Pulmonary embolism index prediction method and device based on blood examination and electronic equipment
By constructing a blood test-based prediction method for pulmonary embolism indexes in primary medical institutions, using machine learning models to integrate biochemical index data, the accuracy and automation of pulmonary embolism diagnosis in primary medical institutions are solved, and the accuracy and universality of early predictions are achieved.
Patent Information
- Application Number
- CN202510217865.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-26
- Publication Date
- 2025-07-04
AI Technical Summary
The lack of CT equipment and professional doctors in primary medical institutions has made it difficult to diagnose pulmonary embolism. It is difficult for existing technology to use the existing blood test conditions of primary medical institutions to accurately predict pulmonary embolism in early stages.
A pulmonary embolism index prediction method based on blood tests is constructed. By constructing a machine learning model for each examination item, using biochemical index data to train the model, fusing the prediction results of each examination item to generate the final pulmonary embolism prediction results.
It realizes the accuracy and automation of early prediction of pulmonary embolism in primary medical institutions, reduces dependence on high-quality imaging equipment and senior doctors, and improves the robustness and universality of predictions.
Smart Images
Figure CN120260899A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of artificial intelligence, and more particularly, relates to a method, apparatus, and electronic device for predicting pulmonary embolism indicators based on blood tests. Background Art
[0002] Pulmonary Embolism (PE) is one of the three common causes of global cardiovascular death. The incidence and mortality of pulmonary embolism are relatively high, and the clinical manifestations are less specific, resulting in a high misdiagnosis rate, missed diagnosis rate, and fatality rate. At the same time, imaging examinations are time-consuming and expensive. In primary hospitals, it is difficult to carry out examinations due to the lack of CT (Computed Tomography) equipment and professional doctors. In addition, some patients have hemodynamic instability and are not suitable for CT pulmonary angiography (CTPA), which also increases the difficulty of diagnosis and treatment for clinicians, delays the best treatment time for patients, leads to poor later treatment effects, and even clinical complications. Therefore, early indicator analysis and timely intervention for pulmonary embolism are extremely important.
[0003] The existing technologies are mainly divided into two categories: one is to use artificial intelligence algorithms to analyze the CT lung images of patients to achieve the diagnosis of PE. However, there are currently problems with this method: it not only requires expensive CT equipment, but also requires professional doctors to operate the CT imaging and conduct detailed manual review and examination of the artificial intelligence diagnosis results. These reasons make it difficult for primary medical institutions to carry out this method. The other is to use biomarker detection reagents with good sensitivity and specificity to conduct differential diagnosis on patients. However, it is currently difficult to find differential diagnosis reagents with good sensitivity and specificity on the market, and it is even more difficult for primary medical institutions to achieve clinical use in a short period of time.
[0004] Therefore, how to construct an artificial intelligence model using the biochemical indicators obtained from the existing relatively well-equipped blood test conditions in primary medical institutions to predict the pulmonary embolism indicators of patients has become an urgent problem to be solved. Summary of the Invention
[0005] To solve the above technical problems, the present invention provides a method, apparatus, and electronic device for predicting pulmonary embolism indicators based on blood tests.
[0006] In a first aspect, the present invention provides a method for predicting pulmonary embolism indicators based on blood tests, including: For each examination item of a patient, using the biochemical indicator of the examination item as an input variable and the pulmonary embolism examination result of the patient as an output variable, respectively constructing a number of machine learning models and training them; For each machine learning model obtained for each inspection item, a test data set is constructed using the patient's biochemical index data, and the evaluation parameters of each machine learning model on the test data set are calculated respectively to determine the optimal machine learning model corresponding to each inspection item; The patient's biochemical index data is input into the optimal machine learning model corresponding to each inspection item to obtain the pulmonary embolism prediction result corresponding to each inspection item; All pulmonary embolism prediction results are fused according to the number of pulmonary embolisms in the pulmonary embolism prediction results to generate the final pulmonary embolism prediction result of the patient.
[0007] In a second aspect, the present invention provides a pulmonary embolism index prediction device based on blood tests, including a model construction unit, a model test unit, a model application unit and a fusion output unit; The model construction unit is used to construct and train a number of machine learning models respectively for each inspection item of the patient, taking the biochemical indexes of the inspection item as input variables and the pulmonary embolism inspection results of the patient as output variables; The model test unit is used to construct a test data set using the patient's biochemical index data for each machine learning model obtained for each inspection item, calculate the evaluation parameters of each machine learning model on the test data set respectively, and determine the optimal machine learning model corresponding to each inspection item; The model application unit is used to input the patient's biochemical index data into the optimal machine learning model corresponding to each inspection item to obtain the pulmonary embolism prediction result corresponding to each inspection item; The fusion output unit is used to fuse all pulmonary embolism prediction results according to the number of pulmonary embolisms in the pulmonary embolism prediction results to generate the final pulmonary embolism prediction result of the patient.
[0008] In a third aspect, the present invention provides an electronic device, including: A processor and a memory; The memory is used to store computer operation instructions; The processor is used to execute the pulmonary embolism index prediction method based on blood tests by calling the computer operation instructions.
[0009] On the basis of the above technical solutions, the present invention can be further improved as follows.
[0010] Further, the inspection items include liver function, renal function, blood lipid, electrolyte, blood glucose, blood gas analysis, myocardial injury markers, brain natriuretic peptide, inflammatory indexes, blood routine and coagulation function.
[0011] Further, before constructing the machine learning model, the biochemical indexes of the inspection item are preprocessed, including data cleaning, missing value processing and normalization processing.
[0012] Further, the machine learning model includes at least two of a gradient boosting model, an extreme gradient boosting model, a light gradient boosting machine, an adaptive boosting model, a logistic regression model, a random forest model, a support vector machine model, a decision tree model, a naive Bayes model, and a K-nearest neighbor model.
[0013] Further, evaluate the parameters of each machine learning model on the test data set, including: Let the true positive be , the false positive be , the false negative be , the precision be , the recall be , then: ; ; .
[0014] Further, fuse all the pulmonary embolism prediction results according to the number of pulmonary embolism predictions that are pulmonary embolism, including: if the number of patients with pulmonary embolism in the pulmonary embolism prediction results is greater than or equal to the set threshold, the final pulmonary embolism prediction result of the patient is the first output value; otherwise, the final pulmonary embolism prediction result of the patient is the second output value.
[0015] Further, the first output value and the second output value are the probability values that the patient is a pulmonary embolism patient, and the probability value is the ratio of the number of patients with pulmonary embolism in the pulmonary embolism prediction results to the total number of examination items.
[0016] Further, deploy each machine learning model to the hospital's electronic medical record system, receive the biochemical index data in the patient's examination items as input variables in real time, and use the doctor's diagnosis result as the output variable to train each machine learning model.
[0017] The beneficial effects of the present invention are as follows: The present invention realizes an artificial intelligence model construction method based on biochemical indicators and its application in the early prediction of PE, and solves the problem of inaccurate PE detection and analysis due to poor equipment and insufficient diagnostic ability in primary medical institutions; the present invention not only significantly improves the prediction accuracy of the patient's pulmonary embolism index and the robustness of the model, but also does not rely on high-quality imaging and electrocardiogram data and senior CT doctors, realizes the whole process automation, has a wide application prospect, and also has obvious advantages in universality and popularization. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 is the schematic diagram of the pulmonary embolism index prediction method based on blood test provided by Embodiment 1 of the present invention; Figure 2 This is the schematic diagram of the pulmonary embolism index prediction device based on blood test provided in Embodiment 2 of the present invention; Figure 3 This is the schematic diagram of an electronic device provided in Embodiment 3 of the present invention.
[0019] Icons: 30 - Electronic device; 310 - Processor; 320 - Bus; 330 - Memory; 340 - Transceiver. Detailed implementation manners
[0020] To make the objectives, 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 with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. Components of the embodiments of the present invention usually described and illustrated in the drawings here can be arranged and designed in various different configurations.
[0021] Embodiment 1 As an embodiment, as shown in the appendix Figure 1 To solve the above technical problems, this embodiment provides a pulmonary embolism index prediction method based on blood test, including: For each inspection item of the patient, taking the biochemical index of the inspection item as the input variable and the pulmonary embolism inspection result of the patient as the output variable, constructing a number of machine learning models respectively and training them; For each machine learning model obtained for each inspection item, using the biochemical index data of the patient to construct a test data set, calculating the evaluation parameters of each machine learning model on the test data set respectively, and determining the optimal machine learning model corresponding to each inspection item; Inputting the biochemical index data of the patient into the optimal machine learning model corresponding to each inspection item to obtain the pulmonary embolism prediction result corresponding to each inspection item; Fusing all the pulmonary embolism prediction results according to the number of pulmonary embolisms in the pulmonary embolism prediction results to generate the final pulmonary embolism prediction result of the patient.
[0022] Optionally, the inspection items include liver function, kidney function, blood lipid, electrolyte, blood glucose, blood gas analysis, myocardial injury markers, brain natriuretic peptide, inflammatory indicators, blood routine and coagulation function.
[0023] Specifically, the biochemical indicators corresponding to liver function include alanine aminotransferase, aspartate aminotransferase, the ratio of alanine aminotransferase to aspartate aminotransferase, alkaline phosphatase, gamma-glutamyl transferase, cholinesterase, prealbumin, total protein, albumin, globulin, albumin / globulin ratio, total bilirubin, direct bilirubin and indirect bilirubin; the biochemical indicators corresponding to kidney function include creatinine, estimated glomerular filtration rate, uric acid, urea, microglobulin, neutrophil gelatinase-associated lipocalin and retinol-binding protein; the biochemical indicators corresponding to blood lipids include triglyceride, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, apolipoprotein A, apolipoprotein B and lipoprotein; the biochemical indicators corresponding to electrolytes include total carbon dioxide combining power, total calcium, inorganic phosphorus, magnesium, potassium, sodium, chloride, serum iron, total iron-binding capacity, unsaturated iron-binding capacity and serum iron saturation; the biochemical indicator corresponding to blood glucose is glucose; the biochemical indicators corresponding to blood gas analysis include pH value, partial pressure of carbon dioxide, partial pressure of oxygen, arterial blood pH, potassium ion, sodium ion, chloride ion, calcium ion, blood glucose, lactate, hematocrit, alveolar-arterial oxygen difference, anion gap, standard bicarbonate, standard bicarbonate, oxygen saturation, total carbon dioxide, partial pressure of oxygen at 50% oxygen saturation of hemoglobin, carboxyhemoglobin, deoxyhemoglobin, methemoglobin, oxyhemoglobin, extracellular fluid base excess, whole blood base excess, oxygen content, respiratory index, oxygen flow, osmotic pressure, buffer base and base excess at current oxygen saturation; the biochemical indicators corresponding to myocardial injury markers include high-sensitivity troponin, myoglobin and creatine kinase isoenzyme; the biochemical indicator corresponding to B-type natriuretic peptide is N-terminal pro-B-type natriuretic peptide; the biochemical indicators corresponding to inflammatory indicators include procalcitonin, interleukin-6, high-sensitivity C-reactive protein and serum amyloid A; the biochemical indicators corresponding to blood routine include white blood cells, absolute neutrophil count, absolute lymphocyte count, absolute monocyte count, absolute eosinophil count, absolute basophil count, neutrophil percentage, lymphocyte percentage, monocyte percentage, eosinophil percentage, basophil percentage, red blood cells, hemoglobin, hematocrit, mean corpuscular volume, mean hemoglobin content, mean hemoglobin concentration, red blood cell distribution width CV, red blood cell distribution width SD, platelet count, mean platelet volume, large platelet ratio, plateletcrit and platelet distribution width; the biochemical indicators corresponding to coagulation function include prothrombin time, international normalized ratio, prothrombin activity, prothrombin ratio, activated partial thromboplastin time, fibrinogen, thrombin time and D-dimer. The test items and the corresponding biochemical indicators are shown in Table 1.
[0024] Table 1 Test items and the biochemical indicators corresponding to each test item
[0025] Optionally, before constructing the machine learning model, the biochemical indicators of the inspection items are preprocessed, including data cleaning, missing value handling, and normalization. During the actual application process, the data is cleaned, missing values are handled, and normalization is performed for model training.
[0026] Optionally, the machine learning model includes at least two of the gradient boosting model, extreme gradient boosting model, light gradient boosting machine, adaptive boosting model, logistic regression model, random forest model, support vector machine model, decision tree model, naive bayes model, and k-nearest neighbor model.
[0027] During the actual application process, based on the gradient boosting algorithm (Gradient Boosting, GB for short), extreme gradient boosting algorithm (XGBoost, XGB for short), light gradient boosting machine (LightGBM, LGBM for short), adaptive boosting algorithm (AdaBoost, AdaB for short), logistic regression algorithm (Logistic Regression, LR for short), random forest algorithm (Random Forest, RF for short), support vector machine algorithm (Support Vector Machine, SVM for short), decision tree algorithm (Decision Tree, DT for short), naive bayes algorithm (Naive Bayes, NB), and k-nearest neighbor algorithm (K-Nearest Neighbor, KNN for short), the above ten machine learning models, namely the GB model, XGB model, LGBM model, AdaB model, LR model, RF model, SVM model, DT model, NB model, and KNN model, are constructed. Each inspection item corresponds to ten machine learning models.
[0028] For the biochemical indicators in the liver function inspection items, let alanine aminotransferase be X1, aspartate aminotransferase be X2, the ratio of alanine aminotransferase to aspartate aminotransferase be X3, alkaline phosphatase be X4, gamma-glutamyl transferase be X5, cholinesterase be X6, prealbumin be X7, total protein be X8, albumin be X9, globulin be X10, albumin / globulin ratio be X11, total bilirubin be X12, direct bilirubin be X13, and indirect bilirubin be X14. The biochemical indicators corresponding to the liver function inspection items are used as input variables, i.e., (X1, X2, X3, X4, X5, X6, X7, X8, X9, X10, X11, X12, X13, X14), and the pulmonary embolism results of the patients diagnosed manually are used as output variables, such as (PE, non-PE). The above ten machine learning models are constructed respectively to obtain ten different liver function machine learning models.
[0029] Using the same principle as that of liver function test items, machine learning models are constructed for kidney function, blood lipid, electrolyte, blood glucose, blood gas analysis, myocardial injury markers, brain natriuretic peptide, inflammatory indicators, blood routine and coagulation function test items respectively. Let the biochemical indicators of kidney function test items be X15~X21, the biochemical indicators of blood lipid test items be X22~X28, the biochemical indicators of electrolyte test items be X29~X39, the biochemical indicator of blood glucose test item be X40, the biochemical indicators of blood gas analysis test items be X41~X70, the biochemical indicators of myocardial injury marker test items be X71~X73, the biochemical indicator of brain natriuretic peptide test item be X74, the biochemical indicators of inflammatory indicator test items be X75~X78, the biochemical indicators of blood routine test items be X79~X102, and the biochemical indicators of coagulation function test items be X103~X110. The biochemical indicators corresponding to the liver function test items are used as input variables, and the pulmonary embolism results (PE, non-PE) of the patients diagnosed manually are used as output variables.
[0030] Optionally, calculate the evaluation parameters of each machine learning model on the test data set, including: Let the true positive be , the false positive be , the false negative be , the precision be , the recall be , then: ; ; .
[0031] Use the F1 score of the model on the test data set as the optimal model. The F1 score is an important indicator for comprehensively evaluating the performance of classification models in machine learning. Especially when dealing with imbalanced data sets, the F1 score can balance the effects of precision and recall. The F1 score is the harmonic mean of precision and recall, and can find the balance point between these two indicators. Let the F1 score of the GB model be F1(GB), the F1 score of the XGB model be F1(XGB), the F1 score of the LGBM model be F1(LGBM), the F1 score of the AdaB model be F1(AdaB), the F1 score of the LR model be F1(LR), the F1 score of the RF model be F1(RF), the F1 score of the SVM model be F1(SVM), the F1 score of the DT model be F1(DT), the F1 score of the NB model be F1(NB), and the F1 score of the KNN model be F1(KNN). Therefore, assuming the optimal machine learning model corresponding to the liver function test items is represented as Optimal Model1, then: Optimal Model1 = MAX{F1(GB), F1(XGB), F1(LGBM), F1(AdaB), F1(LR), F1(RF), F1(SVM), F1(DT), F1(NB), F1(KNN)}. Select the model with the highest F1 score from these 10 models, which indicates that this algorithm is most suitable for constructing a PE classification model based on liver function test items. In the same way, the optimal machine learning models for the remaining 10 test items are constructed respectively. Let the optimal machine learning model for renal function test items be Optimal Model2, the optimal machine learning model for blood lipid test items be Optimal Model3, the optimal machine learning model for electrolyte test items be Optimal Model4, the optimal machine learning model for blood glucose test items be Optimal Model5, the optimal machine learning model for blood gas analysis test items be Optimal Model6, the optimal machine learning model for myocardial injury marker test items be Optimal Model7, the optimal machine learning model for B-type natriuretic peptide test items be Optimal Model8, the optimal machine learning model for inflammatory index test items be Optimal Model9, the optimal machine learning model for blood routine test items be Optimal Model10, and the optimal machine learning model for coagulation function test items be Optimal Model11. As shown in Table 2, the input variables of the model are the biochemical indexes of each test item, and the output variable is the pulmonary embolism result of the patient diagnosed manually, such as YN (PE, non-PE), where N is the model serial number.
[0032] Table 2 Optimal machine learning model table for each test item
[0033] Optionally, fuse all pulmonary embolism prediction results according to the number of pulmonary embolism predictions that are pulmonary embolism, including: if the number of patients with pulmonary embolism in the pulmonary embolism prediction results is greater than or equal to the set threshold, then the final pulmonary embolism prediction result of the patient is the first output value; otherwise, the final pulmonary embolism prediction result of the patient is the second output value.
[0034] Optionally, the first output value and the second output value are the probability values that the patient is a pulmonary embolism patient, and this probability value is the ratio of the number of patients with pulmonary embolism in the pulmonary embolism prediction results to the total number of test items.
[0035] In actual application, after the patient is predicted by the above 11 optimal machine learning models, 11 prediction results are obtained, which are set as Y1, Y2, Y3, Y4, Y5, Y6, Y7, Y8, Y9, Y10 and Y11, and the combination of prediction results is Y[Y1, Y2, Y3, Y4, Y5, Y6, Y7, Y8, Y9, Y10, Y11]. Among the 11 prediction results, if the number of predicted results for PE is greater than or equal to 6, the probability that the patient is a PE patient is high; if the number of predicted results for PE is less than 6, the probability that the patient is a PE patient is low. When the probability of predicting a PE patient is high, the doctor immediately takes medical measures (further imaging examination, surgery or transfer to another hospital) to gain the best time window for timely treatment of the patient.
[0036] Optionally, each machine learning model is deployed to the hospital's electronic medical record system, which receives the patient's biochemical index data in the examination items as input variables in real time and uses the doctor's diagnosis results as output variables to train each machine learning model.
[0037] In the actual application process, the parameters of the machine learning models are continuously adjusted by training each machine learning model, and a machine learning model with better performance is obtained through multiple rounds of model training. The optimized single machine learning model is integrated and deployed into the electronic medical record system of the grassroots hospital, which receives the patient's biochemical index data in real time and outputs the PE prediction results. Doctors can intervene and treat patients early based on the prediction results, thereby reducing the mortality rate of PE.
[0038] The present invention realizes a method for constructing an artificial intelligence model based on biochemical indicators and its application in the early prediction of PE, solving the problem of inaccurate PE detection and analysis due to poor equipment and insufficient diagnostic capabilities in primary medical institutions. The present invention not only significantly improves the prediction accuracy of patients' pulmonary embolism indicators and the robustness of the model, but also does not need to rely on high-quality imaging and electrocardiogram data and experienced CT doctors, realizes full process automation, has broad application prospects, and also has obvious advantages in universality and scalability.
[0039] 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 pulmonary embolism index prediction device based on blood test, including a model building unit, a model testing unit, a model application unit and a fusion output unit; A model building unit is used to build and train several machine learning models for each examination item of the patient, taking the biochemical index of the examination item as an input variable and the patient's pulmonary embolism examination result as an output variable; A model testing unit, which is used to construct a test data set by using the biochemical index data of patients for each machine learning model obtained for each inspection item, calculate the evaluation parameters of each machine learning model on the test data set respectively, and determine the optimal machine learning model corresponding to each inspection item; A model application unit, which is used to input the biochemical index data of patients into the optimal machine learning model corresponding to each inspection item to obtain the pulmonary embolism prediction result corresponding to each inspection item; A fusion output unit, which is used to fuse all the pulmonary embolism prediction results according to the number of pulmonary embolisms in the pulmonary embolism prediction results to generate the final pulmonary embolism prediction result of the patient.
[0040] Optionally, the inspection items include liver function, kidney function, blood lipids, electrolytes, blood glucose, blood gas analysis, myocardial injury markers, brain natriuretic peptide, inflammatory indicators, blood routine and coagulation function.
[0041] Optionally, the biochemical indicators corresponding to liver function include alanine aminotransferase, aspartate aminotransferase, the ratio of alanine aminotransferase to aspartate aminotransferase, alkaline phosphatase, gamma-glutamyl transferase, cholinesterase, prealbumin, total protein, albumin, globulin, albumin / globulin ratio, total bilirubin, direct bilirubin and indirect bilirubin; the biochemical indicators corresponding to kidney function include creatinine, estimated glomerular filtration rate, uric acid, urea, microglobulin, neutrophil gelatinase-associated lipocalin and retinol-binding protein; the biochemical indicators corresponding to blood lipids include triglyceride, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, apolipoprotein A, apolipoprotein B and lipoprotein; the biochemical indicators corresponding to electrolytes include total carbon dioxide combining power, total calcium, inorganic phosphorus, magnesium, potassium, sodium, chloride, serum iron, total iron-binding capacity, unsaturated iron-binding capacity and serum iron saturation; the biochemical indicators corresponding to blood glucose include glucose; the biochemical indicators corresponding to blood gas analysis include pH value, partial pressure of carbon dioxide, partial pressure of oxygen, arterial blood pH, potassium ion, sodium ion, chloride ion, calcium ion, blood glucose, lactate, hematocrit, alveolar-arterial oxygen gradient, anion gap, standard bicarbonate, standard bicarbonate, oxygen saturation, total carbon dioxide, partial pressure of oxygen at 50% oxygen saturation of hemoglobin, carboxyhemoglobin, deoxyhemoglobin, methemoglobin, oxyhemoglobin, extracellular fluid base excess, whole blood base excess, oxygen content, respiratory index, oxygen flow, osmotic pressure, buffer base and base excess at current oxygen saturation; the biochemical indicators corresponding to myocardial injury markers include high-sensitivity troponin, myoglobin and creatine kinase isoenzyme; the biochemical indicators corresponding to B-type natriuretic peptide include N-terminal pro-B-type natriuretic peptide; the biochemical indicators corresponding to inflammation indicators include procalcitonin, interleukin-6, high-sensitivity C-reactive protein and serum amyloid A; the biochemical indicators corresponding to blood routine include white blood cells, absolute neutrophil count, absolute lymphocyte count, absolute monocyte count, absolute eosinophil count, absolute basophil count, neutrophil percentage, lymphocyte percentage, monocyte percentage, eosinophil percentage, basophil percentage, red blood cells, hemoglobin, hematocrit, mean corpuscular volume, mean hemoglobin content, mean hemoglobin concentration, red blood cell distribution width CV, red blood cell distribution width SD, platelet count, mean platelet volume, large platelet ratio, plateletcrit and platelet distribution width; the biochemical indicators corresponding to coagulation function include prothrombin time, international normalized ratio, prothrombin activity, prothrombin ratio, activated partial thromboplastin time, fibrinogen, thrombin time and D-dimer.
[0042] Optionally, the machine learning models include at least two of gradient boosting model, extreme gradient boosting model, light gradient boosting machine, adaptive boosting model, logistic regression model, random forest model, support vector machine model, decision tree model, naive Bayes model and K-nearest neighbor model.
[0043] Optionally, evaluate the parameters of each machine learning model on the test data set, including: setting the true positive example as , the false positive example as , the false negative example as , the precision as , the recall as , then: ; ; .
[0044] Optionally, fuse all the pulmonary embolism prediction results according to the number of pulmonary embolism prediction results that are pulmonary embolism, including: if the number of patients with pulmonary embolism in the pulmonary embolism prediction results is greater than or equal to the set threshold, the final pulmonary embolism prediction result of the patient is the first output value; otherwise, the final pulmonary embolism prediction result of the patient is the second output value.
[0045] Optionally, the first output value and the second output value are the probability values that the patient is a pulmonary embolism patient, and the probability value is the ratio of the number of patients with pulmonary embolism in the pulmonary embolism prediction results to the total number of examination items.
[0046] Optionally, deploy each machine learning model to the hospital's electronic medical record system, receive the biochemical index data in the patient's examination items as input variables in real time, and use the doctor's diagnosis result as the output variable to train each machine learning model.
[0047] Embodiment 3 Based on the same principle as the method shown in the embodiments of the present invention, an electronic device is also provided in the embodiments of the present invention, as shown in the appendix Figure 3 . The electronic device may include, but is not limited to: a processor and a memory; the memory is used to store a computer program; the processor is used to execute the pulmonary embolism index prediction method based on blood tests shown in the embodiments of the present invention by calling the computer program.
[0048] In an alternative embodiment, an electronic device is provided, Figure 3 The electronic device 40 shown in
[0049] includes: a processor 310 and a memory 330. Among them, the processor 310 and the memory 330 are connected, such as through a bus 320.
[0050] The processor 310 can be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in connection with the disclosure of the present invention. The processor 310 can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.
[0051] The bus 320 can include a path for transmitting information between the above components. The bus 320 can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 320 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 only a thick line is shown in the figure, but it does not mean that there is only one bus or one type of bus.
[0052] The memory 330 can be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, or it can also be an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic storage media, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.
[0053] The memory 330 is used to store the application program code (computer program) for implementing the solution of the present invention, and is controlled by the processor 310 for execution. The processor 310 is used to execute the application program code stored in the memory 330 to implement the content shown in the foregoing method embodiments.
[0054] The above are only the 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 changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.
Claims
1. A method for predicting pulmonary embolism indicators based on blood tests, characterized in that, including: For each examination item of the patient, using the biochemical indexes of the examination item as input variables and the pulmonary embolism examination results of the patient as output variables, constructing and training several machine learning models respectively; For each machine learning model obtained for each examination item, using the biochemical index data of the patient to construct a test data set, calculating the evaluation parameters of each machine learning model on the test data set respectively, and determining the optimal machine learning model corresponding to each examination item; Inputting the biochemical index data of the patient into the optimal machine learning model corresponding to each examination item to obtain the pulmonary embolism prediction result corresponding to each examination item; Fusing all the pulmonary embolism prediction results according to the number of pulmonary embolisms in the pulmonary embolism prediction results to generate the final pulmonary embolism prediction result of the patient.
2. The method for predicting pulmonary embolism indicators based on blood tests according to claim 1, wherein The examination items include liver function, kidney function, blood lipid, electrolyte, blood glucose, blood gas analysis, myocardial injury markers, B-type natriuretic peptide, inflammatory indexes, blood routine and coagulation function.
3. The method for predicting pulmonary embolism indicators based on blood tests according to claim 2, wherein Before constructing the machine learning model, preprocessing the biochemical indexes of the examination items, including data cleaning, missing value processing and normalization processing.
4. The method for predicting pulmonary embolism indicators based on blood tests according to claim 1, wherein The machine learning models include at least two of gradient boosting model, extreme gradient boosting model, light gradient boosting machine, adaptive boosting model, logistic regression model, random forest model, support vector machine model, decision tree model, naive Bayes model and K-nearest neighbor model.
5. The method for predicting pulmonary embolism indicators based on blood tests according to claim 1, wherein Calculate the evaluation parameters of each machine learning model on the test data set, including: true positive is , false positive is , false negative is , precision is , recall is , then: ； ； 。 6. The method for predicting pulmonary embolism indicators based on blood tests according to claim 1, wherein Fusing all the pulmonary embolism prediction results according to the number of pulmonary embolisms in the pulmonary embolism prediction results, including: if the number of pulmonary embolisms in the pulmonary embolism prediction results for the patient is greater than or equal to the set threshold, the final pulmonary embolism prediction result for the patient is the first output value; otherwise, the final pulmonary embolism prediction result for the patient is the second output value.
7. The method for predicting pulmonary embolism indicators based on blood tests according to claim 6, wherein The first output value and the second output value are the probability values that the patient is a pulmonary embolism patient, and this probability value is the ratio of the number of pulmonary embolisms in the pulmonary embolism prediction results for the patient to the total number of examination items.
8. The method for predicting pulmonary embolism indicators based on blood tests according to claim 1, wherein Deploying each machine learning model to the hospital's electronic medical record system, receiving the biochemical index data in the patient's examination items as input variables in real time, and using the doctor's diagnosis result as the output variable to train each machine learning model.
9. A pulmonary embolism index prediction device based on blood tests, characterized in that, including a model construction unit, a model testing unit, a model application unit and a fusion output unit; The model construction unit is used for, for each examination item of the patient, using the biochemical indexes of the examination item as input variables and the pulmonary embolism examination results of the patient as output variables, constructing and training several machine learning models respectively; The model testing unit is used for, for each machine learning model obtained for each examination item, using the biochemical index data of the patient to construct a test data set, calculating the evaluation parameters of each machine learning model on the test data set respectively, and determining the optimal machine learning model corresponding to each examination item; The model application unit is used for inputting the biochemical index data of the patient into the optimal machine learning model corresponding to each examination item to obtain the pulmonary embolism prediction result corresponding to each examination item; The fusion output unit is used for fusing all the pulmonary embolism prediction results according to the number of pulmonary embolisms in the pulmonary embolism prediction results to generate the final pulmonary embolism prediction result of the patient.
10. An electronic device, characterized in that, including: Processor and memory; The memory is used for storing computer operation instructions; The processor is used for executing the method for predicting pulmonary embolism indexes based on blood tests according to any one of claims 1 to 8 by invoking the computer operation instructions.