Human-machine collaborative diagnosis and treatment method

Abstract

The application provides a man-machine collaborative diagnosis and treatment method, comprising the following steps: obtaining a doctor clone intelligent agent of a target doctor; collecting multi-source diagnosis and treatment events in a diagnosis and treatment session, and constructing a session state based on the multi-source diagnosis and treatment events; calling the doctor clone intelligent agent according to the session state and a real-time operation state of the target doctor, and generating auxiliary information conforming to the personalized characteristics of the target doctor; when it is detected that the target doctor edits the diagnosis and treatment content, suspending the output of the auxiliary information to the editing area of the target doctor, and temporarily storing the newly generated auxiliary information; when it is detected that the target doctor finishes editing the diagnosis and treatment content, merging the temporarily stored auxiliary information and the edited content of the target doctor, updating the session state based on the merging result, and realizing true man-machine collaboration by closely combining artificial intelligence assistance with the doctor workflow, thereby effectively improving the experience of man-machine collaboration.

Description

Technical Field

[0001] This invention relates to the field of smart medical technology, and in particular to a human-machine collaborative diagnosis and treatment method. Background Technology

[0002] With the continuous development of artificial intelligence (AI) technology in medical settings, AI-assisted diagnosis, medical record organization, diagnostic analysis, and treatment suggestion generation have become important development directions in the field of smart healthcare. Especially in outpatient settings, online audio-visual consultations, and follow-up visits, doctors need to quickly receive patient information, form diagnoses, record treatment content, and make subsequent decisions within a limited timeframe. Therefore, higher demands are placed on the real-time performance, continuity, and collaboration of diagnostic and treatment support systems.

[0003] Existing medical AI applications typically exist as single-point auxiliary tools, such as standalone modules for speech-to-text transcription, medical record generation, assisted diagnosis, or recommendations. While these modules can alleviate the information processing burden on doctors to some extent, most remain at the "task-based invocation" level, meaning they receive input and output results at a specific moment, leaving the decision of whether to adopt them to the doctor.

[0004] However, in real-world medical scenarios, relying solely on one-time call-based assistance methods is insufficient to adapt to the dynamic and continuous nature of the treatment process. When doctors manually edit the content, conflicts or overlaps in results often occur because the doctors' operations cannot be perceived, leading to a disconnect between AI assistance and the doctor's workflow, thus failing to achieve true human-machine collaboration. Summary of the Invention

[0005] This invention provides a human-machine collaborative diagnosis and treatment method to address the shortcomings of existing technologies where artificial intelligence assistance cannot achieve true human-machine collaboration.

[0006] This invention provides a human-machine collaborative diagnosis and treatment method, comprising: Obtain the target doctor's avatar AI agent; During the diagnosis and treatment session, multi-source diagnosis and treatment events are collected, and the session state is constructed based on the multi-source diagnosis and treatment events. Based on the session state and the real-time operation state of the target doctor, the doctor's avatar AI is invoked to generate auxiliary information that conforms to the personalized characteristics of the target doctor; When the target doctor is detected editing the diagnosis content, the output of the auxiliary information to the target doctor's editing area is paused, and the newly generated auxiliary information is temporarily stored; or, When the target doctor finishes editing the treatment content, the temporarily stored auxiliary information is merged with the content edited by the target doctor, and the session state is updated based on the merging result.

[0007] According to a human-machine collaborative diagnosis and treatment method provided by the present invention, the step of obtaining the doctor's avatar intelligent agent of the target doctor includes: The doctor avatar intelligent agent is constructed, which includes multiple skill modules, each of which is used to characterize the personalized features of the target doctor in the corresponding diagnosis and treatment task; The multiple skill modules include at least one of the following: consultation interaction module, medical record generation module, diagnostic analysis module, examination and test recommendation module, treatment and medication suggestion module, risk warning module, and follow-up and medical order module.

[0008] According to a human-machine collaborative diagnosis and treatment method provided by the present invention, the consultation interaction module is used to characterize the questioning order, follow-up questioning method, information focus and communication style of the target doctor in the consultation process; The medical record generation module is used to characterize the target physician's organization and writing habits of chief complaint, present illness, past medical history, physical examination, auxiliary examinations, diagnostic opinions and treatment suggestions; The diagnostic analysis module is used to characterize the target physician's judgment methods and diagnostic expression preferences regarding symptoms, medical history, and examination results. The examination and testing recommendation module is used to characterize the target doctor's examination suggestion habits, recommendation logic, and output boundaries at different stages of diagnosis and treatment. The treatment and medication recommendation module is used to characterize the target physician's expression and preferences regarding treatment recommendations, intervention recommendations, and medication recommendations. The risk alert module is used to characterize the target doctor's alert style and triggering rules for contraindications, risk factors, and key reminders. The follow-up and medical advice module is used to characterize the target doctor's output habits in terms of follow-up reminders, lifestyle suggestions, observation suggestions, and follow-up arrangements.

[0009] According to a human-machine collaborative diagnosis and treatment method provided by the present invention, the multi-source diagnosis and treatment events include at least one of the following: consultation voice events, speech-to-text events, patient input text events, patient upload attachment events, examination result events, doctor operation events, and auxiliary result events generated internally by the system.

[0010] According to a human-machine collaborative diagnosis and treatment method provided by the present invention, the step of constructing a session state based on the multi-source diagnosis and treatment events includes: The multi-source diagnosis and treatment events are structured and converted into session-level diagnosis and treatment contexts; Maintain the session state memory of the diagnosis and treatment session, which includes at least one of the following: confirmed content, content to be updated, intermediate auxiliary results, candidate result cache, historical version information, doctor's manual intervention traces, module trigger records, and final session results.

[0011] According to a human-machine collaborative diagnosis and treatment method provided by the present invention, the step of invoking the doctor's avatar intelligent agent based on the session state and the real-time operation state of the target doctor includes: The collaborative orchestration module dynamically decides whether to trigger an AI-assisted task and the type of task to be triggered based on at least one of the following factors: the current stage of diagnosis and treatment, the type and importance of newly entered information, whether there is content to be updated in the session state, whether the target doctor is in a manual editing state, and the triggering conditions of each skill module.

[0012] According to a human-machine collaborative diagnosis and treatment method provided by the present invention, the triggered task types include at least one of the following: consultation suggestion generation task, medical record content completion task, phased diagnosis prompt task, examination and test suggestion task, treatment and management suggestion task, risk prompt task, and conversation phase summary task.

[0013] According to a human-machine collaborative diagnosis and treatment method provided by the present invention, the step of pausing the output of the auxiliary information to the editing area of ​​the target doctor and temporarily storing the newly generated auxiliary information includes: In response to the detection that the target doctor has edited the diagnosis and treatment content, a manual intervention protection mechanism is activated; During the period when the manual intervention protection mechanism is in effect, one or more of the following processes are performed: suspending direct writing to the doctor's currently edited field, prohibiting the overwriting of confirmed content, converting newly generated content into a candidate result cache, recording the range and content of the field modified by the doctor in this manual intervention, and updating the manual intervention flag in the current session state.

[0014] According to a human-machine collaborative diagnosis and treatment method provided by the present invention, the step of merging the temporarily stored auxiliary information with the content edited by the target doctor includes: In response to the detection that the target doctor has finished editing, the recovery merging mechanism is initiated; In the recovery and merging mechanism, the latest results modified by the doctor and the candidate results cached during the temporary storage period are read; Compare the differences between the latest result and the candidate results; Based on the principle of prioritizing doctors' manual work and the aforementioned differences, the candidate results are inherited, discarded, retained, or partially merged to generate a merged result.

[0015] The human-machine collaborative diagnosis and treatment method provided by the present invention further includes: After the consultation session ends, the session is archived to obtain archived content. The archived content includes at least one of the following: the original event sequence of the session, the trigger records and output results of each skill module, the doctor's adoption and modification behaviors, the location and timing of manual intervention, and the final consultation result. The doctor avatar is updated based on the archived content. The update includes adjusting at least one of the following: personalized parameters of each skill module, doctor expression style characteristics, doctor decision-making preference characteristics, module trigger priority, and human intervention boundary.

[0016] According to the present invention, a human-machine collaborative diagnosis and treatment method is applied to at least one medical service scenario, including outpatient consultation, online audio and video consultation, text and image consultation, and follow-up visit.

[0017] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the human-machine collaborative diagnosis and treatment method as described above.

[0018] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the human-machine collaborative diagnosis and treatment method as described above.

[0019] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the human-machine collaborative diagnosis and treatment method as described above.

[0020] This invention provides a human-machine collaborative diagnosis and treatment method that introduces a doctor-avatar intelligent agent, enabling the generated auxiliary information to be adapted to the personalized characteristics of the target doctor, significantly improving the acceptability and practicality of the auxiliary results. The intervention behavior of the avatar intelligent agent is dynamically adjusted based on the session state and the real-time operation state of the target doctor, keeping the auxiliary process synchronized with the progress of diagnosis and treatment, thus improving the naturalness and smoothness of human-machine collaboration. Output is paused and new results are temporarily saved while the target doctor is editing, and then merged and updated after editing, fundamentally solving the problem of interference between generated content and manual doctor operations, ensuring that the target doctor always maintains a dominant position in the diagnosis and treatment process. Through the saving and merging mechanism, the target doctor can continue to collaborate on the original basis after editing, achieving a seamless and continuous collaborative experience, greatly improving the user experience of human-machine collaboration. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the human-machine collaborative diagnosis and treatment method provided by the present invention; Figure 2This is a schematic diagram of the doctor clone intelligent agent module structure provided by the present invention; Figure 3 This is a schematic flowchart of the combined mechanism of physician-assisted intervention for protection and recovery provided by the present invention; Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0024] Figure 1 This is a flowchart illustrating the human-machine collaborative diagnosis and treatment method provided by the present invention. Figure 2 This is a schematic diagram of the doctor clone intelligent agent module structure provided by the present invention. Figure 3 This is a schematic flowchart of the combined mechanism of doctor's manual intervention for protection and recovery provided by the present invention.

[0025] like Figure 1 As shown, this embodiment provides a human-machine collaborative diagnosis and treatment method that achieves dynamic collaboration between the doctor and an intelligent agent, ensuring the doctor's dominant position in the diagnosis and treatment process. The method mainly includes the following steps: 101. Obtain the target doctor's doctor clone AI agent.

[0026] Specifically, a personalized doctor avatar is pre-built for the target doctor. This avatar is not a generalized model, but rather constructed based on the target doctor's historical medical records, past medical texts, the doctor's adoption and modification results of the AI ​​system's output, the doctor's operation trajectory on the consultation interface, and archived data from historical consultation sessions. Its structure is a multi-skill modular structure; for specific module components, please refer to [reference needed]. Figure 2 The structure diagram of the multi-skill module of the doctor's avatar intelligent agent can fully represent the personalized characteristics of the target doctor's diagnosis and treatment.

[0027] The doctor avatar can lay the foundation for generating personalized auxiliary information that fits the target doctor's work habits, thus avoiding the generalization of AI-assisted information and its disconnect from the doctor's actual needs.

[0028] 102. During the diagnosis and treatment session, collect multi-source diagnosis and treatment events and construct the session state based on the multi-source diagnosis and treatment events.

[0029] Once a treatment session is initiated, all multi-source treatment events related to the current treatment are continuously collected in real time, rather than being collected as static information in a single instance. After collection, various multi-source treatment events undergo unified structured processing, converting unstructured raw events into session-level treatment context. Based on this treatment context, the session state memory is maintained, ultimately forming a complete session state. This session state reflects the current treatment progress, information reserves, and doctor's operational status in real time. This enables continuous reception and unified management of dynamic incremental information during the treatment process, allowing the AI ​​system to continuously and accurately understand the treatment process and preventing auxiliary results from becoming disconnected from the current treatment stage.

[0030] 103. Based on the session status and the target doctor's real-time operation status, invoke the doctor's avatar AI to generate auxiliary information that matches the target doctor's personalized characteristics.

[0031] Specifically, the collaborative orchestration module within the system acts as the core scheduling unit, continuously reading the current session state and the doctor's real-time operational status, such as whether they are idle, editing, or confirming content. Based on this information, it dynamically decides whether to trigger an AI-assisted task. If triggered, it invokes the corresponding skill module within the doctor's AI avatar. This module then generates auxiliary information by combining the target doctor's personalized characteristics. This auxiliary information is tailored to the target doctor's diagnostic habits in terms of expression style, decision-making tendencies, and content organization. This ensures a high degree of match between the timing and content of AI-assisted intervention and the stage of diagnosis and treatment, as well as the doctor's operational status, achieving synchronization between the assistance process and the progress of diagnosis and treatment, and enhancing the naturalness and fluency of human-machine collaboration.

[0032] 104. When it is detected that the target doctor is editing the diagnosis and treatment content, pause the output of auxiliary information to the target doctor's editing area and temporarily save the newly generated auxiliary information.

[0033] Specifically, the system monitors doctors' actions on the diagnostic interface in real time. When it detects that a doctor is editing or revising medical records, diagnoses, treatment suggestions, or other diagnostic content, a manual intervention protection mechanism is immediately activated. Figure 3 As shown, this is the core of the mechanism for combining doctor's manual intervention for protection and recovery. On one hand, it directly suspends the output of any new auxiliary information to the doctor's current editing area to avoid interfering with the doctor's manual editing. On the other hand, if the system still generates new auxiliary information at this time, it caches the information as a candidate result and does not directly write it into the diagnosis and treatment content. This fundamentally avoids the problem of mutual interference and overwriting between AI-generated content and doctor's manual editing operations, ensuring the doctor's dominant position in the diagnosis and treatment process.

[0034] 105. When the target doctor finishes editing the diagnosis and treatment content, merge the temporarily stored auxiliary information with the content edited by the target doctor, and update the session state based on the merging result.

[0035] Specifically, once the system detects that the doctor has completed editing operations and confirmed them, such as switching editing areas, the editing is considered complete. The recovery and merging mechanism is immediately initiated. Following the principle of prioritizing doctor intervention, the temporarily stored auxiliary information is compared and merged with the doctor's latest edited treatment content to generate a merged result. Subsequently, based on this merged result, the current treatment context and session state memory are comprehensively updated, ensuring that the session state remains consistent with the latest treatment situation, providing an accurate state foundation for subsequent collaborative assistance. This achieves seamless human-computer collaboration after doctor editing; doctors do not need to restart the auxiliary function to continue providing collaborative assistance based on the doctor's latest treatment intentions, significantly improving the human-computer collaboration experience.

[0036] Furthermore, based on the above embodiments, this embodiment obtains the doctor avatar intelligent agent of the target doctor, including: constructing the doctor avatar intelligent agent, which includes multiple skill modules, each skill module being used to characterize the personalized features of the target doctor in the corresponding diagnosis and treatment task; wherein, the multiple skill modules include at least one of the following: consultation interaction module, medical record generation module, diagnostic analysis module, examination and test recommendation module, treatment and medication suggestion module, risk warning module, and follow-up and medical order module.

[0037] Specifically, the core action of acquiring a doctor avatar AI agent for a target doctor is to construct a dedicated doctor avatar AI agent. This AI agent adopts a multi-skill modular architecture design, with each skill module used to individually represent the personalized characteristics of the target doctor in the corresponding diagnostic and treatment task. The specific functions of each module and the personalized characteristics it represents are as follows, and the module composition can be referenced. Figure 2 Doctor clone intelligent agent multi-skill module structure diagram.

[0038] The consultation interaction module is used to characterize the target doctor's questioning order, follow-up questioning methods, information focus, and communication style during the consultation process.

[0039] Medical record generation module: used to characterize the target physician's organization and writing habits of medical record content such as chief complaint, present illness, past medical history, physical examination, auxiliary examinations, diagnostic opinions and treatment suggestions.

[0040] Diagnostic Analysis Module: Used to characterize the target physician's judgment methods and diagnostic expression preferences regarding symptoms, medical history, and examination results.

[0041] Examination and testing recommendation module: used to characterize the target physician's examination recommendation habits, recommendation logic, and output boundaries at different stages of diagnosis and treatment.

[0042] Treatment and Medication Recommendation Module: Used to characterize the target physician's expression and preferences for treatment recommendations, intervention recommendations, and medication recommendations.

[0043] Risk alert module: Used to characterize the target doctor's alert style and triggering rules for contraindications, risk factors, and key reminders.

[0044] Follow-up and Medical Orders Module: This module is used to characterize the target physician's output habits in terms of follow-up reminders, lifestyle suggestions, observation suggestions, and follow-up arrangements.

[0045] The aforementioned skill modules can be flexibly selected and combined in one or more ways according to the target doctor's diagnosis and treatment position and work needs. Moreover, the modules do not run in isolation, but share the current diagnosis and treatment context and session state memory to achieve collaborative work between modules.

[0046] The doctor's avatar intelligent agent includes at least the following types of individualized characteristics: Expression characteristics include word preferences, sentence style, result organization, and key point presentation.

[0047] Decision-making characteristics include diagnostic tendencies, examination triggering habits, risk warning methods, and the degree of conservatism in recommendations.

[0048] Process characteristics include consultation sequence, stage switching logic, and priority of different tasks.

[0049] Revision features include common types of modifications doctors make to AI results, modification locations, adoption boundaries, and rejection patterns.

[0050] Interaction characteristics include doctors' operational habits in actions such as manual confirmation, editing, temporary saving, and continued generation.

[0051] The doctor's avatar AI agent can be organized using structured feature templates. The feature template includes at least one or more of the following fields: (1) basic doctor identification information; (2) consultation process preference features; (3) medical record structure and writing style features; (4) diagnostic analysis expression features; (5) examination and testing recommendation preference features; (6) treatment and medication suggestion expression features; (7) risk warning preference features; (8) human intervention boundary features; and (9) historical adoption and revision statistics features. During system execution, the collaborative orchestration module calls the corresponding task feature subset from the feature template based on the current task to be executed, thereby constraining the expression method, result organization method, and output boundary of the AI ​​output content. Different skill modules can call different feature subsets, thus achieving module-level individualized collaboration.

[0052] By breaking down doctors' personalized treatment characteristics into various sub-tasks, the doctor's AI avatar can accurately match the doctor's work habits in different treatment stages. It can build and update auxiliary information by using historical treatment records, previous medical records, doctors' adoption and modification results of system outputs, doctors' operation trajectories on the interface, and session archive data, thereby improving the relevance and practicality of auxiliary information.

[0053] Furthermore, based on the above embodiments, the multi-source diagnosis and treatment events in this embodiment include at least one of the following: consultation voice events, voice-to-text events, patient input text events, patient upload attachment events, examination result events, doctor operation events, and auxiliary result events generated internally by the system.

[0054] Specifically, multi-source diagnostic events refer to all diagnostic and treatment-related events generated during the diagnostic and treatment session, including at least one or more of the following. The system captures and records these events in real time through channels such as the diagnostic and treatment interface, audio and video acquisition devices, and file upload interfaces: Voice events during consultations: Voice information during consultations between doctors and patients, including the doctor's questions and the patient's description of their condition.

[0055] Speech-to-text events: Text information obtained by the system after transcribing the consultation speech event using ASR (Automatic Speech Recognition).

[0056] Patient input text events: Text information directly entered by the patient through the treatment interface, such as supplementary information about the condition, questions, etc.

[0057] Patient-uploaded attachments: Patient-uploaded attachments related to their condition, such as images of examination reports, medical records, and imaging data.

[0058] Examination Result Events: New patient examination and test results information added during the diagnosis and treatment process.

[0059] Doctor's action events: All actions performed by the doctor on the diagnosis and treatment interface, such as inputting, editing, confirming, deleting, saving, and switching diagnosis and treatment interfaces.

[0060] Internal auxiliary result events generated by the system: Internal events such as phased auxiliary information and candidate suggestions generated by the system during the collaboration process.

[0061] By collecting information from all dimensions during the diagnosis and treatment process, we can ensure that the artificial intelligence system can obtain complete diagnostic and treatment evidence, thus avoiding inaccurate auxiliary results due to missing information.

[0062] Furthermore, based on the above embodiments, this embodiment constructs a session state based on multi-source diagnosis and treatment events, including: performing structured processing on multi-source diagnosis and treatment events and converting them into session-level diagnosis and treatment context; maintaining the session state memory of the diagnosis and treatment session, the session state memory including at least one of confirmed content, content to be updated, intermediate auxiliary results, candidate result cache, historical version information, doctor's manual intervention traces, module trigger records, and final session results.

[0063] Specifically, instead of simply piecing together each collected multi-source medical event, the system first performs event classification, field mapping, incremental merging, and conflict resolution. Similar events are grouped into the same information dimension, unstructured events are mapped to pre-defined structured fields, and incremental information from newly added events is merged into existing information. Inconsistent event information is then resolved. Finally, all processed information is organized into a session-level medical context. This context is structured data, including at least six major field groups: patient basic information, current session progress, clinical judgment information, document results information, human intervention information, and collaborative scheduling information. This clearly and comprehensively reflects all core information of the current medical treatment.

[0064] The patient information group includes patient age, gender, chief complaint, medical history, allergy history, and previous examination results. The current conversation progress group includes the current consultation round, current treatment stage, discussed issues, and follow-up questions. The clinical judgment information group includes the stage-specific diagnostic tendency, confirmed diagnostic information, and pending judgment items. The document results information group includes the current draft medical record, confirmed medical record segments, and suggested supplementary segments. The human intervention information group includes the doctor's current editing location, confirmed fields, and the most recent human modification history. The collaborative scheduling information group includes whether auxiliary functions are allowed, the types of tasks allowed to be triggered, and the candidate result cache status.

[0065] Based on a structured diagnostic and treatment context, a session state memory is established and maintained in real time, continuously saving and updating key state information during the diagnostic and treatment process. Confirmed content refers to content explicitly confirmed by the doctor and not suitable for direct modification in subsequent generation. Content awaiting update includes content that may continue to change as new information arrives. Intermediate auxiliary results are content generated by the system at stages but not yet included in the final result. Candidate result cache contains auxiliary information temporarily stored during doctor editing. Historical version information is a historical record of diagnostic and treatment content and status, facilitating backtracking. Doctor manual intervention traces record doctor actions such as editing, deleting, and replacing system results. Module trigger records the start time and output results of each skill module. The final session result is the final diagnostic and treatment content determined after the diagnosis and treatment concludes.

[0066] By converting chaotic, multi-source diagnostic and treatment events into structured information that the artificial intelligence system can understand and access, and by using session state memory to distinguish and manage various states during the diagnostic and treatment process, an accurate and comprehensive state foundation is provided for collaborative orchestration, human intervention protection, and result recovery and merging, ensuring that the artificial intelligence maintains consistent understanding throughout multiple rounds of diagnostic and treatment processes.

[0067] Furthermore, based on the above embodiments, this embodiment calls the doctor's clone intelligent agent according to the session state and the real-time operation state of the target doctor. This includes: the collaborative orchestration module dynamically deciding whether to trigger an artificial intelligence-assisted task and the type of task to be triggered based on at least one of the following factors: the current diagnosis and treatment stage, the type and importance of the newly entered information, whether there is content to be updated in the session state, whether the target doctor is in a manual editing state, and the triggering conditions of each skill module.

[0068] Specifically, the core decision-making body for invoking the doctor's AI avatar is the collaborative orchestration module. As the core scheduling unit of the system, this module reads the current session state and the target doctor's real-time operation state in real time, and makes a comprehensive judgment based on multiple factors to dynamically decide whether to trigger an AI-assisted task and the type of task to be triggered. The factors to be considered include at least one or more of the following: Current stage of diagnosis and treatment: such as early consultation, middle consultation, diagnosis stage, treatment recommendation stage, etc.

[0069] The type and importance of newly entered information: such as whether the newly collected event is a core description of the patient's condition or a general inquiry.

[0070] Does the session status contain content awaiting updates? For example, are there incomplete medical records or unconfirmed diagnostic tendencies?

[0071] Is the target doctor in manual editing mode? If the doctor is in editing mode, the auxiliary task will not be triggered at this time.

[0072] Triggering conditions for each skill module: Each skill module has a preset triggering condition. For example, the consultation interaction module is triggered when the consultation information is insufficient.

[0073] Based on a comprehensive assessment of the above factors, the collaborative orchestration module determines whether to activate AI assistance. If activated, it selects and invokes the corresponding skill module to ensure that the assistance task matches the current diagnostic and treatment needs. This prevents the triggering of AI-assisted tasks from being a fixed, task-oriented invocation, but rather dynamically adjusted according to the actual diagnostic and treatment situation, avoiding the blind triggering of assistance tasks and ensuring that the assistance process is highly synchronized with the progress of diagnosis and treatment.

[0074] Furthermore, based on the above embodiments, the triggered task types in this embodiment include at least one of the following: consultation suggestion generation task, medical record content completion task, phased diagnosis prompt task, examination and testing suggestion task, treatment and management suggestion task, risk prompt task, and conversation phase summary task.

[0075] Specifically, the collaborative orchestration module triggers at least one or more of the following AI-assisted task types based on the comprehensive judgment results. Each task type corresponds one-to-one with the skill modules of the doctor's AI avatar, covering the entire diagnosis and treatment process: Consultation suggestion generation task: Adapted to stages where consultation information is insufficient, this task is executed by the consultation interaction module to generate suggestions for further consultation.

[0076] Medical record content completion task: Adapted to the medical record drafting and improvement stages, this task is executed by the medical record generation module to complete any incomplete medical record content.

[0077] Phased diagnostic prompts: As diagnostic and treatment information becomes more comprehensive, this task is performed by the diagnostic analysis module to generate phased diagnostic tendencies and prompts.

[0078] The task of recommending examinations and tests is performed by the examination and testing recommendation module when the diagnosis needs further verification.

[0079] Treatment and Management Recommendations: This task is performed by the Treatment and Medication Recommendations module after a definitive diagnosis, generating treatment, management, and medication recommendations.

[0080] Risk alert task: During the treatment recommendation and medical order formulation stage, the risk alert module performs the task and generates risk and contraindication alerts related to the condition.

[0081] The conversation phase summary task adapts to the nodes of each stage of diagnosis and treatment, and is executed collaboratively by the corresponding modules to summarize the diagnosis and treatment information of the current stage.

[0082] For example, in the early stages of a consultation, the consultation interaction module can be triggered first to generate suggestions for further follow-up questions; as the diagnostic information becomes more comprehensive, the system can trigger the medical record generation module and the diagnostic analysis module to form a preliminary draft; once the diagnosis becomes clearer, the examination and testing recommendation module, the treatment and medication suggestion module, and the risk warning module can be further triggered to generate auxiliary results. Through this collaborative orchestration mechanism, this invention enables different skill modules to no longer work in isolation, but rather to be scheduled and coordinated at the task level around the progress of the conversation and the doctor's status. By providing precise assistance to each stage of the entire diagnosis and treatment process, and offering targeted auxiliary tasks at different stages, artificial intelligence assistance can be integrated throughout the entire diagnosis and treatment process, improving the efficiency of doctors' diagnostic and treatment work.

[0083] Furthermore, based on the above embodiments, this embodiment suspends the output of auxiliary information to the target doctor's editing area and temporarily stores the newly generated auxiliary information, including: in response to detecting that the target doctor is editing the diagnosis and treatment content, a manual intervention protection mechanism is activated; during the period when the manual intervention protection mechanism is in effect, one or more of the following processes are performed: suspending direct writing to the doctor's currently edited field, prohibiting the overwriting of confirmed content, converting the newly generated content into a candidate result cache, recording the field range and content modified by the doctor this time, and updating the manual intervention flag in the current session state.

[0084] Specifically, upon detecting that the target doctor has edited the treatment content, the system immediately responds and activates a manual intervention protection mechanism. During the period when this mechanism is in effect, the system performs one or more of the following processing operations to ensure that the doctor's manual editing operation is not interfered with: Suspend direct writing to the field the doctor is currently editing: Prohibit any auxiliary information from being directly written to the field the doctor is editing, thus avoiding overwriting at the source.

[0085] Overwriting confirmed content is prohibited: Even if the doctor edits other fields, the system is prohibited from making any modifications to the diagnosis and treatment content that the doctor has confirmed.

[0086] Convert newly generated content into candidate result cache: If the system generates new auxiliary information at this time, store it in the candidate result cache area without displaying or writing it.

[0087] Record the scope and content of the fields modified by the doctor: keep a detailed record of the doctor's editing behavior to form a trace of the doctor's manual intervention, and provide a basis for subsequent result merging.

[0088] Update the manual intervention flag in the current session state: Mark the current stage as doctor manual intervention in the session state memory, so that the collaborative orchestration module can pause triggering new auxiliary tasks accordingly.

[0089] By establishing a comprehensive human intervention and protection mechanism, conflicts and overlaps between artificial intelligence systems and doctors' manual editing operations can be completely avoided at the operational level, ensuring doctors' dominant editing rights over diagnostic and treatment content and improving their user experience.

[0090] Furthermore, based on the above embodiments, this embodiment merges the temporarily stored auxiliary information with the content edited by the target doctor, including: in response to detecting that the target doctor has finished editing, starting the recovery merging mechanism; in the recovery merging mechanism, reading the latest result modified by the doctor and the candidate results cached during the temporary storage period; comparing the differences between the latest result and the candidate results; and according to the doctor's manual priority principle and differences, inheriting, discarding, retaining or partially merging the candidate results to generate the merged result.

[0091] Specifically, upon detecting that the target doctor has finished editing the treatment content, the system immediately responds and initiates a recovery and merging mechanism, performing the result merging operation according to the doctor's manual intervention priority principle. It reads the doctor's latest modified result and the candidate results cached during the temporary storage period, retrieves the doctor's final edited treatment content from the system, and all candidate auxiliary information cached during the manual intervention protection mechanism's effective period. It compares the differences between the latest result and the candidate results, performing field-level and content-level comparisons to identify consistent, conflicting, and complementary parts. Based on the doctor's manual intervention priority principle and the aforementioned differences, it inherits, discards, retains, or partially merges the candidate results, generating a merged result. Candidate results consistent with the latest result are directly inherited; candidate results conflicting with the latest result are directly discarded; candidate results that do not affect the latest result are retained; and candidate results that complement the latest result are partially merged. The entire process adheres to the doctor's manual content priority principle, ensuring that the doctor's edited results are not altered.

[0092] By intelligently integrating temporary auxiliary information with the doctor's edited content, the value of the system's auxiliary information is preserved while strictly adhering to the doctor's diagnostic and treatment intentions. After the doctor finishes editing, the system can continue to provide collaborative assistance based on the latest diagnostic and treatment content, achieving seamless human-machine collaboration.

[0093] Furthermore, based on the above embodiments, this embodiment also includes: after the diagnosis and treatment session ends, archiving the session to obtain archived content, which includes at least one of the following: the original event sequence of the session, trigger records and output results of each skill module, doctor's adoption and modification behavior, location and timing of manual intervention, and final diagnosis and treatment results; updating the doctor's avatar agent based on the archived content, which includes adjusting at least one of the following: personalized parameters of each skill module, doctor's expression style characteristics, doctor's decision-making preference characteristics, module trigger priority, and manual intervention boundary.

[0094] Specifically, after the consultation session is completed, all information about the entire consultation process is comprehensively archived. The archived content is not just the final consultation result, but includes relevant information about the entire consultation process, including at least one or more of the following: the original event sequence of the session, the trigger records and output results of each skill module, the doctor's adoption and modification behavior, the location and timing of manual intervention, and the final consultation result, to ensure that the archived information can completely restore the collaborative process of this consultation.

[0095] Based on the archived content of this diagnosis and treatment, the target doctor's avatar AI agent undergoes targeted and personalized updates, rather than general model retraining. The updates include adjustments to at least one of the following: personalized parameters for each skill module, doctor's expression style characteristics, doctor's decision-making preferences, module trigger priorities, and human intervention boundaries. If the system output is directly adopted by the doctor, the trigger priority of the corresponding module in similar contexts is enhanced. If the system output is partially modified by the doctor, the modification differences are recorded and the module's expression template is updated. If the system output is rejected by the doctor, the priority of that output method in similar scenarios is reduced.

[0096] By forming a continuous learning loop tailored to specific doctors, the doctor's AI avatar can gradually adapt to the target doctor's diagnostic and treatment habits through continuous use, making subsequent auxiliary information increasingly accurate and improving the long-term effectiveness of human-machine collaboration.

[0097] Furthermore, based on the above embodiments, the human-machine collaborative diagnosis and treatment method is mainly applied to at least one medical service scenario among outpatient consultation, online audio and video consultation, text and image consultation, and follow-up visits.

[0098] Specifically, in outpatient consultations: adapting to real-time diagnosis and treatment scenarios in offline outpatient clinics, the system collects multi-source diagnosis and treatment events through outpatient diagnosis and treatment terminals to provide doctors with real-time on-site collaborative assistance.

[0099] Online audio and video consultation: Adapted to remote diagnosis and treatment scenarios with online audio and video, the system acquires consultation voice events through audio and video acquisition and transcription functions, and collects multi-source events by combining online text input and attachment upload functions to achieve remote human-computer collaboration.

[0100] Text-based online consultation: Adapted to lightweight online consultation scenarios using text and images, the system collects patients' text and image inputs, attachment uploads, and doctors' text operation events, and provides targeted auxiliary information such as medical record generation and diagnostic prompts.

[0101] Follow-up visits: Adapted to scenarios of patient follow-up visits and postoperative follow-ups, the system combines the patient's historical medical records and multiple events of this follow-up visit to provide doctors with auxiliary information such as follow-up suggestions and reminders for re-examination of the condition.

[0102] In each application scenario, slight adjustments are made to the collection channels for multi-source diagnosis and treatment events and the output format of auxiliary information according to the characteristics of the scenario to ensure adaptability.

[0103] In practical applications, taking a respiratory medicine outpatient setting as an example, doctor A is conducting a consultation with a patient who complains of "coughing for three days, accompanied by a low-grade fever." The consultation mainly includes the following steps: Step 1: The system pre-constructs a doctor avatar intelligent agent corresponding to Doctor A. This doctor avatar intelligent agent includes at least a consultation interaction module, a medical record generation module, a diagnostic analysis module, an examination and test recommendation module, a treatment and medication suggestion module, and a risk warning module. Based on Doctor A's historical consultation records, previous medical record texts, and Doctor A's adoption and revision behavior of system suggestions, the system extracts Doctor A's question order preference, medical record writing habits, diagnostic expression methods, and suggestion output boundaries.

[0104] Step 2: After the start of this outpatient session, the system continuously receives the patient's voice, speech-to-text, supplementary text from the patient, images of previous examination reports, and the doctor's input in the medical record editing interface.

[0105] Step 3: The system processes the above multi-source events in a unified manner, organizing them into a structured diagnostic context. The context includes at least the patient's basic information, current chief complaint, fragments of present medical history, information already questioned, information to be questioned, stage-specific diagnostic tendencies, and the current draft medical record.

[0106] Step 4: The system establishes the state memory for this session. The state memory records the confirmed patient's basic information, the medical record content to be improved, intermediate diagnostic prompts, historical version information, and whether the doctor is in manual editing mode.

[0107] Step 5: As the consultation progresses, the collaborative orchestration module determines from the context that the patient's information regarding cough and low-grade fever is insufficient, triggering the consultation interaction module to generate suggestions for further follow-up questions, such as the specific temperature of the fever, sputum characteristics, contact history, and past respiratory medical history. This suggestion is not a generic output but is generated based on Doctor A's preferred consultation sequence and expression.

[0108] Step 6: After the patient adds "significant cough at night with yellow phlegm", the system updates the structured diagnosis and treatment context and further triggers the medical record generation module and the diagnostic analysis module to generate a staged medical record draft and diagnostic prompts.

[0109] Step 7: When Doctor A begins manually revising the description of the present illness in the medical record, the system recognizes that the doctor is in manual editing mode and activates the manual intervention protection mechanism. At this time, the system no longer directly overwrites the doctor's current editing area, but instead caches the newly generated suggestions as candidate results.

[0110] Step 8: After Doctor A completes the editing and confirms, the system starts the recovery and merging mechanism, reads Doctor A's latest editing results and the candidate results cached during the protection period, merges them according to the principle of prioritizing manual editing by doctors, and continues to trigger the examination and testing recommendation module and risk warning module based on the new session state.

[0111] Step 9: After the outpatient visit concludes, the system archives the event sequence of the entire session, the output of each module, the adoption and revision status of Doctor A, and the final medical record results, and updates the corresponding doctor avatar AI agent for Doctor A accordingly. If the system detects that Doctor A frequently modifies the examination suggestions in the scenario of "yellow sputum, low-grade fever, and nighttime cough" to a specific expression, it will prioritize generating results that better match Doctor A's habits in subsequent similar scenarios.

[0112] As can be seen from the specific examples, this invention does not simply call artificial intelligence to generate text in outpatient clinics, but rather forms a continuous collaborative diagnosis and treatment process for specific doctors through mechanisms such as doctor avatar modeling, multi-source event structuring, state memory, task orchestration, manual intervention protection, and session archive updates.

[0113] The method of the present invention has the following advantages: First, this invention does not treat artificial intelligence as an auxiliary tool outside the diagnosis and treatment process, but rather constructs it as a collaborative entity that can continuously operate around the doctor's real work process, thereby enabling artificial intelligence to be integrated more naturally into the doctor's diagnosis and treatment workflow and reducing the sense of disconnect and interruption in the process.

[0114] Secondly, this invention achieves continuous understanding and unified management of dynamic incremental information in the diagnosis and treatment process through multi-source event reception, structured context construction, and session state memory mechanism, enabling the AI ​​output to be continuously updated as the consultation progresses, and better match the doctor's real-time decision-making rhythm.

[0115] Third, by introducing a doctor-avatar intelligent agent, this invention enables the AI ​​output to be personalized according to different doctors' consultation habits, expression styles, medical record organization methods, and decision-making preferences, thereby improving the consistency between the auxiliary results and the actual needs of doctors and increasing the acceptability and practicality of AI results.

[0116] Fourth, this invention sets up a doctor's manual intervention protection mechanism and a subsequent recovery and merging mechanism, so that the artificial intelligence will not interfere with or cover key content during the doctor's editing, and can continue to collaborate based on the doctor's latest intention after the editing is completed, thereby effectively solving the problem of easy conflict between artificial intelligence and doctor's manual operation in the prior art.

[0117] Fifth, through a session archiving and continuous optimization mechanism, this invention enables the artificial intelligence system to continuously learn from and gradually adapt to specific doctors, forming a closed-loop optimization path of "doctor use - system feedback - capability enhancement," which is conducive to improving the synergistic effect of the system in long-term application.

[0118] Sixth, this invention is applicable to a variety of medical service scenarios, has good versatility and scalability, and can provide a unified methodological basis for the practical application of the "doctor + AI" collaborative work model in medical scenarios.

[0119] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention.

[0120] like Figure 4 As shown, the electronic device may include a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, communications interface 420, and memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a human-machine collaborative diagnosis and treatment method. This method includes: acquiring a doctor's avatar intelligent agent for the target doctor; collecting multi-source diagnosis and treatment events during the diagnosis and treatment session, and constructing a session state based on the multi-source diagnosis and treatment events; calling the doctor's avatar intelligent agent according to the session state and the real-time operation state of the target doctor to generate auxiliary information that conforms to the personalized characteristics of the target doctor; when it is detected that the target doctor is editing the diagnosis and treatment content, pausing the output of the auxiliary information to the target doctor's editing area and temporarily storing the newly generated auxiliary information; when it is detected that the target doctor has finished editing the diagnosis and treatment content, merging the temporarily stored auxiliary information with the content edited by the target doctor, and updating the session state based on the merging result.

[0121] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0122] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the human-computer collaborative diagnosis and treatment method provided by the above methods. The method includes: acquiring a doctor avatar intelligent agent of a target doctor; collecting multi-source diagnosis and treatment events during the diagnosis and treatment session, and constructing a session state based on the multi-source diagnosis and treatment events; calling the doctor avatar intelligent agent according to the session state and the real-time operation state of the target doctor to generate auxiliary information that conforms to the personalized characteristics of the target doctor; when it is detected that the target doctor is editing the diagnosis and treatment content, pausing the output of the auxiliary information to the editing area of ​​the target doctor, and temporarily storing the newly generated auxiliary information; when it is detected that the target doctor has finished editing the diagnosis and treatment content, merging the temporarily stored auxiliary information with the content edited by the target doctor, and updating the session state based on the merging result.

[0123] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the human-computer collaborative diagnosis and treatment method provided by the above methods. The method includes: acquiring a doctor avatar intelligent agent of a target doctor; during a diagnosis and treatment session, collecting multi-source diagnosis and treatment events and constructing a session state based on the multi-source diagnosis and treatment events; according to the session state and the real-time operation state of the target doctor, invoking the doctor avatar intelligent agent to generate auxiliary information that conforms to the personalized characteristics of the target doctor; when it is detected that the target doctor is editing the diagnosis and treatment content, pausing the output of the auxiliary information to the editing area of ​​the target doctor and temporarily storing the newly generated auxiliary information; when it is detected that the target doctor has finished editing the diagnosis and treatment content, merging the temporarily stored auxiliary information with the content edited by the target doctor, and updating the session state based on the merging result.

[0124] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0125] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A human-machine collaborative diagnosis and treatment method, characterized in that, include: Obtain the target doctor's avatar AI agent; During the diagnosis and treatment session, multi-source diagnosis and treatment events are collected, and the session state is constructed based on the multi-source diagnosis and treatment events. Based on the session state and the real-time operation state of the target doctor, the doctor's avatar AI is invoked to generate auxiliary information that conforms to the personalized characteristics of the target doctor; When the target doctor is detected to be editing the diagnosis and treatment content, the output of the auxiliary information to the target doctor's editing area is paused, and the newly generated auxiliary information is temporarily stored. or, When the target doctor finishes editing the treatment content, the temporarily stored auxiliary information is merged with the content edited by the target doctor, and the session state is updated based on the merging result.

2. The human-machine collaborative diagnosis and treatment method according to claim 1, characterized in that, The intelligent agent that acquires the target doctor's avatar includes: The doctor avatar intelligent agent is constructed, which includes multiple skill modules, each of which is used to characterize the personalized features of the target doctor in the corresponding diagnosis and treatment task; The multiple skill modules include at least one of the following: consultation interaction module, medical record generation module, diagnostic analysis module, examination and test recommendation module, treatment and medication suggestion module, risk warning module, and follow-up and medical order module.

3. The human-machine collaborative diagnosis and treatment method according to claim 2, characterized in that, The consultation interaction module is used to characterize the target doctor's questioning order, follow-up questioning methods, information focus, and communication style during the consultation process; The medical record generation module is used to characterize the target physician's organization and writing habits of chief complaint, present illness, past medical history, physical examination, auxiliary examinations, diagnostic opinions and treatment suggestions; The diagnostic analysis module is used to characterize the target physician's judgment methods and diagnostic expression preferences regarding symptoms, medical history, and examination results. The examination and testing recommendation module is used to characterize the target doctor's examination suggestion habits, recommendation logic, and output boundaries at different stages of diagnosis and treatment. The treatment and medication recommendation module is used to characterize the target physician's expression and preferences regarding treatment recommendations, intervention recommendations, and medication recommendations. The risk alert module is used to characterize the target doctor's alert style and triggering rules for contraindications, risk factors, and key reminders. The follow-up and medical advice module is used to characterize the target doctor's output habits in terms of follow-up reminders, lifestyle suggestions, observation suggestions, and follow-up arrangements.

4. The human-machine collaborative diagnosis and treatment method according to claim 1, characterized in that, The multi-source diagnosis and treatment events include at least one of the following: consultation voice events, voice-to-text events, patient input text events, patient upload attachment events, examination result events, doctor operation events, and auxiliary result events generated internally by the system.

5. The human-machine collaborative diagnosis and treatment method according to claim 1, characterized in that, The construction of session state based on the multi-source diagnosis and treatment events includes: The multi-source diagnosis and treatment events are structured and converted into session-level diagnosis and treatment contexts; Maintain the session state memory of the diagnosis and treatment session, which includes at least one of the following: confirmed content, content to be updated, intermediate auxiliary results, candidate result cache, historical version information, doctor's manual intervention traces, module trigger records, and final session results.

6. The human-machine collaborative diagnosis and treatment method according to claim 1, characterized in that, The step of invoking the doctor's avatar AI agent based on the session state and the target doctor's real-time operation state includes: Based on at least one of the following factors: the current stage of diagnosis and treatment, the type and importance of newly entered information, whether there is content to be updated in the session, whether the target doctor is in a manual editing state, and the triggering conditions of each skill module, a dynamic decision is made on whether to trigger an AI-assisted task and the type of task to be triggered.

7. The human-machine collaborative diagnosis and treatment method according to claim 6, characterized in that, The triggered task types include at least one of the following: consultation suggestion generation task, medical record content completion task, phased diagnosis prompt task, examination and test suggestion task, treatment and management suggestion task, risk prompt task, and conversation phase summary task.

8. The human-machine collaborative diagnosis and treatment method according to claim 1, characterized in that, The step of pausing the output of the auxiliary information to the target doctor's editing area and temporarily storing the newly generated auxiliary information includes: In response to the detection that the target doctor has edited the diagnosis and treatment content, a manual intervention protection mechanism is activated; During the period when the manual intervention protection mechanism is in effect, one or more of the following processes are performed: suspending direct writing to the doctor's currently edited field, prohibiting the overwriting of confirmed content, converting newly generated content into a candidate result cache, recording the range and content of the field modified by the doctor in this manual intervention, and updating the manual intervention flag in the current session state.

9. The human-machine collaborative diagnosis and treatment method according to claim 1, characterized in that, The step of merging the temporarily stored auxiliary information with the content edited by the target doctor includes: In response to the detection that the target doctor has finished editing, the recovery merging mechanism is initiated; In the recovery and merging mechanism, the latest results modified by the doctor and the candidate results cached during the temporary storage period are read; Compare the differences between the latest result and the candidate results; Based on the principle of prioritizing doctors' manual work and the aforementioned differences, the candidate results are inherited, discarded, retained, or partially merged to generate a merged result.

10. The human-machine collaborative diagnosis and treatment method according to any one of claims 1-9, characterized in that, Also includes: After the consultation session ends, the session is archived to obtain archived content. The archived content includes at least one of the following: the original event sequence of the session, the trigger records and output results of each skill module, the doctor's adoption and modification behaviors, the location and timing of manual intervention, and the final consultation result. The doctor avatar is updated based on the archived content. The update includes adjusting at least one of the following: personalized parameters of each skill module, doctor expression style characteristics, doctor decision-making preference characteristics, module trigger priority, and human intervention boundary.

Images