Medical determination support device and medical determination support program

Abstract

To calculate an analysis result in consideration of the importance of each computing unit when calculating an analysis result on an examination object region of a subject on the basis of a plurality of kinds of medical information by using a plurality of computing units corresponding to each of the plurality of kinds of medical information.SOLUTION: An analysis unit 38 outputs an analysis result by using a computing unit 26 group for performing analysis on the basis of a corresponding kind of medical information, in correspondence with each of a plurality of kinds of medical information acquired by a medical information processing unit 36. A memory 22 stores importance information 28, which is the information indicating importance of each kind of the medical information when outputting an analysis result on an analysis item. Prior to the analysis, the analysis unit 38 refers to the importance information 28, specifies importance of each medical information (in other words, each computing unit 26), and calculates the analysis result on the basis of partial analysis information output by each computing unit 26 when each medical information is input to the corresponding computing unit 26, and the specified importance of each of the computing unit 26.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] This specification discloses an improvement to a medical decision support device and a medical decision support program. [Background technology]

[0002] Conventionally, doctors make judgments (e.g., diagnoses) on a target part of a subject based on multiple types of medical information on the subject acquired by one or more medical devices. In this specification, judgment on a target part means judging a progress state and prognosis (prediction of prevention, prediction of progress, risk of onset (including recurrence), etc.) of a certain lesion on the target part. Here, a doctor may perform an analysis along one or more analysis items based on multiple types of medical information, and judge the target part based on the analysis results of one or more analysis items.

[0003] For example, when the part to be assessed is the heart and the assessment content is a assessment regarding ischemic heart disease, the doctor performs an analysis based on multiple types of medical information about the heart, which is the part to be assessed (e.g., ultrasound images, CT (Computed Tomography) images, nuclear magnetic resonance images, etc.), along one or more analysis items related to the heart, and makes a assessment regarding ischemic heart disease based on the analysis results.

[0004] Conventionally, a device for supporting diagnosis of a subject using a plurality of medical information (a plurality of image data) has been proposed. For example, Patent Document 1 discloses an image diagnosis support system that processes a plurality of image data of a subject's test site acquired by different medical devices (e.g., mammography and ultrasound diagnostic devices) to generate judgment data of the test site as to whether the test site is benign or malignant, and generates a probability value of malignancy of each detected lesion based on characteristic data of the detected lesion, and calculates a total probability value by weighting according to a combination of the contents of signals indicating the presence or absence of the lesion or by referring to a look-up table (LUT). Patent Document 2 discloses an image diagnosis support technology that supports disease discrimination using a plurality of types of measurement values ​​acquired by a medical image acquisition device. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent No. 4633298 specification [Patent Document 2] JP 2020-192068 A Summary of the Invention [Problem to be solved by the invention]

[0006] Consider performing an analysis based on a plurality of types of medical information to obtain an analysis result on a target part of a subject using a computing unit corresponding to each of the plurality of types of medical information. The computing unit may be, but is not limited to, a learning model or a LUT (Look Up Table). Here, the importance of each of the plurality of types of medical information when obtaining the analysis result may differ from each other. That is, the importance of the computing units corresponding to each of the plurality of types of medical information may differ from each other. For example, when obtaining an analysis result on the cardiac function of the heart as a target part of the examination, if an ultrasound image of the heart at rest as medical information is more important than an ultrasound image of the heart at stress as medical information, the importance of the computing unit that processes the ultrasound image of the heart at rest is greater than that of the computing unit that processes the ultrasound image of the heart at stress.

[0007] The purpose of the medical decision support device disclosed in this specification is to use multiple calculators, each corresponding to multiple types of medical information, to calculate analysis results for the test area of ​​a subject based on multiple types of medical information, while taking into account the importance of each calculator. [Means for solving the problem]

[0008] The medical decision support device disclosed in this specification is a medical decision support device including an analysis unit that outputs an analysis result regarding a test target part of a subject based on multiple types of medical information regarding the subject, the analysis unit outputting the analysis result using a group of calculators that respectively correspond to the types of medical information and output partial analysis information for calculating the analysis result based on the corresponding types of medical information, the analysis unit identifying the importance of each of the calculators included in the group of calculators based on importance information indicating the importance of each type of medical information when outputting the analysis result, and calculating the analysis result based on the partial analysis information output by each of the calculators when the corresponding medical information is input and the identified importance of each of the calculators.

[0009] The importance information associates the importance of each type of medical information with the diagnostic status of the subject, and the analysis unit identifies the importance of each of the arithmetic units included in the group of arithmetic units based on the importance information and the current diagnostic status of the subject.

[0010] The analysis section may calculate the analysis result based on the identified importance of each of the computing units, without using some of the computing units in the group of computing units.

[0011] The analysis unit may have a plurality of groups of calculators corresponding to a plurality of analysis items related to the examination target area, and may identify the importance of each of the calculators included in the plurality of groups of calculators for each analysis item based on the importance information indicating the importance of each type of medical information, and may output a plurality of analysis results for the plurality of analysis items by calculating the analysis result for each analysis item based on the partial analysis information output by each of the calculators when corresponding medical information is input and the identified importance of each of the calculators.

[0012] The importance of the calculator for one of the analysis items may be different from the importance of a calculator corresponding to the same type of medical information as the calculator for another of the analysis items.

[0013] It is preferable to further include an overall judgment unit that judges the state of the inspection target area based on the output of the learning model for overall judgment when the multiple analysis results output by the analysis unit are input to the learning model for overall judgment, which has been trained to predict and output the state of the inspection target area based on the multiple analysis results for the multiple analysis items.

[0014] The part to be examined may be the heart, the analysis unit may output a disease state for each of a plurality of disease types related to the heart as the plurality of analysis results, and the comprehensive assessment unit may make a comprehensive assessment of the condition of the subject's heart based on the disease state for each of the plurality of disease types.

[0015] The medical information analysis device may further include a display control unit that displays, on a display unit, the medical information on which the analysis is based and the judgment result of the comprehensive judgment unit.

[0016] The display control unit may cause the display unit to display a screen for selecting either the plurality of pieces of medical information that were the basis of analysis by the analysis unit, or the judgment result of the comprehensive judgment unit.

[0017] The display control unit may display the medical information in a manner that enables the importance of the medical information to be determined based on the importance information.

[0018] Furthermore, the medical decision support program disclosed in the present specification causes a computer to function as an analysis unit that outputs an analysis result regarding a test target site of a subject based on multiple types of medical information regarding the subject, the analysis unit outputting the analysis result using a group of calculators that correspond to the types of medical information and are each trained to output partial analysis information for calculating the analysis result based on the corresponding types of medical information, and the analysis unit identifies the importance of each of the calculators included in the group of calculators based on importance information indicating the importance of each type of medical information when outputting the analysis result, and calculates the analysis result based on the partial analysis information output by each of the calculators when the corresponding medical information is input and the identified importance of each of the calculators. Effect of the Invention

[0019] According to the medical decision support device disclosed in this specification, when multiple calculators each corresponding to multiple types of medical information are used to calculate analysis results for the test area of ​​a subject based on multiple types of medical information, the analysis results can be calculated taking into account the importance of each calculator. [Brief description of the drawings]

[0020] [Figure 1] 1 is a schematic diagram illustrating the configuration of a medical decision support system according to an embodiment of the present invention. [Diagram 2]1 is a schematic diagram illustrating the configuration of a medical decision support device according to an embodiment of the present invention. [Diagram 3] FIG. 2 is a functional block diagram of a medical information processing unit. [Figure 4] FIG. 2 is a functional block diagram of an analysis unit. [Diagram 5] FIG. 13 is a conceptual diagram showing a group of computing units for each combination of type of medical information and analysis item. [Figure 6] FIG. 2 is a conceptual diagram showing the contents of importance information. [Figure 7] FIG. 2 is a conceptual diagram showing an example of a calculator used in processing to obtain analysis results for each analysis item. [Figure 8] FIG. 13 is a conceptual diagram showing how a learning device is formed by combining structural information and learning information. [Figure 9] FIG. 11 is a diagram showing a first display example of the processing result of the medical decision support device. [Figure 10] FIG. 11 is a diagram showing a second display example of the processing result of the medical decision support device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] 1 is a schematic diagram of a medical decision support system 10 according to this embodiment. The medical decision support system 10 includes one or more medical devices 12, a user terminal 14, and a medical decision support device 16. In this embodiment, the medical decision support system 10 includes multiple medical devices 12, but the number of medical devices 12 may be one. The multiple medical devices 12, the user terminal 14, and the medical decision support device 16 are communicatively connected via a communication line 18 including a WAN (Wide Area Network) or a LAN (Local Area Network).

[0022] The medical devices 12 are used to obtain medical information about the subject (particularly the part to be judged). The medical decision support system 10 has a plurality of medical devices 12 of different types. For example, the plurality of medical devices 12 are an ultrasound diagnostic device, an X-ray CT device, an MRI (Magnetic Resonance Imaging) device, a nuclear medicine examination device, a blood test device, a genetic test device, an electrocardiogram, or a cardiac catheter examination device. The different types of medical devices 12 obtain different medical information about the part to be judged. For example, an ultrasound diagnostic device obtains ultrasound image data of the part to be judged, an X-ray CT device obtains CT image data of the part to be judged, and an MRI device obtains a nuclear magnetic resonance image of the part to be judged. The medical information obtained by each medical device 12 is sent to the medical decision support device 16 via a communication line 18.

[0023] The user terminal 14 is a terminal used by a user of the medical decision support system 10. The user terminal 14 may be, for example, a personal computer or a tablet terminal. The user terminal 14 includes a processor, a memory, a communication interface, an input interface, and a display.

[0024] 2 is a schematic diagram of the configuration of the medical decision support device 16. In this embodiment, the medical decision support device 16 is a server, but the medical decision support device 16 may be any device as long as it can perform the functions described below. In addition, the medical decision support device 16 may realize each of the functions described below by cooperation between multiple physically separated devices.

[0025] The communication interface 20 is composed of, for example, a network adapter. The communication interface 20 exerts a function of communicating with other devices (medical devices 12 and user terminal 14) via a communication line 18. Particularly, in this embodiment, the communication interface 20 receives medical information from a plurality of medical devices 12. The communication interface 20 also receives instructions from the user terminal 14 and transmits information indicating a processing result to the user terminal 14.

[0026] The memory 22 includes a hard disk drive (HDD), a solid state drive (SSD), an embedded multi media card (eMMC), a read only memory (ROM), or a random access memory (RAM). A medical decision support program for operating each unit of the medical decision support device 16 is stored in the memory 22. The medical decision support program can also be stored in a computer-readable non-transitory storage medium such as a universal serial bus (USB) memory or a CD-ROM. The medical decision support device 16 can read and execute the medical decision support program from such a storage medium.

[0027] 2, the memory 22 stores a medical information DB (DataBase) 24, a calculator 26, importance information 28, a learning model for comprehensive judgment 30, and a prognosis information DB 32. The medical information DB 24 is a database in which medical information received from each medical device 12 is accumulated and stored. The medical information transmitted from each medical device 12 is added with attribute information representing the attributes of the medical information, such as a device ID for uniquely identifying the medical device 12 that acquired it, a subject ID for uniquely identifying the subject, judgment target part information indicating the judgment target part, and operating condition information indicating the operating conditions (e.g., operating mode, etc.) of the medical device 12 when the medical information was acquired. In the medical information DB 24, the medical information and the attribute information are stored in association with each other. The medical information stored in the medical information DB 24 also includes information (e.g., electronic medical records, etc.) obtained by a doctor by interviewing the subject.

[0028] The calculator 26, the importance information 28, the comprehensive judgment learning model 30, and the prognosis information DB 32 will be described later.

[0029] The processor 34 includes at least one of a general-purpose processor (e.g., a CPU (Central Processing Unit)) and a dedicated processor (e.g., a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a programmable logic device). The processor 34 may not be a single processing device, but may be a device formed by the cooperation of multiple processing devices located at physically separate locations.

[0030] 2, the processor 34 performs the functions of a medical information processing unit 36, an analysis unit 38, a comprehensive judgment unit 40, and a display control unit 42 in accordance with a medical judgment support program stored in the memory 22. As will be described in detail later, the processor 34 judges the state of the judgment target site of the subject by performing the above-mentioned functions.

[0031] First, an overview of the determination process of the state of the subject's determination target part as the processor 34 as a whole will be described. The processor 34 receives, from the user terminal 14, the subject to be determined, the subject's determination target part, and determination instruction information including the determination content. In this embodiment, the determination target part is the heart, and the determination content is to determine the subject's heart state, particularly the state related to ischemic heart disease (more specifically, the lesion state of ischemic heart disease and the prognosis (prediction of prevention, prediction of progression, risk of recurrence, etc.)). Of course, the determination target part and the determination content are not limited to this. When the processor 34 receives the determination instruction information, it performs an analysis on multiple different analysis items for determining the determination content (determination of the state related to ischemic heart disease in this embodiment) based on the medical information on the subject stored in the medical information DB 24, and obtains multiple analysis results for the multiple analysis items. In this embodiment, the multiple analysis items for determining the state related to ischemic heart disease are coronary artery stenosis, ischemic myocardium, myocardial viability, cardiac function, and ischemic heart disease risk. Thereafter, the processor 34 obtains a judgment result regarding the judgment contents by comprehensively judging the multiple analysis results for the multiple analysis items. Hereinafter, the details of each function performed by the processor 34 will be described.

[0032] 3 is a functional block diagram of the medical information processing unit 36. The medical information processing unit 36 ​​performs a process of providing information required for analysis to an analysis unit 38, which will be described later, for each analysis item for making a judgment on the judgment content indicated by the judgment instruction information.

[0033] First, the medical information processor 36 identifies multiple different analysis items for making a judgment on the judgment content indicated by the received judgment instruction information. For example, by storing analysis item information in which the judgment content and multiple analysis items are associated in advance in the memory 22, the medical information processor 36 can identify multiple analysis items corresponding to the judgment content by referring to the analysis item information.

[0034] As shown in FIG. 3, the medical information processing unit 36 ​​has functions as a detection unit 36a and a measurement unit 36b.

[0035] The detector 36a extracts information necessary for analysis of the specified multiple analysis items related to the subject indicated by the received determination information from the medical information DB 24. The information necessary for analysis of each analysis item may also be associated in the above-mentioned analysis item information, and the detector 36a can identify the information necessary for analysis of each analysis item by referring to the analysis item information and the attribute information associated with the medical information. For convenience, in FIG. 3, the detectors 36a (coronary artery stenosis information detector 36a-1, ischemic myocardial information detector 36a-2, cardiac function information detector 36a-3, and ischemic heart disease risk information detector 36a-4) that extract necessary information are shown separately for each analysis item.

[0036] The coronary artery stenosis information detection unit 36a-1 extracts information necessary for analysis of the analysis item "coronary artery stenosis" from the medical information DB 24. For example, the coronary artery stenosis information detection unit 36a-1 extracts from the medical information DB 24 the following information: Resting echocardiograms, stress echocardiograms, or intravascular ultrasound images (IVUS (IntraVascular UltraSound)) acquired by an ultrasound diagnostic device, Coronary CT angiography (CCTA) or fractional flow reserve (FFR)-CT images obtained by an X-ray CT scanner, A coronary artery MRA (Magnetic Resonance Angiography) image or a stress myocardial perfusion MRI image acquired by an MRI device, or Results of cardiac nuclear medicine examinations obtained by nuclear medicine examination devices, Extract at least one of the following:

[0037] The ischemic myocardium information detection unit 36a-2 extracts information necessary for the analysis of the analysis items "ischemic myocardium" and "myocardium viability" from the medical information DB 24. For example, the ischemic myocardium information detection unit 36a-2 extracts from the medical information DB 24 the following information: A resting echocardiogram, a stress echocardiogram, a myocardial contrast echogram, or a myocardial strain echogram acquired by an ultrasound diagnostic device; Dual energy CT images, photon counting CT images, delayed contrast CT images, coronary CT angiography (CCTA (Coronary CT Angiography)), FFR (Fractional Flow Reserve)-CT images, or stress perfusion CT images acquired by X-ray CT devices. A cine MRI image, delayed contrast MRI image, or stress myocardial perfusion MRI image acquired by an MRI device, or Results of cardiac nuclear medicine examinations obtained by nuclear medicine examination devices, Extract at least one of the following:

[0038] The cardiac function information detection unit 36a-3 extracts information necessary for the analysis of the analysis item "cardiac function" from the medical information DB 24. For example, the cardiac function information detection unit 36a-3 extracts from the medical information DB 24 the following information: Echocardiograms acquired by ultrasound diagnostic equipment, Cardiac CT images or FFR (Fractional Flow Reserve)-CT images acquired by an X-ray CT device, Cardiac MRI images or delayed contrast MRI images acquired by an MRI device, or Results of cardiac nuclear medicine examinations obtained by nuclear medicine examination devices, Extract at least one of the following:

[0039] The ischemic heart disease risk information detection unit 36a-4 extracts information necessary for analyzing the analysis item "ischemic heart disease risk" from the medical information DB 24. For example, the ischemic heart disease risk information detection unit 36a-4 extracts from the medical information DB 24 biochemical information and genetic factor information acquired from a blood tester, a genetic tester, etc., and information obtained by a doctor interviewing a subject (such as electronic medical records).

[0040] As described above, the detector 36a may acquire multiple types of medical information acquired from multiple types of medical devices 12 as information required for analysis of each analysis item.

[0041] The measuring unit 36b measures parameters necessary for analysis of the identified multiple analysis items based on the information detected by the detecting unit 36a. The parameters necessary for analysis of each analysis item may also be associated in the above-mentioned analysis item information, and the measuring unit 36b can identify the parameters necessary for analysis of each analysis item by referring to the analysis item information. For convenience, in FIG. 3, the measuring units 36b (coronary artery stenosis measuring unit 36b-1, ischemic myocardium measuring unit 36b-2, myocardial viability measuring unit 36b-3, and cardiac function measuring unit 36b-4) that measure the necessary parameters are shown separately for each analysis item.

[0042] The coronary artery stenosis measuring unit 36b-1 measures parameters necessary for analyzing the analysis item "coronary artery stenosis" based on the information extracted by the coronary artery stenosis information detecting unit 36a-1. For example, the coronary artery stenosis measuring unit 36b-1 measures at least one of the following: left ventricular ejection fraction (LVEF), left ventricular end-systolic volume, left ventricular longitudinal systolic function index (GLS (Global Longitudinal Strain), S', MAPSE (Mitral Annular Plane Systolic Excursion)), coronary artery calcification score (CACS (Coronary Artery Calcium Score)), napkin ring sign, lesion vascular diameter increase value, coronary flow reserve (CFR (Coronary Flow Reserve)), or fractional coronary flow reserve (FFR).

[0043] The ischemic myocardium measuring unit 36b-2 measures parameters required for the analysis of the analysis item "ischemic myocardium" based on the information extracted by the ischemic myocardium information detecting unit 36a-2 or the ischemic heart disease risk information detecting unit 36a-4. For example, the ischemic myocardium measuring unit 36b-2 measures at least one of parameters related to the subject's medical history or physical examination, parameters related to an exercise electrocardiogram test, parameters related to a Holter electrocardiogram, or parameters related to a cardiac catheter.

[0044] The myocardial viability measuring unit 36b-3 measures parameters required for the analysis of the analysis item "myocardial viability" based on the information extracted by the ischemic myocardial information detecting unit 36a-2. For example, the myocardial viability measuring unit 36b-3 measures at least one of parameters related to myocardial blood flow, wall motion, myocardial fat metabolism, contractile reserve, myocardial fibrosis, myocardial extracellular volume fraction (ECV) measurement, wall thickness, fractional coronary flow reserve (FFR), left ventricular output (LVEF), end systolic volume (ESV), end diastolic volume (EDV), myocardial weight, fatty degeneration, calcification, cardiac catheter, or electrocardiogram.

[0045] The cardiac function measuring unit 36b-4 measures parameters necessary for the analysis of the analysis item "cardiac function" based on the information extracted by the cardiac function information detecting unit 36a-3. For example, the cardiac function measuring unit 36b-4 measures at least one of parameters related to left ventricular output (LVEF), stroke volume or cardiac output, left ventricular wall motion score index (WMSI (Wall Motion Score Index)), left ventricular longitudinal systolic function index (GLS, S', MAPSE), left ventricular pressure increase rate, left ventricular inflow blood flow velocity waveform (maximum velocity ratio of early diastolic wave and atrial systolic wave or deceleration time of early diastolic wave), pulmonary venous blood flow velocity waveform, systolic positive wave, early diastolic positive wave, atrial systolic negative wave, maximum early diastolic mitral annular movement velocity, maximum tricuspid regurgitation blood flow velocity, systolic pulmonary artery pressure, right ventricular systolic function, right ventricular diastolic function, myocardial strain analysis, and distinction between ischemic and non-ischemic myocardial damage.

[0046] Note that the above parameters are known parameters, and the measurement method thereof can be a conventional method, so the details of the measurement method of each parameter will be omitted here. In this specification, the term "medical information" is a concept that includes information necessary for analysis of multiple analysis items detected by the detection unit 36a and parameters for multiple analysis items measured by the measurement unit 36b. Among the medical information, information necessary for analysis of multiple analysis items detected by the detection unit 36a and parameters for multiple analysis items measured by the measurement unit 36b are particularly referred to as "medical information acquired by the medical information processing unit 36".

[0047] 4 is a functional block diagram of the analysis unit 38. The analysis unit 38 performs analysis for a plurality of analysis items based on a plurality of types of medical information related to the subject (particularly the region to be determined) acquired by the medical information processing unit 36, and outputs a plurality of analysis results related to the region to be determined of the subject for the plurality of analysis items. In particular, in this embodiment, the analysis unit 38 outputs disease states related to a plurality of disease types related to the heart, which is the region to be determined, as a plurality of analysis results. For convenience, in FIG. 4, each analysis unit 38 (coronary artery stenosis analysis unit 38-1, ischemic myocardium analysis unit 38-2, myocardial viability analysis unit 38-3, cardiac function analysis unit 38-4, and ischemic heart disease risk analysis unit 38-5) that extracts necessary information is shown separately for each analysis item.

[0048] In this embodiment, each analysis unit 38 corresponds to each of the multiple types of medical information acquired by the medical information processing unit 36, and outputs the analysis result using a group of calculators 26 that perform analysis based on the corresponding type of medical information. In this specification, the information output by each calculator 26 is referred to as partial analysis information.

[0049] A detailed description will be given with reference to FIG. 5. FIG. 5 is a conceptual diagram showing a group of calculators 26 for each combination of types of medical information and analysis items. The group of calculators 26 shown in FIG. 5 is stored in the memory 22. In FIG. 5, a plurality of analysis items (coronary artery stenosis, ischemic myocardium, myocardial viability, and cardiac function) are arranged horizontally, and a plurality of types of medical information (resting echocardiogram, stress echocardiogram, CCTA, FFR-CT, coronary artery MRA, stress myocardial perfusion MRI, nuclear cardiac examination, and IVUS) are arranged vertically. In this embodiment, the plurality of types of medical information shown in FIG. 5 includes medical information acquired by a plurality of different medical devices 12 (for example, an ultrasound diagnostic device, an X-ray CT device, an MRI device, etc.). In particular, in this embodiment, the plurality of types of medical information shown in FIG. 5 includes a plurality of medical images acquired by a plurality of different medical devices 12.

[0050] 5 correspond to each analysis item and each type of medical information, and are different calculators 26. For example, calculator 26-1 that outputs partial analysis information on coronary artery stenosis based on resting echocardiogram, calculator 26-2 that outputs partial analysis information on ischemic myocardium based on resting echocardiogram, and calculator 26-3 that outputs partial analysis information on coronary artery stenosis based on stress echocardiogram are different calculators 26.

[0051] Here, in this embodiment, each calculator 26 is an analytical learning model. That is, each calculator 26, which is an analytical learning model, is trained in advance so as to predict and output partial analysis information for a corresponding analysis item with high accuracy when a corresponding type of medical information is input. The analytical learning model may be, for example, a neural network, but the analytical learning model may be any model as long as it exhibits the functions described below. In addition, each calculator 26 does not necessarily have to be a learning model. Each calculator 26 may be any model as long as it can output partial analysis information for a corresponding analysis item based on a corresponding type of medical information. For example, each calculator may be an LUT.

[0052] The coronary artery stenosis analysis unit 38-1 obtains a plurality of partial analysis information by inputting the corresponding types of medical information to a plurality of calculators 26 (a plurality of calculators 26 arranged vertically corresponding to the analysis item "coronary artery stenosis" in FIG. 5) corresponding to the analysis item "coronary artery stenosis". Then, the coronary artery stenosis analysis unit 38-1 calculates an analysis result for the analysis item "coronary artery stenosis" based on the plurality of partial analysis information output by the plurality of calculators 26.

[0053] The ischemic myocardium analysis unit 38-2 obtains a plurality of partial analysis information by inputting the corresponding types of medical information to a plurality of calculators 26 (a plurality of calculators 26 arranged vertically corresponding to the analysis item "ischemic myocardium" in FIG. 5) corresponding to the analysis item "ischemic myocardium". Then, the ischemic myocardium analysis unit 38-2 calculates the analysis result for the analysis item "ischemic myocardium" based on the plurality of partial analysis information output by the plurality of calculators 26.

[0054] The myocardial viability analysis unit 38-3 obtains a plurality of partial analysis information by inputting the corresponding types of medical information to a plurality of calculators 26 corresponding to the analysis item "myocardial viability" and each type of medical information (a plurality of calculators 26 arranged vertically corresponding to the analysis item "myocardial viability" in FIG. 5). Then, the myocardial viability analysis unit 38-3 calculates an analysis result for the analysis item "myocardial viability" based on the plurality of partial analysis information output by the plurality of calculators 26.

[0055] The cardiac function analysis unit 38-4 obtains a plurality of partial analysis information by inputting the corresponding types of medical information to a plurality of calculators 26 corresponding to the analysis item "cardiac function" and each type of medical information (a plurality of calculators 26 arranged vertically corresponding to the analysis item "cardiac function" in FIG. 5). Then, the cardiac function analysis unit 38-4 calculates the analysis result for the analysis item "cardiac function" based on the plurality of partial analysis information output by the plurality of calculators 26.

[0056] Note that FIG. 5 (as well as FIGS. 6 and 7 described below) does not show the calculator 26 used by the ischemic heart disease risk analysis unit 38-5 that outputs the analysis results related to the analysis item "ischemic heart disease risk". However, like the other analysis units 38, the ischemic heart disease risk analysis unit 38-5 also outputs the analysis results using the calculator (analytical learning model or LUT) 26. Specifically, the ischemic heart disease risk analysis unit 38-5 inputs the information extracted by the ischemic heart disease risk information detection unit 36a-4 to the calculator 26, thereby performing an analysis related to the analysis item "ischemic heart disease risk" and outputting the analysis results.

[0057] 6 is a conceptual diagram showing the contents of the importance information 28 stored in the memory 22. As described above, each analysis unit 38 calculates the analysis result for each analysis item based on multiple types of medical information, but the importance of each type of medical information differs for each analysis item. For example, when analyzing the analysis item "coronary artery stenosis," the medical information "FFR-CT" is more important than other types of medical information (i.e., it has a greater effect on the analysis result than other types of medical information), and the medical information "IVUS" is less important than other types of medical information (i.e., it has a smaller effect on the analysis result than other types of medical information).

[0058] The importance information 28 is information indicating the importance of each type of medical information when outputting the analysis result for each analysis item. In this embodiment, the importance information 28 indicates three levels of importance, "high," "medium," and "low," for each combination of analysis item and type of medical information. In FIG. 6, the combination shown in shaded areas (i.e., calculator 26a) indicates a "high" level of importance, the combination shown in white (i.e., calculator 26b) indicates a "medium" level of importance, and the combination shown in solid black (i.e., calculator 26c) indicates a "low" level of importance.

[0059] For example, for the analysis item "coronary artery stenosis," the importance of the medical information "resting echocardiogram," "stress echocardiogram," "CCTA," "coronary artery MRA," "stress myocardial perfusion MRI," and "nuclear cardiac examination" is "medium," the importance of the medical information "FFR-CT" is "high," and the importance of the medical information "IVUS" is "low." On the other hand, for the analysis item "myocardial viability," the importance of the medical information "stress echocardiogram" and "nuclear cardiac examination" is "medium," the importance of the medical information "resting echocardiogram" is "high," and the importance of the medical information "CCTA," "FFR-CT," "coronary artery MRA," "stress myocardial perfusion MRI," and "IVUS" is "low." In this way, the importance of a certain calculator 26 (herein referred to as the attention calculator) and the importance of other calculators 26 corresponding to the same type of medical information as the attention calculator for other analysis items may be different from each other. That is, for the medical information "echocardiogram at rest", the importance of the analysis item "coronary artery stenosis" is "medium", but the importance of the analysis item "myocardial viability" is "high".

[0060] Prior to analysis, the analysis unit 38 refers to the importance information 28 to identify the importance of each piece of medical information (in other words, each calculator 26). In this embodiment, since a plurality of analysis units 38 are provided according to the analysis items, the analysis unit 38 identifies the importance of the plurality of calculators 26 for each analysis item based on the importance information 28. Then, the analysis unit 38 calculates an analysis result for each analysis item based on the partial analysis information output by each calculator 26 when each piece of medical information is input to the corresponding calculator 26 and the identified importance of each calculator 26.

[0061] For example, when the importance information 28 is as shown in Fig. 6, the importance of the medical information "FFR-CT" is "high", so the coronary artery stenosis analysis unit 38-1 assigns a higher weight (at least higher than medical information with "medium" importance) to the partial analysis information output by the calculator 26 corresponding to the medical information "FFR-CT" when calculating the analysis result for the coronary artery stenosis, and calculates the analysis result for the analysis item "coronary artery stenosis" based on the partial analysis information output by the multiple calculators 26 corresponding to the multiple medical information. Also, the importance of the medical information "IVUS" is "low", so the coronary artery stenosis analysis unit 38-1 assigns a lower weight (at least lower than medical information with "medium" importance) to the partial analysis information output by the calculator 26 corresponding to the medical information "IVUS" when calculating the analysis result for the coronary artery stenosis, and calculates the analysis result for the analysis item "coronary artery stenosis" based on the partial analysis information output by the multiple calculators 26 corresponding to the multiple medical information. The other analysis sections 38 calculate the analysis results in the same manner.

[0062] Each analysis unit 38 may calculate the analysis result without using some of the calculators 26 of the group of calculators 26 based on the importance of each identified medical information (i.e., each calculator 26). For example, as shown in FIG. 7, each analysis unit 38 may calculate the analysis result without using medical information with a "low" importance. For example, the myocardial viability analysis unit 38-3 may calculate the analysis result for the analysis item "myocardial viability" using only the medical information "resting echocardiogram", "stress echocardiogram", and "nuclear cardiac examination". In this case, the myocardial viability analysis unit 38-3 does not use the calculators 26 corresponding to the medical information "CCTA", "FFR-CT", "coronary artery MRA", "stress myocardial perfusion MRI", and "IVUS" to calculate the analysis result.

[0063] Here, the importance of each type of medical information for each analysis item may vary depending on the diagnostic situation (in other words, the condition) of the subject (particularly the part to be determined). Therefore, in the importance information 28, the importance of each type of medical information may be associated with the diagnostic situation of the subject. In this case, each analysis unit 38 specifies the importance of each piece of medical information (in other words, each calculator 26) based on the importance information 28 and the diagnostic situation of the subject (particularly the part to be determined) prior to analysis.

[0064] In this embodiment, the analysis unit 38 calculates the analysis results using a plurality of calculators 26 corresponding to a plurality of analysis items and a plurality of medical information. Therefore, as shown in FIG. 5 and the like, it becomes necessary to prepare a large number of calculators 26. If the number of analysis items and types of medical information increases, the number of calculators 26 may become enormous. In particular, when the calculators 26 are analytical learning models, storing a large number of calculators 26 in advance in the memory 22 puts a strain on the storage capacity of the memory 22.

[0065] In order to address such a problem, in this embodiment, as shown in FIG. 8, the calculator 26, which is the analytical learning model, is divided into a structural model 50 and learning information 52 and stored in the memory 22. The structural model 50 is data that defines the structure of the analytical learning model (e.g., the number of layers of a neural network, the number of neurons in one layer, etc.). In other words, the structural model 50 is data that is defined by hyperparameters (parameters that are not adjusted by the learning process) of the analytical learning model. The memory 22 may store a plurality of types of structural models 50. The learning information 52 is parameters that are adjusted by the learning process (e.g., the weight of each node of a neural network and the bias of each neuron, etc.). The learning information 52 is information obtained by the learning process.

[0066] The analysis unit 38 combines the structural model 50 and the learning information 52 to form the calculator 26, which is an analytical learning model. For example, when it is desired to obtain partial analysis information based on the medical information "echocardiogram at rest" for the analysis item "coronary artery stenosis", the coronary artery stenosis analysis unit 38-1 combines the structural model 50 for the combination of the analysis item "coronary artery stenosis" and the medical information "echocardiogram at rest" with the learning information 52 for the combination to form the calculator 26. When it is desired to obtain partial analysis information based on the medical information "echocardiogram at rest" for the analysis item "ischemic myocardium", the ischemic myocardium analysis unit 38-2 combines the structural model 50 for the combination of the analysis item "ischemic myocardium" and the medical information "echocardiogram at rest" with the learning information 52 for the combination to form the calculator 26.

[0067] The learning information 52 is different for each combination of analysis items and types of medical information and therefore needs to be prepared separately, but the structural model 50 can be shared among a plurality of combinations of analysis items and types of medical information. Therefore, by storing the structural model 50 and the learning information 52 separately in the memory 22 and combining the structural model 50 and the learning information 52 to form the necessary calculator 26 when the analysis unit 38 performs calculations, it is possible to reduce the storage capacity of the memory 22, at least compared to the case where all calculators 26 are stored in the memory 22.

[0068] In the invention in which the analysis unit 38 specifies the importance of each piece of medical information based on the importance information 28 and calculates the analysis result based on the partial analysis information output by each calculator 26 while taking into account the specified importance, the analysis unit 38 does not necessarily have to analyze a plurality of analysis items. In other words, the invention can be applied even when the analysis unit 38 analyzes only one analysis item.

[0069] 2, the comprehensive judgment unit 40 judges the state of the judgment target part based on the analysis results for multiple analysis items related to the judgment target part obtained by the analysis unit 38. Specifically, the comprehensive judgment unit 40 judges the state of the judgment target part based on the output of the comprehensive judgment learning model 30 when the multiple analysis results output by the analysis unit 38 are input to the comprehensive judgment learning model 30 stored in the memory 22.

[0070] The learning model for comprehensive judgment 30 may be, for example, a neural network, but any model may be used as the learning model for comprehensive judgment 30 as long as it exhibits the functions described below. The learning model for comprehensive judgment 30 is previously trained to predict and output the state of the judgment target part based on the analysis results of multiple analysis items related to the judgment target part. The learning model for comprehensive judgment 30 is trained by a learning device, which is the medical judgment support device 16 or another device. For example, the learning device inputs the analysis results of multiple analysis items related to the judgment target part as learning data to the learning model for comprehensive judgment 30. The learning model for comprehensive judgment 30 outputs a prediction result (the state of the judgment target part) for the learning data. The learning device calculates the difference between the output of the learning model for comprehensive judgment 30 and the state of the judgment target part as teacher data, and adjusts the parameters of the learning model for comprehensive judgment 30 so that the difference becomes smaller. The learning process progresses by repeating such processing. The trained comprehensive assessment learning model 30 becomes able to predict and output with high accuracy the state of the assessment target part based on the analysis results of multiple analysis items related to the assessment target part.

[0071] As described above, in this embodiment, the part to be assessed is the heart, and the analysis unit 38 outputs the disease states for each of a number of disease types related to the heart as a number of analysis results, and therefore the learning model for comprehensive assessment 30 comprehensively assesses the condition of the heart, which is the part to be assessed, based on the disease states for each of the multiple disease types.

[0072] If the medical judgment support device 16 does not have the analysis unit 38, all of the medical information acquired by the medical information processing unit 36 ​​will be input to the learning model for comprehensive judgment 30. Even in this way, functionally, the learning model for comprehensive judgment 30 can predict and output the state of the judgment target part based on the medical information acquired by the medical information processing unit 36. However, if this is done, the number of inputs to the learning model for comprehensive judgment 30 will be enormous, and the amount of calculation for predicting the state of the judgment target part will also be enormous, so the scale of the learning model for comprehensive judgment 30 will also be enormous. However, in this embodiment, since processing by the analysis unit 38 is performed prior to the processing of the learning model for comprehensive judgment 30, the scale of the learning model for comprehensive judgment 30 can be reduced.

[0073] The comprehensive judgment unit 40 may judge the prognosis of the judgment target part based on the prognosis information DB32 stored in the memory 22 and the judged state of the judgment target part. The prognosis information DB32 is a database that accumulates and stores state change information indicating time-dependent changes in the state of the judgment target part (heart in this embodiment) of multiple subjects in the past, the history of treatment, etc. The comprehensive judgment unit 40 identifies state change information indicating the judged state of the judgment target part from the state change information stored in the prognosis information DB32, and judges the prognosis of the judgment target part based on the state of the judgment target part after the judged state indicated by the identified state change information. The prognosis of the judgment target part includes, for example, the prognosis in the case where no treatment is performed as is and the prognosis in the case where appropriate treatment (treatment included in the state change information) is performed.

[0074] The display control unit 42 performs processing to display the processing results of the medical decision support device 16 on the display unit. For example, the display control unit 42 displays the processing results of the medical decision support device 16 on the display of the user terminal 14. In particular, the display control unit 42 displays the medical information that was the basis of the analysis by the analysis unit 38 and the judgment result of the comprehensive judgment unit 40 on the display of the user terminal 14.

[0075] FIG. 9 is a diagram showing a first display example of the processing result of the medical decision support device 16. In the example of FIG. 9, an ultrasound image USI and a coronary CT contrast image CCTA as medical information on which the analysis by the analysis unit 38 is based, a report R as the judgment result of the overall judgment unit 40, and a menu area MA are displayed. The menu area MA is a GUI for the user to select or input an instruction for enlargement from the medical information on which the analysis by the analysis unit 38 is based and the judgment result of the overall judgment unit 40, and thumbnails and titles are displayed. The report R shows the condition of the subject's heart, including the presence of coronary artery calcification, the absence of unstable plaque, stenosis of the left anterior descending artery (LAD) of 0.78, and abnormality in stress myocardial perfusion. Furthermore, report R indicates that the severity risk, which is the prognosis predicted by the comprehensive assessment unit 40, is that the current incidence of infarction five years from now is 34 to 45%, and that the incidence of infarction after appropriate treatment is 8 to 25%.

[0076] FIG. 10 is a diagram showing a second display example of the processing result of the medical judgment support device 16. In the example of FIG. 10, the examination flow for obtaining the medical information on which the analysis by the analysis unit 38 is based, the diagnostic evidence (class, level, grade), and the report R' are displayed as the judgment result of the comprehensive judgment unit 40. In particular, the examination flow of the medical information related to the importance information 28 and the diagnostic evidence may be displayed in a manner that allows the importance of the medical information to be distinguished by highlighting or the like. Also, the diagnostic evidence may be sorted in the display order from top to bottom in descending order of importance. In the report R', blood pressure, blood sugar level, habitual smoking, presence or absence of lipid abnormality, etc. are shown as risk factors for onset. Furthermore, in the report R', the onset prediction predicted by the comprehensive judgment unit 40 shows that the current onset rate of infarction 10 years from now is 30 to 65%, and that the onset rate of infarction when appropriate prevention is performed is 13 to 30%.

[0077] Although the embodiment of the present invention has been described above, the present invention is not limited to the above embodiment, and various modifications are possible without departing from the spirit of the present invention. [Explanation of symbols]

[0078] 10 medical decision support system, 12 medical equipment, 14 user terminal, 16 medical decision support device, 18 communication line, 20 communication interface, 22 memory, 24 medical information DB, 26 calculator, 28 importance information, 30 learning model for overall decision, 32 prognosis information DB, 34 processor, 36 medical information processing unit, 38 analysis unit, 40 overall decision unit, 42 display control unit.

Claims

1. an analysis unit that outputs an analysis result regarding an examination target site of a subject based on a plurality of types of medical information regarding the subject, the analysis unit outputting the analysis result using a group of calculators that respectively correspond to the types of medical information and output partial analysis information for calculating the analysis result based on the corresponding types of medical information; A medical decision support device comprising: The analysis unit includes: identifying the importance of each of the computing units included in the computing unit group based on importance information indicating the importance of each type of the medical information when the analysis result is output; Calculating the analysis result based on the partial analysis information output by each of the computing units when corresponding medical information is input and the importance of each of the computing units identified. A medical decision support device comprising:

2. The importance information is a information relating to the importance of each type of medical information with respect to a diagnostic condition of a subject; The analysis unit includes: determining the importance of each of the computing units included in the computing unit group based on the importance information and a current diagnosis status of the subject; 2. The medical decision support device according to claim 1.

3. the analysis unit calculates the analysis result without using some of the computing units in the group of computing units based on the identified importance of each of the computing units.

3. The medical decision support device according to claim 1 or 2.

4. The analysis unit includes: a plurality of the computing unit groups corresponding to a plurality of analysis items related to the inspection target site; Identifying the importance of each of the computing units included in the plurality of computing unit groups based on the importance information indicating the importance of each type of medical information for each of the analysis items; outputting a plurality of analysis results for the plurality of analysis items by calculating the analysis results based on the partial analysis information output by each of the computing units when corresponding medical information is input for each of the analysis items and the importance of each of the computing units identified; 3. The medical decision support device according to claim 1 or 2.

5. the importance of the calculator for one of the analysis items is different from the importance of a calculator corresponding to the same type of medical information as the calculator for another of the analysis items; 5. The medical decision support device according to claim 4.

6. a comprehensive judgment unit that judges the state of the inspection target site based on an output of the comprehensive judgment learning model when the multiple analysis results output by the analysis unit are input to the comprehensive judgment learning model that has been trained to predict and output the state of the inspection target site based on the multiple analysis results for the multiple analysis items; Further comprising:

5. The medical decision support device according to claim 4.

7. the examination target site is the heart, the analysis unit outputs, as the plurality of analysis results, disease states for each of a plurality of heart-related disease types; The comprehensive assessment unit comprehensively assesses a cardiac condition of the subject based on a disease state related to each of a plurality of disease types.

7. The medical decision support device according to claim 6.

8. a display control unit that causes a display unit to display the medical information that is the basis of the analysis by the analysis unit and the judgment result of the comprehensive judgment unit; The medical decision support device according to claim 6 or 7, further comprising:

9. The display control unit causes the display unit to display a screen for selecting either the plurality of pieces of medical information that are the basis of the analysis by the analysis unit or the judgment result of the comprehensive judgment unit.

9. The medical decision support device according to claim 8.

10. The display control unit displays the medical information in a manner that allows the importance of the medical information to be determined based on the importance information.

9. The medical decision support device according to claim 8.

11. Computer, an analysis unit that outputs an analysis result regarding an examination target site of a subject based on a plurality of types of medical information regarding the subject, the analysis unit outputting the analysis result using a group of calculators that correspond to the types of medical information, respectively, and that are each trained to output partial analysis information for calculating the analysis result based on the corresponding types of medical information; Function as a The analysis unit includes: identifying the importance of each of the computing units included in the computing unit group based on importance information indicating the importance of each type of the medical information when the analysis result is output; Calculating the analysis result based on the partial analysis information output by each of the computing units when corresponding medical information is input and the importance of each of the computing units identified. A medical decision support program comprising:

Images