Program, information processing device, information processing method, and model generation method

Abstract

This program acquires attribute information of a patient, surgical information regarding a surgery scheduled for the patient and hemostatic tool information regarding a hemostatic tool to be used to stop bleeding at a puncture site of the patient. Then, the program inputs the thus acquired attribute information, surgical information and hemostatic tool information to a model, said model having learned training data so that, when the attribute information, surgical information and hemostatic tool information are input thereto, the model outputs the hemostasis time required to stop bleeding at the puncture site. Thus, the program makes a computer perform the process of outputting the hemostasis time.

Description

[Technical Field]

[0001] The present invention relates to a program, an information processing device, an information processing method, and a model generation method. [Background technology]

[0002] When a medical device such as a catheter is inserted into a patient's blood vessel for treatment or examination, for example, an introducer sheath is placed at the puncture site and the medical device is inserted into the blood vessel through the introducer sheath to treat the lesion. In this case, a specific hemostatic device is used to stop bleeding at the puncture site where the introducer sheath is inserted. For example, Patent Document 1 discloses a balloon-pressure type hemostatic device that enhances hemostatic effect by providing an auxiliary balloon between a band wrapped around the patient's limb and a balloon inside the band that presses against the puncture site. The introducer sheath is a component of the introducer. The introducer is composed of an introducer sheath and a dilator inserted into the lumen of the introducer sheath. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2004-41599 Summary of the Invention [Problem to be solved by the invention]

[0004] When performing hemostasis at a puncture site, hemostasis must be continued for a certain period of time. However, because the ease of hemostasis varies from patient to patient, it is difficult for medical professionals to predict how long it will take. This requires timely monitoring of the patient's hemostasis status, which places a heavy burden on medical professionals.

[0005] In one aspect, an object of the present invention is to provide a program or the like that can suitably predict the hemostasis time required to stop bleeding at a puncture site. [Means for solving the problem]

[0006] A program according to one aspect acquires a patient's attribute information, surgery information relating to the surgery the patient will undergo, and hemostatic device information relating to the hemostatic device used to stop bleeding at the patient's puncture site, and causes a computer to execute a process of inputting the acquired attribute information, surgery information, and hemostatic device information into a model that has learned training data so as to output the hemostatic time required to stop bleeding at the puncture site when the attribute information, surgery information, and hemostatic device information are input, and outputting the hemostatic time. [Effects of the Invention]

[0007] In one aspect, the time required for hemostasis at the puncture site can be suitably predicted. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is an explanatory diagram showing an example of the configuration of a hemostasis time prediction system. [Figure 2] FIG. 2 is a block diagram illustrating an example of the configuration of a server. [Figure 3] FIG. 2 is a block diagram illustrating an example of the configuration of a terminal. [Figure 4] 10 is an explanatory diagram showing an example of the record layout of a hemostatic device DB, a scoring table, and an implementation procedure table. FIG. [Figure 5] FIG. 1 is an explanatory diagram showing an overview of a first embodiment. [Figure 6] FIG. 10 is an explanatory diagram illustrating an example of an input screen. [Figure 7] FIG. 10 is an explanatory diagram showing an output image of implementation procedure information. [Figure 8] 10 is a flowchart illustrating a procedure for generating a prediction model. [Figure 9] 10 is a flowchart showing the procedure of a hemostasis time prediction process. [Figure 10]FIG. 10 is an explanatory diagram showing an overview of a second embodiment. [Figure 11] 10 is a flowchart showing the procedure of a hemostasis time prediction process according to the second embodiment. [Figure 12] 10 is a flowchart illustrating a procedure for updating a prediction model. DETAILED DESCRIPTION OF THE INVENTION

[0009] The present invention will be described in detail below with reference to the drawings showing embodiments thereof. (Embodiment 1) Fig. 1 is an explanatory diagram showing an example of the configuration of a hemostasis time prediction system. In this embodiment, a hemostasis time prediction system is described that predicts the hemostasis time required to stop bleeding at a puncture site where a therapeutic instrument such as a catheter is inserted for a patient undergoing surgery using the therapeutic instrument. The hemostasis time prediction system includes an information processing device 1, a terminal 2, and an electronic medical record server 3. Each device is communicatively connected to a network N such as the Internet or a LAN (Local Area Network).

[0010] The information processing device 1 is an information processing device capable of various information processing and sending and receiving information, such as a server computer or a personal computer. In this embodiment, the information processing device 1 is assumed to be a server computer, and for simplicity, will be referred to as server 1 below. The server 1 functions as a generation device that performs machine learning to learn predetermined training data, and generates a prediction model 50 (see FIG. 5 ) that predicts (outputs) the hemostasis time required to stop bleeding at the puncture site, using as input attribute information of a patient undergoing surgery, surgical information related to the surgery the patient will undergo, and hemostatic device information related to the hemostatic device used to stop bleeding at the puncture site.

[0011] Terminal 2 is an information processing terminal used by a user (medical worker) of this system, such as a tablet terminal, a personal computer, a smartphone, etc. In this embodiment, a prediction model 50 generated by server 1 is installed in terminal 2, and terminal 2 predicts the bleeding time based on the prediction model 50.

[0012] In this embodiment, the local terminal 2 predicts the time required for hemostasis, but the server 1 on the cloud may also predict the time required for hemostasis.

[0013] The electronic medical record server 3 is a database server that manages the medical records of patients and stores information about each patient. The terminal 2 acquires the attribute information of the patient from the electronic medical record server 3, inputs it into the prediction model 50, and predicts the time to stop bleeding.

[0014] In this embodiment, vascular surgery using a catheter, such as PCI (Percutaneous Coronary Intervention), will be described as an example of surgery. That is, an introducer sheath is placed at a puncture site such as the wrist or foot, a catheter is inserted into a blood vessel via the introducer sheath, and vascular treatment such as balloon vasodilation and stent placement will be described as an example. Note that in vascular surgery using a catheter, a guiding sheath may be used instead of an introducer. A guiding sheath is an instrument consisting of a guiding catheter and a dilator inserted into the lumen of the guiding catheter. When a guiding sheath is used, a guiding catheter is placed at a puncture site such as the wrist or foot, and a catheter is inserted into a blood vessel via the guiding catheter to perform vascular treatment such as balloon vasodilation and stent placement.

[0015] The location of the lesion to be treated in vascular surgery is not limited to the coronary artery. For example, the femoral artery or popliteal artery of the lower limb may be treated in vascular treatment.

[0016] In this specification, an instrument used to stop bleeding at a puncture site is referred to as a "hemostatic instrument," an instrument placed at the puncture site (in the case of an introducer, an introducer sheath, and in the case of a guiding sheath, a guiding catheter) is referred to as a "sheath," and instruments other than hemostatic instruments and sheaths are collectively referred to as "treatment instruments."

[0017] As described above, terminal 2 uses prediction model 50 to predict the hemostatic time required to stop bleeding at the puncture site. As an example of a hemostatic device, the lower left of Figure 1 illustrates a balloon-pressure hemostatic device 4 that uses a balloon to press against the puncture site. The hemostatic device 4 illustrated in Figure 1 is a band-type hemostatic device, and a balloon 42 into which a fluid such as air or liquid is injected is provided at a curved portion 41 in the center of the band. When using hemostatic device 4 to stop bleeding, a medical professional attaches hemostatic device 4 to the patient's limb so that it covers the puncture site and presses against the puncture site by injecting a fluid into the balloon. The medical professional reduces the pressure of hemostatic device 4 over time to prevent the blood vessel at the puncture site from becoming blocked.

[0018] The hemostatic device 4 described above is just one example, and there are also hemostatic devices that use a hard material such as plastic for pressing instead of a balloon. The form of the hemostatic device is not particularly limited, and any hemostatic device may be selected.

[0019] Terminal 2 inputs the hemostatic device information about the hemostatic device, as well as the patient's attribute information and surgery information, into the prediction model 50 to predict the time it will take to stop bleeding. Terminal 2 presents the predicted time it will take to stop bleeding to the user and guides the user through the steps to perform the hemostatic procedure using the hemostatic device. Predicting the time it will take to stop bleeding makes it easier for the user to predict the timing of the procedure (decompression in the case of a hemostatic device that presses against the puncture site), thereby reducing the burden on the user (healthcare worker).

[0020] 2 is a block diagram showing an example of the configuration of the server 1. The server 1 includes a control unit 11, a main memory unit 12, a communication unit 13, and an auxiliary memory unit . The control unit 11 has one or more arithmetic processing devices such as a CPU (Central Processing Unit), an MPU (Micro-Processing Unit), a GPU (Graphics Processing Unit), etc., and performs various information processing, control processing, etc. by reading and executing a program P1 stored in the auxiliary storage unit 14. The main storage unit 12 is a temporary storage area such as an SRAM (Static Random Access Memory), a DRAM (Dynamic Random Access Memory), or a flash memory, and temporarily stores data necessary for the control unit 11 to execute arithmetic processing. The communication unit 13 is a communication module for performing communication-related processing, and transmits and receives information to and from the outside.

[0021] The auxiliary storage unit 14 is a non-volatile storage area such as a large-capacity memory or a hard disk, and stores the program P1 and other data required for the control unit 11 to execute processing. The auxiliary storage unit 14 also stores a prediction model 50, a hemostatic device DB 141, a scoring table 142, and an implementation procedure table 143. The prediction model 50 is a machine learning model that has learned training data and inputs patient attribute information, surgery information, and hemostatic device information, and outputs a hemostatic time. The prediction model 50 is intended to be used as a program module constituting part of artificial intelligence software. The hemostatic device DB 141 is a database that stores hemostatic device information. The scoring table 142 is a table referenced when scoring data for specific input items entered into the prediction model 50, as described below. The implementation procedure table 143 is a table referenced when outputting implementation procedure information that guides the implementation procedure of a hemostatic procedure based on the hemostatic time output from the prediction model 50, as described below.

[0022] The auxiliary storage unit 14 may be an external storage device connected to the server 1. The server 1 may be a multi-computer consisting of multiple computers, or may be a virtual machine virtually constructed by software.

[0023] In this embodiment, the server 1 is not limited to the above configuration, and may include, for example, an input unit for accepting operation input, a display unit for displaying images, etc. The server 1 may also include a reading unit for reading a portable storage medium 1a such as a CD (Compact Disk)-ROM or a DVD (Digital Versatile Disc)-ROM, and may read and execute the program P1 (program product) from the portable storage medium 1a. Alternatively, the server 1 may read the program P1 from a semiconductor memory 1b.

[0024] 3 is a block diagram showing an example of the configuration of the terminal 2. The terminal 2 includes a control unit 21, a main memory unit 22, a communication unit 23, a display unit 24, an input unit 25, and an auxiliary memory unit 26. The control unit 21 has one or more arithmetic processing units such as a CPU, an MPU, or the like, and performs various information processing, control processing, and the like by reading and executing the program P2 stored in the auxiliary storage unit 26. The main storage unit 22 is a temporary storage area such as a RAM, and temporarily stores data necessary for the control unit 21 to execute arithmetic processing. The communication unit 23 is a communication module for performing communication-related processing, and transmits and receives information to and from the outside. The display unit 24 is a display screen such as an LCD display, and displays images. The input unit 25 is an operation interface such as a touch panel, and accepts operation inputs from the user. The auxiliary storage unit 26 is a non-volatile storage area such as a hard disk, large-capacity memory, or the like, and stores the program P2, the prediction model 50, and other data necessary for the control unit 21 to execute processing.

[0025] The terminal 2 may be provided with a reading unit for reading a portable storage medium 2a such as a CD-ROM, and may read and execute the program P2 (program product) from the portable storage medium 2a. Alternatively, the terminal 2 may read the program P2 from the semiconductor memory 2b.

[0026] FIG. 4 is an explanatory diagram showing an example of the record layout of the hemostatic instrument DB 141, the scoring table 142, and the implementation procedure table 143. Hemostatic device DB141 includes a device ID column, a device name column, a type column, and a device information column. The device ID column stores a device ID for identifying each hemostatic device. The device name column, type column, and device information column each store, in association with the device ID, the name (product name) and type of hemostatic device, and other hemostatic device information (for example, in the case of a balloon compression type hemostatic device, the pressure applied when the balloon is inflated, the contact area of the balloon that comes into contact with the patient's limb, etc.).

[0027] The scoring table 142 includes an item string, an input string, and a score string. The item string stores the items of input data to be scored. The input string and score string store, in association with the input items, the input data before being scored and the score into which the input data is converted. For example, if the patient's age is 90 years or older, the input data related to age is converted to 5 points.

[0028] The implementation procedure table 143 includes a predicted time column, a used instrument column, a puncture site column, and an implementation procedure column. The predicted time column, the used instrument column, and the puncture site column each store a predicted value of hemostasis time, a type of hemostasis instrument, and a puncture site that can be output from the prediction model 50. The implementation procedure column stores implementation procedures for hemostasis work in association with hemostasis time, a type of hemostasis instrument, and a puncture site. The implementation procedure column stores text indicating the work to be performed at each elapsed time, for example, in association with the time elapsed since the start of hemostasis.

[0029] Fig. 5 is an explanatory diagram showing an overview of the first embodiment. Fig. 5 conceptually illustrates how attribute information of a patient and the like is input into a prediction model 50 to predict the time required for hemostasis, and how the procedure for performing the hemostasis operation is guided based on the predicted time required for hemostasis. An overview of this embodiment will be described based on Fig. 5.

[0030] The prediction model 50 is a machine learning model that has learned predetermined training data, such as a neural network generated by deep learning. The prediction model 50 has an input layer that accepts input data, an intermediate layer that extracts features of the input data, and an output layer that outputs a predicted value of the hemostasis time based on the features extracted in the intermediate layer. The terminal 2 inputs patient attribute information, surgery information, and hemostasis device information into the prediction model 50 to predict the hemostasis time.

[0031] The attribute information is basic information about the patient undergoing surgery, and includes the patient's age, sex, race, body shape (e.g., BMI (Body Mass Index)), and disease name. The disease name is the name of the disease that the patient undergoing surgery is suffering from. For example, the disease name determines whether the disease that the patient undergoing surgery is suffering from is ACS (Acute Coronary Syndrome). Note that these are examples of attribute information, and other patient attributes may also be included. For example, the terminal 2 obtains attribute information from an electronic medical record server 3 that manages patient medical records, and inputs the attribute information into the prediction model 50.

[0032] The surgical information is data related to the surgery to be performed on the patient, and includes, for example, the puncture site, the sheath size of the sheath to be placed at the puncture site, and the ACT value (Activated Coagulation Time). Note that these are examples of surgical information, and the information may also include data related to other surgeries. For example, the terminal 2 accepts input of each surgical information from the user before and during the surgery, and inputs the information into the prediction model 50. Note that, when the instrument to be placed at the puncture site is an introducer, the sheath size indicates the maximum outer diameter of the therapeutic instrument that can be inserted into the lumen of the introducer sheath of the introducer (the French size of the introducer sheath). Furthermore, when the instrument to be placed at the puncture site is a guiding sheath, the sheath size indicates the maximum outer diameter of the therapeutic instrument that can be inserted into the lumen of the guiding catheter of the guiding sheath.

[0033] The hemostatic device information is data related to the hemostatic device used to stop bleeding at the puncture site, and includes, for example, the type of hemostatic device (balloon compression type, hard member compression type, etc.), the name of the hemostatic device (product name), and other information (for example, in the case of a balloon compression type, the pressure exerted by the balloon, the contact area of the balloon, etc.). Information on each hemostatic device is stored in the hemostatic device DB 141, and, for example, the terminal 2 obtains the hemostatic device information from the server 1 and inputs it into the prediction model 50.

[0034] There are no particular limitations on the method of acquiring the attribute information, surgery information, and hemostatic device information to be input into the prediction model 50. For example, the terminal 2 may acquire not only the attribute information but also the surgery information and hemostatic device information from the electronic medical record server 3. Alternatively, the user may manually input all input data into the terminal 2.

[0035] The server 1 learns from training data including the above-mentioned various data and generates a prediction model 50. The training data is data in which the correct value of the hemostasis time is assigned to training attribute information, surgery information, and hemostatic device information. For example, the server 1 acquires from the electronic medical record server 3 attribute information of patients who have actually undergone surgery, surgery information related to the surgery undergone by the patient, hemostatic device information related to the hemostatic device used to stop bleeding at the puncture site, and the actual value of the hemostatic time required to stop bleeding at the puncture site. The server 1 provides the above-mentioned various data as training data to the prediction model 50 and performs learning.

[0036] Specifically, the server 1 inputs training attribute information, surgery information, and hemostatic device information into the prediction model 50 and obtains a predicted value of hemostasis time as an output. As shown in FIG. 5 , in this embodiment, the prediction model 50 predicts hemostasis time in five levels: “within 60 minutes,” “60 to 120 minutes,” “120 to 180 minutes,” “180 to 240 minutes,” and “240 to 300 minutes.” The prediction model 50 may also make predictions in three or fewer levels or six or more levels. It may also make predictions without specifying specific values for hemostasis time, such as “long,” “normal,” and “short.” While the present embodiment treats hemostasis time prediction as a classification problem, it may also be treated as a regression problem that predicts hemostasis time as a continuous value. The server 1 compares the predicted value of hemostasis time obtained from the prediction model 50 with the correct answer and optimizes the parameters of the prediction model 50 (e.g., weights between neurons) so that the predicted values approximate each other. In this manner, the server 1 generates the prediction model 50. The server 1 transmits the data of the generated prediction model 50 to the terminal 2, which then installs the model.

[0037] When performing the above-mentioned learning, the server 1 learns different training data for each predetermined category and generates multiple prediction models 50, 50, 50... corresponding to each category. The categories are, for example, the puncture site into which a catheter is inserted and / or the hemostatic device. The server 1 categorizes the training data according to the puncture site and the hemostatic device, and provides the training data for each category to a separate prediction model 50 to generate a prediction model 50 corresponding to each category (puncture site and hemostatic device).

[0038] The categories of puncture sites may be classified, for example, as "wrist" (radial artery), "foot" (femoral artery), or "elbow" (cubital artery), or may be classified by more detailed locations. Furthermore, the categories of hemostatic devices may be classified according to the type of hemostatic device, such as "balloon compression type" or "hard member compression type," or may be classified by individual hemostatic device (product). Thus, the specific categories of puncture sites and hemostatic devices are not particularly limited.

[0039] The time to stop bleeding depends on the location of the puncture site (the thickness of the blood vessels and the force of the blood flow differ between the arm, leg, etc., so the ease of stopping bleeding differs), and the pressure applied to the puncture site depends on the hemostatic device. Therefore, by categorizing the puncture site and the hemostatic device, learning and prediction of the time to stop bleeding can be performed effectively.

[0040] When actually predicting the time required for hemostasis, terminal 2 acquires the patient's attribute information, etc. from electronic medical record server 3, etc. Then, terminal 2 selects a prediction model 50 according to the puncture site of the patient undergoing surgery and the hemostatic instrument used to stop the bleeding, and inputs the attribute information, etc. into prediction model 50 to predict the time required for hemostasis.

[0041] Fig. 6 is an explanatory diagram showing an example of an input screen. Fig. 6 illustrates an input screen displayed on terminal 2 for inputting information necessary for predicting the hemostasis time. As shown in Fig. 6, the input screen includes input fields for inputting the date and time of the procedure (surgery), the patient's name, and other information about the procedure (surgery) and the hemostatic device. For example, terminal 2 accepts input on this screen such as the product name of the sheath to be placed at the puncture site, the sheath size of the sheath to be placed at the puncture site, the puncture site, the ACT value, the type and product name of the hemostatic device, etc.

[0042] The user inputs various information into terminal 2 before and during surgery. For example, before surgery, the user inputs the patient's name, the type of hemostatic device, the product name, etc. The user also inputs the remaining information (sheath size, puncture site, etc.) when the procedure is complete.

[0043] Terminal 2 acquires patient attribute information corresponding to the patient name input on the input screen from electronic medical record server 3. Terminal 2 also accepts input of surgical information such as the sheath size of the sheath to be placed at the puncture site, as shown in Fig. 6. Terminal 2 also acquires hemostatic device information related to the hemostatic device input on the input screen from server 1.

[0044] For example, the terminal 2 displays a predetermined alert (not shown) if there is a missing input item in the patient attribute information acquired from the electronic medical record server 3 among the input items to be input into the prediction model 50. In this case, the terminal 2 displays the input items to be input, prompts the user to input them, and accepts input of attribute information related to the items from the user.

[0045] Continuing the explanation, returning to Figure 5, terminal 2 selects a prediction model 50 according to the puncture site indicated by the surgery information and the hemostatic device indicated by the hemostatic device information. Terminal 2 then inputs the patient's attribute information, surgery information, and hemostatic device information into the selected prediction model 50, and predicts the hemostatic time.

[0046] In this case, the terminal 2 scores some or all of the input items to be input to the prediction model 50, which are included in the patient's attribute information and / or surgery information, according to a predetermined rule and inputs the scores. The input items are the patient's age, sex, race, ACT value, sheath size, whether the patient is suffering from ACS, and body shape. Note that the input items to be scored may be some of these input items, or may include other input items.

[0047] In this embodiment, each input item is scored according to the scoring table 142 (see FIG. 4 ) so that the prediction model 50 can effectively learn and predict each input item. Specifically, because the risk of bleeding increases with age, the patient's age is scored so that the older the patient, the higher the score. Furthermore, because the male-female ratio and the risk of bleeding complications are higher in women undergoing PCI, the scores for women are higher than those for men. Furthermore, because the bleeding risk varies depending on race (especially East Asians), Asian patients are given higher scores. Furthermore, because the ACT value is associated with a higher bleeding risk, the higher the score, the higher the ACT value. Furthermore, because the larger the puncture site, the higher the bleeding risk, the larger the sheath size of the sheath placed at the puncture site, the higher the score. Furthermore, because patients suffering from ACS are at a higher risk of bleeding complications after surgery, patients with ACS are given higher scores. Furthermore, because thin patients are at a higher risk of bleeding complications, patients with a BMI below a threshold are given higher scores.

[0048] As described above, terminal 2 scores each input item in light of vascular surgery (PCI, etc.), and inputs the scored data for each input item into prediction model 50 to predict the hemostasis time. As already mentioned, terminal 2 predicts the hemostasis time in multiple stages, such as "within 60 minutes," "60 to 120 minutes," "120 to 180 minutes," etc.

[0049] Terminal 2 outputs implementation procedure information that guides the implementation procedure of the hemostasis work using the hemostasis instrument according to the predicted hemostasis time. Specifically, as shown on the right side of Fig. 5, terminal 2 outputs text data representing the implementation procedure according to the predicted hemostasis time. Terminal 2 communicates with server 1 and acquires implementation procedure information from implementation procedure DB 143 that corresponds to the hemostasis time output from prediction model 50 and that also corresponds to the puncture site of the patient and the hemostasis instrument that was treated.

[0050] FIG. 7 is an explanatory diagram showing an output image of implementation procedure information. FIG. 7 illustrates an output image of text data representing the implementation procedure. For example, terminal 2 outputs the implementation procedure information acquired from server 1 to a printer (not shown), causing a document describing the implementation procedure to be printed. For example, as shown in FIG. 7, the document may include the patient's name, predicted time for hemostasis, predicted start and end times for hemostasis, etc. The document also includes text describing the implementation procedure for hemostasis work.

[0051] Specifically, the document contains text describing the tasks to be performed at each elapsed time according to the amount of time that has elapsed since the start of hemostasis. For example, when using a pressure-type hemostasis device, the document describes the decompression tasks to be performed at each elapsed time (time) and the observation points to be observed during the tasks, as shown in Figure 7. Terminal 2 reduces the workload of the user by providing chronological guidance on the decompression tasks to be performed at each elapsed time.

[0052] In this embodiment, the implementation procedure information is described as being printed out, but it goes without saying that the implementation procedure information may also be displayed on a screen.

[0053] 8 is a flowchart showing the procedure of the process for generating the prediction model 50. The process for generating the prediction model 50 by machine learning will be described with reference to FIG. The control unit 11 of the server 1 acquires training data for generating the prediction model 50 (step S11). For example, the training data is data in which the actual performance value of the time required for hemostasis is assigned as a correct answer value to attribute information of a patient who actually underwent surgery, surgery information related to the surgery underwent by the patient, and hemostatic device information related to the hemostatic device used during the surgery. The control unit 11 refers to the scoring table 142 and scores the data of some or all of the input items included in the attribute information and surgery information (step S12).

[0054] Based on the training data, the control unit 11 generates a prediction model 50 that outputs a hemostasis time when patient attribute information, surgery information, and hemostasis device information are input (step S13). Specifically, as described above, the control unit 11 generates a neural network as the prediction model 50. The control unit 11 provides training data in which some or all of the patient attribute information and surgery information have been scored to the prediction model 50 to output a predicted value of the hemostasis time, compares the predicted value with a correct value (actual value), and optimizes parameters such as weights so that the two approximate each other. The control unit 11 learns according to the puncture site where the treatment device (catheter) is inserted and the hemostasis device, and generates a prediction model 50 corresponding to each puncture site and hemostasis device. The control unit 11 then completes the series of processes.

[0055] 9 is a flowchart showing the procedure of the hemostasis time prediction process. The process of predicting the hemostasis time using the prediction model 50 will be described with reference to FIG. The control unit 21 of the terminal 2 accepts input of information such as the patient name, surgery information, and hemostatic device (step S31). The control unit 21 acquires patient attribute information corresponding to the input patient name from the electronic medical record server 3 and hemostatic device information from the server 1 (step S32). The control unit 21 determines whether any input items required for input to the prediction model 50 are missing from the patient attribute information acquired in step S32 (step S33). If it is determined that any input items are missing (S33: YES), the control unit 21 displays an alert indicating that information corresponding to the items should be entered, and accepts input of information corresponding to the items (step S34).

[0056] After executing the process of step S34, or if the answer is NO in step S33, control unit 21 scores the data for some or all of the input items included in the attribute information and / or surgery information (step S35). Control unit 21 inputs the attribute information and surgery information, for which the data for some or all of the input items has been scored, and the hemostatic device information into prediction model 50, and predicts the time to stop bleeding (step S36). In this case, control unit 21 selects prediction model 50 according to the puncture site and the hemostatic device, and inputs various information into the selected prediction model 50 to predict the time to stop bleeding.

[0057] The control unit 21 communicates with the server 1 and outputs implementation procedure information corresponding to the hemostasis time predicted in step S36, which information corresponds to the puncture site on the patient and the hemostasis device to be used for the hemostasis procedure (step S37). For example, the control unit 21 outputs text data representing the implementation procedure information to a printer, and causes a document to be printed out describing the work to be performed (e.g., decompression work and observation points) at each elapsed time after the start of hemostasis. The control unit 21 then completes the series of processes.

[0058] As described above, according to the first embodiment, it is possible to suitably predict the hemostasis time required to stop bleeding at the puncture site.

[0059] Furthermore, according to the first embodiment, by preparing a plurality of prediction models 50, 50, 50 . . . according to the puncture site and / or the hemostatic device, it is possible to more appropriately predict the hemostatic time.

[0060] Furthermore, according to the first embodiment, by scoring the age, sex, etc. of the patient in light of catheter vascular surgery, it is possible to more appropriately learn and predict the hemostasis time.

[0061] Furthermore, according to the first embodiment, by presenting implementation procedure information according to the hemostasis time (and the puncture site and hemostasis device), it is possible to suitably support the user performing the hemostasis work.

[0062] (Embodiment 2) In the first embodiment, a prediction model 50 is selected depending on the hemostatic device used in the hemostatic operation. In the present embodiment, a plurality of prediction models 50, 50, 50... are used to predict the hemostatic time required for each hemostatic device and present the prediction to the user. Note that the same reference numerals are used to designate content that overlaps with the first embodiment, and explanations thereof will be omitted.

[0063] Fig. 10 is an explanatory diagram showing an overview of embodiment 2. Fig. 10 conceptually illustrates how patient attribute information and the like are input into prediction models 50, 50, 50... corresponding to hemostatic instruments A, B, C..., and how the hemostatic time when each hemostatic instrument is used is predicted.

[0064] 10, in this embodiment, terminal 2 inputs patient attribute information, surgery information, and hemostatic device information into each prediction model 50, and predicts the time required for hemostasis when each hemostatic device is used. Naturally, each prediction model 50 is input with the hemostatic device information for the corresponding hemostatic device A, B, C, etc.

[0065] For example, terminal 2 displays on the screen the predicted value of the hemostasis time output from each prediction model 50, and presents to the user the hemostasis time when each hemostasis device is used. In addition to the hemostasis time, hemostasis device information for each hemostasis device (type of hemostasis device, product name, etc.) may also be displayed. Terminal 2 accepts input of the selection of the hemostasis device to be used in the hemostasis procedure, and outputs information on the procedure for performing the hemostasis procedure using the selected hemostasis device.

[0066] 11 is a flowchart showing the procedure of the hemostasis time prediction process according to embodiment 2. After executing the process of step S35, terminal 2 executes the following process. It is assumed that terminal 2 has already acquired hemostasis device information of all hemostasis devices from server 1 in step S32.

[0067] The control unit 21 of the terminal 2 inputs the patient's attribute information, surgery information, and hemostatic device information into a plurality of prediction models 50, 50, 50... corresponding to each hemostatic device, and predicts the hemostatic time when each hemostatic device is used (step S201). The control unit 21 displays the hemostatic time for each hemostatic device, and accepts a selection input to select the hemostatic device to be used for the hemostatic procedure (step S202). The control unit 21 outputs implementation procedure information for the selected hemostatic device (step S203), and the series of processes ends.

[0068] As described above, according to the second embodiment, by presenting to the user the time required for hemostasis when using each of a plurality of hemostasis devices, it is possible to assist in the selection of a hemostasis device.

[0069] (Embodiment 3) In this embodiment, a form in which the prediction model 50 is re-learned will be described.

[0070] In this embodiment, after surgery, the server 1 re-learns the prediction model 50 based on the patient's attribute information, etc. input into the prediction model 50 and the actual value of the hemostatic time required to stop bleeding at the actual puncture site for that patient. That is, after outputting (predicting) the hemostatic time, the server 1 updates the prediction model 50 by using the actual value of the hemostatic time as the correct value for re-learning.

[0071] For example, after surgery, terminal 2 transmits various data including the actual hemostasis time (for example, the start and end times of the hemostasis procedure) to electronic medical record server 3, which records the data as part of the patient's medical record. Server 1 accesses electronic medical record server 3, acquires various data including the actual hemostasis time (patient attribute information, surgery information, etc.), and uses the data as training data for re-learning. Note that the training data for re-learning may be acquired, for example, from terminal 2, and the acquisition method is not particularly limited.

[0072] The server 1 updates the prediction model 50 based on the training data for re-learning. That is, the server 1 inputs attribute information of the patient who underwent surgery, surgery information, and hemostatic device information into the prediction model 50 to predict the hemostasis time, compares the predicted value with the correct value (actual value) for re-learning, and updates the parameters of the prediction model 50 so that the two approximate each other. This makes it possible to improve the prediction accuracy through operation of this system.

[0073] 12 is a flowchart showing the procedure of the update process for the prediction model 50. After executing the process of step S37, the terminal 2 executes the following process. After completion of hemostasis, the control unit 21 of the terminal 2 receives from the user an input of the actual hemostasis time required for hemostasis (step S301). The control unit 21 transmits data such as surgery information and hemostasis device information, including the actual hemostasis time, to the electronic medical record server 3 and records it (step S302).

[0074] The control unit 11 of the server 1 acquires various data, including the actual value of the hemostasis time, from the electronic medical record server 3 as training data for re-learning (step S303). The control unit 11 updates the prediction model 50 based on the training data for re-learning (step S304). That is, the control unit 11 inputs attribute information of the patient who underwent surgery, surgery information, and hemostasis instrument information into the prediction model 50, compares the predicted value of the hemostasis time output from the prediction model 50 with the correct value (actual value), and updates parameters such as weights. The control unit 11 then completes the series of processes.

[0075] As described above, according to the third embodiment, the prediction accuracy can be improved through the operation of this system.

[0076] The embodiments disclosed herein are illustrative in all respects and should not be considered limiting. The scope of the present invention is defined by the claims, not by the above meaning, and is intended to include all modifications within the meaning and scope of the claims. [Explanation of symbols]

[0077] 1. Server (information processing device) 11 Control section 12 Main memory 13 Communications Department 14 Auxiliary storage P1 Program 50 Predictive Models 141 Hemostatic device DB 142 Scoring Table 143 Implementation Procedure Table 2. Terminal 21 Control section 22 Main memory 23 Communications Department 24 Display 25 Input section 26 Auxiliary storage P2 Program

Claims

1. Acquire attribute information of a patient, surgery information relating to a surgery the patient will undergo, and hemostatic device information relating to a hemostatic device to be used to stop bleeding at a puncture site of the patient; When the attribute information, surgery information, and hemostatic device information are input, the acquired attribute information, surgery information, and hemostatic device information are input into a model that has learned training data so as to output the hemostatic time required to stop bleeding at the puncture site, and the hemostatic time is output. A program that causes a computer to perform a process.

2. Among the plurality of models that have learned different training data according to the puncture site, the attribute information, surgery information, and hemostatic device information are input into the model corresponding to the puncture site of the patient, and the hemostatic time is output. The program according to claim 1.

3. Among the plurality of models that have learned different training data according to the hemostatic device, the attribute information, surgery information, and hemostatic device information are input into the model corresponding to the hemostatic device used to stop bleeding at the puncture site of the patient, and the hemostatic time is output. The program according to claim 1 or 2.

4. The attribute information, surgery information, and hemostatic device information are input to each of the plurality of models that have learned different training data according to the hemostatic device, and the hemostatic time when each of the plurality of hemostatic devices is used is output. The program according to claim 1 or 2.

5. Among the input items included in the attribute information or surgery information, data of some or all of the input items is scored according to a predetermined rule; The scored data of the input items is input into the model. The program according to any one of claims 1 to 4.

6. the surgery is a vascular surgery using a catheter, The input items to be scored include the patient's age, sex, race, activated whole blood clotting time, sheath size of the sheath to be placed at the puncture site of the patient, whether the patient suffers from acute coronary syndrome, or body shape. The program according to claim 5.

7. outputting implementation procedure information that guides the implementation procedure of a hemostasis operation using the hemostasis instrument according to the hemostasis time output from the model; The program according to any one of claims 1 to 6.

8. the hemostatic device is a device that presses against the puncture site, The implementation procedure information that guides the decompression work at each elapsed time after the start of hemostasis is output according to the hemostasis time. The program according to claim 7.

9. After outputting the hemostasis time, a performance value of the hemostasis time actually required to stop bleeding at the puncture site of the patient is acquired; The model is updated based on the attribute information, surgery information, and hemostatic device information input to the model, and the performance values. The program according to any one of claims 1 to 8.

10. an acquisition unit that acquires attribute information of a patient, surgery information regarding a surgery the patient will undergo, and hemostatic device information regarding a hemostatic device used to stop bleeding at a puncture site of the patient; an output unit that inputs the acquired attribute information, surgery information, and hemostatic device information into a model that has learned training data so as to output a hemostatic time required to stop bleeding at the puncture site when the attribute information, surgery information, and hemostatic device information are input, and outputs the hemostatic time; An information processing device comprising:

11. Acquire attribute information of a patient, surgery information relating to a surgery the patient will undergo, and hemostatic device information relating to a hemostatic device to be used to stop bleeding at a puncture site of the patient; When the attribute information, surgery information, and hemostatic device information are input, the acquired attribute information, surgery information, and hemostatic device information are input into a model that has learned training data so as to output the hemostatic time required to stop bleeding at the puncture site, and the hemostatic time is output. An information processing method in which processing is performed by a computer.

12. acquiring training data including attribute information of a patient, surgical information relating to a surgical operation the patient has undergone, hemostatic device information relating to a hemostatic device used to stop bleeding at a puncture site of the patient, and a hemostatic time required to stop bleeding at the puncture site; A trained model is generated based on the training data, and the trained model outputs the hemostasis time when the attribute information, surgery information, and hemostasis device information are input. A model generation method that causes a computer to execute processing.

Images