Medical information processing device and program

Abstract

To reduce the burden on subjects associated with determination of disease.SOLUTION: A medical information processing device according to an embodiment comprises an acquisition unit, determination unit, and reception unit. The acquisition unit acquires test values of biomarkers collected from a test target subject. The determination unit determines a first disease type the subject may have contracted on the basis of an inference result of a first learning model obtained by entering the test values acquired by the acquisition unit to the first learning model designed to infer a first disease type contracted by a subject in response to input of test values. The reception unit receives an input of a test result associated with the first disease type in the subject tested by a method different from biomarkers. When the test result of the subject negates the determination result, the determination unit uses a second learning model obtained by excluding elements associated with inference of the first disease type from the first learning model to determine a second disease type the subjected may have contracted.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] Embodiments of the present invention relate to a medical information processing apparatus and a program. [Background Art]

[0002] In recent years, liquid biopsy, a test that uses body fluids such as blood to perform diagnosis of diseases and the like, has attracted attention as a test that imposes a low burden on a subject such as a patient. As a liquid biopsy, a technique for determining the presence or absence of a disease and the type of the disease based on detection results of biomarkers such as proteins and genes is known.

[0003] For example, conventionally, by using a trained model that has learned a relationship between biomarker detection results and a plurality of types of diseases such as cancer, a suspected disease is determined from detection values of biomarkers collected from a subject. Such a technique is known.

[0004] However, with conventional techniques, depending on the type of disease, the accuracy of determination may be lowered, and misdetermination is likely to occur. In such a case, a detailed examination that imposes a heavy burden on the subject may be required to identify the type of disease actually contracted. That is, even when liquid biopsy is used, there have been cases where the burden on the subject cannot be reduced when considered comprehensively. [Prior Art Document] [Patent Document]

[0005] [Patent Document 1] Japanese National Publication of International Patent Application No. 2020-530290

[0006] [Non-Patent Document 1] Christopher M. Bishop, "Pattern Recognition and Machine Learning," (USA), 1st edition, Springer, 2006, pp. 225-290. [Overview of the project] [Problems that the invention aims to solve]

[0007] The problem that this invention aims to solve is to reduce the burden on subjects involved in disease diagnosis. [Means for solving the problem]

[0008] The medical information processing device according to this embodiment comprises an acquisition unit, a determination unit, and a reception unit. The acquisition unit acquires biomarker test values ​​collected from a subject being tested. The determination unit receives the input of test values ​​and, based on the inference result of the first trained model obtained by inputting the test values ​​acquired by the acquisition unit to a first trained model that is functioned to infer the first disease type of the disease the subject has contracted, determines the first disease type that the subject may have contracted. The reception unit receives input of test results related to the first disease type of the subject, which are tested using a method different from that of biomarkers. Furthermore, if the test results of the subject contradict the determination result, the determination unit uses a second trained model obtained by removing the elements related to the inference of the first disease type from the first trained model to determine the second disease type that the subject may have contracted. [Brief explanation of the drawing]

[0009] [Figure 1] Figure 1 shows an example of the configuration of a medical information processing device according to the first embodiment. [Figure 2] Figure 2 shows an example of a primary decision model according to the first embodiment. [Figure 3] Figure 3 is an explanatory diagram showing an example of a method for generating a first-order decision model according to the first embodiment. [Figure 4]Figure 4 shows an example of the prediction result of the tumor development site using the determination function according to the first embodiment. [Figure 5] Figure 5 shows an example of the prediction results of the tumor development site in breast cancer patients using the determination function according to the first embodiment. [Figure 6] Figure 6 is a flowchart showing an example of a process performed by the medical information processing device according to the first embodiment. [Figure 7] Figure 7 shows an example of the configuration of a medical information processing device according to the second embodiment. [Figure 8] Figure 8 is a flowchart showing an example of a process performed by the medical information processing device according to the second embodiment. [Modes for carrying out the invention]

[0010] (First Embodiment) Figure 1 is a block diagram showing an example of the configuration of a medical information processing device 5 according to the first embodiment. For example, as shown in Figure 1, the medical information processing device 5 according to the first embodiment is included in a medical information processing system 100 that is communicably connected to a medical image diagnostic device 1, a specimen testing device 2, a terminal device 3, and a medical information storage device 4 via a network 200.

[0011] Here, each device included in the medical information processing system 100 is able to communicate with each other directly or indirectly, for example, via a hospital LAN (Local Area Network) installed within the hospital. Note that devices other than those shown may also be connected to the medical information processing system 100 shown in Figure 1 in a communication-enabled manner.

[0012] For example, the medical information processing system 100 may include various systems such as a hospital information system (HIS), a radiology information system (RIS), a diagnostic report system, a picture archiving and communication system (PACS), and a laboratory information system (LIS).

[0013] Medical imaging diagnostic device 1 captures images of a subject and collects medical images. Then, medical imaging diagnostic device 1 transmits the collected medical images to terminal device 3, medical information storage device 4, and medical information processing device 5. For example, medical imaging diagnostic device 1 may be an X-ray diagnostic device, an X-ray CT (Computed Tomography) device, an MRI (Magnetic Resonance Imaging) device, an ultrasound diagnostic device, a SPECT (Single Photon Emission Computed Tomography) device, a PET (Positron Emission Computed Tomography) device, etc.

[0014] The specimen testing device 2 is operated by a clinical laboratory technician or similar person and performs specimen tests to analyze specimens such as blood and urine obtained from patients. Specimen tests include, for example, pathological tests, hematological and biochemical tests, liquid biopsy tests, general tests of urine and stool, immunoserological tests, genetic tests, microbiological tests, and tests related to blood transfusions and organ transplants. The specimen testing device 2 then transmits the test results (measurement data, etc.) of the specimen tests to the terminal device 3, the medical information storage device 4, and the medical information processing device 5.

[0015] For example, the specimen test apparatus 2 is an automatic biochemical analyzer, an automatic immunoassay analyzer, a flow cytometer, a gene analysis apparatus, a protein analysis apparatus, an extracellular vesicle analysis apparatus, or a circulating tumor cell detection apparatus, for example. A flow cytometer is an apparatus that performs flow cytometry. Flow cytometry is an analysis method in which a suspension of a measurement target is formed into a high-speed fluid, laser light, mercury light, or the like is irradiated onto the fluid to measure scattered light and fluorescence generated thereby, and the size, amount, and the like of the measurement target are measured.

[0016] Further, the gene analysis apparatus is an apparatus that analyzes gene sequences. For example, the gene analysis apparatus is an apparatus that amplifies nucleic acid molecules extracted from a biological sample by the PCR (Polymerase Chain Reaction) method or the like to detect the presence or absence of and quantitate a specific gene sequence, or an apparatus that analyzes sequence information of nucleic acid molecules. The protein analysis apparatus is, for example, an automatic immunoassay analyzer, or a high-sensitivity protein detection apparatus (for example, the Single Molecule Assay apparatus SIMOA (registered trademark) manufactured by Quanterix), or the like.

[0017] Further, the extracellular vesicle analysis apparatus is, for example, a single extracellular vesicle analysis apparatus (for example, EXOVIEW (registered trademark) manufactured by NanoView Biosciences), or the like. Further, the circulating tumor cell detection apparatus is, for example, a circulating tumor cell detection apparatus based on the nanomagnetic bead method (for example, the CELLSEARCH (registered trademark) system manufactured by Veridex), or the like.

[0018] The terminal apparatus 3 is an apparatus operated by a doctor or a clinical laboratory technologist working in a hospital. For example, the terminal apparatus 3 is implemented by a personal computer (PC), a tablet PC, a PDA (Personal Digital Assistant), a mobile phone, or the like.

[0019] The terminal device 3 displays the medical image received from the medical image diagnostic apparatus 1, the specimen testing apparatus 2, or the medical information storage device 4 on its own display, and accepts various operations via the input interface of its own apparatus. For example, the terminal device 3 accepts input operations such as various test results and diagnostic information, and transmits the same to the medical information storage device 4 and the medical information processing apparatus 5.

[0020] The medical information storage device 4 stores various types of medical information in the medical information processing system 100. Specifically, the medical information storage device 4 stores medical information including medical information including medical images received from the medical image diagnostic apparatus 1, measurement data received from the specimen testing apparatus 2, various test results and diagnostic information received via the terminal device 3, and various information received from other devices connected to the medical information processing system 100.

[0021] Here, the medical information storage device 4 stores each piece of information included in the medical information in association with a patient ID or the like. For example, the medical information storage device 4 is implemented by a computer device such as a DB (Database) server, and stores medical information in semiconductor memory elements such as RAM (Random Access Memory) and flash memory, and storage circuits such as hard disks and optical disks.

[0022] Although one medical information storage device 4 is illustrated in FIG. 1, the embodiment is not limited thereto, and a plurality of DB servers may cooperate to function as the medical information storage device 4.

[0023] The medical information processing apparatus 5 acquires various types of information from the medical image diagnostic apparatus 1, the specimen testing apparatus 2, the terminal device 3, and the medical information storage device 4, and performs various types of information processing using the acquired information. For example, the medical information processing apparatus 5 is implemented by a computer device such as a server, a workstation, a personal computer, or a tablet terminal.

[0024] As shown in Figure 1, the medical information processing device 5 includes a communication interface 51, a storage circuit 52, an input interface 53, a display 54, and a processing circuit 55.

[0025] The communication interface 51 is connected to the processing circuit 55 and controls communication between the medical information processing system 100 and each device. Specifically, the communication interface 51 receives various types of information from each device and outputs the received information to the processing circuit 55. For example, the communication interface 51 can be implemented using a network card, network adapter, NIC (Network Interface Controller), etc.

[0026] The memory circuit 52 is connected to the processing circuit 55 and stores various types of data. Specifically, the memory circuit 52 stores various types of information received from the medical image diagnostic device 1, the terminal device 3, and the medical information storage device 4, as well as information input via the input interface 53 and the processing results of the medical information processing device 5. For example, as shown in Figure 1, the memory circuit 52 stores medical data 521, the 0th-order judgment model 522, the 1st-order judgment model 523, the 2nd-order judgment model 524, and the 3rd-order judgment model 525.

[0027] Medical data 521 includes medical images received from the medical imaging diagnostic device 1, specimen testing device 2, terminal device 3, and medical information storage device 4, as well as analysis information of medical images, measurement data from specimen testing, and medical information. For example, medical data 521 is biomarker data. Biomarker data refers to the measured values ​​of biomarkers.

[0028] Here, a biomarker refers to a substance, such as a protein, measured in bodily fluids like blood, whose concentration reflects the presence or progression of a particular disease. Generally, biomarkers are indicators of specific disease conditions or the state of an organism. For example, biomarkers are proteins or genes whose blood concentrations change in response to the presence or progression of various cancers, such as breast cancer and colorectal cancer. Biomarkers can be detected using flow cytometers or gene analyzers.

[0029] For example, known biomarkers for breast cancer include the FGFR1 and ESR1 genes, which are detected as circulating tumor DNA (ctDNA), and the tumor marker (protein) CA15-3; for colorectal cancer, the NRAS and FBXW7 genes, which are detected as ctDNA, and the tumor marker CA125; and for lung cancer, the MET and ALK genes, which are detected as ctDNA, and the tumor marker CYFRA21-1.

[0030] In this embodiment, the medical information processing device 5 determines the type of tumor (cancer) (tumor site). A tumor (cancer) is an example of a disease. The tumor site is an example of a disease type.

[0031] Furthermore, the medical data 521 is acquired from the medical imaging diagnostic device 1, specimen testing device 2, terminal device 3, and medical information storage device 4 via the communication interface 51 and stored in the memory circuit 52. In addition, the medical data 521 is acquired by processing in the processing circuit 55 and stored in the memory circuit 52.

[0032] The 0th-order judgment model 522, the 1st-order judgment model 523, the 2nd-order judgment model 524, and the 3rd-order judgment model 525 are pre-trained models generated by machine learning using the medical data 521 as training data.

[0033] Specifically, the zero-order judgment model 522, the first-order judgment model 523, the second-order judgment model 524, and the third-order judgment model 525 are generated by machine learning using the medical data of multiple subjects and the disease type of each subject as training data, and are stored in the memory circuit 52. Each of the zero-order judgment model 522, the first-order judgment model 523, the second-order judgment model 524, and the third-order judgment model 525 has different conditions for the medical data 521 used in their generation.

[0034] The zero-order determination model 522 is used to determine whether a subject has a disease based on the test values ​​of biomarkers in that subject. Hereafter, the determination of the presence or absence of disease using the zero-order determination model 522 will also be referred to as "zero-order determination".

[0035] Primary, secondary, and tertiary diagnostic models 523, 524, and 525 are used to determine the type of disease based on the biomarker test values ​​of the subject being tested, when the subject is determined to have a disease in the primary diagnostic test. Hereafter, the determination of the type of disease using primary diagnostic model 523 will be referred to as "primary diagnostic," the determination of the type of disease using secondary diagnostic model 524 will be referred to as "secondary diagnostic," and the determination of the type of disease using tertiary diagnostic model 525 will be referred to as "tertiary diagnostic." Primary, secondary, and tertiary diagnostic models 523, 524, and 525 will be described later.

[0036] In the first embodiment, the zero-order determination model 522, the first-order determination model 523, the second-order determination model 524, and the third-order determination model 525 are generated by an external device located outside the medical information processing system 100. However, the zero-order determination model 522, the first-order determination model 523, the second-order determination model 524, and the third-order determination model 525 may also be generated by the processing circuit 55.

[0037] Furthermore, the memory circuit 52 stores various programs that the processing circuit 55 reads and executes to realize various functions. For example, the memory circuit 52 can be implemented using semiconductor memory elements such as RAM (Random Access Memory) or flash memory, or by hard disks, optical discs, etc.

[0038] The input interface 53 is connected to the processing circuit 55 and receives various instructions and information inputs from the operator. Specifically, the input interface 53 converts the input operations received from the operator into electrical signals and outputs them to the processing circuit 55.

[0039] For example, the input interface 53 can be implemented by a trackball, switch buttons, mouse, keyboard, touchpad for input operations by touching the operating surface, touchscreen with an integrated display screen and touchpad, non-contact input circuit using an optical sensor, and audio input circuit.

[0040] In this specification, the input interface 53 is not limited to those equipped with physical operating components such as a mouse or keyboard. For example, an electrical signal processing circuit that receives an electrical signal corresponding to an input operation from an external input device provided separately from the device and outputs this electrical signal to a control circuit is also included as an example of the input interface 53.

[0041] The display 54 is connected to the processing circuit 55 and displays various information and images. Specifically, the display 54 converts the information and image data sent from the processing circuit 55 into electrical signals for display and outputs them. For example, the display 54 can be implemented as an LCD monitor, a CRT (Cathode Ray Tube) monitor, a touch panel, etc.

[0042] The processing circuit 55 controls the operation of the medical information processing device 5 in response to input operations received from the operator via the input interface 53. For example, the processing circuit 55 is implemented by a processor. As shown in Figure 1, the processing circuit 55 performs a control function 551, a determination function 552, and a selection function 553. Here, the control function 551 is an example of an acquisition unit and a reception unit. The determination function 552 and the selection function 553 are examples of a determination unit.

[0043] The control function 551 controls the execution of processing in response to various requests input via the input interface 53. For example, the control function 551 controls the transmission and reception of medical images, examination results, etc. via the communication interface 51, the storage of various information in the memory circuit 52, and the display of information (for example, medical images and processing results from each function) on the display 54.

[0044] For example, the control function 551 acquires medical data from the medical imaging diagnostic device 1, the specimen testing device 2, the terminal device 3, and the medical information storage device 4, and stores it in the memory circuit 52. Specifically, the control function 551 acquires the test values ​​of biomarkers collected from the subject being tested from the specimen testing device 2, and stores them in the memory circuit 52 as medical data 521.

[0045] Furthermore, for example, the control function 551 controls the GUI for executing processing on medical data and displays the processing results of each function on the display 54.

[0046] The judgment function 552 receives the test values ​​as input to a zero-order judgment model 522, which is designed to output whether or not the subject has a disease. Based on the inference results of the zero-order judgment model 522 obtained by inputting the test values ​​acquired by the control function 551, the judgment function 552 determines whether or not the subject has a disease.

[0047] Specifically, the judgment function 552 inputs biomarker measurement data collected from the subject being tested into a zero-order judgment model 522 that has been trained using biomarker data from multiple subjects, and determines whether or not the subject has cancer (tumor) based on the inference results of the zero-order judgment model 522.

[0048] Here, we will explain the zero-order judgment model 522. The zero-order judgment model 522 is a trained model that takes biomarker data as input and outputs whether or not a disease is present as an inference result. The zero-order judgment model 522 uses biomarker data collected from multiple subjects and information indicating whether or not a subject has a disease as training data.

[0049] Furthermore, the judgment function 552 may determine whether or not the subject has a disease based on the inference result of the primary judgment model 523 obtained by inputting the test values ​​acquired by the control function 551 to the primary judgment model 523, without using the primary judgment model 522.

[0050] Furthermore, if a disease is detected, the judgment function 552 receives the test values ​​from the primary judgment model 523, which is designed to infer the type of disease the subject may have, and determines the type of disease the subject may have based on the inference result of the primary judgment model 523 obtained by inputting the test values ​​acquired by the control function 551.

[0051] Specifically, if the zero-order judgment determines that a tumor is present, the judgment function 552 inputs the biomarker measurement data collected from the subject being tested into the primary judgment model 523, which has been trained using biomarker data from multiple subjects. Based on the inference results of the primary judgment model 523, it determines the type of cancer (tumor site) that the subject may have.

[0052] Here, the primary diagnosis model 523 will be explained with reference to Figures 2 and 3. Figure 2 is a diagram showing an example of the primary diagnosis model 523. As shown in Figure 2, the primary diagnosis model 523 is a trained model that, in response to biomarker data input, outputs the disease type of the suspected disease as an inference result.

[0053] Figure 3 is an explanatory diagram showing an example of a method for generating the primary diagnosis model 523. As shown in Figure 3, the primary diagnosis model 523 uses biomarker data collected from multiple subjects and the disease type of each subject as training data. Here, it is preferable to use multiple biomarker data of different types, such as "trace proteins" and "genes," as training data.

[0054] In the case of tumors, the disease type corresponds to the site of tumor development, so it is labeled "tumor site" in Figure 3. Note that "tumor site" is information identified from the results of a definitive diagnosis such as a biopsy. Furthermore, the primary assessment model 523 uses biomarker data from all subjects collected for training, regardless of the tumor site.

[0055] By performing machine learning using the above training data, a primary decision model 523 can be generated that has learned the relationship between multiple biomarker data and tumor sites. Specifically, a trained model (primary decision model) is generated that is functionally configured to output tumor sites in response to biomarker data input.

[0056] Specifically, the external device that generates the trained model (hereinafter also referred to as the training device) inputs biomarker data such as "trace proteins" and "genes," as well as "tumor development sites," from multiple subjects into a machine learning engine as training data, and performs machine learning. The training device then generates a trained model that is configured to output tumor development sites in response to the input biomarker data.

[0057] Here, as a machine learning engine, for example, a neural network described in the publicly known non-patent document "Pattern recognition and machine learning" by Christopher M. Bishop (USA), 1st edition, Springer, 2006, pp. 225-290 can be applied.

[0058] In addition to the neural networks mentioned above, the machine learning engine may also utilize various algorithms such as deep learning, logistic regression analysis, nonlinear discriminant analysis, support vector machines (SVMs), random forests, and naive Bayes.

[0059] As a result of this machine learning, the learning device generates a primary decision model 523 that, in response to inputs such as "biomarker data (trace proteins)" and "biomarker data (genes)" that are input to the trained algorithm, outputs a numerical value as an inference result indicating the likelihood of being affected for each tumor site.

[0060] In this embodiment, the primary determination model 523 outputs a numerical value as an inference result indicating the likelihood of having each of the following cancers: breast cancer (mammary gland), colorectal cancer, liver cancer, lung cancer, ovarian cancer, pancreatic cancer, and upper gastrointestinal cancer (stomach, esophagus).

[0061] The judgment function 552 then determines which cancers are most likely to be present based on the inference results output from the primary judgment model 523. For example, if the judgment function 552 finds a cancer whose numerical value indicating the likelihood of being present exceeds a predetermined threshold, it determines that the cancer is most likely to be present.

[0062] Furthermore, the judgment function 552 may determine that there is no tumor if the threshold is not exceeded for any of the cancers. Also, the judgment function 552 may determine that the cancer with the highest value is the cancer with the highest probability of being present. Also, the judgment function 552 may determine that the cancer with the highest value is the cancer with the highest probability of being present only if the value indicating the probability of being present for any of the cancers does not exceed a predetermined threshold.

[0063] Figure 4 shows an example of the inference results for tumor site by the primary assessment model 523. As shown in Figure 4, the primary assessment model 523 outputs numerical values ​​indicating the likelihood of having a tumor for each of the following cancers: breast cancer, colorectal cancer, liver cancer, lung cancer, ovarian cancer, pancreatic cancer, and upper gastrointestinal cancer, such that the sum of the values ​​equals 1. For example, if the cancer with the highest numerical value is determined to be the cancer with the highest probability of being present, in the example in Figure 4, the assessment function 552 determines that the subject has a high probability of having breast cancer.

[0064] In this case, the subject will undergo a detailed examination for breast cancer. This detailed examination uses methods different from those used for biomarker detection. Examples of detailed breast cancer examinations include mammography, ultrasound, MRI, and PET. Examples of detailed colorectal cancer examinations include colonoscopy and X-ray examination. Examples of detailed liver cancer examinations include ultrasound, X-ray CT, and MRI.

[0065] Further detailed examinations for lung cancer include, for example, chest X-ray, X-ray CT, and PET. Further detailed examinations for ovarian cancer include, for example, ultrasound, X-ray CT, and MRI. Further detailed examinations for pancreatic cancer include, for example, X-ray CT, MRI, ultrasound, and PET. Further detailed examinations for upper gastrointestinal cancer include endoscopy, X-ray examination, X-ray CT, and PET.

[0066] The results of the detailed examination are entered by medical professionals such as clinical laboratory technologists via terminal device 3. These examination results are then received by the control function 551.

[0067] For example, in the example shown in Figure 4, the judgment function 552 determines that the subject has a high probability of having breast cancer, so the subject undergoes mammography, ultrasound, MRI, PET, etc.

[0068] Medical professionals input the results of a detailed breast cancer examination via terminal device 3. For example, if a medical professional inputs an examination result indicating the detection of breast cancer, the control function 551 accepts the input of the examination result. The judgment function 552 then confirms that the detailed examination result affirms the breast cancer detected in the initial assessment, and therefore determines that breast cancer is a cancer that the subject is highly likely to have.

[0069] Incidentally, depending on the tumor site, the accuracy of the primary diagnosis model 523 may be low, potentially leading to misdiagnosis. Therefore, even if a detailed examination is performed on the most likely tumor site, it is possible that no tumor will be detected in that area. In such cases, a detailed examination will be performed on the next most likely tumor site, but depending on the predicted tumor site, the number of detailed examinations may increase significantly. The following explanation of cases where the number of detailed examinations increases will be given using Figure 5.

[0070] Figure 5 shows an example of the tumor site determination result by the diagnostic function 552 for a breast cancer patient. For example, if the cancer with the highest value is determined as the cancer with the highest probability of being present, in the example in Figure 5, the diagnostic function 552 determines that the subject has a high probability of having pancreatic cancer.

[0071] However, since the subject is a breast cancer patient, pancreatic cancer will not be detected even with further examination. Therefore, if we proceed with further examinations in order of likelihood of developing cancer, we would have to perform further examinations for upper gastrointestinal cancer, colorectal cancer, liver cancer, and breast cancer. Repeated examinations like this would place a significant financial and physical burden on the subject. Therefore, it is desirable to reduce the number of further examinations as much as possible.

[0072] Therefore, if the results of the detailed examination of the subject contradict the results of the primary assessment, the judgment function 552 uses the secondary assessment model 524 selected by the selection function 553 to determine the type of disease the subject may have contracted.

[0073] The selection function 553, when the control function 551 receives an inspection result that rejects the primary judgment result, selects a specific secondary judgment model 524 from among multiple secondary judgment models 524 to be used for the secondary judgment. The secondary judgment models 524 are described below.

[0074] The secondary diagnosis model 524 is a trained model obtained by removing elements related to inference for specific disease types from the primary diagnosis model 523. Specifically, the secondary diagnosis model 524 is a trained model generated using the remaining training data obtained by removing elements related to inference for specific disease types from the training data used to generate the primary diagnosis model 523.

[0075] More specifically, the secondary diagnosis model 524 is a trained model generated by inputting the training data of the remaining subjects into a machine learning engine, after excluding the training data of subjects suffering from a specific disease type from the training data of multiple subjects. A corresponding secondary diagnosis model 524 is generated for each disease type determined by the primary diagnosis.

[0076] For example, let's consider the case where the "tumor site" is one of the following: breast cancer, colorectal cancer, liver cancer, lung cancer, ovarian cancer, pancreatic cancer, or upper gastrointestinal cancer. We will now explain the selection process of the secondary determination model 524 using the selection function 553.

[0077] In this case, the secondary determination model 524 consists of seven trained models that have been trained by excluding the biomarker data of patients with a specific type of cancer, such as a trained model trained on the biomarker data of patients with breast cancer, a trained model trained on the biomarker data of patients with colorectal cancer, and so on.

[0078] The selection function 553 selects a secondary determination model 524 that excludes elements related to disease type inference if the results of the detailed examination of the subject related to the primary determination contradict the determination result of the primary determination for the disease type.

[0079] For example, let's consider a case where the initial assessment by the judgment function 552 indicates a high probability of "breast cancer," and the control function 551 receives input indicating that "breast cancer" was not detected in the detailed breast cancer examination.

[0080] In this case, the selection function 553 selects the secondary determination model 524 from among the secondary determination models 524 of 7 that was trained on the biomarker data of subjects excluding the biomarker data of breast cancer patients, as the secondary determination model 524 to be used for secondary determination. Then, the determination function 552 performs secondary determination using this secondary determination model 524.

[0081] Here, the secondary determination model 524, which is trained on the biomarker data of subjects excluding the biomarker data of breast cancer patients, will not output "breast cancer" as the "tumor site" or determine that there is a high probability of "breast cancer" because the training data does not include data where the "tumor site" is "breast cancer". In other words, the possibility of the secondary determination by the determination function 552 incorrectly determining the "tumor site" as "breast cancer" is eliminated.

[0082] Furthermore, the judgment function 552 uses a pre-trained model as the secondary judgment model 524, which was trained on data excluding data from breast cancer patients that are thought to have been a factor in the misjudgment in the primary judgment. Therefore, an improvement in the accuracy of the judgment in the secondary judgment can be expected.

[0083] Furthermore, if the results of the detailed examinations from the second-stage assessment onward contradict the previously determined disease type, the judgment function 552 will use a trained model obtained by removing the elements related to the inference of the disease type from the trained model used to determine the disease type, to newly determine the disease type.

[0084] For example, if the control function 551 receives an input indicating that no tumor was detected in the detailed examination after the secondary assessment, the assessment function 552 performs a tertiary assessment using the tertiary assessment model 525 selected by the selection function 553. The tertiary assessment model 525 will be described below.

[0085] The third-order diagnosis model 525 is a trained model obtained by removing elements related to inference for specific disease types from the second-order diagnosis model 524. Specifically, the third-order diagnosis model 525 is a trained model generated using the remaining training data obtained by removing elements related to inference for specific disease types from the training data used to generate the second-order diagnosis model 524.

[0086] More specifically, the tertiary diagnosis model 525 is a trained model generated by inputting the training data of the remaining subjects, after excluding the training data of subjects suffering from a specific disease type from the training data of the secondary diagnosis model 524, into a machine learning engine. A corresponding tertiary diagnosis model 525 is generated for each disease type determined by the secondary diagnosis.

[0087] For example, let's consider the case where the "tumor site" is one of the following: breast cancer, colorectal cancer, liver cancer, lung cancer, ovarian cancer, pancreatic cancer, or upper gastrointestinal cancer. We will now explain the selection process of the tertiary determination model 525 using the selection function 553.

[0088] In this case, the tertiary determination model 525 consists of 21 trained models that have been trained by excluding biomarker data from patients with two specific cancers, such as a trained model trained on biomarker data from patients with breast cancer and colorectal cancer, a trained model trained on biomarker data from patients with breast cancer and liver cancer, and so on.

[0089] The selection function 553 selects a tertiary assessment model 525 that excludes elements related to the inference of the disease type if the results of the detailed examination of the subject related to the disease type contradict the assessment result of the secondary assessment.

[0090] For example, let's consider a case where the primary assessment by the judgment function 552 indicates a high probability of "breast cancer," the control function 551 receives input from the detailed examination for "colon cancer" stating that "colon cancer" was not detected, the secondary assessment indicates a high probability of "colon cancer," and the control function 551 receives input from the detailed examination for "colon cancer" stating that "colon cancer" was not detected.

[0091] In this case, the selection function 553 selects a tertiary determination model 525 from among the 21 tertiary determination models 525 that has been trained on the biomarker data of the subject, excluding the biomarker data of patients with "breast cancer" and "colorectal cancer," as the tertiary determination model 524 to be used for tertiary determination. The determination function 552 then performs tertiary determination using this tertiary determination model 525.

[0092] In this embodiment, if a tumor is not detected even after the detailed examination for the third-stage assessment, the assessment function 552 will confirm the diagnosis (assessment) result as "no tumor," but this is not limited to this. For example, the assessment function 552 may sequentially perform the fourth-stage assessment, fifth-stage assessment, etc., in the same manner as described above, by using pre-trained models for the fourth-stage assessment, fifth-stage assessment, etc.

[0093] Next, the processing performed by the medical information processing device 5 according to the first embodiment will be described. Figure 6 is a flowchart showing an example of the processing performed by the medical information processing device 5 according to the first embodiment.

[0094] First, the control function 551 acquires biomarker data (step S1). Specifically, the control function 551 acquires biomarker data from the specimen testing device 2 and stores it in the memory circuit 52 as medical data 521.

[0095] Next, the judgment function 552 performs a zero-order judgment to determine the presence or absence of a tumor (cancer) (step S2). Specifically, the judgment function 552 inputs the biomarker data acquired by the control function 551 into the zero-order judgment model 522 and determines the presence or absence of cancer based on the output inference result. If the zero-order judgment determines that there is no cancer (step S1: No), the judgment function 552 confirms the result of no tumor and terminates this process (step S6).

[0096] On the other hand, if cancer is determined to be present in the initial assessment (Step S1: Yes), the assessment function 552 inputs the biomarker data acquired by the control function 551 into the primary assessment model 523. Next, the assessment function 552 determines the tumor site based on the inference results of the primary assessment model 523 (Step S3). After this, the subject undergoes a detailed examination of the tumor site determined in the primary assessment using a medical imaging diagnostic device 1, etc.

[0097] Next, the control function 551 receives input of the detailed examination results for the primary determination (step S4). Specifically, the control function 551 receives input of the detailed examination results entered by a clinical laboratory technician or the like via the terminal device 3.

[0098] Next, the judgment function 552 confirms whether a tumor has been detected based on the results of the detailed examination for the primary judgment (step S5). If a tumor is detected (step S5: Yes), the judgment function 552 confirms the tumor site determined in the primary judgment as the judgment result and terminates this process (step S7).

[0099] On the other hand, if no tumor is detected (Step S5: No), the selection function 553 selects a secondary determination model 524 based on the results of the primary determination and the detailed examination (Step S8). Next, the determination function 552 performs a zero-level determination, similar to Step S2, to determine the presence or absence of a tumor (cancer) (Step S9).

[0100] If the zero-order test determines that there is no cancer (step S9: No), the process proceeds to step S6.

[0101] On the other hand, if cancer is determined to be present in the initial assessment (Step S9: Yes), the assessment function 552 inputs the information into the secondary assessment model 524 selected by the selection function 553. Then, the assessment function 552 determines the tumor site based on the inference results of the secondary assessment model 524 (Step S10). After this, the subject undergoes a detailed examination of the tumor site determined in the secondary assessment using a medical imaging diagnostic device 1, etc.

[0102] Next, the control function 551 receives input of the results of the detailed examination for the secondary determination (step S11). Then, the determination function 552 checks whether a tumor has been detected based on the results of the detailed examination for the secondary determination (step S12). If a tumor is detected (step S12: Yes), the determination function 552 proceeds to the process in step S7.

[0103] On the other hand, if no tumor is detected (Step S12: No), the selection function 553 selects the tertiary determination model 525 based on the results of the secondary determination and the detailed examination (Step S13). Next, the determination function 552 performs a zero-level determination, similar to Step S2, to confirm the presence or absence of a tumor (Step S14). If the zero-level determination determines that there is no cancer (Step S14: No), the process proceeds to Step S6.

[0104] On the other hand, if cancer is determined to be present in the initial assessment (step S14: Yes), the assessment function 552 inputs the information into the tertiary assessment model 525 selected by the selection function 553. Then, the assessment function 552 determines the tumor site based on the inference results of the tertiary assessment model 525 (step S15). After this, the subject undergoes a detailed examination of the tumor site determined in the tertiary assessment using a medical imaging diagnostic device 1, etc.

[0105] Next, the control function 551 receives input of the results of the detailed examination for the tertiary assessment (step S16). Then, the assessment function 552 checks whether a tumor has been detected based on the results of the detailed examination for the tertiary assessment (step S17). If a tumor is detected (step S17: Yes), the assessment function 552 proceeds to the process in step S7. On the other hand, if no tumor is detected (step S17: No), the process proceeds to the process in step S6.

[0106] As described above, the medical information processing device 5 according to the first embodiment performs a primary determination of the presence or absence of a tumor and the site of tumor development using a primary determination model 523 that has been trained using biomarker data of all subjects. If a tumor is not detected in a detailed examination of the site, a secondary determination is performed using a secondary determination model 524 that has been trained by excluding the biomarker data of subjects in whom a tumor developed at that site.

[0107] This makes it possible to use the secondary diagnosis model 524, which has been trained by excluding the subject's biomarker data that is thought to have caused the initial diagnosis to be misdiagnosed, for the secondary diagnosis, and is expected to improve the accuracy of determining the tumor site. Furthermore, by improving the accuracy of determining the tumor site, it is possible to reduce the number of detailed examinations performed on the subject. In other words, the medical information processing device 5 according to the first embodiment can reduce the burden on the subject regarding disease diagnosis.

[0108] (Modification 1 of the first embodiment) In the first embodiment described above, if a tumor is not detected in a detailed examination for a tumor site determined in the primary determination, a secondary determination is performed using a model trained on data excluding data from patients with tumors in that site. However, the model used for the secondary determination is not limited to this.

[0109] Therefore, in this modified example, we will describe a case in which a secondary determination is performed using a model that has been trained by excluding biomarker data that has a high contribution to the primary determination of the tumor site. In the following, we will mainly describe the differences from the embodiments described above, and will omit detailed explanations of points that are common with what has already been described. Furthermore, each embodiment described below may be implemented individually or in combination as appropriate.

[0110] The following describes the secondary determination model 524 of this modified example. The secondary determination model 524 of this modified example is a trained model that was trained using the test values ​​of the remaining types of biomarkers collected from multiple subjects, excluding the test values ​​of biomarkers related to the disease type determined in the primary determination.

[0111] More specifically, the secondary determination model 524 in this modified example is a trained model that performs machine learning by inputting "biomarker data (protein)", "biomarker data (gene)", etc., and "tumor site" as training data into a machine learning engine, excluding biomarker data that is considered to have a high specificity for a particular disease (cancer). The model is then configured to output a numerical value indicating the probability of being affected for each tumor site as an inference result, in response to the input of biomarker data.

[0112] Biomarkers that are considered to have high specificity for specific cancers include, for example, the FGFR1 gene and ESR1 gene detected as ctDNA in the blood for breast cancer, the NRAS gene and FBXW7 gene detected as ctDNA for colorectal cancer, and the MET gene and ALK gene detected as ctDNA for lung cancer.

[0113] For example, if the primary assessment by the judgment function 552 indicates a high probability of "breast cancer," but a detailed examination for "breast cancer" does not detect "breast cancer," the selection function 553 selects a pre-trained model 524 as the secondary assessment model, which is trained on data that excludes biomarker data specific to breast cancer, such as the FGFR1 gene and ESR1 gene, detected as ctDNA. The judgment function 552 then performs a secondary assessment using this secondary assessment model 524.

[0114] Here, the secondary diagnosis model 524, which was trained on data excluding biomarker data considered specific to breast cancer, is less likely to be diagnosed with "breast cancer" in the secondary diagnosis because it was trained on excluding biomarker data that had a high contribution to the primary diagnosis result of "breast cancer".

[0115] Furthermore, since the trained model 524, which was trained using data that excluded biomarker data highly contributing to misjudgments in the primary assessment, is used as the secondary assessment model, an improvement in the accuracy of the secondary assessment can be expected.

[0116] Therefore, the medical information processing device 5 according to this modified example is expected to improve the accuracy of secondary determination of tumor sites, similar to the first embodiment, and thus reduce the number of detailed examinations performed on the subject. In other words, the medical information processing device 5 according to this modified example can reduce the burden on the subject in relation to disease diagnosis.

[0117] (Modification 2 of the first embodiment) In the modified example described above, for instance, if the primary diagnosis is incorrectly identified as "breast cancer," the selection function 553 selects the secondary diagnosis model 524, which has been trained on data that excludes biomarker data considered specific to breast cancer.

[0118] However, since the training data for the secondary diagnosis model 524 includes biomarker data from breast cancer patients, the possibility of misdiagnosis as breast cancer in the secondary diagnosis cannot be ruled out.

[0119] Therefore, in this modified example, a pre-trained model is used as the secondary determination model 524, which is trained on data excluding biomarker data from patients with a specific cancer and biomarker data considered to be specific to that cancer.

[0120] According to this modified example, for instance, if a patient is misdiagnosed as having "breast cancer" in the primary assessment, a trained model that has been trained on data excluding data from breast cancer patients and biomarker data specific to breast cancer, which are considered to have contributed to the misdiagnosis in the primary assessment, is used as the secondary assessment model 524. This further improves the accuracy of the secondary assessment.

[0121] (Second Embodiment) In the first embodiment described above, a configuration was described in which the trained model used for the first-order judgment and the trained model used for the second-order judgment and subsequent judgments are different. In the second embodiment, a case is described in which the trained model used for the first-order judgment and the second-order judgment and subsequent judgments are the same trained model. In the following, we will mainly describe the differences from the embodiment described above, and will omit detailed explanations of points that are common with what has already been described. Furthermore, each embodiment described below may be implemented individually or in combination as appropriate.

[0122] Figure 7 is a block diagram showing an example of the configuration of the medical information processing device 5 according to the second embodiment. The medical information processing device 5 according to the second embodiment differs from the first embodiment in that the memory circuit 52 does not store the secondary determination model 524 and the tertiary determination model 525, the memory circuit 52 stores the processing table 526, the processing circuit 55 does not have a selection function 553, and the processing circuit 55 has a processing function 554.

[0123] The processing table 526 is a data table for processing biomarker data to be input into the primary assessment model 523 during the secondary and tertiary assessments. For example, the processing table 526 is a data table that associates tumor sites with biomarkers specific to those tumor sites and the upper limit, lower limit, or standard value of those biomarkers.

[0124] Specifically, processing table 526 is a data table that associates the tumor site, a biomarker specific to that tumor site, and the upper or lower limit of that biomarker, such as "breast cancer" - "CA15-3" - "0 U / mL". Note that there may be multiple biomarkers specific to the tumor site.

[0125] The processing function 554 invalidates the test values ​​of the biomarkers related to the disease type determined in the primary assessment from among the multiple types of biomarkers collected from the subject, if the test results of the detailed examination for the primary assessment contradict the assessment result.

[0126] Specifically, the processing function 554 processes the subject's biomarker data based on the results of the primary assessment by the judgment function 552 and the processing table 526. For example, if the primary assessment by the judgment function 552 indicates a high probability of "breast cancer," but "breast cancer" is not detected in a detailed examination for "breast cancer," the processing function 554 refers to the processing table 526 and identifies a biomarker specific to "breast cancer" (CA15-3 in the above example).

[0127] Next, the processing function 554 converts the biomarker data of the subject for the identified biomarker into a lower limit value associated with the processing table 526 (in the example above, "0 U / mL").

[0128] In this example, processing function 554 changes the biomarker data related to the tumor site ("breast cancer") determined in the primary assessment to the lower limit, but it may also be possible to change the biomarker data to the upper limit. This is because substances that are normally present in the blood may decrease in quantity and become undetectable due to the development of a tumor. Alternatively, the biomarker data may be changed to standard values, that is, the amount of the substance present in the blood to the values ​​of a healthy person (median, upper limit, lower limit of the healthy range, etc.).

[0129] Furthermore, the processing function 554 may delete the biomarker data related to the tumor site ("breast cancer") determined in the primary assessment. Note that converting biomarker data to an upper or lower limit, or deleting biomarker data, are examples of invalidating test results.

[0130] The judgment function 552 inputs the biomarker data converted by the processing function 554 into the primary judgment model 523 and performs a secondary judgment. Here, by processing the biomarker data that is considered specific to breast cancer into an upper limit, lower limit, or standard value, it is thought that the influence of biomarker data that caused the primary judgment result to be "breast cancer" can be eliminated. Therefore, an improvement in the accuracy of the secondary judgment can be expected.

[0131] Furthermore, if the results of subsequent detailed examinations negate the previously determined disease type, the judgment function 552 inputs the test values ​​of all previously determined disease types related to biomarkers from among the multiple types of biomarkers collected from the subject into the primary judgment model 523. Based on the inference results obtained from this input, the judgment function 552 newly determines the disease type.

[0132] For example, if the control function 551 receives an input of test results indicating that no tumor was detected in the detailed examination after the secondary assessment, the processing function 554 performs a tertiary assessment using the same processing as the secondary assessment. Specifically, if the primary assessment indicates a high probability of "breast cancer" and the secondary assessment indicates a high probability of "colorectal cancer," and neither cancer is detected in the detailed examination, the processing function 554 refers to the processing table 526 to identify biomarkers specific to "breast cancer" and biomarkers specific to "colorectal cancer," and converts each of the identified biomarkers into an upper limit, lower limit, or standard value.

[0133] In this embodiment, if a tumor is not detected even after the detailed examination of the third stage of assessment, the assessment function 552 will confirm the diagnosis (assessment) result as "no tumor," but it is not limited to this. For example, the assessment function 552 may sequentially perform fourth-stage assessments, fifth-stage assessments, etc., in the same manner as described above.

[0134] Next, the processing performed by the medical information processing device 5 according to the second embodiment will be described. Figure 8 is a flowchart showing an example of the processing performed by the medical information processing device 5 according to the second embodiment.

[0135] Steps S21 to S27 are the same processes as steps S1 to S7 in Figure 6, so their explanation is omitted.

[0136] The processing function 554 refers to the processing table 526 and identifies the biomarker to be processed (step S28). For example, if the initial assessment indicates a high probability of "breast cancer," but "breast cancer" is not detected in the detailed examination for "breast cancer," the processing function 554 refers to the processing table 526 and identifies a biomarker specific to "breast cancer."

[0137] Next, the processing function 554 converts (processes) the biomarker data of the subject for the identified biomarkers into upper or lower limits associated with the processing table 526 (step S29). The judgment function 552 inputs the biomarker data of the subject processed by the processing function 554 into the primary judgment model 523.

[0138] Steps S30 to S33 are the same processes as steps S10 to S12 in Figure 6, so their explanation is omitted. Also, steps S34 to S35 are the same processes as steps S28 to S29, so their explanation is omitted. Also, steps S36 to S39 are the same processes as steps S14 to S17 in Figure 6, so their explanation is omitted.

[0139] In the medical information processing device 5 according to the second embodiment, a primary determination is made using a primary determination model 523 that has been trained using biomarker data collected from multiple subjects to determine the presence or absence of a tumor and the site of tumor development. When the control function 551 receives input of an examination result indicating that a tumor could not be detected in a detailed examination of the site, the biomarker data for biomarkers specific to the site is converted to an upper limit, lower limit, or standard value, and then input into the primary determination model 523 to perform a secondary determination.

[0140] This eliminates the need to prepare a separate trained model for subsequent assessments, thus reducing the burden of model generation. Furthermore, by converting biomarker data that may have caused misjudgments in the primary assessment into values ​​that do not affect the assessment, it is believed that factors contributing to misjudgments can be eliminated, leading to an expected improvement in the accuracy of secondary assessments.

[0141] Therefore, the medical information processing device 5 according to the second embodiment is expected to improve the accuracy of secondary determination of tumor sites, similar to the first embodiment, and thus reduce the number of detailed examinations performed on the subject. In other words, the medical information processing device 5 according to the second embodiment reduces the burden of the process of generating the learning model and also reduces the burden on the subject regarding disease determination.

[0142] In the embodiments described above, the judgment function 552 was described in an example where it determines the tumor site as the disease type, but the embodiments are not limited to this. For example, the judgment function 552 may determine myocardial infarction or angina pectoris as the disease type of heart disease, or chronic obstructive pulmonary disease or interstitial pneumonia as the disease type of lung disease.

[0143] In the embodiments described above, examples were given in which the acquisition unit, determination unit, and receiving unit in this specification are implemented by the control function 551, determination function 552, selection function 553, and processing function 554 of the processing circuit 55, respectively. However, the embodiments are not limited to these. For example, in addition to being implemented by the control function 551, determination function 552, selection function 553, and processing function 554 described in the embodiments, the acquisition unit, determination unit, and receiving unit in this specification may also be implemented by hardware alone or by a combination of hardware and software.

[0144] As described above, when the processing circuit 55 is implemented by a processor, each processing function of the processing circuit 55 is stored in the memory circuit 52 in the form of a program that can be executed by a computer. The processing circuit 55 then reads each program from the memory circuit 52 and executes it, thereby realizing the function corresponding to each program. In other words, the processing circuit 55, in the state where each program has been read, will have the functions shown in the processing circuit 55 of Figure 1.

[0145] Although Figure 1 describes the implementation of each processing function by a single processor, it is also possible to configure a processing circuit by combining multiple independent processors, with each processor executing a program to realize the function. Furthermore, the processing functions of the processing circuit 55 may be implemented by appropriately distributing or integrating them across one or more processing circuits.

[0146] Furthermore, although the example shown in Figure 1 describes a single memory circuit 52 that stores programs corresponding to each processing function, it is also possible to have a configuration in which multiple memory circuits are distributed and the processing circuit reads the corresponding program from each individual memory circuit.

[0147] The term "processor" used in the above explanation refers to circuits such as CPUs (Central Processing Units), GPUs (Graphics Processing Units), Application Specific Integrated Circuits (ASICs), and Programmable Logic Devices (e.g., Simple Programmable Logic Devices (SPLDs), Complex Programmable Logic Devices (CPLDs), and Field Programmable Gate Arrays (FPGAs)).

[0148] The processor performs its functions by reading and executing the program stored in the memory circuit 52. Alternatively, instead of storing the program in the memory circuit 52, the processor may be configured to directly incorporate the program into its own circuitry. In this case, the processor performs its functions by reading and executing the program incorporated into the circuitry. Furthermore, the processor in the above-described embodiment is not limited to being configured as a single circuit; it may be configured as a single processor by combining multiple independent circuits, and its functions may be achieved through this combination.

[0149] Here, the program executed by the processor is provided pre-installed in ROM (Read Only Memory) or memory circuits. Alternatively, this program may be provided as a file in an installable or executable format on a computer-readable storage medium such as a CD (Compact Disc)-ROM, FD (Flexible Disk), CD-R (Recordable), or DVD (Digital Versatile Disc).

[0150] Furthermore, this program may be provided or distributed by being stored on a computer connected to a network such as the Internet and downloaded via the network. For example, this program is composed of modules, each containing the functional parts described above. In terms of actual hardware, the CPU reads the program from a storage medium such as ROM and executes it, thereby loading each module into main memory and creating it in main memory.

[0151] According to at least one embodiment described above, the burden on the subject involved in disease diagnosis can be reduced.

[0152] While several embodiments have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These embodiments can be implemented in a variety of other forms, and various omissions, substitutions, modifications, and combinations of embodiments are possible without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents. [Explanation of Symbols]

[0153] 100 Medical Information Processing Systems 1. Medical imaging diagnostic equipment 2. Specimen testing equipment 3 Terminal devices 4. Medical Information Storage Device 5. Medical Information Processing Device 51 Communication Interface 52 Memory circuit 521 Medical Data 522 0th Order Decision Model 523 Primary Decision Model 524 Second-order decision models 525 Third-order decision model 526 Processing Table 53 Input Interfaces 54 displays 55 Processing Circuit 551 Control Functions 552 Judgment function 553 Selection function 554 Machining functions

Claims

1. An acquisition unit that acquires test values ​​of biomarkers collected from the subject being tested, A determination unit determines the possible first disease type the subject may have, based on the inference result of the first trained model obtained by inputting the test values ​​acquired by the acquisition unit into a first trained model that is functioned to infer the first disease type of the disease the subject has contracted, upon receiving the aforementioned test values ​​as input. A reception unit that receives input of test results related to the first disease type of the subject, which are tested using a method different from the biomarker, Equipped with, If the test results of the subject rule out the first disease type, the determination unit uses a second trained model, which is a trained model that was trained using the test values ​​of the remaining types of biomarkers after excluding the test values ​​of the biomarkers related to the first disease type from the test values ​​of multiple types of biomarkers collected from multiple subjects, to determine the second disease type that the subject may have contracted. Medical information processing device.

2. The determination unit receives the test values ​​as input and, based on the inference result of the third trained model obtained by inputting the test values ​​acquired by the acquisition unit, determines whether or not the subject is suffering from the disease. If it is determined that the subject is suffering from the disease, the test values ​​acquired by the acquisition unit are input to the first trained model. The medical information processing device according to claim 1.

3. An acquisition unit that acquires test values ​​of biomarkers collected from the subject being tested, A determination unit determines the possible first disease type the subject may have, based on the inference result of the trained model obtained by inputting the test values ​​acquired by the acquisition unit into a trained model that is functionally configured to infer the first disease type of the disease the subject has contracted, upon receiving the aforementioned test values ​​as input. A reception unit that receives input of test results related to the first disease type of the subject, which are tested using a method different from the biomarker, Equipped with, If the test results of the subject negate the first disease type, the determination unit determines the second disease type that the subject may have based on the inference result obtained by inputting the test values ​​obtained by invalidating the test values ​​of the types of biomarkers related to the first disease type from among the test values ​​of multiple types of biomarkers collected from the subject into the trained model. Medical information processing device.

4. If the test results of the subject negate the second disease type, the determination unit newly determines the second disease type based on the inference result obtained by inputting the test values ​​obtained by invalidating the test values ​​of the types of biomarkers related to the first disease type and the second disease type from among the test values ​​of multiple types of biomarkers collected from the subject into the trained model. The medical information processing device according to claim 3.

5. The aforementioned disease is cancer. A medical information processing device according to any one of claims 1 to 4.

6. On the computer, An acquisition step to obtain test values ​​of biomarkers collected from the subject being tested, A determination step in which, based on the inference result of the first trained model obtained by inputting the test values ​​obtained in the acquisition step into a first trained model that is functioned to infer the first disease type of the disease the subject may have contracted upon receiving the test values, A reception step for receiving input of test results related to the first disease type of the subject, which have been tested using a method different from the biomarker; A program that executes, The determination step, if the test results of the subject rule out the first disease type, determines the second disease type that the subject may have by using a second trained model, which is a trained model that was trained using the test values ​​of the remaining types of biomarkers after excluding the test values ​​of the biomarkers related to the first disease type from the test values ​​of multiple types of biomarkers collected from multiple subjects. program.

7. On the computer, An acquisition step to obtain test values ​​of biomarkers collected from the subject being tested, A determination step in which, based on the inference result of the first trained model obtained by inputting the test values ​​obtained in the acquisition step into a first trained model that is functioned to infer the first disease type of the disease the subject may have contracted upon receiving the test values, A reception step for receiving input of test results related to the first disease type of the subject, which have been tested using a method different from the biomarker; A program that executes, The determination step involves, if the test results of the subject negate the first disease type, inputting test values ​​obtained by invalidating the test values ​​of the types of biomarkers related to the first disease type from among the test values ​​of multiple types of biomarkers collected from the subject into the first trained model, and then determining the second disease type that the subject may have contracted based on the inference results obtained. program.

Images