Remote medical inquiry system based on AI identification

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a remote medical inquiry system based on AI recognition, which comprises a voice structure analysis module, a semantic association mapping module, a knowledge graph reasoning module, an inquiry path adjustment module and a disease traceability analysis module. According to the method, the logic relation of a description chain is accurately obtained by performing specific value mapping on the time delay and the semantic frequency of the voice information, the structural perception ability of the semantic information is enhanced, the symptom frequency and the health index trend are combined, accurate measurement is performed on abnormal fluctuation dependency, inquiry priority grading is optimized, and the key symptom recognition sensitivity is improved; through induction of a periodic fluctuation trend, formation of a dynamic evolution sequence, enhancement of the monitoring capability of health risk evolution, improvement of decision precision and optimization of a diagnosis chain, overall processing is performed through time sequence and semantic association analysis, the recognition and intervention capability of a complex health state is enhanced, and the health assessment adaptability and accuracy in multiple rounds of interaction are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a remote medical consultation system based on AI recognition. Background Art

[0002] The field of artificial intelligence (AI) encompasses various technical means and methods for using computers to simulate human intelligence, primarily involving perception, cognition, decision-making, and execution. Core elements of this field include machine learning-based pattern recognition, natural language processing, human-computer interaction, and the realization of autonomous learning and reasoning capabilities. In telemedicine, AI is driving the intelligent development of services such as medical consultations, auxiliary diagnosis, and health management through voice recognition, image analysis, and language understanding. AI encompasses multiple aspects, including data collection and preprocessing, model training and optimization, and the design of human-computer interaction mechanisms, gradually enabling the implementation of intelligent applications in various scenarios.

[0003] Among them, the telemedicine consultation system refers to an information processing and interaction system based on artificial intelligence technology, which is used to realize remote and intelligent medical consultation services. The technical matters targeted by the system cover the automatic collection and structured organization of patient health information, semantic understanding and automatic classification of consultation content, knowledge matching and reasoning analysis for preliminary diagnosis of symptoms, etc. It specifically converts user expressions into structured text by deploying voice recognition, and then extracts key diagnosis and treatment information through semantic analysis technology. It combines knowledge graphs and rule-based reasoning mechanisms to match and feedback symptoms. The system also realizes intelligent communication between doctors and patients based on natural language processing, and completes functions such as multiple rounds of question and answer and health advice push.

[0004] Existing speech-to-text processing relies heavily on standard recognition and shallow semantic extraction, lacking in-depth exploration of time delays and logical connections between descriptions. This makes it difficult to accurately restore the contextual relationships between some symptoms. For example, when there are multiple consecutive symptom descriptions, the system struggles to distinguish primary and secondary logic, resulting in semantic misalignment. The handling of abnormal health indicator fluctuations is limited to threshold judgments, ignoring the dependency structure between them and symptom semantics. This can easily lead to a one-sided interpretation of abnormal signals and affect the reasonable determination of the consultation level. For high-frequency descriptions in multiple rounds of Q&A, the system lacks a cyclical evolution mechanism, making it difficult to accurately identify progressive risk trends. Path planning is based on static knowledge matching, failing to incorporate dynamic factors such as temporality and volatility, easily leading to suboptimal diagnostic paths and delaying the processing of key nodes. For symptom tracing, the system relies on static rules or templates and lacks parallel analysis of response differences and fluctuation characteristics, resulting in interruptions in the diagnostic information chain and affecting the continuity and accuracy of cause identification. These deficiencies are particularly prominent in high-concurrency and complex disease scenarios, directly impacting the professionalism and reliability of intelligent consultation services. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the existing technology and propose a remote medical consultation system based on AI recognition.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solutions: A remote medical consultation system based on AI recognition includes: The speech structure analysis module obtains the patient's speech expression, symptom description, and health indicator data, analyzes the semantic relationship, determines the semantic direction and description chain length based on the time delay and semantic sequence between the speech expression and symptom description, and maps the ratio of semantic frequency to time delay to generate a health status description association model; The semantic association mapping module selects the symptom description frequency and health indicator change trend of key nodes based on the health status description association model, matches and compares the abnormal fluctuation frequency with the semantic dependency degree, identifies the key level of consultation, and obtains the node consultation priority scale; The knowledge graph reasoning module extracts the time interval and duration of symptom descriptions based on the node consultation priority scale, determines the distribution stage of high-frequency descriptions, classifies the periodic trends of similar symptom descriptions, and generates a health risk evolution sequence; The consultation path adjustment module calls the health risk evolution sequence, combines the abnormal fluctuation value and response time delay in the section under its jurisdiction, screens the target diseases in the high trend section, revises the consultation path value in sections, and obtains the consultation path optimization table.

[0007] As a further solution of the present invention, the health status description association model includes semantic direction classification, description chain path structure, and semantic delay ratio distribution; the node consultation priority scale includes consultation level label, node risk coefficient, and priority score interval; the health risk evolution sequence includes description interval distribution, description continuous frequency, and periodic classification results; the consultation path optimization table includes a target disease list, path revision parameters, and delay association factors.

[0008] As a further solution of the present invention, the speech structure analysis module includes: The speech and symptom recognition submodule obtains the patient's speech expression, symptom description, and health indicator data, aligns the three signals by timestamp, extracts symptom description mutation points, locates changes in speech expression and health indicators, identifies semantic differences, and generates symptom description response deviation; The semantic lag analysis submodule extracts the speech signal delay segment and the symptom signal response sequence based on the symptom description response deviation, calculates the delay length and response sequence, determines the signal semantic direction, measures the number of description chain trigger sequences within the lag segment, and generates the semantic signal lag structure quantity; The symptom description structure generation submodule extracts the linkage period between the health indicator change points and the description chain signals based on the semantic signal hysteresis structure, compares the signal response frequency and delay, identifies the linkage times and the average delay ratio, analyzes the semantic relationship and linkage frequency mapping, and generates a health status description association model.

[0009] As a further solution of the present invention, the semantic association mapping module includes: The fluctuation extraction submodule extracts the frequency of key node symptom descriptions based on the health status description association model, filters out sequences with fluctuations exceeding the benchmark value, analyzes the proportion of abnormal fluctuations, and generates an abnormal fluctuation frequency ratio; The frequency dependency matching submodule calls the abnormal fluctuation frequency ratio, combines the node description frequency and the number of dependency paths, identifies the frequency dependency comparison item, adjusts the node abnormal offset according to the proportion of tasks on the dependency path, and calculates the fluctuation coupling priority index; The priority determination submodule extracts the numerical interval corresponding to the node based on the fluctuation coupling priority index, sets the node level division interval group, assigns the level identifier according to the index value falling into the interval, sorts and determines the consultation level, and obtains the node consultation priority scale.

[0010] As a further solution of the present invention, the knowledge graph reasoning module includes: The description record screening submodule collects the time points of symptom description triggering and ending according to the node consultation priority scale, identifies the time intervals between adjacent descriptions, filters out records exceeding the benchmark value, and generates a description time interval sequence; The description period division submodule calls the description time interval sequence, counts the number of consecutive descriptions of the same type of description events, calculates the description period distribution trend value based on the description duration, interval time and health indicator changes, and establishes the description period distribution stage; The periodic trend classification submodule calls the periodic distribution description stage, identifies description events of the same periodic trend, identifies trend fluctuation differences, classifies and archives them according to difference thresholds, and generates a health risk evolution sequence.

[0011] As a further solution of the present invention, the consultation path adjustment module includes: The trend sequence identification submodule calls the health risk evolution sequence, extracts the incremental change value, determines the trend rising interval based on the cumulative increase and fluctuation amplitude, marks the key symptoms, and generates an automated consultation trend anomaly identification list; The response characteristic extraction submodule calls the disease number in the automated medical consultation trend abnormality identification list, calculates the abnormal fluctuation value within the segment, identifies the response delay, combines the delay distribution and the fluctuation frequency, analyzes the disease response difference, and obtains the automated medical consultation response characteristic difference value; The path revision submodule identifies the path revision item structure based on the difference value of the automated medical consultation response characteristics and the path corresponding to the trend segment, analyzes the path deviation degree and frequency distribution amplitude, calculates the path revision deviation degree, and adjusts the medical consultation frequency and time distribution in combination with the original path structure difference to obtain the medical consultation path optimization table.

[0012] As a further solution of the present invention, the system also includes a disease tracing and analysis module: The symptom traceability analysis module collects the response differences of high-frequency symptom consultations and the fluctuation characteristics of health indicators based on the consultation path optimization table, determines whether the fluctuation characteristics are consistent, and identifies the symptom attribution nodes to form health risk traceability analysis results; The health risk traceability analysis results include a fluctuation consistent feature group, a symptom node positioning result, and a traceability path structure diagram.

[0013] As a further solution of the present invention, the disease tracing analysis module includes: The response difference extraction submodule checks the response time of high-frequency symptoms and the change sequence of health indicators based on the consultation path optimization table, identifies the synchronization deviation between the response time difference and the change rate of health indicators, selects the time points and symptom numbers where the deviation exceeds the stable range, and generates an automated consultation response deviation list; The fluctuation consistency identification submodule calls the health indicator fluctuation sequence according to the automated medical inquiry response deviation list, compares the fluctuation amplitude and direction at the deviation time point, identifies the continuous consistent fluctuation segment and records the interval overlapping with the deviation time point, and generates a response fluctuation consistent segment table; The disease attribution identification submodule extracts the node path of the disease number according to the response fluctuation consistent segment table, tracks the node response sequence and signal transmission in the overlapping segment, identifies the abnormal response frequency of the signal source node, and forms the health risk traceability analysis result.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by mapping the ratio of the time delay and semantic frequency of voice information, the logical relationship of the description chain is accurately obtained, and a correlation model between health status and expression characteristics is constructed, which effectively enhances the structural perception ability of semantic information. By combining the frequency of symptom descriptions with the trend of health indicators, the dependence of abnormal fluctuations is accurately measured at the semantic level, and the refined classification of consultation priorities is achieved, thereby improving the sensitivity of key symptom identification. In terms of processing the time distribution and description persistence of high-frequency symptoms, by summarizing the trend of periodic fluctuations, a dynamic evolution sequence is formed to enhance the ability to monitor the evolution of potential health risks. By combining the response time and symptom manifestations of high-risk fluctuation segments, the consultation path is revised based on trend prediction, effectively avoiding interference from low-correlation paths and improving decision-making accuracy and timeliness. The response difference of the symptom attribution node is compared with the fluctuation of health data for consistency, further tracing the core symptom inducement and optimizing the integrity of the diagnostic chain. In the overall processing process, the ambiguity and ambiguity in the symptom description are analyzed through time series and semantic association to enhance the hierarchical identification and intervention capabilities of complex health states, so that health assessment in multiple rounds of interaction has stronger adaptability and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a system flow chart of the present invention; Figure 2 This is a flow chart of the speech structure analysis module in the present invention; Figure 3 This is a flow chart of the semantic association mapping module in the present invention; Figure 4 This is a flow chart of the knowledge graph reasoning module in the present invention; Figure 5 This is a flow chart of the consultation path adjustment module in the present invention; Figure 6 This is a flow chart of the disease tracing analysis module in the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0018] See also Figure 1 , the remote medical consultation system based on AI recognition includes: The speech structure analysis module obtains the patient's speech expression, symptom description, and health indicator data, analyzes the semantic relationship, determines the semantic direction and description chain length based on the time delay and semantic sequence between the speech expression and symptom description, and maps the ratio of semantic frequency to time delay to generate a health status description association model; The semantic association mapping module, based on the health status description association model, selects the symptom description frequency of key nodes and the change trend of health indicators, matches and compares the abnormal fluctuation frequency with the semantic dependency degree, identifies the key level of consultation, and obtains the node consultation priority scale; The knowledge graph reasoning module extracts the time interval and duration of symptom descriptions based on the node consultation priority scale, determines the distribution stage of high-frequency descriptions, classifies the cyclical trends of similar symptom descriptions, and generates a health risk evolution sequence; The consultation path adjustment module calls the health risk evolution sequence, combines the abnormal fluctuation values ​​and response time delays within the jurisdiction, screens the target diseases in the high-trend segments, and revises the consultation path values ​​in sections to obtain the consultation path optimization table; The disease traceability analysis module is based on the consultation path optimization table, collects the response differences of high-frequency consultation diseases and the fluctuation characteristics of health indicators, determines whether the fluctuation characteristics are consistent, and identifies the disease attribution nodes to form the health risk traceability analysis results.

[0019] The health status description association model includes semantic direction classification, description chain path structure, and semantic delay ratio distribution. The node consultation priority scale includes consultation level label, node risk coefficient, and priority score interval. The health risk evolution sequence includes description interval distribution, description continuous frequency, and periodic classification results. The consultation path optimization table includes target symptom list, path revision parameters, and delay correlation factors. The health risk traceability analysis results include fluctuation consistent feature group, symptom node positioning results, and traceability path structure diagram.

[0020] See also Figure 2 , the speech structure analysis module includes: The speech and symptom recognition submodule obtains the patient's speech expression, symptom description, and health indicator data, aligns the three signals by timestamp, extracts symptom description mutation points, locates changes in speech expression and health indicators, identifies semantic differences, and generates symptom description response deviation; Obtain the patient's voice description of "I have been coughing recently, feeling a little chest tightness, and waking up suddenly at night" recorded through the mobile terminal App. The symptom descriptions checked by the patient on the App include "coughing", "chest tightness", and "waking up suddenly at night", as well as the health indicator data uploaded by the patient, such as heart rate and blood oxygen saturation. For example, the recorded heart rate data is {t1: 78, t2: 80, t3: 85} bpm, and the blood oxygen saturation is {t1: 98%, t2: 97%, t3: 95%}. Align the timestamp of the voice description, the recording time of the symptom description, and the collection time of the health indicator data. If "coughing" The recording time of the symptom description is t1, "chest tightness" is t2, and "nighttime waking up" is t3. The time points when the symptom description changes from nothing to something, worsens or alleviates are extracted as mutation points. For example, if the patient starts recording cough symptoms at time t1, then t1 is the mutation starting point of the cough symptom, and the voice expression located near t1 is the voice segment describing the cough. The heart rate value located near t1 is 78bpm, and the blood oxygen saturation is 98%. If the patient records the nighttime waking up symptom at time t3 and has not recorded it before, then t3 is the mutation starting point of the nighttime waking up symptom, and the voice expression located near t3 is the voice segment describing the nighttime waking up symptom. The heart rate value near position t3 is 85bpm, and the blood oxygen saturation is 95%. Compare the semantic differences between the words "cough", "chest tightness", "wake up due to suffocation" in the patient's voice description and the symptom description checked by the patient. For example, if "shortness of breath" is mentioned in the voice but it is not checked in the symptom description, there is a semantic difference. Calculate the degree of semantic deviation between the voice description and the symptom description. For example, use word vectors to calculate the semantic distance between the two. If the distance exceeds the preset threshold of 0.2, it is considered that there is a response deviation. For each symptom description mutation point, calculate the deviation of its corresponding voice description and health indicator change relative to the baseline. For example, For example, taking the patient's usual heart rate as 75bpm as the baseline, the heart rate deviation at t1 is +3bpm, and the heart rate deviation at t3 is +10bpm. The speech expression semantic deviation and health indicator change deviation corresponding to each symptom description mutation point are integrated to generate a response deviation for each symptom description. For example, the response deviation of the symptom description of coughing at t1 is {speech semantic deviation: 0.15, heart rate deviation: +3, blood oxygen saturation deviation: -1}, and the response deviation of the symptom description of waking up at night due to suffocation at t3 is {speech semantic deviation: 0.05, heart rate deviation: +10, blood oxygen saturation deviation: -2}.

[0021] The semantic lag analysis submodule extracts the speech signal delay segment and the symptom signal response sequence based on the symptom description response deviation, calculates the delay length and response sequence, determines the signal semantic direction, measures the number of description chain trigger sequences within the lag segment, and generates the semantic signal lag structure. According to the response deviation of the symptom description, for example, the response deviation of the symptom description of cough at time t1 is {speech semantic deviation: 0.15, heart rate deviation: +3, blood oxygen saturation deviation: -1}, and the response deviation of the symptom description of waking up at night at time t3 is {speech semantic deviation: 0.05, heart rate deviation: +10, blood oxygen saturation deviation: -2}, the delay time of the speech segment related to each symptom description is extracted from the patient's continuous speech signal. For example, the patient first recorded the cough symptom and mentioned it in the speech 5 seconds later, then the speech signal delay segment is 5 seconds. Determine the order in which the patient records different symptoms, for example, cough first, then chest tightness, and then waking up at night. Count the time length of each speech signal delay segment, for example, cough delay 5 seconds, chest tightness delay 3 seconds, and waking up at night delay 1 second. Record the response order of symptom signals, that is, cough-chest tightness-waking up at night, and judge whether the semantic direction of the speech signal is consistent with the response order of the symptom signal, for example, first in the speech If a patient describes a cough and then chest tightness, and the order is consistent with the recording, then the semantic direction is consistent. The order and number of triggering events in the patient's description chain within the identified lag segment are counted. For example, within a 5-second delay after the onset of coughing, the patient first said "throat discomfort", then said "a little dry", and finally said "cough". The number of description chain triggering sequences is 3. The delay length, response order, signal semantic direction, and number of description chain triggering sequences within the lag segment of each symptom description are integrated to generate a semantic signal lag structure quantity. For example, the semantic signal lag structure quantity for cough is {delay length: 5 seconds, response order: 1, semantic direction: consistent, number of triggering sequences: 3}, the semantic signal lag structure quantity for chest tightness is {delay length: 3 seconds, response order: 2, semantic direction: consistent, number of triggering sequences: 1}, and the semantic signal lag structure quantity for waking up at night due to suffocation is {delay length: 1 second, response order: 3, semantic direction: consistent, number of triggering sequences: 0}.

[0022] The symptom description structure generation submodule extracts the linkage period between health indicator change points and description chain signals based on the semantic signal hysteresis structure, compares the signal response frequency and delay, identifies the linkage number and average delay ratio, analyzes the mapping between semantic relationship and linkage frequency, and generates a health status description association model. Based on the semantic signal lag structure quantity, for example, the semantic signal lag structure quantity of cough is {delay length: 5 seconds, response order: 1, semantic direction: consistent, number of trigger sequences: 3}, the semantic signal lag structure quantity of chest tightness is {delay length: 3 seconds, response order: 2, semantic direction: consistent, number of trigger sequences: 1}, and the semantic signal lag structure quantity of nighttime awakening is {delay length: 1 second, response order: 3, semantic direction: consistent, number of trigger sequences: 0}. The change points where the patient's health indicator data (such as heart rate and blood oxygen saturation) change significantly are extracted. For example, if the heart rate suddenly rises from 78bpm to 85bpm, then this time point is the heart rate change point. The time period when each health indicator change point is linked to the description chain signal (i.e., the symptom description recorded by the patient) is recorded. For example, if the heart rate continues to rise within 1 minute after the nighttime awakening is recorded, then this 1 minute is the linkage period. During the linkage period, the number of responses and time delay of each symptom signal, for example, cough was mentioned twice during the linkage period, with an average delay of 4 seconds. The specific number of times the health indicator changes were linked to the description chain signal during each linkage period was identified. For example, waking up at night due to shortness of breath was linked to increased heart rate once. The average delay ratio in each linkage relationship was calculated. For example, the average delay of increased heart rate relative to the delay ratio of waking up at night due to shortness of breath was 1 / 1=1. The mapping relationship between the semantic relationship described by the patient (such as causality, accompaniment, progression, etc.) and the linkage frequency of health indicators was analyzed. For example, if a certain semantic relationship is frequently accompanied by increased heart rate, then there is a high-frequency linkage mapping between this semantic relationship and increased heart rate. The semantic relationship and linkage frequency mapping obtained by analysis were integrated to form a health status description association model, which records different symptom descriptions, changes in health indicators, and the linkage relationship, frequency and delay information between them.

[0023] See also Figure 3 , the semantic association mapping module includes: The fluctuation extraction submodule extracts the frequency of symptom descriptions of key nodes based on the health status description association model, filters out sequences with fluctuations exceeding the baseline value, analyzes the proportion of abnormal fluctuations, and generates the abnormal fluctuation frequency ratio; Based on the health status description association model, the frequency of occurrence of the symptom description corresponding to each key node in the knowledge graph in the patient's consultation record is extracted. For example, the node "cough" appears 15 times, "chest tightness" appears 10 times, and "fever" appears 2 times. A benchmark frequency value is set. For example, the benchmark value is set to the average description frequency of similar patients in the past week. For each key node, the sequences whose description frequency fluctuations exceed the benchmark value are screened. For example, the benchmark frequency of "cough" is 10 times, and it actually appears 15 times, exceeding 5 times. The number of fluctuations exceeding the benchmark value is recorded, and the proportion of the number of abnormal fluctuations of each node to its total number of descriptions is calculated. For example, the total number of descriptions of "cough" is 20 times, and the abnormal fluctuation is 5 times, then the abnormal fluctuation frequency ratio is 5 / 20=0.25.

[0024] The frequency dependency matching submodule calls the abnormal fluctuation frequency ratio, combines the node description frequency and the number of dependency paths, identifies the frequency dependency comparison item, and adjusts the node abnormal offset according to the proportion of tasks on the dependency path. The formula is: ; The wave coupling priority index is calculated; in, represents the fluctuation coupling priority index, Representative The frequency of tasks, represents the average value of task frequency, Representative The proportion of tasks in the dependency path, Representative The number of dependent paths for a task, represents the variance of the node description frequency, represents the standard deviation of the node description frequency, Represents the total number of tasks; The Volatility Coupling Priority Index is an indicator that measures the volatility relationship between different assets or markets within a market. By analyzing the volatility linkage between different markets or assets, it helps investors determine the relative volatility of a particular asset or market within a specific time period. The core of this indicator is to reveal the coupling relationship between market fluctuations, focusing on the synchronization of volatility between different assets or markets. When volatility coupling is strong, it indicates that the price fluctuations of assets within the market are highly consistent, leading to risk resonance. When volatility coupling is weak, it indicates that the volatility of each asset is highly independent, providing opportunities for diversified investment. Through this indicator, investors can better understand market dynamics and make more accurate investment decisions. Call the abnormal fluctuation frequency ratio of each key node, for example, "cough" is 0.25, "chest tightness" is 0.10, and "fever" is 0.50. Combined with the actual frequency of each node in the patient's description, for example, "cough" 15 times, "chest tightness" 10 times, and "fever" 2 times, and the number of dependent paths of the node in the knowledge graph, for example, "cough" depends on 3 paths, "chest tightness" depends on 2 paths, and "fever" depends on 1 path, identify the items that need frequency dependency comparison, that is, all key nodes, and adjust the abnormal deviation degree of each node according to the proportion of tasks on its dependent path. For example, if the total number of tasks in the three dependent paths of "cough" is 10, of which "cough" itself accounts for 2 tasks, then the proportion is 2 / 10=0.2, using the formula Calculate the wave coupling priority index; Suppose there are three tasks: "cough" (i=1), "chest tightness" (i=2), and "fever" (i=3). Their frequencies are , , , average frequency , accounting for , , , the number of dependent paths , , ; Frequency variance ; Standard deviation ; but ; The calculated fluctuation coupling priority index is 9.36. By considering the degree of deviation between the node description frequency and the average frequency, the importance of the node in the dependency path, and the complexity of the node dependency path, combined with the discrete degree of the node description frequency, the abnormal fluctuation priority of the key nodes can be more accurately evaluated, thereby providing a more reliable basis for the subsequent consultation path optimization.

[0025] The priority level determination submodule extracts the numerical interval corresponding to the node based on the fluctuation coupling priority index, sets the node level to divide the interval group, assigns the level identifier according to the index value falling into the interval, sorts and determines the consultation level, and obtains the node consultation priority scale; Based on the fluctuation coupling priority index, whose value is 9.36, the preset numerical interval corresponding to each node in the knowledge graph is extracted. For example, the interval of "cough" is (8, 12), the interval of "chest tightness" is (5, 9), and the interval of "fever" is (15, 20). The node level is set to divide the interval group, for example, low priority: <6, medium priority: 6-12, high priority: >12, and the fluctuation coupling priority index value of each node is compared with the set interval. For example, the index value of cough is 9.36, which falls into the medium priority interval, the index value of chest tightness is assumed to be 5.2, which falls into the low priority interval, and the index value of fever is assumed to be 16, which falls into the high priority interval. According to the interval in which the index value falls, each node is assigned a corresponding level identifier, for example, cough: medium, chest tightness: low, fever: high, and the levels of all nodes are sorted to determine the level order of this consultation, for example, fever>cough>chest tightness. The consultation priority scale containing the nodes and their corresponding consultations is obtained, as shown in Table 1; Table 1: Node consultation priority scale As shown in Table 1, the priority order of fever, cough, and chest tightness in this consultation was determined based on the fluctuation coupling priority index. The results indicate that fever should be given priority attention in this consultation, followed by cough, and finally chest tightness. The node consultation priority scale provides a basis for subsequent consultation path planning.

[0026] See also Figure 4 , the knowledge graph reasoning module includes: The description record screening submodule collects the trigger and end time points of symptom descriptions based on the node consultation priority scale, identifies the time intervals between adjacent descriptions, filters out records that exceed the benchmark value, and generates a description time interval sequence; According to the node consultation priority scale, the triggering time and end time of each symptom description of the patient during the consultation process were collected. For example, the triggering time of the first cough description was 10:00:00 and the end time was 10:00:15, and the triggering time of the second cough description was 10:05:30 and the end time was 10:05:40. The time interval between two adjacent descriptions was identified. For example, the interval from the end of the first cough description to the triggering of the second cough description was 5 minutes and 15 seconds. Records with time intervals exceeding the set benchmark value (for example, 5 minutes) were screened out, and the screened description time interval records were combined into a description time interval sequence, for example, {315 seconds, 480 seconds, 60 seconds}.

[0027] The description period division submodule calls the description time interval sequence to count the number of consecutive descriptions of the same type of description events. According to the description duration, interval time and health indicator changes, the formula is used: ; Calculate the trend value of the description period distribution and establish the description period distribution stage; in, Represents the trend value describing the period distribution, represents the change in the k-th time interval describing the event, Represents the number of consecutive descriptions of the k-th event type description, Represents the change in health indicators corresponding to the k-th description event, represents the duration of the k-th described event, represents the interval time of the kth described event, represents the number of total described events; The periodic distribution trend value is an indicator used to describe the distribution and change trends of a phenomenon or data within a specific time period. By analyzing the distribution characteristics of data across different periods, it reveals the regularity and changing trends of a phenomenon in the time series. The periodic distribution trend value can help understand the fluctuation pattern and periodic characteristics of data or phenomena and predict future changes. Specifically, the focus is on the fluctuation range of data within the period, the time of peak occurrence, and the relative relationship between periods. Using the periodic distribution trend value, analysts can identify potential periodic fluctuations and trend change points, providing a basis for decision-making. Call the description time interval sequence, for example, {315 seconds, 480 seconds, 60 seconds}, and count the number of consecutive descriptions of the same type of description event (such as coughing). For example, if the patient coughs 2 times in a row, the duration of each description is calculated based on the duration of each description ( ), interval time ( ) and the corresponding changes in health indicators ( ), using the formula Calculate the trend value of the description period distribution and establish the description period distribution stage; Assume there are three described events, and the detailed data are shown in Table 2; Table 2: Detailed data table describing the event but ; The calculated trend value of the descriptive period distribution is 1.49. If the preset trend value interval is: stable (T<0.8), rising (0.8≤T≤1.5), and significantly rising (T>1.5), then the descriptive period distribution is in an upward stage. The descriptive period distribution stage is established as "rising", which comprehensively considers the changes in the time intervals, continuity, changes in health indicators, and the duration of descriptions and intervals of the described events. It can more comprehensively evaluate the distribution trend of the descriptive periodicity, thereby more accurately judging changes in patients' health risks.

[0028] The periodic trend classification submodule calls the description period distribution stage, identifies the description events of the same periodic trend, identifies the trend fluctuation differences, classifies and archives them according to the difference threshold, and generates the health risk evolution sequence; Call the description cycle distribution stage, for example, "rising", identify all description events belonging to the same cycle trend (such as all rising trends), identify the differences in trend fluctuations of similar description events, for example, cough and wheezing, which are both rising trends, have a faster increase in cough frequency, and classify and archive the description events based on the preset difference threshold (for example, the difference in frequency change rate exceeds 10%). For example, a frequency change rate of 10%-20% is classified as a mild increase, 20%-30% is classified as a moderate increase, and >30% is classified as a severe increase, generating a sequence reflecting the evolution of the patient's health risk over time, for example, {"cough":"mild increase","chest tightness":"stable","wheezing":"moderate increase"}.

[0029] See also Figure 5 , the consultation path adjustment module includes: The trend sequence identification submodule calls the health risk evolution sequence, extracts the incremental change value, determines the rising trend interval based on the cumulative increase and fluctuation amplitude, marks key diseases, and generates an automated consultation trend anomaly identification list; Call the health risk evolution sequence, for example, {"cough":"mild increase","chest tightness":"stable","wheezing":"moderate increase"}, extract the incremental change value of each symptom, for example, the frequency of "cough" increased by 5% compared with the last time, and the frequency of "wheezing" increased by 15%. Based on the set cumulative increase threshold (for example, cumulative increase >20%) and fluctuation amplitude threshold (for example, single fluctuation >10%), determine which symptoms are in the trend increase range and mark the symptoms as key symptoms. For example, if the cumulative increase of "wheezing" exceeds 20%, it is marked as a key symptom. Generate an automated medical consultation trend anomaly recognition list containing the marking information, for example, {"cough":"mild increase","chest tightness":"stable","wheezing":"moderate increase","key symptoms": ["wheezing"]}.

[0030] The response feature extraction submodule calls the disease number in the automated medical consultation trend anomaly identification list using the formula: ; Calculate the abnormal fluctuation value within the segment, identify the response delay, combine the delay distribution and fluctuation frequency, analyze the difference in symptom response, and obtain the difference value of the automated medical consultation response characteristics; in, Represents the abnormal fluctuation value within the segment, Represents the total number of data points in the segment, Representative The deviation value of the data point, represents the mean of the data points, Representative The delay value of the data point, Represents the average value of the delay value, represents the standard deviation of the delay values; Abnormal intra-segment fluctuations refer to abnormal fluctuations in data or market prices within a specific time period. These fluctuations are either too large or too small compared to normal fluctuations within that period. Abnormal fluctuations are caused by unexpected events, dramatic changes in market sentiment, or external factors. They do not conform to normal fluctuation patterns and indicate market instability or potential risks. By identifying and analyzing abnormal intra-segment fluctuations, investors and analysts can promptly detect anomalies in the market and take appropriate measures to prevent potential risks or seize investment opportunities. Call the symptom number in the list of abnormal trend identification of automated medical consultation, for example, extract the number of "wheezing", and use the formula Calculate the abnormal fluctuation value within the segment, identify the response delay, combine the delay distribution and fluctuation frequency, analyze the difference in symptom response, and obtain the difference value of the automated medical consultation response characteristics; Assume that for the "wheezing" symptom, within a time period containing 5 data points, the deviation value is The sequence is {2, -1, 3, -2, 1}, and the average deviation value , delay value The sequence is {1, 0, 2, 1, 0} seconds, and the average delay value is Second; Standard deviation of delay values: ; but ; The difference value of the automated medical consultation response characteristics was obtained to be 6.91. The benefit of the formula is that it not only takes into account the degree of deviation of the data points, but also introduces the delay value and its degree of deviation from the average value, and normalizes it by the standard deviation. It can more comprehensively evaluate the abnormal fluctuations within the segment and the impact of response delays, thereby more accurately identifying the characteristic differences in disease responses.

[0031] The path revision submodule identifies the path revision item structure based on the difference value of the automated medical consultation response characteristics and the path corresponding to the trend segment, analyzes the path deviation degree and frequency distribution amplitude, calculates the path revision deviation degree, and adjusts the medical consultation frequency and time distribution based on the difference in the original path structure to obtain the medical consultation path optimization table; Based on the difference value of the response characteristics of the automated consultation, for example, "wheezing" is 6.91, and the trend segment corresponding to the disease in the health risk evolution sequence (for example, an upward segment), the path revision item structure related to the trend segment in the knowledge graph is identified, the degree of deviation of the disease path (for example, the difference between the actual consultation path and the standard path) and the distribution amplitude of the deviation frequency are analyzed, and the path revision deviation degree is calculated. For example, the edit distance is used to measure the degree of path deviation, and the frequency distribution amplitude is calculated by counting the number of deviations that occur within a certain period of time. The difference value of the response characteristics of the automated consultation, the path revision deviation degree and the difference in the original path structure (for example, the length and number of nodes of the original path) are combined to adjust the frequency and time distribution of consultations related to the disease in subsequent consultations. For example, for "wheezing" with a high response characteristic difference value and an upward trend, the consultation frequency is increased and the consultation interval is shortened to obtain an updated consultation path optimization table, which contains the optimized consultation frequency and time point recommendations for each disease.

[0032] See also Figure 6 , the disease traceability analysis module includes: The response difference extraction submodule checks the response time of high-frequency symptoms and the change sequence of health indicators based on the consultation path optimization table, identifies the synchronization deviation between the response time difference and the change rate of health indicators, selects the time points and symptom numbers where the deviation exceeds the stable range, and generates an automated consultation response deviation list; Based on the consultation path optimization table, the patient response time and the change sequence of the patient's health indicators for high-frequency symptoms (for example, symptoms with a consultation frequency higher than the average) are examined, and the synchronization deviation between the response time difference (the patient's feedback delay on the symptom) and the health indicator change rate are identified. For example, if the patient's feedback delay on "wheezing" exceeds 3 seconds and the blood oxygen saturation decrease rate accelerates at this time, there is a synchronization deviation. The time points and corresponding symptom numbers where the synchronization deviation exceeds the preset stable range (for example, the response time difference >3 seconds and the blood oxygen saturation decrease rate <-0.5% / minute) are screened to generate an automated consultation response deviation list, which records the time points and symptoms where significant response deviations occur, for example, {time point: 10:15:30, symptom: "wheezing"}.

[0033] The fluctuation consistency discrimination submodule calls the health indicator fluctuation sequence based on the automated questionnaire response offset list, compares the fluctuation amplitude and direction at the offset time point, identifies the continuous consistent fluctuation segments and records the intervals overlapping with the offset time point, and generates a response fluctuation consistent segment table; According to the automated medical inquiry response offset list, for example, {time point: 10:15:30, symptom: "wheezing"}, call the patient's health indicator fluctuation sequence. For example, around 10:15:30, the blood oxygen saturation dropped rapidly from 96% to 93%. Compare the health indicator fluctuation amplitude and direction at the offset time point. For example, the blood oxygen saturation dropped by 3%, and the direction was negative. Identify the segments of continuous and consistent fluctuations, and record the intervals overlapping with the offset time points. For example, the blood oxygen saturation continued to drop from 10:15:00 to 10:16:00. Generate a response fluctuation consistent segment table, which records the time intervals that overlap with the response offset time points and have consistent health indicator fluctuation trends, for example, {symptom: "wheezing", time interval: [10:15:00-10:16:00], fluctuation type: "decreased blood oxygen saturation"}.

[0034] The disease attribution identification submodule extracts the node path of the disease number based on the response fluctuation consistent segment table, tracks the node response sequence and signal transmission in the overlapping segments, identifies the abnormal response frequency of the signal source node, and forms the health risk traceability analysis results; According to the response fluctuation consistent segment table, for example, {symptom: "wheezing", time interval: [10:15:00-10:16:00], fluctuation type: "decreased blood oxygen saturation"}, extract the node path of the offset symptom number ("wheezing") in the knowledge graph, trace the response sequence and signal transmission process of the relevant nodes in the knowledge graph within the overlapping segment, identify the abnormal response frequency of the node as the signal source (for example, the upper respiratory tract infection node that may cause wheezing and decreased blood oxygen) (for example, the frequency of abnormal health indicators related to the node), integrate the analysis results, and form a traceability analysis result for the patient's health risk, for example, "the patient's current wheezing and decreased blood oxygen saturation are related to upper respiratory tract infection, and it is recommended to further inquire about related symptoms."

[0035] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. The remote medical consultation system based on AI recognition is characterized by: The system comprises: The speech structure analysis module obtains the patient's speech expression, symptom description, and health indicator data, analyzes the semantic relationship, determines the semantic direction and description chain length based on the time delay and semantic sequence between the speech expression and symptom description, and maps the ratio of semantic frequency to time delay to generate a health status description association model; The semantic association mapping module selects the symptom description frequency and health indicator change trend of key nodes based on the health status description association model, matches and compares the abnormal fluctuation frequency with the semantic dependency degree, identifies the key level of consultation, and obtains the node consultation priority scale; The knowledge graph reasoning module extracts the time interval and duration of symptom descriptions based on the node consultation priority scale, determines the distribution stage of high-frequency descriptions, classifies the periodic trends of similar symptom descriptions, and generates a health risk evolution sequence; The consultation path adjustment module calls the health risk evolution sequence, combines the abnormal fluctuation value and response time delay in the section under its jurisdiction, screens the target diseases in the high trend section, revises the consultation path value in sections, and obtains the consultation path optimization table.

2. The AI-based remote medical consultation system according to claim 1 is characterized in that: The health status description association model includes semantic direction classification, description chain path structure, and semantic delay ratio distribution; the node consultation priority scale includes consultation level label, node risk coefficient, and priority score interval; the health risk evolution sequence includes description interval distribution, description continuous frequency, and periodic classification results; the consultation path optimization table includes target disease list, path revision parameters, and delay association factors.

3. The remote medical consultation system based on AI recognition according to claim 1 is characterized in that: The speech structure analysis module includes: The speech and symptom recognition submodule obtains the patient's speech expression, symptom description, and health indicator data, aligns the three signals by timestamp, extracts symptom description mutation points, locates changes in speech expression and health indicators, identifies semantic differences, and generates symptom description response deviation; The semantic lag analysis submodule extracts the speech signal delay segment and the symptom signal response sequence based on the symptom description response deviation, calculates the delay length and response sequence, determines the signal semantic direction, measures the number of description chain trigger sequences within the lag segment, and generates the semantic signal lag structure quantity; The symptom description structure generation submodule extracts the linkage period between the health indicator change points and the description chain signals based on the semantic signal hysteresis structure, compares the signal response frequency and delay, identifies the linkage times and the average delay ratio, analyzes the semantic relationship and linkage frequency mapping, and generates a health status description association model.

4. The remote medical consultation system based on AI recognition according to claim 3 is characterized in that: The semantic association mapping module includes: The fluctuation extraction submodule extracts the frequency of key node symptom descriptions based on the health status description association model, filters out sequences with fluctuations exceeding the benchmark value, analyzes the proportion of abnormal fluctuations, and generates an abnormal fluctuation frequency ratio; The frequency dependency matching submodule calls the abnormal fluctuation frequency ratio, combines the node description frequency and the number of dependency paths, identifies the frequency dependency comparison item, adjusts the node abnormal offset according to the proportion of tasks on the dependency path, and calculates the fluctuation coupling priority index; The priority determination submodule extracts the numerical interval corresponding to the node based on the fluctuation coupling priority index, sets the node level division interval group, assigns the level identifier according to the index value falling into the interval, sorts and determines the consultation level, and obtains the node consultation priority scale.

5. The AI-based remote medical consultation system according to claim 4 is characterized in that: The knowledge graph reasoning module includes: The description record screening submodule collects the time points of symptom description triggering and ending according to the node consultation priority scale, identifies the time intervals between adjacent descriptions, filters out records exceeding the benchmark value, and generates a description time interval sequence; The description period division submodule calls the description time interval sequence, counts the number of consecutive descriptions of the same type of description events, calculates the description period distribution trend value based on the description duration, interval time and health indicator changes, and establishes the description period distribution stage; The periodic trend classification submodule calls the periodic distribution description stage, identifies description events of the same periodic trend, identifies trend fluctuation differences, classifies and archives them according to difference thresholds, and generates a health risk evolution sequence.

6. The AI-based remote medical consultation system according to claim 5, characterized in that: The consultation path adjustment module includes: The trend sequence identification submodule calls the health risk evolution sequence, extracts the incremental change value, determines the trend rising interval based on the cumulative increase and fluctuation amplitude, marks the key symptoms, and generates an automated consultation trend anomaly identification list; The response characteristic extraction submodule calls the disease number in the automated medical consultation trend abnormality identification list, calculates the abnormal fluctuation value within the segment, identifies the response delay, combines the delay distribution and the fluctuation frequency, analyzes the disease response difference, and obtains the automated medical consultation response characteristic difference value; The path revision submodule identifies the path revision item structure based on the difference value of the automated medical consultation response characteristics and the path corresponding to the trend segment, analyzes the path deviation degree and frequency distribution amplitude, calculates the path revision deviation degree, and adjusts the medical consultation frequency and time distribution in combination with the original path structure difference to obtain the medical consultation path optimization table.

7. The remote medical consultation system based on AI recognition according to claim 1 is characterized in that: The system also includes a disease tracing and analysis module: The symptom traceability analysis module collects the response differences of high-frequency symptom consultations and the fluctuation characteristics of health indicators based on the consultation path optimization table, determines whether the fluctuation characteristics are consistent, and identifies the symptom attribution nodes to form health risk traceability analysis results; The health risk traceability analysis results include a fluctuation consistent feature group, a symptom node positioning result, and a traceability path structure diagram.

8. The AI-based remote medical consultation system according to claim 7, characterized in that: The disease tracing analysis module includes: The response difference extraction submodule checks the response time of high-frequency symptoms and the change sequence of health indicators based on the consultation path optimization table, identifies the synchronization deviation between the response time difference and the change rate of health indicators, selects the time points and symptom numbers where the deviation exceeds the stable range, and generates an automated consultation response deviation list; The fluctuation consistency identification submodule calls the health indicator fluctuation sequence according to the automated medical inquiry response deviation list, compares the fluctuation amplitude and direction at the deviation time point, identifies the continuous consistent fluctuation segment and records the interval overlapping with the deviation time point, and generates a response fluctuation consistent segment table; The disease attribution identification submodule extracts the node path of the disease number according to the response fluctuation consistent segment table, tracks the node response sequence and signal transmission in the overlapping segment, identifies the abnormal response frequency of the signal source node, and forms the health risk traceability analysis result.

Images