Blood purification system, blood purification device, model generation method, model evaluation method, and program

Abstract

To support the use of appropriate control models for blood purification therapy. [Solution] The blood purification system 100 includes a display unit 14 that displays a model creation screen on which treatment-related data used to generate a control model for controlling the blood purification device 10 can be set, and a model generation unit 56 that generates a control model by machine learning using the treatment-related data set on the model creation screen as input.

Description

Technical Field

[0001] The present invention relates to a blood purification system, a blood purification apparatus, a model generation method and a program. Background Art

[0002] In blood purification therapy, a patient's blood is extracorporeally circulated through a blood circuit, and waste products and excess water in the blood are removed using a blood purifier. During treatment, blood pressure values and the like indicating the patient's condition are measured periodically, and blood flow rate, dialysate flow rate and the like are controlled according to the patient's condition. For example, a technique is known in which blood information is input, control parameters are output from a trained model, and the operation of a blood purification system is controlled based on the control parameters (see, for example, Patent Document 1). Prior Art Documents Patent Documents

[0003] Patent Document 1 Japanese Unexamined Patent Application Publication No. 2024-12445 Summary of the Invention Problem to be Solved by the Invention

[0004] In blood purification therapy, treatment contents and treatment policies may differ for each patient or each facility, and it is difficult to prepare a universal control model applicable to all treatments.

[0005] The present invention has been made in view of these problems, and one exemplary object thereof is to provide a technique that supports the use of a control model suitable for blood purification therapy. Means for Solving the Problem

[0006] A blood purification system according to one aspect of the present invention is a blood purification system comprising a blood purification device, a display unit that displays a model creation screen on which treatment-related data used for generating a control model for controlling the blood purification device can be set, and a model generation unit that generates a control model by machine learning using the treatment-related data set on the model creation screen as input.

[0007] Another aspect of the present invention is a blood purification device. This device comprises a display unit and a communication unit, the display unit displaying a model creation screen on which treatment-related data used to generate a control model for controlling the blood purification device can be set, and the communication unit transmits the settings on the model creation screen.

[0008] Another aspect of the present invention is a model generation method. This method comprises the steps of: displaying a model creation screen on the display unit of a blood purification device, on which treatment-related data to be used to generate a control model for controlling a blood purification device can be set; and generating a control model by machine learning using the treatment-related data set on the model creation screen as input.

[0009] Another aspect of the present invention is a program. This program enables a computer to perform the following functions: display a model creation screen on the display unit of the blood purification device, on which treatment-related data used to generate a control model for controlling the blood purification device can be set; and transmit the settings on the model creation screen. [Effects of the Invention]

[0010] According to the present invention, it is possible to support the use of appropriate control models for blood purification therapy. [Brief explanation of the drawing]

[0011] [Figure 1] This is a perspective view showing the external appearance of a blood purification device according to an embodiment. [Figure 2] This diagram schematically shows the configuration of a blood purification device according to an embodiment. [Figure 3]This diagram schematically shows the functional configuration of the blood purification system according to the embodiment. [Figure 4] This figure shows an example of the model creation screen. [Figure 5] This figure shows an example of the model evaluation screen display. [Figure 6] This figure shows an example of the model selection screen. [Modes for carrying out the invention]

[0012] Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the drawings. In the description, the same elements will be denoted by the same reference numerals, and redundant explanations will be omitted as appropriate. To aid in understanding the description, the dimensional ratios of each component in each drawing do not necessarily correspond to the actual dimensional ratios.

[0013] Figure 1 is a perspective view showing the external appearance of a blood purification device 10 according to an embodiment. The blood purification device 10 is used in blood purification therapies such as hemodialysis and is also called a dialysis machine or dialysis monitoring device. Figure 1 shows the device without consumables such as blood circuits and blood purifiers attached. The blood purification device 10 comprises a main body 12, a display unit 14, and an infusion pole 16.

[0014] The display unit 14 is located on the upper part of the main unit 12. The display unit 14 displays a screen showing the operating status of the blood purification device 10, etc. The display unit 14 is, for example, composed of a touch panel. The user can input user operations by touching the surface of the display unit 14.

[0015] The IV pole 16 is located on the side of the main body 12. The IV pole 16 is equipped with a blood purifier holder (not shown) for attaching a blood purifier, also known as a dialyzer. An infusion hook is provided at the top of the IV pole 16 for suspending an infusion bag, such as a saline solution bag.

[0016] A blood circuit is attached to the front surface of the main body 12. A blood pump 18 is provided on the front surface of the main body 12. The blood pump 18 is, for example, a peristaltic pump, and accommodates a pump tube provided in the blood circuit. The blood pump 18 generates a flow of blood or replacement fluid in the blood circuit by squeezing the accommodated pump tube.

[0017] Figure 2 is a diagram schematically showing the configuration of the blood purification apparatus 10 according to the embodiment. The blood purification apparatus 10 uses a blood purifier 20 and a blood circuit 22. The blood purifier 20 and the blood circuit 22 are consumables and are disposed of after each treatment. The blood purification apparatus 10 includes a dialysate supply unit 24, a replacement fluid supply unit 26, and a control unit 28.

[0018] In terms of hardware, each block shown in the block diagram of the present disclosure can be implemented by elements such as a processor such as a CPU (Central Processing Unit) of a computer, a memory such as ROM (Read Only Memory) or RAM (Random Access Memory), or other elements and mechanical devices, and in terms of software, it is implemented by a computer program or the like. Here, functional blocks implemented by cooperation of hardware and software are depicted. Those skilled in the art will understand that these functional blocks can be implemented in various forms by a combination of hardware and software.

[0019] The blood purifier 20 has a blood introduction port 20a and a blood delivery port 20b connected to the blood circuit 22, and a dialysate introduction port 20c and a dialysate delivery port 20d connected to the dialysate supply unit 24. The blood purifier 20 has hollow fibers accommodated therein. The inside of the hollow fibers serves as a blood flow path connecting between the blood introduction port 20a and the blood delivery port 20b. The outside of the hollow fibers serves as a dialysate flow path connecting between the dialysate introduction port 20c and the dialysate delivery port 20d. A large number of micropores (pores) are formed in the hollow fibers, and the hollow fibers are configured such that waste products and excess water in the blood permeate into the dialysate through the large number of micropores.

[0020] The blood circuit 22 is composed of a flexible tube that serves as a blood flow path. The blood circuit 22 includes a blood removal-side blood circuit 22a and a blood return-side blood circuit 22b. The blood removal-side blood circuit 22a is a blood removal tube that connects between a patient's blood vessel (e.g., an artery) and the blood introduction port 20a of the blood purifier 20. The blood removal-side blood circuit 22a is attached to the blood pump 18. The blood return-side blood circuit 22b is a blood return tube that connects between a patient's blood vessel (e.g., a vein) and the blood outlet port 20b of the blood purifier 20.

[0021] The dialysate supply unit 24 is configured to supply dialysate to the blood purifier 20 and discharge dialysate from the blood purifier 20. The dialysate supply unit 24 includes a supply line L1, a discharge line L2, a bypass line L3, a duplex pump 30, and a water removal pump 32.

[0022] The supply line L1 is a main pipe through which the dialysate supplied to the blood purifier 20 flows, and is connected to the dialysate introduction port 20c of the blood purifier 20. The dialysate flowing through the supply line L1 is supplied from a dialysate supply device. The discharge line L2 is a main pipe through which the dialysate discharged from the blood purifier 20 flows, and is connected to the dialysate outlet port 20d of the blood purifier 20. The duplex pump 30 is provided midway through the supply line L1 and the discharge line L2, and controls the flow rate of the dialysate flowing through the supply line L1 and the discharge line L2. The water removal pump 32 is provided in the bypass line L3 that bypasses the duplex pump 30 in the discharge line L2.

[0023] The replacement fluid supply unit 26 is configured to supply replacement fluid to the blood circuit 22. The replacement fluid supply unit 26 is connected to, for example, a branch portion of the blood removal-side blood circuit 22a. The replacement fluid supply unit 26 supplies replacement fluid supplied from, for example, a replacement fluid bag to the blood circuit 22. The replacement fluid supply unit 26 may supply the dialysate supplied from the dialysate supply unit 24 to the blood circuit 22. The replacement fluid supply unit 26 can supply replacement fluid for the purpose of emergency fluid replacement during treatment, for example.

[0024] The control unit 28 controls the overall operation of the blood purification device 10. The control unit 28 includes, for example, a processor such as a CPU (Central Processing Unit), memory such as RAM (Random Access Memory) or ROM (Read Only Memory), and storage devices such as an HDD (Hard Disk Drive) or SSD (Solid State Drive). The control unit 28 operates according to a program, for example, by the processor executing a program stored in memory. The processor may execute a program stored in any storage device other than memory, or a program obtained from any recording medium by a reading device, or a program obtained via a network.

[0025] The control unit 28 comprises a dialysis control unit 34, a measurement unit 36, a data management unit 38, a model management unit 40, and a communication unit 42.

[0026] The dialysis control unit 34 controls the operation of the blood pump 18, dialysate supply unit 24, and replacement fluid supply unit 26 according to the treatment conditions set for each patient. Treatment conditions include, but are not limited to, dialysis time, amount of fluid removed, fluid removal rate, blood flow rate, and treatment mode. Treatment modes include, but are not limited to, hemodiafiltration (HDF) mode, hemodialysis (HD) mode, and hemofiltration (HF) mode. Treatment conditions are set, for example, via the touch panel of the display unit 14 before the start of blood purification treatment. Depending on the treatment conditions, the replacement fluid bag may or may not be connected to the blood circuit 22.

[0027] The measurement unit 36 ​​measures the patient using the measurement device 44. The measurement device 44 is, for example, a cuff that is attached to the patient. The cuff is, for example, wrapped around the patient's upper arm and configured to pressurize the patient's blood vessels. The measurement unit 36 ​​measures blood pressure using, for example, the oscillometric method. The measurement unit 36 ​​sets the internal pressure of the cuff to be higher than the systolic blood pressure and detects the pressure oscillations superimposed on the internal pressure as a pulse wave signal during the process of gradually decreasing the internal pressure. Based on the detected pulse wave signal, the measurement unit 36 ​​obtains the pulse wave amplitude value with respect to the cuff pressure. Based on the detected pulse wave signal, the measurement unit 36 ​​obtains the patient's blood pressure value. The measurement unit 36 ​​may obtain the systolic blood pressure, diastolic blood pressure, and mean blood pressure as the blood pressure value. The measurement unit 36 ​​may obtain the pulse rate based on the detected pulse wave signal.

[0028] The measuring device 44 may be a weighing scale. The measuring unit 36 ​​may measure the patient's weight using a weighing scale. The patient's weight may be measured before the start and after the end of treatment. The measuring device 44 may be a wearable device worn by the patient. The measuring unit 36 ​​may acquire the patient's vital data (body temperature, pulse, blood pressure, etc.) through the wearable device.

[0029] The data management unit 38 manages treatment-related data concerning blood purification therapy. The data management unit 38 manages performance data generated for each blood purification therapy treatment given to a patient, as well as patient data related to the patient.

[0030] The performance data includes a patient ID that identifies the patient being treated and the date on which the blood purification treatment was performed. The performance data may also include data measured by the measuring device 44 before and after the blood purification treatment, and may include parameters such as the patient's pre-dialysis blood pressure, post-dialysis blood pressure, pre-dialysis body weight, and post-dialysis body weight. The performance data may also include treatment conditions for the blood purification treatment, and may include parameters such as dialysis time, amount of fluid removed, fluid removal rate, blood volume, treatment mode, and reference area (range from the lower limit to the upper limit of ΔBV). The performance data may also include time-series data acquired during the execution of the blood purification treatment, and may include time-series values ​​of parameters such as amount of fluid removed, fluid removal rate, blood volume, blood pressure (e.g., systolic blood pressure, diastolic blood pressure, and mean blood pressure), pulse rate, pulse wave signal, ΔBV, Kt / V, DDM absorbance, and extracorporeal blood flow rate (LDQb). The performance data may also include logs of alerts issued by the control unit 28 (such as a drop in blood pressure). Patient data may include patient attributes such as age, sex, weight (dry weight), dialysis history, blood test information, medical history, lifestyle, and medical condition. The data managed by the data management unit 38 is transmitted to the information processing device 50, which will be described later, and stored in the information processing device 50.

[0031] The model management unit 40 manages the control model used by the dialysis control unit 34. The control model is, for example, a trained model generated by machine learning, and has at least one input and at least one output. Any parameters included in the treatment-related data can be used as the parameters for the input and output. Time-series data included in the treatment-related data can also be used as the input.

[0032] The model management unit 40 displays a model creation screen 70 (described later in Figure 4) on the display unit 14 for generating a control model. The model management unit 40 displays a model evaluation screen 80 (described later in Figure 5) on the display unit 14 for evaluating the control model. The model management unit 40 displays a model selection screen 90 (described later in Figure 6) on the display unit 14 for selecting a control model to be used by the dialysis control unit 34. Details of these screens will be described separately.

[0033] The communication unit 42 communicates with the information processing device 50 via the network 48. The communication unit 42 transmits data managed by the data management unit 38, for example, to the information processing device 50. The communication unit 42 receives information from the information processing device 50 regarding the control model managed by the model management unit 40. Details of the information processing device 50 will be described later with reference to Figure 3.

[0034] Figure 3 is a schematic diagram showing the configuration of a blood purification system 100 according to an embodiment. The blood purification system 100 comprises a plurality of blood purification devices 10, an information processing device 50, and a user terminal 60. Each of the plurality of blood purification devices 10 is configured in the same way as the blood purification device 10 shown in Figure 2.

[0035] Network 48 connects to multiple blood purification devices 10, information processing devices 50, and user terminals 60. Network 48 is, for example, a local area network (LAN) installed within a facility where blood purification treatment is performed.

[0036] The information processing device 50 is a management device that manages the blood purification device 10, and is, for example, a server device. The information processing device 50 includes a communication unit 52, a storage unit 54, a model generation unit 56, and a model evaluation unit 58.

[0037] The communication unit 52 communicates with the blood purification device 10 via the network 48. The communication unit 52, for example, acquires treatment-related data from the blood purification device 10. The communication unit 52 transmits to the blood purification device 10 information regarding the control model generated by the model generation unit 56 and information regarding the evaluation of the control model evaluated by the model evaluation unit 58.

[0038] The memory unit 54 stores treatment-related data acquired from the blood purification device 10. The memory unit 54 stores information about the control model generated by the model generation unit 56. The memory unit 54 stores information about the evaluation of the control model evaluated by the model evaluation unit 58.

[0039] The model generation unit 56 generates a control model for use by the model management unit 40. For example, based on the creation specifications set in the model management unit 40, the model generation unit 56 performs machine learning using treatment-related data stored in the memory unit 54 and generates a trained model. The control model generated by the model generation unit 56 is stored in the memory unit 54.

[0040] The model evaluation unit 58 evaluates the control model generated by the model generation unit 56. The model evaluation unit 58 inputs treatment-related data stored in the storage unit 54 into the control model based on evaluation specifications set in the model management unit 40, for example, and calculates an evaluation value for the control model. The evaluation value calculated by the model evaluation unit 58 is stored in the storage unit 54.

[0041] The user terminal 60 is used by a user who is a medical professional. The user terminal 60 is a general-purpose computer such as a personal computer, or a portable terminal such as a smartphone or tablet computer. The user terminal 60 includes a communication unit 62 and a display unit 64.

[0042] The communication unit 62 communicates with the information processing device 50 via the network 48. The display unit 64 displays information acquired from the information processing device 50. The display unit 64 may display at least one of the following: a model creation screen 70 (described later in Figure 4), a model evaluation screen 80 (described later in Figure 5), and a model selection screen 90 (described later in Figure 6), similar to the display unit 14 of the blood purification device 10.

[0043] Next, we will explain the creation, evaluation, and use of control models by the Model Management Unit 40.

[0044] Figure 4 shows an example of the display of the model creation screen 70. The model management unit 40 displays the model creation screen 70 on the display unit 14 for creating a control model to be used by the dialysis control unit 34. The model creation screen 70 includes a basic setting unit 72, an input setting unit 74, an output setting unit 76, and a creation execution unit 78.

[0045] The basic setting unit 72 sets the machine learning specifications for creating a control model. The basic setting unit 72 includes, for example, a learning method setting unit 72a, an algorithm setting unit 72b, a target patient setting unit 72c, and a target period setting unit 72d.

[0046] The learning method setting section 72a allows you to set learning methods such as supervised learning and unsupervised learning. The algorithm setting section 72b allows you to set machine learning algorithms such as random forest, linear regression, support vector machine, gradient boosting, and neural network.

[0047] In the target patient setting unit 72c, conditions related to patients can be set as conditions for treatment-related data used for learning. Patient conditions can be set to a specific patient (e.g., patient A), specific patient attributes, or the entire facility (i.e., all patients). Multiple patient attributes may be combined and set in the target patient setting unit 72c; for example, it may be possible to set a female with arrhythmia.

[0048] In the target period setting unit 72d, conditions related to the period can be set as conditions for the treatment-related data used for learning. As conditions related to the period, it is possible to set a period starting from the present time, such as the most recent 1 month, the most recent 6 months, or the most recent 1 year, or a period from an arbitrary start date to an end date (for example, AAAA / AA / AA to BBBB / BB).

[0049] The input setting unit 74 sets the treatment-related data to be used as input for the control model. The input setting unit 74 can set, for example, multiple parameters 74a to 74e included in the treatment-related data as inputs. In the example shown in Figure 4, the first parameter 74a is set to the fluid removal rate, the second parameter 74b is set to the fluid removal amount, the third parameter 74c is set to body weight, the fourth parameter 74d is set to gender, and the fifth parameter 74e is set to blood pressure. The number of inputs that can be set in the input setting unit 74 is not limited to 5, but may be 4 or less, or 6 or more.

[0050] The output setting unit 76 sets the dialysis-related data to be used for outputting the control model. The output setting unit 76 can set, for example, one parameter 76a included in the treatment-related data as an output. In the example shown in Figure 4, blood pressure reduction is set as the output parameter 76a. Note that the number of outputs that can be set in the output setting unit 76 is not limited to one, but may be multiple.

[0051] The creation execution unit 78 is an operation button for starting the creation of a control model. After setting the necessary items in the basic setting unit 72, input setting unit 74, and output setting unit 76, pressing the creation execution unit 78 sends settings indicating the specifications for creating the control model to the information processing device 50. The information processing device 50 (i.e., the model generation unit 56) performs machine learning to create the control model according to the settings indicating the creation specifications.

[0052] Figure 5 shows an example of the display of the model evaluation screen 80. The model management unit 40 displays the model evaluation screen 80 for evaluating the control model created by machine learning on the display unit 14. The model evaluation screen 80 includes a basic setting unit 82, an input display unit 84, an output display unit 86, and an evaluation value display unit 88.

[0053] The basic setting unit 82 sets the evaluation specifications for the control model. The basic setting unit 82 includes, for example, an evaluation model setting unit 82a, an evaluation index setting unit 82b, a target patient setting unit 82c, and a target period setting unit 82d.

[0054] The evaluation model setting unit 82a selects the control model to be evaluated. The control model to be evaluated can be selected from multiple control models stored in the storage unit 54, and can be selected by keyword search or patient information search.

[0055] The evaluation metric setting unit 82b selects evaluation metrics for the selected model. Examples of evaluation metrics include, but are not limited to, accuracy, precision, recall, mean absolute error (MAE), mean absolute percent error (MAPE), weighted absolute percent error (WAPE), mean squared error (MSE), and mean squared error (RMSE). The evaluation metrics may be selected according to the algorithm of the control model.

[0056] In the target patient setting unit 82c, conditions related to patients can be set as conditions for treatment-related data used for evaluation. Patient conditions can be set as a specific patient (e.g., patient A), specific patient attributes, or the entire facility (i.e., all patients). In the target patient setting unit 72c, multiple patient attributes may be set in combination; for example, it may be possible to set a target as a male with a dialysis history of 3 years or more.

[0057] In the target period setting unit 82d, you can set period-related conditions as conditions for the treatment-related data used for evaluation. As period-related conditions, you can set periods starting from the present time, such as the most recent 1 month, the most recent 6 months, or the most recent 1 year, or time periods from an arbitrary start date to an end date (for example, AAAA / AA / AA to BBBB / BB).

[0058] The input display unit 84 displays a list of inputs used for the selected control model. The output display unit 86 displays the outputs used for the selected control model. The parameters displayed in the input display unit 84 and the output display unit 86 are the same as the parameters displayed in the input setting unit 74 and the output setting unit 76 when the control model is created.

[0059] The evaluation value display unit 88 displays the evaluation value calculated based on the evaluation specifications set in the basic setting unit 82. In the example in Figure 5, the evaluation value 88a displays that the accuracy rate of predicting the occurrence of blood pressure drop is 51%. The evaluation value 88a is calculated by the model evaluation unit 58 based on the evaluation specifications set in the basic setting unit 82.

[0060] The evaluation value 88a may vary depending on the treatment-related data input into the control model for evaluation. In other words, the evaluation value 88a may vary depending on the settings of the target patient setting unit 82c and the target period setting unit 82d. The user can, for example, determine whether the control model being evaluated is suitable for a specific patient A. For example, if the evaluation value 88a is good, it may be preferable to apply the evaluated control model to the blood purification treatment of specific patient A. On the other hand, if the evaluation value 88a is not good, it may be preferable not to apply the evaluated control model to the blood purification treatment of specific patient A. Furthermore, if the treatment plan is changed, the user can determine whether the control model being evaluated is suitable for the changed blood purification treatment by evaluating it using dialysis-related data for the period after the change. For example, the user can apply a more appropriate control model to blood purification treatment by evaluating a model they created on the model creation screen 70 on the model evaluation screen 80. The user can also adopt a more appropriate control model by evaluating a model created by someone else on the model evaluation screen 80.

[0061] Figure 6 shows an example of the display of the model selection screen 90. The model management unit 40 displays the model selection screen 90 on the display unit 14 for selecting a control model to be used by the dialysis control unit 34. The model selection screen 90 includes a list display unit 92, a detailed description unit 94, and a model search unit 96.

[0062] The list display unit 92 displays a list of control models available to the model management unit 40. The control models available to the model management unit 40 are obtained from the information processing device 50. The list display unit 92 may prioritize displaying control models suitable for patient A who is to be treated. The calculation result of the model evaluation unit 58 can be used as an evaluation value to determine whether or not a control model is suitable for patient A who is to be treated.

[0063] In the example shown in Figure 6, the ultrafiltration setting model (for patient A only) 92a, the ultrafiltration setting model (general purpose) 92b, the blood pressure drop prediction model 92c, and the alert prediction model 92d are displayed. Pressing the buttons corresponding to each model 92a to 92d allows you to individually switch between the ON state (enabled) and the OFF state (disabled) for each model 92a to 92d. In the example shown in Figure 6, the ultrafiltration setting model (for patient A only) 92a and the blood pressure drop prediction model 92c are enabled (on), while the ultrafiltration setting model (general purpose) 92b and the alert prediction model 92d are disabled (off). Pressing the details button 98, indicated by "?" to the right of each model 92a to 92d, displays detailed information about the corresponding model in the details section 94.

[0064] The detailed description section 94 displays detailed information about the control model selected in the list display section 92. The detailed description section 94 includes the creation specifications set on the model creation screen 70 when the control model is created, and the evaluation values ​​displayed on the model evaluation screen 80 when the control model is evaluated. For example, the detailed description section 94 displays the model name, target patient, target period, input parameters, output parameters, control content based on the model, evaluation values, creation date, and creator. The control content based on the model indicates the control content executed by the dialysis control unit 34 when the model is used. In the example shown in Figure 6, if the blood pressure drop prediction model predicts the occurrence of a blood pressure drop, control is executed to reduce the ultrafiltration rate below the set value in order to prevent the occurrence of a blood pressure drop. The ultrafiltration rate can be changed by changing the operation of the ultrafiltration pump 32.

[0065] The model search unit 96 includes operation buttons for displaying a search screen to narrow down the control models. In the example shown in Figure 6, the keyword search unit 96a and the patient information search unit 96b are displayed. When the keyword search unit 96a is pressed, a keyword input screen is displayed, and control models that match the entered keyword are displayed. For example, control models that match the keyword with the display content of the detailed description unit 94 (model name, input parameters, output parameters, control content, etc.) are narrowed down. The search screen may have a keyword input section for each display item of the detailed description unit 94. When the patient information search unit 96b is pressed, patient data indicating the attributes of patient A to be treated is displayed in a list. When at least one of the displayed patient data is selected, control models suitable for the selected patient's attributes (e.g., age, gender, lifestyle, medical history) are narrowed down.

[0066] The dialysis control unit 34 can control the operation of the blood purification device 10 based on the control model selected on the model selection screen 90. For example, the dialysis control unit 34 can control the operation of the blood purification device 10 based on the output of the control model by inputting treatment-related data acquired during treatment into the selected control model. When using the ultrafiltration setting model, the operation of the ultrafiltration pump 32 can be set based on the output of the ultrafiltration setting model. When using the blood pressure drop prediction model, the operation of the ultrafiltration pump 32 can be changed by predicting the timing of the occurrence of a drop in blood pressure.

[0067] According to this embodiment, blood purification therapy can be performed using an appropriate control model for each patient, each treatment, or each facility. For example, in the case of a patient with a long treatment history, since treatment-related data from the past long period is accumulated in the memory unit 54, there is a high possibility that a patient-specific control model can be created with high accuracy. On the other hand, in the case of a patient with a short treatment history, since there is little treatment-related data accumulated in the memory unit 54, if an attempt is made to create a patient-specific control model, the amount of data may be insufficient, and sufficient accuracy may not be obtained. In such cases, it may be advantageous to use a general-purpose model based on treatment-related data from multiple patients. According to this embodiment, it is possible to create an individual model specific to a particular patient, and it is also possible to create a general-purpose model that is not specific to a particular patient. Furthermore, it is possible to evaluate whether the created control model has sufficient accuracy and select a control model based on the evaluation results. This improves user convenience.

[0068] In the above-described embodiment, the information processing device 50 is configured to include a model generation unit 56 and a model evaluation unit 58. In a modified example, at least one of the model generation unit 56 and the model evaluation unit 58 may be provided in the blood purification device 10.

[0069] The present invention has been described above based on examples. Those skilled in the art will understand that the present invention is not limited to the above embodiments, that various design changes are possible, and that various modifications are possible, and that such modifications also fall within the scope of the present invention.

[0070] Several embodiments of the present invention will be described below.

[0071] A first aspect of the present invention is a blood purification system comprising a blood purification device, the system comprising: a display unit that displays a model creation screen on which treatment-related data used to generate a control model for controlling the blood purification device can be set; and a model generation unit that generates the control model by machine learning using the treatment-related data set on the model creation screen as input. According to the first aspect, a control model using treatment-related data can be easily created by operating the model creation screen. This makes it possible to support the use of a control model suitable for each patient or facility.

[0072] A second aspect of the present invention is the blood purification system according to the first aspect, wherein the model creation screen allows setting at least one of the learning method, algorithm, input data, and output data of the control model. According to the second aspect, since the specifications of the control model can be set on the model creation screen, it is possible to support the use of a control model suitable for each patient or facility.

[0073] A third aspect of the present invention is a blood purification system according to the first or second aspect, wherein the model creation screen allows setting at least one of patient-related conditions and period-related conditions as conditions for treatment-related data used in machine learning of the control model. According to the third aspect, since patient and period-related conditions can be set on the model creation screen, it is possible to support the use of a control model that is more suitable for a specific patient.

[0074] A fourth aspect of the present invention is a blood purification system according to any one of the first to third aspects, further comprising a model evaluation unit for calculating an evaluation value of the control model, wherein the display unit can set the control model to be evaluated and the conditions for the treatment-related data used for evaluation, and displays a model evaluation screen capable of displaying the evaluation value calculated by the model evaluation unit. According to the fourth aspect, since the control model can be evaluated by setting conditions on the model evaluation screen, it is possible to support the use of a control model suitable for each patient or facility.

[0075] A fifth aspect of the present invention is a blood purification system according to the fourth aspect, wherein the model evaluation screen allows setting at least one of patient-related conditions and period-related conditions as conditions for treatment-related data used for evaluation. According to the fifth aspect, since patient and period-related conditions can be set on the model evaluation screen, it is possible to support the use of a control model that is more suitable for a specific patient.

[0076] A sixth aspect of the present invention is a blood purification system according to any one of the first to fifth aspects, further comprising a storage unit for storing a plurality of control models, wherein the display unit displays a model selection screen on which at least one of the plurality of control models can be selected, and the blood purification device controls the operation of the blood purification device based on the output of at least one control model selected on the model selection screen. Because the control model to be used for control can be easily selected on the model selection screen, it is possible to support the use of a control model suitable for each patient or facility.

[0077] A seventh aspect of the present invention is a blood purification system according to the sixth aspect, wherein the display unit prioritizes displaying the control model most suitable for the patient being treated from among the plurality of control models. According to the seventh aspect, since the control model most suitable for the patient is prioritized, it is possible to support the use of a control model more suitable for a specific patient.

[0078] An eighth aspect of the present invention is a blood purification system according to the seventh aspect, wherein the preferred control model is based on evaluation values ​​obtained when evaluating the plurality of control models using treatment-related data of the patient to be treated. According to the eighth aspect, a control model suitable for the patient is preferred based on the evaluation value for each patient, thereby supporting the use of a control model more suitable for a specific patient.

[0079] A ninth aspect of the present invention is a blood purification system according to any one of the sixth to eighth aspects, wherein the plurality of control models include individual models generated by machine learning using treatment-related data of a specific patient as input, and general-purpose models generated by machine learning using treatment-related data of multiple patients as input. According to the ninth aspect, since a control model can be selected from both individual models and general-purpose models, it is possible to support the use of a control model suitable for the patient.

[0080] A tenth aspect of the present invention is a blood purification device comprising a display unit and a communication unit, wherein the display unit displays a model creation screen on which treatment-related data used to generate a control model for controlling the blood purification device can be set, and the communication unit transmits the settings on the model creation screen. According to the tenth aspect, a control model using treatment-related data can be easily created by operating the model creation screen displayed on the display unit of the blood purification device. This makes it possible to support the use of a control model suitable for each patient or facility.

[0081] An eleventh aspect of the present invention is a model generation method comprising the steps of: displaying a model creation screen on the display unit of a blood purification device, on which treatment-related data used to generate a control model for controlling a blood purification device can be set; and generating the control model by machine learning using the treatment-related data set on the model creation screen as input. According to the eleventh aspect, a control model using treatment-related data can be easily created by operating the model creation screen. This makes it possible to support the use of a control model that is suitable for each patient or facility.

[0082] A twelfth aspect of the present invention is a program that enables a computer to implement a function for displaying a model creation screen on the display unit of a blood purification device, on which treatment-related data used to generate a control model for controlling the blood purification device can be set, and a function for transmitting the settings on the model creation screen. According to the twelfth aspect, a control model using treatment-related data can be easily created by operating the model creation screen displayed on the display unit of the blood purification device. This makes it possible to support the use of a control model that is suitable for each patient or facility. [Explanation of Symbols]

[0083] 10...Blood purification device, 14...Display unit, 18...Blood pump, 20...Blood purifier, 22...Blood circuit, 24...Dialysis fluid supply unit, 26...Intravenous fluid supply unit, 28...Control unit, 30...Dual pump, 32...Water removal pump, 34...Dialysis control unit, 42...Communication unit, 50...Information processing unit, 54...Storage unit, 56...Model generation unit, 58...Model evaluation unit, 70...Model creation screen, 80...Model evaluation screen, 90...Model selection screen, 100...Blood purification system.

Claims

1. A blood purification system equipped with a blood purification device, A display unit that displays a model creation screen on which treatment-related data can be set for generating a control model to control the blood purification device, The system includes a model generation unit that generates the control model by machine learning using treatment-related data set on the model creation screen as input. Blood purification system.

2. The model creation screen allows setting at least one of the learning method, algorithm, input data, and output data of the control model. The blood purification system according to claim 1.

3. The model creation screen allows setting at least one of the following conditions as conditions for treatment-related data used in machine learning of the control model: a condition related to the patient and a condition related to the period. The blood purification system according to claim 1.

4. The system further includes a model evaluation unit that calculates an evaluation value of the control model, The display unit can set the control model to be evaluated and the conditions for the treatment-related data used for evaluation, and displays a model evaluation screen that can display the evaluation value calculated by the model evaluation unit. The blood purification system according to claim 1.

5. The aforementioned model evaluation screen allows setting at least one of the following conditions for the treatment-related data used in the evaluation: a condition related to the patient and a condition related to the time period. The blood purification system according to claim 4.

6. It further includes a memory unit that stores multiple control models, The display unit displays a model selection screen that allows the selection of at least one of the plurality of control models. The blood purification device controls its operation based on the output of at least one control model selected on the model selection screen. The blood purification system according to claim 1.

7. The display unit prioritizes displaying the control model most suitable for the patient being treated from among the multiple control models. The blood purification system according to claim 6.

8. The preferred control model is based on evaluation values ​​obtained when evaluating the multiple control models using treatment-related data of the patient being treated. The blood purification system according to claim 7.

9. The aforementioned multiple control models include individual models generated by machine learning using treatment-related data of specific patients as input, and general-purpose models generated by machine learning using treatment-related data of multiple patients as input. A blood purification system according to any one of claims 6 to 8.

10. A blood purification device comprising a display unit and a communication unit, The display unit displays a model creation screen on which treatment-related data used to generate a control model for controlling the blood purification device can be set. The communication unit transmits the settings on the model creation screen. Blood purification device.

11. The steps include: displaying a model creation screen on the display unit of the blood purification device, which allows setting treatment-related data to be used in generating a control model for controlling the blood purification device; The process includes the step of generating the control model by machine learning using treatment-related data set on the model creation screen as input. Model generation method.

12. A function to display a model creation screen on the display unit of the blood purification device, which allows setting treatment-related data used to generate a control model for controlling the blood purification device, The function of sending the settings on the aforementioned model creation screen, and to enable the computer to perform this function, program.

Images