiries reduce medical efficiency and make it difficult to meet the high expectations of Internet hospital patients for medical services. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a rural medical service method and system based on retired doctor resources and artificial intelligence.

[0007] In a first aspect, the present invention provides a rural medical service method based on retired doctor resources and artificial intelligence, including: Establish health records for left-behind elderly people and collect their historical medical data, including their basic information, medical history data, and diagnosis and treatment data; Collect the case data of all patients that each Silver Age doctor has treated, including outpatient data and inpatient data, and establish a Silver Age doctor diagnosis and treatment database; Construct a neural network model based on deep learning, input the historical medical data of left-behind elderly people and the case data of all patients who have been treated by each silver-age doctor in the silver-age doctor diagnosis and treatment database as sample data into the neural network model, extract features from the historical medical data of left-behind elderly people by the neural network model to obtain first vector data, extract features from the case data of all patients who have been treated by all silver-age doctors in the silver-age doctor diagnosis and treatment database by the neural network model to obtain second vector data, and calculate the first vector similarity between the first vector data and the second vector data; The neural network model is trained using sample data, back propagated according to the value of the loss function, and the parameters of the neural network model are adjusted to obtain a vector similarity model; Obtain the historical medical data of the current left-behind elderly and input it into the vector similarity model to obtain the first vector similarity; Filter the second vector data corresponding to the first vector similarity greater than the first set threshold, and use the silver-aged doctor corresponding to the second vector data as the target doctor for the left-behind elderly; The case data of the current left-behind elderly and the case data of all patients who have been treated by the target doctor are obtained, and the data are input into a vector similarity model. The vector similarity model performs feature extraction on the case data of the current left-behind elderly to obtain third vector data. The vector similarity model performs feature extraction on the case data of all patients who have been treated by the target doctor to obtain fourth vector data. The second vector similarity between the third vector data and the fourth vector data is calculated. Screening the second vector similarity greater than the second set threshold, and taking the case data corresponding to the second vector similarity as the target case data; Select the final treatment plan for the current patient from the treatment plans corresponding to the target case data.

[0008] In a second aspect, the present invention provides a rural medical service system based on retired doctor resources and artificial intelligence, including a historical medical data collection unit, a case data collection unit, a model building unit, a model training unit, an input unit, a first screening unit, a data processing unit, a second screening unit and a selection unit; The historical medical data collection unit is used to establish the health records of left-behind elderly people and collect their historical medical data, including their basic information, medical history data and diagnosis and treatment data; The case data collection unit is used to collect the case data of all patients that each Silver Age doctor has treated, including outpatient data and inpatient data, and to establish a Silver Age doctor diagnosis and treatment database; A model building unit is used to build a neural network model based on deep learning, and inputs the historical medical data of the left-behind elderly and the case data of all patients who have been treated by each silver-age doctor in the silver-age doctor diagnosis and treatment database as sample data into the neural network model, and the neural network model extracts features from the historical medical data of the left-behind elderly to obtain first vector data, and the neural network model extracts features from the case data of all patients who have been treated by all silver-age doctors in the silver-age doctor diagnosis and treatment database to obtain second vector data, and calculates the first vector similarity between the first vector data and the second vector data; A model training unit is used to train the neural network model using sample data, perform back propagation according to the value of the loss function, and adjust the parameters of the neural network model to obtain a vector similarity model; An input unit, used to obtain the historical medical data of the current left-behind elderly and input it into the vector similarity model to obtain a first vector similarity; A first screening unit is used to screen second vector data corresponding to the first vector similarity greater than a first set threshold, and select the silver-aged doctor corresponding to the second vector data as the target doctor for the left-behind elderly; A data processing unit is used to obtain the case data of the current left-behind elderly and the case data of all patients who have been treated by the target doctor, and input them into a vector similarity model, the vector similarity model performs feature extraction on the case data of the current left-behind elderly to obtain third vector data, the vector similarity model performs feature extraction on the case data of all patients who have been treated by the target doctor to obtain fourth vector data, and calculates the second vector similarity between the third vector data and the fourth vector data; A second screening unit, used for screening second vector similarities greater than a second set threshold, and taking case data corresponding to the second vector similarities as target case data; The selection unit is used to select a final diagnosis and treatment plan for the current patient from the diagnosis and treatment plans corresponding to the target case data.

[0009] Based on the above technical solution, the present invention can also be improved as follows.

[0010] Furthermore, feature extraction is performed on the case data of all patients who have been treated by all the silver-age doctors in the silver-age doctor diagnosis and treatment database. When the second vector data is obtained, similarity calculation is performed on the case data of all patients who have been treated by a silver-age doctor with the first vector data of the historical medical data of a left-behind elderly person to obtain several similarity coefficients, and the average value of all similarity coefficients is calculated to obtain the second vector data.

[0011] Furthermore, the neural network model based on deep learning is a ResNet network model, which includes several residual blocks; the residual block includes a main path and a shortcut connection path; the main path includes a convolution layer, a batch normalization layer and an activation function layer; the shortcut connection path connects the input to the output of the main path; the convolution layer in each residual block uses different convolution kernel sizes and step sizes to extract features; the output of the residual block passes through a global average pooling layer to convert the feature map into a feature vector of fixed length, and then connects a fully connected layer and a Softmax activation function for classification.

[0012] Furthermore, in the residual block, firstly, a convolutional layer to reduce the number of channels, followed by a The convolutional layer is used to extract features, and finally a The convolutional layer is used to restore the number of channels. The ResNet network model gradually extracts features from the input data through multiple residual blocks.

[0013] Furthermore, in the batch normalization layer, the batch data is set to , is the number of data, For the data, the mean of the data is , the data variance is , the data normalization value is , the output after the convolution operation is , is a constant, and the mean of the data is , the data variance is , the data normalization value is .

[0014] Further, the first vector data includes the first vector, and the second vector data includes the second vector; by calculating the inner product of the first vector and the second vector and dividing it by the product of the modulus of the first vector and the modulus of the second vector, a value between -1 and 1 is obtained as the calculation result of the vector similarity; if the directions of the two vectors are exactly the same, the inner product of the vectors is equal to the product of the moduli of the vectors, and the value of the vector similarity is 1; if the directions of the two vectors are completely opposite, the inner product is the negative of the product of the moduli of the vectors, and the value of the vector similarity is -1; if the two vectors are orthogonal, the inner product is 0, and the value of the vector similarity is 0.

[0015] Further, let the first vector be , the nth feature is , , the second vector is , the nth feature is , , the magnitude of the first vector is , , the magnitude of the second vector is , , the vector similarity between the first vector and the second vector is ,but: .

[0016] Furthermore, among all the target doctors, the first several target doctors are selected, and the appointment time is booked: an appointment application information is sent to the doctor with the largest second vector data. If the doctor accepts the appointment application information within the set time, the appointment application is successfully matched, and the system returns the successful appointment information to the left-behind elderly, including the treatment time and the treating doctor information. If the doctor does not accept the appointment application information after the set time, the appointment application information is sent to the target doctor with the second second vector data; if the target doctor accepts the appointment application information within the set time, the appointment application is successfully matched, and the system automatically returns the successful appointment information to the left-behind elderly, otherwise, the appointment application information is sent to the target doctor with lower second vector data until all the target doctors are applied for or the appointment application is successfully matched; if all the target doctors are applied for but the appointment application is not successfully matched, the staff will manually designate the target doctor to determine the target doctor and the successful appointment information, and the system will return the successful appointment information to the left-behind elderly.

[0017] The beneficial effects of the present invention are as follows: the present invention can collect the historical medical data of left-behind elderly people and the case data of all patients treated by doctors, extract vectors through a neural network model, calculate the condition similarity data through vector similarity, and analyze the case data treated by doctors and the patient's condition similarity data, so as to obtain the most matching doctor for treatment effect, thereby improving the patient's treatment efficiency and the patient's satisfaction with the treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A schematic diagram of a rural medical service method based on retired doctor resources and artificial intelligence provided in Example 1 of the present invention; Figure 2 This is a schematic diagram of a rural medical service system based on retired doctor resources and artificial intelligence provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0020] Example 1 As an example, Figure 1 As shown, in order to solve the above technical problems, this embodiment provides a rural medical service method based on retired doctor resources and artificial intelligence, including: Establish health records for left-behind elderly people and collect their historical medical data, including their basic information, medical history data, and diagnosis and treatment data; Collect the case data of all patients that each Silver Age doctor has treated, including outpatient data and inpatient data, and establish a Silver Age doctor diagnosis and treatment database; Construct a neural network model based on deep learning, input the historical medical data of left-behind elderly people and the case data of all patients who have been treated by each silver-age doctor in the silver-age doctor diagnosis and treatment database as sample data into the neural network model, extract features from the historical medical data of left-behind elderly people by the neural network model to obtain first vector data, extract features from the case data of all patients who have been treated by all silver-age doctors in the silver-age doctor diagnosis and treatment database by the neural network model to obtain second vector data, and calculate the first vector similarity between the first vector data and the second vector data; The neural network model is trained using sample data, back propagated according to the value of the loss function, and the parameters of the neural network model are adjusted to obtain a vector similarity model; Obtain the historical medical data of the current left-behind elderly and input it into the vector similarity model to obtain the first vector similarity; Filter the second vector data corresponding to the first vector similarity greater than the first set threshold, and use the silver-aged doctor corresponding to the second vector data as the target doctor for the left-behind elderly; The case data of the current left-behind elderly and the case data of all patients who have been treated by the target doctor are obtained, and the data are input into a vector similarity model. The vector similarity model performs feature extraction on the case data of the current left-behind elderly to obtain third vector data. The vector similarity model performs feature extraction on the case data of all patients who have been treated by the target doctor to obtain fourth vector data. The second vector similarity between the third vector data and the fourth vector data is calculated. Screening the second vector similarity greater than the second set threshold, and taking the case data corresponding to the second vector similarity as the target case data; Select the final treatment plan for the current patient from the treatment plans corresponding to the target case data.

[0021] In the actual application process, the historical medical data of hospital patients (such as past medical records, current medical history, allergy history, test results, diagnosis results and treatment plans, etc.) are input into the neural network model based on deep learning to extract features from the historical medical data of the left-behind elderly, and then the first vector data of the historical medical data of the left-behind elderly is obtained, and the case data of all patients treated by all silver-age doctors are extracted through the vector extraction model to obtain the second vector data; then the first vector data and the second vector data are calculated by vector similarity; and the vector similarity model is obtained through model training. The case data of all patients treated by the silver-age doctors and the historical medical data of the left-behind elderly are used as the input of the neural network model, and the second vector data corresponding to the first vector similarity greater than the first set threshold is screened. The silver-age doctor corresponding to the second vector data is used as the target doctor for the left-behind elderly. The silver-age doctor and the left-behind elderly can be matched according to the case data of all patients treated by the doctor, and the target doctor with better diagnosis and treatment effect can be selected, which is conducive to finding the most suitable doctor for the left-behind elderly and improving the diagnosis and treatment effect.

[0022] Optionally, feature extraction is performed on the case data of all patients who have been treated by all the silver-age doctors in the silver-age doctor diagnosis and treatment database. When the second vector data is obtained, similarity calculation is performed on the case data of all patients who have been treated by a silver-age doctor with the first vector data of the historical medical data of a left-behind elderly person to obtain several similarity coefficients, and the average of all similarity coefficients is calculated to obtain the second vector data.

[0023] Assuming that a senior doctor has treated 1,000 patients, the vector data of the case data of these 1,000 patients are respectively similarly calculated with the first vector data of the historical medical data of a left-behind elderly person to obtain 1,000 similarity coefficients, and the average value of all similarity coefficients is calculated to obtain the second vector data of the senior doctor.

[0024] Optionally, the neural network model based on deep learning is a ResNet network model, which includes several residual blocks; the residual block includes a main path and a shortcut connection path; the main path includes a convolution layer, a batch normalization layer and an activation function layer; the shortcut connection path connects the input to the output of the main path; the convolution layer in each residual block uses different convolution kernel sizes and step sizes to extract features; the output of the residual block passes through a global average pooling layer to convert the feature map into a feature vector of fixed length, and then connects a fully connected layer and a Softmax activation function for classification.

[0025] In traditional deep neural networks, as the number of network layers increases, the training error and test error of the model will first decrease and then increase. This phenomenon is called network degradation. This is not due to overfitting, because the error increases even on the training set. For example, a shallow network can fit the data well, but when more layers are stacked on this shallow network to build a deep network, the performance of the deep network may be worse than the shallow network. The deep network ResNet network model converts the learning mapping function into a learning residual function, which can easily learn a simple identity mapping, thereby avoiding the degradation problem of the deep network.

[0026] Optionally, in the residual block, first pass a convolutional layer to reduce the number of channels, followed by a The convolutional layer is used to extract features, and finally a The convolutional layer is used to restore the number of channels. The ResNet network model gradually extracts features from the input data through multiple residual blocks.

[0027] Taking the bottleneck structure as an example, let the output of the residual block be , the output calculation process of the residual block is: Assuming the input is ,go through Convolutional layer (convolution kernel is , the bias is ) is the output after ,go through Convolutional layer (convolution kernel is , the bias is ) is the output after , after the last Convolutional layer (convolution kernel is , the bias is ) is the output after , Conv represents the convolution operation. , ( represents the activation function and batch normalization operation), , , if the number of channels connected by the shortcut is equal to The final output of the residual block is ; If the number of channels is different, assume that The output of the shortcut connection after convolution adjustment is ,but . Assume that the feature vector after global average pooling is , in the fully connected layer: let is the weight matrix of the fully connected layer, is the bias vector, . Let C be the number of categories, is the output of the fully connected layer. elements, Represents input Belongs to category The probability of , then the Softmax function: . Optionally, in the batch normalization layer, let the batch data be , is the number of data, For the data, the mean of the data is , the data variance is , the data normalization value is , the output after the convolution operation is , is a constant, and the mean of the data is , the data variance is , the data normalization value is .

[0028] Optionally, the first vector data includes the first vector, and the second vector data includes the second vector; a value between -1 and 1 is obtained by calculating the inner product of the first vector and the second vector and dividing it by the product of the modulus of the first vector and the modulus of the second vector, as the calculation result of the vector similarity; if the directions of the two vectors are exactly the same, the inner product of the vectors is equal to the product of the moduli of the vectors, and the value of the vector similarity is 1; if the directions of the two vectors are completely opposite, the inner product is the negative of the product of the moduli of the vectors, and the value of the vector similarity is -1; if the two vectors are orthogonal, the inner product is 0, and the value of the vector similarity is 0.

[0029] Optionally, let the first vector be , the nth feature is , , the second vector is , the nth feature is , , the magnitude of the first vector is , , the magnitude of the second vector is , , the vector similarity between the first vector and the second vector is ,but: .

[0030] Optionally, among all target doctors, select the first several target doctors and make an appointment for the treatment time: send an appointment application information to the doctor with the largest second vector data. If the doctor accepts the appointment application information within the set time, the appointment application is successfully matched, and the system returns the successful appointment information to the left-behind elderly, including the treatment time and the treating doctor information. If the doctor does not accept the appointment application information after the set time, the appointment application information is sent to the target doctor with the second second vector data; if the target doctor accepts the appointment application information within the set time, the appointment application is successfully matched, and the system automatically returns the successful appointment information to the left-behind elderly; otherwise, the appointment application information is sent to the target doctor with lower second vector data until all target doctors are applied for appointment or the appointment application is successfully matched; if all target doctors are applied for appointment but the appointment application is not successfully matched, the staff will manually designate the target doctor to determine the target doctor and the successful appointment information, and the system will return the successful appointment information to the left-behind elderly.

[0031] When selecting the final treatment plan for the current patient from the treatment plans corresponding to the target case data, the patient's recovery status is evaluated based on the patient's biochemical indicator data, and the patient's treatment costs are counted, so as to select a treatment plan with better treatment effect and lower treatment cost.

[0032] The present invention can collect the historical medical data of left-behind elderly people and the case data of all patients treated by doctors, extract vectors through a neural network model, calculate the condition similarity data through vector similarity, and analyze the case data treated by doctors and the patient's condition similarity data, so as to obtain the most matching doctor for treatment effect, thereby improving the patient's diagnosis and treatment efficiency and the patient's satisfaction with the treatment.

[0033] Example 2 Based on the same principle as the method shown in Example 1 of the present invention, as shown in the attached Figure 2 As shown, an embodiment of the present invention also provides a rural medical service system based on retired doctor resources and artificial intelligence, including a historical medical data collection unit, a case data collection unit, a model building unit, a model training unit, an input unit, a first screening unit, a data processing unit, a second screening unit and a selection unit; The historical medical data collection unit is used to establish the health records of left-behind elderly people and collect their historical medical data, including their basic information, medical history data and diagnosis and treatment data; The case data collection unit is used to collect the case data of all patients that each Silver Age doctor has treated, including outpatient data and inpatient data, and to establish a Silver Age doctor diagnosis and treatment database; A model building unit is used to build a neural network model based on deep learning, and inputs the historical medical data of the left-behind elderly and the case data of all patients who have been treated by each silver-age doctor in the silver-age doctor diagnosis and treatment database as sample data into the neural network model, and the neural network model extracts features from the historical medical data of the left-behind elderly to obtain first vector data, and the neural network model extracts features from the case data of all patients who have been treated by all silver-age doctors in the silver-age doctor diagnosis and treatment database to obtain second vector data, and calculates the first vector similarity between the first vector data and the second vector data; A model training unit is used to train the neural network model using sample data, perform back propagation according to the value of the loss function, and adjust the parameters of the neural network model to obtain a vector similarity model; An input unit, used to obtain the historical medical data of the current left-behind elderly and input it into the vector similarity model to obtain a first vector similarity; A first screening unit is used to screen second vector data corresponding to the first vector similarity greater than a first set threshold, and select the silver-aged doctor corresponding to the second vector data as the target doctor for the left-behind elderly; A data processing unit is used to obtain the case data of the current left-behind elderly and the case data of all patients who have been treated by the target doctor, and input them into a vector similarity model, the vector similarity model performs feature extraction on the case data of the current left-behind elderly to obtain third vector data, the vector similarity model performs feature extraction on the case data of all patients who have been treated by the target doctor to obtain fourth vector data, and calculates the second vector similarity between the third vector data and the fourth vector data; A second screening unit, used for screening second vector similarities greater than a second set threshold, and taking case data corresponding to the second vector similarities as target case data; The selection unit is used to select a final diagnosis and treatment plan for the current patient from the diagnosis and treatment plans corresponding to the target case data.

[0034] Optionally, feature extraction is performed on the case data of all patients who have been treated by all the silver-age doctors in the silver-age doctor diagnosis and treatment database. When the second vector data is obtained, similarity calculation is performed on the case data of all patients who have been treated by a silver-age doctor with the first vector data of the historical medical data of a left-behind elderly person to obtain several similarity coefficients, and the average of all similarity coefficients is calculated to obtain the second vector data.

[0035] Optionally, the neural network model based on deep learning is a ResNet network model, which includes several residual blocks; the residual block includes a main path and a shortcut connection path; the main path includes a convolution layer, a batch normalization layer and an activation function layer; the shortcut connection path connects the input to the output of the main path; the convolution layer in each residual block uses different convolution kernel sizes and step sizes to extract features; the output of the residual block passes through a global average pooling layer to convert the feature map into a feature vector of fixed length, and then connects a fully connected layer and a Softmax activation function for classification.

[0036] Optionally, in the residual block, first pass a convolutional layer to reduce the number of channels, followed by a The convolutional layer is used to extract features, and finally a The convolutional layer is used to restore the number of channels. The ResNet network model gradually extracts features from the input data through multiple residual blocks.

[0037] Optionally, in the batch normalization layer, let the batch data be , is the number of data, For the data, the mean of the data is , the data variance is , the data normalization value is , the output after the convolution operation is , is a constant, and the mean of the data is , the data variance is , the data normalization value is .

[0038] Optionally, the first vector data includes the first vector, and the second vector data includes the second vector; a value between -1 and 1 is obtained by calculating the inner product of the first vector and the second vector and dividing it by the product of the modulus of the first vector and the modulus of the second vector, as the calculation result of the vector similarity; if the directions of the two vectors are exactly the same, the inner product of the vectors is equal to the product of the moduli of the vectors, and the value of the vector similarity is 1; if the directions of the two vectors are completely opposite, the inner product is the negative of the product of the moduli of the vectors, and the value of the vector similarity is -1; if the two vectors are orthogonal, the inner product is 0, and the value of the vector similarity is 0.

[0039] Optionally, let the first vector be , the nth feature is , , the second vector is , the nth feature is , , the magnitude of the first vector is , , the magnitude of the second vector is , , the vector similarity between the first vector and the second vector is ,but: .

[0040] Optionally, among all target doctors, select the first several target doctors and make an appointment for the treatment time: send an appointment application information to the doctor with the largest second vector data. If the doctor accepts the appointment application information within the set time, the appointment application is successfully matched, and the system returns the successful appointment information to the left-behind elderly, including the treatment time and the treating doctor information. If the doctor does not accept the appointment application information after the set time, the appointment application information is sent to the target doctor with the second second vector data; if the target doctor accepts the appointment application information within the set time, the appointment application is successfully matched, and the system automatically returns the successful appointment information to the left-behind elderly; otherwise, the appointment application information is sent to the target doctor with lower second vector data until all target doctors are applied for appointment or the appointment application is successfully matched; if all target doctors are applied for appointment but the appointment application is not successfully matched, the staff will manually designate the target doctor to determine the target doctor and the successful appointment information, and the system will return the successful appointment information to the left-behind elderly.

[0041] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A rural medical service method based on retired doctors’ resources and artificial intelligence, characterized by: include: Establish health records for left-behind elderly people and collect their historical medical data, including their basic information, medical history data, and diagnosis and treatment data; Collect the case data of all patients that each Silver Age doctor has treated, including outpatient data and inpatient data, and establish a Silver Age doctor diagnosis and treatment database; Construct a neural network model based on deep learning, input the historical medical data of left-behind elderly people and the case data of all patients who have been treated by each silver-age doctor in the silver-age doctor diagnosis and treatment database as sample data into the neural network model, extract features from the historical medical data of left-behind elderly people by the neural network model to obtain first vector data, extract features from the case data of all patients who have been treated by all silver-age doctors in the silver-age doctor diagnosis and treatment database by the neural network model to obtain second vector data, and calculate the first vector similarity between the first vector data and the second vector data; The neural network model is trained using sample data, back propagated according to the value of the loss function, and the parameters of the neural network model are adjusted to obtain a vector similarity model; Obtain the historical medical data of the current left-behind elderly and input it into the vector similarity model to obtain the first vector similarity; Filter the second vector data corresponding to the first vector similarity greater than the first set threshold, and use the silver-aged doctor corresponding to the second vector data as the target doctor for the left-behind elderly; The case data of the current left-behind elderly and the case data of all patients who have been treated by the target doctor are obtained, and the data are input into a vector similarity model. The vector similarity model performs feature extraction on the case data of the current left-behind elderly to obtain third vector data. The vector similarity model performs feature extraction on the case data of all patients who have been treated by the target doctor to obtain fourth vector data. The second vector similarity between the third vector data and the fourth vector data is calculated. Screening the second vector similarity greater than the second set threshold, and taking the case data corresponding to the second vector similarity as the target case data; Select the final treatment plan for the current patient from the treatment plans corresponding to the target case data.

2. The rural medical service method based on retired doctor resources and artificial intelligence according to claim 1 is characterized in that: Feature extraction is performed on the case data of all patients who have been treated by all the silver-age doctors in the silver-age doctor diagnosis and treatment database to obtain the second vector data. The case data of all patients who have been treated by a silver-age doctor are respectively similarly calculated with the first vector data of the historical medical data of a left-behind elderly person to obtain several similarity coefficients, and the average value of all similarity coefficients is calculated to obtain the second vector data.

3. The rural medical service method based on retired doctor resources and artificial intelligence according to claim 1 is characterized in that: The neural network model based on deep learning includes the ResNet network model, which includes several residual blocks; the residual block includes a main path and a shortcut connection path; the main path includes a convolution layer, a batch normalization layer and an activation function layer; the shortcut connection path connects the input to the output of the main path; the convolution layer in each residual block uses different convolution kernel sizes and step sizes to extract features; the output of the residual block passes through a global average pooling layer to convert the feature map into a feature vector of fixed length, and then connects a fully connected layer and a Softmax activation function for classification.

4. The rural medical service method based on retired doctors' resources and artificial intelligence according to claim 3 is characterized in that: In the residual block, we first pass a convolutional layer to reduce the number of channels, followed by a The convolutional layer is used to extract features, and finally a The convolutional layer is used to restore the number of channels; the ResNet network model gradually extracts features from the input data through multiple residual blocks.

5. The rural medical service method based on retired doctors' resources and artificial intelligence according to claim 3 is characterized in that: In the batch normalization layer, the batch data is set to , is the number of data, For the data, the mean of the data is , the data variance is , the data normalization value is , the output after the convolution operation is , is a constant, and the mean of the data is , the data variance is , the data normalization value is .

6. The rural medical service method based on retired doctors resources and artificial intelligence according to claim 1 is characterized in that: The first vector data includes the first vector, and the second vector data includes the second vector; by calculating the inner product of the first vector and the second vector and dividing it by the product of the modulus of the first vector and the modulus of the second vector, a value between -1 and 1 is obtained as the calculation result of the vector similarity; if the directions of the two vectors are exactly the same, the inner product of the vectors is equal to the product of the moduli of the vectors, and the value of the vector similarity is 1; if the directions of the two vectors are completely opposite, the inner product is the negative of the product of the moduli of the vectors, and the value of the vector similarity is -1; if the two vectors are orthogonal, the inner product is 0, and the value of the vector similarity is 0.

7. The rural medical service method based on retired doctors resources and artificial intelligence according to claim 6 is characterized in that: The first vector is , the nth feature is , , the second vector is , the nth feature is , , the magnitude of the first vector is , , the magnitude of the second vector is , , the vector similarity between the first vector and the second vector is ,but: 。 8. The rural medical service method based on retired doctors resources and artificial intelligence according to claim 1 is characterized in that: Among all the target doctors, select the first several target doctors and make an appointment for the treatment time: send an appointment application information to the doctor with the largest second vector data. If the doctor accepts the appointment application information within the set time, the appointment application is successfully matched, and the system returns the successful appointment information to the left-behind elderly, including the treatment time and the treating doctor information. If the doctor does not accept the appointment application information after the set time, the appointment application information is sent to the target doctor with the second second vector data; if the target doctor accepts the appointment application information within the set time, the appointment application is successfully matched, and the system automatically returns the successful appointment information to the left-behind elderly, otherwise, the appointment application information is sent to the target doctor with lower second vector data until all the target doctors are applied for or the appointment application is successfully matched; if all the target doctors are applied for but the appointment application is not successfully matched, the staff will manually designate the target doctor to determine the target doctor and the successful appointment information, and the system will return the successful appointment information to the left-behind elderly.

9. A rural medical service system based on retired doctors’ resources and artificial intelligence, characterized by: It includes a historical medical data collection unit, a case data collection unit, a model building unit, a model training unit, an input unit, a first screening unit, a data processing unit, a second screening unit and a selection unit; The historical medical data collection unit is used to establish the health records of left-behind elderly people and collect their historical medical data, including their basic information, medical history data and diagnosis and treatment data; The case data collection unit is used to collect the case data of all patients that each Silver Age doctor has treated, including outpatient data and inpatient data, and to establish a Silver Age doctor diagnosis and treatment database; A model building unit is used to build a neural network model based on deep learning, and inputs the historical medical data of the left-behind elderly and the case data of all patients who have been treated by each silver-age doctor in the silver-age doctor diagnosis and treatment database as sample data into the neural network model, and the neural network model extracts features from the historical medical data of the left-behind elderly to obtain first vector data, and the neural network model extracts features from the case data of all patients who have been treated by all silver-age doctors in the silver-age doctor diagnosis and treatment database to obtain second vector data, and calculates the first vector similarity between the first vector data and the second vector data; A model training unit is used to train the neural network model using sample data, perform back propagation according to the value of the loss function, and adjust the parameters of the neural network model to obtain a vector similarity model; An input unit, used to obtain the historical medical data of the current left-behind elderly and input it into the vector similarity model to obtain a first vector similarity; A first screening unit is used to screen second vector data corresponding to the first vector similarity greater than a first set threshold, and select the silver-aged doctor corresponding to the second vector data as the target doctor for the left-behind elderly; A data processing unit is used to obtain the case data of the current left-behind elderly and the case data of all patients who have been treated by the target doctor, and input them into a vector similarity model, the vector similarity model performs feature extraction on the case data of the current left-behind elderly to obtain third vector data, the vector similarity model performs feature extraction on the case data of all patients who have been treated by the target doctor to obtain fourth vector data, and calculates the second vector similarity between the third vector data and the fourth vector data; A second screening unit, used for screening second vector similarities greater than a second set threshold, and taking case data corresponding to the second vector similarities as target case data; The selection unit is used to select a final diagnosis and treatment plan for the current patient from the diagnosis and treatment plans corresponding to the target case data.

Images