Medical data query method and system based on electronic medical record

Abstract

The invention relates to the field of medical data query, in particular to a medical data query method and system based on an electronic medical record, and the system comprises a user interface operation module, a data collection and processing module, a modeling module and a disease risk prediction module. According to the system, firstly, a user interface operation module designs a user interface by utilizing a front-end and back-end development technology and is used for collecting user input data and inquiring medical record data of a patient according to the input data, and a data collection and processing module performs feature conversion on the inquired data and collects historical medical data; the modeling module is used for carrying out feature conversion and label processing on the historical medical data, carrying out model training according to the historical medical data by using a decision tree algorithm and sending a trained model to the disease risk prediction module; and the disease risk prediction module performs disease risk prediction on the query data by using the trained model, and sends a prediction result to the user interface operation module for visual output.

Description

Technical Field

[0001] The present invention relates to the field of medical data query, and specifically to a medical data query method and system based on electronic medical records. Background Art

[0002] Medical data query methods often involve doctors querying and screening paper medical records to determine patients' medical data. However, traditional medical data query methods have some drawbacks.

[0003] On the one hand, traditional medical data query methods rely on patients' paper medical records when querying medical data, which not only consumes a large amount of time but also may result in data loss due to damage to the paper medical records.

[0004] On the other hand, traditional medical data query methods require doctors to judge patients' disease risks based on medical records and cannot predict the risk of a specific disease in real time, affecting the treatment time. Summary of the Invention

[0005] The purpose of the present invention is to provide a medical data query method and system based on electronic medical records to solve the problems raised in the above background art.

[0006] To achieve the above purpose, one of the purposes of the present invention is to provide a medical data query system based on electronic medical records, which includes a user interface operation module, a data collection and processing module, a modeling module, and a disease risk prediction module, wherein:

[0007] The user interface operation module designs a user interface using front-end and back-end development technologies, is used to collect user input data, query patient medical record data according to the input data, and send the queried patient data to the data collection and processing module;

[0008] The data collection and processing module performs feature transformation on the query data, collects historical medical data, performs feature transformation and label processing on the historical medical data, and sends the processed query data and historical medical data to the modeling module and the disease risk prediction module respectively;

[0009] The modeling module uses the decision tree algorithm to train a model based on historical medical data and sends the trained model to the disease risk prediction module;

[0010] The disease risk prediction module uses the model trained by the modeling module to predict the disease risk of the query data sent by the data collection and processing module and sends the prediction result to the user interface operation module for visual output.

[0011] As a further improvement of this technical solution, the user interface operation module includes a front-end and back-end development unit and a query data receiving unit. The front-end and back-end development unit designs a user interface using front-end and back-end development technologies for users to input data and query patient medical record data, and sends the queried patient data to the query data receiving unit. The query data receiving unit is used to receive the query data sent by the front-end and back-end development unit and send the query data to the feature conversion unit in the data collection and processing module.

[0012] As a further improvement of this technical solution, the data collection and processing module includes a historical data collection unit, a feature conversion unit, and a label processing unit. The historical data collection unit is used to collect historical medical data and send the data to the feature conversion unit. The feature conversion unit receives the query data and historical medical data sent by the query data receiving unit and the historical data collection unit respectively, and is used to convert the data into the model input type, and send the query data and historical medical data to the disease risk prediction unit and the label processing unit in the disease risk prediction module respectively. The label processing unit is used to use the diseases suffered by the patients in the historical medical data as the label column of the data and send the processed data to the historical data receiving unit in the modeling module.

[0013] As a further improvement of this technical solution, the modeling module includes a historical data receiving unit and a model training unit. The historical data receiving unit receives the historical medical data sent by the label processing unit, is used to standardize the historical medical data, and sends the historical medical data to the model training unit. The model training unit uses the decision tree algorithm to perform model training based on the historical medical data sent by the historical data receiving unit and sends the trained model to the disease risk prediction unit in the disease risk prediction module.

[0014] As a further improvement of this technical solution, the disease risk prediction module includes a prediction data receiving unit and a disease risk prediction unit. The prediction data receiving unit receives the query data sent by the feature conversion unit and is used to send the query data to the disease risk prediction unit. The disease risk prediction unit uses the model trained by the model training unit to predict the disease risk of the query data sent by the prediction data receiving unit and sends the prediction result to the disease risk visualization unit.

[0015] As a further improvement of this technical solution, the user interface operation module includes a disease risk visualization unit. The disease risk visualization unit receives the prediction result sent by the disease risk prediction unit, converts the prediction result into text numerical type data, and uses a visualization tool to perform visualization operations on the data and send it to the user's front-end page.

[0016] As a further improvement of the technical solution, the front-end and back-end development unit designs a user page through front-end interface design, front-end operation, back-end processing, back-end database operation, and front-end and back-end data interaction.

[0017] As a further improvement of the technical solution, the feature conversion unit converts the numerical type of the image data, and uses the feature points and feature descriptors in the image as the feature columns of the image data.

[0018] The second object of the present invention is to provide a method for a medical data query system based on electronic medical records, including the following method steps

[0019] S1. The user interface operation module uses front-end and back-end development technologies to design a user interface for collecting user input data, querying patient medical record data according to the input data, and sending the queried patient data to the data collection and processing module;

[0020] S2. The data collection and processing module performs feature conversion on the query data, collects historical medical data, and performs feature conversion and label processing on the historical medical data;

[0021] S3. The modeling module uses the decision tree algorithm to perform model training based on historical medical data;

[0022] S4. The disease risk prediction module uses the trained model to predict the disease risk of the query data, and sends the prediction result to the user interface operation module for visual output.

[0023] Compared with the prior art, the beneficial effects of the present invention are:

[0024] 1. The medical data query method and system based on electronic medical records uses front-end interface design, front-end operation, back-end processing, back-end database operation, and front-end and back-end data interaction to design a user page. Medical staff can query the medical record data of patients according to the page and input data, which not only saves the time for querying paper medical records, but also greatly reduces the possibility of data loss caused by the damage of paper data.

[0025] 2. The medical data query method and system based on electronic medical records uses the decision tree algorithm to train the model according to the historical medical data in each hospital database, and then uses the trained model to predict the disease risk of the patient data queried by medical staff on the user page, so as to predict the disease suffered by the patient in real time, and send the prediction result to the medical staff for reference, saving the time cost for medical staff to judge the patient's disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic diagram of the overall module of the present invention;

[0027] Figure 2 Schematic diagrams of various modular units of the present invention;

[0028] Figure 3 Schematic diagram of the overall method flow element of the present invention;

[0029] In the figure: 100, user interface operation module; 101, front-end and back-end development unit; 102, query data receiving unit; 103, disease risk visualization unit; 200, data collection and processing module; 201, historical data collection unit; 202, feature transformation unit; 203, label processing unit; 300, modeling module; 301, historical data receiving unit; 302, model training unit; 400, disease risk prediction module; 401, prediction data receiving unit; 402, disease risk prediction unit. Specific embodiments

[0030] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0031] Please refer to Figures 1-3 , one of the purposes of this embodiment is that a medical data query system based on electronic medical records includes a user interface operation module 100, a data collection and processing module 200, a modeling module 300, and a disease risk prediction module 400.

[0032] For the convenience of doctors to query patients' medical data, the front-end and back-end development unit 101 in the user interface operation module 100 will design a user page using front-end and back-end development technologies, specifically including:

[0033] Front-end interface design: Create a form containing input fields for name and card number so that users can enter query conditions; Add a query button, and when the user clicks this button, a query operation is triggered; Display blank or placeholder text boxes for data such as patients' basic information, symptom descriptions, medical examination results, and medication records on the interface;

[0034] Front-end operations: Obtain the name and card number information entered by the user in the front-end interface; Send the obtained query condition data to the back-end server and use AJAX for data transmission; Receive the medical data returned by the back-end and fill the data into the corresponding front-end text boxes for display;

[0035] Backend processing: Receive the name and card number information sent by the frontend; according to the received query conditions, perform database query operations on the backend to obtain relevant data of the corresponding patient; organize the query results into a JSON data format and return them to the frontend;

[0036] Backend database operation: According to the query conditions, write the corresponding database query statement to obtain data such as the patient's basic information, symptom description, medical examination results, and medication records from the database; process the query results and extract the required data to prepare for returning to the frontend;

[0037] Frontend-backend data interaction and display: The backend returns the processed data to the frontend and uses JSON or other formats for data transmission; the frontend receives the data returned by the backend and fills the data into the corresponding text boxes to display data such as the patient's basic information, symptom description, medical examination results, and medication records.

[0038] The front-end and back-end development unit 101 sends the queried data to the query data receiving unit 102, and the query data receiving unit 102 is used to receive the queried data and send the query data to the feature conversion unit 202 in the data collection module.

[0039] To make the data conform to the input data type of the model, it is necessary to perform feature conversion on the query data, specifically including:

[0040] Gender: Convert gender data into binary data, that is, 0 represents female and 1 represents male;

[0041] Pain location, medical history, surgical history, allergy history, medication use, and skin condition: Convert a categorical feature with n possible values into n binary features, each feature having only two possible values (0 or 1). For a certain sample, the feature can be repeatedly set to 1, and the remaining features are set to 0. For example, if the pain locations include the head, chest, waist, and throat, then [1, 0, 0, 0] represents head pain; [0, 1, 0, 0] represents chest pain; [1, 1, 0, 0] represents head and chest pain;

[0042] ECG, X-ray, MRI: The feature points and feature descriptors in the extracted images are used as the feature columns of the data. For feature points, first convert the image into a grayscale image, and select a pixel point in the grayscale image as the current pixel point; select a circular neighborhood around the current pixel point, and the range of the neighborhood is 16 pixel points. Number the pixel points in the order of symmetric positions, and calculate the grayscale difference between the current pixel point and the pixel points in the neighborhood; if there are more than 12 consecutive pixel points whose grayscale values are greater than or less than the grayscale value of the current pixel point plus or minus the threshold of 20, then the current pixel point is considered a corner point; detect each pixel point in the image according to the above steps, and return the detected corner points as feature points; for feature descriptors, first select a circular area with a radius of 10 pixels as the calculation range, and uniformly select 16 sampling points within the selected circular area; for each sampling point, calculate the grayscale centroid of the pixels in the 3x3 area, and obtain the main rotation direction of the feature point according to the grayscale centroid information; select a set of sampling point pairs in the area around the feature point, where the number of sampling point pairs is set to 128, calculate the difference between the grayscale values of the two sampling points, convert the continuous grayscale values into binary codes, and combine the binary codes of all sampling point pairs to form a descriptor; the specific steps are as follows:

[0043] Feature point detection:

[0044] a. Convert the image into a grayscale image: Let I(x, y) represent the grayscale value of the pixel (x, y) in the original image, and G(x, y) represent the grayscale image;

[0045] b. Define the neighborhood radius R as 16 pixels, and the coordinates of the pixel points in the neighborhood are (x_i, y_i), where i = 1, 2,..., 16;

[0046] c. Calculate the grayscale difference of each pixel point (x, y):

[0047] ΔG(x, y, i) = |G(x, y) - G(x_i, y_i)|

[0048] d. Define a threshold T = 20;

[0049] e. Calculate the condition for a pixel point to be a corner point:

[0050] If there are 12 consecutive pixel points i that satisfy ΔG(x, y, i) > T or ΔG(x, y, i) < -T, then (x, y) is considered a corner point;

[0051] Feature descriptor generation:

[0052] a. Select a circular area with a radius of 10 pixels, centered at (x, y);

[0053] b. Uniformly select 16 sampling points with coordinates (x_i, y_i), where i = 1, 2,..., 16;

[0054] c. Calculate the main rotation direction θ of the feature point:

[0055] θ = atan2(∑i(G(x_i, y_i)*sin(2θ_i)), ∑i(G(x_i, y_i)*cos(2θ_i)))

[0056] d. Select 128 pairs of sampling points around the feature point, where the coordinates of each pair of sampling points are (x_i, y_i) and (x_j, y_j), where i, j = 1, 2,..., 128;

[0057] e. Calculate the gray difference of each pair of sampling points:

[0058] ΔG(i, j) = G(x_i, y_i) - G(x_j, y_j)

[0059] f. Convert the continuous gray values into binary codes:

[0060] B(i, j) = 1 if ΔG(i, j) > 0, otherwise B(i, j) = 0;

[0061] g. Combine the binary codes of all sampling point pairs to form a descriptor:

[0062] D = [B(1, 1), B(1, 2),..., B(128, 128)].

[0063] The feature conversion unit 202 sends the converted query data to the prediction data receiving unit 401 in the disease risk prediction module 400. The prediction data receiving unit 401 is used to receive the converted query data and send the data to the disease prediction risk unit for disease risk prediction.

[0064] Before prediction, a model needs to be established, so historical data needs to be collected for model training. The historical data collection unit 201 in the data collection and processing module 200 collects historical medical data using the public application interfaces in each hospital database. The historical medical data includes the queried data and the diseases suffered by patients. The collected historical medical data is sent to the feature conversion unit 202 for feature conversion, and the converted data is sent to the label processing unit 203. The label processing unit 203 takes the diseases suffered by patients in the data as the label column of the data, so that the model can be trained supervised, and converts the diseases suffered by patients into numerical types. Each different value of the categorical feature is mapped to an integer code, and each value is assigned a unique integer number to represent the category. For example, the disease classification includes heart disease, hypertension, diabetes, asthma, and fracture. Then 0 represents heart disease; 1 represents hypertension; 2 represents diabetes; 3 represents asthma; 4 represents fracture.

[0065] The label processing unit 203 sends the processed data to the historical data receiving unit 301 in the modeling module 300. To reduce the impact of outliers and noise on the model, improve the robustness of the model to outliers, reduce the interference with model training and prediction, and thus improve the performance and stability of the model, by calculating the mean and standard deviation of the data feature columns, subtracting the mean from the original value and then dividing by the standard deviation, the data is converted into a distribution with a mean of 0 and a standard deviation of 1, keeping the feature weights balanced, and sending the standardized data to the model training unit 302.

[0066] The model training unit 302 uses the decision tree algorithm in machine learning to train the model. The training process is as follows:

[0067] Data partitioning: The data set is split into two parts, 80% is the training set and 20% is the test set. The training set is used for model training, and the test set is used for model evaluation.

[0068] Model training: Select a feature with the smallest Gini coefficient in the data as the root node. The calculation formula of the Gini coefficient is: Gini = 1 - ∑(pi)2, where p is the sample proportion of each category in a node, and Σ represents the summation over all categories.

[0069] Next, for each child node under the root node, continue to partition according to the size of the Gini coefficient, and iterate continuously until the node reaches a pure state, that is, the samples in the node belong to the same category, then stop growing.

[0070] Model evaluation: Calculate the number of samples accurately predicted between the label array predicted by the model and the label array of the test set. Divide the number of accurately predicted samples by the number of samples in the test set and multiply by the percentage to obtain the accuracy rate. When the accuracy rate is greater than or equal to 90%, the model training is successful, and the trained model is sent to the disease risk prediction unit 402 for model application; otherwise, model tuning is performed.

[0071] Model tuning: For each leaf node, calculate the difference in prediction performance on the test set before and after pruning, that is, the change in the accuracy rate.

[0072] If the performance does not decrease after pruning, prune this leaf node to become the parent node of a leaf node, and predict this node as the category with the highest frequency on the test set; otherwise, judge the performance after pruning of the next leaf node.

[0073] Prune step by step upwards until no more pruning can be done, that is, until the performance after pruning of all leaf nodes results in an accuracy rate < 90%.

[0074] The disease risk prediction unit 402 predicts the disease risk for the prediction data receiving unit 401 based on the model trained by the model training unit 302. The prediction process is as follows:

[0075] Starting from the root node, according to the value of the feature, select the corresponding branch according to the decision condition.

[0076] After entering the next node, according to the decision condition of the current node and the value of the feature, select the corresponding branch to enter the next node.

[0077] From the root node to the leaf node, keep making branch selections according to the value of the feature until reaching the leaf node, and the leaf node corresponds to the value of the disease risk.

[0078] Finally, output the predicted disease risk label value, that is, the value corresponding to the leaf node.

[0079] The decision tree pruning process is as follows:

[0080] For each leaf node L, calculate the difference in prediction performance on the test set before and after pruning:

[0081] A. ΔAccuracy(L) = Accuracy(L) - Accuracy_after_pruning(L)

[0082] B. If the performance does not decrease after pruning (ΔAccuracy(L) >= 0), then prune this leaf node to become the parent node of a leaf node, and predict this node as the category with the highest frequency on the test set. Let P(L) be the parent node of node L.

[0083] C. Otherwise, judge the performance after pruning of the next leaf node.

[0084] D. Prune step by step upwards until no more pruning can be done, that is, stop when the performance after pruning of all leaf nodes results in an accuracy rate < 90%.

[0085] The disease risk prediction process is as follows:

[0086] A. Starting from the root node, according to the value of the feature value, select the corresponding branch according to the judgment condition. Let R represent the root node of the decision tree.

[0087] B. After entering the next node, according to the judgment condition of the current node and the value of the feature value, select the corresponding branch to enter the next node. Let N represent the current node.

[0088] C. From the root node to the leaf node, continuously perform branch selection according to the value of the feature value until reaching the leaf node L, where L corresponds to the value of the disease risk.

[0089] D. Finally, output the predicted disease risk label value, that is, the value corresponding to the leaf node L.

[0090] The disease risk prediction unit 402 sends the predicted result to the disease risk visualization unit 103. The disease risk visualization unit 103 converts the predicted result into text numerical type data and performs visualization operations on the data using a visualization tool and sends it to the front-end page of the user for medical staff to refer to.

[0091] The second object of this embodiment lies in a method for a medical data query system based on electronic medical records, including the following method steps:

[0092] S1. The user interface operation module 100 designs a user interface using front-end and back-end development technologies, which is used to collect user input data, query patient medical record data according to the input data, and send the queried patient data to the data collection and processing module 200.

[0093] S2. The data collection and processing module 200 performs feature conversion on the query data, collects historical medical data, and performs feature conversion and label processing on the historical medical data.

[0094] S3. The modeling module 300 uses the decision tree algorithm to perform model training according to historical medical data.

[0095] S4. The disease risk prediction module 400 uses the trained model to predict the disease risk of the query data and sends the predicted result to the user interface operation module 100 for visual output.

[0096] The foregoing has shown and described the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. A medical data query system based on electronic medical records, characterized by: It comprises a user interface operation module (100), a data collection and processing module (200), a modeling module (300) and a disease risk prediction module (400), wherein: The user interface operation module (100) uses front-end and back-end development technologies to design a user interface, and is used to collect user input data, query patient medical record data based on the input data, and send the queried patient data to the data collection and processing module (200); The data collection and processing module (200) performs feature conversion on the query data, collects historical medical data, performs feature conversion and label processing on the historical medical data, and sends the processed query data and historical medical data to the modeling module (300) and the disease risk prediction module (400) respectively; The modeling module (300) uses a decision tree algorithm to perform model training based on historical medical data, and sends the trained model to the disease risk prediction module (400); The disease risk prediction module (400) uses the model trained by the modeling module (300) to predict the disease risk of the query data sent by the data collection and processing module (200), and sends the prediction result to the user interface operation module (100) for visual output.

2. The medical data query system based on electronic medical records according to claim 1, characterized in that: The user interface operation module (100) comprises a front-end and back-end development unit (101) and a query data receiving unit (102). The front-end and back-end development unit (101) uses front-end and back-end development technologies to design a user interface for a user to input data and query patient medical record data, and sends the queried patient data to the query data receiving unit (102); the query data receiving unit (102) is used to receive the query data sent by the front-end and back-end development unit (101), and send the query data to the feature conversion unit (202) in the data collection and processing module (200).

3. The medical data query system based on electronic medical records according to claim 1, characterized in that: The data collection and processing module (200) comprises a historical data collection unit (201), a feature conversion unit (202) and a label processing unit (203); the historical data collection unit (201) is used to collect historical medical data and send the data to the feature conversion unit (202); the feature conversion unit (202) receives the query data and historical medical data sent by the query data receiving unit (102) and the historical data collection unit (201) respectively, is used to convert the data into a model input type, and sends the query data and historical medical data to the disease risk prediction unit (402) and the label processing unit (203) in the disease risk prediction module (400) respectively; the label processing unit (203) is used to use the disease suffered by the patient in the historical medical data as a label column of the data, and send the processed data to the historical data receiving unit (301) in the modeling module (300).

4. The medical data query system based on electronic medical records according to claim 1, characterized in that: The modeling module (300) comprises a historical data receiving unit (301) and a model training unit (302); the historical data receiving unit (301) receives historical medical data sent by the label processing unit (203), is used to perform standardization processing on the historical medical data, and sends the historical medical data to the model training unit (302); the model training unit (302) uses a decision tree algorithm to perform model training based on the historical medical data sent by the historical data receiving unit (301), and sends the trained model to the disease risk prediction unit (402) in the disease risk prediction module (400).

5. The medical data query system based on electronic medical records according to claim 1, characterized in that: The disease risk prediction module (400) comprises a prediction data receiving unit (401) and a disease risk prediction unit (402); the prediction data receiving unit (401) receives query data sent by the feature conversion unit (202) and is used to send the query data to the disease risk prediction unit (402); the disease risk prediction unit (402) uses the model trained by the model training unit (302) to predict the disease risk of the query data sent by the prediction data receiving unit (401), and sends the prediction result to the disease risk visualization unit (103).

6. The medical data query system based on electronic medical records according to claim 1, characterized in that: The user interface operation module (100) comprises a disease risk visualization unit (103), wherein the disease risk visualization unit (103) receives the prediction result sent by the disease risk prediction unit (402), converts the prediction result into textual numerical type data, and uses a visualization tool to perform a visualization operation on the data and sends the data to a user's front-end page.

7. The medical data query system based on electronic medical records according to claim 2, characterized in that: The front-end and back-end development unit (101) designs the user page through front-end interface design, front-end operation, back-end processing, back-end database operation and interaction of front-end and back-end data.

8. The medical data query system based on electronic medical records according to claim 3, characterized in that: The feature conversion unit (202) performs numerical type conversion on the image data, and uses the feature points and feature descriptors in the image as feature columns of the image data.

9. A method for using the medical data query system based on electronic medical records according to any one of claims 1 to 8, characterized in that: The method comprises the following steps: S1, the user interface operation module (100) uses the front-end and back-end development technologies to design a user interface for collecting user input data, querying patient medical record data based on the input data, and sending the queried patient data to the data collection and processing module (200); S2, the data collection and processing module (200) performs feature conversion on the query data, collects historical medical data, and performs feature conversion and label processing on the historical medical data; S3, the modeling module (300) uses a decision tree algorithm to perform model training based on historical medical data; S4. The disease risk prediction module (400) uses the trained model to predict the disease risk of the query data, and sends the prediction result to the user interface operation module (100) for visual output.

Images