Patient care plan determination system, and patient care plan determination method

Abstract

To recommend inspection having high validity, according to a result at each timing of inspection, diagnosis, treatment and progress observation, in a whole workflow of a patient management.SOLUTION: A patient care plan determination system includes: a medical information server for holding attribute information and inspection results of a plurality of patients; an information terminal for providing a patient management policy indicating a classification of a target patient; and a calculator for acquiring predetermined information from the medical information server and the information terminal and determining a test item candidate related to a disease of the target patient. The calculator acquires the attribute information of the target patient and the patient management policy of the target patient from the medical information server, and applies patient attribute information of the target patient, a test result of the target patient, and the patient management policy of the target patient, to a test plan learned model for calculating a test value of each test item for specifying the disease according to the patient management policy, and executes estimation processing of a test value of a future test item candidate of the target patient and output processing of the test item candidate and the test value of the target patient.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to a patient care plan decision system and a patient care plan decision method. [Background technology]

[0002] The task of a doctor in ordering tests is to integrate and interpret a large amount of clinical information and select appropriate tests for each patient. Specifically, a doctor (i) examines the patient and collects patient information, lists suspected diseases based on the examination results, and prioritizes the diseases, (ii) selects and performs necessary tests from all possible tests, taking into account their contribution to disease identification and the burden on the patient (issuing a test order), (iii) interprets the test results and updates the list of suspected diseases, and (iv) repeats steps (ii) and (iii) until the disease is identified.

[0003] In many cases, doctors need a wealth of experience to place such test orders accurately. For this reason, for example, Patent Document 1 proposes a system that supports the creation of tests and / or treatment plans to be performed on a given patient by displaying case information registered in a database on a screen. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2007-287027 A Summary of the Invention [Problem to be solved by the invention]

[0005] It is extremely difficult to provide appropriate tests to patients based on the position of the test in the overall picture of patient management, such as initial consultation, follow-up visit, hospital follow-up, emergency care, etc., and the desired effect. Treatment and diagnosis of patients in hospitals are carried out in accordance with management policies. Here, management policies refer to information that specifies the tests, diagnoses, and treatments to be carried out according to patient classification (patient positioning), such as initial consultation patients, follow-up visit patients, hospitalized patients, patients under observation, and emergency patients.

[0006] The system disclosed in Patent Document 1 can only refer to past cases registered in a database, and cannot support examinations and treatments in line with patient management policies. In other words, even if the symptoms are the same, the examination and treatment methods will differ depending on the patient's position, such as an initial visit or a follow-up visit, so appropriate examinations and treatments cannot be performed simply by referring to past cases. Therefore, if there was a system that would consider the overall picture of patient management and recommend highly appropriate examination items at the right time, appropriate examinations could be performed on various patients without relying on the doctor's experience.

[0007] In light of this situation, this disclosure proposes technology that recommends highly appropriate tests based on the results at each stage of testing, diagnosis, treatment, and follow-up observation in the overall workflow of patient management. [Means for solving the problem]

[0008] In order to solve the above problem, the present disclosure provides, as an example, a patient care plan determination system that determines at least test item candidates related to a disease of a target patient, comprising: A medical information server that stores at least attribute information and test results of a plurality of patients; an information terminal that provides a patient management policy indicating a classification of target patients; A computer that acquires predetermined information from a medical information server and an information terminal and determines candidate test items related to a disease of a target patient, The computer, A process of acquiring attribute information of the target patient and a patient management policy of the target patient from a medical information server; A test value indicating information capable of identifying a disease according to a patient management policy, and a process of estimating the test value of future test item candidates of the target patient by applying the patient attribute information of the target patient, the test results of the target patient, and the patient management policy of the target patient to a test plan trained model for calculating the test value for each test item; A process of outputting candidate test items and test values ​​for a target patient; We propose a patient care planning decision system that implements the above.

[0009] Further features related to the present disclosure will become apparent from the description of the present specification and the accompanying drawings. Also, aspects of the present disclosure may be realized and realized by the elements and combinations of various elements and aspects of the following detailed description and the appended claims. The descriptions in this specification are exemplary and illustrative only and are not intended to limit the scope or application of the present disclosure in any way. Effect of the Invention

[0010] According to the technology disclosed herein, it becomes possible to recommend highly appropriate tests based on the results at each stage of testing, diagnosis, treatment, and follow-up observation in the overall workflow of patient management. [Brief description of the drawings]

[0011] [Figure 1] FIG. 1 is a diagram showing an example of a schematic configuration of an information processing system (also called a patient care plan determination system) 10 according to the present embodiment. [Diagram 2] 1 is a diagram illustrating an example of a functional configuration (example of a software configuration) of an information processing system (patient care plan determination system) 10 according to a first embodiment. [Diagram 3] 13 is a diagram for explaining a data flow in a learning process by the inspection learning unit 300 of the inspection planning unit 1012. FIG. [Figure 4]5 is a flowchart illustrating details of a learning parameter determination process according to the first embodiment. [Diagram 5] 13 is a flowchart for explaining details of the process of determining (learning) a threshold value of an inspection value when constructing a trained model. [Figure 6] 13 is a diagram for explaining a data flow in prediction processing by the inspection prediction unit 600 of the inspection planning unit 1012. FIG. [Figure 7] 11 is a flowchart for explaining details of determining examination item t+1 at time point t+1 and estimating the probability of being diagnosed with disease t+1 (probability of diagnosis of disease t+1) according to the first embodiment. [Figure 8] FIG. 8 is a diagram showing an example of the configuration of patient information 800 according to this embodiment (common to the first and second embodiments). [Figure 9] 13 is a flowchart for explaining an inspection cost estimation process executed by an inspection plan unit 1012. [Figure 10] FIG. 2 is a diagram showing an example of the configuration of a GUI 1000 for an examination order displayed on a display screen of an information terminal 110 according to the first embodiment. [Figure 11] FIG. 11 is a diagram illustrating an example of a functional configuration (example of a software configuration) of an information processing system (patient care plan determination system) 10 according to a second embodiment. [Figure 12] FIG. 11 is a diagram showing an example of detailed internal logical configurations of a medical care plan unit 1100 and an examination plan unit 1012 in the patient care plan determination system 10 of the second embodiment. [Figure 13] 13 is a diagram for explaining the data flow in the learning process by the treatment learning units t1301_1 to t+N1301_N of the treatment planning unit 1100. FIG. [Figure 14] 13 is a flowchart for explaining a learning parameter determination process (learned model generation process) by clinical learning units t1301_1 to t+N_1301_N in the second embodiment. [Figure 15] 13 is a diagram for explaining a data flow in treatment prediction processing by the treatment prediction unit 1500 of the treatment planning unit 1100. FIG. [Figure 16] 13 is a flowchart for explaining the process of estimating multiple patterns of patient management policy, medical treatment method, and patient disease information from time t+1 (time step t+1) to time t+N (time step t+N) according to the second embodiment. [Figure 17] FIG. 11 is a diagram for explaining a data flow in a learning process by an inspection learning unit 1700 of the inspection planning unit 1012 according to the second embodiment. [Figure 18] 10 is a flowchart illustrating details of a learning parameter determination process according to the second embodiment. [Figure 19] FIG. 19 is a diagram showing a configuration example of an output screen (UI: User Interface) 1900 for a medical treatment plan and an examination plan according to the second embodiment. [Figure 20] FIG. 11 is a diagram showing an example of the configuration of a GUI 2000 for an examination order displayed on the display screen of an information terminal 110 according to the second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] This embodiment discloses an information processing system (patient care plan determination system) that predicts a treatment plan including the processes and results of the patient management phase, examination, diagnosis, treatment, and follow-up observation based on patient information, and estimates the test items required to execute the treatment plan.

[0013] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. In the accompanying drawings, functionally identical elements may be indicated by the same numbers. Note that the accompanying drawings show specific embodiments and implementation examples according to the principles of the present disclosure, but these are for understanding the present disclosure and are by no means used to interpret the present disclosure in a limiting manner.

[0014] In the present embodiment, the description is given in sufficient detail for a person skilled in the art to implement the present disclosure, but it should be understood that other implementations and forms are possible, and that changes in configuration and structure and substitutions of various elements are possible without departing from the scope and spirit of the technical ideas of the present disclosure. Therefore, the following description should not be interpreted as being limited thereto.

[0015] (1) First embodiment <Example of hardware configuration of information processing system> 1 is a diagram showing an example of a schematic configuration of an information processing system (also called a patient care plan determination system) 10 according to this embodiment. Note that the hardware configuration example of the information processing system 10 is common to the first embodiment and the second embodiment described later.

[0016] The information processing system (patient care plan decision system) 10 includes, for example, a computer 100 provided on the cloud, an information terminal 110 installed in a doctor's examination room, and an external recording device (medical information server) 111, all of which are connected via a network (hospital network) 109.

[0017] The computer 100 includes a processor 101 such as a CPU, a main memory device 102, a secondary memory device 103, a network adapter 104, an input device 105, and an output device 106. The main memory device 102 holds, for example, a program for determining an examination order, which will be described later. The secondary memory device 103 stores, for example, information (electronic medical record information, medical receipt information, guideline information, etc.) acquired from an external recording device 111. The processor 101 reads the program from the main memory device 102, expands it in an internal memory (not shown), and realizes each processing unit (various functions for determining an examination order), which will be described later.

[0018] The input device 105 is composed of devices for inputting information to the computer 100, such as a keyboard, a mouse, and a touch panel. The output device 106 is composed of devices for outputting information, such as a display and a printer. The network adapter 104 receives information from the information terminal 110 via the network 109 and passes it to the processor 101 or stores it in the secondary storage device 103. In addition, the network adapter 104 transmits determined examination order information to the information terminal 110 via the network 109 in response to a command from the processor 101, for example.

[0019] <Example of functional configuration of information processing system> FIG. 2 is a diagram showing an example of a functional configuration (an example of a software configuration) of the information processing system (patient care plan determination system) 10 according to the first embodiment.

[0020] The information processing system 10 has, as its functions, a test ordering system 1110 provided in an external recording device (medical information server) 111, a patient information acquisition unit 1011 and a test planning unit 1012 set up in a processor 101 in a computer 100, and a test order output unit 1101 provided in an information terminal 110.

[0021] The information processing system 10 also handles electronic medical record information 1111, medical receipt information 1112, guideline information 1113, patient information 201, patient management policy 202, future test item candidates 203, and test value 204, all of which are stored in an external recording device (medical information server) 111. The medical receipt information 1112 includes medical fee (point) information. The guideline information 1113 stores information on test items medically determined by symptoms (e.g., perform test X in the case of symptoms A and B). The patient management policy 202 is information indicating the classification of patients, such as first visit, follow-up visit, hospitalization, follow-up observation, emergency, etc.

[0022] The patient information acquisition unit 1011 accesses the electronic medical record information 1111 of the external recording device (medical information server) 111, acquires information on past examinations and past diagnoses corresponding to the target patient, and acquires them as patient information 201. The patient information acquisition unit 1011 also acquires (receives) and outputs the patient management policy 202 input (selected) by the doctor from the information terminal 110. The patient information 201 includes, for example, patient attribute information (such as gender and age), previously taken examination items and examination result information, and patient disease information identified by the previous examination. In this specification, examination result information at time t (time step t) is represented as examination result information t, and examination result information at time t+1 (time step t+1) is represented as examination result information t+1. Furthermore, disease information confirmed by examination t+1 performed at time t+1 (time step t+1) is represented as patient disease information t+1. When a disease is unconfirmed by examination t at time t, the disease is identified by performing examination t+1.

[0023] The examination planning unit 1012 uses the patient information 201 and the patient management policy to calculate the examination value 204 indicating the amount of information (probability that a disease can be identified by each examination item) according to the disease type taking into account the patient management policy, and specifies examination items with a large amount of information (high probability value) as future examination item candidates 203. The examination planning unit 1012 transmits the specified future examination item candidates 203 to the information terminal 110 as recommended examination information.

[0024] The examination order output unit 1101 receives the future examination item candidates 203 and displays them on a display screen (not shown) of the information terminal 110. When a doctor checks the future examination item candidates 203 displayed on the display screen and selects a necessary examination item from the candidates, the examination order output unit 1013 transmits the examination order to the examination ordering system 1110. The test ordering system 1110 issues test orders received from the information terminal 110 to the testing room.

[0025] <Data flow of learning process by the inspection learning unit 300 of the inspection planning unit 1012> Fig. 3 is a diagram for explaining a data flow in a learning process by the inspection learning unit 300 of the inspection planning unit 1012. The inspection planning unit 1012 includes an inspection learning unit 300 that generates a trained model, and an inspection prediction unit 400 that infers inspection items and the like using the trained model. In Fig. 3, the inspection learning unit 300 generates various data and outputs a trained model 310 generated based on the data.

[0026] When generating a trained model, the test learning unit 300 acquires, as input data, guideline information, patient management policy t at time t (time step t), patient attribute information, test items, test result information t at time t, test result information t+1 at time t+1 (time step t+1), and patient disease information t+1 at time t+1. The guideline information and patient attribute information are fixed information when generating a trained model, but other information changes at each time. In other words, the patient management policy t at a certain time t (for example, if the patient was a first-time patient last time, this time the patient is diagnosed for the second time, so the management policy changes), the test result at time t, the test result information t+1 at the next time t+1, and the patient disease information t+1 determined based on the test result information t+1 are information that changes from time to time. In this way, the test learning unit 300 executes learning using information at multiple times to build the trained model 310.

[0027] The test learning unit 300 uses a probability distribution model to aggregate test result data for each test item and estimate a probability distribution 301 (estimation of probability distribution 301 in test results). As the amount of data increases, the probability distribution 301 approaches a normal distribution. Furthermore, the inspection learning unit 300 generates a probability distribution function by learning based on the estimated probability distribution, and generates a restored inspection result 302 using the probability distribution function. Next, the inspection learning unit 300 compares the restored inspection result 302 with the input inspection result to estimate an inspection value 303 for each inspection item.

[0028] Furthermore, the inspection learning unit 300 estimates a learning parameter 304 for estimating a threshold of the inspection value, that is, a threshold of the inspection value that identifies the inspection item, and stores the learning parameter threshold 304 for each inspection item in the main memory device 102 or the secondary memory device 103. Then, the test learning unit 300 estimates the diagnosis probability for the disease information t+1 at time t+1 and the risk of the disease (disease diagnosis probability and risk 305).

[0029] The test learning unit 300 determines learning parameters and outputs them as a trained model 310. The learning parameters include, for example, parameters of a neural network for extracting features from patient information, parameters for estimating a probability distribution, parameters for restoring test results, and parameters for estimating a test value.

[0030] <Details of learning parameter determination process> 4 is a flowchart for explaining the details of the learning parameter determination process according to the first embodiment. In the following explanation, the subject of operation of each step is the inspection learning unit 300, but since the inspection planning unit 1012 including the inspection learning unit 300 is realized by deploying a program in the processor 101, the processor 101 may be the subject of operation.

[0031] (i) S401 The testing learning unit 300 accepts patient information necessary to generate the trained model 310. It is preferable to obtain as much patient information as possible in order to generate a more accurate trained model 310. Here, the patient information is composed of various information of various patients, and each piece of information includes guideline information, patient management policy t at time t (time step t), patient information, test items, test result information t at time t, test result information t+1 at time t+1 (time step t+1), and patient disease information t+1 at time t+1, as described above.

[0032] (ii) S402 The testing learning unit 300 integrates the patient information acquired in S401 and extracts the features of the patient information. Specifically, the testing learning unit 300 converts the received patient information into text and into numerical values ​​so that it can be processed by a neural network (integration of patient information), and calculates the features of the numerical information by applying the integrated patient information to a neural network (e.g., deep learning).

[0033] (iii)S403 The test learning unit 300 aggregates a plurality of test results for each test item, and estimates a probability distribution (function) for each test item using a probability distribution model.

[0034] (iv)S404 The test learning unit 300 restores the test result based on the probability distribution estimated in S403. Since the test result as the correct value is already known, this is a process to confirm whether the correct value can be obtained from the estimated probability distribution.

[0035] (v)S405 The test learning unit 300 estimates the test value from the test results restored in S404. If the original test results can be restored from the estimated probability distribution, the estimated probability distribution is deemed to have been correctly estimated. In S405, the test value, which is the value that will be obtained when new patient information is input, is estimated.

[0036] (vi)S406 The inspection learning unit 300 estimates a learning error in the inspection value estimated in S405. Here, the learning error is the error between the original inspection result and the restored inspection result.

[0037] (vii)S407 The inspection learning unit 300 determines whether the learning error value estimated in S406 has converged. Whether or not the learning error value has converged is determined based on whether or not the value is smaller than a preset threshold. If the learning error value has converged (Yes in S407), the process proceeds to S408. If the learning error value has not converged (No in S407), the process proceeds to S409. When the learning error is small, a trained model has been constructed that can predict a patient's test value when new patient information is input.

[0038] (viii)S408 The inspection learning unit 300 sets the learning parameters with which the learning error has converged to the trained model. Here, the learning parameters include the parameters of the neural network in S402, the parameters used in the probability distribution estimation, the parameters used in the inspection result reconstruction, and the parameters used in the inspection value estimation.

[0039] (ix)S409 The test learning unit 300 updates the learning parameters (for example, parameters of a neural network) in accordance with an optimization function for updating so as to reduce the learning error.

[0040] <Details of the process to determine the threshold> FIG. 5 is a flowchart for explaining details of the process of determining (learning) a threshold value of the inspection value when constructing a trained model.

[0041] (i) S501 The test learning unit 300 acquires the test value for each test item and the guideline information 1113 .

[0042] (ii) S502 The test learning unit 300 receives test item t+1 at time point t+1 (time step t+1).

[0043] (iii)S503 The inspection learning unit 300 estimates a threshold value of the inspection value that identifies the inspection item. It is possible to determine a learning parameter for estimating the threshold value of the inspection value by determining an initial value (fixed value) of the threshold value and changing the parameter so as to reduce the learning error.

[0044] Since it is assumed that the threshold will change depending on the management policy (patient classification: first visit, follow-up visit, hospitalized, under observation, emergency, etc.), the test value threshold is estimated in S503. The threshold is estimated corresponding to the test item, and the final test value threshold is determined (estimated) by learning to match the management policy in the subsequent processing.

[0045] (iv) S504 The test learning unit 300 estimates the diagnostic probability and risk (the level of risk of the diagnosed disease itself) for disease information t+1 obtained from test result t+1 corresponding to test item t+1. Although the threshold estimated in S503 may not be accurate, the diagnostic probability and risk of a specific disease are estimated using the test result restored from the probability distribution estimated in the process of Figure 4 and the estimated threshold.

[0046] (v) S505 The test learning unit 300 estimates a management policy t at time t (time step t) from the diagnostic probability and risk information estimated in S504. That is, in S505, it is estimated which category the patient belongs to (first visit, return visit, hospitalized, under observation, emergency, etc.).

[0047] (vi) S506 The test learning unit 300 acquires an actual patient management policy t.

[0048] (vii) S507 The test learning unit 300 compares the patient management policy t estimated in S505 with the actual patient management policy t acquired in S506, and estimates the learning error.

[0049] (viii)S508 The inspection learning unit 300 determines whether the learning error value estimated in S507 has converged. Whether or not the learning error value has converged is determined based on whether or not the value is smaller than a preset threshold. If the learning error value has converged (Yes in S508), the process proceeds to S509. If the learning error value has not converged (No in S508), the process proceeds to S510.

[0050] (ix)S509 The inspection learning unit 300 determines the parameters of the threshold of the inspection value estimated in S503 as learning parameters for estimating the threshold of the inspection value.

[0051] (x)S510 The inspection learning unit 300 updates the learning parameters (neural network parameters) for estimating the threshold of the inspection value according to the optimization function. Then, in S503, the process of estimating the threshold of the inspection value for identifying the inspection item is executed again using the updated learning parameters.

[0052] <Data flow of prediction processing by the test prediction unit 600 of the test planning unit 1012> Fig. 6 is a diagram for explaining a data flow in prediction processing by the test prediction unit 600 of the test planning unit 1012. In Fig. 6, the test prediction unit 600 applies data at time point t (time step t) to a trained model, and outputs test item t+1 at time point t+1 (time step t+1) and the diagnosis probability and risk for disease t+1 at time point t+1.

[0053] When predicting a test item t+1, etc., the test prediction unit 600 acquires, as input data, a patient management policy t at time t, patient attribute information, and test items and test results t at time t. For example, based on the most recent patient management policy and test results, it predicts the test items to be performed at the next time point, the probability of being diagnosed with a certain disease, and the risk level of that disease.

[0054] The test prediction unit 600 applies the patient management policy t at time t, the patient attribute information, and the test items and test results t at time t to the trained model generated by the test learning unit 300, thereby generating candidates for test item t+1 at time t+1 and their test results 601, test values ​​602 for each of the candidate test items t+1, and a threshold value 603 for the test value. As described above, for example, the test item t can be the most recent test item, the test result t can be the most recent test result corresponding to the test item, and the candidate test item t+1 can be the candidate test item to be taken at the next time.

[0055] The test prediction unit 600 further determines whether the test value (test value of each test item candidate) 602 of the test item t+1 candidate is greater than a threshold 603 for the test value, determines the test item t+1 whose test value is greater than the threshold, and estimates the probability (diagnosis probability) of being diagnosed with the disease at time t+1 (the disease at the next time point).

[0056] <Details of the test item determination (prediction) process> 7 is a flowchart for explaining the details of determining an examination item t+1 at time t+1 (time step t+1) and estimating the probability of being diagnosed with disease t+1 (probability of being diagnosed with disease t+1) according to the first embodiment. In the following explanation, the examination prediction unit 600 is the subject of operation for each step, but since the examination planning unit 1012 including the examination prediction unit 600 is realized by deploying a program in the processor 101, the processor 101 may be the subject of operation.

[0057] (i) S701 The test prediction unit 600 receives (acquires) the patient information to be processed, the trained model 310, and the patient management policy t, which are input by, for example, an operator.

[0058] (ii) S702 The test prediction unit 600 integrates the patient information acquired in S701 and extracts features of the patient information. Specifically, the test prediction unit 600 converts the received patient information into text and into numerical values ​​so that it can be processed by a neural network (integration of patient information), and calculates features of the numerical information by applying the integrated patient information to a neural network (e.g., deep learning).

[0059] (iii)S703 The test prediction unit 600 estimates the test results for multiple test items by applying the features extracted in S702 for each candidate for test item t+1 to the trained model 310.

[0060] (iv) S704 The test prediction unit 600 digitizes each test result estimated in S703 and estimates the test value of each candidate test item t+1. The test value is a probability value that indicates how much the probability of diagnosing a disease can be improved if the corresponding test item candidate is selected. The test value is expressed as an amount of information, and is obtained by calculation using the trained model 310.

[0061] (v)S705 The test prediction section 600 identifies the test item whose value is the maximum among the test values ​​estimated in S704.

[0062] (vi)S706 The test prediction unit 600 estimates a threshold for determining the value of testing according to the management policy t and the amount of information on the value of testing.

[0063] (vii) S707 The test prediction unit 600 determines whether the test value identified in S705 is greater than the threshold value estimated in S706. If the test value is greater than the threshold value (Yes in S707), the process proceeds to S708. If the test value is equal to or less than the threshold value (No in S707), the process proceeds to S710.

[0064] (viii)S708 The test prediction unit 600 determines the test item with the maximum test value as test item t+1_610.

[0065] (ix)S709 The test prediction section 600 estimates the probability and risk 620 of being diagnosed with disease t+1 when test item t+1 is performed. Specifically, the test prediction unit 600 restores the test value from the test value probability distribution information of the trained model 310, and estimates the disease diagnosis probability. Furthermore, the test prediction unit 600 estimates the test value based on the estimated disease diagnosis probability. Then, the test prediction unit 600 identifies the test item that maximizes the test value (disease diagnosis probability). The disease diagnosis probability corresponding to the identified test item becomes the probability of being diagnosed with the disease t+1. In this way, the test value is estimated based on the disease probability estimation, so the disease probability is also estimated in the process of estimating the test value. Furthermore, the test prediction unit 600 refers to guideline information and identifies risk (disease risk) 620 on a rule basis.

[0066] (x)S710 The test prediction unit 600 adds the estimated test results of the test items identified in S705 to the test results t.

[0067] (xi)S711 The test prediction unit 600 updates the candidates for the test item t+1. For example, when k test items (m is a positive integer) are candidates, the next group of m test item candidates are set as new test items to be processed. Then, the processing from S703 onwards is executed for the new test item candidates.

[0068] <Example of patient information> FIG. 8 is a diagram showing an example of the configuration of patient information 800 according to this embodiment (common to the first and second embodiments).

[0069] The patient information includes identification information (ID) 801, patient attribute information 802, date and time 803, a patient management policy 804, a group of results 805 of multiple tests (test items A to Z), and disease information 806 as constituent items.

[0070] Identification information (ID) 801 is information for uniquely identifying and identifying a patient. Patient attribute information 802 is information indicating the patient's gender and age. Date and time 803 is information indicating the consultation date and time. Patient management policy 804 is information indicating the patient classification, such as first visit, follow-up visit, hospitalized, under observation, emergency, etc., as described above. A group of results 805 of multiple tests (test items A to Z) is information indicating the test values ​​and judgment results (abnormal (abnormal (H) because the value is too high or abnormal (L) because the value is too low), normal) corresponding to each test item. Disease information 806 is information indicating the type of disease diagnosed by the doctor based on the test results.

[0071] <Test cost estimation process> 9 is a flowchart for explaining the inspection cost estimation process executed by the inspection planning unit 1012. Note that the inspection cost estimation process is a rule-based process and does not involve learning.

[0072] (i) S901 The examination planning unit 1012 receives medical receipt information 1112 including information on medical fee calculation conditions.

[0073] (ii) S902 The inspection planning unit 1012 receives information on the inspection item t+1_610 predicted by the inspection prediction unit 600.

[0074] (iii) S903 The examination plan unit 1012 receives a patient management policy t for the target patient.

[0075] (iv) S904 The examination planning unit 1012 estimates (calculates) the medical fee points for the examination item t+1_610 based on the medical fee calculation conditions acquired in S901.

[0076] (v)S905 The examination planning unit 1012 estimates (calculates) the cost to be borne by the hospital based on the medical fee points estimated (calculated) in S904. Information on the estimated cost to be borne by the hospital is transmitted to the information terminal 110 and can be displayed on the display screen of the information terminal 110.

[0077] <Example of GUI Configuration Displayed on the Display Screen of Information Terminal 110> 10 is a diagram showing a configuration example of a test order GUI 1000 displayed on the display screen of the information terminal 110 according to the first embodiment. Through the test order GUI 1000, a patient management policy is selected and displayed (selected by a doctor), and disease determination results and test items are output.

[0078] The test order GUI 1000 includes, for example, a management policy selection / display section 1001, a disease probability etc. display section 1002, and a disease test item display section 1003 as components.

[0079] The management policy selection / display unit 1001 displays the patient management policies for the target patient that the doctor can select when the computer 100 of the patient care plan determination system 10 is provided with recommended test information, and the currently selected patient management policy.

[0080] The disease probability display section 1002 provides information on the disease name, disease probability, and disease risk (how much risk there is of the disease itself) obtained by the examination and disease prediction process (prediction simulation). For example, the prediction process shows that disease A has a 70% probability of being correct, and that the risk of the disease is medium.

[0081] The disease test item display section 1003 displays, for example, test items for diseases with a disease probability higher than a predetermined value or test items for high-risk diseases, patient costs indicating the cost paid by the patient for the test, and hospital costs indicating the cost borne by the hospital. In the example of Fig. 10, test items for disease A with a probability of 70% and disease C with a high risk are displayed.

[0082] (2) Second embodiment The second embodiment proposes a patient care plan determination system 10 that provides a medical treatment plan for a target patient in addition to the examination plan according to the first embodiment. The hardware configuration of the patient care plan determination system 10 according to the second embodiment is as shown in FIG. 1, and therefore the description thereof will be omitted.

[0083] <Example of functional configuration of information processing system> FIG. 11 is a diagram showing an example of a functional configuration (example of a software configuration) of an information processing system (patient care plan determination system) 10 according to the second embodiment.

[0084] The information processing system 10 has, as its functions, a test ordering system 1110 provided in an external recording device (medical information server) 111, a patient information acquisition unit 1011, a medical treatment planning unit 1100, and a test planning unit 1012 set up in a processor 101 in a computer 100, and an information terminal 110 and a test order output unit 1101.

[0085] The information processing system 10 also handles, as information, electronic medical record information 1111, medical receipt information 1112, guideline information 1113, patient information 201, patient management policy 202, and examination plan i_10121 (including, for example, the above-mentioned future examination item candidates 203 and examination value 204), all of which are stored in an external recording device (medical information server) 111. The medical receipt information 1112 includes medical fee (point) information. The guideline information 1113 stores information on examination items medically determined by symptoms (e.g., in the case of symptoms A and B, examination X is performed).

[0086] As in the first embodiment, the patient information acquisition unit 1011 accesses electronic medical record information 1111 in the external recording device (medical information server) 111, acquires information on past examinations and past diagnoses corresponding to the target patient, and acquires this as patient information 201. The patient information acquisition unit 1011 also acquires (receives) and outputs a patient management policy 202 input (selected) by a doctor from the information terminal 110. The patient information 201 is composed of information as shown in FIG.

[0087] The medical treatment plan unit 1100 uses the patient information 201 and the patient management policy 202 to predict (estimate) a future management policy, medical treatment method, and disease information, and outputs it as a patient medical treatment plan 11001. For example, in the case of an inpatient, the patient medical treatment plan 11001 includes information such as whether or not the current treatment will be continued, and if the treatment method is to be changed, which method will be used.

[0088] The examination planning unit 1012 determines (predicts) an examination plan 10121 using the patient information 201 and the patient management policy 202, as in the first embodiment. The examination plan 10121 includes, as information, future examination items, examination results, and predictions of examination value, as in the first embodiment. The examination plan 10121 includes future examination item candidates 203 and examination value 204 (see FIG. 2). The examination planning unit 1012 calculates an examination value 204 indicating an amount of information (probability that a disease can be identified by each examination item) according to the disease type, taking into account the patient management policy 202, and specifies an examination item with a large amount of information (high probability value) as a future examination item candidate 203. The examination planning unit 1012 transmits the specified future examination item candidate 203 to the information terminal 110 as recommended examination information.

[0089] The examination order output unit 1013 receives the future examination item candidates 203 and displays them on a display screen (not shown). When a doctor checks the future examination item candidates 203 displayed on the display screen and selects a necessary examination item from the candidates, the examination order output unit 1013 transmits an examination order to the examination ordering system 1110. The test ordering system 1110 issues test orders received from the information terminal 110 to the testing room.

[0090] <Example of detailed internal logical configuration of the medical care planning unit 1100 and the examination planning unit 1012> FIG. 12 is a diagram showing an example of detailed internal logical configurations of the treatment plan unit 1100 and the examination plan unit 1012 in the patient care plan determination system 10 of the second embodiment.

[0091] (i) The medical treatment planning unit 1100 includes N+1 medical treatment prediction units, namely, a medical treatment prediction unit t_1100_1 at time t (time step t), a medical treatment prediction unit t+1_1100_2 at time t+1 (time step t+1), a medical treatment prediction unit t+2_1100_3 at time t+2 (time step t+2), and a medical treatment prediction unit t+1_1100_N+1 at time t+N (time step t+N).

[0092] One treatment prediction unit t+k_1100_k+1 (k = 0, 1, ..., N) acquires (accepts) patient management policy and patient information (see Figure 8) from time tk (time step tk) to time t (time step t) (acquires information prior to treatment prediction), and uses this information to generate treatment plan prediction information consisting of patient management policy t+k, treatment method t+k, and patient disease information t+k at time t+k (time step t+k), and outputs it to the examination planning unit 1012.

[0093] (ii) The inspection planning unit 1012 includes N+1 inspection prediction units, namely, an inspection prediction unit t_1012_1 at time t (time step t), an inspection prediction unit t+1_1012_2 at time t+1 (time step t+1), an inspection prediction unit t+2_1012_3 at time t+2 (time step t+2), and an inspection prediction unit t+1_1012_N+1 at time t+N (time step t+N).

[0094] The test prediction unit t_1012_1 acquires patient information (FIG. 8), test items, and test result information t from the medical information server 111, and, as in the first embodiment, estimates the probability distribution of the test results for each test item → restores the test results at the next time step t+1 → estimates the test value, the disease diagnosis probability, and its risk. The test prediction unit t+k_1012_k+1 (k=0,1,...,N) acquires patient information, test items, and test result information t+k at time t+k (time step t+k) (test prediction result estimated by the test prediction unit t+k-1), and estimates the probability distribution of the test results for each test item → restores the test results at the next time step t+k+1 → estimates the test value, the disease diagnosis probability, and its risk, and outputs the information.

[0095] (iii) Furthermore, when an examination is actually performed, the patient's examination result 1200_1 at time t (time step t), the patient's examination result 1200_2 at time t+1 (time step t+1), the patient's examination result 1200_3 at time t+2 (time step t+2), ..., the patient's examination result 1200_N+1 at time t+N (time step t+N) are output and aggregated and stored in the medical information server 111 as examination results t, ..., t+N_1201.

[0096] <Data flow of learning process by the treatment learning unit 1301 of the treatment planning unit 1100> Fig. 13 is a diagram for explaining the data flow in the learning process by the medical treatment learning unit t1301_1 to the medical treatment learning unit t+N1301_N of the medical treatment planning unit 1100. The medical treatment planning unit 1100 includes the medical treatment learning unit t1301_1 to the medical treatment learning unit t+N1301_N that generate trained models, and the medical treatment prediction unit 1500 (see Fig. 15) that infers a medical treatment method and the like using the trained models. In Fig. 13, the medical treatment learning unit t1301_1 to the medical treatment learning unit t+N1301_N generate various data and output the trained model 1303 generated based on them.

[0097] The medical treatment learning unit t1301_1 acquires, as input data, a patient management policy at time t (time step t), medical treatment method information t at time t (time step t), patient attribute information, examination items, examination result information t at time t (time step t), and patient disease information t at time t (time step t). The patient attribute information and examination items are fixed information when generating a trained model, but other information changes at each time. In other words, the patient management policy t at a certain time t (time step t) (for example, if the patient was a first-time patient last time, this time the patient is a second-time patient, so the management policy changes), the examination result at time t (time step t), the examination result information t+1 at the next time t+1 (time step t+1), and the patient disease information t+1 determined based on the examination result information t+1 are information that changes from time to time. The medical treatment learning units 1301_1 to 1301_N sequentially execute learning using information at multiple times in this way to build a trained model 1303.

[0098] The clinical learning unit t1301 at time t (time step t) and the clinical learning unit t+k1301_k at time t+k (k=1, , N) (time step t+k) estimate the patient management policy t+k+1, clinical method t+k+1, and patient disease information t+k+1 at time t+k+1 (time step t+k+1) using patient attribute information, patient management policies t and t+k at time t and t+k (time step t and time step t+k), and clinical method information t+k. The clinical learning unit t+k1301_k then compares the actual patient management policy t+k+1, the implemented clinical method t+k+1, and the actual patient disease information t+k+1 with the predicted (estimated) patient management policy t+k+1, the predicted clinical method t+k+1, and the predicted patient disease information t+k+1 to estimate the error t+k1302_k of the trained model. Furthermore, the medical learning unit t+k 1301_k repeats the learning process to converge the error t+k of the trained model and determine the learning parameters of the trained model 1303.

[0099] <Processing for generating trained models by the clinical learning unit> 14 is a flowchart for explaining the learning parameter determination process (trained model generation process) by the clinical learning unit t1301_1 to the clinical learning unit t+N_1301_N in the second embodiment. In the following explanation, the clinical learning unit is the subject of operation of each step, which is a general term for the clinical learning unit t1301_1 or the clinical learning unit t+k1301_k (k is an integer from 1 to N). In addition, since the clinical planning unit 1100 including the clinical learning unit 1301 is realized by deploying a program in the processor 101, the processor 101 may be the subject of operation.

[0100] (i) S1401 The medical care learning unit 1301 acquires, for example, information on a plurality of patients from the medical information server 111, such as patient information and patient management policies from time tk to time t (time step tk to time step t) for the plurality of patients. Here, the patient information from time step tk to t includes patient attribute information, examination item and examination result information from time tk to time t (time step tk to t), patient disease information, and medical care method information (see FIG. 8 for the configuration items of patient information).

[0101] (ii) S1402 The medical care learning unit 1301 integrates the patient information and patient management policy acquired in S1401 and extracts the features of that information. Specifically, the medical care learning unit 1301 converts the received patient information and patient management information into text and into numerical values ​​so that they can be processed by a neural network (information integration), and extracts the features of the numerical information by applying the integrated information to a neural network (e.g., deep learning).

[0102] (iii) S1403 The medical care learning unit 1301 estimates the patient management policy, medical care method, and patient disease information from time t+1 to time t+N (time step t+1 to time step t+N) using the feature amount extracted in S1402. For example, the estimation can be performed by machine learning using a neural network such as deep learning. For example, by using an RNN (Recurrent Neural Network) or the like to perform time-series learning on the feature amounts of the patient information, management policy, etc. at each time point in the past, the patient management policy, medical care method, and patient disease information at each time point in the future can be predicted. More specifically, the machine learning model used for the prediction calculation uses the patient information and patient management policy at each time point in the past as time-series observation data, and predicts the change in the patient disease information as the treatment effect of a specific medical care method.

[0103] (iv) S1404 The medical treatment learning unit 1301 compares the patient management policy, medical treatment method, and patient disease information (medical treatment plan) from time step t+1 to time step t+N obtained in S1403 with the actual patient management policy, medical treatment method, and patient disease information from time step t+1 to time step t+N, and estimates the learning error.

[0104] Here, the learning error indicates the error in the estimation results for treatment plan patterns 1 to M (M is an integer equal to or greater than 2). The learning error may be a weighted average of the actual error and the virtual error. The actual error represents, for example, the difference from the patient information acquired from a treatment plan actually performed for patient A (e.g., treatment plan 1). Meanwhile, the virtual error represents the difference from virtual patient information acquired from a treatment plan that was not actually performed (e.g., treatment plan 2). The virtual patient information is acquired from another patient (e.g., patient B) for whom treatment plan 2 was actually performed. Patient B is selected from a group of patients whose patient information features are similar to those of patient A.

[0105] (v)S1405 The medical learning unit 1301 compares the learning error estimated in S1404 with a preset threshold value to determine whether the learning error has converged. If the learning error has converged (Yes in S1405), the process proceeds to S1406. On the other hand, if the learning error has not converged (No in S1405), the process proceeds to S1407.

[0106] (vi)S1406 The medical treatment learning unit 1301 determines the parameters of the trained model used for extracting features in S1402 and estimating the patient management policy, medical treatment method, and patient disease information in S1403 as parameters of the trained model.

[0107] (vii) S1407 The medical care learning unit 1301 updates the learning parameters so as to reduce the learning error according to the optimization function. The updated learning parameters are used when estimating information such as a patient management policy again in the repeated processing.

[0108] <Data flow of prediction processing by the medical treatment prediction unit 1500 of the medical treatment planning unit 1100> 15 is a diagram for explaining a data flow in treatment prediction processing by the treatment prediction unit 1500 of the treatment planning unit 1100. The treatment prediction unit 1500 applies input data to the trained model 1303 to predict the diagnosis of tuberculosis.

[0109] The medical treatment prediction unit 1500 acquires, as input data, a patient management policy tk at time tk (time step tk) to a patient management policy t at time t (time step t), medical treatment method information tk at time tk (time step tk) to medical treatment method information t at time t (time step t), patient attribute information, examination items and examination result information tk at time tk (time step tk) to examination result information t at time t (time step t), and patient disease information tk at time tk (time step tk) to patient disease information t at time t (time step t). In the second embodiment, a medical treatment plan t+1 at time t+1 (time step t+1) to a medical treatment plan t+N at time t+N (time step t+N) are predicted using information from time tk (time step tk) to time t (time step t) of the patient to be processed.

[0110] The medical treatment prediction unit 1500 calculates a predicted patient management policy t+N, a predicted medical treatment method information t+N, and a predicted patient disease information t+N from the predicted patient management policy t+1, the predicted medical treatment method information t+1, and the predicted patient disease information t+1 for medical treatment plan patterns 1 to M (M is an integer of 2 or more) by applying the patient management policy tk at time tk (time step tk) to the patient management policy t at time t (time step tk), the medical treatment method information tk at time tk (time step tk) to the medical treatment method information t at time t (time step t), the patient attribute information, the test items and the test result information tk at time tk (time step tk) to the test result information t at time t (time step t), and the patient disease information tk at time tk (time step tk) to the patient disease information t at time t (time step t) to the trained model 1303.

[0111] Here, treatment plan patterns 1 to M are patterns obtained by extracting possible treatment methods for the target patient, and include not only treatment methods predicted from the features of the target patient, but also treatment plan patterns adopted for patients with features similar to those of the target patient. This is to predict the treatment effects (disease information) of treatment plan patterns that can become treatment options, so that doctors can select the optimal treatment option.

[0112] <Details of the test item determination (prediction) process> 16 is a flowchart for explaining the process of estimating multiple patterns of patient management policy, medical treatment method, and patient disease information from time t+1 (time step t+1) to time t+N (time step t+N) according to the second embodiment. In the following explanation, the medical treatment prediction unit 1500 is the subject of operation of each step, but since the medical treatment planning unit 1100 including the medical treatment prediction unit 1500 is realized by deploying a program in the processor 101, the processor 101 may be the subject of operation.

[0113] (i) S1601 The medical treatment prediction unit 1500 acquires (accepts) patient information (patient information of the patient to be processed) from time step tk to time step t and patient management policy from time step tk to time step t from the medical information server 111, for example. The patient information from time step tk to time step t includes patient attribute information, medical treatment method information from time step tk to time step t, test result information from time step tk to time step t, and patient disease information from time step tk to time step t. In addition, the medical treatment prediction unit 1500 sets up the trained model 1303 (prepares the trained model 1303 for use).

[0114] (ii) S1602 The medical care prediction unit 1500 integrates the patient information acquired in S1601 and extracts features of the patient information. Specifically, the medical care prediction unit 1500 converts the received patient information into text and into numerical values ​​so that it can be processed by a neural network (integration of patient information), and calculates features of the numerical information by applying the integrated patient information to a neural network (e.g., deep learning).

[0115] (iii)S1603 The treatment prediction unit 1500 applies the features of the patient information acquired in S1602 to the trained model 1303 to estimate the patient management policy, patient disease information, and treatment method as a treatment plan from time step t+1 to time step t+N. The treatment plan is estimated for multiple patterns (Pattern 1 to Pattern M) at each of time steps t+1 to t+N.

[0116] (iv) S1604 The output device 106 receives the information estimated in S1603 (patient management policy and treatment method for each patient disease information from time step t+1 to time step t+N) from the treatment prediction unit 1500, and transmits this information to the information terminal 110 as a treatment plan (pattern 1 to pattern M). The information terminal 110 displays, for example, the received information on the medical treatment plan on a display screen, allowing the user (doctor) to select a desired medical treatment plan.

[0117] <Data flow of learning process by the inspection learning unit 1700 of the inspection planning unit 1012> Fig. 17 is a diagram for explaining a data flow in a learning process by the inspection learning unit 1700 of the inspection planning unit 1012 of the second embodiment. The inspection planning unit 1012 includes an inspection learning unit 1700 that generates a trained model, and an inspection prediction unit 400 that infers inspection items and the like using the trained model. In Fig. 17, the inspection learning unit 1700 generates various data and outputs a trained model 1701 generated based on the data.

[0118] When generating a trained model, the test learning unit 1700 acquires as input data guideline information, a patient management policy t at time t (time step t) to a patient management policy t+N at time t+N (time step t+N), patient attribute information, test items and test result information t at time t (time step t) to test result information t+N at time t+N (time step t+N), test result information t+N at time t+1 (time step t+1), test result information t+1 to t+N (time step t+N), patient disease information t+N at time t+1 (time step t+1), patient disease information t+N from time t+1 to time t+N (time step t+N), and treatment method t+N at time t+1 (time step t+1) to time t+N (time step t+N).

[0119] The test learning unit 1700 uses a probability distribution model to aggregate test result data for each test item and estimate a probability distribution 1702 in the test results (estimation of probability distribution in the test results). As the number of data increases, the probability distribution approaches a normal distribution.

[0120] In addition, the testing learning unit 1700 generates a probability distribution function by learning based on the estimated probability distribution 1702, and uses it to generate reconstructed testing results 1703_1 at time t+1 (time step t+1) to reconstructed testing results 1703_N at time t+N (time step t+N), testing values ​​1704_1 for each testing item at time t+1 (time step t+1) to testing values ​​1704_N for each testing item at time t+N (time step t+N), and disease diagnosis probability and risk 1705_1 at time t+1 (time step t+1) to disease diagnosis probability and risk 1705_N at time t+N (time step t+N). Then, the testing learning unit 1700 estimates from the information a patient management policy 1706_1 at time t+1 (time step t+1) to a patient management policy 1706_N at time t+N (time step t+N), patient disease information 1707_1 at time t+1 (time step t+1) to a patient disease information 1707_N at time t+N (time step t+N), and a treatment method 1708_1 at time t+1 (time step t+1) to a treatment method 1708_N at time t+N (time step t+N). Furthermore, the testing learning unit 1700 compares the estimated patient management policy, patient disease information, and treatment method with the actual patient management policy, patient disease information, and treatment method, and estimates a learning error 1709_1 at time t (time step t) to a learning error 1709_N at time t+N (time step t+N). When the learning error t to the learning error t+N has converged, the test learning unit 1700 outputs a trained model 1701 having parameters related to the converged learning error.

[0121] <Details of learning parameter determination process> 18 is a flowchart for explaining the details of the learning parameter determination process according to the second embodiment. In the following explanation, the testing and learning unit 300 is the subject of operation for each step, but since the testing plan unit 1012 including the testing and learning unit 1700 is realized by loading a program into the processor 101, the processor 101 may be the subject of operation.

[0122] (i) S1801 The test learning unit 1700 receives patient attribute information, test item information, and test result information t at time point t (time step t) required to generate a trained model 1701.

[0123] (ii) S1802 The testing learning unit 1700 integrates the information acquired in S1801 and extracts the feature quantities of the information. Specifically, the testing learning unit 1700 converts the received patient attribute information, information on test items, and test result information t at time t (time step t) into text and digitizes it so that it can be processed by a neural network (information integration), and calculates the feature quantities of the numerical information by applying the integrated information to a neural network (e.g., deep learning).

[0124] (iii)S1803 The test learning unit 1700 aggregates a plurality of test results for each test item, and estimates a probability distribution (function) for each test item using a probability distribution model.

[0125] (iv) S1804 The test learning unit 1700 reconstructs the test result in the next time step (t→t+1) based on the probability distribution estimated in S1703. Since the test result as the correct value is already known, this is a process to check whether the correct value can be obtained from the estimated probability distribution.

[0126] (v)S1805 The test learning unit 1700 estimates the test value, disease diagnosis probability, and risk from the test results restored in S1804. If the original test results can be restored from the estimated probability distribution, the estimated probability distribution is deemed to have been correctly estimated. In S1805, the test value, which indicates the value that will be obtained when new patient information is input, and the disease diagnosis probability and risk are estimated.

[0127] (vi) S1806 The inspection learning unit 1700 estimates the learning error of the inspection result of the next time step (t→t+1) using the restored inspection result estimated in S1804. Here, the learning error is the error between the original inspection result and the restored inspection result.

[0128] (vii) S1807 The test learning unit 1700 estimates patient management policy, disease information, and treatment method for the next time step (t → t + 1) by applying the test value estimated in S1805 and the disease diagnosis probability and risk to, for example, a deep learning method for time series learning (RNN).

[0129] (viii) S1808 The test learning unit 1700 estimates the learning error between the actual treatment method and the treatment method of the next time step (time step t+1) estimated in S1807.

[0130] (ix)S1809 The inspection learning unit 1700 judges whether the processing is completed for all time steps (t to t+N). If the processing is completed (Yes in S1809), the processing proceeds to S1810. If there is information on time steps that have not yet been processed (No in S1809), the processing proceeds to S1811.

[0131] (x)S1810 The inspection learning unit 1700 determines whether the values ​​of the learning error estimated in S1806 and the learning error estimated in S1808 have converged. Whether or not the values ​​have converged can be determined, for example, by whether or not they are smaller than a preset threshold. If the learning error values ​​have converged (Yes in S1810), the process proceeds to S1813. If the learning error values ​​have not converged (No in S1810), the process proceeds to S1814.

[0132] (xi)S1811 The checking and learning unit 1700 updates the current time step (eg, t) to the next time step (eg, t+1).

[0133] (xii) S1812 The inspection learning unit 1700 acquires inspection item and inspection result information in the current time step + 1 (for example, t + 1). Then, the process proceeds to S1802, where information integration and feature extraction are executed again.

[0134] (xiii)S1813 The test learning unit 1700 sets the learning parameters for which the learning error has converged to the learned model. Here, the learning parameters include a patient management policy, medical treatment method information, and neural network parameters within the RNN for predicting disease information.

[0135] (xiv)S1814 Test learning unit 1700 updates the learning parameters (for example, parameters of a neural network) according to an optimization function for updating so as to reduce the learning error, and executes the process again from S1801.

[0136] <Treatment plan and examination plan output screen> 19 is a diagram showing a configuration example of an output screen (UI: User Interface) 1900 of a medical treatment plan and an examination plan according to the second embodiment. The output screen 1900 can be displayed on the display screen of the information terminal 110, for example.

[0137] The output screen 1900 is a screen showing recommended contents of medical treatment plans and examination plans, specifically, a screen showing contents of examination items (estimated) and treatment methods (estimated) in the process from "suspected" to "complete cure" of a specific disease. The output screen 1900 is composed of, for example, an initial visit information display field 1901, a re-examination information display field 1902, a follow-up observation information display field I_1903, a follow-up observation information display field II_1904, an update patient information button 1905, an update plan button 1906, and an output examination order button 1907.

[0138] The patient information update button 1905 is a button that is pressed when, for example, the target patient has changed and the patient information of the target patient is to be input.

[0139] An update plan button 1906 is a button that is pressed when re-executing the process of creating (predicting) a medical treatment plan.

[0140] The test order output button 1907 is a button that is pressed when ordering a test to an examination room (not shown) via the test ordering system 1110.

[0141] The output screen 1900 in FIG. 19 shows that when disease A is suspected at the initial visit, the treatment plan and test plan are predicted as follows: (i) at the follow-up visit, a definitive diagnosis of disease A is made, and tests 4, 5, and 6 are recommended, as well as prescriptions a, b, and c as treatment methods; (ii) during treatment, tests 3, 4, and 5 are recommended, as well as prescriptions a, b, and c as treatment methods; and (iii) by continuing the tests and treatments shown in follow-up I, the patient will be completely cured.

[0142] <Example of GUI Configuration Displayed on the Display Screen of Information Terminal 110> 20 is a diagram showing a configuration example of a test order GUI 2000 displayed on the display screen of the information terminal 110 according to the second embodiment. Through the test order GUI 2000, a patient management policy is selected and displayed (selected by a doctor), and disease determination results and test items are output.

[0143] The test order GUI 2000 includes, as components, for example, a management policy selection / display section 1001, a disease probability etc. display section 2002, and a disease test item display section 2003. The difference from the test order GUI 1000 according to the first embodiment is that the disease test item display section 2003 of the test order GUI 2000 is configured to enable comparison of patient costs and hospital charges between now and one month from now (future).

[0144] The management policy selection / display unit 2001 displays the patient management policies of the target patient that the doctor can select when the computer 100 of the patient care plan determination system 10 is provided with recommended test information, and the currently selected patient management policy.

[0145] The disease probability display section 2002 provides information on the disease name, disease probability, and disease risk (degree of risk of the disease itself) obtained by the examination and disease prediction process (prediction simulation). For example, it shows that disease A obtained as a result of the prediction process is 70% likely to be true, and that the risk of the disease is medium.

[0146] The disease test item display unit 2003 displays, for example, test items for diseases with a disease probability higher than a predetermined value or test items for diseases with a high risk, patient costs indicating the expenses paid by the patient for the test, and hospital costs indicating the expenses borne by the hospital. In the example of Fig. 10, test items for disease A with a probability of 70% and disease C with a high risk are displayed. For disease A with the highest probability, it is shown that the patient costs will be the same now and one month from now, while the hospital costs will be zero one month from now. In contrast, for disease C with a low probability but the highest risk, it is shown that there will be no change in patient costs now and one month from now, and the hospital costs will be zero.

[0147] (3) Summary (i) The patient care plan determination system 10 according to the present embodiment is a system that uses a trained model to estimate future test item candidates and their values ​​(test values) from the patient management policy, patient information, and information on test items and test results. As shown in FIG. 1 and FIG. 2, the patient care plan determination system 10 can be configured by a computer 100, an information terminal 110, and a medical information server (external recording device) 111. The computer 100 acquires predetermined information from the medical information server and the information terminal, and performs processing to determine test item candidates related to the disease of the target patient. Specifically, the computer estimates the test values ​​of future test item candidates of the target patient by applying the patient attribute information of the target patient, the test results of the target patient, and the patient management policy of the target patient to a trained model (test plan trained model) 310 for calculating the test value for each test item, which is a test value indicating information that can identify a disease according to the patient management policy, and outputs the test item candidates and the test values ​​of the target patient. The output destination can be, for example, an information terminal 110 operated by a user (such as a doctor). In addition, the candidate test items can be those with the highest test value. By doing so, it becomes possible to recommend tests with high validity according to the results at each timing of examination, diagnosis, treatment, and follow-up in the overall workflow of patient management.

[0148] (ii) The trained model 310 includes, as training parameters, a parameter for estimating a probability distribution in the test results for each test item, a parameter for restoring the test results from the estimated probability distribution, and a parameter for estimating the test value. The outline of the training parameter generation process is as follows. That is, the computer 100 integrates patient information of multiple patients (training sample data) to extract features, and estimates the probability distribution in the test results of multiple patients for each test item. Then, the computer 100 restores the test results based on the estimated probability distribution to generate restored test results, and estimates the test value based on the restored test results. The computer 100 also estimates a training error, which is the error between the test value obtained from the actual test results and the estimated test value, and when the training error has converged, sets the training parameter that provides the training error to the training parameter of the trained model 310. In this way, it is possible to construct the trained model 310 for calculating future test item candidates and test value information indicating their validity by applying information on the patient to be determined (patient attribute information, the test results of the target patient, and the patient management policy of the target patient). The convergence of the learning error can be judged using a threshold. When making the judgment, it is necessary to use an appropriate threshold. Therefore, the computer 100 estimates a patient management policy using the diagnosis probability of a specific disease estimated based on the reconstruction test result and information on the risk of the disease, and estimates a threshold for judging whether the learning error has converged or not based on the comparison result between the estimated patient management policy and the actual management policy. This makes it possible to set the threshold used for judging the convergence of the learning error to an appropriate value.

[0149] (iii) The computer 100 estimates the medical fee points based on the candidate test items, the patient management policy of the target patient, and the medical fee calculation conditions obtained from the medical information server, and estimates the cost to be borne by the hospital from the medical fee points. In this way, it becomes possible to determine future test items taking into account cost-effectiveness.

[0150] (iv) The second embodiment describes a patient care plan determination system 10 that provides not only an examination plan but also a medical treatment plan. Specifically, the computer 100 acquires from the medical information server 111 patient examination and medical treatment history information including the patient management policy of the target patient in the past (time step tk) and at the present time (time step t), the medical treatment method information of the target patient in the past and at the present time, the patient attribute information of the target patient, the examination item and examination result information of the target patient in the past and at the present time, and the patient disease information of the target patient in the past and at the present time, and applies it to a trained model (trained medical treatment plan model) 1303 for predicting the medical treatment plan of the target patient from the patient examination and medical treatment history information of the target patient, thereby estimating a medical treatment plan including the patient management policy and medical treatment method in the future (from time step t+1 to time step t+N) of the target patient. In this way, in addition to an appropriate examination, the doctor can propose an appropriate treatment method at each future time point to the patient (even if he / she has little experience). Furthermore, the computer 100 estimates multiple patterns of medical treatment plans for the target patient at each future time point (time step t+1 to time step t+N) (see FIG. 15). The information terminal 110 then displays the information on the medical treatment plan for the target patient received from the computer 100 on the display screen in chronological order. For example, the information on the medical treatment plan is displayed on the display screen in the order of first visit, follow-up visit, and follow-up observation.

[0151] (v) An overview of the process of generating learning parameters of the trained model 1303 is as follows: That is, the computer 100 acquires from the medical information server 111 patient management policies (samples for generating trained models) before a predetermined time point (from time step tk to time step t), medical treatment method information before a predetermined time point for multiple patients, patient attribute information for multiple patients, test item and test result information before a predetermined time point for multiple patients, and patient disease information before a predetermined time point for multiple patients, as patient examination and medical treatment history information for generating trained models, extracts features of the information, performs machine learning based on the features, and estimates patient management policies, medical treatment methods, and patient disease information after the predetermined time point (from time step t+1 to time step t+N). In addition, the computer 100 estimates a learning error for the medical treatment plan by comparing the actual patient management policy, medical treatment method, and patient disease information of multiple patients after a predetermined time point with the estimated patient management policy, medical treatment method, and patient disease information after a predetermined time point, and if the learning error has converged, sets the learning parameters providing the learning error for the medical treatment plan as the learning parameters of the trained model 1303. If the learning error has not converged, the computer 100 updates the learning parameters providing the learning error for the medical treatment plan using an optimization function, and executes a process of re-estimating using the updated learning parameters. In this way, a trained model 1303 can be constructed for appropriately proposing a future medical treatment plan by applying information of the patient for whom the medical treatment plan is to be created (patient attribute information, the test results of the target patient, and the patient management policy of the target patient).

[0152] (vi) The functions of the present embodiment and each example can also be realized by software program code. In this case, a storage medium on which the program code is recorded is provided to a system or device, and the computer (or CPU or MPU) of the system or device reads the program code stored in the storage medium. In this case, the program code itself read from the storage medium realizes the functions of the above-mentioned embodiment, and the program code itself and the storage medium on which it is stored constitute the present disclosure. Examples of storage media for supplying such program code include flexible disks, CD-ROMs, DVD-ROMs, hard disks, optical disks, magneto-optical disks, CD-Rs, magnetic tapes, non-volatile memory cards, and ROMs.

[0153] Also, an operating system (OS) running on a computer may perform all or a part of the actual processing based on the instructions of the program code, and the functions of the above-mentioned embodiments may be realized by the processing. Furthermore, after the program code read from a storage medium is written into a memory on a computer, a CPU of the computer may perform all or a part of the actual processing based on the instructions of the program code, and the functions of the above-mentioned embodiments may be realized by the processing.

[0154] Furthermore, the program code of the software that realizes the functions of the embodiments and each example may be distributed via a network and stored in a storage means such as a hard disk or memory of the system or device, or in a storage medium such as a CD-RW or CD-R, so that when used, the computer (or CPU or MPU) of the system or device reads out and executes the program code stored in the storage means or storage medium.

[0155] The processes and techniques described herein are not inherently related to any particular device, and may be implemented by a combination of components. Various types of general-purpose devices may also be added. A dedicated device may be constructed to execute the functions of the present embodiment and each example. Various functions may also be formed by appropriately combining multiple components disclosed in the present embodiment and each example. For example, some components may be deleted from all the components shown in the embodiment and each example, or components across different examples may be appropriately combined.

[0156] In this disclosure, specific examples are described, but they are in all respects for the purpose of explanation (understanding the technology of the present disclosure) and not for the purpose of limitation. It is understood by those skilled in the art that there are many combinations of hardware, software, and firmware suitable for implementing the technology of the present disclosure. For example, the described software can be implemented in a wide range of program or script languages, such as assembler, C / C++, perl, Shell, PHP, Java (registered trademark), etc.

[0157] Furthermore, in the above-described embodiment, the control lines and information lines are those that are considered necessary for the explanation, and not all of the control lines and information lines in the product are necessarily shown. All of the components may be connected to each other.

[0158] In addition, other implementations of the present disclosure will become apparent to those skilled in the art from consideration of the present embodiments and examples. The specification and examples are exemplary only, with the scope and spirit of the present disclosure being indicated by the following claims. [Explanation of symbols]

[0159] 10 Patient Care Planning Decision System 100 calculator 101 Processor 102 Main storage 103 Secondary storage device 104 Network Adapter 105 Input Device 106 Output Device 109 Network 110 Information terminal 111 External recording device, medical information server 1011 Patient Information Acquisition Department 1012 Inspection Planning Department 1101 Test order output unit 1100 Medical Planning Department 1110 Test ordering system

Claims

1. A patient care plan determination system that determines at least test item candidates related to a disease of a target patient, a medical information server that stores at least patient attribute information and test results of a plurality of patients; an information terminal that provides a patient management policy indicating the classification of the target patient; a computer that acquires predetermined information from the medical information server and the information terminal and determines candidate test items related to the disease of the subject patient; The computer A process of acquiring patient attribute information of the target patient and a patient management policy of the target patient from the medical information server; A test value indicating information capable of identifying a disease according to the patient management policy, wherein the test value of future candidate test items for the target patient is estimated by applying the patient attribute information of the target patient, the test results of the target patient, and the patient management policy of the target patient to a test plan trained model for calculating the test value for each test item; a process of outputting the candidate test items and the test value of the target patient; A patient care planning decision system that implements

2. In claim 1, The computer is a patient care plan determination system that selects the test item with the greatest test value as the candidate test item for the target patient.

3. In claim 1, A patient care plan determination system, wherein the test plan trained model includes, as learning parameters, parameters for estimating a probability distribution of test results for each test item, parameters for restoring test results from the estimated probability distribution, and parameters for estimating the test value.

4. In claim 3, The computer A process of acquiring patient information of a plurality of patients from the medical information server, the patient information including attribute information of the patients, the test results, and the patient management policy; A process of integrating the patient information of the plurality of patients and extracting features; a process of estimating a probability distribution of the test results of the plurality of patients for each of the test items; a process of reconstructing the inspection result based on the estimated probability distribution to generate a reconstructed inspection result; A process of estimating the inspection value based on the restoration inspection result; A process of estimating a learning error, which is an error between an inspection value obtained from the actual inspection result and the estimated inspection value; When the learning error has converged, a process of setting a learning parameter that provides the learning error as a learning parameter of the inspection plan trained model; A patient care planning decision system that implements

5. In claim 4, The computer estimates a patient management policy using the diagnostic probability of a specified disease estimated based on the reconstruction test results and information on the risk of the disease, and estimates a threshold for determining whether the learning error has converged based on the results of a comparison between the estimated patient management policy and the actual management policy.

6. In claim 1, The computer further A process of acquiring medical receipt information including information on medical fee calculation conditions from the medical information server; A process of estimating medical fee points based on the candidate test items, the patient management policy of the target patient, and the medical fee calculation conditions; A process of estimating hospital costs from the medical fee points; A patient care planning decision system that implements

7. In claim 1, The computer further A process of acquiring patient examination and medical treatment history information from the medical information server, the patient examination and medical treatment history information including the past and current patient management policy of the target patient, the past and current medical treatment method information of the target patient, the patient attribute information of the target patient, the past and current test item and test result information of the target patient, and the past and current patient disease information of the target patient; A process of estimating a medical treatment plan including a future patient management policy and medical treatment method for the target patient by applying the patient's past and current examination and medical treatment history information of the target patient to a medical treatment plan trained model for predicting the medical treatment plan for the target patient from the patient's examination and medical treatment history information; A process of outputting a medical treatment plan for the target patient; A patient care planning decision system that implements

8. In claim 7, further comprising: the computer executes a process of extracting features of the patient examination and medical history information; A patient care plan determination system that applies the features of the patient examination and medical history information to the medical plan trained model.

9. In claim 7, The computer estimates multiple patterns of the treatment plan at each future point in time for the target patient.

10. In claim 7, The information terminal receives information on the treatment plan of the target patient from the computer and displays the contents of the treatment plan in chronological order on a display screen.

11. In claim 10, The information terminal displays the contents of the medical treatment plan on the display screen in the order of initial visit, follow-up visit, and follow-up observation.

12. In claim 7, The computer A process of acquiring, from the medical information server, patient management policies for a plurality of patients prior to a predetermined time point, medical treatment method information for the plurality of patients prior to the predetermined time point, patient attribute information for the plurality of patients, test item and test result information for the plurality of patients prior to the predetermined time point, and patient disease information for the plurality of patients prior to the predetermined time point as patient examination and medical treatment history information for generating a trained model; A process of integrating the patient examination and medical history information for generating the trained model and extracting features for generating the trained model; A process of performing machine learning based on the trained model generation features and estimating a patient management policy, a medical treatment method, and patient disease information after the predetermined time point; a process of estimating a learning error for a medical treatment plan by comparing actual patient management policies, medical treatment methods, and patient disease information of the plurality of patients after the predetermined time point with the estimated patient management policies, medical treatment methods, and patient disease information after the predetermined time point; When the learning error for the treatment plan has converged, a process of setting the learning parameters that provide the learning error for the treatment plan as learning parameters of the treatment plan trained model; A patient care plan decision system that generates the treatment plan trained model by executing

13. In claim 12, A patient care plan determination system, wherein if the learning error for the treatment plan has not converged, the computer updates the learning parameters that provide the learning error for the treatment plan using an optimization function, and performs the estimation process again using the updated learning parameters.

14. A patient care plan determination method for determining at least test item candidates related to a disease of a target patient, a computer that acquires predetermined information from a medical information server that stores at least patient attribute information and test results of a plurality of patients and an information terminal that provides a patient management policy indicating the classification of the target patient, and determines test item candidates related to the disease of the target patient, and acquires the patient attribute information of the target patient and the patient management policy of the target patient from the medical information server; The computer estimates the test value of future test item candidates for the target patient by applying the patient attribute information of the target patient, the test results of the target patient, and the patient management policy of the target patient to a test plan trained model for calculating the test value for each test item, the test value indicating information that can identify a disease according to the patient management policy; The computer outputs candidate test items and the test values ​​for the target patient.

Images