Computer program, method for generating large-scale language model, information processing method, and information processing device

Abstract

To provide a computer program for assisting medial care by a doctor, a method for generating a large-scale language model, an information processing method, and an information processing device.SOLUTION: A computer program makes a computer execute processing for: acquiring a prompt related to acute phase medical care and medical examination information related to a patient in an acute phase; acquiring a sentence related to an acute phase medical care according to the acquired prompt and medical examination information from a large-scale language model that uses data including a medical examination guideline about the acute phase medical care; and outputting the acquired sentence.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present invention relates to a computer program for performing medical assistance, a method for generating a large-scale language model, an information processing method, and an information processing device. [Background technology]

[0002] The acute phase is the period until the pathological condition of a patient who is in an unstable state due to an acute onset of a disease or injury, or an acute exacerbation of a chronic disease, stabilizes to a certain extent through treatment. Medical care given to patients in the acute phase is called acute phase medical care. In acute phase medical care, examinations and tests are performed according to the patient's symptoms, and treatment and other procedures are performed. Conventionally, technology has been developed to assist doctors using computers. Patent Document 1 discloses an example of technology to assist doctors using computers. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2022-170354 A Summary of the Invention [Problem to be solved by the invention]

[0004] In acute care, doctors need to quickly and accurately determine what treatment should be given to patients based on the results of examinations or tests. However, the amount of information that doctors must consider or refer to in order to make a quick and accurate decision is extremely large, and it is not easy to make a decision even for specialists. Therefore, there is a demand for technology that supports doctors' medical treatment so that doctors can make quick, accurate decisions with reproducibility.

[0005] The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide a computer program, a method for generating a large-scale language model, an information processing method, and an information processing device for assisting doctors in their medical treatment. [Means for solving the problem]

[0006] A computer program according to one embodiment of the present invention is characterized in that it causes a computer to execute a process of acquiring a prompt related to acute medical care and medical information related to a patient in an acute phase, acquiring a sentence related to acute medical care corresponding to the acquired prompt and the medical information from a large-scale language model using data including clinical guidelines related to acute medical care, and outputting the acquired sentence.

[0007] A method for generating a large-scale language model according to one embodiment of the present invention is characterized in that it acquires training data including clinical guidelines related to acute care, and generates a large-scale language model based on the training data that outputs sentences related to acute care when a prompt related to acute care including clinical information about a patient in the acute phase is input.

[0008] An information processing method according to one embodiment of the present invention is characterized in that it acquires a prompt related to acute medical care and medical information related to a patient in an acute phase, acquires a sentence related to acute medical care corresponding to the acquired prompt and the medical information from a large-scale language model using data including clinical guidelines related to acute medical care, and outputs the acquired sentence.

[0009] An information processing device according to one embodiment of the present invention includes a calculation unit that acquires a prompt related to acute medical care and medical information related to a patient in an acute phase, acquires a sentence related to acute medical care corresponding to the acquired prompt and the medical information from a large-scale language model using data including clinical guidelines related to acute medical care, and outputs the acquired sentence.

[0010] In one embodiment of the present invention, a large-scale language model is used to acquire and output sentences related to acute medical care in response to prompts related to acute medical care and medical information related to a patient in an acute phase. The output sentences conform to clinical guidelines related to acute medical care and are appropriate sentences equivalent to those written by an experienced doctor. The doctor performs medical care while referring to the output sentences, thereby providing assistance to the medical care. Effect of the Invention

[0011] The present invention has an excellent effect in that a doctor in charge of medical treatment can receive medical assistance as if he or she had received advice from an experienced specialist. [Brief description of the drawings]

[0012] [Figure 1] FIG. 1 is a schematic diagram showing a configuration example of a medical assistance system. [Diagram 2] 1 is a block diagram showing an example of the internal configuration of an information processing device according to a first embodiment. [Diagram 3] 2 is a block diagram showing an example of the internal configuration of a terminal device 2. FIG. [Figure 4] 10 is a flowchart showing an example of a procedure for medical assistance processing executed by an information processing device. [Diagram 5] 1 is a conceptual diagram showing an overview of a process for acquiring sentences related to acute medical care from a large-scale language model in embodiment 1. FIG. [Figure 6] 11 is a chart showing an example of time-series vital data. [Figure 7] 1 is a diagram showing an example of the results of enrichment analysis of hemocyte RNA sequencing. [Figure 8] This is a chart showing an example of the results of analyzing up-regulated factors by performing RNA sequencing on blood cells. [Figure 9] FIG. 13 is a schematic diagram showing an example of an input screen for initial treatment information. [Figure 10] 13 is a schematic diagram showing an example of a sentence relating to acute medical care that is output according to medical information. FIG. [Figure 11]FIG. 13 is a schematic diagram showing an example of an input screen for sequentially changing medical information. [Figure 12] 13 is a schematic diagram showing an example of a plurality of sentences relating to acute medical care that are output in accordance with updated medical information. FIG. [Figure 13] FIG. 13 is a schematic diagram showing an example in which a plurality of types of text relating to acute medical care are displayed side by side. [Figure 14] FIG. 11 is a block diagram showing an example of the internal configuration of an information processing device according to a second embodiment. [Figure 15] FIG. 11 is a conceptual diagram illustrating the function of a large-scale language model according to the second embodiment. [Figure 16] FIG. 2 is a block diagram showing an example of the internal configuration of a learning device. [Figure 17] 11 is a flowchart illustrating an example of a processing procedure executed by the learning device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] The present invention will now be described in detail with reference to the drawings showing embodiments thereof. <Embodiment 1> FIG. 1 is a schematic diagram showing an example of the configuration of a medical support system 100. The medical support system 100 is a system that performs processing to assist medical treatment by a doctor during acute care. The medical support system 100 includes an information processing device 1. The information processing device 1 is connected to a communication network 3 such as the Internet. The medical support system 100 also includes a terminal device 2. The terminal device 2 is communicatively connected to the information processing device 1 via the communication network 3 using wired communication or wireless communication.

[0014] The information processing device 1 and the terminal device 2 are configured using a computer. The terminal device 2 is operated by a user 4, such as a doctor involved in acute care. For example, the terminal device 2 is provided in a medical institution where acute care is performed. The medical assistance system 100 may include a plurality of terminal devices 2 connected to the information processing device 1 via a communication network 3. The terminal device 2 receives information about a patient in an acute phase and transmits the input information to the information processing device 1, and the information processing device 1 uses a large-scale language model to generate text related to the information from the terminal device 2. The information processing device 1 transmits the generated text to the terminal device 2, and the terminal device 2 outputs the text transmitted from the information processing device 1. The user 4 obtains the text related to the patient in an acute phase. The text obtained in this manner is used for medical assistance.

[0015] FIG. 2 is a block diagram showing an example of the internal configuration of the information processing device 1 according to the first embodiment. The information processing device 1 is configured using a computer such as a server device. The information processing device 1 executes an information processing method. The information processing device 1 includes a calculation unit 11, a memory 12, a storage unit 13, a reading unit 14, and a communication unit 15. The calculation unit 11 is configured using, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or a multi-core CPU. The calculation unit 11 may be configured using a quantum computer. The memory 12 stores temporary data generated in association with calculation. The memory 12 is, for example, a RAM (Random Access Memory). The storage unit 13 is non-volatile, for example, a hard disk or a non-volatile semiconductor memory. The reading unit 14 reads information from a recording medium 10 such as an optical disk or a portable memory. The communication unit 15 communicates with the outside of the information processing device 1 via a communication network 3 by wired communication or wireless communication. Specifically, the communication unit 15 communicates with a terminal device 2.

[0016] The calculation unit 11 causes the reading unit 14 to read the computer program 131 recorded on the recording medium 10, and stores the read computer program 131 in the storage unit 13. The calculation unit 11 executes processing for realizing the functions of the information processing device 1 in accordance with the computer program 131. The computer program 131 may be a program product. The computer program 131 may be stored in advance in the storage unit 13, or may be downloaded from outside the information processing device 1. In this case, the information processing device 1 does not need to include the reading unit 14.

[0017] The computer program 131 can be deployed to be executed on a single computer, or on multiple computers located at one site or distributed across multiple sites and interconnected by a communications network. That is, the information processing device 1 may be composed of multiple computers, and the computer program 131 may be executed on multiple computers connected via a communications network.

[0018] The information processing device 1 includes a large-scale language model 132. The large-scale language model 132 is a language generation model that, when a prompt consisting of an instruction sentence requesting some kind of answer and information accompanying the prompt are input, outputs a sentence according to the input prompt and information. For example, the large-scale language model 132 is BERT, GPT-4, Bard, LLaMA, or the like. The large-scale language model 132 is trained in advance.

[0019] The large-scale language model 132 is realized by the calculation unit 11 executing information processing according to the computer program 131. The storage unit 13 stores data for realizing the large-scale language model 132. The large-scale language model 132 may be configured with hardware. The large-scale language model 132 may be realized using a quantum computer. Alternatively, the large-scale language model 132 may be provided outside the information processing device 1, and the information processing device 1 may execute processing using the external large-scale language model 132. For example, the large-scale language model 132 may be realized using a plurality of computers connected via a communication network, or may be realized using a cloud.

[0020] The storage unit 13 stores reference data 133 including information referenced by the large-scale language model 132. The reference data 133 is information related to acute medical care. The reference data 133 may be stored in a distributed manner in multiple computers. A part or all of the reference data 133 may be stored in a computer external to the information processing device 1.

[0021] The reference data 133 includes clinical guidelines for acute medical care. For example, clinical guidelines for acute medical care include sepsis guidelines (https: / / www.jsicm.org / pdf / jjsicm28Suppl.pdf), ARDS (Acute Respiratory Distress Syndrome) guidelines (https: / / www.jsicm.org / publication / pdf / 220728JSICM_ihardsg.pdf), and trauma guidelines (https: / / www.herusu-shuppan.co.jp / 014-2 / ). The data included in the reference data 133 may be vectorized.

[0022] Furthermore, the reference data 133 may include a doctor's order book for acute medical care. The doctor's order book for acute medical care is used as a means of communicating the contents of medical assistance, treatment, care, etc. from a doctor to medical professionals such as nurses, and is determined according to the medical policy of the medical institution, etc., for each medical institution, region, etc. For example, the doctor's order book for acute medical care includes the following infusion instructions, ARDS instructions, and sepsis instructions. By including the doctor's order book for acute medical care in the reference data 133, the user 4 of the information processing device 1 can obtain documents related to patients in the acute phase that correspond in detail to the medical policy determined by the medical institution, etc. to which the user belongs.

[0023] ·Infusion instructions The amount of infusion is determined in order of priority. Priority 4: Heart rate ÷ systolic blood pressure is W 14 (predetermined value) or more, or W 14 If lactate levels are expected to be greater than X 14 (Predetermined value) or more or X 14 or if urine volume (ml / hr) is greater than or equal to body weight divided by Y 14 Less than (predetermined value) or weight divided by X 14 In these cases, lactated Ringer's solution via intravenous route should be administered in total Z 14 Administer at (predetermined value) ml / hr. Priority 3: Heart rate ÷ systolic blood pressure is W 13 (predetermined value) or more, or W 13 If lactate levels are expected to be greater than X 13 (Predetermined value) or more or X 13 or if urine volume (ml / hr) is greater than or equal to body weight divided by Y 13 Less than (predetermined value) or weight divided by Y 13 In these cases, lactated Ringer's solution via intravenous route should be administered in total Z 13 Administer at (predetermined value) ml / hr. Priority 2: Heart rate ÷ systolic blood pressure is W 12(predetermined value) or more, or W 12 If lactate levels are expected to be greater than X 12 (Predetermined value) or more or X 12 or if urine volume (ml / hr) is greater than or equal to body weight divided by Y 12 Less than (predetermined value) or weight divided by Y 12 In these cases, lactated Ringer's solution via intravenous route should be administered in total Z 12 Administer at (predetermined value) ml / hr. Priority 1: If the conditions of Priority 2, Priority 3, and Priority 4 are not met, lactated Ringer's solution should be administered intravenously. 11-1 (Specified value) ml / hr should be administered. However, for those aged 80 or older, lactated Ringer's solution should be administered at a total of Z 11-2 Administer at (predetermined value) ml / hr.

[0024] ·ARDS instructions Priority 4: Severe ARDS: PaO2 / FiO2 ratio is V 24 (Predetermined value) or less (Normal PEEP≧W in apneic state) 24 If the FiO2 is X cmH2O, the ventilator settings are as follows: 24 (Predetermined value)%, PEEP is Y 24 (Predetermined value) cmH2O, PS is Z 24 (Predetermined value) cmH2O. The amount of infusion should follow the infusion instructions. Priority 3: Moderate ARDS: PaO2 / FiO2 ratio is V 24 From V 23 (Predetermined value) (Normal PEEP ≧ W in apnea state) 23 If the FiO2 is X (predetermined value) cmH2O, the ventilator settings are as follows: 23 (Predetermined value)%, PEEP is Y 23 (Predetermined value) cmH2O, PS is Z 23 (Predetermined value) cmH2O. The amount of infusion should follow the infusion instructions. Priority 2: Mild ARDS: PaO2 / FiO2 ratio is V 23 From V 22 (Predetermined value) (Normal PEEP or CPAP in apneic state ≧ W 22If the FiO2 is X (predetermined value) cmH2O, the ventilator settings are as follows: 22 (Predetermined value)%, PEEP is Y 22 (Predetermined value) cmH2O, PS is Z 22 (Predetermined value) cmH2O. The amount of infusion should follow the infusion instructions. Priority 1: If Priority 2, Priority 3, or Priority 4 does not apply, SpO2 (percutaneous arterial oxygen saturation) is X 21 Oxygen is administered to maintain the (predetermined value)%. The amount of oxygen administered is Y 21 (Predetermined value) L / min-Z 21 (Predetermined value) L / min. The amount of infusion should conform to the infusion instructions.

[0025] ·Sepsis instructions Priority 4: In septic shock: In addition to sepsis, circulatory and cellular / metabolic abnormalities are severe enough that vasoactive agents are required to maintain blood pressure and serum lactate levels are high (U 34 If the blood pressure exceeds the prescribed level (mmol / L), the patient is in septic shock. In addition to treating sepsis, a vasopressor should be administered. The first choice is noradrenaline. 34 (Predetermined value) γ to maximum W 34 Use with a volume up to (predetermined value)γ. Target mean arterial pressure (MAP) is X 34 (Predetermined value) mmHg. If noradrenaline does not provide sufficient blood pressure, dobutamine is used as the second choice. 34 (Predetermined value) γ to maximum Z 34 Use with a volume up to (predetermined value)γ. Set the target mean arterial pressure (MAP) to X 34 The target mean arterial pressure (MAP) is set to X mmHg. Steroids are administered continuously in combination with dobutamine. 34 In mmHg.

[0026] FIG. 3 is a block diagram showing an example of the internal configuration of the terminal device 2. The terminal device 2 is a computer such as a smartphone, a tablet computer, or a personal computer. The terminal device 2 includes a calculation unit 21, a memory 22, a storage unit 23, an operation unit 24, a display unit 25, and a communication unit 26. The calculation unit 21 is configured using, for example, a CPU, a GPU, or a multi-core CPU. The calculation unit 21 may be configured using a quantum computer. The memory 22 stores temporary data generated in association with the calculation. The memory 22 is, for example, a RAM. The storage unit 23 is non-volatile, for example, a hard disk or a non-volatile semiconductor memory.

[0027] The operation unit 24 receives input of information such as text by receiving an operation from the user 4. The operation unit 24 is, for example, a touch panel, a keyboard, or a pointing device. The display unit 25 displays images. The display unit 25 is, for example, a liquid crystal display or an EL display (Electroluminescent Display). The operation unit 24 and the display unit 25 may be integrated. The communication unit 26 communicates with the outside of the terminal device 2 by wired or wireless communication. Specifically, the communication unit 26 communicates with the information processing device 1 via the communication network 3. The communication unit 26 is also capable of receiving medical information related to a patient.

[0028] The storage unit 23 stores the computer program 231. For example, the computer program 231 is downloaded using the communication unit 26 and stored in the storage unit 23. For example, the computer program 231 may be stored in the storage unit 23 in advance. For example, the computer program 231 may be read into the terminal device 2 via an interface unit (not shown) and stored in the storage unit 23. The calculation unit 21 executes processing in accordance with the computer program 231.

[0029] The medical assistance system 100 assists doctors in providing medical care during acute care. More specifically, the medical assistance system 100 receives information about a patient in the acute phase and outputs text about acute medical care according to the patient's condition. FIG. 4 is a flowchart showing an example of the procedure of medical assistance processing executed by the information processing device 1. Hereinafter, steps are abbreviated as S. The calculation unit 11 executes information processing according to the computer program 131, causing the information processing device 1 to execute the following processes.

[0030] The information processing device 1 acquires a prompt requesting text related to acute medical care, and medical information related to the patient (S11). The medical information is any information related to medical care for a specific patient, such as information for identifying the patient, the patient's medical history, the results of examining the patient, the contents of tests, the test results, the patient's vital data, the contents of treatment, or the results of treatment. For example, the medical information is initial treatment information that represents a record of the initial treatment for the patient. The initial treatment information includes the history of the initial treatment, the examination results at the time of the initial treatment, the test results at the time of the initial treatment, or the diagnosis results at the time of the initial treatment.

[0031] For example, the initial treatment information includes the following: 68-year-old male - The patient has a medical history of interstitial pneumonia and is taking one tablet of prednisone 5 mg per day. -Since yesterday, the patient had a fever and was having worsening breathing difficulties, so an ambulance was called this morning. At the time of transport to our hospital, SpO2 was 91% (under room air), respiratory rate was 30 breaths / min, heart rate was 120 beats / min, blood pressure was 90 / 40 mmHg, consciousness level was 134 on GCS, pupil diameter was approximately 3 mm on both sides, light reflex was present, and there was no quadriplegia. During initial medical examination, the patient was diagnosed with respiratory failure and placed on artificial respirator. Upon detailed medical examination, a chest X-ray showed decreased opacity on both sides of the chest. The doctor diagnosed the patient with pneumonia. A chest CT (Computed Tomography) showed infiltrative shadows on the posterior side of the lower lobes on both sides of the chest. The doctor diagnosed the patient with aspiration pneumonia. Blood gases were collected. The ventilator settings were FiO2 60%, PEEP 5cmH2O, and PaO2 80. The heart rate was 125 beats per minute, blood pressure was 87 / 35 mmHg, and the level of consciousness was 123 on the GCS. Blood gas showed a lactate level of 30.

[0032] In S11, the prompt and medical information are input to the terminal device 2. For example, the user 4, who is a doctor who has examined a patient in an acute phase at a medical institution, operates the operation unit 24 to input the prompt and medical information. For example, the calculation unit 21 displays an input image for inputting the prompt and medical information on the display unit 25, and the user 4 performs input using the input image. The user 4 may input the prompt and medical information by inputting text using the input image. The calculation unit 21 may display an input image including multiple prompts or options for medical information on the display unit 25, and the user 4 may operate the operation unit 24 to select one of the options to input the prompt or medical information. The calculation unit 21 may perform a process of adjusting the content of the input prompt so that an answer is more easily obtained, such as adding a fixed phrase, adjusting the word order, or rephrasing.

[0033] A part or all of the medical information may be input from outside the terminal device 2 through the communication unit 26. For example, the terminal device 2 may download a part or all of the medical information from a storage device (such as an electronic medical record system) that stores information about a patient, such as a medical record or test results, through the communication unit 26. A part or all of the medical information may be input through an interface unit (not shown). For example, the medical information is initial treatment information, and the prompt is a prompt requesting a differential diagnosis, such as "Please consider a differential diagnosis for this patient."

[0034] The terminal device 2 transmits the input prompt and medical information to the information processing device 1. At this time, the calculation unit 21 causes the communication unit 26 to transmit the prompt and medical information to the information processing device 1 via the communication network 3. The information processing device 1 acquires the prompt and medical information by receiving the transmitted prompt and medical information via the communication unit 15. The calculation unit 11 stores the acquired prompt and medical information in the memory 12 or the storage unit 13.

[0035] The information processing device 1 acquires sentences related to acute medical care corresponding to the acquired prompt and medical information from the large-scale language model 132 (S12). FIG. 5 is a conceptual diagram showing an overview of a process of acquiring sentences related to acute medical care from the large-scale language model 132 in the first embodiment. In S12, the calculation unit 11 searches for related information related to the acquired prompt and medical information from the reference data 133, and extracts the searched related information from the reference data 133. The calculation unit 11 may search for and extract related information related only to the medical information. For example, the computer program 131 uses ChatwithPDF, which is a plug-in for GPT. For example, the medical information is initial medical care information, and information including contents dealing with the test results or diagnosis results included in the initial medical care information is extracted as related information from the medical care guidelines included in the reference data 133.

[0036] The calculation unit 11 generates a prompt by adding medical information and related information to the acquired prompt. For example, a prompt is generated that presents related information, explains the patient's condition using the medical information, and requests the user to consider a differential diagnosis for the patient. The calculation unit 11 inputs the prompt by adding the medical information and related information to the large-scale language model 132. The large-scale language model 132 performs calculations and outputs sentences in response to the input of the prompt. Since the prompt input to the large-scale language model 132 includes content related to acute medical care, the large-scale language model 132 outputs sentences related to acute medical care.

[0037] In response to a prompt requesting the user to consider a differential diagnosis for the patient, including clinical and related information, the large-scale language model 132 outputs a sentence including the differential diagnosis. For example, the following sentence may be output: This patient's symptoms and test results suggest the following differential diagnoses: 1. Aspiration pneumonia: Chest CT scan showed bilateral lower lobe infiltrates, which is a typical initial finding of aspiration pneumonia. In addition, the patient had a medical history of interstitial pneumonia and was taking steroids, which may have increased the risk of aspiration. 2. Acute Respiratory Distress Syndrome (ARDS): The patient is experiencing dyspnea and hypoxemia (SpO2 91%), which indicates possible ARDS. Blood gas analysis shows a PaO2 of 80, and considering the ventilator settings (FiO2 60%, PEEP 5cmH2O), ARDS is likely. 3. Sepsis / Septic Shock: The patient is feverish and has a high blood gas level of lactate (30), suggesting possible sepsis or septic shock. Additionally, the patient has an increased heart rate and a decrease in blood pressure, both signs of shock. These differential diagnoses are based on the patient’s symptoms, laboratory findings, and medical history. The final diagnosis is confirmed based on further laboratory findings and the clinical course.

[0038] The calculation unit 11 acquires the sentences related to acute medical care output by the large-scale language model 132. The calculation unit 11 stores the acquired sentences related to acute medical care in the memory 12 or the storage unit 13.

[0039] The information processing device 1 outputs the sentences related to acute medical care acquired from the large-scale language model 132 (S13). In S13, the calculation unit 11 causes the communication unit 15 to transmit the acquired sentences related to acute medical care to the terminal device 2. The terminal device 2 receives the sentences related to acute medical care transmitted from the information processing device 1 via the communication unit 26. The calculation unit 21 displays an image including the received sentences related to acute medical care on the display unit 25. In this manner, the sentences related to acute medical care are output. For example, sentences including differential diagnosis results are output.

[0040] A user 4 such as a doctor checks the text related to acute medical care displayed on the display unit 25. For example, the user 4 checks the differential diagnosis result included in the displayed text. For example, a doctor who provides initial treatment to a patient in an acute phase makes a final diagnosis after referring to the outputted differential diagnosis result. The outputted differential diagnosis result conforms to the clinical guidelines related to acute medical care included in the reference data 133. If the reference data 133 includes a doctor's instruction book related to acute medical care, the content conforms to the clinical guidelines related to acute medical care and the doctor's instruction book related to acute medical care included in the reference data 133. Therefore, the differential diagnosis result is an accurate diagnosis result equivalent to a diagnosis made by an experienced doctor. By referring to such a differential diagnosis result by a doctor who provides initial treatment to a patient in an acute phase, the doctor's diagnosis is assisted as if he or she had received advice from an experienced specialist, and the doctor can make a quick and accurate diagnosis.

[0041] After S13 is completed, the medical assistance system 100 and the information processing device 1 end the medical assistance processing. The processing of S11 to S13 is performed at any time as necessary. For example, each time medical treatment is performed at a medical institution, a user inputs a prompt and medical information into the terminal device 2, and the processing of S11 to S13 is performed. The processing of S11 to S13 may be performed multiple times for the medical treatment of the same patient. After the processing of S11 to S13 is performed once, a prompt including part or all of the output text related to acute medical care may be input into the terminal device 2, and the processing of S11 to S13 may be performed again.

[0042] In S11, medical information including a medical image related to a patient may be acquired as the medical information. The medical image may be an X-ray image, a CT image, an MRI (Magnetic Resonance Imaging) image, an echo image, or an electrocardiogram. In S11, the calculation unit 11 may analyze the medical image, calculate a feature amount of the medical image, and use the calculated feature amount as the medical information. The calculation unit 11 may determine the condition of the patient, such as the presence or absence of a lesion, based on the medical image, and use the determination result as the medical information. The processes of S11 to S13 are executed, and a statement regarding acute medical care corresponding to the medical image is output. For example, if a medical image showing a test result is included in the medical information, a differential diagnosis result corresponding to the medical image is output. The test result is often represented by a medical image, and a statement regarding acute medical care can be obtained directly from the medical image without a medical professional such as a doctor having to interpret the medical image, so that a statement regarding acute medical care to assist medical care can be obtained quickly.

[0043] In S11, a prompt requesting a treatment plan may be acquired as the prompt. For example, after a sentence including a differential diagnosis result is output as a sentence related to acute medical care, a prompt requesting a treatment plan for the differential diagnosis result may be input. In this case, in S11, a prompt citing the output differential diagnosis result and including a phrase requesting a treatment plan such as "please decide on a treatment plan" is input to the terminal device 2, and the information processing device 1 acquires the prompt and medical information.

[0044] The processes of S11 to S13 are executed, and in response to the prompt requesting the treatment plan for the differential diagnosis result, a sentence including the treatment plan for the differential diagnosis result is output. For example, the output treatment plan includes the following content: The treatment for this patient will depend on the differential diagnosis and the severity of the condition. The following is a general guide: 1. Aspiration pneumonia: Antibiotic therapy is the main treatment. In the case of aspiration pneumonia, it is often caused by bacteria normally present in the oral cavity, so broad-spectrum antibiotics against these bacteria are selected. Preventive measures against aspiration are also important. 2. Acute Respiratory Distress Syndrome (ARDS): The main treatment is mechanical ventilation. Improvement of hypoxemia and lung protection strategies are important. Lung protection strategies are strategies to minimize lung damage, such as low ventilation strategies and appropriate PEEP settings. 3. Sepsis / Septic Shock: Control of the source of infection (e.g. antibiotic therapy and surgery if necessary) and management of the patient's overall condition (e.g. fluid replacement, blood pressure management, organ support) are important. Blood gas analysis shows high lactate levels, so treatment aimed at improving blood flow may be required. In addition, an increased heart rate and decreased blood pressure have been observed, so medication to maintain blood pressure (e.g. noradrenaline) may be required. These treatment plans are general and may vary depending on the specific circumstances of the patient. The final treatment plan is determined by the doctor's judgment.

[0045] A user 4 such as a doctor checks the treatment plan contained in the output text. For example, a doctor treating a patient in the acute phase refers to the output treatment plan and then decides on an actual treatment plan. The output treatment plan is equivalent to a treatment plan that would be considered by a doctor with a lot of experience. By referring to such a treatment plan, the doctor can decide on a treatment plan quickly and accurately, just as if he or she had received advice from a doctor with a lot of experience.

[0046] In S11, a prompt requesting the setting of a medical device may be acquired as the prompt. A medical device is equipment used in medical treatment such as examination or treatment. For example, a medical device is an artificial ventilator, a syringe pump, an ECMO (extracorporeal membrane oxygenation), an intraaortic balloon pump (IABP), a dialysis machine, or an electric medical bed. The medical device needs to be set according to the condition of the patient. For example, in S11, a prompt requesting the setting of a medical device such as "please give instructions on the settings of the artificial ventilator and the amount of infusion according to the priority" is input to the terminal device 2. The information processing device 1 acquires the prompt and the medical information.

[0047] The processes of S11 to S13 are executed, and a sentence including the setting of the medical device is output in response to the prompt requesting the setting of the medical device. For example, the following sentence including the setting of the medical device is output. In this patient's case, the heart rate was 120 beats per minute, blood pressure was 90 / 40mmHg, the consciousness level changed to 134 on the GCS, and the blood gas lactate level was 30, so there is a high possibility of septic shock. Therefore, the following treatment plan is proposed. Ventilator settings: FiO2: 100% PEEP: 10cmH2O PS: 25cmH2O Infusion management: Lactated Ringer's solution is administered intravenously at 300ml / hr. We also recommend initiating administration of noradrenaline to maintain blood pressure, with a target mean arterial pressure (MAP) of 65 mmHg. If noradrenaline does not provide sufficient blood pressure elevation, dobutamine is used as a second choice, at doses ranging from 1 μg to a maximum of 20 μg. In addition, we also recommend initiating continuous administration of steroids. These treatment plans are general and may vary depending on the specific circumstances of the patient. The final treatment plan is determined by the doctor's judgment.

[0048] A user 4 such as a medical professional checks the settings of the medical device contained in the output text. For example, a medical professional who handles the medical device sets up the medical device by referring to the output settings. The output settings are appropriate for the patient and are equivalent to settings considered by an experienced medical professional. The medical professional can set up the medical device appropriately.

[0049] The medical information may include time-series vital data. The time-series vital data is data that indicates changes in a patient's vital signs. For example, the vital signs are heart rate, respiratory rate, SpO2, arterial blood carbon dioxide partial pressure, blood pressure, or body temperature. The time-series vital data is acquired by repeatedly measuring the vital signs using a measuring device. In S11, the medical information including the time-series vital data and a prompt requesting a predicted value of the patient's time-series vital data may be acquired.

[0050] FIG. 6 is a chart showing an example of time-series vital data. The heart rate, systolic blood pressure, and mean blood pressure measured at each time are recorded. The time-series data consisting of the values ​​of the heart rate, systolic blood pressure, and mean blood pressure measured at a plurality of times is actually acquired time-series vital data of the patient. For example, in S11, the actually acquired time-series vital data of the patient as shown in FIG. 6 and a prompt including a word requesting a predicted value of the patient's time-series vital data, such as "Please predict the heart rate and systolic blood pressure from the time-series values," are input to the terminal device 2. The information processing device 1 acquires the prompt and medical information.

[0051] The processes of S11 to S13 are executed, and in response to the prompt requesting the predicted value of the patient's time-series vital data, a sentence including the predicted value of the patient's time-series vital data is output. For example, a sentence including the predicted value of the time-series vital data as follows is output. Based on the data provided, the following table shows the predicted values ​​for heart rate, systolic blood pressure and mean blood pressure. Time Heart rate Systolic blood pressure Mean blood pressure 15:24 105 87 52 15:25 105 87 52 15:26 105 87 52 15:27 105 87 52 15:28 105 87 52 15:29 105 87 52 15:30 105 87 52 15:31 105 87 52 15:32 105 87 52 15:33 105 87 52 15:34 105 87 52 15:35 105 87 52 15:36 105 87 52 15:37 105 87 52 15:38 105 87 52 15:39 105 87 52 However, since no data is provided after 15:24, the predicted value cannot be predicted from this point onwards. The predicted value is only a guess and may vary depending on the actual condition and progress of treatment. Therefore, please follow the doctor's instructions, properly monitor the patient's condition, and take any necessary measures.

[0052] A user 4 such as a doctor checks the predicted value of the time-series vital data included in the output text. For example, a doctor treating a patient in the acute phase considers a future treatment plan by referring to the predicted value of the output time-series vital data. By referring to the predicted value of the output time-series vital data, the doctor can consider in advance an appropriate treatment plan according to the future transition of the vital data.

[0053] In S11, a prompt requesting a treatment plan according to a time change in the patient's vital signs may be acquired as a prompt. For example, after a sentence including a predicted value of the patient's time-series vital data is output, a prompt requesting a treatment plan according to the predicted value of the output time-series vital data is input. In this case, in S11, a prompt including a word requesting a treatment plan such as "Please give instructions on the setting of the artificial ventilator and the amount of infusion in accordance with the priority, taking into consideration the change in the patient's vital signs after admission to the ICU" is input to the terminal device 2, citing the predicted value of the output time-series vital data. The information processing device 1 acquires the prompt and medical information. A prompt requesting a treatment plan according to the actually acquired time-series vital data of the patient may be acquired.

[0054] The processes of S11 to S13 are executed, and a text including a treatment plan is output in response to a prompt requesting a treatment plan according to a time change in the patient's vital signs. For example, the treatment plan to be output includes the following content: Ventilator settings: 1.FiO2: 60% 2. PEEP: 8cmH2O 3. PS: 20cmH2O Infusion volume: 1. Administer lactated Ringer's solution via the intravenous route at a total rate of 200ml / hr. These instructions are based on changes in the patient's vital signs and guidelines for ARDS and sepsis. If the patient's condition changes, these settings should be adjusted accordingly.

[0055] A user 4 such as a doctor checks the treatment plan contained in the output text. For example, a doctor treating a patient in an acute phase refers to the output treatment plan and then determines an actual treatment plan according to changes in the patient's vital signs over time. By referring to the output treatment plan, the doctor can determine an appropriate treatment plan according to changes in the patient's vital signs over time.

[0056] The information processing device 1 in this embodiment can also perform processing to output text related to acute medical care according to the analysis result of the omics information. The omics information is comprehensive information on biomolecules such as genomes. The reference data 133 includes information related to medical care using the omics information, such as a treatment plan according to the analysis result of the omics information related to a patient.

[0057] The medical information may include blood gene sequence information of the patient. For example, the result of enrichment analysis of the patient's blood cell RNA (ribonucleic acid) sequence is used as the blood gene sequence information. FIG. 7 is a diagram showing an example of the result of enrichment analysis of the blood cell RNA sequence. The analyzed activation signal and inhibitory signal are recorded. In S11, medical information including the result of enrichment analysis of the blood cell RNA sequence as shown in FIG. 7 and a prompt requesting a treatment plan according to the result of enrichment analysis of the blood cell RNA sequence may be acquired. For example, the prompt includes a word requesting a treatment plan such as "Please consider the optimal treatment plan for the patient based on the result of enrichment analysis of the blood cell RNA sequence." Such a prompt and medical information are input to the terminal device 2, and the information processing device 1 acquires the prompt and the medical information.

[0058] The processes of S11 to S13 are executed, and a text including a treatment plan is output in response to a prompt requesting a treatment plan according to the results of enrichment analysis of the patient's blood cell RNA sequence. For example, the output treatment plan may include the following content: The RNA sequencing of this patient suggests chronic inflammatory, autoimmune, and immunosuppressive responses, neutral leukocyte hyperactivity, and induction of cell death. Possible therapeutic strategies for these include: 1. Because the PD-1 and PD-L1 pathways are activated, treatment with immune checkpoint inhibitors (e.g., pembrolizumab and nivolumab) can be considered. However, caution is required as immune-related adverse events (irAEs) may occur as side effects. 2.Activation of the inflammasome pathway, TREM1 signaling, Toll-like receptor signaling, pyroptosis signaling pathway, and IL-6 signaling indicates that the patient is experiencing a chronic inflammatory response. As a treatment for this, the use of anti-inflammatory drugs (e.g., steroids, NSAIDs, IL-6 inhibitors, etc.) can be considered. 3. Activation of the Neutrophil Extracellular Trap Signaling Pathway indicates overactivity of neutral leukocytes. This is seen in some severe infections and autoimmune diseases. The use of immunosuppressants (e.g., cyclosporine) can be considered as a treatment to suppress overactivity of neutral leukocytes. 4. Inhibition of CREB Signaling in Neurons, Endocannabinoid Neuronal Synapse Pathway, IL-23 Signaling Pathway, HER-2 Signaling in Breast Cancer, and HOTAIR Regulatory Pathway indicates immunosuppressive and / or neuroinflammatory responses in patients. Treatment of these responses may include the use of immune-enhancing drugs (e.g., IL-2, interferon, etc.) or specific neuroinflammatory treatments (e.g., anti-neuroinflammatory drugs). However, these treatment plans are based on a comprehensive consideration of the RNA sequencing results and the patient's condition, and may not necessarily be applicable to all patients. When deciding on an actual treatment plan, the patient's overall condition, other medical information, possible side effects, etc. must also be taken into account.

[0059] In S11, a prompt requesting the molecular pathology analysis results and a treatment plan may be acquired as a prompt. In this case, in S11, medical information including the results of enrichment analysis of the patient's blood cell RNA sequence and a prompt requesting the molecular pathology analysis results and a treatment plan are acquired. For example, the prompt includes a word requesting the molecular pathology analysis results and a treatment plan, such as "What molecular pathology (phenotype, subtype, endotype) that has been reported so far does this correspond to, and what treatment is recommended for the corresponding molecular pathology?"

[0060] The processes of S11 to S13 are executed, and in response to a prompt requesting the molecular pathology analysis result and the treatment plan, a text including the molecular pathology analysis result and the treatment plan is output. For example, the output treatment plan includes the following content: The results of this RNA sequencing can indicate chronic inflammation, autoimmune reactions, immunosuppressive reactions, and neutral leukocyte hyperactivity. Each disease phenotype, subtype, and endotype has its own unique treatment. For example: Chronic inflammatory responses correspond to phenotypes such as rheumatoid arthritis and Crohn's disease, and anti-inflammatory drugs (e.g., steroids, NSAIDs, biologics, etc.) are used. Autoimmune reactions correspond to phenotypes such as systemic lupus erythematosus (SLE) and multiple sclerosis (MS), and immunosuppressants (e.g., cyclosporine, azathioprine, methotrexate, etc.) and immunomodulatory drugs (e.g., anti-TNF drugs, IL-6 inhibitors, etc.) are used to treat these diseases. Immunosuppressive reactions correspond to phenotypes of cancer and infectious diseases, and immune checkpoint inhibitors (e.g., pembrolizumab and nivolumab), antiviral drugs, and antibiotics are used. Neutral leukocyte hyperactivity corresponds to the phenotype of severe infections and some autoimmune diseases, and antibiotics and immunosuppressants (e.g., cyclosporine) are used to treat these conditions. However, the definitions of these phenotypes, subtypes, and endotypes, as well as the appropriate treatments, are constantly evolving due to ongoing research and change depending on the individual patient's condition and response to treatment, so the most up-to-date medical information and the patient's overall condition must be taken into consideration when selecting a specific treatment.

[0061] The medical information may include the results of performing hemocyte RNA sequencing using the patient's whole blood and analyzing upregulatory factors. FIG. 8 is a diagram showing an example of the results of performing hemocyte RNA sequencing and analyzing upregulatory factors. As a result of the analysis, activated upregulatory factors and inhibited upregulatory factors are recorded. In S11, as shown in FIG. 8, medical information including the results of performing hemocyte RNA sequencing and analyzing upregulatory factors, and a prompt requesting a treatment plan according to the results of performing hemocyte RNA sequencing and analyzing upregulatory factors may be acquired. For example, the prompt may include a word requesting a treatment plan, such as "Please consider the optimal treatment plan for the patient based on the information on the upregulatory factors revealed by the hemocyte RNA sequence." Such a prompt and medical information are input to the terminal device 2, and the information processing device 1 acquires the prompt and the medical information.

[0062] The processes of S11 to S13 are executed, and a text including a treatment plan is output in response to a prompt requesting a treatment plan according to the results of the analysis of upregulatory factors by performing hemocyte RNA sequencing. For example, the treatment plan to be output includes the following content: RNA sequencing results confirmed the activation or inhibition of several up-regulated factors in the patients' blood cells. These factors may affect the patients' pathology and provide important information for treatment selection. 1. Activation of TNF, IL1B, IL6, IL33, and IL5: These are cytokines that mainly regulate inflammation and immune responses. Their activation indicates that the patient is experiencing a chronic inflammatory response, and anti-inflammatory drugs (e.g., steroids, NSAIDs) or biologics targeting specific cytokines (e.g., TNF inhibitors, IL-6 inhibitors) can be considered. 2. Activation of CEBPA, CSF3, IKZF1, and EP300: These are factors that mainly regulate the differentiation and function of white blood cells. Their activation indicates that the patient's immune system is activated, and the use of immunosuppressants (e.g., cyclosporine, azathioprine) can be considered. 3. Suppression of SOX4, APOE, EBI3, IL7R, N6AMT1, HIVEP1, DUSP1, RBPJ, BNIP4L, CITED2: The roles of these factors are diverse, but some are involved in immune response and cell survival and proliferation. Suppression of these may indicate that the patient's immune system is suppressed. In that case, the use of immune-enhancing drugs (e.g. IL-2, interferon) can be considered.

[0063] A user 4 such as a doctor checks the output treatment plan using the blood gene sequence information of the patient as described above. The doctor refers to the output treatment plan and then decides on an actual treatment plan according to the blood gene sequence information of the patient. By referring to the output treatment plan, the doctor can decide on an appropriate treatment plan according to the blood gene sequence information of the patient.

[0064] In this embodiment, the molecular pathological analysis may be performed using analysis results other than enrichment analysis or upregulatory factor analysis of hemocyte RNA sequencing. For example, the results of activation variation analysis, pathway analysis, causal analysis, or network analysis may be used.

[0065] In this embodiment, other methods of using omics information may be used. For example, preemptive pharmacogenomics may be performed. The information processing device 1 can predict the response to each drug from the viewpoint of preemptive pharmacogenomics based on the genetic information or gene product information of the patient. This makes it possible to select the most suitable drug for each patient and maximize the therapeutic effect.

[0066] As a method of using the omics information, a patient's disease susceptibility may be determined. The information processing device 1 can predict the susceptibility or risk to a specific disease based on the patient's gene information or gene product information by using the relationship between the gene information or gene product information and the disease susceptibility. This makes it possible to preventively manage the disease risk of each individual patient.

[0067] As a method of using omics information, real-time data-based clinical research may be conducted. The medical support system 100 centralizes and accumulates information on patients, including gene information or gene product information, and enables real-time data-based clinical research, such as randomized controlled trials (RCTs), based on real-time data, by a large-scale language model 132 using the accumulated information. This allows for highly immediate clinical research using ongoing data.

[0068] As a method of using omics information, holistic treatment including gene information or gene product information may be performed. The information processing device 1 can predict not only precision treatment for diseases requiring specific treatment but also potential genetic diseases such as cancer genes by using a large-scale language model 132 using the relationship between gene information or gene product information and diseases. This enables the realization of holistic treatment, that is, treatment that takes into account the overall health condition of the patient.

[0069] The prompt acquired in S11 may be a prompt including a request for other acute medical care, and the medical information acquired in S11 may include other information regarding the patient. The processes of S11 to S13 may be repeated for the medical care of the same patient. At this time, the prompt may be changed and input to the terminal device 2, and the processes of S11 to S13 may be repeated.

[0070] Fig. 9 is a schematic diagram showing an example of an input screen for initial treatment information. In S11, an input screen including the initial treatment information as shown in Fig. 9 is displayed on the display unit 25, and the initial treatment information is input to the terminal device 2. The initial treatment information includes the patient's life history, medical history, medical image data, etc. In the example shown in Fig. 9, the input screen includes a "patient ID" and a "patient name", but from the viewpoint of strengthening the protection of personal information, a process may be performed in S11 or S12 to remove personal information that identifies the patient, such as the "patient ID" or the "patient name", from the medical information.

[0071] Fig. 10 is a schematic diagram showing an example of a sentence related to acute medical care that is output in response to medical information. Fig. 10 shows an example of a document that is output when the processes of S11 to S13 are repeatedly performed. In response to a prompt requesting a differential diagnosis, a sentence 51 including a differential diagnosis result is output. Furthermore, in response to a prompt requesting a judgment of the patient's condition, such as the severity, a sentence 51 including a judgment result of the patient's condition is output. The time when each sentence 51 is output is added to each output sentence 51.

[0072] The processes of S11 to S13 are repeated, and a plurality of sentences 51 are sequentially output. In the example shown in FIG. 10, a plurality of sentences 51 are displayed on the display unit 25. In S12, the information processing device 1 acquires each sentence 51, and in S13, the information processing device 1 transmits the sentence 51 to the terminal device 2, and the terminal device 2 displays an image including the plurality of sentences 51 on the display unit 25. The calculation unit 21 generates an image in which the plurality of sentences 51 are arranged in the order in which the information processing device 1 acquired each sentence 51, and displays the generated image on the display unit 25. Thereby, the acquired plurality of sentences 51 are displayed arranged in the order in which they were acquired. The information processing device 1 may generate an image in which the plurality of sentences 51 are arranged, transmit the generated image from the information processing device 1 to the terminal device 2, and the terminal device 2 may display the received image to display the plurality of sentences 51.

[0073] A user 4, such as a doctor, checks a plurality of sentences 51 relating to acute medical care displayed on the display unit 25. As a plurality of sentences 51 relating to acute medical care are provided in sequence, the user 4 can easily check a variety of information relating to the patient. The doctor can use the variety of information relating to the patient in treating the patient. Note that the plurality of sentences 51 may be displayed sequentially rather than simultaneously.

[0074] The medical information input in S11 may be changed sequentially, and the processes of S11 to S13 may be repeated. FIG. 11 is a schematic diagram showing an example of an input screen for medical information that is changed sequentially. In S11, an input screen including the medical information as shown in FIG. 11 is displayed on the display unit 25, and the medical information is input to the terminal device 2. The medical information includes information about the patient that changes over time, such as changes over time in the measured values ​​of a continuous monitor that measures vital signs, and changes in the displayed data of medical images. For example, a patient whose initial treatment information is shown in FIG. 9 is admitted to a hospital, and during the hospitalization, medical information as shown in FIG. 11 is acquired. For example, examinations are performed periodically, the medical information is updated, and the processes of S11 to S13 are repeated.

[0075] FIG. 12 is a schematic diagram showing an example of a plurality of sentences related to acute medical care output in response to updated medical information. FIG. 12 shows an example of a document output when medical information is updated and the processes of S11 to S13 are repeated. The processes of S11 to S13 are repeated, and a plurality of sentences 51 are sequentially acquired. The information processing device 1 sequentially transmits the plurality of sentences 51 to the terminal device 2, and the terminal device 2 generates an image in which the plurality of sentences 51 are arranged in the order in which the information processing device 1 acquired them, and displays the generated image on the display unit 25. The information processing device 1 may transmit an image in which the plurality of sentences 51 are arranged to the terminal device 2, and the terminal device 2 may display the image. A plurality of sentences 51 related to acute medical care are displayed in the order in which they were acquired. The date of illness and the time when each sentence 51 was output are added to each sentence 51. In the example shown in FIG. 12, the condition of the patient to be determined changes in response to the update of the medical information.

[0076] A user 4, such as a doctor, checks a plurality of sentences 51 related to acute medical care displayed on the display unit 25. Since information about the patient is output according to the updated medical information, the user 4 can obtain information according to the patient's condition, which changes daily. The doctor can provide medical care by referring to the latest information about the patient. Furthermore, since the plurality of sentences 51 are displayed in the order in which they were acquired, the user 4 can check the progress of changes over time in the patient's condition according to updates to the medical information. The doctor can provide appropriate medical care according to the progress of changes over time in the patient's condition. Note that the plurality of sentences 51 may be displayed sequentially, rather than simultaneously. For example, only the latest sentence 51 may be displayed.

[0077] The medical assistance system 100 may perform a process of outputting, in parallel, a plurality of types of text related to acute medical care in response to a plurality of types of prompts with different contents. FIG. 13 is a schematic diagram showing an example in which a plurality of types of text 51 related to acute medical care are displayed side by side. In S11, one of a plurality of types of prompts, such as a prompt requesting a diagnosis, a prompt requesting a treatment plan, or a prompt requesting settings of a medical device, and the latest medical information are acquired. The processes of S11 to S13 are repeatedly executed for each type of prompt. The processes of S11 to S13 may be executed in parallel for each type of prompt. By executing the processes of S11 to S13 for each type of prompt, a plurality of types of text 51 are acquired.

[0078] In S13, the information processing device 1 transmits the acquired text 51 to the terminal device 2, and the terminal device 2 generates an image in which a plurality of types of text 51 are arranged, and displays the generated image on the display unit 25. As a result, as shown in FIG. 13, a plurality of types of text 51 related to acute medical care are displayed arranged side by side. Also, in S13, the terminal device 2 displays the latest text 51 corresponding to the latest medical information as each type of text 51. The latest text 51 is the text 51 last acquired for each type of prompt. The terminal device 2 may display the medical information in addition to the plurality of types of text 51 related to acute medical care. In S13, the information processing device 1 may transmit an image in which a plurality of types of text 51 are arranged to the terminal device 2, and the terminal device 2 may display the image.

[0079] A user 4, such as a doctor, checks multiple types of text 51 related to acute medical care displayed on the display unit 25. Since multiple types of text 51 related to acute medical care, such as diagnosis results, treatment plans, or medical device settings, are displayed side by side, the user 4 can comprehensively check information related to the patient's condition. In addition, since each type of text 51 is the latest text 51, the user 4 can check information according to the latest comprehensive information related to the patient's condition. The doctor can provide appropriate medical treatment according to the latest information related to the patient's condition.

[0080] As described above in detail, in this embodiment, the information processing device 1 uses the large-scale language model 132 to acquire and output sentences related to acute medical care in accordance with prompts related to acute medical care and medical information related to a patient in an acute phase. The output sentences conform to the clinical guidelines related to acute medical care and the doctor's instruction book related to acute medical care contained in the reference data 133, and are therefore appropriate sentences equivalent to those written by an experienced doctor. The doctor performs medical care by referring to the output sentences, thereby providing assistance with medical care. The doctor performing the medical care can receive medical assistance as if he or she had received advice from an experienced doctor, and the doctor can perform medical care more quickly and accurately.

[0081] Even in a case where the reference data 133 does not include a doctor's order book for acute medical care, the output text will be in accordance with the clinical guidelines for acute medical care, and will be as appropriate as a text prepared by an experienced doctor. This allows doctors to provide quicker and more accurate medical care.

[0082] <Embodiment 2> FIG. 14 is a block diagram showing an example of the internal configuration of an information processing device 1 according to the second embodiment. The configuration of the medical support system 100 other than the information processing device 1 is the same as that of the first embodiment. The storage unit 13 does not store reference data 133. FIG. 15 is a conceptual diagram showing the function of a large-scale language model 132 according to the second embodiment. A prompt including medical information on a patient in the acute phase is input to the large-scale language model 132. The large-scale language model 132 is trained in advance so as to output a sentence related to acute medical care when a prompt including medical information is input. The configuration of other parts of the information processing device 1 is the same as that of the first embodiment.

[0083] The large-scale language model 132 according to the second embodiment is generated by fine tuning the large-scale language model. The fine tuning is performed by a learning device 6. FIG. 16 is a block diagram showing an example of the internal configuration of the learning device 6. The learning device 6 is a computer such as a server device or a personal computer. The learning device 6 includes a calculation unit 61, a memory 62, a storage unit 63, a reading unit 64, an operation unit 65, and a display unit 66. The calculation unit 61 is configured using, for example, a CPU, a GPU, or a multi-core CPU. The calculation unit 61 may be configured using a quantum computer. The memory 62 stores temporary data generated in association with the calculation. The memory 62 is, for example, a RAM. The reading unit 64 reads information from a recording medium 60 such as an optical disk or a portable memory. The storage unit 63 is non-volatile, for example, a hard disk or a non-volatile semiconductor memory.

[0084] The operation unit 65 receives information input by receiving operations from a user. The operation unit 65 is, for example, a keyboard, a pointing device, or a touch panel. The display unit 66 displays an image. The display unit 66 is, for example, a liquid crystal display or an EL display.

[0085] The calculation unit 61 causes the reading unit 64 to read the computer program 631 recorded on the recording medium 60, and stores the read computer program 631 in the storage unit 63. The calculation unit 61 executes processing for realizing the functions of the learning device 6 in accordance with the computer program 631. The computer program 631 may be a program product. The computer program 631 may be stored in advance in the storage unit 63, or may be downloaded from outside the learning device 6. In this case, the learning device 6 does not need to be equipped with the reading unit 64.

[0086] The computer program 631 can be deployed to be executed on a single computer, or on multiple computers located at one site or distributed across multiple sites and interconnected by a communication network. That is, the learning device 6 may be configured with multiple computers, and the computer program 631 may be executed on multiple computers connected via a communication network. The learning device 6 may be configured using a cloud server.

[0087] The learning device 6 executes a method for generating a large-scale language model. The learning device 6 includes a large-scale language model 632 that is the basis of the large-scale language model 132. The large-scale language model 632 is realized by the calculation unit 61 executing information processing according to a computer program 631. The storage unit 63 stores data for realizing the large-scale language model 632. The large-scale language model 632 may be configured with hardware. The large-scale language model 632 may be realized using a quantum computer. Alternatively, the large-scale language model 632 may be provided outside the learning device 6, and the information processing device 1 may execute processing using the external large-scale language model 632. For example, the large-scale language model 632 may be realized using a plurality of computers connected via a communication network, or may be realized using a cloud.

[0088] The learning device 6 generates the large-scale language model 132 by fine-tuning the large-scale language model 632. The storage unit 63 stores training data 633 for fine-tuning the large-scale language model 632. The training data 633 may be distributed and stored in multiple computers. A part or all of the training data 633 may be stored in a computer external to the learning device 6. The training data 633 is composed of information on acute medical care and includes a clinical guideline for acute medical care. The training data 633 may include a doctor's instruction book for acute medical care in addition to the clinical guideline for acute medical care. For example, the training data 633 includes multiple data sets in which a command sentence requesting some kind of answer and a prompt including clinical information are associated with a sentence for the prompt based on the clinical guideline and the doctor's instruction book.

[0089] 17 is a flowchart showing an example of the procedure of processing executed by the learning device 6. The calculation unit 61 executes information processing according to the computer program 631, whereby the learning device 6 executes the following processing. The calculation unit 61 reads out the training data 633 stored in the storage unit 63, whereby the learning device 6 acquires the training data 633 (S21). In S21, the learning device 6 may acquire the training data 633 by inputting the training data 633 from an external source.

[0090] The learning device 6 then generates the large-scale language model 132 by fine-tuning the large-scale language model 632 using the training data 633 (S22). In S22, the calculation unit 61 inputs a prompt including an instruction sentence and medical information contained in the training data 633 to the large-scale language model 632 that is the source of the large-scale language model 132. The large-scale language model 632 performs calculations in response to the input of the prompt, and outputs a sentence.

[0091] The calculation unit 61 adjusts parameters of the calculation of the large-scale language model 632 so as to reduce an error between a sentence output by the large-scale language model 632 and a sentence associated with a prompt input in the training data 633. The calculation unit 61 repeats processing using a plurality of data sets included in the training data 633 to adjust the parameters of the large-scale language model 632, thereby learning the large-scale language model 632. In this manner, the calculation unit 61 fine-tunes the large-scale language model 632.

[0092] The training data 633 includes unlabeled medical treatment guidelines and a doctor's instruction sheet, and the calculation unit 61 may fine-tune the large-scale language model 632 by performing unlabeled learning using the training data 633. The calculation unit 61 may fine-tune the large-scale language model 632 by performing reinforcement learning. For example, the user may use the operation unit 65 to provide feedback on the output of the large-scale language model 632, and the calculation unit 61 may perform reinforcement learning by adjusting parameters of the calculation using the feedback.

[0093] In this way, the calculation unit 61 generates the large-scale language model 132 according to the second embodiment by performing fine tuning of the large-scale language model 632. The calculation unit 61 stores the adjusted final parameters in the storage unit 63. After S22 ends, the learning device 6 ends the process.

[0094] The large-scale language model 132 generated by S21 to S22 is provided in the information processing device 1. For example, the final parameters adjusted in S22 are input to the information processing device 1 through the communication unit 15 or an interface (not shown) and stored in the storage unit 13. The large-scale language model 132 is realized by the calculation unit 11 executing information processing using the stored parameters. Note that the information processing device 1 may also function as the learning device 6.

[0095] The medical assistance system 100 assists a doctor in providing medical care during acute care. The information processing device 1 executes the processes of S11 to S13. As in the first embodiment, the information processing device 1 acquires a prompt requesting text related to acute care and medical information related to a patient (S11). The information processing device 1 then acquires text related to acute care corresponding to the acquired prompt and medical information from the large-scale language model 132 (S12). In S12, the calculation unit 11 generates a prompt by adding the acquired medical information to the acquired prompt, and inputs the generated prompt including the medical information to the large-scale language model 132. The large-scale language model 132 performs calculations in response to the input of the prompt, and outputs text related to acute care.

[0096] The calculation unit 11 acquires the sentence related to acute medical care output by the large-scale language model 132. The information processing device 1 outputs the sentence related to acute medical care acquired from the large-scale language model 132 (S13). In S13, similar to the first embodiment, the sentence is transmitted from the information processing device 1 to the terminal device 2, and the terminal device 2 displays an image including the sentence related to acute medical care. After S13 ends, the medical assistance system 100 and the information processing device 1 end the medical assistance process. The processes of S11 to S13 are executed at any time as necessary, similar to the first embodiment.

[0097] As described above in detail, in the second embodiment, the information processing device 1 also uses the large-scale language model 132 to acquire and output sentences related to acute medical care in response to prompts related to acute medical care and medical information related to a patient in an acute phase. The large-scale language model 132 is fine-tuned using training data 633 including clinical guidelines related to acute medical care and a doctor's instruction book related to acute medical care. Sentences output using the fine-tuned large-scale language model 132 are appropriate sentences equivalent to those written by an experienced doctor. As in the first embodiment, the doctor's medical care is assisted as if the doctor had given advice from an experienced doctor, and the doctor can provide faster and more accurate medical care.

[0098] Even in a case where the training data 633 does not include a doctor's order book for acute medical care, the sentences output using the large-scale language model 132 will conform to the clinical guidelines for acute medical care. Therefore, appropriate sentences equivalent to those written by experienced doctors are output, enabling doctors to provide prompt and accurate medical care.

[0099] In the first and second embodiments, examples in which a single terminal device 2 is used have been mainly shown. In a configuration in which the medical support system 100 includes a plurality of terminal devices 2, the processes of S11 to S13 according to the first or second embodiment may be executed in parallel using each of the terminal devices 2. In the first and second embodiments, the input of the prompt and medical information and the display of the text related to acute medical care are performed using the same terminal device 2, but the input of the prompt and medical information and the display of the text related to acute medical care may be performed using different terminal devices 2. For example, the medical support system 100 may be used for remote medical care. In the first and second embodiments, the processes are performed using the information processing device 1 and the terminal device 2, but the information processing device 1 may be configured to execute the processes of S11 to S13 independently.

[0100] The present invention is not limited to the contents of the above-described embodiment, and various modifications are possible within the scope of the claims. In other words, embodiments obtained by combining technical means appropriately modified within the scope of the claims are also included in the technical scope of the present invention.

[0101] The matters described in each embodiment can be combined with each other. In addition, the independent claims and dependent claims described in the claims can be combined with each other in any and all combinations regardless of the citation format. Furthermore, the claims use a format in which a claim cites two or more other claims (multiple claim format), but this is not limited to this. A multiple claim (multi-multi claim) that cites at least one multiple claim may also be used. [Explanation of symbols]

[0102] 100 Medical Support System 10 Recording media 1. Information processing device 11 Arithmetic section 13 Storage section 131 Computer Programs 132 Large-scale language models 133 Reference Data 2 Terminal Device 3. Communication Network 4 User 6 Learning Device

Claims

1. Acquire prompts regarding acute care and medical information regarding patients in an acute phase, Obtaining sentences related to acute medical care corresponding to the obtained prompt and the medical information from a large-scale language model using data including clinical guidelines related to acute medical care; Output the obtained sentence A computer program that causes a computer to execute a process.

2. the prompt is a prompt requesting a differential diagnosis; The medical information is initial treatment information representing a record of initial treatment for the patient, Outputting a sentence including a differential diagnosis result obtained from the large-scale language model 2. The computer program product according to claim 1, which causes a computer to execute a process.

3. After outputting the differential diagnosis result, in response to a prompt requesting a treatment plan for the differential diagnosis result, obtain a sentence including a treatment plan for the differential diagnosis result from the large-scale language model; Output a document including a treatment plan for the differential diagnosis result obtained.

3. The computer program according to claim 2, which causes a computer to execute a process.

4. the prompt is a prompt requesting configuration of a medical device; Outputting a sentence including the setting obtained from the large-scale language model 2. The computer program product according to claim 1, which causes a computer to execute a process.

5. The medical information includes time-series vital data of the patient, the prompt is a prompt requesting a predicted value of the patient's time-series vital data; Outputting a sentence including the predicted value obtained from the large-scale language model 2. The computer program product according to claim 1, which causes a computer to execute a process.

6. The medical information includes medical images relating to the patient.

2. The computer program product of claim 1.

7. The medical information includes blood gene sequence information of the patient; the prompt is a prompt requesting a course of treatment for the patient; Outputting a sentence including the treatment plan obtained from the large-scale language model 2. The computer program product according to claim 1, which causes a computer to execute a process.

8. The medical information includes a result of enrichment analysis of the patient's blood cell RNA sequence, The prompt is a prompt requesting a treatment plan according to the result, Outputting a sentence including the treatment plan obtained from the large-scale language model 8. A computer program as claimed in claim 7.

9. The medical information includes information on upregulatory factors identified by RNA sequencing of blood cells of the patient, the prompt is a prompt requesting a treatment plan depending on the information of the upregulatory factor; Outputting a sentence including the treatment plan obtained from the large-scale language model 8. A computer program as claimed in claim 7.

10. The data includes physician's orders for acute care.

2. The computer program product of claim 1.

11. Iteratively acquiring sentences related to acute care from the large-scale language model in response to the prompt and medical information; Display multiple documents related to acute medical care obtained in the order in which they were obtained 2. The computer program product according to claim 1, which causes a computer to execute a process.

12. Repeatedly acquiring from the large-scale language model a plurality of types of prompts and a plurality of types of sentences related to acute medical care in response to time-varying medical information; Display multiple types of the latest documents related to acute medical care side by side 2. The computer program product according to claim 1, which causes a computer to execute a process.

13. Obtain training data including clinical guidelines for acute care, Based on the training data, a large-scale language model is generated that outputs sentences related to acute care when a prompt related to acute care is input, the sentences including medical information related to a patient in an acute phase. A method for generating a large-scale language model.

14. The training data includes a physician's order book for acute care.

14. The method for generating a large scale language model according to claim 13.

15. Acquire prompts regarding acute care and medical information regarding patients in an acute phase, Obtaining sentences related to acute medical care corresponding to the obtained prompt and the medical information from a large-scale language model using data including clinical guidelines related to acute medical care; Output the obtained sentence 23. An information processing method comprising:

16. A calculation unit is provided, The calculation unit is Acquire prompts regarding acute care and medical information regarding patients in an acute phase, Obtaining sentences related to acute medical care corresponding to the obtained prompt and the medical information from a large-scale language model using data including clinical guidelines related to acute medical care; Output the obtained sentence 23. An information processing apparatus comprising:

Images