Inspection item decision-making system and method based on data processing

Abstract

The invention belongs to the field of medical information systems, and provides an examination item decision-making system and method based on data processing, and the system comprises a data collection and preprocessing module which is used for collecting and preprocessing the health data of a patient, and obtaining first health data; the examination item decision module inputs the first health data into a first model to obtain a first examination item; the data generation module is used for sequentially removing one item in the first health data and inputting the first health data into the first model to obtain a second examination item explanation module, and the data generation module is used for calculating contribution values of health features to examination items according to the first health data, the first examination items and the second examination items and outputting the contribution values to the examination items. And explaining the first inspection item according to the contribution value.

Description

Technical Field

[0001] The present invention belongs to the field of medical information systems, and in particular relates to an examination item decision system and method based on data processing. Background Art

[0002] In the existing examination item decision-making system, although data-driven models have been widely used in medical diagnosis, the problem of insufficient interpretability is still prominent. Especially when applying complex machine learning models (such as neural networks, random forests, etc.), although these models can effectively generate recommendations for examination items based on a large amount of patient health data, their black box nature makes it difficult for doctors and patients to understand the logic and basis behind the decision.

[0003] Specifically, the examination item decision models in the existing technology often only output prediction results or a list of recommended examination items, lacking a quantitative explanation of the specific contribution of each health feature in the decision-making process. For example, the model may recommend blood sugar tests and electrocardiograms, but it cannot clearly explain the specific reasons for recommending these items, whether it is because the patient's blood sugar level is abnormal, or because other factors (such as family history or age) have affected the model's prediction. This lack of explanation not only affects the doctor's clinical judgment, but also reduces the patient's trust in the system's recommendation results.

[0004] In addition, due to the complexity and multi-source heterogeneity of medical data, existing examination item decision models usually rely on the analysis of a large amount of feature data, and different features often contribute differently to examination items. However, existing systems fail to fully reveal the impact of each feature on the final recommendation. This not only limits the transparency of doctors in decision-making, but may also lead to unnecessary examination items being recommended, increasing the burden on patients and wasting medical resources. Summary of the invention

[0005] In order to solve the problems in the prior art, the present invention provides an inspection item decision system based on data processing, comprising the following modules:

[0006] A data collection and preprocessing module, used to collect and preprocess the patient's health data to obtain first health data;

[0007] An examination item decision module inputs the first health data into a first model to obtain a first examination item;

[0008] A data generation module is used to remove one item from the first health data in turn and input it into the first model to obtain a second examination item

[0009] An interpretation module is used to calculate a contribution value of a health feature to an examination item based on the first health data, a first examination item, and a second examination item, and to interpret the first examination item based on the contribution value.

[0010] Another aspect of the present invention further provides an inspection item decision method based on data processing, which uses the aforementioned system to make inspection item decisions.

[0011] Through the explanation module, the system can provide the specific contribution value of each feature to the recommended examination items based on the patient's health data. This enables doctors and patients to intuitively understand the role of each feature in the recommended examination items, enhances the interpretability of the system, and effectively solves the problem of opaque decision-making caused by the "black box" problem in existing technologies.

[0012] This system can quickly process a large amount of health data and generate accurate examination recommendations through the automated examination item decision module. Doctors no longer need to rely on manual analysis of the impact of each health feature, saving a lot of time. At the same time, the detailed contribution value explanation provided by the explanation module also helps doctors make more accurate decisions during the examination recommendation process, reducing the risk of misdiagnosis and missed diagnosis.

[0013] Through the explanation module, the system can explain to patients in detail why specific examination items are recommended, solving the problem of patients' doubts and distrust of examination item recommendations in the existing technology. Patients can clearly understand how each health characteristic (such as blood sugar level, blood pressure, family medical history, etc.) affects the recommendation results, which enhances patients' trust in the medical diagnosis system and improves the efficiency of doctor-patient communication.

[0014] In summary, the present invention significantly improves the transparency, personalization and efficiency of examination project decision-making by combining data processing technology, machine learning models and explanatory analysis modules, solves key problems in the prior art, helps to optimize the allocation of medical resources, improves patient experience, and improves the overall quality of medical diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 It is a system block diagram of the method of the present invention. DETAILED DESCRIPTION

[0017] Below, the invention is preferably described in conjunction with the accompanying drawings and specific implementation methods.

[0018] This embodiment solves the above problem through the following steps:

[0019] In one embodiment, reference Figure 1 , the present invention provides an inspection item decision system based on data processing.

[0020] Specifically, it relates to the technical field of automatically generating recommended examination items for individual patients through the collection, analysis and processing of medical data. The system mainly includes a data collection module, a data processing module, a decision module and an interpretation module.

[0021] It refers to the process of using computer systems to collect, clean, format and standardize multi-source heterogeneous data such as patients' electronic health records, laboratory test data, imaging data, medical history records, etc., to ensure the integrity, consistency and availability of the data for subsequent analysis and decision-making.

[0022] During the medical diagnosis or health check-up process, the system recommends specific diagnostic or examination items including but not limited to blood tests, imaging tests (such as CT, MRI), electrocardiogram tests, genetic tests, etc. based on the patient's clinical symptoms, medical history and other relevant factors.

[0023] By establishing a data analysis model, a computer system is used to automatically analyze and generate a list of recommended examination items based on the patient's individualized information, give patients corresponding explanations, and reduce patient misunderstandings.

[0024] Each module is further explained below.

[0025] The data collection and preprocessing module is used to collect and preprocess the patient's health data to obtain the first health data.

[0026] The data acquisition and preprocessing module is used to automatically or semi-automatically collect the patient's health-related information from multiple data sources and perform preprocessing operations on the collected data. The module includes but is not limited to a data input unit, a data cleaning unit, a data format conversion unit, and a data standardization unit.

[0027] Data collection is the process of obtaining health-related information such as patient physiological parameters, laboratory test results, imaging data, medical history records, and drug use from a variety of data sources such as medical devices, hospital information systems (HIS), electronic health records (EHR) systems, and laboratory information management systems (LIMS). Data collection can be done in a variety of ways, including real-time collection, batch upload, or data synchronization based on API interfaces.

[0028] The collected health data is cleaned, denoised, missing values ​​are filled, redundancy is eliminated, data is standardized, and format conversion is performed to ensure the integrity, consistency, and accuracy of the data. Data cleaning includes deleting outliers or invalid data, format conversion includes converting data from different systems into a unified standard format, and standardization is the unit conversion or scale normalization of numerical data for subsequent data analysis and processing.

[0029] Various data related to the patient's health status, including but not limited to physiological parameters (such as body temperature, blood pressure, heart rate), laboratory test data (such as blood routine, urine routine), imaging data (such as X-rays, CT images), medical history information (such as previous illness, allergy history), genetic information and other data information obtained through clinical examinations.

[0030] The data acquisition and preprocessing module can effectively collect and process a variety of health data from different sources, provide high-quality, structured input data for subsequent decision-making modules, and ensure the system's data processing capabilities and decision-making accuracy in complex medical scenarios.

[0031] An inspection item decision module inputs the first health data into a first model to obtain a first inspection item;

[0032] The examination item decision module is used to automatically decide and generate a personalized examination item recommendation list for the patient based on the first model and the preprocessed health data. The module includes a model loading unit, a data input unit, a model reasoning unit and a result output unit, and can dynamically select the optimal examination item according to the patient's health data.

[0033] Model loading unit: used to load and initialize the first model. This unit can dynamically load different types of machine learning models according to demand. The first model can be adjusted according to changes in the patient's health data to ensure the applicability and accuracy of the model.

[0034] Data input unit: used to receive and input pre-processed patient health data, including but not limited to the patient's age, gender, body mass index, medical history, laboratory test results and imaging data, etc. The data input unit can be connected to the data acquisition and pre-processing module to realize automatic data input.

[0035] Model reasoning unit: Analyze and reason the input patient health data based on the first model, automatically identify the patient's health status and potential health risks, and generate a personalized examination item recommendation list based on the analysis results. The reasoning process may include the use of multiple algorithm models such as classification, regression, and clustering.

[0036] Result output unit: used to output the recommended list of examination items generated by the model reasoning unit. The result output unit provides the reasoning results to doctors or patients in a structured form. The recommended examination items can be directly used for clinical decision-making and support further examination adjustment or optimization.

[0037] Preferably, the first model is a machine learning model. Specifically, the machine learning model can be a supervised learning model, an unsupervised learning model or a semi-supervised learning model, which is suitable for different types of health data analysis scenarios. Preferably, the first model is one or more of a random forest model, a logistic regression model, a support vector machine model, and a neural network model. The machine learning model can be trained on the patient's historical data and has the ability to predict and reason about new input data, ensuring that the generated examination item recommendation results can accurately match the patient's actual needs.

[0038] Further preferably, the machine learning model can dynamically adjust weights and parameters according to the health data of different patients to ensure the personalization of examination item recommendations. By allocating weights to different feature data, the system can focus on the patient's key health indicators, thereby generating more accurate examination recommendations. The model reasoning unit can generate different examination item recommendations based on the individual characteristics of different patients to achieve truly personalized medical services.

[0039] For example, historical data is used to train a random forest model, and the input data includes age, gender, blood pressure, body mass index (BMI),

[0040] Blood sugar level, heart rate, family medical history, serum cholesterol, previous diseases, etc. The output data is examination items such as blood sugar test, electrocardiogram (ECG) test, blood lipid analysis, etc.

[0041] It should be noted that all outputs of the machine learning model are probabilistic. Generally, a probability of more than 50% (or set higher) is considered to require examination. For example, the probabilities of blood sugar testing, electrocardiogram (ECG) testing, and blood lipid analysis are {0.92, 0.82, 0.52}. The first examination item output should include the predicted probability of each item.

[0042] The data generation module is used to remove one item from the first health data in turn and input it into the first model to obtain a second examination item.

[0043] In order to quantify the contribution of each feature in the recommended examination items, the features in the health data are removed one by one and input into the first model (i.e., the examination item decision model), and then the changes in the model output are observed, which provides the necessary data preparation for the calculation of the subsequent model. Specifically, the subsequent model calculates the marginal contribution of each feature by comparing the impact of the presence and absence of each feature on the model output (examination item recommendation result).

[0044] Specifically:

[0045] First, obtain the patient's health data as the input of the model. This health data set can include multiple features, such as age, gender, body mass index (BMI), blood pressure, blood sugar level, medical history, etc. Assume that the original health data is:

[0046] X={x1,x2,x3,…,x n},in

[0047] x1(age): 55 years old

[0048] x2(gender): male

[0049] x3 (blood pressure): 150 / 90 mmHg

[0050] x4 (blood sugar level): 8.5mmol / L

[0051] x5(BMI): 28

[0052] x6 (family history): Family history of diabetes

[0053] The original health data X is input into the first model, and the model outputs the recommended examination items based on these features. For example, the model may output the following recommendation result: Y = {y1, y2}

[0054] y1: Blood sugar test

[0055] y2: Electrocardiogram (ECG) examination

[0056] Next, it is necessary to remove each feature in the health data in turn, and then re-input the remaining features into the first model to generate new examination item recommendation results.

[0057] For example, first remove feature x1 (age) to get a new health data set: new ={x2,x3,x4,x5,x6}

[0058] The new dataset X new Input into the model to generate new examination item recommendation results. Assuming that after removing age, the output of the model becomes: Y new={y1}The result shows that after removing the age feature, the model only recommends blood glucose testing but does not recommend electrocardiogram examination.

[0059] Next, remove other features in turn, remove each feature (such as blood sugar level, body mass index, family medical history), and observe the changes in the model output, and record the second examination item each time a feature output is removed.

[0060] An interpretation module is used to calculate a contribution value of a health feature to an examination item based on the first health data, a first examination item, and a second examination item, and to interpret the first examination item based on the contribution value.

[0061] The explanation module aims to enhance the interpretability of the system by making the decision-making process behind the examination item recommendation transparent and providing detailed explanations to help doctors and patients understand the recommendation results. This module uses the second model to analyze the recommendation process of the examination item and generate an explanation report for each examination item.

[0062] Specifically, the interpretation module includes:

[0063] The marginal contribution of each feature to the model output is quantified by comparing the difference in model output with and without the feature. The contribution value is calculated by the following formula:

[0064] φ i =|f(X / {i})-f(X)|

[0065] where φ i Indicates the contribution value of feature i. The larger the value, the greater the influence of the feature on the prediction result. When predicting the test items, φ i Used to explain the specific contribution of a patient's health characteristic (such as blood sugar level, blood pressure, etc.) when recommending the test item.

[0066] X / {i} means excluding feature x from the input feature set X i After that, for example, by removing feature x1 (age) in the previous example, we get a new healthy data set: {x2,x3,x4,x5,x6}.

[0067] f(X) represents the predicted probability value vector obtained using the input feature set X. For example, if the probability of blood glucose testing is 0.86 and the probability of electrocardiogram (ECG) is 0.53, then f(X) = {0.86, 0.53}

[0068] f(X / {i}) means using the input feature set X to remove feature x iGet the predicted probability value vector. For example, after removing a certain feature, the probability of blood sugar testing is predicted to be 0.80, and the probability of electrocardiogram (ECG) examination is 0.41 (usually less than 0.5, no examination is performed), then f(X / {i}) = {0.80, 0.41}

[0069] |f(X / {i})-f(X)| means subtracting f(X / {i}) and f(X) term by term and taking the absolute value. Continuing with the above example, we have

[0070] |f(X / {i})-f(X)|={0.06, 0.12}, then the removed item contributes 0.06 to blood glucose testing and 0.12 to electrocardiogram (ECG) testing.

[0071] Contribution value assigns a specific contribution score to each feature, which explains why the feature affects the model's recommended inspection items.

[0072] For each inspection item, take the relevant features with a preset number of contribution value rankings before them, and explain the inspection item based on the relevant features.

[0073] For this patient's blood glucose test recommendation, the interpretation report might be as follows:

[0074] Blood glucose level x3 (Contribution value: 0.7: The patient's blood glucose level is high (8.5mmol / L), which is the main reason for recommending blood glucose testing.

[0075] Family history x5 (contribution value: 0.2): The patient has a family history of diabetes, which further increases the recommendation weight of blood glucose testing.

[0076] Age x1 (contribution value: 0.1): The patient is older, which also has a certain impact on the recommendation of blood glucose testing.

[0077] Through contribution value calculation, the explanation module can quantify the specific contribution of each health feature to the test item recommendation results. The second model not only explains how the complex model makes decisions, but also ensures that the importance of each feature to the recommendation results is fair and meaningful. In this way, doctors and patients can intuitively understand why the system recommends a specific test item and the specific role of health features in the decision-making process.

[0078] On the other hand, the present invention also discloses a method for making inspection project decisions based on data processing, which uses the system described above to make inspection project decisions.

[0079] The present invention does not particularly specify the structure of some modules, which shall be subject to the contents recorded in the prior art. The prior art mentioned in the above background technology section and the specific embodiment section of the present invention can be used as part of the present invention to understand the meaning of some technical features or parameters. The scope of protection of the present invention shall be subject to the contents actually recorded in the claims.

Claims

1. An inspection item decision system based on data processing, characterized in that: The system includes the following modules: A data collection and preprocessing module, used to collect and preprocess the patient's health data to obtain first health data; An examination item decision module inputs the first health data into a first model to obtain a first examination item; A data generation module is used to remove one item from the first health data in turn and input it into the first model to obtain a second examination item An interpretation module is used to calculate a contribution value of a health feature to an examination item based on the first health data, a first examination item, and a second examination item, and to interpret the first examination item based on the contribution value.

2. The inspection item decision system based on data processing according to claim 1 is characterized in that: The data acquisition and preprocessing module obtains the patient's physiological parameters, laboratory test results, imaging data, medical history records, and information related to drug use from medical equipment, hospital information systems, electronic health record systems, and laboratory information management systems.

3. The inspection item decision system based on data processing according to claim 1 is characterized in that: The preprocessing includes: cleaning, denoising, filling missing values, eliminating redundancy, data standardization, and format conversion.

4. The inspection item decision system based on data processing according to claim 1, characterized in that: The health data includes physiological parameters, laboratory test data, imaging data, medical history information, genetic information and clinical examination data.

5. The inspection item decision system based on data processing according to claim 1 is characterized in that: The first model is a machine learning model.

6. The inspection item decision system based on data processing according to claim 5 is characterized in that: The first model is a random forest model.

7. The inspection item decision system based on data processing according to claim 1, characterized in that: The first model is a machine learning model, and the first inspection items include a predicted probability of each item.

8. The inspection item decision system based on data processing according to claim 1, characterized in that: Explaining the first inspection item according to the contribution value includes: for each inspection item, taking relevant features of a preset number of contribution value rankings before each inspection item, and explaining the inspection item according to the relevant features.

9. A method for inspection item decision-making based on data processing, characterized in that: Use the system described in any one of claims 1 to 8 to make inspection project decisions.

Images