JP7706147B2 - Prediction device, prediction method, prediction program, and prediction system - Google Patents

Description

The present disclosure relates to a prediction device, a prediction method, a prediction program, and a prediction system.

Acute myocardial infarction (AMI) is the leading cause of death worldwide. To date, several methods have been proposed to predict the short- and long-term prognosis of AMI patients.

For example, Patent Document 1 discloses risk scoring methods such as the TIMI (Thrombolysis In Myocardial Ischemia) risk index and the GRACE (Global Registries of Acute Coronary Events) score as methods for predicting the prognosis of AMI.

Special table 2016-532093 publication

The conventional prediction method disclosed in Patent Document 1 has been in existence since before percutaneous transluminal coronary angioplasty using a catheter became the main treatment for AMI. For this reason, it is difficult to say that the conventional prediction method is compatible with the current standard of medical care in which percutaneous transluminal coronary angioplasty has become the main treatment, and there is a demand for a prediction method that can predict the prognosis of AMI patients with higher accuracy.

This disclosure has been made to solve these problems, and its purpose is to provide technology that can predict the prognosis of patients with acute myocardial infarction with greater accuracy.

A prediction device according to an aspect of the present disclosure predicts the prognosis of a patient with acute myocardial infarction. The prediction device includes an acquisition unit, a prediction unit, and an output unit. The acquisition unit acquires at least one feature of the patient. The prediction unit predicts the prognosis based on the at least one feature using a trained prediction model. The output unit outputs the predicted prognosis. The at least one feature is ten types of feature, including the maximum value of creatine phosphokinase in blood, the amount of hemoglobin in blood, heart rate, the concentration of creatinine in blood, systolic blood pressure, blood glucose level, age, a classification result regarding the severity of cardiac dysfunction in acute myocardial infarction, the number of white blood cells in blood, and the amount of C-reactive protein in blood.

A prediction method according to another aspect of the present disclosure is a method for predicting the prognosis of a patient with acute myocardial infarction by a computer. The prediction method includes the steps of (a) acquiring at least one feature of the patient, (b) predicting the prognosis based on the at least one feature using a trained prediction model, and (c) outputting a prediction result of the prognosis. The at least one feature is ten types of feature, including the maximum value of creatine phosphokinase in blood, the amount of hemoglobin in blood, heart rate, the concentration of creatinine in blood, systolic blood pressure, blood glucose level, age, a classification result regarding the severity of cardiac dysfunction in acute myocardial infarction, the number of white blood cells in blood, and the amount of C-reactive protein in blood.

A prediction program according to another aspect of the present disclosure is a program for predicting the prognosis of a patient with acute myocardial infarction. The prediction program causes a computer to execute the steps of (a) acquiring at least one feature of the patient, (b) predicting the prognosis based on the at least one feature using a trained prediction model, and (c) outputting a prediction result of the prognosis. The at least one feature is ten types of feature, including the maximum value of creatine phosphokinase in blood, the amount of hemoglobin in blood, heart rate, the concentration of creatinine in blood, systolic blood pressure, blood glucose level, age, a classification result regarding the severity of cardiac dysfunction in acute myocardial infarction, the number of white blood cells in blood, and the amount of C-reactive protein in blood.

A prediction system according to another aspect of the present disclosure predicts the prognosis of a patient with acute myocardial infarction. The prediction system includes a user terminal and a server device configured to communicate with the user terminal. The server device includes an acquisition unit, a prediction unit, and an output unit. The acquisition unit acquires at least one feature of the patient from the user terminal. The prediction unit predicts the prognosis based on the at least one feature using a trained prediction model. The output unit outputs the prediction result of the prognosis to the user terminal. The at least one feature is ten types of feature, including the maximum value of creatine phosphokinase in blood, the amount of hemoglobin in blood, heart rate, the concentration of creatinine in blood, systolic blood pressure, blood glucose level, age, a classification result regarding the severity of cardiac dysfunction in acute myocardial infarction, the number of white blood cells in blood, and the amount of C-reactive protein in blood.

According to the present disclosure, it is possible to predict prognosis based on a patient's features using a trained prediction model, thereby making it possible to predict the prognosis of patients with acute myocardial infarction with greater accuracy.

1 is a diagram for explaining an application example of a prediction device according to a first embodiment; FIG. 2 is a diagram illustrating a configuration of a control device provided in the prediction device according to the first embodiment. FIG. 2 is a diagram for explaining generation of a prediction model according to the first embodiment. FIG. 11 is a diagram for explaining an example of sample data. FIG. 11 is a diagram for explaining an example of sample data. FIG. 13 is a diagram showing evaluation results of AUC (Area Under the Curve) for each number of types of feature amounts. FIG. 11 is a diagram for explaining the importance of each feature amount. FIG. 11 is a diagram for explaining the importance of each feature amount. FIG. 11 is a diagram showing a comparison result of prediction models in the present embodiment and a comparative example. FIG. 11 is a diagram showing a comparison result of prediction models in the present embodiment and a comparative example. FIG. 11 is a diagram showing a comparison result of prediction models in the present embodiment and a comparative example. FIG. 11 is a diagram showing a comparison result of prediction models in the present embodiment and a comparative example. FIG. 11 is a diagram showing a comparison result of prediction models in the present embodiment and a comparative example. 1 is a diagram for explaining an example of a display of a prediction result by the prediction device according to the first embodiment. FIG. 4 is a flowchart of a prediction process executed by the prediction device according to the first embodiment. FIG. 11 is a diagram for explaining a prediction system according to a second embodiment.

This embodiment will be described in detail with reference to the drawings. Note that the same or equivalent parts in the drawings will be given the same reference numerals, and in principle their description will not be repeated.

<Prediction device according to the first embodiment>
A prediction device 1 according to a first embodiment will be described with reference to FIGS.

[Configuration of prediction device]
1 is a diagram for explaining an application example of a prediction device 1 according to embodiment 1. The prediction device 1 is an information terminal that executes predetermined information processing, such as a desktop personal computer (PC), a laptop PC, a smartphone, a smart watch, a wearable device, or a tablet PC. In embodiment 1, the prediction device 1 is used by a user such as a doctor, and executes processing for predicting the prognosis of a patient with AMI (hereinafter also referred to as "prediction processing").

Specifically, the prediction device 1 includes a prediction model 121 for predicting the prognosis of a patient with AMI. The prediction model 121 is generated by machine learning based on at least one feature value belonging to the patient and whether or not the patient has died in the hospital. The prediction device 1 uses the prediction model 121 to predict the prognosis of the patient based on the feature values of the patient with AMI. By providing the patient with the prediction result by the prediction device 1, the user can provide the patient with medical treatment with higher accuracy than before.

As shown in FIG. 1, the prediction device 1 includes a control device 10, a display device 20, a keyboard 31, and a mouse 32. The display device 20 displays an image on a display 25 under the control of the control device 10. Each of the keyboard 31 and the mouse 32 accepts a user operation and inputs information based on the accepted user operation to the control device 10. The display device 20 may include a touch panel (not shown). The user may input information to the control device 10 by operating the touch panel.

FIG. 2 is a diagram showing the configuration of the control device 10 provided in the prediction device 1 according to the first embodiment. As shown in FIG. 2, the control device 10 includes a calculation device 11, a storage device 12, an input interface 13, and an output interface 14.

The calculation device 11 is an example of a "prediction unit" and is a calculation entity (computer) that executes various processes according to various programs. The calculation device 11 includes, for example, at least one of a CPU (Central Processing Unit), an FPGA (Field Programmable Gate Array), a GPU (Graphics Processing Unit), and an MPU (Multi Processing Unit). Furthermore, the calculation device 11 may include volatile memory such as a DRAM (Dynamic Random Access Memory) and an SRAM (Static Random Access Memory), and non-volatile memory such as a ROM (Read Only Memory) and a flash memory. The calculation device 11 may be configured with a processing circuitry.

The storage device 12 includes non-volatile memory such as a hard disk drive (HDD) and a solid state drive (SSD). The storage device 12 stores various programs and data such as a prediction model 121 and a prediction program 122. The prediction program 122 is a program that causes the calculation device 11 to execute prediction processing using the prediction model 121.

The input interface 13 is an example of an "acquisition unit" and acquires information input from the outside. For example, the input interface 13 acquires information input from each of the keyboard 31 and the mouse 32.

The output interface 14 is an example of an "output unit" and outputs information to the outside in accordance with the control of the computing device 11. For example, the output interface 14 outputs information for displaying an image to the display device 20.

In the control device 10 configured as described above, the input interface 13 acquires at least one characteristic amount of the AMI patient from the outside. The calculation device 11 executes a prediction process according to the prediction program 122 to predict the prognosis of the AMI patient based on the characteristic amount of the AMI patient and the prediction model 121. The output interface 14 outputs image information including the prediction result obtained by the calculation device 11 to the display device 20. The display device 20 displays the prediction result on the display 25 based on the image information acquired from the control device 10.

As a result, a user can input at least one feature of an AMI patient into the prediction device 1, and use the trained prediction model 121 to predict the patient's prognosis.

[Generation of predictive models]
The generation of the prediction model 121 will be described with reference to FIGS.

FIG. 3 is a diagram for explaining the generation of the prediction model 121 according to the first embodiment. The prediction model 121 according to the first embodiment is generated by supervised learning. As shown in FIG. 3, the designer of the prediction model 121 first prepares a plurality of sample data to be used when generating the prediction model 121. For example, the designer prepares data on patients with AMI after percutaneous transluminal coronary intervention became the main treatment for AMI. In this example, the designer prepares data on 2,553 patients with ST-segment elevation acute myocardial infarction who were registered at medical institutions from 2009 to 2015 as sample data.

Each of the multiple sample data includes at least one feature and correct answer data. The at least one feature includes clinical data and biological data of the patient. The correct answer data includes information on whether or not the patient died in the hospital.

Here, the sample data will be specifically described with reference to Figures 4 and 5. Figures 4 and 5 are diagrams for explaining an example of sample data. Figures 4 and 5 show data on surviving patients and data on patients who died in the hospital for each feature.

As shown in Figures 4 and 5, of the 2,553 patients, 2,358 patients survived, and 195 patients died in the hospital. In other words, of the 2,553 pieces of sample data, 2,358 pieces of sample data (Survival) were assigned survival as the correct answer data, and 195 pieces of sample data (Dead) were assigned in-hospital death as the correct answer data. The sample data includes data corresponding to various features for both the patients who survived and the patients who died in the hospital.

As shown in FIG. 4, the features include, for example, age, sex, body mass index (BMI), presence or absence of hypertension, presence or absence of diabetes, presence or absence of dyslipidemia, presence or absence of smoking, presence or absence of family members who have developed AMI, presence or absence of previous AMI, presence or absence of previous cardiovascular disease, presence or absence of hemodialysis treatment, Killip classification results regarding the severity of cardiac dysfunction in AMI, systolic blood pressure (BPs), heart rate (HR), number of white blood cells (WBC), amount of hemoglobin (Hb), blood sugar level (BS), maximum value of creatine phosphokinase in blood (max CPK, hereafter also referred to as "max CPK"), concentration of creatinine (Cr), concentration of C-reactive protein (CRP) in blood, and concentration of creatinine (Cr) in blood. These include the amount of thrombus protein, TIMI grades before and after treatment, whether or not treatment was performed using a stent, and whether or not treatment was performed by thrombus aspiration.

Here, creatine phosphokinase (CPK) is an enzyme present in muscles such as skeletal muscles, cardiac muscles, and smooth muscles, and plays an important role in the energy metabolism of muscle cells. When abnormalities occur in these muscle cells, CPK flows into the blood, and the amount of CPK in the blood increases. When AMI occurs, the CPK value shown by blood tests becomes higher than the normal value and peaks within 24 hours after the onset of AMI. maxCPK is the CPK value at the time of peaking. The unit of CPK value is "IU/L" or "U/L". For these units, for example, the amount of enzyme that can convert 1 μmol of substrate per minute at a specified temperature (for example, 30 degrees, 37 degrees when measured with an automatic biochemical analyzer in clinical testing) in 1 L of sample is defined as one unit.

Furthermore, as shown in FIG. 5, the features include, for example, the site where the AMI occurred, the blood vessel responsible for the AMI, the number of lesions, the season when the AMI occurred, the time when the AMI occurred, the time it took for the patient to resume blood flow after the patient arrived at the hospital, the time it took for the patient to arrive at the hospital after the AMI occurred, and the method of hospitalization. Note that, as shown in parentheses in each of FIG. 4 and FIG. 5, some of the multiple features also include information on the percentage (%) or the range of variation in the data (IQR: Interquartile Range).

Returning to FIG. 3, the designer generates and evaluates the predictive model 121 using the hold-out method. Specifically, the designer divides multiple sample data into training data for generating the predictive model 121 and test data for evaluating the predictive model 121. For example, the designer uses 80% of the multiple sample data as training data and 20% of the multiple sample data as test data. The respective proportions of training data and test data can be set by the designer as appropriate.

Furthermore, the designer generates a predictive model 121 by K-fold cross-validation using multiple training data. Specifically, the designer divides the multiple training data into K pieces (K=5 in this example), and of the K pieces of training data, uses one training data as validation data, and uses the remaining (K-1) training data as learning data. The designer then trains a predictive model using the learning data, and then evaluates the predictive model using the validation data. Next, the designer uses one training data from the K pieces of training data that is different from the training data already used as validation data, as new validation data, and uses the remaining (K-1) training data as new learning data. The designer then trains a predictive model using the learning data, and then evaluates the predictive model using the validation data.

The designer repeats the machine learning and evaluation set for the predictive model as described above K times until all training data has been used as validation data. In this way, the designer generates a predictive model using multiple training data.

Ensemble learning is applied as the machine learning of the prediction model. For example, in the first embodiment, random forest is used as the machine learning algorithm. Random forest is a method of generating multiple decision trees based on at least one feature selected randomly from multiple features included in the training data, and improving prediction accuracy by taking the average or majority vote of each prediction result of the multiple decision trees.

By using a learning machine (random forest classifier) that employs random forest as a learning algorithm, the designer can select the prediction model with the highest prediction accuracy from among the multiple prediction models generated, and select at least one feature that will provide the highest prediction accuracy. The designer then evaluates the selected prediction model using test data. At this time, the designer uses the feature that will provide the highest prediction accuracy as the feature of the test data. In this way, the designer can generate the final prediction model 121.

The machine learning algorithm of the prediction model 121 is not limited to random forest, and known learning algorithms such as support vector machines, neural networks, logistic regression, and xgboost (eXtreme Gradient Boosting) may be applied. Furthermore, the designer may generate the prediction model 121 using the same algorithm as the algorithm used to extract at least one feature, or may generate the prediction model 121 using an algorithm different from the algorithm used to extract at least one feature. For example, the designer may use the random forest algorithm when selecting at least one feature that has the highest prediction accuracy from the training data, while using the selected at least one feature to generate the final prediction model 121, the designer may use another algorithm different from the random forest described above to generate the prediction model 121.

Figure 6 shows the evaluation results of AUC (Area Under the Curve) for each number of types of features. Figure 6 (A) shows the evaluation results of AUC in a prediction model generated by random forest. Figure 6 (B) shows the evaluation results of AUC in a prediction model generated by xgboost. Figure 6 (C) shows the evaluation results of AUC in a prediction model generated by logistic regression. Figure 6 shows a graph with the number of types of features on the horizontal axis and AUC on the vertical axis for each learning algorithm.

As shown in FIG. 9, which will be described later, AUC is an evaluation index indicated by the area under a receiver operating characteristic (ROC) curve plotted on a graph with the false positive rate (1-specificity) on the horizontal axis and the sensitivity on the vertical axis. The closer the AUC is to 1, the higher the prediction accuracy of the prediction model. The false positive rate indicates the proportion of sample data associated with a negative label (e.g., survival) as the correct answer data, but the prediction model erroneously predicts it as a positive label (e.g., in-hospital death). The sensitivity indicates the proportion of sample data associated with a positive label (e.g., in-hospital death) as the correct answer data, but the prediction model correctly predicts it as a positive label (e.g., in-hospital death).

As shown in FIG. 6, a prediction model generated by any of the learning algorithms, random forest, xgboost, and logistic regression, can obtain a high AUC value of 0.85 or more when six or more types of features are used as input. Furthermore, a prediction model generated by any of the learning algorithms can obtain the highest AUC value when ten types of features are used as input.

Figures 7 and 8 are diagrams for explaining the importance of each feature. In Figures 7 and 8, graphs are shown with the importance of the feature on the horizontal axis and the type of feature on the vertical axis. The importance of the feature is calculated by a learning device that uses random forest as a learning algorithm. The higher the importance of the feature used as input data for the prediction model, the higher the prediction accuracy of the prediction model.

As shown in Figure 7, the 10 features that yield the highest AUC values are max CPK, hemoglobin level, heart rate, creatinine concentration, systolic blood pressure, blood glucose level, age, Killip classification result, white blood cell count, and C-reactive protein level.

As shown in Figure 8, when all features are selected as input data for the prediction model, the features with the highest importance are max CPK, systolic blood pressure, age, creatinine concentration, Killip classification results, hemoglobin amount, white blood cell count, heart rate, blood glucose level, body mass index, and C-reactive protein amount.

In both the graphs shown in Figures 7 and 8, maxCPK has the highest importance, and of all the features, maxCPK can most effectively improve the prediction accuracy of the prediction model.

Thus, the inventor, who designed the prediction model 121, found that high prediction accuracy can be maintained by using six or more types of features as patient features used when predicting the prognosis of AMI patients, that prediction accuracy can be most improved by using ten types of features as shown in FIG. 7, and that prediction accuracy can be most improved by using max CPK among all the features. Note that while Python code was used to generate the prediction model 121, the prediction model 121 may be generated using other programming languages.

[Evaluation of predictive models]
Evaluation of the prediction model 121 will be described with reference to Figs. 9 to 13. Figs. 9 to 13 are diagrams showing comparison results of prediction models in this embodiment and a comparative example. In the evaluation results shown in Figs. 9 to 13, prediction results are used when ten types of feature amounts as shown in Fig. 7 are input to the prediction model 121 generated by random forest as this embodiment. As comparative example 1, prediction results using the conventional TIMI risk index are used. As comparative example 2, prediction results using the conventional GRACE score are used.

Figure 9 shows a graph of the ROC (Receiver Operating Characteristic) curve for each of the present embodiment, Comparative Example 1, and Comparative Example 2, with the false positive rate on the horizontal axis and sensitivity on the vertical axis. In the graph, the area below the ROC curve is called the AUC (Area Under the Curve) and takes values between 0 and 1. The higher the prediction accuracy, the closer the AUC value is to 1.

As shown in FIG. 9, the prediction results using the prediction model 121 of this embodiment have an AUC closer to 1 than the prediction results using each of Comparative Example 1 and Comparative Example 2.

As such, a comparison using the ROC curve shows that the prediction model 121 of this embodiment has higher prediction accuracy than both Comparative Example 1 and Comparative Example 2.

Figure 10 shows the confusion matrix when predicting prognosis using 511 sample data for each of the present embodiment, Comparative Example 1, and Comparative Example 2. The confusion matrix is used when evaluating prediction accuracy and includes true negatives (TN), false positives (FP), false negatives (FN), and true positives (TP).

True negatives indicate the number of times that the label of sample data associated with a negative label (e.g., survival) as the correct answer data was correctly predicted as a negative label (e.g., survival), and the higher the prediction accuracy, the larger the value. False positives indicate the number of times that the label of sample data associated with a negative label (e.g., survival) as the correct answer data was incorrectly predicted as a positive label (e.g., in-hospital death), and the higher the prediction accuracy, the smaller the value. False negatives indicate the number of times that the label of sample data associated with a positive label (e.g., in-hospital death) as the correct answer data was incorrectly predicted as a negative label (e.g., survival), and the higher the prediction accuracy, the smaller the value. True positives indicate the number of times that the label of sample data associated with a positive label (e.g., in-hospital death) as the correct answer data was correctly predicted as a positive label (e.g., in-hospital death), and the higher the prediction accuracy, the larger the value.

As shown in FIG. 10, the prediction results of this embodiment have a larger true negative (TN) value than the prediction results of Comparative Example 1 and Comparative Example 2. The prediction results of this embodiment have a smaller false positive (FP) value than the prediction results of Comparative Example 1 and Comparative Example 2. The prediction results of this embodiment have a smaller false negative (FN) value than the prediction result of Comparative Example 2, and this value is the same as the prediction result of Comparative Example 1. The prediction results of this embodiment have a larger true positive (TP) value than the prediction result of Comparative Example 2, and this value is the same as the prediction result of Comparative Example 1.

As such, a comparison using the confusion matrix shows that the prediction model 121 of this embodiment has higher prediction accuracy than both Comparative Example 1 and Comparative Example 2.

Figure 11 (A) shows the results of comparing the sensitivity for each of this embodiment, Comparative Example 1, and Comparative Example 2. Figure 11 (B) shows the results of comparing the specificity for each of this embodiment, Comparative Example 1, and Comparative Example 2. Figure 11 (C) shows the results of comparing the precision for each of this embodiment, Comparative Example 1, and Comparative Example 2. Figure 11 (D) shows the results of comparing the F-value for each of this embodiment, Comparative Example 1, and Comparative Example 2. Figure 11 (E) shows the results of comparing the accuracy for each of this embodiment, Comparative Example 1, and Comparative Example 2.

Sensitivity (recall) is calculated as TP/(TP+FN), and the higher the prediction accuracy, the larger the value. Specificity is calculated as TN/(TN+FP), and the higher the prediction accuracy, the larger the value. Precision is calculated as TP/(TP+FP), and the higher the prediction accuracy, the larger the value. F-measure is calculated as (2*recall*precision)/(recall+precision), and the higher the prediction accuracy, the larger the value. Precision is calculated as (TP+TN)/(TN+FP+FN+FP), and the higher the prediction accuracy, the larger the value.

As shown in FIG. 11, the prediction results of this embodiment are greater than the prediction results of Comparative Example 1 and Comparative Example 2 in terms of sensitivity, specificity, precision, F-measure, and accuracy.

As such, a comparison using sensitivity, specificity, precision, F-measure, and accuracy rate shows that the prediction model 121 of this embodiment has higher prediction accuracy than both Comparative Example 1 and Comparative Example 2.

Figure 12 shows a graph of the PR (Precision Recall) curve for each of the present embodiment, Comparative Example 1, and Comparative Example 2, with the recall rate on the horizontal axis and the precision rate on the vertical axis. In the graph, the area below the PR curve is called the AUPRC (Area Under the Precision Recall Curve) and takes values between 0 and 1. The higher the prediction accuracy, the closer the AUPRC value is to 1.

As shown in Figure 12, the prediction results of this embodiment have an AUPRC closer to 1 than the prediction results of each of Comparative Examples 1 and 2.

As such, a comparison using the PR curve shows that the prediction model 121 of this embodiment has higher prediction accuracy than both Comparative Example 1 and Comparative Example 2.

Figure 13 summarizes the results of comparing this embodiment with Comparative Example 1 and Comparative Example 2. As shown in Figure 13, the prediction model 121 of this embodiment has higher prediction accuracy than both Comparative Example 1 and Comparative Example 2.

The TIMI risk index of Comparative Example 1 and the GRACE score of Comparative Example 2 existed before percutaneous coronary intervention became the primary treatment for AMI, and therefore are not compatible with the current standard of care. In contrast, the prediction model 121 of this embodiment was generated using sample data on AMI patients after percutaneous coronary intervention became the primary treatment for AMI, and therefore is compatible with the current standard of care.

The prediction model 121 of this embodiment uses machine learning by AI (Artificial Intelligence), whereas the TIMI risk index of Comparative Example 1 and the GRACE score of Comparative Example 2 do not use AI.

The prediction model 121 of this embodiment is highly convenient and practical because it can predict the prognosis of AMI patients with higher accuracy than both the TIMI risk index of Comparative Example 1 and the GRACE score of Comparative Example 2 by allowing the user to input the features of AMI patients into the prediction device 1.

[Display prediction results]
An example of display of a prediction result by the prediction device 1 will be described with reference to Fig. 14. Fig. 14 is a diagram for explaining an example of display of a prediction result by the prediction device 1 according to the first embodiment.

As shown in FIG. 14, the display device 20 of the prediction device 1 displays, on the display 25, an image 21 and an image 22 for the user to input the value of at least one feature, and an image 23 including a prediction result calculated based on the value of at least one feature.

Image 21 includes icons that allow the user to input the results of the Killip classification. Image 22 includes icons that allow the user to input max CPK, age, creatinine concentration, C-reactive protein amount, hemoglobin amount, white blood cell count, blood glucose level, systolic blood pressure, and heart rate. Image 23 includes information indicating, for example, the probability of in-hospital death as a prediction result.

The user can obtain the prediction result displayed by image 23 by inputting at least one feature to the prediction device 1 via image 21 and image 22 using a keyboard 31 and a mouse 32, etc.

The prediction device 1 may display information indicating the probability of survival as a prediction result on the display 25 of the display device 20. Furthermore, the prediction device 1 may present the prediction result not as a probability but as points calculated according to a predetermined criterion.

[Prediction Device Processing]
The processing of the prediction device 1 will be described with reference to Fig. 15. Fig. 15 is a flowchart of the prediction processing executed by the prediction device 1 according to the first embodiment. The processing steps (hereinafter abbreviated as "S") shown in Fig. 15 are realized by the calculation device 11 executing the prediction program 122.

As shown in FIG. 15, the prediction device 1 acquires at least one feature via the input interface 13 (S1). The prediction device 1 predicts the prognosis of the patient based on the acquired at least one feature and the trained prediction model 121 (S2). The prediction device 1 outputs the prognosis prediction result to the display device 20 via the output interface 14.

As described above, the prediction device 1 according to the first embodiment can predict the prognosis based on the patient's features and the trained prediction model 121, and therefore can predict the prognosis of AMI patients with higher accuracy. Therefore, the prediction device 1 can be useful for informed consent in clinical settings, and by using the prediction device 1, users can improve the quality of medical care for AMI patients.

<Prediction System According to Second Embodiment>
Hereinafter, a prediction system 1000 according to the second embodiment will be described, focusing only on differences from the prediction device 1 according to the first embodiment. Fig. 16 is a diagram showing the prediction system 1000 according to the second embodiment.

As shown in FIG. 16, the prediction system 1000 includes a user terminal 200 and a server device 300. The server device 300 is configured to communicate with the user terminal 200 via a network 500.

The user terminal 200 is an information terminal that performs specified information processing, such as a desktop PC (personal computer), a laptop PC, a smartphone, a smart watch, a wearable device, and a tablet PC.

Similar to the control device 10 according to the first embodiment, the server device 300 includes an acquisition unit 330, a prediction unit 310, and an output unit 340. The acquisition unit 330 and the output unit 340 may be configured, for example, by a communication device for communicating with the user terminal 200 via the network 500. The prediction unit 310 may be configured, for example, by a calculation device such as a CPU, similar to the control device 10.

In the prediction system 1000 having such a configuration, a user inputs at least one feature of an AMI patient to the user terminal 200. The user terminal 200 outputs the at least one feature to the server device 300 via the network 500. The server device 300 acquires the at least one feature acquired from the user terminal 200, and predicts the prognosis of the patient based on the acquired at least one feature and the trained prediction model. The server device 300 outputs the prediction result to the user terminal 200 via the network 500. The user terminal 200 displays the prediction result acquired from the server device 300 on the display 250.

This allows a user to input the features of an AMI patient into the user terminal 200 and use the server device 300 to predict the patient's prognosis.

The server device 300 may be configured to communicate with multiple user terminals 200, not just one user terminal 200. For example, in the prediction system 1000, the server device 300 may exist in the form of cloud computing. In this case, the server device 300 may predict the prognosis of a patient based on at least one feature acquired from each of the multiple user terminals 200, and output the prediction result to each of the multiple user terminals 200. In this way, for example, it becomes possible to transmit the prediction result from the server device 300 of the core hospital to each user terminal 200 in response to an input from a user terminal 200 installed in each of multiple hospitals in the region.

The embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present disclosure is indicated by the claims rather than the description of the embodiments above, and is intended to include all modifications within the meaning and scope of the claims.

1 Prediction device, 10 Control device, 11 Arithmetic device, 12 Storage device, 13 Input interface, 14 Output interface, 20 Display device, 21, 22, 23 Image, 25, 250 Display, 31 Keyboard, 32 Mouse, 121 Prediction model, 122 Prediction program, 200 User terminal, 300 Server device, 310 Prediction unit, 330 Acquisition unit, 340 Output unit, 500 Network, 1000 Prediction system.

Claims (7)

A prediction device for predicting the prognosis of a patient with acute myocardial infarction, comprising:
An acquisition unit that acquires at least one feature of the patient;
A prediction unit that predicts the prognosis based on the at least one feature amount using a trained prediction model;
An output unit that outputs the prognosis prediction result ,
The at least one feature value is ten types of feature values, namely, maximum value of creatine phosphokinase in blood, amount of hemoglobin in blood, heart rate, concentration of creatinine in blood, systolic blood pressure, blood glucose level, age, classification result regarding the severity of cardiac dysfunction in the acute myocardial infarction, number of white blood cells in blood, and amount of C-reactive protein in blood . The prediction device according to claim 1 , wherein the prediction model is generated by machine learning based on the ten types of feature amounts and whether or not the patient has died in the hospital. The prediction device according to claim 2 , wherein the machine learning algorithm includes a random forest. Further comprising a display device,
The output unit outputs the prediction result to the display device.
4. The prediction device according to claim 1 , wherein the display device displays an image for inputting values of the ten types of feature quantities and an image including the prediction result. A method for predicting the prognosis of a patient with acute myocardial infarction by a computer, comprising:
acquiring at least one characteristic of the patient;
predicting the prognosis based on the at least one feature using a trained prediction model;
outputting the result of the prognosis prediction ;
The at least one feature quantity is ten types of feature quantities, namely, maximum value of creatine phosphokinase in blood, amount of hemoglobin in blood, heart rate, concentration of creatinine in blood, systolic blood pressure, blood glucose level, age, classification result regarding severity of cardiac dysfunction in the acute myocardial infarction, number of white blood cells in blood, and amount of C-reactive protein in blood . A prediction program for predicting the prognosis of a patient with acute myocardial infarction, comprising:
On the computer,
acquiring at least one characteristic of the patient;
predicting the prognosis based on the at least one feature using a trained prediction model;
and outputting the result of the prognosis prediction .
The at least one feature value is ten types of feature values, namely, maximum value of creatine phosphokinase in blood, amount of hemoglobin in blood, heart rate, concentration of creatinine in blood, systolic blood pressure, blood glucose level, age, classification result regarding severity of cardiac dysfunction in the acute myocardial infarction, number of white blood cells in blood, and amount of C-reactive protein in blood . A prediction system for predicting the prognosis of a patient with acute myocardial infarction, comprising:
A user terminal;
A server device configured to communicate with the user terminal,
The server device includes:
an acquisition unit that acquires at least one feature of the patient from the user terminal;
A prediction unit that predicts the prognosis based on the at least one feature amount by using a trained prediction model;
an output unit that outputs the prognosis prediction result to the user terminal ;
The at least one feature value is ten types of feature values, namely, maximum value of creatine phosphokinase in blood, amount of hemoglobin in blood, heart rate, concentration of creatinine in blood, systolic blood pressure, blood glucose level, age, classification result regarding severity of cardiac dysfunction in the acute myocardial infarction, number of white blood cells in blood, and amount of C-reactive protein in blood .

Images