Interrogation method and device and storage medium

Abstract

The embodiment of the invention provides an inquiry method and device and a storage medium. The method comprises the steps that a target medical expression corresponding to the chief complaint content of a patient and consciousness state data of the patient are acquired; determining at least one target red flag symptom from a plurality of preset red flag symptoms according to the target medical expression and the consciousness state data; determining a plurality of inquiry targets according to the target medical expression and the obtained target red flag symptom, and determining an inquiry strategy and an inquiry sequence of each inquiry target according to the consciousness state data; performing interrogation on the patient according to the interrogation strategy and the interrogation sequence of each interrogation target; and outputting the inquiry medical history of the patient according to the answers of the patient to the plurality of inquiry targets. According to the embodiment of the invention, patients with unconsciousness can be accurately inquired.

Description

TECHNICAL FIELD

[0001] The embodiment of the present application relates to, but is not limited to, the technical field of medical technology, and particularly relates to a diagnosis method and device and a storage medium. BACKGROUND

[0002] In the related art, in order to adapt to the intelligent development, the existing diagnosis is mostly configured with a voice diagnosis system. The voice diagnosis system defaults that the patients for diagnosis are all conscious. However, the patients with Alzheimer's disease (AD) and other cognitive impairment often have impaired consciousness when seeking medical treatment, and thus it is difficult for them to accurately and completely describe their own symptoms, which leads to the fact that the existing voice diagnosis system cannot accurately diagnose the patients with impaired consciousness, and the treatment of the patients is delayed. SUMMARY

[0003] The embodiment of the present application provides a diagnosis method and device and a storage medium, which can accurately diagnose the patients with impaired consciousness.

[0004] In one aspect, the embodiment of the present application provides a diagnosis method, which comprises the following steps: obtaining a target medical expression corresponding to a complaint content of a patient and consciousness state data of the patient; determining at least one target red flag symptom from a plurality of preset red flag symptoms according to the target medical expression and the consciousness state data; determining a plurality of diagnosis targets according to the target medical expression and the target red flag symptom obtained, and determining a diagnosis strategy and a diagnosis order of each diagnosis target according to the consciousness state data; diagnosing the patient according to the diagnosis strategy and the diagnosis order of each diagnosis target; outputting a diagnosis history of the patient according to the answers of the patient to a plurality of diagnosis targets.

[0005] In one embodiment, the obtaining of the target medical expression corresponding to the complaint content of the patient comprises: collecting reply audio of the patient in a complaint process, and converting the reply audio into the complaint content; mapping the complaint content into a plurality of candidate medical expressions; performing constraint screening on the plurality of candidate medical expressions to obtain a pending medical expression; determining a first determination score of the pending medical expression according to a confidence of the pending medical expression and a similarity between the pending medical expression and the complaint content; in response to the first determination score being not less than a preset score threshold, determining the pending medical expression as the target medical expression.

[0006] In an embodiment, the determining, according to the target medical expression and the consciousness state data, at least one target red-flag symptom from a plurality of preset red-flag symptoms comprises: For each of the red-flag symptoms, determining a preset professional score of the red-flag symptom, a sensitivity score of the consciousness state data to the red-flag symptom, and a relevance score of the target medical expression to the red-flag symptom, and determining a second determination score of the red-flag symptom according to the preset professional score, the sensitivity score, and the relevance score; Determining the red-flag symptoms with the K highest second determination scores as the target red-flag symptoms, where K is a positive integer greater than or equal to 1.

[0007] In an embodiment, the determining, according to the preset professional score, the sensitivity score, and the relevance score, the second determination score of the red-flag symptom comprises: In response to collecting the physical sign data of the patient, determining a physical sign score for the red-flag symptom according to the physical sign data; Determining the second determination score according to the preset professional score, the sensitivity score, the relevance score, and the physical sign score.

[0008] In an embodiment, the plurality of interrogation targets comprises at least one red-flag interrogation target and at least one refined interrogation target; The determining, according to the target medical expression and the obtained target red-flag symptom, a plurality of interrogation targets comprises: Determining at least one red-flag interrogation target according to the obtained target red-flag symptom; Determining a target symptom site according to the target medical expression, and generating at least one refined interrogation target according to a preset professional cause-effect diagram corresponding to the target symptom site.

[0009] In an embodiment, the interrogating the patient according to the interrogation strategy and the interrogation order of each of the interrogation targets comprises: Obtaining current electroencephalogram data of the patient, and updating the consciousness state data according to the current electroencephalogram data; In response to the patient consciousness state represented by the updated consciousness state data being different from the patient consciousness state represented by the consciousness state data before the update, updating, according to the updated consciousness state data, the interrogation strategy and the interrogation order corresponding to each uninterrogated interrogation target; Interrogating the patient according to each uninterrogated interrogation target, the current corresponding interrogation strategy, and the current corresponding interrogation order.

[0010] In an embodiment, the plurality of interrogation targets comprises a plurality of red-flag interrogation targets; The interrogation strategy and the interrogation sequence according to each of the interrogation targets are used to interrogate the patient, including: In response to completing the interrogation of the plurality of red flag interrogation targets, and the patient feedback indicating that one or more of the target red flag symptoms exist, each uninterrogated interrogation target is updated according to the feedback of the target red flag symptom; The current electroencephalogram data of the patient is obtained, and the consciousness state data is updated according to the current electroencephalogram data; According to the updated consciousness state data, the updated each interrogation target, the current corresponding interrogation strategy and the current corresponding interrogation sequence are determined; According to the updated each interrogation target, the current corresponding interrogation strategy and the current corresponding interrogation sequence, the patient is interrogated.

[0011] In an embodiment, the interrogation history of the patient is output according to the answers of the patient to the plurality of interrogation targets, including: The diagnosis result of the patient is determined according to the answers of the patient to the plurality of interrogation targets; According to the plurality of interrogation targets and the corresponding interrogation sequence, the answers of the patient to the plurality of interrogation targets, and the consciousness state data, an interrogation evidence chain is generated; The preset history template is filled according to the target medical expression, the interrogation evidence chain and the diagnosis result, and the interrogation history is output.

[0012] In another aspect, the embodiments of the present application also provide an interrogation device, including: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the interrogation method as described above is implemented.

[0013] In another aspect, the embodiments of the present application also provide a computer readable storage medium, which stores computer executable instructions, and the computer executable instructions are used to execute the interrogation method as described above.

[0014] In another aspect, the embodiments of the present application also provide a computer program product, which includes a computer program or computer instructions stored in a computer readable storage medium, and a processor of an interrogation device reads the computer program or computer instructions from the computer readable storage medium, and the processor executes the computer program or computer instructions, so that the communication device executes the interrogation method as described above.

[0015] This application provides a consultation method, apparatus, and storage medium. The method includes: acquiring a target medical description corresponding to a patient's chief complaint and the patient's state of consciousness data; determining at least one target red flag symptom from a set of preset red flag symptoms based on the target medical description and the state of consciousness data; determining multiple consultation targets based on the target medical description and the obtained target red flag symptom, and determining a consultation strategy and consultation order for each consultation target based on the state of consciousness data; conducting a consultation with the patient based on the consultation strategy and consultation order for each consultation target; and outputting the patient's medical history based on the patient's responses to the multiple consultation targets. By determining the target red flag symptom through the target medical description corresponding to the patient's chief complaint and the state of consciousness data, the red flag symptom that the patient may actually want to express can be determined based on the patient's current state of consciousness and the expressed chief complaint. Then, by determining a consultation strategy and consultation order that matches the patient's current state of consciousness based on the state of consciousness data, the patient can be accurately consulted. Based on this, when conducting a consultation with an unconscious patient using this method, it is possible to effectively identify the symptoms that the unconscious patient actually wants to address first in their current state of consciousness, and determine a consultation strategy and sequence suitable for the patient's current state of consciousness, thereby accurately conducting a consultation with an unconscious patient. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a consultation method provided in one embodiment of this application; Figure 2 yes Figure 1 A flowchart illustrating an embodiment of a sub-step of step 110; Figure 3 yes Figure 1 A flowchart illustrating an embodiment of a sub-step of step 120; Figure 4 yes Figure 1 A flowchart illustrating an embodiment of a sub-step of step 140; Figure 5 yes Figure 1 A flowchart illustrating another sub-step embodiment of step 140; Figure 6 This is a schematic diagram of the execution framework of the consultation method provided in the embodiments of this application; Figure 7 This is a schematic diagram of the data processing chain of a consultation method provided in one embodiment of this application; Figure 8 This is a schematic diagram of the interface for selecting the affected area during a consultation, provided in an embodiment of this application. Figure 9This is a schematic diagram of the structure of a consultation device provided in one embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0018] It should be noted that although the flowchart shows a logical order, in some cases, the steps shown or described may be executed in a different order than that shown in the flowchart. In the description of the embodiments of this application, "multiple" (or more than) means two or more, "greater than," "less than," and "exceeding" are understood to exclude the number itself, while "above," "below," and "within" are understood to include the number itself. If "first," "second," etc., are described, they are only used to distinguish technical features and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or the order of the technical features indicated. Furthermore, in the description of the embodiments of this application, the various values ​​mentioned (such as first value, second value, etc.) can be flexibly represented as a single numeric code or an enumerated type value.

[0019] In related technologies, patients with Alzheimer's disease (AD) and other cognitive impairments often have difficulty accurately and completely describing their symptoms when seeking medical attention due to impaired cognitive function. For example, patients may only be able to vaguely express "my stomach...is not feeling well" or "it hurts here...", but cannot provide key information such as location, nature, severity, duration, and triggers. This expression impairment requires repeated questioning in clinical practice, and may even rely on family members to "translate" in order to piece together a relatively complete medical history.

[0020] Given the above premises, most existing online consultation systems are equipped with voice consultation systems to adapt to the development of intelligent systems. These systems generally use large language models (LLM) based on natural language processing (NLP) technology to summarize symptoms, thereby mapping symptoms to corresponding medical terms for patient consultation.

[0021] However, the LLM models used in existing voice consultation systems (such as MedMacau, ChatGLM, and multilingual machine learning models) are usually trained directly from collected corpora. Since the patient's level of consciousness is not reflected in the corpus, these systems assume that all patients are conscious and can express their symptoms stably and coherently. Therefore, when encountering patients with AD / delirium who are not fully conscious and whose expressions are disjointed or incoherent, the LLM model, guided by the patient's expression and lacking dynamic perception and contextual interaction of the patient's current cognitive state, can lead to incorrect consultations and potentially cause medical accidents.

[0022] Based on this, for conscious patients, they usually directly express the symptoms that require the most immediate attention during the consultation, so existing voice consultation systems can provide relatively accurate answers. However, unconscious patients often cannot accurately express their most urgent symptoms, and the consultation content output by the LLM (Local Mental Health) system, guided by the patient's expression, does not tend to focus on the patient's most urgent symptoms, which can easily lead to delays in the patient's medical treatment.

[0023] The above analysis clearly shows that existing voice-based consultation systems are unable to accurately diagnose patients with impaired consciousness, thus delaying their treatment.

[0024] To accurately conduct medical interviews with patients of impaired consciousness, embodiments of this application provide a medical interview method, a medical interview device, a computer-readable storage medium, and a computer program product, including: acquiring target medical statements corresponding to the patient's chief complaint and the patient's state of consciousness data; determining at least one target red flag symptom from a set of preset red flag symptoms based on the target medical statement and the state of consciousness data; determining multiple interview targets based on the target medical statement and the obtained target red flag symptom, and determining interview strategies and interview sequences for each interview target based on the state of consciousness data; conducting a medical interview with the patient based on the interview strategies and interview sequences for each interview target; and outputting the patient's medical history based on the patient's responses to the multiple interview targets. By determining target red flag symptoms based on the target medical statements corresponding to the patient's chief complaint and the state of consciousness data, it is possible to determine the red flag symptoms that the patient may actually want to express based on the patient's current state of consciousness and the expressed chief complaint. Then, by determining interview strategies and interview sequences that match the patient's current state of consciousness based on the state of consciousness data, accurate medical interviews can be conducted with the patient. Based on this, when conducting a consultation with an unconscious patient using this method, it is possible to effectively identify the symptoms that the unconscious patient actually wants to address first in their current state of consciousness, and determine a consultation strategy and sequence suitable for the patient's current state of consciousness, thereby accurately conducting a consultation with an unconscious patient.

[0025] See Figure 1 , Figure 1 The flowchart of a diagnostic method provided in one embodiment of this application is illustrated. The diagnostic method provided in this application embodiment can be applied to edge computing devices, terminals, etc., and can be specifically embodied as a diagnostic system in the mentioned hardware device. Specifically, the flowchart of the diagnostic method may include steps 110 to 150.

[0026] Step 110: Obtain the target medical description corresponding to the patient's chief complaint and the patient's state of consciousness data; Step 120: Based on the target medical description and consciousness status data, identify at least one target red flag symptom from a set of pre-defined red flag symptoms; Step 130: Based on the target medical description and the obtained target red flag symptoms, identify multiple consultation targets, and determine the consultation strategy and consultation order for each consultation target based on the consciousness state data; Step 140: Conduct a consultation with the patient according to the consultation strategy and consultation order for each consultation goal; Step 150: Output the patient's medical history based on the patient's answers to multiple consultation objectives.

[0027] In one embodiment, the chief complaint refers to the main symptoms described by the patient at the beginning of the consultation. Specifically, in the medical field, the chief complaint is the core of subsequent consultations, and all subsequent consultations revolve around the chief complaint to determine the specific disease causing the chief symptom. For example, at the beginning of the consultation, the patient is usually asked, "Where do you feel unwell?" The patient's answer, such as "My stomach is upset" or "My chest hurts," is the chief complaint.

[0028] In one embodiment, the target medical expression refers to the professional symptom terminology in the medical field that matches the chief complaint. For example, if the patient's chief complaint is "I don't breathe well," the target medical expression could be "difficulty breathing." It should be noted that chief complaints can take many forms. If the chief complaint is a colloquial expression by the patient, then the target medical expression is the medical expression that is most similar to the chief complaint; if the chief complaint itself is a medical expression, then the chief complaint is the target medical expression.

[0029] In one embodiment, consciousness state data refers to data used to represent the patient's current state of consciousness. This consciousness state data may include at least one of alertness data, cognitive load data, physiological arousal data, and encephalopathy risk data. Arousal data characterizes the patient's focus on the consultation content; a closer arousal value of 1 indicates greater alertness and better ability to follow the conversation, while a closer arousal value of 0 indicates drowsiness or difficulty maintaining attention. Cognitive load data characterizes the patient's understanding of the consultation content; a closer cognitive load value of 1 indicates a heavier burden of understanding and organizing language, while a closer cognitive load value of 0 indicates a lighter burden. Physiological arousal data characterizes the patient's anxiety level; a closer physiological arousal value of 1 indicates higher anxiety or physiological arousal, while a closer physiological arousal value of 0 indicates more stable emotions. Encephalopathy risk data indicates the degree of encephalopathy-like slowing of brain activity; a closer encephalopathy risk value of 1 indicates a higher probability or risk of the EEG showing abnormal patterns such as suspected "diffuse slowing of brain activity," while a closer encephalopathy risk value of 0 indicates a lower risk.

[0030] In one embodiment, the chief complaint can be converted into a target medical expression by an edge computing device outside the consultation system. During the process of obtaining the target medical expression corresponding to the patient's chief complaint, the consultation system can receive information sent by the edge computing device and parse the target medical expression from the sent information.

[0031] In one embodiment, the chief complaint can be automatically converted into a target medical expression by the consultation system. Specifically, in obtaining the target medical expression corresponding to the patient's chief complaint, the patient's audio responses during the complaint process can be collected first and converted into the chief complaint content. Then, the chief complaint content is uploaded to a medical LLM (Medical Library Model), which determines the target medical expression matching the chief complaint content. The medical LLM can be Med-PaLM, BioGPT, etc., and is not specifically limited here.

[0032] In one embodiment, in the process of obtaining the target medical expression corresponding to the patient's chief complaint, the process can first perform voice recall based on the target medical expression to obtain multiple candidate medical expressions, and then refine the target medical expression from the multiple candidate medical expressions through cross-coding.

[0033] In one embodiment, the chief complaint can be automatically converted into the target medical description by the consultation system, see [link to relevant documentation]. Figure 2 The process of obtaining the target medical description corresponding to the patient's chief complaint may include the following steps.

[0034] Step 210: Collect audio recordings of the patient's responses during the chief complaint process and convert the audio recordings into the chief complaint content; Step 220: Map the chief complaint to multiple corresponding candidate medical statements; Step 230: Perform constraint screening on multiple candidate medical expressions to obtain pending medical expressions; Step 240: Determine the first judgment score of the pending medical statement based on the confidence level of the statement and its similarity to the chief complaint. Step 250: In response to the first judgment score being not less than a preset scoring threshold, the pending medical expression is determined as the target medical expression.

[0035] In one embodiment, converting the response audio into the main complaint means performing speech recognition on the response audio to obtain the main complaint in text form. The speech recognition can be implemented using neural network models such as Gaussian Mixture Model (GMM) or DeepSpeech; no specific model is limited here.

[0036] In one embodiment, during the process of collecting the patient's response audio during the chief complaint and converting the response audio into the chief complaint content, a language selection interface can be first displayed to the patient, and the language selection information entered by the patient on the language selection interface can be obtained. Then, the response audio is subjected to speech recognition based on the selected language according to the language selection information to obtain the chief complaint content.

[0037] In one embodiment, candidate medical expressions can be output in text form, or in the form of a concept ID of the candidate medical expression in a specific medical coding standard, etc., without being specifically limited here. For example, different symptom coding standards such as SNOMED CT, MedDRA PT / LLT, ICD-10, and ICD-11 can encode the same medical expression into different concept IDs, and the candidate medical expression can be represented by one of the corresponding concept IDs.

[0038] In one embodiment, during the process of mapping the chief complaint to multiple candidate medical expressions, a search can be performed in a preset medical thesaurus based on the chief complaint to determine multiple candidate medical expressions that match the chief complaint. The medical thesaurus can be MedDRA, MedMacau, etc., and is not specifically limited here.

[0039] In one embodiment, during the process of mapping the chief complaint to multiple corresponding candidate medical expressions, the chief complaint can be uploaded to a medical LLM (Medical LLM Model), which then identifies multiple candidate medical expressions that match the chief complaint. The medical LLM can be Med-PaLM, BioGPT, etc., and is not specifically limited here.

[0040] In one embodiment, constraining and filtering multiple candidate medical expressions to obtain pending medical expressions refers to filtering candidate medical expressions according to preset judgment rules and using the filtered candidate medical expressions as pending medical expressions. The preset judgment rules may include specific specialty rules, description matching rules, etc., which are not specifically limited here. For example, the preset judgment rules may include specific rules such as "pain ≠ discomfort," "burning ≠ bloating," "location × type must be consistent (i.e., the candidate medical expression should be a symptom that can appear at the location of the patient's complaint)," and "medical expression abbreviations must be accurate," to eliminate confusing items among multiple candidate medical expressions.

[0041] In one embodiment, the first judgment score for determining a candidate medical statement can be represented by formula (1) based on the confidence level of the medical statement to be determined and the similarity between it and the content of the chief complaint.

[0042] ; In formula (1), Score can be used to represent the first judgment score; α can be used to represent the weighted value, specifically satisfying: α∈[0.5,0.8]; confPT can be used to represent confidence level; simPT can be used to represent the similarity between the content of the complaint and the main complaint.

[0043] In one embodiment, in response to a first determination score being less than a preset score threshold, a query is generated to confirm with the patient whether the pending medical statement is equivalent to the expressed content. If the query result is yes, the pending medical statement is determined as the target medical statement. If the query result is no, the patient's chief complaint is retrieved again, and step 110 is executed again.

[0044] The above embodiments are for the case where there is only one pending medical expression. In the actual process of determining the target medical expression, the preset judgment rule cannot guarantee that only one pending medical expression can be selected from multiple candidate medical expressions. Sometimes there may be multiple pending medical expressions. In this case, the first judgment score of each pending medical expression can be calculated based on formula (1). Then, it is determined whether the first judgment score of the pending medical expression with the highest first judgment score is not less than the preset score threshold. If yes, the pending medical expression with the highest first judgment score is determined as the target medical expression. If no, information is generated to ask the patient about their chief complaint to determine whether the pending medical expression is equivalent to the content expressed. If the answer is yes, the pending medical expression is determined as the target medical expression. If the answer is no, the patient's chief complaint content is retrieved again, and step 110 is executed again.

[0045] In one embodiment, the consciousness state data can be obtained by an edge-assisted computing device other than the edge computing device executing the consultation method provided in this application embodiment. The edge-assisted computing device can obtain the consciousness state data by recognizing the patient's facial image captured by an external camera, or it can be connected to a certain electroencephalogram (EEG) acquisition device to acquire EEG data, and then convert the EEG data to obtain the consciousness state data, etc., etc., without being specifically limited here. During the process of obtaining the patient's consciousness state data, the edge computing device can receive the information sent by the edge-assisted computing device and parse the consciousness state data from the sent information.

[0046] In one embodiment, during the process of acquiring the patient's state of consciousness data, the edge computing device is connected to an electroencephalogram (EEG) acquisition device. The edge computing device can first acquire the patient's EEG data and then convert the EEG data into state of consciousness data.

[0047] For example, an edge computing device connects to a head-mounted EEG device to collect the patient's brainwave frequencies (equivalent to electroencephalogram data) and determine specific consciousness state data based on the brainwave frequencies. For instance, it determines alertness data based on the ratio between brainwave frequencies θ and δ; it determines cognitive load data based on the ratio between brainwave frequencies θ and α; it determines physiological arousal data based on the relative brainwave frequency β; and it determines encephalopathy risk data based on the peak value of brainwave frequency α.

[0048] In one embodiment, during the acquisition of the patient's state of consciousness data, the edge computing device is connected to an electroencephalogram (EEG) acquisition device. The edge computing device can first acquire the patient's physiological signals and then convert the physiological signals into state of consciousness data.

[0049] For example, an edge computing device connected to a wristband, front-facing camera, skin patch, pulse oximeter, and other devices forms a multimodal acquisition device. This device collects multimodal physiological signals from the patient and determines specific consciousness status data based on these signals. For instance, it uses images captured by the front-facing camera to determine the patient's pupil dilation level to determine alertness; it uses images captured by the front-facing camera to determine the patient's blink frequency to determine cognitive load; it uses Galvanic Skin Response (GSR) fluctuations collected by the wristband to determine physiological arousal data; and it uses blood oxygen concentration and heart rate collected by the pulse oximeter to determine physiological arousal data.

[0050] The following explains the relevant content of this application concerning the red flag symptoms.

[0051] In one embodiment, red flag signs refer to symptoms that indicate the presence of a potentially serious illness requiring further examination and treatment. For example, symptoms requiring rapid emergency treatment, such as sudden myocardial infarction, internal bleeding, or persistent high fever, are all considered red flag signs. The preset multiple red flag signs may include multiple red flag signs from a single specialty or multiple specialties, depending on the application scenario of the consultation method provided in this application. For instance, if the consultation method provided in this application is only applied to the neurology department, then the preset multiple red flag signs will only include multiple red flag signs from the neurology specialty; or, if the consultation method provided in this application is applied to surgery, then the preset multiple red flag signs may include multiple red flag signs from various surgical specialties such as basic surgery and cardiothoracic surgery, etc., and so on. No specific limitation is made here.

[0052] In one embodiment, a target red flag symptom refers to a red flag symptom that may accompany the symptoms described in the target medical description and requires urgent treatment. Specifically, the target medical description is only the symptom that the patient can most intuitively express in their current cognitive state. In reality, the symptoms that accompany a patient are diverse, and the target red flag symptom is one or more of these symptoms that accompany the target medical description. For example, if "dyspnea" is the target medical description, and the preset red flag symptoms include "urinary and fecal incontinence," "hematochezia," "pink frothy sputum," and "cyanosis," in respiratory diseases, acute left heart failure can cause "dyspnea" and "pink frothy sputum," and hypoxia can cause "cyanosis" and "dyspnea." Therefore, the red flag symptom accompanying "dyspnea" may be "pink frothy sputum" or "cyanosis," or both. Therefore, the edge computing device will identify "pink frothy sputum" and "cyanosis" as target red flag symptoms.

[0053] In one embodiment, in the process of determining at least one target red flag symptom from a set of preset red flag symptoms based on the target medical description and state of consciousness data, the target medical description and state of consciousness data can be input into a pre-trained neural network model. The neural network model calculates the probability of each red flag symptom based on the target medical description and state of consciousness data, and determines the K red flag symptoms with the highest probabilities as the target red flag symptoms. Here, K is a positive integer greater than or equal to 1. The pre-trained neural network model can include XGBoost, random forest, etc., and is not specifically limited here.

[0054] In one embodiment, in the process of determining at least one target red flag symptom from a plurality of preset red flag symptoms based on the target medical description and consciousness state data, specifically for each red flag symptom, the similarity between the target medical description and the red flag symptom can be calculated by weighting the consciousness state data; after obtaining the similarity between each red flag symptom and the target medical description, the K red flag symptoms with the highest similarity are determined as target red flag symptoms, where K is an integer greater than or equal to 1.

[0055] See Figure 3 In one embodiment, the process of determining at least one target red flag symptom from a plurality of preset red flag symptoms based on target medical descriptions and state of consciousness data may include the following steps.

[0056] Step 310: For each red flag symptom, determine the preset specialist score for the red flag symptom, the sensitivity score of consciousness state data to the red flag symptom, and the correlation score between the target medical description and the red flag symptom, and determine the second judgment score for the red flag symptom based on the preset specialist score, sensitivity score, and correlation score. Step 320: Identify the K highest second-ranking red flag symptoms as target red flag symptoms, where K is a positive integer greater than or equal to 1.

[0057] In one embodiment, the preset specialist score refers to a preset baseline score for a red flag symptom. The preset specialist score can characterize the treatment priority among various red flag symptoms. For example, in the gastroenterology department, red flag symptoms include "hematemesis" and "abdominal distension". "Hematemesis" is a high-risk signal, so the preset specialist score for "hematemesis" will be higher than the preset specialist score for "abdominal distension".

[0058] In one embodiment, the sensitivity score refers to a score used to indicate how easily the red flag symptom can induce a state of consciousness in the patient as represented by the state of consciousness data. For example, hypoxia can lead to symptoms such as "dyspnea," "cyanosis," and "dizziness." Since state of consciousness data can directly reflect whether the patient's brain activity is active, if the state of consciousness data indicates that the patient's brain activity is not active, the more relevant the red flag symptom "dizziness" may be to the current state of consciousness, the higher the sensitivity score for the red flag symptom "dizziness" will be.

[0059] It is important to note that because the sensitivity score is closely related to the patient's state of consciousness, some red flag symptoms related to the state of consciousness, such as "dizziness," will have a higher sensitivity score. For some red flag symptoms that are not related to the state of consciousness, the sensitivity score will be lower. In this way, the sensitivity score formed by the state of consciousness data can form an output gating, which helps to rule out some diseases such as acute pain and encephalopathy risks that patients cannot express in their current state when they are not in a state of consciousness.

[0060] It is also important to note that there are various ways to determine the sensitivity score. For example, the sensitivity score can be obtained by weighted summing the differences between the alertness data and the preset alertness threshold, the differences between the cognitive load data and the preset cognitive load threshold, the differences between the physiological arousal data and the physiological arousal threshold, and the differences between the encephalopathy risk data and the encephalopathy risk threshold. Alternatively, the sensitivity score can be obtained by weighted summing one or more of the alertness data, cognitive load data, physiological arousal data, and encephalopathy risk data that exceed their respective thresholds, and so on. The specific method is not limited here.

[0061] In one embodiment, the relevance score refers to a score used to represent the degree of association between a target medical statement and a red flag symptom. For example, "melena" easily leads to "cyanosis," therefore, the association between "melena" and the red flag symptom "cyanosis" is high, and the corresponding relevance score is high. The relevance score can be represented by the similarity between the target medical statement and the red flag symptom, or by the adjacent distance between the target medical statement and the red flag symptom in a medical knowledge graph, etc., and is not specifically limited here.

[0062] In one embodiment, the process of determining the second judgment score of the red flag symptom based on the preset specialist score, sensitivity score and relevance score can be represented by the following formula (2).

[0063] ; Where f can be used to represent red flag symptoms; Score(f) can be used to represent the second-order score for red flag symptoms; b_f can be used to represent the preset specialist score for red flag symptoms; EEG_f(g) can be used to represent sensitivity scores; w1 can be used to represent the weighted value of the sensitivity score; text_f can be used to represent relevance scores; w2 can be used to represent the weighted value of the relevance score.

[0064] It is important to note that the more physiological information available about the patient, the more accurate the identification of the target red flag symptom. For example, incorporating easily obtainable patient signs can enhance the identification process.

[0065] In one embodiment, in the process of determining the second judgment score for red flag symptoms based on preset specialist scores, sensitivity scores, and relevance scores, specifically in response to the collection of the patient's physical signs data, the physical signs score for red flag symptoms is determined based on the physical signs data, and then the second judgment score is determined based on the preset specialist scores, sensitivity scores, relevance scores, and physical signs scores.

[0066] In one embodiment, vital signs data refers to data used to characterize a patient's vital signs index. Vital signs data may include at least one of specific data such as the patient's blood pressure, pulse rate, and oxygen saturation (SpO2), etc., without specific limitations herein.

[0067] In one embodiment, a vital sign score refers to the degree of correlation between a patient's current vital sign data and the red flag symptoms currently being assessed for a second judgment. For example, if blood pressure, pulse rate, and blood oxygen saturation are all within normal medical standards, then red flag symptoms indicating respiratory conditions are generally poorly correlated with these vital sign data, and the vital sign scores for these red flag symptoms and the vital sign data are low.

[0068] In one embodiment, the determination of vital signs score can be done in various ways. For example, the difference between blood pressure data and a preset blood pressure threshold, the difference between pulse rate data and a preset pulse rate threshold, and the difference between blood oxygen saturation data and blood oxygen saturation data can be weighted and summed to obtain the vital signs score. Alternatively, the vital signs score can be obtained by weighting and summing one or more of the blood pressure data, pulse rate data, and blood oxygen saturation data that exceed their respective thresholds, etc. The specific method is not limited here.

[0069] In one embodiment, the process of determining the second judgment score based on the preset specialty score, sensitivity score, relevance score and physical sign score can be represented by the following formula (3).

[0070] ; Where f can be used to represent red flag symptoms; Score(f) can be used to represent the second-order score for red flag symptoms; b_f can be used to represent the preset specialist score for red flag symptoms; EEG_f(g) can be used to represent sensitivity scores; w1 can be used to represent the weighted value of the sensitivity score; text_f can be used to represent relevance scores; w2 can be used to represent the weighted value of relevance scores; vitals_f can be used to represent vital signs scores; w3 can be used to represent the weighted value of the physical characteristic score.

[0071] The following section explains the relevant content regarding the consultation objectives involved in this application.

[0072] In one embodiment, the consultation objective refers to indicative information that indicates the target information that needs to be determined during the patient's current diagnosis. The consultation objective may include, but is not limited to, the onset of the disease, the location of the disease, the nature and progression of symptoms, and accompanying symptoms, etc., without further specific limitations.

[0073] In one embodiment, the multiple consultation targets include at least one red flag consultation target and at least one refined consultation target. In the process of determining multiple consultation targets based on the target medical description and the obtained target red flag symptoms, at least one red flag consultation target can be determined first based on the obtained target red flag symptoms. Then, the target symptom location is determined based on the target medical description, and at least one refined consultation target is generated based on the preset specialist cause-effect diagram corresponding to the target symptom location.

[0074] Among them, the "red flag" diagnostic objective refers to the information indicating that it is necessary to determine whether a patient has the target red flag symptom. Additionally, the "refined diagnostic objective" refers to the information indicating that it is necessary to determine more detailed information about the chief complaint.

[0075] For example, if the chief complaint is "feeling short of breath" and the target red flag symptom is "cyanosis", then the red flag consultation goal is to determine whether the patient has "cyanosis". The detailed consultation goals can be "the onset time of the onset of dyspnea", "the duration of the onset of dyspnea", "what happened before the onset of dyspnea", etc.

[0076] In one embodiment, the pre-defined specialist cause-effect diagram of the target symptom site refers to a relationship diagram formed by the diseases that may occur at the target symptom site, the nature of the diseases, the duration of the diseases, the causes, and the symptoms caused.

[0077] In one embodiment, the consultation strategy refers to the specific strategy used to obtain the consultation target. Since consultation generally involves question and answer, the consultation strategy can be embodied in different forms of consultation questions in this embodiment. The consultation strategy may include multiple-choice questions with specific content, open-ended questions, multiple-selection questions, image area selection questions, etc., without further limitation. Furthermore, depending on the complexity, the consultation strategy for a consultation target may be embodied in only one consultation question, or it may be embodied in multiple consultation questions of different forms, etc., without further limitation.

[0078] In one embodiment, the consultation sequence refers to the order in which the questions are asked according to each consultation strategy. For example, suppose the consultation goal is to determine whether a patient has two target red flag symptoms and one accompanying symptom. The consultation strategy for the two target red flag symptoms can be represented as a multiple-choice question (A, B, C, D) with two options, and the consultation strategy for the accompanying symptom can be represented as a multiple-choice question (E) with an image region. In this case, the consultation system can choose to ask the patient question B first, then question A, then question C, then question E, and finally question D. The order in which the patient is asked questions is "B→A→C→E→D", which is the consultation sequence.

[0079] In one embodiment, there are various ways to determine the consultation strategy and consultation order for each consultation target based on consciousness state data, and no specific method is limited here.

[0080] For example, the consciousness state data and each consultation target are input into the decision tree model to obtain the consultation strategy for each consultation target. Then, the understanding difficulty score of the consultation strategy of each consultation target is scored by the consciousness state data, and the consultation order is determined according to the understanding difficulty score of the consultation strategy of each consultation target.

[0081] For another example, by inputting consciousness state data and each consultation target into the LLM, the consultation strategy and consultation order for each consultation target can be obtained directly.

[0082] In one embodiment, target red flag symptoms may include general target red flag symptoms and specialized target red flag symptoms. In determining the consultation strategy and sequence for each consultation target based on consciousness state data, specifically, the consultation strategy for the red flag symptom target corresponding to the general target red flag symptoms, the consultation strategy for the red flag symptom target corresponding to each specialized target red flag symptom, and the consultation strategy for the detailed consultation target corresponding to each specialized target red flag symptom can be determined separately based on consciousness state data. Then, for each specialized target red flag symptom, the consultation sequence between the red flag symptom target and the detailed consultation target corresponding to that specialized target red flag symptom is determined based on consciousness state data. By branching the red flag symptoms, different specialized target red flag symptoms can be accessed based on feedback from the general target red flag symptoms during the consultation process, achieving precise consultation. Furthermore, due to the use of branched consultation, the consultation of other useless items can be reduced, improving consultation efficiency.

[0083] By using consciousness state data to determine the consultation strategies and sequence for each consultation objective, we can formulate consultation strategies and sequences for each consultation question that are appropriate for the patient's current state of consciousness based on the patient's consciousness state reflected in the consciousness state data. This provides a consultation process that is appropriate for the patient's current state of consciousness and helps to complete accurate consultations for patients in different states of consciousness.

[0084] In the actual consultation process, the patient's state of consciousness is dynamic. During the consultation, there may be situations where certain consultation strategies and sequences do not match the patient's latest state of consciousness. Therefore, the consultation strategies and sequences can be updated in real time according to the patient's state of consciousness.

[0085] See Figure 4 In one embodiment, the process of conducting a patient consultation based on the consultation strategy and consultation sequence for each consultation objective may include the following steps.

[0086] Step 410: Obtain the patient's current EEG data and update the consciousness status data based on the current EEG data; Step 420: In response to the difference between the patient's state of consciousness represented by the updated state of consciousness data and the patient's state of consciousness represented by the pre-update state of consciousness data, update the consultation strategy and consultation order corresponding to each unasked consultation target according to the updated state of consciousness data; Step 430: Conduct a consultation with the patient based on the unasked consultation goals, the current consultation strategy, and the current consultation sequence.

[0087] It should be noted that updating consciousness state data based on current EEG data is based on the same principle as determining consciousness state data using EEG data as described above, and will not be elaborated on here.

[0088] In one embodiment, the difference between the patient's state of consciousness represented by the updated state of consciousness data and the patient's state of consciousness represented by the previous state of consciousness data can be determined in a variety of ways, which are not specifically limited here.

[0089] For example, if the level of consciousness data is different before and after the update, it can be assumed that the level of consciousness has changed; if the level of cognitive load data is different before and after the update, it can be assumed that the level of cognitive load has changed; if the level of physiological arousal data is different before and after the update, it can be assumed that the level of physiological arousal has changed; if the level of encephalopathy risk data is different before and after the update, it can be assumed that the level of encephalopathy risk has changed. In this case, if the updated state of consciousness data meets at least one of the above four conditions, it can be considered that the patient's state of consciousness represented by the updated state of consciousness data is different from the patient's state of consciousness represented by the unupdated state of consciousness data.

[0090] For example, if the difference between the updated consciousness data and the updated consciousness data is not less than a preset threshold for alerting changes in consciousness, then the level of consciousness can be considered to have changed; if the difference between the updated cognitive load data and the updated cognitive load data is not less than a preset threshold for alerting changes in cognitive load, then the level of cognitive load can be considered to have changed; if the difference between the updated physiological arousal data and the updated physiological arousal data is not less than a preset threshold for alerting changes in physiological arousal, then the level of physiological arousal can be considered to have changed; if the difference between the updated encephalopathy risk data and the updated encephalopathy risk data is not less than a preset threshold for alerting changes in encephalopathy risk, then the level of encephalopathy risk can be considered to have changed if the updated consciousness data meets at least one of the above four conditions.

[0091] In one embodiment, updating the consultation strategy and consultation order corresponding to each unquestioned consultation target based on the updated state of consciousness data means first identifying the consultation targets that have not yet been questioned, and then re-determining the consultation strategy and consultation order for each unquestioned consultation target based on the updated state of consciousness data.

[0092] In one embodiment, during the process of interviewing the patient based on each unasked interview target, the current corresponding interview strategy, and the current corresponding interview sequence, the patient's electroencephalogram (EEG) data can be collected in real time, and the above-mentioned steps of updating the interview strategy and interview sequence can be executed simultaneously until all interview targets are interviewed.

[0093] By collecting consciousness status data in real time and adjusting the consultation strategy and the order of consultation questions according to the patient's current consciousness status based on the real-time reflection of the consciousness status data, a consultation process that is consistent with the patient's current consciousness status can be dynamically provided, which helps to complete accurate consultations for patients in different consciousness states.

[0094] Based on the above embodiments, all consultation targets can be achieved. However, in reality, red flag symptoms require immediate treatment. That is, when a patient reports that they have a target red flag symptom, it is necessary to immediately inquire about the relevant details of the target red flag symptom for emergency treatment. At this time, the consultation target needs to be changed.

[0095] In one embodiment, conducting a patient consultation based on the consultation strategy and sequence for each consultation objective refers to asking the patient questions that embody the consultation strategy for each consultation objective, following the consultation sequence. It should be noted that if the patient reports no red flag symptoms, it indicates that no emergency treatment is required, and therefore the consultation can be completed in the manner described above.

[0096] See Figure 5 In one embodiment, the multiple consultation targets include multiple red flag consultation targets, and the process of conducting consultations on patients according to the consultation strategies and consultation order of each consultation target may include the following steps.

[0097] Step 510: In response to the completion of one or more red flag consultation goals and the patient's feedback of one or more target red flag symptoms, update each unconsulted consultation goal based on the feedback of target red flag symptoms; Step 520: Obtain the patient's current EEG data and update the consciousness status data based on the current EEG data; Step 530: Based on the updated consciousness state data, determine the updated consultation goals, the corresponding consultation strategies, and the corresponding consultation order; Step 540: Conduct a consultation with the patient based on the updated consultation goals, the current consultation strategy, and the current consultation sequence.

[0098] In one embodiment, completing a consultation for one or more red flag diagnostic goals and receiving feedback from the patient regarding the presence of one or more target red flag symptoms may include one of the following situations: completing a consultation for one red flag diagnostic goal and receiving feedback from the patient regarding the presence of that target red flag symptom; completing consultations for multiple red flag diagnostic goals and receiving feedback from the patient regarding the presence of one or more of those target red flag symptoms.

[0099] In one embodiment, updating the consultation targets for each unconsulted patient based on the feedback target red flag symptoms refers to adjusting the currently unconsulted consultation targets to consultation targets that inquire about the details of the existing target red flag symptoms, based on the target red flag symptoms present in the patient.

[0100] It should be noted that, based on the updated consciousness state data, determining the updated consultation goals, the current consultation strategy, and the current consultation order is the same principle as in step 140, and will not be repeated here.

[0101] The following explains the relevant content of this application concerning the output of medical history.

[0102] In one embodiment, outputting the patient's medical history based on the patient's answers to multiple medical history targets refers to the operation of filling the answers to multiple medical history targets into the corresponding slots of a preset medical history template to generate medical history text.

[0103] In one embodiment, in the process of outputting the patient's medical history based on the patient's answers to multiple medical questions, the patient's diagnosis can be determined first based on the patient's answers to multiple medical questions. Then, a chain of medical evidence is generated based on the multiple medical questions and their corresponding order, the patient's answers to the multiple medical questions, and the consciousness status data. Finally, a preset medical history template is filled with the target medical description, the chain of medical evidence, and the diagnosis, and the medical history is output.

[0104] In one embodiment, in the process of generating a chain of medical evidence based on multiple medical inquiry goals and their corresponding order, the patient's answers to the multiple medical inquiry goals, and consciousness state data, the chain of medical evidence can be obtained by aligning the answers to the multiple medical inquiry goals and consciousness state data based on timestamps according to the multiple medical inquiry goals and their corresponding order.

[0105] In one embodiment, the medical history can be uploaded to the local medical system via a transmission method such as HTTPS / TLS encrypted transmission, or it can be cached locally so that the local system can decide when to upload it, etc., and the specific method is not limited here.

[0106] By standardizing the target medical descriptions, the chain of evidence in the medical history interview, and the diagnostic results using pre-set medical history templates, the process of rewriting patient medical histories can be completed conveniently, effectively improving the efficiency of medical history entry. Furthermore, the addition of a chain of evidence to the medical history allows doctors to easily review the interview process, effectively enhancing the credibility of the medical history content.

[0107] Several specific embodiments are given below. It should be noted that the embodiments given below are based on the following... Figure 6The diagnostic method execution framework shown consists of an edge computing device, a front-end interactive device, and a storage system for receiving and recording medical history. The interactive device may include a language selection and diagnostic device (such as a tablet or all-in-one machine), a voice acquisition device (such as a microphone), an EEG acquisition device (such as a head-mounted EEG machine), an input device (such as a large button), and a voice playback device (such as an external speaker or bone conduction headphones).

[0108] Example 1: See Figure 7 , Figure 7 This describes a data processing chain from patient input to electronic medical record output. At the start of the consultation, the system asks the patient to wear an EEG collection device. Afterwards, the system displays a screen allowing the patient or their family to select the language for the consultation, preventing misinterpretation or incorrect questions. Specifically, the screen provides multiple language buttons; after clicking, the system asks again in the same language, "Is this the correct selection?" for confirmation.

[0109] Next, the system asks the patient about their discomfort and captures their speech using a microphone. This speech is then converted into the patient's main complaint using Automatic Speech Recognition (ASR) and output as segmented text with timestamps. Afterward, a controlled medical thesaurus (such as MedDRA) is used to retrieve candidate medical expressions (Top-M, default 5-10) related to the main complaint. Only the most suitable target medical expression is selected from the candidates, and its unique concept ID and confidence level are output.

[0110] During the process of voice acquisition and confirmation of the target medical statement, the EEG acquisition device collects the patient's electroencephalogram data and converts it into consciousness state data.

[0111] Following this, the conceptual ID and confidence level of the target medical expression, along with the consciousness state data, are input into the Red Flag and Causality Subsystem. The Red Flag and Causality Subsystem has the ability to identify the red flag symptom consultation target and its subsequent detailed consultation targets. Therefore, the Red Flag and Causality Subsystem can determine multiple red flag symptom consultation targets and multiple detailed consultation targets based on the conceptual ID and confidence level of the target medical expression and the consciousness state data.

[0112] The Red Flag and Causality subsystem then outputs multiple Red Flag symptom consultation targets and multiple detailed consultation targets to the strategy selector, and the system also inputs consciousness state data into the strategy selector. The strategy selector is responsible for "translating" consultation targets into consultation strategies and determining the consultation order among various consultation questions.

[0113] After obtaining the consultation strategies and sequence for each consultation objective, various hardware devices are used to interact with the patient, thereby obtaining various information. Following this, the information obtained is assembled into a "medical history card" using a pre-set medical history template. This template includes standard slots for location, nature, duration, precipitating factors, and accompanying conditions, as well as slots for the chain of evidence. Information can be filled into the corresponding slots in the medical history template. After the medical history is generated, it is written into the Electronic Medical Record (EMR) system.

[0114] Example 2: When taking a medical history of an AD patient who presents with upper abdominal discomfort, the medical history system first eliminates risks and then refines the description of symptoms. Without a diagnosis, the medical history system can first eliminate risks and then write the symptoms of "upper abdomen + heartburn + ≥3 days + acid reflux" into a rewritable and traceable description using standard terminology and evidence chain.

[0115] Starting point: An elderly AD patient enters the outpatient clinic, stating that "my stomach is not feeling well."

[0116] First, four large buttons appear on the tablet's home screen (Simplified Chinese / Cantonese / English / Portuguese). The family member selects "Cantonese," and the system announces in Cantonese, "Is this the correct choice?" → [Yes] [No] (large buttons). Ultimately, the patient selects Cantonese as the language.

[0117] The patient says, "My stomach doesn't feel well," which is captured by the microphone. The ASR (Automatic Surgery) system generates the segmented text "My stomach doesn't feel well." During this process, a head-mounted EEG device simultaneously collects EEG data and outputs the following consciousness status data: Cognitive Load = 0.72, Physiological Arousal = 0.65, Awareness = 0.55, and Brain Disorder Risk = 0.20. This data will ultimately be used by the system to determine that the patient's cognitive load is high. The system will then use a "two-choice," "graphical," and "large button" questioning approach. The questioning speed should be slowed down, and at least 10 seconds should be allowed for the patient to understand the information.

[0118] Next, candidate medical expressions are retrieved from a controlled medical terminology database by searching the text “stomach discomfort”. Then, they are filtered according to preset specialty rules to determine the pending medical expressions. Finally, the target medical expression is determined by a mixed scoring method.

[0119] Then, based on the text and consciousness status data, the target red flag symptoms were identified as "vomiting blood," "melena," and "chest pain at rest." Subsequently, several more detailed consultation targets were determined based on the target medical description.

[0120] Then, based on the consciousness status data, we determined to first rule out red flag symptoms before refining the consultation. We also determined that the consultation for red flag symptoms would use a "large button + two options" approach, while the consultation for refined consultation targets would use a "diagram + two options" approach.

[0121] Therefore, the following interaction process is performed through the screen.

[0122] The screen displays "Have you vomited blood in the past 24 hours?" and two button controls: "Yes" and "No". The patient selects "No".

[0123] The screen displays "Is there black stool?" and two button controls: "Yes" and "No". The patient selects "No".

[0124] The screen displays the question "Do you experience chest pain at rest?" along with two button controls: "Yes" and "No". The patient selects "No".

[0125] Next, the screen will display as follows Figure 8 The diagram shows a human abdomen and displays the text "Please select the location where you feel uncomfortable." The patient clicks on the "upper abdomen" location.

[0126] The screen displays the inquiry-like text "More like heartburn or bloating," along with a corresponding illustration of the nature as a selection control. The patient selects "heartburn."

[0127] The screen displays the text "Has it lasted more than three days?" followed by two button controls: "Yes" and "No". The patient clicks "Yes".

[0128] The screen displays the question "Do you have acid reflux?" along with two button controls: "Yes" and "No". The patient selects "Yes".

[0129] Finally, the following medical history was generated based on the patient's answers.

[0130] Location: Upper abdomen (PT / LLT, evidence: click on upper abdomen, t=22.1s, gating: cognitive load data=0.72); Properties: heartburn (PT / LLT, evidence: graphical choice, t=25.0s); Timeframe: ≥3 days (Evidence: Two-way choice, t=27.3s); Accompanying: regurgitation / acid reflux (PT / LLT, t=29.8s); Red flag: Hematemesis (negative), melena (negative), chest pain at rest (negative); Consciousness state gating tracking: full-process gating sampling.

[0131] Example 3: Conduct an interview with an AD patient. The AD patient has dyspnea at rest. The interview system first rules out risks and then details the symptom description. Without a diagnosis, the interview system can rule out risks first, quickly verify when detecting red flag symptoms, and standardize and write back the results and evidence chain to the EMR to assist doctors in timely treatment.

[0132] S0: Four large buttons (Simplified Chinese / Cantonese / English / Portuguese) appear on the first screen of the tablet. The family member clicks "Cantonese", and the system broadcasts in Cantonese "Is it selected correctly?" → [Yes] [No] (large buttons), and finally the patient selects Cantonese as the language.

[0133] S1: The patient says "Qi hao wu shun", which is collected by the microphone, and the ASR generates segmented text "Qi hao wu shun". During this process, EEG is synchronously collected to obtain electroencephalogram data, and the following conscious state data is output: cognitive load data = 0.66, physiological arousal data = 0.71, wakefulness data = 0.55, and encephalopathy risk data = 0.68. Such data will finally be determined by the system that the patient has a relatively high cognitive load and encephalopathy risk. Specifically, the interview methods of "choose one of two", "diagram", and "large button" are required. The speaking speed of the interview questions needs to be slowed down, and at least 10S of understanding time needs to be reserved for the patient.

[0134] S2: Retrieve candidate medical expressions through the text "Qi hao wu shun" in the controlled medical term library, then screen according to the preset specialty rules to determine the pending medical expressions, and then determine the target medical expressions through mixed scoring.

[0135] S3: Determine the target red flag symptoms as "dyspnea at rest", "cyanosis", and "massive hemoptysis" among multiple red flag symptoms according to the text and conscious state data. Then, determine multiple subsequent refined interview targets according to the target medical expressions.

[0136] S4: Determine to rule out red flag symptoms first and then refine the interview according to the conscious state data, and determine that the interview for the red flag symptom target adopts the method of "large button + choose one of two", and the interview for the refined interview target adopts the method of "diagram + choose one of two".

[0137] Therefore, the following interaction process is carried out through the screen.

[0138] The screen shows "Is it also difficult to breathe at rest?" and two button controls "Yes" and "No". The patient clicks "Yes".

[0139] The screen shows "It is recommended to measure blood oxygen". The doctor measures the vital signs of the patient and obtains the sign data SpO2 = 90% (low). At this time, the screen will show "Please go to the doctor as soon as possible / transfer to the emergency department for evaluation".

[0140] S5: Generate the following medical history according to the patient's answer.

[0141] Nature: Dyspnea (PT / LLT, evidence: verbal statement); Red flag: resting shortness of breath (positive, t=12.0s), SpO2 90% (low)), hemoptysis (negative / not tested), cyanosis (negative / not tested); Recommendation: Emergency assessment suggestion.

[0142] In addition to the embodiments described above, one embodiment of this application also provides a medical consultation device. See also Figure 9 , Figure 9 This is a schematic diagram of the structure of a medical consultation device provided in one embodiment of this application. Figure 9 As shown, the consultation device includes a memory 1100 and a processor 1200. The number of memories 1100 and processors 1200 can be one or more. Figure 9 Taking a memory 1100 and a processor 1200 as an example; Figure 9 The memory 1100 and processor 1200 can be connected via a bus or other means. Figure 9 Taking the example of a connection between China and Israel via a bus.

[0143] The memory 1100, as a computer-readable storage medium, can be used to store one or more software programs, computer-executable programs, and modules, such as the programs, instructions, or modules corresponding to the information processing methods provided in any embodiment of this application. The processor 1200 implements the diagnostic method provided in any embodiment of this application by executing one or more computer programs, instructions, and modules stored in the memory 1100.

[0144] The memory 1100 may primarily include a program storage area and a data storage area, wherein the program storage area may store the operating system and computer programs required for at least one function. Furthermore, the memory 1100 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 1100 may further include memory remotely located relative to the processor 1200, and these remote memories can be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0145] In addition to the embodiments described above, one embodiment of this application also provides a computer-readable storage medium storing computer-executable instructions for performing the consultation method as described in any of the preceding embodiments.

[0146] Furthermore, one embodiment of this application also provides a computer program product, including a computer program or computer instructions stored in a computer-readable storage medium. The processor of the diagnostic device reads the computer program or computer instructions from the computer-readable storage medium and executes the computer program or computer instructions, causing the communication device to perform the diagnostic method as described in any of the preceding embodiments.

[0147] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0148] The above is a detailed description of the preferred embodiments of this application. However, this application is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for taking medical history, characterized in that, Includes the following steps: Obtain the target medical description corresponding to the patient's chief complaint and the patient's state of consciousness data; Based on the target medical description and the consciousness state data, at least one target red flag symptom is determined from a set of preset red flag symptoms; Based on the target medical description and the obtained target red flag symptoms, multiple consultation targets are identified, and the consultation strategy and consultation order for each consultation target are determined based on the consciousness state data. The patient is interviewed according to the interview strategy and interview sequence for each of the interview objectives; The patient's medical history is output based on the patient's answers to multiple diagnostic objectives.

2. The method according to claim 1, characterized in that, The acquisition of the target medical description corresponding to the patient's chief complaint includes: The audio of the patient's responses during the chief complaint process is collected, and the audio responses are converted into the content of the chief complaint; The chief complaint is mapped to multiple corresponding candidate medical statements; The candidate medical expressions are subjected to constraint screening to obtain the pending medical expressions; A first judgment score for the pending medical statement is determined based on the confidence level of the statement and its similarity to the content of the chief complaint. In response to the first determination score being not less than a preset score threshold, the pending medical expression is determined as the target medical expression.

3. The method according to claim 1, characterized in that, The step of determining at least one target red flag symptom from a set of preset red flag symptoms based on the target medical description and the consciousness state data includes: For each of the aforementioned red flag symptoms, a preset specialist score for the red flag symptoms, a sensitivity score of the consciousness state data to the red flag symptoms, and a correlation score between the target medical description and the red flag symptoms are determined. A second judgment score for the red flag symptoms is then determined based on the preset specialist score, the sensitivity score, and the correlation score. The K highest second judgment scores of the red flag symptoms are determined as the target red flag symptoms, where K is a positive integer greater than or equal to 1.

4. The method according to claim 3, characterized in that, The second determination score for the red flag symptom, based on the preset specialist score, the sensitivity score, and the relevance score, includes: In response to the collection of the patient's vital signs data, a vital signs score for the red flag symptom is determined based on the vital signs data; The second judgment score is determined based on the preset specialty score, the sensitivity score, the relevance score, and the physical sign score.

5. The method according to claim 1, characterized in that, The multiple consultation objectives include at least one red flag consultation objective and at least one refined consultation objective; The process involves determining multiple diagnostic targets based on the target medical description and the obtained target red flag symptoms, including: Based on the obtained target red flag symptoms, at least one of the red flag diagnostic targets is determined; The target symptom location is determined based on the target medical description, and at least one refined consultation target is generated based on the preset specialist cause-effect diagram corresponding to the target symptom location.

6. The method according to claim 1, characterized in that, The process of conducting a patient consultation based on the consultation strategy and consultation order for each consultation objective includes: Acquire the patient's current electroencephalogram (EEG) data, and update the consciousness state data based on the current EEG data; In response to the difference between the patient's state of consciousness represented by the updated state of consciousness data and the patient's state of consciousness represented by the unupdated state of consciousness data, the consultation strategy and consultation order corresponding to each unasked consultation target are updated according to the updated state of consciousness data. The patient is interviewed based on the unasked consultation objectives, the current consultation strategy, and the current consultation sequence.

7. The method according to claim 1, characterized in that, The multiple consultation goals mentioned include multiple red flag consultation goals; The process of conducting a patient consultation based on the consultation strategy and consultation order for each consultation objective includes: In response to the completion of one or more of the red flag diagnostic targets and the patient reporting the presence of one or more of the target red flag symptoms, the diagnostic targets that have not been consulted are updated based on the reported target red flag symptoms; Acquire the patient's current electroencephalogram (EEG) data, and update the consciousness state data based on the current EEG data; Based on the updated consciousness state data, the updated consultation objectives, the corresponding consultation strategies, and the corresponding consultation order are determined. The patient is interviewed based on the updated consultation goals, the current consultation strategy, and the current consultation sequence.

8. The method according to claim 1, characterized in that, The step of outputting the patient's medical history based on the patient's answers to multiple medical inquiry targets includes: The diagnosis of the patient is determined based on the patient's answers to multiple of the stated diagnostic objectives; A chain of diagnostic evidence is generated based on multiple diagnostic objectives and their corresponding order, the patient's responses to the multiple diagnostic objectives, and the consciousness state data. The preset medical history template is filled in based on the target medical description, the chain of evidence for medical history taking, and the diagnostic results, and the medical history taking is output.

9. A diagnostic device, characterized in that, include: At least one processor; At least one memory for storing at least one program; The diagnostic method as described in any one of claims 1 to 8 is implemented when at least one of the programs is executed by at least one of the processors.

10. A computer-readable storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are used to perform the consultation method as described in any one of claims 1 to 8.

Images