Information processing apparatus, computer program, information processing method, and learning model generation method

Abstract

Provided are an information processing device, a computer program, an information processing method, and a learning model generation method capable of predicting the amount of decrease in platelets during surgery or after surgery before surgery.SOLUTION: The information processing device includes a control unit, and the control unit acquires pre-surgery examination data of a surgery using an artificial heart and lung apparatus of a patient, inputs the acquired examination data to a learning model that predicts a platelet loss amount during the surgery or after the surgery in a case where the pre-surgery examination data is input, and predicts the platelet loss amount of the patient.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

[0002] In recent years, the number of surgeries for elderly people has increased, resulting in an increase in the amount of blood transfusions required. Meanwhile, the number of people donating blood has been declining, resulting in a shortage of donors and raw materials, leading to a shortage of blood products. Furthermore, despite the high cost per unit of blood product, some medical institutions are overusing blood products and discarding unused blood products at high rates. Appropriate use of blood products is required to ensure a stable supply of blood products.

[0003] Blood products (transfusion products) include red blood cell products, plasma products, platelet products, and whole blood products. Platelet products have an extremely short shelf life, so they must be prepared in the operating room immediately before surgery. Patent Document 1 discloses a storage method for platelet products that can store them while suppressing platelet aggregation by configuring a platelet product storage unit to be rotatable on two axes and uniformly stirring the platelets in three dimensions with a constant load. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2015-139487 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the amount of platelet transfusion required during surgery varies depending on the type of surgery and the patient's illness, making it difficult to predict the amount of platelet preparation required before surgery.

[0006] The present invention has been made in consideration of the above circumstances, and aims to provide an information processing device, a computer program, an information processing method, and a learning model generation method that can predict the amount of platelet reduction during or after surgery before surgery. [Means for solving the problem]

[0007] The present application includes multiple means for solving the above-mentioned problems, and one example is an information processing device that includes a control unit that acquires pre-operative test data for a patient undergoing surgery using an artificial heart-lung machine, and when the pre-operative test data is input, inputs the acquired test data into a learning model that predicts the amount of platelet reduction during or after surgery, thereby predicting the amount of platelet reduction in the patient. [Effects of the Invention]

[0008] According to the present invention, it is possible to predict the amount of platelet reduction during or after surgery before surgery. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a diagram illustrating an example of a configuration of an information processing system according to an embodiment of the present invention. [Figure 2] FIG. 10 is a diagram illustrating an example of patient information. [Figure 3] FIG. 10 is a diagram illustrating an example of test data. [Figure 4] FIG. 10 is a diagram showing an example of prediction of the amount of platelet reduction during surgery using a learning model. [Figure 5] FIG. 10 is a diagram illustrating an example of training data used to generate a learning model. [Figure 6] FIG. 10 is a diagram showing an example of data calculated for use in platelet transfusion management. [Figure 7] FIG. 10 is a diagram illustrating an example of a platelet transfusion amount management process performed by an information processing device. [Figure 8] FIG. 10 is a diagram illustrating an example of a platelet transfusion amount management process performed by an information processing device. [Figure 9] FIG. 10 is a diagram showing an example of detailed symptom description. [Figure 10] FIG. 10 is a diagram illustrating an example of a learning model generation process performed by an information processing device. [Figure 11] FIG. 10 is a diagram illustrating an example of a display screen displayed by the information processing device. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing an example of the configuration of an information processing system of this embodiment. The information processing system includes information processing device 50. Patient data server 100, test data server 200, blood test device 300, and transfusion product inventory management server 400 are connected to information processing device 50 via communication network 1. Note that in the example of FIG. 1, information processing device 50 and blood test device 300 are separate devices, but this configuration is not limited to this. For example, information processing device 50 may be incorporated into blood test device 300. Furthermore, information processing device 50 is not limited to a single device, but may be composed of multiple devices.

[0011] The patient data server 100 includes a patient information DB 110. In response to access from an external device, the patient data server 100 performs processes such as reading and writing of patient information in the patient information DB 110. Details of the patient information will be described later.

[0012] The test data server 200 includes a test result DB 210. The test data server 200 performs processes such as reading and writing test results (test data) in the test result DB 210 in response to access from an external device. The test data is managed by categorizing them by patient and test date. Details of the test data will be described later.

[0013] Blood test apparatus 300 is installed in a hospital and is capable of performing blood tests such as the number and morphology of red blood cells, white blood cells, and platelets, which are typical components of blood, and the amount of hemoglobin. Test data (blood test data) obtained by testing with blood test apparatus 300 is recorded in test result DB 210.

[0014] The transfusion product inventory management server 400 is a server for managing the inventory of transfusion products, such as transfusion platelet products, stored in medical institutions such as hospitals. The transfusion product inventory management server 400 can process transactions for receiving transfusion products to storage locations, shipping transactions to operating rooms and other locations within hospitals, and ordering transactions for transfusion products from blood centers.

[0015] The information processing device 50 can predict the amount of platelet reduction during or after surgery based on the patient information of the patient and test data from tests performed on the patient before surgery. The information processing device 50 includes a control unit 51 that controls the entire device, a communication unit 52, a memory 53, a display unit 54, an operation unit 55, a storage unit 56, a template unit 59, and a recording medium reading unit 60. The information processing device 50 can be configured as a server, a computer, or the like.

[0016] The control unit 51 may be configured by incorporating a required number of central processing units (CPUs), micro-processing units (MPUs), graphics processing units (GPUs), etc. The control unit 51 may also be configured by combining digital signal processors (DSPs), field-programmable gate arrays (FPGAs), etc.

[0017] The communication unit 52 includes a communication module and has the function of communicating with the patient data server 100, the test data server 200, and the transfusion product inventory management server 400 via the communication network 1. The communication unit 52 can acquire patient information of a patient from the patient data server 100. The communication unit 52 can also acquire pre-operative test data of the patient from the test data server 200.

[0018] The display unit 54 can be configured with a liquid crystal panel, an organic EL (Electro Luminescence) display, etc. Instead of the display unit 54, an external display device may be provided.

[0019] The operation unit 55 is configured with a touch panel or the like, and can be used to operate icons displayed on the display unit 54, move and operate a cursor, input characters, etc. The operation unit 55 may also be a mouse or a keyboard.

[0020] The storage unit 56 can be configured with a semiconductor memory, a hard disk, or the like, and stores a computer program 57 (program product), a learning model 58, and required information.

[0021] The computer program 57 can be stored in the storage unit 56 by reading the computer program 57 recorded on a storage medium (for example, an optically readable disk storage medium such as a CD-ROM) M using the storage medium reading unit 60. The computer program 57 may also be downloaded from an external device via the communication unit 52 and stored in the storage unit 56.

[0022] The memory 53 can be configured with a semiconductor memory such as a static random access memory (SRAM), a dynamic random access memory (DRAM), or a flash memory. A computer program 57 can be loaded into the memory 53, and the control unit 51 can execute the computer program 57. The control unit 51 can execute processing defined by the computer program 57. In other words, processing by the control unit 51 is also processing by the computer program 57.

[0023] The template unit 59 includes multiple standard phrases for generating detailed symptom descriptions to supplement the medical treatment details listed on medical receipts (medical fee statements). The detailed symptom descriptions are text linked to medical receipts using the patient record number, patient name, insurer number, etc., and are documents that explain the progression of symptoms, the reason for testing, specific medical treatment details, and their necessity based on treatment data and test data. The template unit 59 collects standard phrases categorized by, for example, the patient's illness, symptoms, or surgical details (surgery name), and stores them in a database that can identify appropriate standard phrases depending on the patient's illness, symptoms, or surgical details. Note that, instead of the template unit 59, a large language model (LLM), described below, may be used to generate the detailed symptom descriptions.

[0024] FIG. 2 is a diagram showing an example of patient information. As shown in FIG. 2, the patient information includes the patient ID, name, date of birth, age, sex, BMI (Body Mass Index), height, weight, blood type, medication history, surgery history, medical history, blood transfusion history, pregnancy history, liver transplant history, etc. The medication history includes whether or not the patient has taken antiplatelet drugs or anticoagulants, and the duration of their use. Antiplatelet drugs are used to prevent angina pectoris, cerebral infarction, etc., and prevent blood clots from forming in the arteries. Anticoagulants are used to prevent blood clots from forming in the heart or veins. It is expected that taking antiplatelet drugs or anticoagulants will suppress the function of platelets.

[0025] As described above, the patient information includes the patient's age, sex, and body mass index (BMI). The patient information also includes whether the patient is taking an antiplatelet drug or an anticoagulant. Note that the patient information is not limited to the example shown in FIG. 2.

[0026] 3 is a diagram showing an example of test data. As shown in FIG. 3, the test name of the blood test includes, for example, red blood cell distribution width (RDW), platelet distribution width (PDW), mean platelet volume (MPV), red blood cell count (RBC) ( / μL), white blood cell count (WBC) ( / μL), hemoglobin concentration (HGB), platelet count (PLT) ( / μL), platelet concentration (Pct), mean corpuscular volume (MCV), etc., and preferably includes blood cell morphology tests such as red blood cell distribution width (RDW), platelet distribution width (PDW), and mean platelet volume (MPV). The test data may also include test data obtained by a device other than blood test device 300 (e.g., a biochemical analyzer), such as protein amount.

[0027] The red blood cell distribution width (RDW) is the variation in the size of red blood cells and is used as a quantitative indicator of red blood cell size anomaly and an indicator of microcytic anemia. An increased red blood cell distribution width (RDW) may indicate decreased red blood cell production or iron deficiency.

[0028] Platelet distribution width (PDW) is the variation in platelet size and reflects the distribution width of platelet volume. The platelet distribution width (PDW) value increases when there is a wide distribution from small to large platelets. A large platelet distribution width (PDW) value suggests an abnormality in platelet production and a high risk of platelet reduction.

[0029] The mean platelet volume (MPV) is the average platelet volume and represents the size of platelets. It is known that the value of mean platelet volume (MPV) increases when platelet production is increased.

[0030] The red blood cell count (RBC) is the number of red blood cells per unit volume of blood being analyzed. The white blood cell count (WBC) is the number of white blood cells per unit volume of blood being analyzed. The hemoglobin concentration (HGB) is the concentration of hemoglobin per unit volume of blood being analyzed. Hemoglobin is the main component of red blood cells and is the protein responsible for transporting oxygen. The platelet count (PLT) is the number of platelets per unit volume of blood being analyzed. Platelets are the blood cell component central to the hemostasis mechanism. The platelet concentration (Pct) is the concentration of platelets in the blood expressed as a percentage. The mean corpuscular volume (MCV) is the average volume of red blood cells.

[0031] As described above, the test data includes blood test data, preferably a blood cell morphology test. More preferably, the test data includes at least one of a platelet distribution width (PDW) and a mean platelet volume (MPV). Even more preferably, the test data includes both PDW and MPV. By combining PDW and MPV, the amount of platelet reduction can be predicted with higher accuracy. The test data also includes a red blood cell distribution width (RDW). Note that the test data is not limited to the example shown in FIG. 3.

[0032] FIG. 4 is a diagram showing an example of prediction of the amount of platelet reduction during surgery using the learning model 58. When the patient's preoperative test data and patient information are input, the learning model 58 can predict the amount of platelet reduction during surgery for that patient and estimate (predict) risk factors that affect the platelet reduction. The test data includes RDW, PDW, MPV, etc. Note that RDW is not required test data, but may be included in the test data. The patient information includes information such as age, sex, BMI, and whether or not the patient is taking oral antiplatelet drugs or anticoagulants. The amount of platelet reduction can be expressed, for example, as the number of platelets per microliter (μL).

[0033] Risk factors are factors that affect platelet reduction during surgery and can be expressed as positive or negative values. The number N of risk factors output by the learning model 58 can be set appropriately. For example, risk factors whose values ​​are equal to or greater than a predetermined value may be output, or the top risk factors from among multiple risk factors may be output. Risk factors are extracted, for example, from patient information and test data.

[0034] In this embodiment, the surgery may be performed using, for example, a heart-lung machine. The heart-lung machine takes in blood (venous blood) that has been returned to the heart from the patient's entire body using a roller pump through a tube from the vena cava. The taken-in blood is oxygenated in an artificial lung, and the blood (arterial blood) that has had impurities removed by filtration is then sent to the entire body through a tube from the aorta. It is thought that the decrease in platelets during or after surgery is due to the flow of blood through the tubes and pump as the blood passes through the heart-lung machine, as well as physical and chemical stress caused by coagulation substances.

[0035] In this embodiment, "post-surgery" includes immediately after surgery and a certain period after surgery.

[0036] The learning model 58 may predict the platelet count during or after surgery instead of predicting the amount of platelet reduction during or after surgery. The learning model 58 may also predict the risk of platelet reduction depending on the amount of platelet reduction during or after surgery. In this case, for example, a large amount of platelet reduction may be deemed a high risk, a small amount of platelet reduction may be deemed a low risk, and a medium amount of platelet reduction may be deemed a medium risk.

[0037] As described above, the learning model 58 can predict (output) the amount of platelet reduction during or after surgery. Furthermore, by predicting the amount of platelet reduction during or after surgery, the learning model 58 can predict the patient's risk of platelet reduction during or after surgery. Furthermore, the learning model 58 can predict the amount of platelet reduction during or after surgery, and output risk factors that explain why such a platelet reduction was predicted. This allows medical professionals, such as doctors, to understand the basis for the amount of platelet reduction during or after surgery predicted by the information processing device 50.

[0038] The learning model 58 can be configured using models and methods such as SHAP (Shapley Additive Explanations), LIME (Local Interpretable Model-Agnostic Explanations), neural network interpreter, AM (Activation Maximization), decision tree, random forest, etc.

[0039] As described above, the control unit 51 acquires the patient's pre-operative test data, and when the pre-operative test data is input, the control unit 51 inputs the acquired test data into the learning model 58 that predicts the amount of platelet reduction during or after surgery, thereby predicting the amount of platelet reduction in the patient.

[0040] In addition, when the control unit 51 acquires the patient information of the patient and inputs the patient information and test data, it can input the acquired patient information and test data into a learning model 58 that predicts the amount of platelet reduction during or after surgery, and predict the amount of platelet reduction during or after surgery for the patient.

[0041] In addition, the control unit 51 acquires the patient's pre-operative test data, and when the pre-operative test data is input, inputs the acquired test data into a learning model 58 that predicts the amount of platelet reduction during or after surgery and outputs risk factors that affect platelet reduction, thereby predicting the amount of platelet reduction during or after surgery for the patient and estimating risk factors that will affect the predicted platelet reduction.

[0042] In addition, the control unit 51 acquires the patient information of the patient, and when the patient information and test data are input, the acquired patient information and test data are input into a learning model 58 which predicts the amount of platelet reduction during or after surgery and outputs risk factors that affect platelet reduction, thereby predicting the amount of platelet reduction during or after surgery for the patient and estimating risk factors that affect the predicted platelet reduction.

[0043] Next, a method for generating the learning model 58 (learning method) will be described.

[0044] Fig. 5 is a diagram showing an example of training data used to generate the learning model 58. In Fig. 5, the training data includes input data for learning and teacher data for learning. To prepare the training data, data is collected in the following procedure.

[0045] (1) First, preoperative data for multiple patients is collected. The preoperative data includes preoperative examination data and patient information. The examination data includes PDW, MPV, RDW, etc. The patient information includes age, sex, BMI, and whether or not the patient is taking oral antiplatelet or anticoagulant drugs.

[0046] (2) Next, when the multiple patients actually undergo surgery, intraoperative or postoperative data is collected. The intraoperative or postoperative data includes, for example, data on the amount of platelet reduction. Note that the intraoperative or postoperative data is not limited to the example of FIG. 5 and may include data related to platelet reduction. For example, the amount of platelet reduction may be a value obtained by subtracting the postoperative platelet count from the preoperative platelet count, a value calculated from the platelet count monitored during surgery, or a combination of these values. Furthermore, it may be a value calculated taking into account the amount of platelets transfused during surgery.

[0047] (3) The collected preoperative data and intraoperative or postoperative data are applied to a logistic regression analysis equation that shows the relationship between preoperative data and intraoperative or postoperative data to extract factors that affect thrombocytopenia as risk factors. Risk factors can be extracted from preoperative data. For example, if the risk factors are x1, x2, x3, ..., and the amount of thrombocytopenia is R, and the logistic regression analysis equation is "x1·w1+x2·w2+x3·w3+...=R," the influence of each risk factor x1, x2, x3, ... on the amount of thrombocytopenia R can be expressed by the magnitude of the coefficients w1, w2, w3, .... For example, the risk factor with the largest coefficient can be considered to have the greatest impact on thrombocytopenia.

[0048] In the logistic regression analysis, the preoperative data corresponds to the explanatory variable, and the amount of platelet reduction corresponds to the objective variable. Note that when the learning model 58 predicts the risk of platelet reduction instead of the amount of platelet reduction, the formula for the logistic regression analysis can be preset by treating each of the collected intraoperative data as an event, classifying the amount of platelet reduction as a binary value of "1" (if exceeded) or "0" (if not exceeded) depending on whether it exceeds a predetermined range, and calculating the probability of the event "1" to determine the risk of platelet reduction.

[0049] (4) The preoperative data of each patient is associated with the amount of platelet reduction and risk factors of that patient, thereby obtaining preoperative data as input data for learning, and the amount of platelet reduction and risk factors as training data for learning, for each of multiple patients.

[0050] The control unit 51 acquires training data including the patient's pre-operative test data and the amount of platelet reduction during or after the surgery, and based on the acquired training data, can generate a learning model 58 that predicts the amount of platelet reduction during or after the surgery when the pre-operative test data is input.

[0051] In addition, the control unit 51 acquires training data including the patient's pre-operative test data, the patient's patient information, and the amount of platelet reduction during the patient's surgery, and based on the acquired training data, when the pre-operative test data and patient information are input, it can generate a learning model 58 that predicts the amount of platelet reduction during or after the surgery.

[0052] The control unit 51 acquires training data including the patient's pre-operative test data and the amount of platelet reduction during or after the surgery, and based on the acquired training data, when the pre-operative test data is input, can generate a learning model 58 that predicts the amount of platelet reduction during or after the surgery and outputs risk factors that affect the platelet reduction.

[0053] In addition, the control unit 51 acquires training data including the patient's pre-operative test data, the patient's patient information, and the amount of platelet reduction during or after the surgery, and based on the acquired training data, when the pre-operative test data and patient information are input, it can generate a learning model 58 that predicts the amount of platelet reduction during or after the surgery and outputs risk factors that affect the platelet reduction.

[0054] The risk factors contained in the training data can be calculated from a logistic regression analysis formula in which the preoperative data is used as an explanatory variable and the amount of platelet reduction is used as a response variable.

[0055] As described above, the learning model 58 may be generated (learned) by the information processing device 50, or may be generated by another external learning device (not shown), and the generated learning model 58 may be acquired by the information processing device 50.

[0056] Next, a method for managing the amount of platelet transfusion during surgery based on the amount of platelet decrease (or the risk of platelet decrease) predicted by the information processing device 50 will be described.

[0057] Fig. 6 is a diagram showing an example of data calculated for use in platelet transfusion management. As shown in Fig. 6, the calculated data includes, for example, "predicted postoperative platelet count," "predicted post-transfusion platelet increase," and "predicted post-transfusion platelet count."

[0058] The "predicted postoperative platelet count ( / μL)" is the predicted postoperative platelet count if no platelet transfusion is administered, and can be calculated using the formula {pre-transfusion platelet count - platelet decrease amount}. The "pre-transfusion platelet count" can be obtained in advance by conducting a blood test on the patient before transfusion. The "platelet decrease amount" is the amount of platelet decrease predicted by the information processing device 50.

[0059] "Predicted post-transfusion platelet increase ( / μL)" is the predicted platelet increase immediately after platelet transfusion, and is expressed as {(total number of transfused platelets) / (circulating blood volume (mL) × 10 3 Circulating blood volume can be calculated using the formula: )] × 2 / 3}. The circulating blood volume can be set at 70 mL / kg, and varies depending on the patient's weight. For example, 10 units of platelet concentrate (approximately 200 mL, containing 2.0 × 10 platelets) 11When 10 units or more of platelets are transfused into a patient with a circulating blood volume of 4,900 mL (body weight 70 kg), the platelet count is expected to increase by 27,200 / μL or more immediately (for example, within 10 minutes to 1 hour after platelet transfusion) compared to the pre-transfusion level. In practice, a single platelet dose of 10 units is typically administered. Children weighing 25 kg or less are transfused with 10 units over a period of 3 to 4 hours. Furthermore, as can be seen from the formula in Figure 6, the greater the patient's body weight, the smaller the predicted platelet increase at the time of transfusion.

[0060] "Predicted post-transfusion platelet count ( / μL)" is the predicted post-transfusion platelet count, and can be calculated using the formula {predicted post-operative platelet count + predicted post-transfusion platelet increase}.

[0061] The platelet transfusion volume is managed by determining the platelet transfusion volume at the time of surgery so that the predicted post-transfusion platelet count ( / μL) is within the required range, and then determining whether the number of stored platelet products is excessive or insufficient based on the determined transfusion volume and the number of platelet products available for surgery. If there is a shortage, additional platelet products are ordered. In this embodiment, "ordering" refers to an order from a medical institution such as a hospital where surgery is performed to a blood center.

[0062] Perioperative management during surgery using a cardiopulmonary bypass can be performed, for example, as follows: (1) If the platelet count ( / μL) remains within the required range (e.g., 30,000 / μL) throughout the surgery and postoperative period, platelet transfusion is deemed appropriate. (2) In the case of complex cardiovascular surgery requiring the use of a cardiopulmonary bypass for a long period (e.g., 3 hours or more), surgery requiring extensive adhesion removal due to reoperation, or surgery involving chronic kidney or liver disease in which significant bleeding is expected, platelet transfusions should be administered to maintain the platelet count ( / μL) within the required range (e.g., 50,000 / μL to 100,000 / μL).

[0063] The "recommended number of platelet preparations to be used in surgery" can be identified based on the calculation formula shown in Fig. 6. That is, the control unit 51 calculates the predicted postoperative platelet count based on the amount of platelet decrease predicted by the learning model 58, calculates the predicted post-transfusion platelet count based on the calculated predicted postoperative platelet count and the predicted post-transfusion platelet increase, and calculates the recommended number of platelet preparations so that the calculated predicted post-transfusion platelet count falls within, for example, the required range described above.

[0064] Furthermore, when determining whether the predicted post-transfusion platelet count is within a required range, the control unit 51 can calculate the number of platelet preparations to be used for transfusion based on the total number of transfused platelets when calculating the predicted post-transfusion platelet increase. For example, when the total number of transfused platelets is 2.0×10 11 In this case, the number of platelet preparations is 10 units, and the total number of transfused platelets is 1.0 × 10 11 In this case, the number of platelet preparations is 5 units.

[0065] Furthermore, the control unit 51 can acquire weight data of the patient and calculate the predicted post-transfusion platelet increase based on the acquired weight data.

[0066] 7 and 8 are diagrams showing an example of a platelet transfusion amount management process performed by the information processing device 50. The control unit 51 acquires pre-operative examination data of a patient from the examination data server 200 (S11), and acquires patient information of the patient from the patient data server 100 (S12). The control unit 51 inputs the acquired examination data and patient information into the learning model 58 (S13).

[0067] The control unit 51 acquires the amount of platelet reduction during or after surgery output by the learning model 58 (S14), and calculates the predicted postoperative platelet count (S15). The control unit 51 determines whether the risk of platelet reduction is high (S16). In this case, if the amount of platelet reduction is greater than a predetermined amount, the control unit 51 can determine that the risk of platelet reduction is high.

[0068] If the risk of platelet reduction is high (YES in S16), the control unit 51 acquires the number of reserved platelet preparations (S17) and calculates the number of recommended platelet preparations (S18). As described above, the predicted postoperative platelet count is calculated based on the amount of platelet reduction predicted by the learning model 58, the predicted post-transfusion platelet count is calculated based on the calculated predicted postoperative platelet count and the predicted post-transfusion platelet increase count, and the recommended number of platelet preparations can be calculated so that the calculated predicted post-transfusion platelet count falls within a required range.

[0069] The control unit 51 determines whether the number of recommended platelet preparations is greater than the number of reserved platelet preparations (S19), and if the number of recommended platelet preparations is greater than the number of reserved platelet preparations (YES in S19), outputs a warning message urging the customer to order additional platelet preparations (S20), and performs the processing of step S21 described below.If the number of recommended platelet preparations is not greater than the number of reserved platelet preparations (NO in S19), the control unit 51 performs the processing of step S21 described below.

[0070] The control unit 51 determines whether the surgery can be postponed (S21). Whether the surgery can be postponed can be determined based on, for example, the patient's pre-surgery examination data, patient information, etc. Alternatively, the determination may be made by examining the patient's physical data, such as the patient's illness, condition, surgery content, and whether the patient's body is able to withstand the surgery. If the surgery can be postponed (YES in S21), the control unit 51 outputs a warning message recommending that the surgery be postponed (S22), and ends the process.

[0071] If the surgery cannot be postponed (NO in S21), the control unit 51 determines whether a platelet transfusion was performed at the time of the surgery (S23). If a platelet transfusion was performed (YES in S23), the control unit 51 acquires test data for symptom description and the number of platelet preparations transfused (S24), generates a symptom description (S25), and ends the process.

[0072] If platelets were not transfused during surgery (NO in S23), the control unit 51 ends the process. If the risk of platelet reduction is not high (low) (NO in S16), the control unit 51 notifies that the surgery can be performed (S26) and ends the process.

[0073] FIG. 9 shows an example of a detailed symptom description. When a medical institution such as a hospital submits a medical receipt (medical fee statement) to an examination and payment agency, the process involves the following steps: (1) input of medical information, (2) creation and printing of the medical receipt, (3) review of the medical receipt, (4) confirmation by a doctor, and (5) submission of the medical receipt to the examination and payment agency. If the examination and payment agency determines that the medical treatment details listed on the medical receipt are insufficient, medical fees are reduced. Therefore, detailed symptom information is generated to supplement the details listed on the medical receipt and submitted to the examination and payment agency along with the medical receipt. When a medical institution's staff member inputs the details of the medical treatment into a medical receipt computer (medical receipt computer), the computer automatically generates and outputs the medical receipt, but detailed symptom information previously had to be entered manually. The information processing device 50 of this embodiment can automatically generate detailed symptom information without the staff member having to enter the information manually, thereby reducing the staff member's workload.

[0074] That is, when a surgery is performed and the platelet count decreases after the surgery, the control unit 51 can obtain the number of platelets administered at the time of the surgery, obtain the details of the treatment related to the surgery, and generate a detailed symptom description based on the number of platelets administered and the details of the treatment. More specifically, the control unit 51 can identify a standard document corresponding to the patient's disease, symptoms, and surgery details (surgery name) from the template unit 59, and apply the number of platelets administered at the time of the surgery and the details of the treatment related to the surgery (for example, blood-related test data, data on drugs administered during the surgery, procedures, etc.) to the identified standard document to generate a detailed symptom description.

[0075] The control unit 51 may also generate symptom descriptions using a large-scale language model (LLM) (not shown). In this case, the large-scale language model is a machine learning model that uses natural language processing techniques to learn language patterns from existing text data and generate or understand text, such as sentences or conversations. In this embodiment, the large-scale language model first imports multiple symptom descriptions related to various surgeries as text data, divides the imported text data into small chunks, and performs an embedding process on each of the divided chunks to generate a vector representation (embedding) of each chunk. The embedding process includes quantifying the type of chunk and the positional relationship between the chunks. As a result, each chunk is represented as a point in a high-dimensional space, where chunks with similar meanings are located close to each other and chunks with different meanings are located far from each other. The large-scale language model can numerically capture the semantic relevance and similarity between chunks and understand the semantics, i.e., meaning, of the text data. In addition, large-scale language models are equipped with an attention mechanism, and when linearly combining the numerical representations of each chunk calculated by the embedding process, the magnitude of the linear combination coefficients can be appropriately adjusted to obtain a numerical representation that expresses the complex structure of the sentence.

[0076] The large-scale language model is composed of a deep neural network, and can use, for example, models such as GPT-4, GPT-3.5, BERT, LaMDA, PaLM, LLaMA, NVIDIA, or new models to be developed and operated in the future. The information processing device 50 of this embodiment can acquire a detailed symptom description generated by the large-scale language model by outputting a required question to the large-scale language model. In this case, the control unit 512 inputs the question for generating the detailed symptom description as a prompt to the large-scale language model, along with the number of platelets administered during surgery and treatment details related to the surgery (e.g., blood-related test data, drug data and treatments administered during surgery, etc.).

[0077] In the example in Figure 9, the detailed symptoms were as follows: "The patient was 82 years old, and his D-dimer was elevated at 8.9 before surgery, suggesting an increased fibrinolytic system and difficulty in hemostasis." After weaning from cardiopulmonary bypass, a blood viscoelasticity test was performed. The CRT maximum amplitude (MA), which indicates overall coagulation function, was low at 36.0 (normal value above 52). The CFF, which indicates fibrinogen coagulation function, was 7.9 (normal value above 15), and the CRT-CFF, which indicates platelet coagulation function, was low at 23.9 (normal value above 35). The patient was in a state of deficiency in platelets, coagulation factors, and fibrinogen. After weaning from cardiopulmonary bypass, the peripheral blood platelet count was critically low at 46,000, as predicted before surgery. The PT-INR was 1.80 and the APTT was prolonged at 48, suggesting a deficiency in coagulation factors. The FDP was 14, and the D-dimer was also elevated at 8.8, indicating an increased fibrinolytic system. Therefore, hemostatic control was achieved by transfusing 188 units of MAP, 40 units of platelets, 20 units of FFP, and 12 units of cryoprecipitate. The "Guidelines for the Appropriate Use of Blood Products in Cases of Massive Bleeding" recommend a 1:1:1 ratio of FFP:platelets:red blood cells, but in cardiac surgery, a higher ratio of FFP:red blood cells than 1:1 is strongly recommended (Class IC), advising the use of larger amounts of FFP. In this case, blood viscoelasticity testing indicated that FFP was necessary, and it was essential for saving the patient's life. The detailed symptom description is merely an example and is not limited to the example in Figure 9.

[0078] 10 is a diagram showing an example of a process for generating a learning model 58 by the information processing device 50. The control unit 51 collects pre-operative data (examination data, patient information) of multiple patients (S31), and collects intra-operative or post-operative data (amount of platelet reduction) of the multiple patients (S32). The control unit 51 performs logistic regression analysis on the collected post-operative data, intra-operative or post-operative data, and estimates (predicts) risk factors for platelet reduction (S33).

[0079] The control unit 51 acquires the collected preoperative data, intraoperative or postoperative data, and estimated risk factors as training data (S34). Based on the acquired training data, the control unit 51 generates a learning model 58 that predicts the amount of platelet reduction during or after surgery and risk factors for platelet reduction when preoperative data is input (S35), and then ends the processing. The generated learning model 58 can be stored in the memory unit 56.

[0080] Next, a display screen displayed on the display unit 54 by the information processing device 50 will be described.

[0081] FIG. 11 is a diagram illustrating an example of a display screen displayed by the information processing device 50. As shown in FIG. 11, the display screen displays the patient ID, name, scheduled surgery date, surgery name, and attending physician. The display screen also displays the predicted postoperative platelet count (50,000 / μL in the example of FIG. 11), the risk of platelet reduction ("severe reduction," i.e., high risk, in the example of FIG. 11), the number of recommended platelet preparations (25 units in the example of FIG. 11), and the number of reserved platelet preparations (20 units in the example of FIG. 11). The display screen also displays a warning message. In the example of FIG. 11, a message such as "The number of recommended platelet preparations has exceeded the number of reserved platelet preparations. Please reserve additional platelet preparations." The display screen also displays risk factors that affect platelet reduction. In the example of FIG. 11, PDW, PLT, and liver transplant history are displayed as risk factors, along with the numerical values ​​for PDW and PLT and the presence of a liver transplant history (+). The display screen also displays an "Inventory Management App" icon for accepting additional platelet preparation reservations. By operating the "Inventory Management App" icon, the user can transition to the platelet product ordering screen. On the transitioned ordering screen, a medical professional such as a doctor can send an order transaction for the required number of platelet products to the transfusion product inventory management server 400.

[0082] As described above, when the predicted platelet reduction rate exceeds a predetermined amount, the control unit 51 can output the recommended number of platelet preparations to be used in surgery. When the platelet reduction rate exceeds the predetermined amount, for example, the risk of platelet reduction is high.

[0083] Furthermore, the control unit 51 can output the predicted postoperative platelet count, the number of recommended platelet preparations to be used in surgery, and the number of reserved platelet preparations to be used in surgery.

[0084] The control unit 51 can output a warning message when the number of recommended platelet preparations exceeds the number of reserved platelet preparations. Also, the control unit 51 can accept transition to a platelet preparation ordering screen when the number of recommended platelet preparations exceeds the number of reserved platelet preparations.

[0085] Furthermore, the control unit 51 can output risk factors that affect the decrease in platelets.

[0086] By viewing the display screen shown in Figure 11, doctors and other medical professionals can understand the degree of risk of post-operative platelet reduction for a patient before surgery. Following the warning message, they can reserve and order the appropriate amount of platelet preparations. This allows for proper inventory management of platelet preparations, preventing the use of excessive amounts of blood preparations and the disposal of unused blood preparations.

[0087] (Appendix 1) The information processing device includes a control unit that acquires pre-operative test data for a patient undergoing surgery using an artificial heart-lung machine, and when the pre-operative test data is input, inputs the acquired test data into a learning model that predicts the amount of platelet reduction during or after surgery, thereby predicting the amount of platelet reduction for the patient.

[0088] (Supplementary Note 2) In the information processing device according to Supplementary Note 1, the test data includes blood cell morphology test data.

[0089] (Supplementary Note 3) In the information processing device according to Supplementary Note 1 or Supplementary Note 2, the test data includes at least one of a platelet distribution width (PDW) and a mean platelet volume (MPV).

[0090] (Supplementary Note 4) In the information processing device according to any one of Supplementary Note 1 to Supplementary Note 3, the test data further includes a red blood cell distribution width (RDW).

[0091] (Appendix 5) In the information processing device, in any one of Appendices 1 to 4, the control unit acquires related information about the patient, and when further related information is input, inputs the acquired test data into a learning model that predicts the amount of platelet reduction during or after surgery, thereby predicting the amount of platelet reduction in the patient.

[0092] (Appendix 6) In the information processing device of Appendix 5, the related information includes one or more pieces of information selected from the group consisting of the patient's height, weight, age, sex, and body mass index (BMI), and whether or not the patient is taking oral antiplatelet drugs or anticoagulants.

[0093] (Supplementary Note 7) In the information processing device according to any one of Supplementary Note 1 to Supplementary Note 6, the control unit outputs the number of recommended platelet preparations to be used in the surgery when the predicted amount of platelet reduction exceeds a predetermined amount.

[0094] (Appendix 8) In the information processing device according to any one of Appendices 1 to 7, the control unit acquires a predicted post-transfusion platelet increase count, calculates a predicted post-operative platelet count based on the predicted platelet decrease amount, calculates a predicted post-transfusion platelet count based on the calculated predicted post-operative platelet count and the acquired predicted post-transfusion platelet increase count, and calculates a recommended number of platelet preparations to be used in the surgery so that the calculated predicted post-transfusion platelet count falls within a required range.

[0095] (Appendix 9) In the information processing device according to any one of Appendices 1 to 8, the control unit acquires the total number of transfused platelets contained in the platelet preparation used in the surgery, acquires weight data of the patient, and calculates the predicted post-transfusion platelet increase based on the acquired weight data and the total number of transfused platelets.

[0096] (Appendix 10) In the information processing device according to any one of Appendices 1 to 9, the control unit outputs a predicted postoperative platelet count, a recommended number of platelet preparations to be used in the surgery, and a reserved number of platelet preparations to be used in the surgery.

[0097] (Appendix 11) In the information processing device, in any one of Appendices 1 to 10, the control unit outputs a warning message and accepts transition to a platelet preparation ordering screen if the number of recommended platelet preparations exceeds the number of reserved platelet preparations.

[0098] (Supplementary Note 12) In the information processing device according to any one of Supplementary Note 1 to Supplementary Note 11, the control unit outputs risk factors that affect thrombocytopenia.

[0099] (Appendix 13) In the information processing device, in any one of Appendices 1 to 12, the control unit acquires the number of recommended platelet preparations to be used in the surgeries for multiple patients and the timing of the surgeries for the multiple patients, and calculates the order amount of platelet preparations for each timing of the surgeries based on the acquired number of recommended platelet preparations and the timing of the surgeries.

[0100] (Appendix 14) In the information processing device, in any one of Appendices 1 to 13, when the postoperative platelet count decreases due to the surgery, the control unit obtains the number of platelets administered at the time of the surgery, obtains the treatment details related to the surgery, and generates a detailed symptom description based on the number of platelet administrations and the treatment details.

[0101] (Appendix 15) The computer program acquires pre-operative test data for a patient undergoing surgery using an artificial heart-lung machine, and when the pre-operative test data is input, inputs the acquired test data into a learning model that predicts the amount of platelet reduction during or after surgery, thereby causing the computer to execute a process of predicting the amount of platelet reduction in the patient.

[0102] (Appendix 16) The information processing method obtains pre-operative test data for a patient undergoing surgery using an artificial heart-lung machine, and when the pre-operative test data is input, inputs the obtained test data into a learning model that predicts the amount of platelet reduction during or after surgery, thereby predicting the amount of platelet reduction in the patient.

[0103] (Appendix 17) The learning model generation method acquires training data including pre-operative test data of a patient undergoing surgery using an artificial heart-lung machine and the amount of platelet reduction during or after the surgery of the patient, and generates a learning model that predicts the amount of platelet reduction during or after the surgery based on the acquired training data when the pre-operative test data is input.

[0104] The matters described in each embodiment can be combined with each other. In addition, the independent claims and dependent claims described in the claims can be combined with each other in any combination, regardless of the reference format. Furthermore, although the claims use a format in which a claim references two or more other claims (multiple claim format), this is not limited to this format. A multiple claim (multi-multi claim) that references at least one other multiple claim may also be used. [Explanation of symbols]

[0105] 1. Communication Network 50 Information processing equipment 51 Control section 52 Communications Department 53 Memory 54 Display section 55 Operation section 56 Memory section 57 Computer Programs 58 Learning Model 59 Template Section 60 Recording medium reading unit 100 Patient Data Server 110 Patient information DB 200 Inspection Data Server 210 Test Results DB 300 Blood testing equipment 400 Transfusion product inventory management server

Claims

1. A control unit is provided, The control unit obtaining preoperative examination data for a patient's cardiopulmonary bypass surgery; When pre-operative test data is input, the acquired test data is input into a learning model that predicts the amount of platelet reduction during or after surgery, and the amount of platelet reduction of the patient is predicted. Information processing device.

2. The test data includes blood cell morphology test data. The information processing device according to claim 1 .

3. The test data includes at least one of a platelet distribution width (PDW) and a mean platelet volume (MPV). The information processing device according to claim 1 .

4. The test data further includes red blood cell distribution width (RDW). The information processing device according to claim 3 .

5. The control unit obtaining relevant information about the patient; inputting the acquired test data into a learning model that predicts the amount of platelet reduction during or after surgery when related information is further input, and predicting the amount of platelet reduction in the patient; The information processing device according to claim 1 .

6. The related information includes one or more pieces of information selected from the group consisting of the patient's height, weight, age, sex, and body mass index (BMI), and whether or not the patient is taking an antiplatelet drug or an anticoagulant drug. The information processing device according to claim 5 .

7. The control unit If the predicted platelet reduction amount exceeds a predetermined amount, the number of platelet preparations recommended for use in the surgery is output. The information processing device according to claim 1 .

8. The control unit Obtain the predicted post-transfusion platelet increase number, Calculating the predicted postoperative platelet count based on the predicted platelet reduction amount; Calculating the predicted post-transfusion platelet count based on the calculated predicted post-operative platelet count and the obtained predicted post-transfusion platelet increase count; Calculating the number of platelet preparations recommended to be used in the surgery so that the calculated predicted post-transfusion platelet count falls within a required range. The information processing device according to claim 1 .

9. The control unit Obtain the total number of transfused platelets contained in the platelet preparation used in the surgery; obtaining weight data of the patient; Calculating the predicted post-transfusion platelet increase based on the acquired weight data and the total number of transfused platelets. The information processing device according to claim 8 .

10. The control unit outputting a predicted postoperative platelet count, a recommended number of platelet preparations to be used in the surgery, and a reserved number of platelet preparations to be used in the surgery; The information processing device according to claim 1 .

11. The control unit If the number of recommended platelet preparations exceeds the number of reserved platelet preparations, a warning message is output and the user is allowed to proceed to the platelet preparation ordering screen. The information processing device according to claim 1 .

12. The control unit Output risk factors that affect thrombocytopenia, The information processing device according to claim 1 .

13. obtaining preoperative examination data for a patient's cardiopulmonary bypass surgery; When pre-operative test data is input, the acquired test data is input into a learning model that predicts the amount of platelet reduction during or after surgery, and the amount of platelet reduction of the patient is predicted. A computer program that causes a computer to perform a process.

14. obtaining preoperative examination data for a patient's cardiopulmonary bypass surgery; When pre-operative test data is input, the acquired test data is input into a learning model that predicts the amount of platelet reduction during or after surgery, and the amount of platelet reduction of the patient is predicted. Information processing methods.

15. obtaining training data including preoperative examination data of a patient undergoing a heart-lung machine operation and a platelet reduction level during or after the operation of the patient; Based on the acquired training data, a learning model is generated that predicts the amount of platelet reduction during or after surgery when preoperative test data is input. Learning model generation method.

Images