A triage method and system based on multimodal triboelectric analysis

Abstract

The present invention discloses a triage method and system based on multimodal triboelectric analysis, which belongs to the field of medical triage technology, and specifically includes: providing an interactive model with a surface covered with triboelectric material, the interactive model dividing a number of anatomical mapping areas; detecting user touch events and parsing the specific anatomical mapping area positions; selecting a corresponding target sensor combination and feature analysis mode according to the type of touched area; activating corresponding sensors according to the target sensor combination to collect the user's biophysical signal data stream; executing the calculation process specified by the feature analysis mode to extract a multidimensional feature parameter set; inputting the multidimensional feature parameter set into a clinically verified symptom classification decision structure, generating a triage priority code, and transmitting it to a medical system terminal through an encrypted communication protocol; the present invention uses triboelectric technology to locate the touch position, combines biophysical signal acquisition and multimodal analysis, and quickly generates a preliminary assessment result of the condition.

Description

Technical Field

[0001] The present invention relates to the technical field of medical triage, and in particular to a triage method and system based on multimodal triboelectric analysis. Background Art

[0002] With global medical resources becoming increasingly scarce and the aging population intensifying, efficient and accurate triage systems are crucial for improving the quality of healthcare. According to the World Health Organization, the average wait time for emergency rooms worldwide reached 2.7 hours in 2022, with 35% of patients experiencing worsening conditions due to triage delays. Traditional manual triage relies on the experience of medical staff, is subject to high subjectivity, and is time-consuming. This is particularly true during public health emergencies, where it can be inefficient and pose a high risk of cross-infection. Therefore, the development of automated, contactless, intelligent triage technology has become a research hotspot in the field of medical information technology.

[0003] In recent years, significant progress has been made in medical-assisted diagnostic systems based on sensor technology and artificial intelligence. Wearable devices and smart interactive terminals have shown great potential in collecting physiological signals. For example, some systems use pressure sensors to detect pressure feedback from painful areas and combine them with temperature sensors to monitor inflammatory responses. Other research is using triboelectric power generation to achieve self-powered biosignal acquisition. These technologies offer new approaches to non-invasive medical testing.

[0004] Existing systems typically use fixed sensor combinations and are unable to dynamically adjust acquisition strategies based on the location of a patient's symptoms, resulting in both data redundancy and the loss of critical information. Furthermore, most solutions only analyze a single type of signal independently, lacking the ability to integrate and process multimodal data. Summary of the Invention

[0005] The purpose of the present invention is to provide a triage method and system based on multimodal triboelectric analysis to solve the following technical problems:

[0006] Existing systems typically use fixed sensor combinations and are unable to dynamically adjust acquisition strategies based on the location of a patient's symptoms, resulting in both data redundancy and the loss of critical information. Furthermore, most solutions only analyze a single type of signal independently, lacking the ability to integrate and process multimodal data.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] A triage method based on multimodal triboelectric analysis comprises the following steps:

[0009] Provides an interactive model of a surface covered with triboelectric materials. The interactive model is divided into several anatomical mapping areas, each of which is equipped with multiple types of sensor access ports.

[0010] Detecting user touch events through triboelectric materials and analyzing the specific anatomical mapping area where the touch event occurred;

[0011] Selecting a corresponding signal acquisition strategy according to the type of the touched anatomical mapping area, the signal acquisition strategy including a target sensor combination identifier and a feature analysis mode identifier;

[0012] Activate the corresponding sensor according to the target sensor combination identifier to collect the user's biophysical signal data stream; execute the calculation process specified by the feature analysis mode identifier to extract a multi-dimensional feature parameter set from the biophysical signal data stream;

[0013] The multidimensional feature parameter set is input into the clinically verified symptom classification decision structure to generate a triage priority code and transmit it to the medical system terminal through an encrypted communication protocol.

[0014] As a further solution of the present invention, the process of parsing the specific anatomical mapping area location where the touch event occurs is:

[0015] A high-density capacitive sensing network is deployed beneath the triboelectric material layer. The network nodes form a coordinate system that maps the boundaries of the anatomically mapped region. The network captures the distribution of capacitance value mutations caused by touch events and calculates the spatial Euclidean distance between the peak coordinates of the mutation and the nearest anatomically mapped region boundary.

[0016] When the spatial Euclidean distance is less than a preset tolerance threshold, the touch event is determined to belong to the anatomically mapped area; the entry in the signal acquisition strategy library is indexed according to the type of the anatomically mapped area, and the entry records the target sensor combination activation sequence and feature analysis mode call sequence corresponding to the area.

[0017] As a further solution of the present invention: the signal acquisition strategy library includes:

[0018] Collect typical pathological characterization data from different anatomical mapping areas and establish a mapping relationship table between regional pathological characteristics and biophysical signal types. Based on the mapping relationship table, generate target sensor combination configuration rules: configure a dynamic pressure signal acquisition combination for pain-sensitive areas and a temperature distribution signal acquisition combination for areas with high inflammation incidence.

[0019] Activation conditions are set for each target sensor combination: the first touch triggers the basic signal acquisition combination, and the continuous touch timeout triggers the enhanced signal acquisition combination; the basic signal acquisition combination includes core biophysical sensors, and the enhanced signal acquisition combination adds auxiliary cross-validation sensors.

[0020] As a further solution of the present invention: the determination logic of the continuous touch is:

[0021] Recording a sequence of continuous touch events on the same anatomically mapped area over a preset time period; verifying that the time intervals between touch events are within the normal operating interval range for the human body;

[0022] Calculate the spatial position distribution density of continuous touch points, and confirm the validity of the operation when the density value exceeds the set ratio of the area; start the enhanced signal acquisition combination activation instruction for the effective continuous touch action, and extend the feature analysis time window.

[0023] As a further solution of the present invention: the calculation process of the feature analysis mode includes:

[0024] Perform time series decomposition on dynamic pressure signals to extract the slope change curve of the pressure rise, the peak maintenance time, and the oscillation decay period parameters; perform spatial gradient calculation on temperature distribution signals to generate thermal diffusion rate values ​​and temperature field symmetry indexes; perform time-frequency domain correlation analysis on multi-source mixed signals to calculate the phase synchronization coefficient matrix and energy coupling strength values ​​across signal channels;

[0025] All characteristic parameters are normalized by the regional model to eliminate the individual response differences of sensors and generate a comparable multidimensional characteristic parameter set.

[0026] As a further solution of the present invention: the symptom classification decision structure includes:

[0027] An independent clinical validation database was constructed to store the multidimensional feature parameter sets and final triage conclusions of medically confirmed cases. A supervised learning algorithm was used to train a tree-like decision structure, with the root node corresponding to the anatomical mapping region type and the branch nodes associated with the confidence intervals of the multidimensional feature parameters. Leaf nodes were bound to triage priority codes, and the code generation rules followed the urgency grading principles of the International Classification of Diseases.

[0028] The decision structure is updated through a sliding time window mechanism: the main branches of regional types are retained, the confidence interval boundaries of feature parameters are optimized, and low-frequency leaf nodes are merged.

[0029] As a further solution of the present invention: the generation process of the triage priority code is:

[0030] A feature analysis channel is selected based on the activated target sensor combination type, and each channel outputs a feature confidence score value. The feature confidence score value is input into the symptom classification decision structure and traverses along the branch node path to the leaf node. When a leaf node maps multiple triage priority codes, a conflict arbitration protocol is executed: a medical urgency weight comparison table is called.

[0031] The triage priority code with the highest weight value is selected as the output result; before output, a pathology logic compatibility check is performed to verify whether there is any contradiction between the multidimensional feature parameter set and the known pathology representation of the anatomical mapping area.

[0032] As a further solution of the present invention: the specific process of the pathological logic compatibility check is:

[0033] Construct a knowledge base of pathological features of the anatomical mapping area. The knowledge base stores typical pathological representations verified by medical literature and their corresponding feature parameter constraint ranges. Split the currently extracted multidimensional feature parameter set into independent feature vectors according to parameter categories. Each feature vector contains data in the time evolution dimension, spatial distribution dimension, and multimodal association dimension. Perform a knowledge base comparison operation on each independent feature vector:

[0034] Step 1: Locate the pathological feature entry set of the anatomical mapping region to which the feature vector belongs;

[0035] Step 2: Calculate the deviation between the data of each dimension of the feature vector and the pathological representation constraint range in the entry set;

[0036] Step 3: When the deviation of the time evolution dimension exceeds the first threshold, it is marked as a temporal logic conflict;

[0037] Step 4: When the spatial distribution dimension deviation exceeds the second threshold, it is marked as a spatial distribution conflict;

[0038] Step 5: When the multimodal correlation dimension deviation exceeds the third threshold, it is marked as a cross-validation conflict;

[0039] The conflict marking results of all independent feature vectors are counted to generate a conflict type distribution map; a differentiated processing process is initiated based on the conflict type distribution map: confidence reduction processing is performed on the feature vectors of single-dimensional conflict to reduce their weight coefficient in the symptom classification decision structure; re-collection instructions are triggered for feature vectors of multi-dimensional composite conflict, the collection time of the target sensor combination is extended and auxiliary verification sensors are added; the decision priority of non-conflict feature vectors is increased, and they are given priority in participating in the triage priority code generation operation.

[0040] As a further solution of the present invention, the triage result verification process is also included:

[0041] Receive actual confirmed data fed back by the medical system, parse the disease classification code and emergency classification label therein, and generate a reference feature parameter set; calculate the deviation dimension by dimension between the reference feature parameter set and the multidimensional feature parameter set, and screen the feature dimensions whose deviation value exceeds the tolerance threshold;

[0042] For the extracted paths in the feature analysis mode of the exceeded dimensions, verify whether the signal processing parameters are consistent with the biophysical characteristics of the current anatomical mapping area; check whether the acquisition time of the target sensor combination meets the minimum requirements for data integrity; and evaluate the rationality of the weight distribution of the multidimensional feature parameters in the symptom classification decision structure;

[0043] Generate optimization instructions based on the verification results, adjust the feature analysis calculation coefficients, extend the signal acquisition time window, or reconstruct the confidence interval of the decision structure; use historical case data sets to perform simulation backtesting, and compare the change in the difference between the triage priority code and the actual confirmed data before and after optimization; when the difference reduction ratio reaches the set benchmark, the optimization parameters are synchronously updated to the signal acquisition strategy library.

[0044] The present invention also includes a triage system based on multimodal triboelectric analysis, which is used to implement the above-mentioned triage method based on multimodal triboelectric analysis, comprising:

[0045] An interactive model, wherein the surface of the interactive model is covered with a triboelectric material, and the interactive model is divided into a number of anatomical mapping areas, each area being configured with multiple types of sensor access ports;

[0046] A region detection module, which detects user touch events through triboelectric materials and analyzes the specific anatomical mapping region where the touch event occurred;

[0047] a strategy selection module, wherein the strategy selection module selects a corresponding signal acquisition strategy according to the type of the touched anatomical mapping area, wherein the signal acquisition strategy includes a target sensor combination identifier and a feature analysis mode identifier;

[0048] A parameter acquisition module, which activates corresponding sensors according to the target sensor combination identifier to collect the user's biophysical signal data stream; executes the calculation process specified by the feature analysis mode identifier to extract a multidimensional feature parameter set from the biophysical signal data stream;

[0049] A triage generation module inputs a multi-dimensional feature parameter set into a clinically verified symptom classification decision structure, generates a triage priority code, and transmits it to a medical system terminal through an encrypted communication protocol.

[0050] Beneficial effects of the present invention:

[0051] The present invention effectively solves the problems of static sensor configuration, insufficient multimodal fusion capability and poor clinical adaptability in the prior art by constructing a multimodal triboelectric analysis system based on anatomical mapping areas. The present invention uses triboelectric materials and high-density capacitive sensing networks to achieve precise positioning of touch events and dynamic matching of anatomical areas, and automatically activates the corresponding target sensor combination according to the type of touched area, solving the problems of data redundancy and key information loss caused by traditional fixed sensor combinations; through dynamic pressure signal time series decomposition, temperature distribution signal spatial gradient calculation and multi-source signal time-frequency domain correlation analysis, a multidimensional feature set including parameters such as pressure slope change, thermal diffusion rate, phase synchronization coefficient, etc. is constructed, combined with a clinically verified tree-type decision structure to achieve multimodal Deep fusion analysis of dynamic biophysical signals; establish a signal acquisition strategy library and pathological logic compatibility check mechanism, dynamically switch the basic and enhanced signal acquisition combinations according to continuous touch actions, verify the temporal, spatial and multimodal correlation dimensions of characteristic parameters through the anatomical regional pathological feature knowledge base, effectively deal with complex symptoms such as referred pain and discomfort in multiple parts, and reduce the cross-regional misdiagnosis rate; use self-powered friction power generation technology to power the sensor, combine encrypted communication protocol with sliding time window optimization algorithm, while realizing non-contact and low-power detection, dynamically update the decision structure confidence interval through clinical diagnosis data, so that the reliability of the system in real medical scenarios is improved; the present invention significantly improves the defects of traditional manual triage, which is highly subjective, inefficient and insufficiently adaptable to existing intelligent system scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The present invention will be further described below with reference to the accompanying drawings.

[0053] Figure 1 It is a flow chart of a triage method based on multimodal triboelectric analysis of the present invention. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0055] See also Figure 1 As shown, the present invention is a triage method based on multimodal triboelectric analysis, comprising the following steps:

[0056] 1. Interaction Model Construction and Touch Positioning

[0057] An interactive model covered with triboelectric material is provided. Its surface is divided into multiple mapping areas based on human anatomy, such as the chest, abdomen, and joints. When a user touches the model, the triboelectric material generates an electric charge signal. The system then analyzes the touch event using a built-in high-precision positioning network, accurately identifying the specific area touched and the type of action, such as pressing, sliding, or tapping. This allows the user to determine the body part of interest, providing a positioning basis for subsequent signal acquisition.

[0058] 2. Dynamic Signal Acquisition Strategy

[0059] Depending on the anatomical area touched, the system automatically activates the corresponding combination of hand sensors:

[0060] Touch the chest area: The pressure sensor is activated to detect changes in hand pressure, the temperature sensor collects palm skin temperature, and the microphone captures breathing rate or abnormal sounds to comprehensively determine whether there are symptoms such as chest tightness and fever;

[0061] Abdominal area touch: The pressure sensor records the peak pressure, the vibration sensor analyzes the hand tremor frequency, and the skin conductance sensor is added during continuous touch to evaluate the nerve stress response, helping to identify abdominal pain or inflammation-related problems;

[0062] Touch the joint area: Use the pressure sensor to sense the resistance of joint pressing, and the vibration sensor to capture the vibration signal generated by sliding friction. Combined with the abnormal joint noise recorded by the microphone, it can preliminarily judge abnormal joint movement or inflammation risk.

[0063] 3. Multimodal Signal Feature Extraction

[0064] Deeply process multi-source signals such as pressure, temperature, vibration, and voice collected from the hands:

[0065] Analyze the correlation between pressure intensity and tremor frequency to identify physiological reactions such as pain or muscle tension; integrate palm temperature and skin conductance data to establish an assessment model for fever or infection risk; associate voice descriptions with hand movement timing, such as identifying the correspondence between keywords such as "tingling" and specific touch movements, to improve the accuracy of symptom expression.

[0066] Individual differences are eliminated through standardization processing, and a set of characteristic parameters containing multi-dimensional information is generated to provide comprehensive data support for disease analysis.

[0067] IV. Clinical Decision-Making and Triage Output

[0068] A pre-trained intelligent decision-making framework integrates extensive clinical case data, starting with the touch area and combining extracted multi-dimensional feature parameters. It then filters through pre-set decision branches, ultimately generating a triage priority code that conforms to the International Classification of Diseases. This code is transmitted to the medical system terminal in real time via an encrypted protocol, enabling the dispatch of emergency resources. For example, high-risk symptoms trigger an emergency treatment process, while low-risk symptoms are assigned to a regular waiting list.

[0069] In a preferred embodiment of the present invention, the process of parsing the specific anatomical mapping area location where the touch event occurs is as follows:

[0070] Beneath the triboelectric material layer, a high-density capacitive sensing network consisting of a micron-scale electrode array is deployed. The network nodes are evenly spaced, forming a Cartesian coordinate system that maps the boundaries of human anatomical regions, such as the precordial area and McBurney's point in the right lower abdomen. When a user touches the interactive model, the triboelectric material contacts the skin, generating charge transfer and causing a sudden change in capacitance in the corresponding area. The system captures this capacitance change data across the entire area in real time through the sensor array, generating a dynamic distribution map of the sudden changes.

[0071] The positioning algorithm achieves accurate matching through the following steps:

[0072] First, the maximum amplitude point in the mutation distribution map is identified as the core action point of the touch event, and its coordinate value in the coordinate system is calculated; then the minimum spatial Euclidean distance between this coordinate point and the boundary of the preset anatomical region is extracted, and the preset tolerance threshold is set to conform to the average error range of the human finger touch pressure area; when the distance value is less than the threshold, the touch event is determined to belong to this area. For example, a touch event on the upper left chest triggers the precordial area label, and the heart-related signal acquisition strategy is synchronously activated.

[0073] In a preferred embodiment of the present invention, the signal acquisition strategy library includes:

[0074] The signal acquisition strategy library is built based on a large amount of clinical case data, achieving intelligent matching through a bidirectional mapping between pathological features and signal types. First, typical pathological characterization data from different anatomical regions is collected to establish a mapping relationship table between regional pathological features and biophysical signal types. For example, a dynamic pressure signal acquisition combination is configured for pain-sensitive areas such as the knee joint and lumbar spine, while a temperature distribution signal acquisition combination is configured for areas with high inflammation risk, such as the tonsils and appendix projection area.

[0075] Activation conditions are set for each target sensor combination: the first touch triggers the basic signal acquisition combination, which includes core biophysical sensors. For example, a touch event in the right lower abdomen activates a pressure sensor to detect the pressing depth and a temperature sensor to monitor local skin temperature. After a continuous touch timeout, an enhanced signal acquisition combination is triggered, and auxiliary cross-validation sensors are added. For example, if the right lower abdomen is continuously touched for more than a certain period of time, a vibration sensor is added to detect bowel sound conduction and a bioimpedance sensor is added to assess the risk of ascites.

[0076] In another preferred embodiment of the present invention, the determination logic of the continuous touch is:

[0077] The system ensures the accuracy of complex symptom collection through a triple-verification mechanism. First, it records a sequence of continuous touch events in the same anatomical area over a preset time period, verifying that the time intervals between touch events are within the normal range of human operation and eliminating interference from accidental touches. It then calculates the spatial distribution density of continuous touch points. When the density exceeds a set ratio of the area, the operation is confirmed to be valid. For example, if a palm continuously massages a certain area, resulting in a dense distribution of touch points, this is considered active symptom feedback. Finally, it initiates enhanced signal acquisition combined activation instructions for valid continuous touch actions, while also extending the feature analysis time window to obtain more comprehensive biophysical signals.

[0078] In another preferred embodiment of the present invention, the calculation process of the feature analysis mode includes:

[0079] 1. Analysis of the time series characteristics of dynamic pressure signals. Using advanced time series decomposition algorithms to process pressure signals, we extract three key parameters:

[0080] The pressure rising slope change curve reflects the rate of change of the pressing force per unit time and can be used to determine the urgency of the touch and pressure action;

[0081] Peak maintenance time refers to the duration of pressure after reaching its peak, which can indicate whether the muscle is in a state of continuous tension or spasm;

[0082] The oscillation attenuation period parameter helps identify abnormal conditions of joints or deep tissues by analyzing the frequency and attenuation rate of pressure signal oscillations.

[0083] 2. Model the spatial characteristics of the temperature distribution signal and perform spatial gradient calculation on the temperature distribution signal to generate two important indicators:

[0084] The thermal diffusion rate value is used to measure the expansion speed of the temperature abnormality area per unit time. An increase in this value usually indicates an intensified inflammatory response;

[0085] The temperature field symmetry index can effectively mark local pathological characteristics, such as unilateral joint inflammation or tissue damage, by comparing the temperature differences between left and right symmetrical areas of the human body.

[0086] 3. Time-frequency domain correlation analysis of multi-source mixed signals. For multi-source signals such as pressure, temperature, and bioelectricity, a time-frequency domain joint analysis method is used to calculate two core parameters:

[0087] The phase synchronization coefficient matrix is ​​used to describe the consistency of phase differences across signal channels. A higher coefficient indicates a stronger pathological correlation between signals.

[0088] The energy coupling intensity value can identify the characteristics of complex diseases with multi-system involvement by analyzing the interaction intensity of signal energy in the time and frequency domain.

[0089] 4. Standardization of feature parameters. All extracted feature parameters must be calibrated using a regional standardization model. This model is built based on a large amount of individual sensor response data and can effectively eliminate performance differences between different devices. Ultimately, it generates a unified format multi-dimensional feature parameter set that can be compared across devices.

[0090] In another preferred embodiment of the present invention, the symptom classification decision structure includes:

[0091] 1. Construction of an independent clinical validation database

[0092] First, we integrate a massive collection of confirmed medical cases, each containing a detailed set of multidimensional feature parameters and the final triage conclusion. Data is extracted from electronic medical records using natural language processing technology and undergoes multiple verifications by experienced clinical experts to ensure high accuracy in data annotation. This database covers typical conditions across diverse anatomical regions, providing a solid foundation for subsequent model training.

[0093] 2. The training and reasoning mechanism of the tree-type decision structure uses a supervised learning algorithm to train the tree-type decision structure. Its architecture is designed as follows:

[0094] The root node corresponds to the anatomical mapping area type and is directly triggered by the touch positioning result to ensure the accuracy of the decision starting point;

[0095] The branch nodes are associated with the confidence intervals of the multi-dimensional feature parameters. Each branch represents the judgment condition of a key feature indicator, such as temperature gradient, pressure slope, etc.

[0096] Leaf nodes are bound to triage priority codes. The code generation rules strictly adhere to the urgency grading principles of the International Classification of Diseases, covering a multi-level classification from immediate emergency care to routine waiting. During the inference process, the input multidimensional feature parameter set is filtered layer by layer along the decision tree branches, ultimately matching the most appropriate leaf node and generating the corresponding triage conclusion.

[0097] 3. Dynamic optimization mechanism of decision-making structure, which continuously updates the decision-making structure through sliding time window technology, including:

[0098] Main trunk retention: Maintain the main framework of anatomical regional classification to ensure the stability and interpretability of clinical logic;

[0099] Branch optimization: Adjust the boundaries of the confidence intervals of feature parameters based on the latest clinical data, so that the model can adapt to subtle changes in disease characteristics;

[0100] Leaf node merging: Cluster and merge leaf nodes that are triggered infrequently, simplifying the model complexity while retaining the core diagnostic path and improving decision-making efficiency.

[0101] In another preferred embodiment of the present invention, the generation process of the triage priority code is as follows:

[0102] Dynamic adaptation of feature analysis channels

[0103] The system automatically matches the corresponding feature analysis channel based on the type of target sensor combination currently activated, such as the pressure sensor combination configured for pain-sensitive areas and the temperature sensor combination configured for areas with high inflammation. Each channel integrates a dedicated algorithm module. For example, the pressure signal channel uses wavelet transform to analyze the dynamic changes in pressure intensity and generate a feature confidence score value between 0 and 1. The higher the score, the higher the degree of match between the current signal feature and the target disease. Taking the precordial touch event as an example, if the pressure channel detects a significant increase in pressure intensity in a short period of time, the score can reach 0.92, strongly indicating the possibility of acute chest pain-related diseases.

[0104] Decision structure traversal and conflict resolution mechanism

[0105] The confidence scores output by each feature analysis channel are input into a pre-trained tree-like decision structure. Starting from the root node (anatomical region type), the system filters through branch nodes (such as "significant change in pressure slope" and "abnormal temperature gradient") layer by layer, ultimately reaching the leaf nodes. When a single leaf node corresponds to multiple triage priority codes, such as angina pectoris and pleurisy, the system calls a medical urgency weight comparison table based on the International Classification of Diseases. This table presets urgency weights for different conditions, such as 9.0 for acute myocardial infarction and 6.5 for common pneumonia. The system automatically selects the code with the highest weight as the temporary output result to ensure that urgent conditions are treated first.

[0106] Pre-verification of pathological logical compatibility

[0107] Before generating the final triage code, the system performs a rigorous pathology logic compatibility check to avoid misdiagnosis due to signal interference or feature inconsistencies. For example, if the feature vector generated for an abdominal touch event contains both "local temperature increase" and "normal pressure," the system verifies that these two features conform to the known pathology logic. Clinical experience shows that appendicitis is often accompanied by tenderness and temperature increase. Temperature increase alone is more likely to indicate superficial inflammation rather than deeper organ disease. This type of verification can eliminate conclusions that do not conform to pathological principles.

[0108] In a preferred embodiment of the present invention, the specific process of the pathological logic compatibility check is as follows:

[0109] The construction and application of the pathological feature knowledge base integrates more than 3,000 typical pathological features verified by medical literature. Each entry contains:

[0110] Anatomical region labels, such as "right lower abdomen" and "medial side of knee joint";

[0111] Characteristic parameter constraint range, for example, appendicitis corresponds to "temperature gradient not less than 0.8℃ / cm" and "pressure peak not less than 18N";

[0112] Time evolution dimension constraints, such as the temperature increase caused by inflammation must last for more than 5 minutes; spatial distribution dimension constraints, such as joint lesions usually require the bilateral temperature difference to not exceed 1°C; multimodal association constraints, such as precordial pain is often accompanied by abnormal ECG signals.

[0113] The system splits the extracted multi-dimensional feature parameters into independent feature vectors according to their attributes. Each vector contains three-dimensional data:

[0114] The time evolution dimension records the changing patterns of signal characteristics over time, such as the dynamic curve of the pressure signal and the duration of temperature anomalies;

[0115] The spatial distribution dimension describes the distribution characteristics of the signal within the anatomical region, such as the symmetry of the temperature field and the concentration of the pressure peak;

[0116] The multimodal correlation dimension reflects the interactive relationship between different types of signals, such as the frequency correlation between pressure and vibration signals, and the energy coupling between temperature and bioelectric signals.

[0117] The comparison operation is carried out in five steps: first, the corresponding pathology entry set in the knowledge base is located according to the touch area. For example, the "left chest" entry set includes diseases such as angina pectoris and pleurisy. Then, the Euclidean distance algorithm is used to calculate the deviation of the data in each dimension from the entry constraint range, that is, the absolute difference between the measured value and the constraint median divided by the constraint range. Then, the deviation of the time evolution, spatial distribution, and multimodal association dimensions is compared with the preset thresholds. When the deviation of the time dimension exceeds the first threshold, the spatial dimension exceeds the second threshold, and the association dimension exceeds the third threshold, the temporal logic conflict, spatial distribution conflict, and cross-validation conflict are marked respectively.

[0118] Intelligent execution of differentiated processing flow, the system initiates targeted processing based on the conflict type distribution map:

[0119] If there is only a single dimension conflict, such as deviation in the time evolution dimension, the system performs confidence downgrade processing on the feature vector, for example, reducing the original score of 0.8 to 0.6, thereby reducing its influence weight in the decision-making structure;

[0120] If a multi-dimensional conflict occurs, such as a simultaneous deviation in both time and space, the system triggers a signal re-collection instruction, extending the sensor collection time from 10 seconds to 30 seconds and adding auxiliary sensors. For example, in the case of abdominal conflict, the bioimpedance sensor is activated to detect ascites to obtain more comprehensive signal data.

[0121] For non-conflicting feature vectors, the system increases their decision priority and adds a 20% weight coefficient when generating triage codes to ensure that they participate in the calculation first.

[0122] In another preferred embodiment of the present invention, a triage result verification process is further included:

[0123] 1. In-depth analysis and feature comparison of clinical feedback data

[0124] The triage result verification process has established a closed-loop iterative system of "data feedback-deviation analysis-model optimization" to ensure that the system continues to be close to actual clinical needs. First, the system receives confirmed data fed back by the medical system in real time through a secure encrypted interface. These data include disease classification codes, emergency classification labels, and detailed biophysical signal records in electronic medical records. Disease classification codes follow the internationally accepted ICD-11 standard, and emergency classification labels cover different levels such as immediate rescue, consultation within 30 minutes, and routine waiting. The system uses natural language processing technology to deeply analyze medical record texts, extract key feature parameters such as body temperature, pressure strength, vibration frequency, voice features, etc., and generate a three-dimensional reference feature set containing time series, spatial distribution, and multimodal associations.

[0125] The system then calculates the dimension-by-dimension deviation between the reference feature set and the multidimensional feature parameters extracted during this triage process. Taking the temperature feature of the McBurney point in the right lower abdomen as an example, if the reference data shows a temperature gradient of 1.2°C / cm in this area, while the value extracted by the system during triage is 0.9°C / cm, and the deviation between the two exceeds the preset tolerance threshold (e.g., 0.2°C / cm), the temperature dimension will be marked as an "out-of-standard dimension." This deviation calculation covers all feature parameters, including dozens of indicators such as the slope of the pressure rise edge, the thermal diffusion rate, and the phase synchronization coefficient, ensuring full dimensional verification of the data.

[0126] 2. Traceability Analysis and Integrity Verification of Exceeding Dimensions

[0127] For the marked out-of-standard dimensions, the system activates a three-layer traceability mechanism to investigate the root causes layer by layer:

[0128] Signal processing parameter verification: This tool tracks the algorithm execution path in the feature analysis mode to verify that the signal processing parameters are consistent with the biophysical characteristics of the current anatomical region. For example, vibration signal analysis in the knee joint area requires an 8-15Hz bandpass filter to highlight synovial hyperplasia. If the actual parameter is set to 5-20Hz, high-frequency signal noise may interfere with the detection of key pathological features. In this case, the system will automatically identify and recommend adjusting the filter parameters.

[0129] Data Collection Integrity Check: Verifies that the acquisition duration of the target sensor combination meets the minimum data integrity requirements. Different anatomical regions and sensor combinations have different preset acquisition durations, such as at least 10 seconds for the Basic signal acquisition combination and at least 30 seconds for the Enhanced combination. If, during an abdominal touch event, the pressure sensor only acquires data for 8 seconds before stopping, resulting in missing key parameters such as peak duration, the system will flag this as "Insufficient Acquisition Duration" and automatically extend the acquisition window for subsequent similar events.

[0130] Decision Weight Reasonability Assessment: This system analyzes the weight coefficients of the overweighted dimensions in the symptom classification decision structure to determine whether they reasonably reflect the importance of the feature in clinical diagnosis. For example, in the diagnosis of appendicitis, if the diagnostic weight of temperature gradient is lower than that of pressure, the inflammation feature may be weakened. The system will recalculate and adjust the weight distribution by comparing a large amount of clinical data to ensure that key features have a reasonable proportion in the decision.

[0131] 3. Multi-dimensional Optimization Instruction Generation and Effect Verification

[0132] Based on the traceability results, the system generates three types of optimization instructions to improve diagnostic performance in a targeted manner:

[0133] Algorithm parameter adjustments: Addressing signal processing issues, key parameters of the feature extraction algorithm were optimized. For example, the Fast Fourier Transform (FFT) analysis window for joint vibration signals was extended from 200ms to 300ms to improve the accuracy of identifying high-frequency harmonics. The threshold for calculating the temperature field symmetry index was adjusted from a tolerance of 1.5°C to 1.2°C to reduce false positive results due to individual differences.

[0134] Acquisition strategy optimization: For scenarios with insufficient acquisition time or incomplete sensor configuration, signal acquisition time is extended and auxiliary sensors are added. For example, the baseline signal acquisition time for the abdominal area is extended from 10 seconds to 15 seconds, the enhanced acquisition time is extended from 30 seconds to 45 seconds, and a bowel sound microphone is added as an auxiliary sensor to ensure more complete biophysical signals are obtained.

[0135] Decision structure reconstruction: The decision tree model was retrained using a machine learning algorithm, and the confidence intervals of feature parameters were adjusted. For example, the confidence level of "precordial pressure slope > 15 N / s" as a criterion for myocardial infarction was increased from 0.8 to 0.85. Low-frequency leaf nodes with fewer than 10 activations per quarter were merged, simplifying the model while retaining the core diagnostic pathway.

[0136] After the optimization instructions are generated, the system performs simulated backtesting using a dataset containing 100,000 historical cases. Taking myocardial infarction diagnosis as an example, the backtesting process includes: inputting pre-optimization feature parameters, generating triage codes, and comparing them with actual confirmed data; importing the optimized parameters for retesting, and calculating the magnitude of the difference between the two results. When the misdiagnosis rate for key conditions decreases by 15% or more (e.g., the misdiagnosis rate for myocardial infarction drops from 9% to 7.5%), and the deviation values ​​for more than 80% of cases are significantly reduced, the system determines that the optimization effect has met the target and automatically updates the optimized parameters to the signal acquisition strategy library, completing one round of iterative optimization.

[0137] The present invention also includes a triage system based on multimodal triboelectric analysis, which is used to implement the above-mentioned triage method based on multimodal triboelectric analysis, comprising:

[0138] An interactive model, wherein the surface of the interactive model is covered with a triboelectric material, and the interactive model is divided into a number of anatomical mapping areas, each area being configured with multiple types of sensor access ports;

[0139] A region detection module, which detects user touch events through triboelectric materials and analyzes the specific anatomical mapping region where the touch event occurred;

[0140] a strategy selection module, wherein the strategy selection module selects a corresponding signal acquisition strategy according to the type of the touched anatomical mapping area, wherein the signal acquisition strategy includes a target sensor combination identifier and a feature analysis mode identifier;

[0141] A parameter acquisition module, which activates corresponding sensors according to the target sensor combination identifier to collect the user's biophysical signal data stream; executes the calculation process specified by the feature analysis mode identifier to extract a multidimensional feature parameter set from the biophysical signal data stream;

[0142] A triage generation module inputs a multi-dimensional feature parameter set into a clinically verified symptom classification decision structure, generates a triage priority code, and transmits it to a medical system terminal through an encrypted communication protocol.

[0143] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A triage method based on multimodal triboelectric analysis, characterized in that: The following steps are involved: Provides an interactive model of a surface covered with triboelectric materials. The interactive model is divided into several anatomical mapping areas, each of which is equipped with multiple types of sensor access ports. Detecting user touch events through triboelectric materials and analyzing the specific anatomical mapping area where the touch event occurred; Selecting a corresponding signal acquisition strategy according to the type of the touched anatomical mapping area, the signal acquisition strategy including a target sensor combination identifier and a feature analysis mode identifier; Activate the corresponding sensor according to the target sensor combination identifier and collect the user's biophysical signal data stream; Execute the computational process specified by the feature analysis pattern identifier to extract a multidimensional feature parameter set from the biophysical signal data stream; Input the multi-dimensional feature parameter set into the clinically verified symptom classification decision structure to generate a triage priority code and transmit it to the medical system terminal through an encrypted communication protocol; The process of resolving the specific anatomical mapping area where a touch event occurred is: A high-density capacitive sensing network is deployed beneath the triboelectric material layer. The network nodes form a coordinate system that maps the boundaries of the anatomically mapped region. The network captures the distribution of capacitance value mutations caused by touch events and calculates the spatial Euclidean distance between the peak coordinates of the mutation and the nearest anatomically mapped region boundary. When the spatial Euclidean distance is less than a preset tolerance threshold, it is determined that the touch event belongs to the anatomical mapping area; The entries in the signal acquisition strategy library are indexed according to the type of the anatomical mapping area, and the entries record the target sensor combination activation sequence and feature analysis mode calling sequence corresponding to the area; The calculation process of the feature analysis mode includes: Perform time series decomposition on dynamic pressure signals to extract the slope change curve of the pressure rise, the peak maintenance time, and the oscillation decay period parameters; perform spatial gradient calculation on temperature distribution signals to generate thermal diffusion rate values ​​and temperature field symmetry indexes; perform time-frequency domain correlation analysis on multi-source mixed signals to calculate the phase synchronization coefficient matrix and energy coupling strength values ​​across signal channels; All characteristic parameters are standardized by the regional model to eliminate the individual response differences of sensors and generate a comparable multi-dimensional characteristic parameter set; The symptom classification decision structure includes: An independent clinical validation database was constructed to store the multidimensional feature parameter sets and final triage conclusions of medically confirmed cases. A supervised learning algorithm was used to train a tree-like decision structure, with the root node corresponding to the anatomical mapping region type and the branch nodes associated with the confidence intervals of the multidimensional feature parameters. Leaf nodes were bound to triage priority codes, and the code generation rules followed the urgency grading principles of the International Classification of Diseases. The decision structure is updated through a sliding time window mechanism, the main branches of regional types are retained, the confidence interval boundaries of feature parameters are optimized, and low-frequency leaf nodes are merged.

2. A triage method based on multimodal triboelectric analysis according to claim 1, characterized in that: The signal acquisition strategy library includes: Collect typical pathological characterization data from different anatomical mapping areas and establish a mapping relationship table between regional pathological characteristics and biophysical signal types. Based on the mapping relationship table, generate target sensor combination configuration rules: configure a dynamic pressure signal acquisition combination for pain-sensitive areas and a temperature distribution signal acquisition combination for areas with high inflammation incidence. Activation conditions are set for each target sensor combination: the first touch triggers the basic signal acquisition combination, and the continuous touch timeout triggers the enhanced signal acquisition combination; the basic signal acquisition combination includes core biophysical sensors, and the enhanced signal acquisition combination adds auxiliary cross-validation sensors.

3. A triage method based on multimodal triboelectric analysis according to claim 2, characterized in that: The determination logic of the continuous touch is: Recording a sequence of continuous touch events on the same anatomically mapped area over a preset time period; verifying that the time intervals between touch events are within the normal operating interval range for the human body; Calculate the spatial position distribution density of continuous touch points, and confirm the validity of the operation when the density value exceeds the set ratio of the area; start the enhanced signal acquisition combination activation instruction for the effective continuous touch action, and extend the feature analysis time window.

4. The triage method based on multimodal triboelectric analysis according to claim 1, characterized in that: The generation process of the triage priority code is as follows: Select a feature analysis channel based on the activated target sensor combination type, and output a feature confidence score value for each channel; Input the feature confidence score value into the symptom classification decision structure and traverse along the branch node path to the leaf node; When a leaf node maps multiple triage priority codes, a conflict arbitration protocol is executed: the medical urgency weight comparison table is called; Select the triage priority code with the highest weight value as the output result; A pathological logic compatibility check is performed before output to verify whether there is any contradiction between the multidimensional feature parameter set and the known pathological representation of the anatomical mapping area.

5. A triage method based on multimodal triboelectric analysis according to claim 4, characterized in that: The specific process of the pathological logic compatibility check is as follows: Construct a knowledge base of pathological features of the anatomical mapping area. The knowledge base stores typical pathological representations verified by medical literature and their corresponding feature parameter constraint ranges. Split the currently extracted multidimensional feature parameter set into independent feature vectors according to parameter categories. Each feature vector contains data in the time evolution dimension, spatial distribution dimension, and multimodal association dimension. Perform a knowledge base comparison operation on each independent feature vector: Step 1: Locate the pathological feature entry set of the anatomical mapping region to which the feature vector belongs; Step 2: Calculate the deviation between the data of each dimension of the feature vector and the pathological representation constraint range in the entry set; Step 3: When the deviation of the time evolution dimension exceeds the first threshold, it is marked as a temporal logic conflict; Step 4: When the spatial distribution dimension deviation exceeds the second threshold, it is marked as a spatial distribution conflict; Step 5: When the multimodal correlation dimension deviation exceeds the third threshold, it is marked as a cross-validation conflict; The conflict marking results of all independent feature vectors are counted to generate a conflict type distribution map; a differentiated processing process is initiated based on the conflict type distribution map: confidence reduction processing is performed on the feature vectors of single-dimensional conflict to reduce their weight coefficient in the symptom classification decision structure; re-collection instructions are triggered for feature vectors of multi-dimensional composite conflict, the collection time of the target sensor combination is extended and auxiliary verification sensors are added; the decision priority of non-conflict feature vectors is increased, and they are given priority in participating in the triage priority code generation operation.

6. The triage method based on multimodal triboelectric analysis according to claim 1, characterized in that: It also includes the triage result verification process: Receive actual confirmed data fed back by the medical system, parse the disease classification code and emergency classification label therein, and generate a reference feature parameter set; calculate the deviation dimension by dimension between the reference feature parameter set and the multidimensional feature parameter set, and screen feature dimensions whose deviation value exceeds the tolerance threshold; For the extracted paths in the feature analysis mode of the exceeded dimensions, verify whether the signal processing parameters are consistent with the biophysical characteristics of the current anatomical mapping area; check whether the acquisition time of the target sensor combination meets the minimum requirements for data integrity; and evaluate the rationality of the weight distribution of the multidimensional feature parameters in the symptom classification decision structure; Generate optimization instructions based on the verification results, adjust the feature analysis calculation coefficients, extend the signal acquisition time window, or reconstruct the confidence interval of the decision structure; use historical case data sets to perform simulation backtesting, and compare the change in the difference between the triage priority code and the actual confirmed data before and after optimization; when the difference reduction ratio reaches the set benchmark, the optimization parameters are synchronously updated to the signal acquisition strategy library.

7. A triage system based on multimodal triboelectric analysis, used to implement a triage method based on multimodal triboelectric analysis according to any one of claims 1 to 6, characterized in that: include: An interactive model, wherein the surface of the interactive model is covered with a triboelectric material, and the interactive model is divided into a number of anatomical mapping areas, each area being configured with multiple types of sensor access ports; A region detection module, which detects user touch events through triboelectric materials and analyzes the specific anatomical mapping region where the touch event occurred; a strategy selection module, wherein the strategy selection module selects a corresponding signal acquisition strategy according to the type of the touched anatomical mapping area, wherein the signal acquisition strategy includes a target sensor combination identifier and a feature analysis mode identifier; A parameter acquisition module, which activates corresponding sensors according to the target sensor combination identifier and collects a biophysical signal data stream of the user; Execute the computational process specified by the feature analysis pattern identifier to extract a multidimensional feature parameter set from the biophysical signal data stream; A triage generation module inputs a multi-dimensional feature parameter set into a clinically verified symptom classification decision structure, generates a triage priority code, and transmits it to a medical system terminal through an encrypted communication protocol.

Images