Medical systems and medical information processing devices

Abstract

To provide a new technology to detect the state of a circulatory system of a patient in a non-invasive manner.SOLUTION: A data acquisition unit of a medical system acquires data from the ocular fundus of a patient using at least one optical method, and a data processing unit processes the data acquired by the data acquisition unit for generating information on the circulatory system of the patient. The data processing unit includes an inference processing part. Using a learned model constructed by machine learning using training data including second data generated by processing first data acquired from the ocular fundus using the at least one optical method, and diagnostic result data, the inference processing part performs inference processing with the data generated by processing the data acquired from the ocular fundus of the patient by the data acquisition unit as input, and the information on the circulatory system of the patient as output.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] This invention relates to a medical system and a medical information processing device.

Background Art

[0002] Symptoms of diseases and signs of exacerbation are complex, and various techniques have been developed to detect them. For example, Patent Document 1 discloses a technique for determining the risk of infectious diseases without using advanced medical knowledge, which determines the risk based on the presence or absence of abnormalities in each of arterial oxygen saturation, body temperature, and heart rate.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] One object of this invention is to provide a new technique for non-invasively detecting the state of a patient's circulatory system.

Means for Solving the Problems

[0005] A medical system according to some exemplary embodiments includes a data acquisition unit and a data processing unit. The data acquisition unit acquires data from the fundus of a patient using at least one optical method. The data processing unit processes the data acquired by the data acquisition unit to generate information regarding the patient's circulatory system.

[0006] In some exemplary embodiments, the information regarding the circulatory system includes information regarding the tendency to form blood clots.

[0007] In some exemplary embodiments, the information regarding the tendency to form blood clots includes information regarding the properties of blood.

[0008] In some exemplary embodiments, information regarding blood properties includes information indicating changes in blood properties due to increased blood coagulation and fibrinolysis.

[0009] In some exemplary embodiments, information relating to the circulatory system includes information relating to thrombotic symptoms.

[0010] In some exemplary embodiments, information regarding thrombotic symptoms includes information indicating the distribution of blood flow velocity within the blood vessels.

[0011] In some exemplary embodiments, information regarding thrombotic symptoms includes information regarding structures formed within blood vessels.

[0012] In some exemplary embodiments, information concerning the circulatory system includes information concerning the circulatory system condition associated with the infection.

[0013] In some exemplary embodiments, information relating to the circulatory system includes at least one of the following: information relating to a condition relating to sepsis, information relating to a condition relating to disseminated intravascular coagulation (DIC), information relating to a condition relating to thrombosis, and information relating to a condition relating to vascular occlusion.

[0014] In some exemplary embodiments, at least one optical technique includes at least one of optical coherence tomography blood flow measurement (OCT blood flow measurement), optical coherence tomography angiography (OCT-A), and color fundus photography.

[0015] In some exemplary embodiments, at least one optical technique includes OCT blood flow measurement, and a data processing unit generates information about the blood coagulation and fibrinolytic system based at least on the blood flow data obtained by the OCT blood flow measurement.

[0016] In some exemplary embodiments, the data acquisition unit includes an OCT device and a calculation unit. The OCT device collects data by applying an optical coherence tomography (OCT) scan to the patient's fundus. The calculation unit calculates blood flow velocity and vessel diameter based at least on the data collected by the OCT device. The data processing unit generates information about the blood coagulation and fibrinolytic systems based at least on the blood flow velocity and vessel diameter calculated by the calculation unit.

[0017] In some exemplary embodiments, the data processing unit includes a WSR calculation unit. The WSR calculation unit calculates the wall shear velocity (WSR) based at least on the blood flow velocity and the vessel diameter.

[0018] In some exemplary embodiments, the data processing unit includes a storage unit and a WSS calculation unit. The storage unit stores previously acquired blood viscosity information. The WSS calculation unit calculates the wall shear stress (WSS) based at least on the wall shear rate and the blood viscosity information.

[0019] In some exemplary embodiments, the data acquisition unit includes an OCT device and a blood flow information generation unit. The OCT device repeatedly applies optical coherence tomography (OCT) scans to a predetermined area of ​​the patient's fundus to collect time-series data. The blood flow information generation unit generates blood flow information representing the spatial distribution and temporal changes of blood flow velocity, at least based on the time-series data collected by the OCT device. The data processing unit generates information regarding the blood coagulation and fibrinolytic systems, at least based on the blood flow information generated by the blood flow information generation unit.

[0020] In some exemplary embodiments, the data processing unit generates information about structures formed within blood vessels, based at least on blood flow information.

[0021] In some exemplary embodiments, the data processing unit includes a WSR information generation unit. The WSR information generation unit generates WSR information representing the spatial distribution and temporal variation of wall shear velocity (WSR), based at least on blood flow information.

[0022] In some exemplary embodiments, the data processing unit generates information regarding a structure formed in a blood vessel based at least on blood flow information and WSR information.

[0023] In some exemplary embodiments, the data processing unit includes a storage unit and a WSS information generation unit. The storage unit stores blood viscosity distribution information acquired in advance. The WSS information generation unit generates WSS information representing the spatial distribution and temporal change of wall shear stress (WSS) based at least on WSR information and blood viscosity distribution information.

[0024] In some exemplary embodiments, the data processing unit generates information regarding a structure formed in a blood vessel based at least on blood flow information and WSS information.

[0025] In some exemplary embodiments, the data processing unit includes a first inference processing unit. The first inference processing unit performs an inference process that takes, as an input, data acquired from the fundus of a patient by the data acquisition unit and outputs information regarding the circulatory system of the patient, using a first trained model constructed by machine learning using first training data including the first data acquired from the fundus using at least one optical method and diagnostic result data.

[0026] In some exemplary embodiments, the data processing unit includes a second inference processing unit. The second inference processing unit performs an inference process that takes, as an input, data generated by processing data acquired from the fundus of a patient by the data acquisition unit and outputs information regarding the circulatory system of the patient, using a second trained model constructed by machine learning using second training data including the second data generated by processing the first data acquired from the fundus using at least one optical method and diagnostic result data.

[0027] A medical system according to some exemplary embodiments further includes a transmission unit. The transmission unit transmits information regarding the circulatory system generated by the data processing unit toward a doctor terminal located at a remote position with respect to the data acquisition unit.

[0028] Some exemplary medical systems further include a physician terminal.

[0029] A medical system according to some exemplary embodiments further includes an operating unit for remotely controlling a data acquisition unit.

[0030] A medical information processing device according to some exemplary embodiments includes a data receiving unit and a data processing unit. The data receiving unit receives data acquired from the patient's fundus using at least one optical method. The data processing unit processes the data received by the data receiving unit to generate information about the patient's circulatory system.

[0031] A medical information processing device according to several exemplary embodiments further includes a first transmitting unit. The first transmitting unit transmits information about the circulatory system generated by the data processing unit to a physician terminal located remotely from the location where the data was acquired.

[0032] A medical system according to several exemplary embodiments includes a medical information processing device according to an exemplary embodiment and a physician's terminal.

[0033] A medical system according to some exemplary embodiments further includes a data acquisition device and a second transmission unit. The data acquisition device acquires data from the patient's fundus using at least one optical method. The second transmission unit transmits the data acquired by the data acquisition device to a medical information processing device. A data receiving unit receives the data transmitted by the second transmission unit. A data processing unit processes the data transmitted by the second transmission unit and received by the data receiving unit to generate information about the patient's circulatory system. [Effects of the Invention]

[0034] According to exemplary embodiments, it is possible to provide a new technology for non-invasively detecting the state of a patient's circulatory system. [Brief explanation of the drawing]

[0035] [Figure 1]This is a schematic diagram illustrating an example of the configuration of a medical system according to an exemplary embodiment. [Figure 2] This is a schematic diagram illustrating an example of the data structure processed by a medical system according to an exemplary embodiment. [Figure 3] This is a schematic diagram illustrating an example of the configuration of a medical system according to an exemplary embodiment. [Figure 4] This is a schematic diagram illustrating an example of the configuration of a medical system according to an exemplary embodiment. [Figure 5] This is a schematic diagram illustrating an example of the configuration of a medical system according to an exemplary embodiment. [Figure 6] This is a schematic diagram illustrating an example of the configuration of a medical system according to an exemplary embodiment. [Figure 7] This is a schematic diagram illustrating an example of the configuration of a medical system according to an exemplary embodiment. [Figure 8] This is a flowchart illustrating an example of the operation of a medical system relating to an exemplary embodiment. [Figure 9] This is a schematic diagram illustrating an example of the configuration of a medical system according to an exemplary embodiment. [Figure 10] This is a schematic diagram illustrating an example of the configuration of a medical system according to an exemplary embodiment. [Figure 11] This is a schematic diagram illustrating an example of the configuration of a medical system according to an exemplary embodiment. [Figure 12] This is a schematic diagram illustrating an example of the configuration of a medical system according to an exemplary embodiment. [Figure 13] This is a schematic diagram illustrating an example of the configuration of a medical information processing device and a medical system including the same, according to an exemplary embodiment. [Modes for carrying out the invention]

[0036] This disclosure describes several exemplary embodiments of medical systems and medical information processing devices. A person with ordinary skill in the art will understand that the embodiments relating to this disclosure provide various variations and equivalents, and that the embodiments relating to this disclosure or their variations or equivalents provide various other embodiments such as medical methods, methods for controlling systems, methods for controlling devices, programs, recording media, etc.

[0037] Several exemplary embodiments involve generating information about a patient's circulatory system by processing data acquired from the patient's fundus using at least one optical method (optical modality) with a computer. This computer processing may include inference. This inference may be performed, for example, by an algorithm using a trained model (inference model) built by machine learning, by an algorithm without a trained model, or by a combination of these.

[0038] The data subjected to computer processing in some exemplary embodiments may be data obtained from any ophthalmic examination, for example, data obtained from any ophthalmic modality device. The ophthalmic modality device may be, for example, an optical coherence tomography (OCT) device, a fundus camera, a scanning laser ophthalmoscope, a slit-lamp microscope, or a surgical microscope. In some exemplary embodiments, the optical coherence tomography device is used, for example, for optical coherence tomography blood flow measurement or optical coherence tomography angiography (OCT-A). In some exemplary embodiments, fundus imaging devices such as fundus cameras, scanning laser ophthalmoscopes, slit-lamp microscopes, and surgical microscopes are used, for example, for color fundus imaging. The data subjected to computer processing is not limited to these, and may further include, for example, other types of examination data, electronic medical record data, interview data, and patient background information (age, treatment history, medical history, medication history, surgical history, etc.).

[0039] Exemplary embodiments are configured to generate predetermined information about a patient's circulatory system from such data. The information generated by some exemplary embodiments may include at least one of quantitative and qualitative information, for example, any of the following: information on thrombosis tendencies; information on thrombotic symptoms; information on circulatory system conditions associated with infection; information indicating conditions related to sepsis; information indicating conditions related to disseminated intravascular coagulation (DIC); information indicating conditions related to thrombosis; and information indicating conditions related to vascular occlusion.

[0040] Information regarding thrombosis tendency refers to information indicating a tendency for thrombi to form in the patient's circulatory system (intravascular, intracardiac), and includes, for example, information regarding thrombosis risk. Information regarding thrombosis tendency may include either information regarding blood properties or information indicating changes in blood properties due to increased coagulation and fibrinolysis. Information regarding blood properties includes information regarding the nature of the blood and / or information regarding the state of the blood. Information indicating changes in blood properties due to increased coagulation and fibrinolysis includes information indicating changes in blood properties resulting from the activation of the blood coagulation system (coagulation system, blood coagulation factors), and / or information indicating changes in blood properties resulting from the activation of the blood clot dissolving system (fibrinolysis system, fibrinolysis system). Information regarding thrombosis tendency may include, for example, viscosity, wall shear stress, wall shear rate, amount or proportion of specific components, ratios between specific components, information indicating changes in any of these, and information indicating the distribution of any of these.

[0041] Information regarding thrombotic symptoms is information regarding symptoms caused by thrombosis and may include, for example, information showing the distribution of blood flow velocity within a blood vessel, and information regarding structures formed within a blood vessel. The distribution of blood flow velocity within a blood vessel may be, for example, one or more combinations of one-dimensional, two-dimensional, three-dimensional, and temporal distributions. Structures formed within a blood vessel may be, for example, white thrombi, red thrombi, mixed thrombi, hyaline thrombi, or structures related to the formation mechanism of any of these (e.g., intermediate products).

[0042] Information regarding the cardiovascular condition associated with infectious diseases includes, for example, information regarding diseases and conditions associated with or caused by infectious diseases, such as vasculitis, thrombosis, blood coagulation, sepsis, DIC, pneumonia, lymphadenitis, and lymphangitis. The infectious diseases covered may be any viral infection, any bacterial infection, or any fungal infection, such as the 2020 pandemic of Coronavirus Disease 2019 (COVID-19), Severe Acute Respiratory Syndrome (SARS), Middle East Respiratory Syndrome (MERS), influenza, and infectious endocarditis.

[0043] Sepsis is a very serious condition caused by the systemic spread of infection, leading to circulatory shock, disseminated intravascular coagulation (DIC), and multiple organ failure. Information describing a sepsis condition includes, for example, information about symptoms such as inflammation and circulatory failure caused by sepsis.

[0044] Disseminated intravascular coagulation (DIC) is a syndrome in which blood coagulation reactions, which should normally occur only at the site of bleeding, occur uncontrolledly throughout the blood vessels of the body. The pathophysiology of DIC involves persistent and marked coagulation activation throughout the blood vessels, leading to the formation of numerous microthrombi. As it progresses, organ damage due to microcirculatory impairment leads to consumptive coagulation disorders and bleeding. Furthermore, fibrinolytic activation occurs along with coagulation activation, resulting in excessive fibrinolysis of thrombi and promoting bleeding. Information describing the state of disseminated intravascular coagulation (DIC) includes, for example, information describing the above-mentioned pathophysiological features of DIC (hypercoagulation, hyperfibrinolysis, thrombosis, bleeding, etc.).

[0045] Information indicating the status of thrombi may be any information regarding thrombi present or potentially present in the circulatory system (intravascular, intracardiac), including, for example, the presence or absence of thrombi, the degree of thrombosis, the distribution of thrombi, the number of thrombi, and the probability of thrombosis formation.

[0046] Information indicating the status of vascular occlusion may be any information relating to vascular occlusion occurring or potentially occurring in the circulatory system, and may include, for example, the presence or absence of vascular occlusion, the degree of vascular occlusion, the distribution of vascular occlusion locations, the number of vascular occlusion locations, and the probability of vascular occlusion occurring.

[0047] Thus, exemplary embodiments enable non-invasive detection of a patient's circulatory system status by generating information about the patient's circulatory system based on data acquired from the patient's fundus using one of the exemplary optical modalities described above. Several exemplary embodiments can generate information about one or more of the following: thrombosis tendency (e.g., blood properties, and / or changes in blood properties due to increased coagulation and fibrinolysis), thrombotic symptoms (e.g., blood flow velocity distribution, and / or intravascular structures), circulatory system status and / or changes in status associated with infection, sepsis, DIC, thrombosis, vascular occlusion, matters similar to any of these, matters arising from any of these, and matters relating to any of these mechanisms. The types of information that can be generated by exemplary embodiments are not limited to these, and may be any type of information that can be generated (e.g., derived, estimated, etc.) by a combination of the optical modality employed and the data processing employed.

[0048] Several exemplary embodiments were devised with the following background in mind and can achieve corresponding effects. Healthcare workers such as doctors and nurses are exposed to the risk of hospital-acquired infections. For example, during the COVID-19 pandemic in 2020, cluster infections occurred in medical institutions overwhelmed with many patients, highlighting the significant risk of infection to healthcare workers. It should be noted that the risk of infection to healthcare workers can increase not only during infectious disease outbreaks but also during disasters and major accidents. Generally, maintaining sufficient distance between people, so-called social distancing, is considered important for reducing the risk of infection, but achieving this in standard medical practice is not easy. For example, when performing tests, doctors and other medical professionals are often in close proximity to patients.

[0049] Some exemplary embodiments may be configured to provide information generated by computer processing of data acquired by an optical modality to a physician's terminal located remotely. Furthermore, some exemplary embodiments may be configured to allow remote operation of the testing device (optical modality device) and computer. These configurations make it possible to use data obtained from tests that previously could only be performed by being in close proximity to the patient for diagnostic purposes. In other words, some exemplary embodiments allow for maintaining social distancing between patients and healthcare workers, while enabling non-invasive and highly accurate detection of complex physiological events such as symptoms and signs of worsening conditions.

[0050] Here, "remote location" refers to any location that allows for social distancing between the patient and healthcare worker. For example, the physician's terminal may be located in a different room from the testing equipment or in a different facility. Similarly, the device for remotely controlling the testing equipment (operating device, control unit) may be located in a different room from the testing equipment or in a different facility. Furthermore, social distancing may not be required if the test is conducted under sufficient infection control measures, such as when wearing full protective clothing.

[0051] Exemplary embodiments can be modified by the information contained herein or by any other known art. Such modifications may include, for example, addition, combination, substitution, deletion, omission, and other modifications.

[0052] At least some of the functions of the elements described in this disclosure may be implemented using a circuitry or processing circuitry. The circuitry or processing circuitry may be a general-purpose processor, dedicated processor, integrated circuit, CPU (Central Processing Unit), GPU (Graphics Processing Unit), ASIC (Application Specific Integrated Circuit), programmable logic device (e.g., SPLD (Simple Programmable Logic Device), CPLD (Complex Programmable Logic Device), FPGA (Field Programmable Gate) configured and / or programmed to perform at least some of the disclosed functions. A processor may include an array, a conventional circuit configuration, and any combination thereof. A processor may be considered a processing circuit configuration or circuit configuration, including transistors and / or other circuit configurations. In this disclosure, terms such as circuit configuration, circuit, computer, processor, unit, means, part, or similar terms may include hardware that performs at least a portion of the disclosed functions, and / or hardware programmed to perform at least a portion of the disclosed functions. Hardware may be hardware disclosed herein, or known hardware programmed and / or configured to perform at least a portion of the disclosed functions. If the hardware is a processor that can be considered a certain type of circuit configuration, then terms such as circuit configuration, circuit, computer, processor, unit, means, part, or similar terms may be a combination of hardware and software, the software may be used to constitute the hardware and / or processor.

[0053] The exemplary embodiments described below may be combined in any way. For example, two or more exemplary embodiments can be combined at least partially.

[0054] <Configuration of the medical system> Several examples of the configuration of an exemplary medical system are described below. The exemplary medical system 1 shown in Figure 1 includes a data acquisition unit 10, a data processing unit 20, and an output unit 30. The medical system 1 may further include an operating device 2.

[0055] In a typical example, the data acquisition unit 10 and the data processing unit 20 are connected via a communication line. This communication line may, for example, form a network within a medical institution, or it may form a network spanning multiple facilities. The communication technology applied to this communication line may be arbitrary and may be any of the various known communication technologies such as wired communication, wireless communication, or short-range communication. The connection configuration between the data processing unit 20 and the output unit 30 may be similar. Alternatively, the data processing unit 20 and the output unit 30 may be functional units mounted on the same computer.

[0056] The operating device 2 is used by medical professionals to remotely operate the data acquisition unit 10 (examination device, optical modality device). The operating device 2 is also used by medical professionals (examiners) to provide instructions to patients (subjects) undergoing examinations using the data acquisition unit 10. Furthermore, the operating device 2 may be used to remotely operate the data processing unit 20. The operating device 2 includes, for example, a computer, an operation panel, etc.

[0057] The data acquisition unit 10 is configured to acquire data from the patient's fundus using at least one optical modality. The data acquisition unit 10 includes any optical fundus imaging modality device, such as an optical coherence tomography device, fundus camera, scanning laser ophthalmoscope, slit lamp microscope, or surgical microscope. The data acquisition unit 10 may also be capable of acquiring other types of examination data, electronic medical record data, interview data, patient background information, etc.

[0058] The optical coherence tomography apparatus and / or fundus camera may be, for example, an automated apparatus for various imaging preparation operations, as described in Japanese Patent Publication No. 2020-44027. These imaging preparation operations are performed to optimize imaging conditions, and examples include alignment adjustment, focus adjustment, optical path length adjustment, polarization adjustment, and light intensity adjustment. Furthermore, the apparatus may be capable of automatically performing operations to maintain the favorable imaging conditions achieved through the imaging preparation operations. Such operations include automatic alignment adjustment (tracking) in accordance with eye movements and automatic optical path length adjustment (Z-lock) in accordance with eye movements. These automated operations are effective, for example, in examinations performed without the presence of an examiner.

[0059] The data acquired by the optical coherence tomography device (optical coherence tomography data) may be, for example, at least one of the following: 3D image data obtained by applying a 3D scan to the fundus, projection image data of 3D image data, optical coherence tomography angiography image data, or optical coherence tomography blood flow data.

[0060] Optical coherence tomography angiography is an optical modality that visualizes blood vessels using motion contrast technology, enabling visualization of even fine vessels. Optical coherence tomography angiography image data is acquired using optical coherence tomography devices described, for example, in Japanese Patent Publication No. 2019-58495 and Japanese Patent Publication No. 2019-154988.

[0061] Optical coherence tomography is an optical modality for measuring blood flow (blood flow dynamics). Optical coherence tomography blood flow data is acquired using optical coherence tomography devices described, for example, in Japanese Patent Publication No. 2019-54994 and Japanese Patent Publication No. 2020-48730. In some exemplary embodiments, optical coherence tomography can acquire optical coherence tomography blood flow data such as blood flow velocity, blood flow rate, vessel diameter, waveform data representing time-series changes (time changes, time-dependent changes) of blood flow velocity, and waveform data representing time-series changes of blood flow rate. Typically, the waveform data is a time-series graph of blood flow velocity represented by a two-dimensional coordinate system with time on the horizontal axis and blood flow velocity on the vertical axis. Furthermore, the optical modality used for measuring fundus blood flow is not limited to optical coherence tomography blood flow measurement, but may also be laser speckle flowgraphy (LSFG), as described in publication No. 2008 / 069062, for example.

[0062] Image data obtainable by a fundus camera (fundus camera image data) includes, for example, color fundus image data, infrared fundus image data, and fluorescence angiography fundus image data (fluorescein angiography image data, indocyanine green angiography image data, etc.). In some exemplary embodiments, color fundus image data is obtained using a fundus camera.

[0063] The scanning laser ophthalmoscope may be, for example, the device described in Japanese Patent Publication No. 2014-226156. Examples of image data (scanning laser image data) that can be acquired by a scanning laser ophthalmoscope include color fundus image data, monochromatic fundus image data, and fluorescence angiography fundus image data. In some exemplary embodiments, color fundus image data is acquired using a scanning laser ophthalmoscope.

[0064] The slit lamp microscope may be, for example, a device effective for remote imaging as described in Japanese Patent Publication No. 2019-213734. The image data acquired by the slit lamp microscope may be, for example, at least one of color fundus image data, anterior segment cross-sectional image data, and anterior segment three-dimensional image data. In some exemplary embodiments, color fundus image data is acquired using a slit lamp microscope.

[0065] The surgical microscope may be, for example, a device effective for remote surgery as described in Japanese Patent Publication No. 2002-153487. In some exemplary embodiments, color fundus image data is acquired using the surgical microscope.

[0066] In this embodiment, at least one of the inspection devices (e.g., optical coherence tomography device, fundus camera, etc.) included in the data acquisition unit 10 may be remotely operated and / or remotely controlled.

[0067] For example, to consider the risk of infection to healthcare workers, the testing room where the testing equipment is used can be separated from the control room where the equipment is operated. In addition to the testing equipment, the testing room is equipped with speakers and displays to output instructions (voice, images, video, etc.) from the operator in the control room, a video camera to film the subject (patient) in the testing room, a microphone to input the subject's voice, and a computer connected to the testing equipment.

[0068] Meanwhile, the control room is equipped with control device 2 for remotely operating the testing equipment. Control device 2 includes a computer, control panel, display, video camera, microphone, etc. The computer performs processing for remote operation. The computer is connected to the testing equipment in the testing room. The control panel, video camera, and microphone are used to input instructions to the patient. The display shows data acquired by the testing equipment and information for remote operation (screen, information from the testing room, etc.).

[0069] With this configuration, operators (healthcare workers) in the control room can remotely control testing equipment in the testing room using, for example, an application programming interface (API), and can also send instructions to the patient using video conferencing. As a result, the patient can undergo the test alone following the instructions of the operator in a remote location, and consequently, the risk of infection from the patient to the operator can be significantly reduced.

[0070] To facilitate examinations performed by a single patient (subject), an automated examination device (as described above) with automated preparation procedures can be used. In this case, it is possible to perform the examination without requiring instructions from an operator. In some cases, it may not even be necessary to have an assistant (such as an operator) present. However, since it is anticipated that some patients may have difficulty undergoing the examination alone, an assistant may be stationed at a remote location, or the assistant may monitor the examination status from a remote location. The assistant (such as an operator) that sends instructions to the patient may be an anthropomorphic computer system (typically an automated response system using artificial intelligence technology).

[0071] The data processing unit 20 performs various data processing operations. In this embodiment, the data processing unit 20 is configured to process the data acquired by the data acquisition unit 10 in order to generate information about the patient's circulatory system.

[0072] The information generated by the data processing unit 20 in this embodiment may be, for example, at least one of the following: information on thrombus formation tendency (information on blood properties, and / or information indicating changes in blood properties due to enhanced blood coagulation and fibrinolysis); information on thrombotic symptoms (a method indicating blood flow velocity distribution within blood vessels, and / or information on structures formed within blood vessels); information on the state (and / or changes in state) of the circulatory system associated with infection; information indicating a state related to sepsis; information indicating a state related to DIC; information indicating a state related to thrombosis; information indicating a state related to vascular occlusion.

[0073] Some examples of the processing performed by the data processing unit 20 will be described in the embodiments described below. The data processing unit 20 may or may not use a trained model (inference model) constructed using machine learning.

[0074] Figure 2 shows an example of a data structure for processing (recording, transmitting, etc.) data generated by the data processing unit 20. The data structure 100 in this example includes a thrombus formation tendency data section 110, a thrombotic symptom data section 120, an infection-related data section 130, a sepsis data section 140, a DIC data section 150, a thrombus data section 160, and a vascular occlusion data section 170.

[0075] The thrombus formation tendency data section 110 is an area (such as a folder) where information regarding thrombus formation tendency generated by the data processing unit 20 is recorded. The thrombus formation tendency data section 110 includes the blood properties data section 111. The blood properties data section 111 is an area where information regarding blood properties generated by the data processing unit 20 is recorded. The blood properties data section 111 includes the blood properties change data section 112. The blood properties change data section 112 is an area where information indicating changes in blood properties due to increased blood coagulation and fibrinolysis, generated by the data processing unit 20, is recorded.

[0076] The thrombotic symptom data section 120 is an area where information regarding thrombotic symptoms generated by the data processing unit 20 is recorded. The thrombotic symptom data section 120 includes a blood flow velocity distribution data section 121 and an intravascular structure data section 122. The blood flow velocity distribution data section 121 is an area where information indicating the distribution of blood flow velocity within a blood vessel, generated by the data processing unit 20, is recorded. The intravascular structure data section 122 is an area where information regarding structures formed within a blood vessel, generated by the data processing unit 20, is recorded.

[0077] The Infection-Related Data Unit 130 is an area where information regarding the cardiovascular state associated with infection, generated by the data processing unit 20, is recorded. The Sepsis Data Unit 140 is an area where information indicating the state of sepsis, generated by the data processing unit 20, is recorded. The DIC Data Unit 150 is an area where information indicating the state of DIC, generated by the data processing unit 20, is recorded. The Thrombosis Data Unit 160 is an area where information indicating the state of thrombosis, generated by the data processing unit 20, is recorded. The Vascular Occlusion Data Unit 170 is an area where information indicating the state of vascular occlusion, generated by the data processing unit 20, is recorded.

[0078] In some exemplary embodiments, the data structure 100 includes at least one of the data sections 110 to 170 described above. In some exemplary embodiments, the data structure 100 may include data sections other than the data sections 110 to 170 described above. For example, the data structure 100 may include a fundus data section in which data acquired from the fundus by the data acquisition unit 10 is recorded, a processed data section in which data obtained by applying predetermined processing to the data acquired from the fundus by the data acquisition unit 10 is recorded, and an arbitrary data section in which any type of data is recorded. The arbitrary data section may record, for example, data acquired by any examination device, electronic medical record data, interview data, patient information (e.g., patient identifier, patient background information), etc.

[0079] We will explain some of the background to this embodiment, which is configured to generate these data. In "Is Disseminated Intravascular Coagulation (DIC) Involved in Deaths from COVID-19?" (Japan Medical Journal website: https: / / www.jmedj.co.jp / Journal / paper / detail.php?id=14500), it is stated that COVID-19-induced DIC (thromboembolism due to DIC) is considered to be one of the causes of death from severe COVID-19 pneumonia, that myocarditis may also occur, that cardiac and vascular echocardiography, which is performed in a closed and close-contact environment, is rarely performed and therefore the formation of thrombi in deep veins and within the heart is unknown, and that DIC... It has been pointed out that myocarditis can easily lead to the formation of intracardiac thrombi, potentially causing thromboembolism and multiple organ failure; that COVID-19 infection can cause sepsis; that imaging diagnosis of microthrombotic disorders caused by sepsis and other conditions is difficult; that early circulatory abnormalities are difficult to diagnose because they involve microvessels; that blood coagulation tests, including D-dimer, and cardiac and vascular echocardiography are considered effective for patients with COVID-19 pneumonia; that if DIC is diagnosed, dramatic improvement in symptoms can be expected with anticoagulation therapy; and that preventing thrombosis may be fundamental to the treatment of COVID-19 pneumonia.

[0080] The article "It is estimated that many severely ill COVID-19 patients develop sepsis" (Japan Medical Journal website: https: / / www.jmedj.co.jp / Journal / paper / detail.php?id=14563) points out that many severe cases and deaths from COVID-19 are caused by sepsis. This invention provides a non-invasive method for providing information on sepsis.

[0081] The Ministry of Health, Labour and Welfare has pointed out that in sepsis, the balance between blood coagulation and thrombolysis is disrupted, leading to the formation of blood clots throughout the body and bleeding in microvessels, resulting in a syndrome called DIC, which has a poor prognosis. This method provides a non-invasive approach to providing information on DIC, blood coagulation, thrombolysis, thrombosis, and bleeding.

[0082] The article "COVID-19 and Coagulopathy: Frequently Asked Questions" (AMERICAN SOCIETY OF HEMATOLOGY website: https: / / www.hematology.org / covid-19-and-coagulopathy) points out that in patients infected with COVID-19, when DIC occurs, thrombus formation may occur in various microvessels, mainly in the lungs; the correlation between blood test values ​​indicating increased blood coagulation and the severity of the disease strongly suggests this possibility; frequent occurrence of heart and kidney damage in addition to lung damage is also thought to be due to intravascular thrombus formation; and there have been cases that presented with vasculitis symptoms similar to Kawasaki disease as skin symptoms. This invention provides a non-invasive method for providing information on thrombosis, blood coagulation, and vasculitis.

[0083] The "Guidelines for the Treatment of Novel Coronavirus Infection 2020 19-COVID 2nd Edition" (Ministry of Health, Labour and Welfare website: https: / / www.mhlw.go.jp / content / 000631552.pdf) points out that D-dimer, CRP (C-reactive protein), LDH (serum lactate dehydrogenase), ferritin, lymphocytes, and creatinine may be useful as markers of severe COVID-19. In particular, a correlation has been noted between blood test values ​​indicating increased blood coagulation and the severity of the disease. In addition to these, some literature points out the usefulness of cardiac toporonin (Tn), IL-1β, IL-6, IL-8, TNFa, and IFNa. This invention provides a non-invasive method for providing information on changes in blood properties that are reflected in these severe disease markers.

[0084] The data processing unit 20 in this embodiment may be designed and configured with this background in mind. Several exemplary embodiments of the data processing unit 20 will be described later. It should be noted that the data processing unit 20 (and data structure 100) can be designed and configured in the same manner when targeting other infectious diseases.

[0085] Figure 3 shows an example of the configuration of the data processing unit 20 in this embodiment. In this example, the data processing unit 20 includes an ocular image data processing unit 21 and an ocular blood flow data processing unit 22.

[0086] The ocular image data processing unit 21 may include, for example, a processor that operates according to a program created based on at least the medical knowledge described above. In this case, the ocular image data processing unit 21 can generate information about the patient's circulatory system by processing the image data (ocular image data) acquired from the patient's fundus by the data acquisition unit 10 with at least this processor.

[0087] The ocular image data input to the processor may be, for example, optical coherence tomography image data or color fundus image data. The information output from the processor may be, for example, any of the following, as described above: information on thrombosis tendency, information on thrombotic symptoms, information on the state of the circulatory system associated with infection, information indicating the state of sepsis, information indicating the state of DIC, information indicating the state of thrombosis, and information indicating the state of vascular occlusion.

[0088] The ocular image data processing unit 21 may include, for example, a trained model constructed by machine learning based on medical knowledge as described above. In this case, the ocular image data processing unit 21 can generate information about the patient's circulatory system by processing the image data (ocular image data) acquired from the patient's fundus by the data acquisition unit 10 using at least this trained model.

[0089] The ocular image data input to the trained model may be, for example, optical coherence tomography image data or color fundus image data. The information output from the trained model may be, for example, any of the following, as described above: information on thrombosis tendency, information on thrombotic symptoms, information on the state of the circulatory system associated with infection, information indicating the state of sepsis, information indicating the state of DIC, information indicating the state of thrombosis, and information indicating the state of vascular occlusion.

[0090] Figure 4 shows an example of an ocular image data processing unit 21 configured using machine learning. In this example, the ocular image data processing unit 21 includes an inference processing unit 210. The inference processing unit 210 is configured to perform inference processing to derive information about the patient's cardiovascular system from ocular image data acquired by the data acquisition unit 10, using a trained model constructed by machine learning using training data including clinical data (ocular image data and diagnostic result data).

[0091] The ocular image data included in the training data is, for example, image data acquired using the same optical modality as the optical modality of the data acquisition unit 10, but other modalities may also be used. Other modalities include optical modalities different from the optical modality of the data acquisition unit 10, ultrasound modalities, electrical modalities, magnetic modalities, electromagnetic modalities, etc. The diagnostic result data included in the training data may be, for example, data obtained by a physician or another inference model (trained model) based on the relevant ocular image data.

[0092] Through machine learning (supervised learning) based on such training data, a trained model (inference model) can be created that takes ocular image data acquired by the data acquisition unit 10 as input and outputs estimated diagnostic data related to the cardiovascular system. The training data used in machine learning may include data created by a computer based on clinical data. Machine learning may also include transfer learning.

[0093] The inference processing unit 210 includes the trained model obtained in this manner, inputs the eye image data acquired by the data acquisition unit 10 into the trained model, and sends the estimated diagnostic data output from the trained model to the output unit 30.

[0094] The machine learning algorithms that can be used in exemplary embodiments are not limited to supervised learning, but may be any algorithm such as unsupervised learning, semi-supervised learning, reinforcement learning, transduction, multitask learning, or any combination of two or more algorithms.

[0095] The machine learning techniques that can be used in exemplary embodiments are arbitrary and may include any techniques such as neural networks, support vector machines, decision tree learning, correlation rule learning, genetic programming, clustering, Bayesian networks, representation learning, and extreme learning machines, or any combination of two or more techniques.

[0096] Figure 5 shows an example of the configuration of the inference processing unit 210. In this example, the inference processing unit 210 includes a first trained model 211 and a second trained model 212. In some exemplary embodiments, the inference processing unit 210 may include only one of the first trained model 211 and the second trained model 212.

[0097] The first pre-trained model 211 is constructed by machine learning using training data including ocular image data and diagnostic result data. For example, the first pre-trained model 211 includes a convolutional neural network (CNN). This convolutional neural network includes, for example, an input layer into which ocular image data is input, a convolutional layer that applies filtering (convolution) to the input ocular image data to create a feature map, a pooling layer that compresses the data while retaining the features obtained in the convolutional layer, a fully connected layer that extracts and determines characteristic findings from all the data obtained in the pooling layer, and an output layer that outputs the data obtained in the fully connected layer. By inputting the ocular image data acquired by the data acquisition unit 10 into the first pre-trained model 211, information regarding the circulatory system that takes predetermined features into consideration is generated.

[0098] The features considered by the first pre-trained model 211 may include, for example, features related to the visualization state and features related to the visualization target. Features related to the visualization state include color tone and brightness. Features related to the visualization target include features related to the retinal blood vessels, features related to the optic nerve head, and features related to the macula.

[0099] In this embodiment, in order to generate information about the circulatory system, features of the retinal blood vessels are considered in particular. Features of the retinal blood vessels include distribution, thickness (vascular diameter), curvature (course), and hemorrhage. For example, features such as rupture, hemorrhage, and abnormal course of retinal microvessels may be detected from ocular imaging data of patients with sepsis or DIC.

[0100] As one embodiment, we will describe an example where image data representing the morphology (structure) of the fundus, such as optical coherence tomography angiography image data, is acquired. In this case, the first trained model 211 includes, for example, a convolutional neural network constructed by machine learning using training data including optical coherence tomography angiography image data and diagnostic result data. Note that the training data may include any image data, such as fluorescence angiography fundus image data. The convolutional neural network in this embodiment includes, for example, an input layer into which optical coherence tomography angiography image data is input, a convolutional layer that applies filtering (convolution) to the input optical coherence tomography angiography image data to create a feature map related to the vascular structure, a pooling layer that compresses data while preserving the features of the vascular structure obtained in the convolutional layer, a fully connected layer that extracts and determines characteristic findings of the vascular structure from all the data obtained in the pooling layer, and an output layer that outputs the data obtained in the fully connected layer. By inputting the optical coherence tomography angiography image data acquired by the data acquisition unit 10 into the first trained model 211, information regarding the circulatory system that takes into account the vascular structure of the fundus is generated.

[0101] As another embodiment, an example in which color fundus image data is acquired will be described. In this case, the first trained model 211 includes, for example, a convolutional neural network constructed by machine learning using training data including color fundus image data and diagnostic result data. The convolutional neural network in this embodiment includes, for example, an input layer into which color fundus image data is input, a convolutional layer that applies filtering (convolution) to the input color fundus image data to create a feature map relating to color information (e.g., R value, G value, B value), a pooling layer that compresses data while preserving the features of the color information obtained in the convolutional layer, a fully connected layer that extracts and determines characteristic findings of color information from all the data obtained in the pooling layer, and an output layer that outputs the data obtained in the fully connected layer. By inputting the color fundus image data acquired by the data acquisition unit 10 into the first trained model 211, information relating to the circulatory system that takes into account the color tone of the fundus is generated.

[0102] The second pre-trained model 212 was constructed using machine learning with training data that included data generated by processing ocular image data acquired from the fundus using a predetermined modality, as well as diagnostic result data.

[0103] If the data generated by processing ocular image data is image data, the second trained model 212 includes, for example, a convolutional neural network. This convolutional neural network includes, for example, an input layer to which the image data generated by processing ocular image data is input, a convolutional layer that applies filtering (convolution) to the input ocular image data to create a feature map, a pooling layer that compresses the data while retaining the features obtained in the convolutional layer, a fully connected layer that extracts and determines characteristic findings from all the data obtained in the pooling layer, and an output layer that outputs the data obtained in the fully connected layer. By inputting the data generated by processing the ocular image data acquired by the data acquisition unit 10 into the second trained model 212, information about the circulatory system that takes predetermined features into account is generated. The features considered by the second trained model 212 may be the same as or different from the features considered by the first trained model 211.

[0104] The data input to the second pre-trained model 212 is not limited to image data; for example, it may be numerical data, distribution data, time series data, etc. When data other than image data is input to the second pre-trained model 212, the second pre-trained model 212 is constructed according to the form of the input data and the features to be considered. For example, when processing time series data such as waveform data or time-series data, the second pre-trained model 212 may include a recurrent neural network (RNN).

[0105] If the data input to the inference processing unit 210 is video data, the trained model for processing the video data may have a structure that combines, for example, a convolutional neural network and a recurrent neural network.

[0106] The ocular blood flow data processing unit 22 may include, for example, a processor that operates according to a program created based on at least the medical knowledge described above. In this case, the ocular blood flow data processing unit 22 can generate information about the patient's circulatory system by processing the data (ocular blood flow data) acquired from the patient's fundus by the data acquisition unit 10 with at least this processor.

[0107] The ocular blood flow data input to the processor may be, for example, data acquired by optical coherence tomography blood flow measurement. The data acquired by optical coherence tomography blood flow measurement may be, for example, waveform image data representing the time-series changes of hemodynamics (blood flow velocity, blood flow rate, etc.), map image data representing the spatial distribution of hemodynamics, image data representing both the spatial distribution and time-series changes of hemodynamics, a series of pairs of numerical values ​​and time representing the time-series changes of hemodynamics, a series of pairs of numerical values ​​and coordinates representing the spatial distribution of hemodynamics, a series of triplets of numerical values, coordinates and time representing both the spatial distribution and time-series changes of hemodynamics. The information output from the processor may be, for example, any of the above-mentioned information regarding thrombus formation tendencies, information regarding thrombotic symptoms, information regarding the state of the circulatory system associated with infection, information indicating the state of sepsis, information indicating the state of DIC, information indicating the state of thrombosis, and information indicating the state of vascular occlusion.

[0108] The ocular blood flow data processing unit 22 may include, for example, a trained model constructed by machine learning based on medical knowledge as described above. In this case, the ocular blood flow data processing unit 22 can generate information about the patient's circulatory system by processing the data (ocular blood flow data) acquired from the patient's fundus by the data acquisition unit 10 using at least this trained model.

[0109] The ocular blood flow data input to the trained model may be data acquired by optical coherence tomography blood flow measurement, for example, as in the case of the processor described above. The information output from the trained model may also be the same information as in the case of the processor described above.

[0110] Figure 6 shows an example of an ocular blood flow data processing unit 22 configured using machine learning. In this example, the ocular blood flow data processing unit 22 includes an inference processing unit 220. The inference processing unit 220 is configured to perform inference processing to derive information about the patient's circulatory system from the ocular blood flow data acquired by the data acquisition unit 10, using a trained model constructed by machine learning using training data including clinical data (ocular blood flow data and diagnostic result data).

[0111] The ocular blood flow data included in the training data is, for example, data acquired using the same optical modality as the optical modality of the data acquisition unit 10, but other modalities may also be used. Other modalities include optical modalities different from the optical modality of the data acquisition unit 10, ultrasonic modalities, electrical modalities, magnetic modalities, electromagnetic modalities, etc. The diagnostic result data included in the training data may be, for example, data obtained by a physician or another inference model (trained model) based on the relevant ocular blood flow data.

[0112] Through machine learning (supervised learning) based on such training data, a trained model (inference model) can be created that takes ocular blood flow data acquired by the data acquisition unit 10 as input and outputs estimated diagnostic data related to the circulatory system. The training data used for machine learning may include data created by a computer based on clinical data. Machine learning may include transfer learning. The machine learning algorithm and machine learning techniques may be the same as those used in the ocular image data processing unit 21.

[0113] The inference processing unit 220 includes the trained model obtained in this manner, inputs the ocular blood flow data acquired by the data acquisition unit 10 into the trained model, and sends the estimated diagnostic data output from the trained model to the output unit 30.

[0114] Figure 7 shows an example of the configuration of the inference processing unit 220. In this example, the inference processing unit 220 includes a first trained model 221 and a second trained model 222. In some exemplary embodiments, the inference processing unit 220 may include only one of the first trained model 221 and the second trained model 222. Unless otherwise specified, various aspects of the trained model provided in the inference processing unit 220 may be the same as the corresponding aspects of the trained model provided in the inference processing unit 210.

[0115] The first pre-trained model 221 is constructed by machine learning using training data including ocular blood flow data and diagnostic result data. The first pre-trained model 221 includes, for example, models that correspond to the type (nature) of input data and the type (nature) of output data. For example, the first pre-trained model 221 may include a convolutional neural network similar to the first pre-trained model 211 of the ocular image data processing unit 21. By inputting the ocular blood flow data acquired by the data acquisition unit 10 into the first pre-trained model 221, information regarding the circulatory system is generated. The features considered by the first pre-trained model 221 may be the same as or different from the features considered by the first pre-trained model 211 of the ocular image data processing unit 21.

[0116] The second pre-trained model 222 is constructed by machine learning using training data that includes data generated by processing data acquired from the fundus using a predetermined modality, and diagnostic result data. The type of data generated by processing data acquired from the fundus can be arbitrary and may include, for example, image data, numerical data, distribution data, time series data, etc. The second pre-trained model 222 includes, for example, a model corresponding to the type (mode) of input data and the type (mode) of output data. By inputting the data obtained by processing the data acquired by the data acquisition unit 10 into the second pre-trained model 222, information regarding the circulatory system is generated. The features considered by the second pre-trained model 222 may be the same as or different from the features considered by the second pre-trained model 212 of the ocular image data processing unit 21. Furthermore, the data input to the second pre-trained model 222 may be, for example, any of the following types: ocular blood flow data generated by processing ocular blood flow data acquired from the fundus using a predetermined modality; data of a type other than ocular blood flow data generated by processing ocular blood flow data acquired from the fundus using a predetermined modality; or ocular blood flow data generated by processing data of a type other than ocular blood flow data acquired from the fundus using a predetermined modality.

[0117] The output unit 30 outputs the results of the processing performed by the data processing unit 20. The mode of output processing is arbitrary and may be, for example, transmission, display, recording, or printing. The information output by the output unit 30 may be the results of the processing performed by the data processing unit 20 itself (information about the patient's circulatory system), information including the processing results, or information obtained by processing the processing results. For example, the medical system 1 may further include a report creation unit (not shown) that creates a report based on the information about the circulatory system obtained by the data processing unit 20. In this case, the output unit 30 can output the created report.

[0118] The output unit 30 illustrated in Figure 1 includes a transmission unit 31. The transmission unit 31 transmits the results of the processing performed by the data processing unit 20 to the physician terminal 3. The physician terminal 3 is located remotely from the data acquisition unit 10.

[0119] The transmission of data from the output unit 30 to the physician terminal 3 may be direct or indirect. Direct transmission involves sending the processing results (information related to the cardiovascular system, reports, etc.) from the output unit 30 to the physician terminal 3. Indirect transmission involves sending the processing results to a device other than the physician terminal 3 (server, database, etc.) and providing the processing results to the physician terminal 3 via that device.

[0120] As in this example, by positioning the physician terminal 3 at a remote location relative to the data acquisition unit 10, and configuring the system so that the data processing unit 20 generates information (or information based thereon) based on the data acquired by the data acquisition unit 10 from the patient's fundus and provides this information to the physician terminal 3, it is possible to ensure social distancing between the physician (healthcare worker) and the patient, thereby reducing the risk of infection for the physician (healthcare worker).

[0121] <Usage patterns of medical systems> The following describes an exemplary usage of the medical system 1. The flowchart in Figure 8 shows an example of how the medical system 1 is used. In this example, a pre-trained model is used, but in examples where a pre-trained model is not used, the construction and implementation of the pre-trained model (steps S1 and S2) are unnecessary. For example, instead, a processing program is created and implemented.

[0122] (S1: Build a pre-trained model) As preparation for the operation of medical system 1, a trained model to be used in the data processing unit 20 is constructed. Note that the processing performed at this stage may be an update (parameter adjustment / update) of an already operational trained model.

[0123] (S2: Load the trained model into the data processing unit) As further preparation for the operation of medical system 1, the trained model built in step S1 is loaded into the data processing unit 20. In this process, for example, the trained model built in step S1 is transmitted to medical system 1 via a communication line.

[0124] (S3: Data is obtained from the patient's fundus.) The subjects may be, for example, patients who have been definitively diagnosed with COVID-19, or patients suspected of having COVID-19. The data acquisition unit 10 of the medical system 1 acquires data from the patient's fundus using at least one optical modality.

[0125] The data acquisition unit 10 can, for example, apply optical coherence tomography and / or color fundus photography to the fundus. The data acquired by optical coherence tomography may be, for example, 3D image data, projection image data, optical coherence tomography angiography image data, and optical coherence tomography blood flow data. The data acquired by color fundus photography may be, for example, color frontal image data representing the morphology of the fundus.

[0126] At least a portion of the inspections performed in this process may be remote inspections using the operating device 2.

[0127] (S4: Enter data into the data processing unit) The data acquired in step S3 is sent to the data processing unit 20. In this example, at least a portion of the data input to the data processing unit 20 is input to the trained model constructed in step S1.

[0128] (S5: Generate information about the circulatory system) The data processing unit 20 processes the data entered in step S4 to generate information about the patient's circulatory system. This provides, for example, at least one of the following pieces of information: information about thrombosis tendency (information about blood properties, and / or information indicating changes in blood properties due to increased coagulation and fibrinolysis); information about thrombotic symptoms (a method indicating the blood flow velocity distribution within blood vessels, and / or information about structures formed within blood vessels); information about the state (and / or changes in state) of the circulatory system associated with infection; information indicating the state of sepsis; information indicating the state of DIC; information indicating the state of thrombosis; and information indicating the state of vascular occlusion.

[0129] The information generated by the data processing unit 20 is recorded, for example, according to the data structure 100 in Figure 2. This provides a data package related to the patient's circulatory system.

[0130] (S6: Create a report) Medical system 1 (the report generation unit, not shown in the diagram above) generates a report based on the patient's circulatory system information generated in step S5.

[0131] (S7: Submit report) The transmission unit 31 of the output unit 30 transmits the report created in step S6 to the physician terminal 3 located remotely from the data acquisition unit 10, or to a computer capable of providing information to the medical terminal 3. The physician terminal 3 is not limited to a computer used by a physician, but may also be a computer used by a medical professional other than a physician (medical professional terminal).

[0132] Such a medical system 1 makes it possible to ensure social distancing between healthcare workers and patients and reduce the risk of infection from patients to healthcare workers. Furthermore, since medical system 1 is configured to acquire data from the patient's fundus using non-invasive optical modalities such as optical coherence tomography and color fundus photography, and to generate information about the patient's circulatory system from this data, it is possible to provide a technology for non-invasively detecting the state of the patient's circulatory system. The circulatory system state detected in this way is based on the aforementioned medical knowledge and / or other medical knowledge, such as symptoms, signs of worsening, and the risk of worsening.

[0133] <First Embodiment of a Healthcare System> An exemplary embodiment of the medical system 1 described above will now be described. In this embodiment, the case in which the data acquisition unit 10 performs optical coherence tomography, and in particular the case in which optical coherence tomography blood flow measurement is performed, will be described. Based on the medical knowledge described above, this embodiment is configured to generate information on the blood coagulation and fibrinolysis system from ocular blood flow data acquired by optical coherence tomography blood flow measurement. Unless otherwise specified, the medical system of this embodiment may have the same configuration as the medical system 1 described above.

[0134] Figure 9 shows an example of the configuration of the medical system according to this embodiment. The medical system 1A in this example includes a data acquisition unit 10A, a data processing unit 20A, and an output unit 30. The output unit 30 and the transmission unit 31 are the same as the output unit 30 and transmission unit 31 in the medical system 1 described above. The same applies to the operating device 2 and the physician terminal 3.

[0135] The data acquisition unit 10A is an example of the data acquisition unit 10 of the medical system 1 described above, and includes an optical coherence tomography (OCT) device 11 and a calculation unit 12.

[0136] The optical coherence tomography apparatus 11 applies a scan to the patient's fundus for optical coherence tomography blood flow measurement. The calculation unit 12 obtains ocular blood flow data based on the data collected by the optical coherence tomography apparatus 11 through this scan. The ocular blood flow data includes blood flow velocity and vessel diameter at the location where the scan is applied. The scanning method performed by the optical coherence tomography apparatus 11 and the calculation method performed by the calculation unit 12 may be any known method, for example, the method described in Japanese Patent Application Publication No. 2020-48730 can be used.

[0137] The data processing unit 20A processes the ocular blood flow data acquired by the data acquisition unit 10A in order to generate information about the blood coagulation and fibrinolysis system. An example of the configuration of the data processing unit 20A is shown in Figure 10. In this example, the data processing unit 20A includes a wall shear rate (WSR) calculation unit 231, a storage unit 232, a wall shear stress (WSS) calculation unit 233, and an information generation unit 234.

[0138] The WSR calculation unit 231 calculates the wall shear velocity (WSR) based on the blood flow velocity and vessel diameter calculated by the calculation unit 12 of the data acquisition unit 10A. The method for calculating the wall shear velocity from the blood flow velocity and vessel diameter is arbitrary, and for example, the method described in the following document can be used: Taiji Nagaoka and Akitoshi Yoshida, "Noninvasive Evaluation of Wall Shear Stress on Retinal Microcirculation in Humans," IOVS.2006, Vol.47, 1113-1119. In this document, the blood flow velocity and vessel diameter are measured using laser Doppler velocimetry (LDV), but it will be obvious to those skilled in the art that the same wall shear velocity calculation method can be applied to the blood flow velocity and vessel diameter obtained using optical coherence tomography, as in this embodiment.

[0139] Based on this literature, the optical coherence tomography device 11 of the data acquisition unit 10A collects data by applying a scan for at least one cardiac cycle. The calculation unit 12 calculates the time average (V) of the (centerline) blood flow velocity in one cardiac cycle as the blood flow velocity. mean The calculation unit 12 calculates the vessel diameter (D) from the cross-sectional image data constructed from the data collected by the scan over one cardiac cycle described above. The WSR calculation unit 231 calculates the wall shear rate (WSR) using the following formula: WSR = 8 × V mean / D.

[0140] The calculation unit 12 can calculate the vascular cross-sectional area (Area) from the cross-sectional image data constructed from the data collected during the scan over one cardiac cycle. Furthermore, the calculation unit 12 calculates the time-averaged blood flow velocity (V mean Blood flow (BF) can be calculated by multiplying the blood vessel cross-sectional area (Area) by the blood vessel cross-sectional area (BF = V). mean ×Area.

[0141] The memory unit 232 stores blood viscosity information 232a. The blood viscosity information 232a includes the blood viscosity value η. The blood viscosity value η may be an actual measured value or a standard value. Blood viscosity is measured, for example, using a conical plate viscometer. Alternatively, blood viscosity may be estimated from data obtained by blood tests such as hematocrit (Ht), red blood cell count, and red blood cell indices (mean corpuscular volume (MCV), mean corpuscular pigment (MCH), etc.). Or, blood viscosity may be estimated by substituting specified values ​​for blood parameters such as plasma viscosity. As standard values, normal values ​​or disease values ​​determined from the range of blood viscosity values ​​obtained from clinical data or experimental data can be used.

[0142] The WSS calculation unit 233 calculates the wall shear stress (WSS) based at least on the wall shear rate calculated by the WSR calculation unit 231 and the blood viscosity value included in the blood viscosity information 232a. The method for calculating the wall shear rate from the wall shear rate and blood viscosity value may be arbitrary. For example, using the method described in the above-mentioned literature (Nagaoka and Yoshida), the WSS calculation unit 233 calculates the wall shear stress (WSS) by the following equation: WSS = η × WSR.

[0143] The information generation unit 234 generates information about the patient's circulatory system based at least on the wall shear stress calculated by the WSS calculation unit 233. In this embodiment, information about the blood coagulation and fibrinolysis system can be generated as information about the patient's circulatory system.

[0144] According to Michael R. Condon et al., "Appearance of an erythrocyte population with decreased deformability and hemoglobin content following sepsis," Am J Physiol Heart Circ Physiol 284:H2177-2184, 2003, the deformability of erythrocytes is impaired and wall shear stress is increased in the blood of sepsis model animals. Increased wall shear stress promotes vascular endothelial damage and is therefore thought to be associated with a tendency to form thrombi. Based on this background, the information generation unit 234 can evaluate the wall shear stress value calculated by the WSS calculation unit 233 and generate information including the result.

[0145] Furthermore, if a standard value (default value) is adopted as the blood viscosity, that is, if η is assumed to be constant in the above formula "WSS = η × WSR", then the WSR value and the WSS value correspond one-to-one. Therefore, in this case, it is not necessary to provide the storage unit 232 and the WSS calculation unit 233. Moreover, the information generation unit 234 may be configured to generate information about the patient's circulatory system (information about the blood coagulation and fibrinolysis system) based on the wall shear velocity value calculated by the WSR calculation unit 231.

[0146] <Second Embodiment of the Healthcare System> Other exemplary embodiments of medical system 1 are described below. In this embodiment, similar to the first embodiment, the data acquisition unit 10 performs optical coherence tomography blood flow measurement and generates information about the blood coagulation and fibrinolysis system from the ocular blood flow data acquired thereby, as well as information about structures formed within blood vessels. Unless otherwise specified, the medical system of this embodiment may have the same configuration as medical system 1 and / or 1A described above.

[0147] Figure 11 shows an example of the configuration of the medical system according to this embodiment. The medical system 1B in this example includes a data acquisition unit 10B, a data processing unit 20B, and an output unit 30. The output unit 30 and the transmission unit 31 are the same as the output unit 30 and the transmission unit 31 in the medical system 1 described above. The same applies to the operating device 2 and the physician terminal 3.

[0148] The data acquisition unit 10B is an example of the data acquisition unit 10 of the medical system 1 described above, and includes an optical coherence tomography (OCT) device 13 and a blood flow information generation unit 14.

[0149] The optical coherence tomography device 13 applies a scan for optical coherence tomography blood flow measurement to the patient's fundus. The optical coherence tomography device 13 repeatedly applies the optical coherence tomography scan to a predetermined area of ​​the patient's fundus to collect time-series data. This optical coherence tomography scan includes an A scan for at least one location (A line), and may be, for example, A scans, B scans, and circle scans for multiple locations. This provides time-series data corresponding to each scan application location.

[0150] The blood flow information generation unit 14 generates ocular blood flow data based on the time series collected by the optical coherence tomography device 13. The ocular blood flow data includes blood flow information representing the spatial distribution and temporal changes of blood flow velocity. The space in which the distribution of blood flow velocity is defined may be a one-dimensional space, a two-dimensional space, or a three-dimensional space. The temporal changes in blood flow velocity are defined for each point (each position) in space. Thus, the blood flow information obtained by the blood flow information generation unit 14 represents the temporal changes in blood flow velocity at each point in the one-dimensional, two-dimensional, or three-dimensional space inside the blood vessel to which optical coherence tomography blood flow measurement is applied. Such blood flow information is described, for example, in Robert S. Reneman and Arnold PG Hoeks, "Wall shear stress as measured in vivo: consequences for the design of the arterial system," Med Biol Eng Comput (2008) 46:499-507.

[0151] The data processing unit 20B can generate information regarding the blood coagulation and fibrinolysis system by processing the ocular blood flow data (blood flow information) acquired by the data acquisition unit 10A. Furthermore, the data processing unit 20B can generate information regarding structures formed within blood vessels by processing the ocular blood flow data (blood flow information) acquired by the data acquisition unit 10A.

[0152] Figure 12 shows an example of the configuration of the data processing unit 20B. In this example, the data processing unit 20B includes a wall shear rate (WSR) information generation unit 235, a storage unit 236, a wall shear stress (WSS) information generation unit 237, and an information generation unit 238.

[0153] The WSR information generation unit 235 generates WSR information representing the spatial distribution and temporal changes of wall shear velocity (WSR) based at least on the blood flow information generated by the blood flow information generation unit 14 of the data acquisition unit 10B. The method for generating WSR information representing the spatial distribution and temporal changes of wall shear velocity from blood flow information representing the spatial distribution and temporal changes of blood flow velocity is arbitrary, and for example, the method described in the above-mentioned literature (Robert S. Reneman and Arnold PG Hoeks) can be used.

[0154] The memory unit 236 stores blood viscosity information 236a. The blood viscosity information 236a may be a single value (η), similar to the blood viscosity information 232a in the first embodiment, or it may be a distribution of blood viscosity values ​​in at least the target space (the space in which the distribution of blood flow velocity is defined). A single blood viscosity value (η) corresponds to the case where the blood viscosity distribution in the target space is uniform (constant).

[0155] The WSS information generation unit 237 generates WSS information representing the spatial distribution and temporal change of wall shear stress (WSS) based at least on the WSR information generated by the WSR information generation unit 235 and the blood viscosity distribution information 236a. The method for generating the WSS information may be arbitrary. For example, the WSS information generation unit 237 can generate WSS information by multiplying the value of the wall shear velocity and the value of the blood viscosity for each point in the target space, similar to the first embodiment.

[0156] The information generation unit 238 can generate information about structures formed within blood vessels (thrombi, thrombus formation tendency, etc.) based at least on the blood flow information generated by the blood flow information generation unit 14 of the data acquisition unit 10B and the WSS information generated by the WSS information generation unit 237. The information generation unit 238 can also generate information about structures formed within blood vessels based at least on the blood flow information generated by the blood flow information generation unit 14 of the data acquisition unit 10B and the WSR information generated by the WSR information generation unit 235. Furthermore, the information generation unit 238 can generate information about the blood coagulation and fibrinolysis system based on any of the blood flow information generated by the blood flow information generation unit 14 of the data acquisition unit 10B, the WSR information generated by the WSR information generation unit 235, and the WSS information generated by the WSS information generation unit 237 (and other information).

[0157] <Medical information processing equipment and medical systems> Examples of exemplary medical information processing devices and medical systems including them are described below. Unless otherwise specified, the elements relating to the following embodiments may be the same as the elements of any of the medical systems 1, 1A, and 1B described above.

[0158] The exemplary medical information processing device 5 shown in Figure 13 includes a data receiving unit 51, a data processing unit 52, and an output unit 53. External to the medical information processing device 5 in this embodiment, a data acquisition device 6, an operating device 7, a communication device 8, and a physician terminal 9 are provided.

[0159] The data acquisition device 6 acquires data from the patient's fundus using at least one optical method. The operating device 7 is used by a healthcare professional to operate the data acquisition device 6 (examination device). In some exemplary embodiments, the operating device 7 is located at a distance from the data acquisition device 6 and is used to remotely operate the data acquisition device 6. The communication device 8 transmits the data acquired by the data acquisition device 6 to the medical information processing device 5. In some exemplary embodiments, the physician terminal 9 is located at a distance from the data acquisition device 6.

[0160] The data receiving unit 51 of the medical information processing device 5 receives data acquired from the patient's fundus using at least one optical method. In this embodiment, the data receiving unit 51 receives data acquired by the data acquisition device 6 and transmitted by the communication device 8. In the example of Figure 13, data is sent from the data acquisition device 6 to the data receiving unit 51 via the communication device 8, but the data input method to the medical information processing device 5 is not limited to this. For example, data acquired by the data acquisition device 6 may be stored in a database, and the data may be sent from this database to the data receiving unit 51. The data receiving unit 51 may include, for example, communication equipment for connecting to a communication line, a drive device for reading data recorded on a recording medium, etc.

[0161] The data processing unit 52 is configured to process data received by the data receiving unit 51 in order to generate information about the patient's circulatory system. The output unit 53 outputs the information about the patient's circulatory system generated by the data processing unit 52. In this example, the output unit 53 includes a transmission unit 54. The transmission unit 54 can transmit the information about the patient's circulatory system generated by the data processing unit 52 to a physician's terminal 9 located remotely from the data acquisition device 6.

[0162] Any of the matters described for any of the aforementioned medical systems 1, 1A, and 1B can be combined with the medical information processing device 5 or a medical system including it.

[0163] Such a medical information processing device 5 and a medical system including it make it possible to ensure social distancing between healthcare workers and patients and reduce the risk of infection from patients to healthcare workers. Furthermore, the medical information processing device 5 and the medical system including it are configured to acquire data from the patient's fundus using non-invasive optical modalities such as optical coherence tomography and color fundus photography, and to generate information about the patient's circulatory system from this data, thereby providing a technology for non-invasively detecting the state of the patient's circulatory system. The state of the circulatory system detected in this way is based on the aforementioned medical knowledge and / or other medical knowledge, such as symptoms, signs of worsening, and the risk of worsening.

[0164] <Summary> As described above, the technology disclosed herein uses non-invasive optical ophthalmic modalities such as optical coherence tomography blood flow measurement, optical coherence tomography angiography, and color fundus photography to detect conditions related to the circulatory system, such as tendencies toward vascular inflammation, tendencies toward thrombus formation, tendencies toward sepsis, and tendencies toward DIC. For example, changes in blood properties in blood coagulation tendencies associated with infections can be detected based on blood flow velocity obtained by optical coherence tomography blood flow measurement, as well as wall shear velocity and wall shear stress determined from blood vessel diameter, or based on the spatial distribution and temporal changes of blood flow velocity in a cross-sectional area of ​​a blood vessel (temporal changes in the blood flow velocity profile). Furthermore, by inputting this data into a trained model constructed using machine learning, it is possible to output indicators related to the severity of infections, etc. This enables non-invasive early detection of changes in disease conditions and the provision of various diagnostic support information.

[0165] For example, it has been pointed out that patients with COVID-19 develop thrombi in the microvessels of various organs, primarily the lungs. Furthermore, a correlation has been noted between various vascular test values ​​indicating increased vascular coagulation, which is a contributing factor to thrombosis, and the severity of the disease. The progression of severe COVID-19 is thought to be as follows: (1) infection; (2) rapid immune response and inflammation; (3) DIC; (4) angiogenesis in multiple organs; (5) death due to cerebral infarction, myocardial infarction, multiple organ failure, etc. Cases in which (2) to (4) progress rapidly are also known. In (2) rapid immune response and inflammation, various coagulation abnormalities are detected by blood tests, such as a decrease in platelets, an increase in D-dimer, a decrease in fibrinogen, and an extension of prothrombin time (PT time).

[0166] The technology disclosed herein enables the non-invasive detection of hemodynamics in retinal blood vessels (blood flow velocity, blood flow rate, blood flow waveform shape, etc.) and the early detection of the risk of severe COVID-19 by evaluating wall shear stress and thrombosis (peripheral vascular occlusion) from the data.

[0167] When performing evaluations using a pre-trained model built with machine learning, for example, the correlation between various blood test values ​​and hemodynamic information (blood flow velocity, blood flow rate, blood flow waveform shape, etc.) is used as training data. This allows for the construction of a pre-trained model that takes hemodynamic information as input and outputs blood test values. By inputting ocular blood flow data (hemodynamic information) obtained from the patient's fundus using optical coherence tomography into this pre-trained model, it becomes possible to estimate blood test values.

[0168] The combination of input and output data is not limited to this example and can be arbitrarily determined based on medical knowledge and background. For example, the input data may include color fundus image data, optical coherence tomography angiography image data, and other optical coherence tomography image data (e.g., morphological image data and / or functional image data). The output data may include severity, magnitude of severity risk, and numerical values ​​from tests other than blood tests.

[0169] Thus, the technology disclosed herein makes it possible to detect abnormalities in microvessels and blood flow in diseases that cause vascular and circulatory disorders throughout the body at an early stage using a non-invasive modality. [Explanation of symbols]

[0170] 1, 1A, 1B Medical System 2 Control device 3. Doctor's terminal 10, 10A, 10B Data acquisition unit 11, 13 Optical coherence tomography device 12 Calculation Section 14 Blood flow information generation section 20, 20A, 20B Data Processing Unit 21. Ocular Image Data Processing Unit 210 Inference Processing Unit 211 First pre-trained model 212 Second pre-trained model 22. Ocular blood flow data processing unit 220 Inference Processing Unit 221 First pre-trained model 222 Second pre-trained model 231 WSR calculation section 232 Storage section 232a Blood viscosity information 233 WSS calculation section 234 Information generation section 235 WSR information generation section 236 Memory section 236a Blood viscosity information 237 WSS information generation section 238 Information generation section 30 Output section 31 Transmitter

Claims

1. A data acquisition unit that acquires data from the patient's fundus using at least one optical method, To generate information regarding the patient's circulatory system, a data processing unit processes the data acquired by the data acquisition unit. Includes, The data processing unit includes an inference processing unit that uses a trained model constructed by machine learning using training data including a second data generated by processing first data acquired from the fundus of the patient using the at least one optical method and diagnostic result data, and takes as input the data generated by processing the data acquired from the fundus of the patient by the data acquisition unit, and outputs information about the patient's circulatory system. The inference processing unit takes as input the ocular blood flow data generated by processing the data acquired from the patient's fundus by the data acquisition unit, and outputs information regarding the patient's circulatory system. Healthcare system.

2. The inference processing unit takes as input the ocular blood flow data generated by processing the data collected by the data acquisition unit by applying optical coherence tomography to the fundus of the patient, and outputs information regarding the patient's circulatory system. The medical system according to claim 1.

3. The ocular blood flow data includes any of the following: waveform image data representing the time-series changes in hemodynamics; map image data representing the spatial distribution of hemodynamics; image data representing both the spatial distribution and time-series changes in hemodynamics; a series of pairs of numerical values ​​and time representing the time-series changes in hemodynamics; a series of pairs of numerical values ​​and coordinates representing the spatial distribution of hemodynamics; and a series of pairs of numerical values, coordinates and time representing both the spatial distribution and time-series changes in hemodynamics. The medical system according to claim 2.

4. The inference processing unit outputs one of the following as information about the patient's circulatory system: information about a tendency to form thrombus, information about thrombotic symptoms, information about the state of the circulatory system associated with infection, information indicating a state related to sepsis, information indicating a state related to disseminated intravascular coagulation, information indicating a state related to thrombosis, and information indicating a state related to vascular occlusion. The medical system according to claim 3.

5. A data receiving unit that receives data acquired from the patient's fundus using at least one optical method, To generate information about the patient's circulatory system, a data processing unit processes the data received by the data receiving unit. Includes, The data processing unit includes an inference processing unit that uses a trained model constructed by machine learning using training data including a second data generated by processing first data acquired from the fundus using at least one optical method and diagnostic result data, to take data generated by processing the data received by the data receiving unit as input and output information about the patient's circulatory system. The inference processing unit takes the ocular blood flow data generated by processing the data received by the data receiving unit as input and performs an inference process that outputs information about the patient's circulatory system. Medical information processing device.

Images