Intelligent pharyngolaryngitis monitoring system and method based on multi-parameter acquisition

Abstract

The invention relates to the technical field of medical intelligence, and discloses a throat inflammation intelligent monitoring system and method based on multi-parameter acquisition, and the system comprises a physiological data acquisition module, a feature separation module, a physiological feature fusion module, an inflammation prediction module and a health report generation module. Synchronously collecting multi-source physiological data of the throat part of the patient; carrying out noisy point cleaning on the multi-source physiological data to obtain standardized multi-modal data of the multi-source physiological data; extracting a temperature anomaly region distribution feature, a sound spectrum resonance frequency band feature and a mucous membrane texture anomaly feature in the standardized multi-modal data; performing feature level fusion on the temperature anomaly region distribution feature, the acoustic spectrum resonance frequency band feature and the mucous membrane texture anomaly feature to obtain a fusion feature vector of the throat part; performing inflammation prediction on the patient according to the fusion feature vector to obtain the throat inflammation grade of the patient; generating a health report of the patient according to the throat inflammation grade; according to the invention, the accuracy of pharyngolaryngitis identification can be improved.

Description

Technical Field

[0001] This invention relates to the field of medical intelligent technology, and in particular to an intelligent monitoring system and method for pharyngeal inflammation based on multi-parameter acquisition. Background Technology

[0002] Current technologies for monitoring throat inflammation often rely on analyzing single types of physiological data, failing to comprehensively integrate diverse features of the throat region. This single-parameter acquisition mode cannot fully reflect the complex physiological state of inflammation, making inflammation identification susceptible to interference from local abnormalities. In cases of similar symptoms or early stages of inflammation, missed or false diagnoses are common, resulting in low overall accuracy.

[0003] Meanwhile, existing technologies lack a systematic processing and fusion mechanism for the collected physiological data. On the one hand, environmental noise and interference signals in the raw data are not effectively cleaned, resulting in insufficient reliability of feature extraction. On the other hand, the lack of scientific weighting and correlation analysis in the integration of multiple features fails to fully leverage the synergistic effect of different features in inflammation assessment, thus limiting the accuracy of inflammation grade prediction and making it difficult to meet clinical requirements for the accuracy and stability of monitoring results. Summary of the Invention

[0004] This invention provides an intelligent monitoring system and method for pharyngeal inflammation based on multi-parameter acquisition, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition, characterized in that the system includes a physiological data acquisition module, a feature separation module, a physiological feature fusion module, an inflammation prediction module, and a health report generation module, wherein:

[0006] The physiological data acquisition module is used to simultaneously collect multi-source physiological data of the throat area in patients; and to perform noise removal on the multi-source physiological data to obtain standardized multimodal data of the multi-source physiological data;

[0007] The feature separation module is used to extract the temperature anomaly region distribution features, acoustic spectrum resonance frequency band features, and mucosal texture anomaly features from the standardized multimodal data.

[0008] The physiological feature fusion module is used to perform feature-level fusion of the temperature anomaly region distribution features, the acoustic spectrum resonance frequency band features, and the mucosal texture anomaly features to obtain the fused feature vector of the pharyngeal region.

[0009] The inflammation prediction module is used to predict the inflammation of the patient based on the fused feature vector to obtain the patient's pharyngeal inflammation level;

[0010] The health report generation module is used to generate a health report for the patient based on the level of throat inflammation.

[0011] In a preferred embodiment, when the physiological data acquisition module performs synchronous acquisition of multi-source physiological data from the pharynx of a patient, it is specifically used for:

[0012] Based on the anatomical features of the pharynx, the key areas for data collection were determined;

[0013] Multi-source data collection was performed on key areas of the patient to obtain physiological data of the pharynx.

[0014] The infrared temperature data, acoustic vibration data, and optical image data in the physiological data are time-stamped and synchronized to obtain multi-source physiological data of the pharynx.

[0015] In a preferred embodiment, when the physiological data acquisition module performs noise removal on the multi-source physiological data to obtain standardized multimodal data of the multi-source physiological data, it is specifically used for:

[0016] Based on a reference temperature point, the influence of environmental temperature fluctuations on the temperature distribution in the multi-source physiological data is eliminated to obtain the temperature data of the pharynx.

[0017] By removing environmental noise and respiratory interference components from the acoustic vibrations in the multi-source physiological data, a filtered acoustic signal of the pharyngeal region is obtained.

[0018] Illumination balancing is performed on the image data from the multi-source physiological data to obtain enhanced image data of the pharyngeal region;

[0019] The temperature data, the filtered acoustic signal, and the enhanced image data are dimensionally aligned to obtain standardized multimodal data of the pharyngeal region.

[0020] In a preferred embodiment, the feature separation module, when extracting temperature anomaly region distribution features, acoustic spectral resonance frequency band features, and mucosal texture anomaly features from the standardized multimodal data, is specifically used for:

[0021] Identify the temperature anomaly regions in the temperature distribution data of the standardized multimodal data to obtain the temperature anomaly distribution of the pharynx;

[0022] The resonant frequency band energy distribution of the acoustic vibration data in the standardized multimodal data is extracted to obtain the acoustic spectrum resonance characteristics of the pharynx.

[0023] Based on the changes in texture parameters of the mucosal region in the standardized multimodal data, abnormal mucosal texture features of the pharynx are generated.

[0024] In a preferred embodiment, when the physiological feature fusion module performs feature-level fusion of the temperature anomaly region distribution features, the acoustic spectrum resonance frequency band features, and the mucosal texture anomaly features to obtain the fused feature vector of the pharynx, it is specifically used for:

[0025] The temperature anomaly region distribution characteristics, the acoustic spectrum resonance frequency band characteristics, and the mucosal texture anomaly characteristics are normalized to obtain the temperature feature vector, acoustic feature vector, and texture feature vector of the pharynx.

[0026] The temperature feature vector, the acoustic feature vector, and the texture feature vector are concatenated to obtain the feature matrix of the pharyngeal region.

[0027] The feature matrix is ​​weighted and fused to obtain the fused feature vector of the pharyngeal region.

[0028] In a preferred embodiment, the physiological feature fusion module performs weighted fusion of the feature matrix to obtain a fused feature vector for the pharyngeal region, specifically for:

[0029] Calculate the attention weights of the eigenvectors in the feature matrix, wherein the formula for calculating the attention weights is as follows: ;

[0030] In the formula, For the first Attention weights for each feature vector. For normalized exponential functions, For query vector, The hyperbolic tangent activation function is used. For the preset weight matrix, For the first 1 eigenvector For bias vectors, This is a flag indicating that the vector has been inverted.

[0031] The feature matrix is ​​weighted and fused based on the attention weights to obtain the fused feature vector of the throat region.

[0032] In a preferred embodiment, when the inflammation prediction module performs inflammation prediction on the patient based on the fused feature vector to obtain the patient's pharyngeal inflammation level, it is specifically used for: Extract the high-level feature representation of the fused feature vector to obtain the deep feature vector of the pharyngeal region;

[0033] The inflammation category probability of the patient is calculated based on the deep feature vector to obtain the inflammation probability distribution of the patient;

[0034] The patient's throat inflammation level is generated based on the inflammation probability distribution.

[0035] In a preferred embodiment, the inflammation prediction module calculates the inflammation probability distribution using the following formula:

[0036] ;

[0037] In the formula, For the first The nth deep feature vector is predicted as the nth The probability of inflammation-like symptoms. It is a natural exponential function. For the first Inflammation-like score, For the first Inflammation-like score, The ordinal number representing the degree of inflammation. This represents the total number of inflammation severity categories.

[0038] In a preferred embodiment, when the health report generation module generates a health report for the patient based on the level of throat inflammation, it is specifically used for:

[0039] Based on the corresponding medical explanation text matched with the level of throat inflammation, the description of the inflammation level of the throat inflammation is obtained;

[0040] Based on the description of the inflammation level, a medical knowledge base was searched to obtain the precautions for the pharyngeal inflammation level.

[0041] The description of the inflammation level and the precautions are processed to optimize the language, resulting in the patient's health report.

[0042] To address the aforementioned problems, this invention also provides a method for intelligent monitoring of pharyngeal inflammation based on multi-parameter acquisition, the method comprising:

[0043] S1. Simultaneously collect multi-source physiological data of the pharynx in patients; and perform noise removal on the multi-source physiological data to obtain standardized multimodal data of the multi-source physiological data;

[0044] S2. Extract the temperature anomaly region distribution features, acoustic spectrum resonance frequency band features, and mucosal texture anomaly features from the standardized multimodal data;

[0045] S3. Perform feature-level fusion on the temperature anomaly region distribution characteristics, the acoustic spectrum resonance frequency band characteristics, and the mucosal texture anomaly characteristics to obtain the fused feature vector of the pharyngeal region;

[0046] S4. Based on the fused feature vector, predict the inflammation of the patient to obtain the pharyngeal inflammation level of the patient;

[0047] S5. Generate a health report for the patient based on the level of throat inflammation.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. This invention acquires comprehensive and reliable standardized multimodal data by simultaneously collecting multi-source physiological data from the pharynx and performing targeted noise removal and standardization. Based on this, the extracted temperature anomaly distribution features, acoustic spectrum resonance band features, and mucosal texture anomaly features accurately capture the characteristic manifestations of pharyngeal inflammation from different physiological dimensions. The synergistic effect of these multi-dimensional features can more comprehensively reflect the inflammatory state, significantly improving the comprehensiveness and accuracy of pharyngeal inflammation identification.

[0050] 2. This invention scientifically integrates multi-dimensional features through feature-level fusion and effectively highlights key features by combining attention weight calculation. The resulting fused feature vector can more accurately characterize the nature of inflammation. Inflammation prediction based on this can generate accurate throat inflammation levels. Combined with the detailed level descriptions and precautions provided by the health report generation module, this not only improves the accuracy of inflammation monitoring but also enhances the practicality of the results, providing reliable support for the assessment and intervention of throat inflammation. Attached Figure Description

[0051] Figure 1 This is a system architecture diagram of an intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition, provided in an embodiment of the present invention.

[0052] Figure 2 This is a flowchart illustrating an intelligent monitoring method for pharyngeal inflammation based on multi-parameter acquisition, provided in an embodiment of the present invention.

[0053] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0056] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0057] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0058] In practice, the server-side equipment deployed in a multi-parameter-based intelligent monitoring system for throat inflammation may consist of one or more devices. This system can be implemented as a business instance, a virtual machine, or hardware devices. For example, it can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, it can be understood as software deployed on a cloud node, providing a multi-parameter-based intelligent monitoring system for throat inflammation to various user terminals. Alternatively, it can be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed to manage various user terminals. Or, it can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more devices configured to provide a multi-parameter-based intelligent monitoring system for throat inflammation to various user terminals.

[0059] In terms of implementation, the intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition and the user terminal are mutually compatible. That is, if the intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition is implemented as an application installed on a cloud service platform, then the user terminal is implemented as a client that establishes a communication connection with the application; or if the intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition is implemented as a website, then the user terminal is implemented as a webpage; or if the intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.

[0060] like Figure 1 The figure shown is a system architecture diagram of an intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition, provided by an embodiment of the present invention.

[0061] The intelligent monitoring system 100 for pharyngeal inflammation based on multi-parameter acquisition described in this invention can be set up in a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed into a website. Depending on the functions implemented, the intelligent monitoring system 100 for pharyngeal inflammation based on multi-parameter acquisition may include a physiological data acquisition module 101, a feature separation module 102, a physiological feature fusion module 103, an inflammation prediction module 104, and a health report generation module 105. The modules described in this invention can also be called units, referring to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0062] In this embodiment of the invention, in a multi-parameter acquisition-based intelligent monitoring system for pharyngeal inflammation, each of the above modules can be implemented independently and can call other modules. Here, "calling" can be understood as one module connecting to multiple modules of another type and providing corresponding services to those connected modules. The multi-parameter acquisition-based intelligent monitoring system for pharyngeal inflammation provided by this embodiment of the invention allows for adjustment of the system's applicability by adding modules and directly calling them, without modifying the program code. This enables cluster-based horizontal expansion, facilitating quick and flexible expansion of the system. In practical applications, the above modules can be set in the same device or different devices, or in virtual devices, such as service instances on a cloud server.

[0063] The following describes, with reference to specific embodiments, each component and its specific workflow of an intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition:

[0064] The physiological data acquisition module 101 is used to simultaneously collect multi-source physiological data of the throat area in patients; and to perform noise removal on the multi-source physiological data to obtain standardized multimodal data of the multi-source physiological data.

[0065] In this embodiment of the invention, when the physiological data acquisition module performs synchronous collection of multi-source physiological data from the pharynx of a patient, it is specifically used for:

[0066] Based on the anatomical features of the pharynx, the key areas for data collection were determined;

[0067] Multi-source data collection was performed on key areas of the patient to obtain physiological data of the pharynx.

[0068] The infrared temperature data, acoustic vibration data, and optical image data in the physiological data are time-stamped and synchronized to obtain multi-source physiological data of the pharynx.

[0069] In this embodiment of the invention, when the physiological data acquisition module performs noise removal on the multi-source physiological data to obtain standardized multimodal data of the multi-source physiological data, it is specifically used for:

[0070] Based on a reference temperature point, the influence of environmental temperature fluctuations on the temperature distribution in the multi-source physiological data is eliminated to obtain the temperature data of the pharynx.

[0071] By removing environmental noise and respiratory interference components from the acoustic vibrations in the multi-source physiological data, a filtered acoustic signal of the pharyngeal region is obtained.

[0072] Illumination balancing is performed on the image data from the multi-source physiological data to obtain enhanced image data of the pharyngeal region;

[0073] The temperature data, the filtered acoustic signal, and the enhanced image data are dimensionally aligned to obtain standardized multimodal data of the pharyngeal region.

[0074] Specifically, the key areas for data collection were determined based on the anatomical features of the pharynx. The pharynx is divided into the nasopharynx, oropharynx, and laryngopharynx from top to bottom. The oropharynx includes structures such as the tonsils and posterior pharyngeal wall, while the laryngopharynx includes structures such as the epiglottis and vocal cords. These structures are common sites of pharyngeal diseases. Therefore, the tonsillar fossa where the tonsils are located, the mucosal surface of the posterior pharyngeal wall, the free edge of the epiglottis, and the surface and edges of the vocal cords were identified as key areas for data collection. The specific locations and extents of these areas in the pharynx were clarified using human anatomical atlases to ensure accurate coverage during data collection.

[0075] Furthermore, multi-source data collection was performed on key areas of the patient to obtain physiological data of the pharynx. An infrared thermal imager was used to capture images of the identified key areas for 10 seconds while the patient maintained natural breathing and did not swallow, obtaining infrared temperature data reflecting the temperature distribution within the area. Simultaneously, an acoustic vibration sensor was placed against the patient's neck, corresponding to the vocal cords, and the patient was asked to emit an "ah" sound for 5 seconds. The sensor recorded acoustic vibration data of the key area, reflecting the amplitude and frequency changes of tissue vibration within the area. In addition, an electronic laryngoscope was inserted through the patient's mouth until the key area was clearly visible, capturing static images and a continuous 5-second dynamic video of the area to obtain optical image data. This data presents the morphology, color, and mucosal surface condition of the area. The aforementioned infrared temperature data, acoustic vibration data, and optical image data together constitute the physiological data of the pharynx.

[0076] Furthermore, the infrared temperature data, acoustic vibration data, and optical image data in the physiological data are time-stamped and synchronized to obtain multi-source physiological data of the pharynx. Before multi-source data acquisition, the infrared thermal imager, acoustic vibration sensor, and electronic laryngoscope are connected to the same computer. A unified time starting point is set through the synchronization control program in the computer. When the program issues a start signal, the three devices start recording data simultaneously and add a timestamp based on the time starting point to their respective recorded data. The timestamp is accurate to milliseconds. After acquisition, the three types of data are imported into the computer, and the parts with the same timestamp in each type of data are extracted by data processing software. For example, the infrared temperature data, acoustic vibration data, and optical image data with timestamps of 0 milliseconds, 1 millisecond, 2 milliseconds, etc., are matched one-to-one and combined to form a set containing the three types of data in the same time dimension, thus obtaining the multi-source physiological data of the pharynx.

[0077] Specifically, the temperature data of the pharynx is obtained by eliminating the influence of environmental temperature fluctuations on the temperature distribution in the multi-source physiological data based on a reference temperature point. A stable mucosal location in a non-critical area of ​​the patient's oral cavity is selected as the reference temperature point. The temperature value of this reference temperature point is continuously recorded while collecting multi-source physiological data. The difference between the temperature value of each data point in the multi-source physiological data and the temperature value of the reference temperature point during the same period is calculated. This difference is used as a correction factor to adjust the temperature distribution data; that is, the original value of each temperature data point is subtracted from the corresponding correction factor. The resulting temperature data of the pharynx after adjustment is the temperature data of the pharynx after eliminating the influence of environmental temperature fluctuations.

[0078] Further, environmental noise and respiratory interference components from the acoustic vibrations in the multi-source physiological data are removed to obtain the filtered acoustic signal of the pharynx. Simultaneously with the acquisition of acoustic vibration data, an environmental noise signal is recorded at a quiet location 1 meter away from the patient using an acoustic vibration sensor of the same model. The acoustic vibration data from the multi-source physiological data is compared with the environmental noise signal, and the portion of the acoustic vibration data consistent with the characteristics of the environmental noise signal is removed. By observing the rhythm of chest rise and fall during the patient's breathing, the time interval of respiratory interference is determined, and the signal within that time interval is deleted from the acoustic vibration data; the remaining portion is the filtered acoustic signal of the pharynx.

[0079] Further, illumination balancing is performed on the image data from the multi-source physiological data to obtain enhanced image data of the pharyngeal region. The darkest and brightest pixels within key areas of the image data are selected, and their brightness difference is calculated. The brightness values ​​of all pixels in the image are adjusted proportionally, increasing the brightness of the darkest pixel to 1.2 times its original brightness and decreasing the brightness of the brightest pixel to 0.8 times its original brightness, while maintaining the original proportional distribution of brightness between the adjusted darkest and brightest values ​​for other pixels. The resulting image data is the enhanced image data of the pharyngeal region.

[0080] Furthermore, the temperature data, the filtered acoustic signal, and the enhanced image data are dimensionally aligned to obtain standardized multimodal data of the pharynx. Using the pixel coordinates of the enhanced image data as a reference, the actual position of the pharynx corresponding to each pixel is determined. The temperature data is mapped to the pixel coordinates of the enhanced image data according to its corresponding actual position, so that each pixel simultaneously corresponds to a temperature value. Based on the position of the pharyngeal structure (such as the vocal cords) reflected by the filtered acoustic signal in the enhanced image, each data point of the filtered acoustic signal is associated with its corresponding pixel coordinates, ensuring a one-to-one correspondence between the temperature data, the filtered acoustic signal, and the enhanced image data within the same pixel coordinate system. The resulting set containing pixel coordinates, temperature values, and acoustic signals constitutes the standardized multimodal data of the pharynx.

[0081] In summary, this invention acquires comprehensive and reliable standardized multimodal data by simultaneously collecting multi-source physiological data from the pharynx and performing targeted noise removal and standardization. Based on this, the extracted temperature anomaly distribution features, acoustic spectrum resonance band features, and mucosal texture anomaly features accurately capture the characteristic manifestations of pharyngeal inflammation from different physiological dimensions. The synergistic effect of these multi-dimensional features more comprehensively reflects the inflammatory state, significantly improving the comprehensiveness and accuracy of pharyngeal inflammation identification.

[0082] In summary, this invention scientifically integrates multi-dimensional features through feature-level fusion and effectively highlights key features by combining attention weight calculation. The resulting fused feature vector can more accurately characterize the nature of inflammation. Inflammation prediction based on this can generate accurate throat inflammation levels. Combined with the detailed level descriptions and precautions provided by the health report generation module, this not only improves the accuracy of inflammation monitoring but also enhances the practicality of the results, providing reliable support for the assessment and intervention of throat inflammation.

[0083] The feature separation module 102 is used to extract the temperature anomaly region distribution features, acoustic spectrum resonance frequency band features, and mucosal texture anomaly features from the standardized multimodal data.

[0084] In this embodiment of the invention, when the feature separation module extracts the temperature anomaly region distribution features, acoustic spectrum resonance frequency band features, and mucosal texture anomaly features from the standardized multimodal data, it is specifically used for:

[0085] Identify the temperature anomaly regions in the temperature distribution data of the standardized multimodal data to obtain the temperature anomaly distribution of the pharynx;

[0086] The resonant frequency band energy distribution of the acoustic vibration data in the standardized multimodal data is extracted to obtain the acoustic spectrum resonance characteristics of the pharynx.

[0087] Based on the changes in texture parameters of the mucosal region in the standardized multimodal data, abnormal mucosal texture features of the pharynx are generated.

[0088] Specifically, multiple healthy individuals without throat diseases were collected. Under the same environmental conditions as patients, the same infrared thermal imager was used to collect temperatures in key areas of their throat, including the tonsillar fossa, posterior pharyngeal wall mucosa, free edge of the epiglottis, and vocal cords. Each collection lasted 10 seconds, and the average temperature of all temperature points in each key area was taken as the representative temperature of that area. The representative temperatures of each key area of ​​these 1000 healthy individuals were statistically analyzed. The highest and lowest 5% of values ​​were removed, and the minimum and maximum values ​​of the remaining 90% constituted the normal temperature range of that area. Each temperature point in the standardized multimodal data was examined. If the temperature value of a certain temperature point exceeded the normal temperature range of that area, the point was marked as a temperature anomaly. All temperature anomalies were classified according to the key areas they were located in, and the specific coordinates and temperature values ​​of each anomaly were recorded, forming a set containing the location of the anomaly area, the coordinates of the anomaly point, and the corresponding temperature value, thus obtaining the abnormal temperature distribution of the throat.

[0089] Furthermore, by analyzing filtered acoustic signals from multiple healthy individuals, the common resonance frequency band for acoustic vibrations in the pharynx was determined to be 100 Hz to 5000 Hz. This range covers the main frequencies of vocal cord vibration and surrounding tissue resonance in healthy individuals. The 100 Hz to 5000 Hz frequency band was then divided into four sub-bands: 100 Hz-500 Hz, 501 Hz-1000 Hz, 1001 Hz-2000 Hz, 2001 Hz-3000 Hz, and 3001 Hz-5000 Hz. Five sub-bands of 00 Hz are used, with each sub-band having a continuous and non-overlapping frequency range. Filtered acoustic signals are extracted from standardized multimodal data, and the vibration amplitude of the signal in each sub-band is analyzed segment by segment using acoustic analysis equipment. The vibration amplitudes at all times within each sub-band are summed to obtain the total energy of that sub-band. The sub-bands are arranged from low to high frequency, and the total energy of each sub-band is recorded sequentially. The resulting energy set arranged in frequency order is the acoustic spectrum resonance characteristic of the throat region.

[0090] Furthermore, several healthy individuals without throat diseases were selected, and enhanced images of their throats were taken using the same electronic laryngoscope as those used by patients. Mucosal images of areas such as the tonsillar fossa and posterior pharyngeal wall were extracted from these images. Image measurement software was used to measure the texture parameters in these mucosal images. This included measuring the width of 100 textures using the software's built-in length measurement tool, averaging the width to determine a normal average texture width of 0.1-0.3 mm, and using an angle measurement tool to determine whether the main direction of the texture was horizontal or vertical (angle with the horizontal line between 0-15 degrees or 75-90 degrees). A counting tool was used to count the number of textures in 10 1 square millimeter areas, averaging the number to determine a normal number of 5-8 textures. The mucosa was delineated within the standardized multimodal enhanced image data. For each region, the texture parameters within that region are measured using the same image measurement software. 100 textures are randomly selected, their widths are measured, and the average value is calculated. The angle between each texture and the horizontal line is measured to determine its direction. Ten 1-square-millimeter regions are selected to count the number of textures. The measurement results are compared with normal texture parameters. If the average width is less than 0.07 mm or greater than 0.39 mm, or if more than 60% of the texture directions have angles exceeding 45 degrees with the horizontal or vertical direction, or if the average number of textures in the 10 regions is less than 3.5 or greater than 10.4, then it is considered an abnormal texture. The type, specific value, and location of the abnormal texture parameters in the mucosal region are recorded. The resulting set containing the abnormal parameters, values, and locations represents the abnormal mucosal texture characteristics of the pharynx.

[0091] In summary, extracting the distribution features of abnormal temperature regions, acoustic spectrum resonance band features, and abnormal mucosal texture features from standardized multimodal data can accurately capture the specific manifestations of pharyngeal inflammation from multiple physiological dimensions. The distribution features of abnormal temperature regions can directly reflect the abnormal local thermal effects caused by inflammation, accurately locating the extent of inflammation; the acoustic spectrum resonance band features can capture the changes in acoustic characteristics of pharyngeal vibration function caused by inflammation, reflecting pathological changes at the functional level; and the abnormal mucosal texture features can present the morphological changes of mucosal tissue caused by inflammation, reflecting pathological features at the structural level.

[0092] In summary, the three elements complement each other from the perspectives of thermodynamics, acoustics, and morphology, comprehensively covering the characteristics of inflammation at different physiological levels. This provides rich and accurate basic data for subsequent feature fusion, enabling the fused feature vector to more completely represent the essence of inflammation, thereby improving the accuracy of inflammation identification and providing a multi-dimensional and reliable basis for accurately judging the state of pharyngeal inflammation.

[0093] The physiological feature fusion module 103 is used to perform feature-level fusion of the temperature abnormality region distribution features, the acoustic spectrum resonance frequency band features, and the mucosal texture abnormality features to obtain the fusion feature vector of the pharyngeal region.

[0094] In this embodiment of the invention, when the physiological feature fusion module performs feature-level fusion of the temperature anomaly region distribution features, the acoustic spectrum resonance frequency band features, and the mucosal texture anomaly features to obtain the fused feature vector of the pharynx, it is specifically used for:

[0095] The temperature anomaly region distribution characteristics, the acoustic spectrum resonance frequency band characteristics, and the mucosal texture anomaly characteristics are normalized to obtain the temperature feature vector, acoustic feature vector, and texture feature vector of the pharynx.

[0096] The temperature feature vector, the acoustic feature vector, and the texture feature vector are concatenated to obtain the feature matrix of the pharyngeal region.

[0097] The feature matrix is ​​weighted and fused to obtain the fused feature vector of the pharyngeal region.

[0098] In this embodiment of the invention, the physiological feature fusion module performs weighted fusion of the feature matrix to obtain a fused feature vector for the pharyngeal region, specifically for:

[0099] Calculate the attention weights of the eigenvectors in the feature matrix, wherein the formula for calculating the attention weights is as follows:

[0100] ;

[0101] In the formula, For the first Attention weights for each feature vector. For normalized exponential functions, For query vector, The hyperbolic tangent activation function is used. For the preset weight matrix, For the first 1 eigenvector For bias vectors, This is a flag indicating that the vector has been inverted.

[0102] The feature matrix is ​​weighted and fused based on the attention weights to obtain the fused feature vector of the throat region.

[0103] Specifically, the elements included in the distribution characteristics of temperature anomalies are determined, including the number of anomalies in each key area, the difference between the average temperature of the anomalies and the normal range, and the maximum and minimum values ​​of these elements in historical data are counted. The current value of each element is subtracted from the minimum value of the corresponding element, and the result is divided by the difference between the maximum and minimum values ​​of the element to obtain the normalized value of each element. These normalized values ​​are arranged in order, and the resulting vector is the temperature feature vector of the throat area.

[0104] Furthermore, the acoustic spectrum resonance frequency band features are determined to include the energy values ​​of five sub-bands. The maximum and minimum values ​​of these five energy values ​​in historical health data are statistically analyzed. The current energy value of each sub-band is subtracted from the minimum value of the corresponding sub-band, and the result is divided by the difference between the maximum and minimum values ​​of the sub-band to obtain the normalized value of each energy value. These five normalized values ​​are arranged in ascending order of sub-band values, and the resulting vector is the acoustic feature vector of the pharynx. The abnormal mucosal texture features are determined to include three parameters: average width, average directional angle, and average number. The maximum and minimum values ​​of each parameter are statistically analyzed in historical health data. The current value of each parameter is subtracted from the minimum value of the corresponding parameter, and the result is divided by the difference between the maximum and minimum values ​​of the parameter to obtain the normalized value of each parameter. These three normalized values ​​are arranged in the order of width, directional angle, and number, and the resulting vector is the texture feature vector of the pharynx.

[0105] Furthermore, the temperature feature vector, acoustic feature vector, and texture feature vector are connected in sequence, with all elements of the temperature feature vector at the beginning, followed by all elements of the acoustic feature vector, and finally all elements of the texture feature vector, forming a new matrix containing all elements of the three vectors. This matrix is ​​the feature matrix of the pharynx.

[0106] Furthermore, fixed weights are assigned to the temperature feature vector, acoustic feature vector, and texture feature vector, respectively. The weight of the temperature feature vector is set to 0.3, the weight of the acoustic feature vector is set to 0.4, and the weight of the texture feature vector is set to 0.3, with the sum of the three weights being 1. Each element in the temperature feature vector is multiplied by 0.3 to obtain the temperature-weighted element; each element in the acoustic feature vector is multiplied by 0.4 to obtain the acoustic-weighted element; and each element in the texture feature vector is multiplied by 0.3 to obtain the texture-weighted element. The temperature-weighted element, acoustic-weighted element, and texture-weighted element at the corresponding positions are added according to the order of the elements in the feature matrix, and the new vector obtained is the fused feature vector of the pharynx.

[0107] Specifically, It is the first Each feature vector originates from the feature matrix, namely, a temperature feature vector, an acoustic feature vector, or a texture feature vector extracted from the feature matrix. It is a preset weight matrix, the values ​​of which are manually set based on a large amount of historical throat feature data, which includes temperature, acoustic and texture feature vectors under different health conditions; It is a bias vector, the value of which is also manually set based on the aforementioned historical feature data, and is used to adjust... and The result after the calculation; It is a query vector, whose value is manually set according to the throat features that need to be focused on at present (such as abnormal temperature, abnormal acoustic resonance, etc.) to reflect the focus on different features.

[0108] Furthermore, the meaning of this formula is to calculate the attention weight of each feature vector (temperature feature vector, acoustic feature vector, texture feature vector) in the feature matrix, that is, by using the feature vectors With the preset weight matrix Add bias vector after calculation After further processing with the hyperbolic tangent activation function, an intermediate value is obtained. This intermediate value is then compared with the query vector. Multiply the components and then process them using a normalized exponential function to obtain the importance and weight of each eigenvector in the overall analysis. The larger the value, the more significant the influence of the feature vector on judging the state of the throat.

[0109] Furthermore, the trend of the formula is: when the eigenvector Features and query vectors included The better the match for the features of interest, through Operations, addition , The intermediate value obtained after processing and For the result of multiplication to be larger, the passage The result after processing The larger it is; conversely, when Features and The lower the feature matching degree, the lower the median value. The smaller the result of multiplication, The smaller it is; at the same time, The function will limit intermediate values ​​to between -1 and 1 to avoid extreme large or small values. The processing will make the sum of the attention weights of all feature vectors equal to 1, ensuring a reasonable weight distribution.

[0110] Specifically, temperature feature vector, acoustic feature vector, and texture feature vector are extracted from the feature matrix. The number of elements contained in each feature vector and their arrangement order in the feature matrix are determined. For example, the temperature feature vector contains 5 elements and is placed in the first 5 positions of the feature matrix, the acoustic feature vector contains 4 elements and is placed in the middle 4 positions, and the texture feature vector contains 3 elements and is placed in the last 3 positions.

[0111] Further, extract the weight values ​​from the attention weights that correspond to the three feature vectors, where the temperature feature vector corresponds to... Acoustic feature vectors correspond to Texture feature vectors correspond to ,and , , The sum of these weights is 1, and these weight values ​​reflect the importance of each feature vector during fusion.

[0112] Furthermore, each element in the temperature feature vector is compared with... Multiplying these results in five temperature-weighted elements. The value of each element is obtained by multiplying the original element value by... Obtained by direct multiplication; each element in the acoustic feature vector is multiplied by... Multiply by 4 to obtain 4 acoustic weighting elements; then multiply each element in the texture feature vector by 4. Multiplying them together yields three texture-weighted elements.

[0113] Furthermore, following the original arrangement order of the elements in the feature matrix, the temperature-weighted elements, acoustic-weighted elements, and texture-weighted elements are arranged sequentially to form a new vector containing 5+4+3=12 elements. This vector is the fusion feature vector of the pharyngeal region.

[0114] In summary, feature-level fusion of temperature anomaly distribution characteristics, acoustic spectrum resonance band characteristics, and mucosal texture anomaly characteristics can eliminate scale differences between different features through normalization, ensuring that multi-dimensional features participate in integration under a unified benchmark. Concatenating the normalized feature vectors to form a feature matrix, and then combining this with attention weight calculations for weighted fusion, can dynamically highlight key features more closely related to inflammation and weaken interference from secondary information.

[0115] In summary, this fusion approach integrates complementary information from thermal, acoustic, and morphological levels, and enhances the specificity of features through scientific weighting. This results in a more comprehensive and accurate capture of the essential characteristics of pharyngeal inflammation in the fused feature vector, effectively improving the completeness and specificity of feature representation and laying a reliable foundation for the accurate prediction of subsequent inflammation levels.

[0116] The inflammation prediction module 104 is used to predict the inflammation of the patient based on the fused feature vector to obtain the pharyngeal inflammation level of the patient.

[0117] In this embodiment of the invention, when the inflammation prediction module performs inflammation prediction on the patient based on the fused feature vector to obtain the patient's pharyngeal inflammation level, it is specifically used for:

[0118] Extract the high-level feature representation of the fused feature vector to obtain the deep feature vector of the pharyngeal region;

[0119] The inflammation category probability of the patient is calculated based on the deep feature vector to obtain the inflammation probability distribution of the patient;

[0120] The patient's throat inflammation level is generated based on the inflammation probability distribution.

[0121] In this embodiment of the invention, the inflammation prediction module calculates the inflammation probability distribution using the following formula:

[0122] ;

[0123] In the formula, For the first The nth deep feature vector is predicted as the nth The probability of inflammation-like symptoms. It is a natural exponential function. For the first Inflammation-like score, For the first Inflammation-like score, The ordinal number representing the degree of inflammation. This represents the total number of inflammation severity categories.

[0124] Specifically, the 10 elements with the highest correlation to pharyngeal inflammation in the fused feature vector are selected as initial features. These elements are selected based on the fact that their numerical differences between healthy and diseased states are most significant in historical inflammation case data. Specifically, they include 4 temperature-related elements, 3 acoustic-related elements, and 3 texture-related elements. These 10 elements are sorted from largest to smallest value, and the first 3 elements with the largest values ​​and the last 3 elements with the smallest values ​​are selected, forming a first set of 6 intermediate features. From the remaining 4 elements, the numerical difference between every two adjacent elements is calculated, and the 4 elements with the largest absolute difference are selected. Then, the difference between every two consecutive elements in these 4 elements is calculated, resulting in 3 difference results as the second set of intermediate features. The 6 original elements of the first set and the 3 difference results of the second set are arranged in the order of temperature, acoustic, and texture to form a set containing 9 elements, which is the depth feature vector of the pharynx.

[0125] Furthermore, the common types of throat inflammation were identified as acute pharyngitis, chronic pharyngitis, and tonsillitis. 500 sets of historical deep feature vector data corresponding to each of these three types of inflammation were collected. All historical data came from clinically diagnosed cases and were collected under the same conditions as the current patient. The current patient's deep feature vector was compared one by one with the 500 sets of historical data for each type of inflammation. For each historical data set, the number of elements in the current feature vector whose values ​​were within the same range (±5% of the historical element value) as the corresponding element in that historical data set was counted. This number was then calculated. The ratio of the current feature vector to the total number of elements (9) in the deep feature vector is used to obtain the matching ratio between the current feature vector and the historical data group. The average of the matching ratios in 500 historical data groups is taken as the initial probability of this type of inflammation. The initial probabilities of the three types of inflammation are adjusted. If the sum of the three is greater than 1, each probability is reduced proportionally to the sum. If the sum is less than 1, each probability is increased proportionally to the sum, until the sum of the three is 1. The three adjusted probabilities are arranged in the order of acute pharyngitis, chronic pharyngitis, and tonsillitis. The set formed is the inflammation probability distribution of the patient.

[0126] Furthermore, examine the inflammation category corresponding to the highest probability value in the inflammation probability distribution. If it is acute pharyngitis, count the number of elements with values ​​exceeding the healthy range among the four temperature-related elements in the deep feature vector. The ratio of this number to 4 is the abnormality ratio of temperature-related elements. When the maximum probability is greater than 0.7 and the abnormality ratio of temperature-related elements is greater than 50%, it is judged as severe acute pharyngitis. When the probability is between 0.4 and 0.7 and the abnormality ratio of temperature-related elements is between 30% and 50%, it is judged as moderate acute pharyngitis. When the probability is less than 0.4 and the abnormality ratio of temperature-related elements is less than 30%, it is judged as mild acute pharyngitis. If the maximum probability corresponds to chronic pharyngitis, count the ratio of the number of abnormal values ​​among the three texture-related elements to 3 as the texture abnormality ratio. Use the same probability interval as above combined with the texture abnormality ratio to determine mild, moderate, or severe. If it is tonsillitis, count the ratio of the number of abnormal values ​​among the three acoustic-related elements to 3 as the acoustic feature abnormality ratio. Similarly, combine the probability interval to determine the level. The final judgment result is the patient's pharyngeal inflammation level.

[0127] Specifically, It is the first A depth feature vector, derived from the depth feature vector of the pharyngeal region; and This represents the ordinal number of the inflammation category, corresponding to categories such as acute pharyngitis, chronic pharyngitis, and tonsillitis. This represents the total number of inflammation severity categories, the value of which is determined by the actual number of inflammation categories (e.g., 3 categories). It is the first The score for inflammation-like characteristics is determined by the number of... The depth feature vector and the first The historical deep feature vectors of inflammation-like structures are compared, and the degree of matching between the two is calculated. The higher the degree of matching, the better. The larger; It is the first The score for inflammation-like symptoms, the calculation method and Same, corresponding to the first Inflammation-like.

[0128] Furthermore, this formula is used to calculate the first... The nth deep feature vector is predicted as the nth The probability of inflammation, through analysis of the first... The score for a given inflammation class is calculated using its natural index, then divided by the sum of the natural indices of all inflammation class scores to obtain a normalized probability value. This probability value reflects the... The depth feature vector belongs to the th . The probability of inflammation of all categories, with the sum of the probabilities of all categories being 1, together constitute the inflammation probability distribution of the patient.

[0129] Furthermore, when the first Inflammation-like score As the natural index increases, its corresponding result increases, increasing its proportion in the total denominator, leading to the [missing information - likely a specific value or percentage]. The nth deep feature vector is predicted as the nth Probability of inflammation Increase; when When the natural index decreases, the corresponding natural index result decreases, and the probability decreases; if Scores in other categories The greater the difference, the more obvious the trend of increasing or decreasing probability becomes, ultimately resulting in the category with the highest score having the highest probability.

[0130] In summary, predicting the severity of throat inflammation based on fused feature vectors allows for full utilization of the multi-dimensional key information integrated within the fused feature vectors. By extracting high-level feature representations from the fused feature vectors, the intrinsic relationships between features can be deeply explored, more accurately capturing the essential characteristics of inflammation. Calculating the probability of inflammation categories based on deep feature vectors quantifies the likelihood of different levels, making the results more scientific. The final throat inflammation level can intuitively reflect the severity of the inflammation.

[0131] In summary, this prediction method, relying on comprehensive and optimized fusion features, effectively improves the accuracy and stability of inflammation level assessment, provides a reliable basis for the generation of subsequent health reports, helps to accurately assess the patient's throat inflammation status, and provides clear guidance for clinical intervention.

[0132] The health report generation module 105 is used to generate a health report for the patient based on the level of throat inflammation.

[0133] In this embodiment of the invention, when the health report generation module generates a health report for the patient based on the level of throat inflammation, it is specifically used for:

[0134] Based on the corresponding medical explanation text matched with the level of throat inflammation, the description of the inflammation level of the throat inflammation is obtained;

[0135] Based on the description of the inflammation level, a medical knowledge base was searched to obtain the precautions for the pharyngeal inflammation level.

[0136] The description of the inflammation level and the precautions are processed to optimize the language, resulting in the patient's health report.

[0137] Specifically, a pre-defined medical explanatory text library is established, covering all possible levels of pharyngeal inflammation. Each level entry includes three parts: symptoms, local signs, and examination characteristics. For example, the entry for "mild acute pharyngitis" is described in detail as "the patient subjectively feels dryness and slight stinging in the throat, with symptoms being more pronounced upon waking in the morning; examination reveals mild congestion of the posterior pharyngeal wall mucosa, no tonsil enlargement, and no secretions; body temperature characteristics show that the percentage of abnormal temperature points is less than 30%." The entry for "severe tonsillitis" is recorded as "the patient experiences severe pain when swallowing, which may radiate to the ear, accompanied by fever (body temperature)." (Temperature exceeding 38.5℃); Examination revealed grade III tonsillar enlargement, with the surface covered by yellowish-white purulent secretions, and exudate visible in the tonsillar fossa; Acoustic characteristics showed that the abnormal energy ratio of the resonance frequency band exceeded 60%; The text library is indexed hierarchically by "inflammation type-severity", such as "acute pharyngitis-mild" and "chronic pharyngitis-moderate", etc. The obtained pharyngeal inflammation level (such as "moderate chronic pharyngitis") is used as the index keyword. The corresponding entries are searched layer by layer in the text library, and all content under the entry is extracted to ensure that the description and level are completely matched, which is the inflammation level description content of the pharyngeal inflammation level.

[0138] Furthermore, a medical knowledge base is constructed, stored in a structured tabular format. Data tables are categorized by "inflammation type-severity," and each table contains four sub-items: "Daily Care," "Dietary Recommendations," "Behavioral Taboos," and "Symptom Monitoring." For example, in the "Mild Acute Pharyngitis" data table, the "Daily Care" sub-item is "Groingle your mouth twice daily with warm salt water (around 35℃), holding it in your mouth for 30 seconds each time before spitting it out." The "Dietary Recommendations" sub-item is "Avoid spicy foods such as chili peppers, ginger, and garlic; reduce the intake of excessively hot (above 60℃) drinks." The "Behavioral Taboos" sub-item is... The first item is "Avoid smoking and exposure to secondhand smoke, and avoid staying up late." The second item, "Symptom monitoring," is "Observe the frequency of sore throat daily, and record any swallowing difficulties promptly." Core information is extracted from the description of the inflammation level, such as "chronic pharyngitis" and "severe" in "severe chronic pharyngitis," as search keywords. The corresponding data table is located in the knowledge base, and all content of the four sub-items is extracted one by one. Each sub-item is checked to see if it matches the clinical characteristics of the inflammation level (e.g., for severe inflammation, "daily monitoring of body temperature changes" should be added). The information is then summarized to form the precautions for the throat inflammation level.

[0139] Furthermore, the descriptions of inflammation levels were translated, replacing technical terms with colloquial expressions. For example, "hyperplasia of lymphoid follicles on the posterior pharyngeal wall" was changed to "small granular protrusions on the inner wall of the pharynx," "vocal cord mucosal congestion" was changed to "redness on the surface of the vocal cords," and "tonsillar crypt effusion" was changed to "yellow pus in the crypts of the tonsils." The statements in the precautions section were also made more relatable, such as changing "avoid high-intensity vocal cord vibration" to "do not sing or shout loudly, and try to lower your volume when speaking," and "it is recommended to consume a liquid diet rich in vitamin C" to "drink more orange juice, pear soup, and other vitamin-rich soups or juices." The text was then updated to reflect a more general approach. The logical structure of the content is reorganized as follows: The first paragraph is "Description of the Condition," which integrates and optimizes the description of the inflammation level, explaining the specific manifestations of the current inflammation; the second paragraph is "Daily Care Methods," listing the care measures in the precautions; the third paragraph is "Dietary Recommendations," extracting relevant dietary information; the fourth paragraph is "Situations Requiring Attention," summarizing situations requiring timely medical attention during symptom monitoring (such as "pain worsening and affecting eating" or "fever lasting more than 3 days"); duplicate statements are removed (e.g., when "avoid spicy food" is mentioned in both care and diet, it is retained only once), ensuring that each paragraph is concise and clear, without any remaining technical jargon, and the final coherent text constitutes the patient's health report.

[0140] In summary, generating a health report based on the severity of throat inflammation transforms professional inflammation assessment results into intuitive and easy-to-understand content. By matching corresponding medical explanatory text, the specific meaning of the inflammation level is clearly presented, allowing patients to accurately understand their own throat condition. The precautions retrieved from the medical knowledge base based on the inflammation level are targeted and professional, providing patients with scientific advice on care, diet, and medical treatment.

[0141] In summary, the language-optimized health reports strike a balance between professionalism and readability, avoiding misunderstandings of technical jargon while ensuring the accuracy of the information. This not only enhances patients' understanding of their condition but also provides clear guidance for their subsequent health management, strengthening the practical value of monitoring results and contributing to the development of effective inflammation intervention plans.

[0142] Reference Figure 2 The diagram shown is a flowchart illustrating a method for intelligent monitoring of pharyngeal inflammation based on multi-parameter acquisition, according to an embodiment of the present invention. In this embodiment, the method for intelligent monitoring of pharyngeal inflammation based on multi-parameter acquisition includes:

[0143] S1. Simultaneously collect multi-source physiological data of the pharynx in patients; and perform noise removal on the multi-source physiological data to obtain standardized multimodal data of the multi-source physiological data;

[0144] S2. Extract the temperature anomaly region distribution features, acoustic spectrum resonance frequency band features, and mucosal texture anomaly features from the standardized multimodal data;

[0145] S3. Perform feature-level fusion on the temperature anomaly region distribution characteristics, the acoustic spectrum resonance frequency band characteristics, and the mucosal texture anomaly characteristics to obtain the fused feature vector of the pharyngeal region;

[0146] S4. Based on the fused feature vector, predict the inflammation of the patient to obtain the pharyngeal inflammation level of the patient;

[0147] S5. Generate a health report for the patient based on the level of throat inflammation.

[0148] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0149] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0150] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A smart monitoring system for pharyngeal inflammation based on multi-parameter acquisition, characterized in that, The system includes a physiological data acquisition module, a feature separation module, a physiological feature fusion module, an inflammation prediction module, and a health report generation module, wherein: The physiological data acquisition module is used to simultaneously collect multi-source physiological data of the throat area in patients; and to perform noise removal on the multi-source physiological data to obtain standardized multimodal data of the multi-source physiological data; The feature separation module is used to extract the temperature anomaly region distribution features, acoustic spectrum resonance frequency band features, and mucosal texture anomaly features from the standardized multimodal data. The physiological feature fusion module is used to perform feature-level fusion of the temperature anomaly region distribution features, the acoustic spectrum resonance frequency band features, and the mucosal texture anomaly features to obtain the fused feature vector of the pharyngeal region. The inflammation prediction module is used to predict the inflammation of the patient based on the fused feature vector to obtain the pharyngeal inflammation level of the patient. The health report generation module is used to generate a health report for the patient based on the level of throat inflammation.

2. The intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition as described in claim 1, characterized in that, When the physiological data acquisition module performs synchronous collection of multi-source physiological data from the pharynx in patients, it is specifically used for: Based on the anatomical features of the pharynx, the key areas for data collection were determined; Multi-source data collection was performed on key areas of the patient to obtain physiological data of the pharynx. The infrared temperature data, acoustic vibration data, and optical image data in the physiological data are time-stamped and synchronized to obtain multi-source physiological data of the pharynx.

3. The intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition as described in claim 2, characterized in that, When the physiological data acquisition module performs noise removal on the multi-source physiological data to obtain standardized multimodal data, it is specifically used for: Based on a reference temperature point, the influence of environmental temperature fluctuations on the temperature distribution in the multi-source physiological data is eliminated to obtain the temperature data of the pharynx. By removing environmental noise and respiratory interference components from the acoustic vibrations in the multi-source physiological data, a filtered acoustic signal of the pharyngeal region is obtained. Illumination balancing is performed on the image data from the multi-source physiological data to obtain enhanced image data of the pharyngeal region; The temperature data, the filtered acoustic signal, and the enhanced image data are dimensionally aligned to obtain standardized multimodal data of the pharyngeal region.

4. The intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition as described in claim 1, characterized in that, The feature separation module, when extracting temperature anomaly region distribution features, acoustic spectrum resonance frequency band features, and mucosal texture anomaly features from the standardized multimodal data, is specifically used for: Identify the temperature anomaly regions in the temperature distribution data of the standardized multimodal data to obtain the temperature anomaly distribution of the pharynx; The resonant frequency band energy distribution of the acoustic vibration data in the standardized multimodal data is extracted to obtain the acoustic spectrum resonance characteristics of the pharynx. Based on the changes in texture parameters of the mucosal region in the standardized multimodal data, abnormal mucosal texture features of the pharynx are generated.

5. The intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition as described in claim 1, characterized in that, When the physiological feature fusion module performs feature-level fusion of the temperature anomaly region distribution features, the acoustic spectrum resonance frequency band features, and the mucosal texture anomaly features to obtain the fused feature vector of the pharynx, it is specifically used for: The temperature anomaly region distribution characteristics, the acoustic spectrum resonance frequency band characteristics, and the mucosal texture anomaly characteristics are normalized to obtain the temperature feature vector, acoustic feature vector, and texture feature vector of the pharynx. The temperature feature vector, the acoustic feature vector, and the texture feature vector are concatenated to obtain the feature matrix of the pharyngeal region. The feature matrix is ​​weighted and fused to obtain the fused feature vector of the pharyngeal region.

6. The intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition as described in claim 5, characterized in that, The physiological feature fusion module performs weighted fusion of the feature matrix to obtain a fused feature vector for the pharyngeal region, specifically for: Calculate the attention weights of the eigenvectors in the feature matrix, wherein the formula for calculating the attention weights is as follows: ； In the formula, For the first Attention weights for each feature vector. For normalized exponential functions, For query vector, The hyperbolic tangent activation function is used. The preset weight matrix, For the first 1 eigenvector For bias vectors, This is a sign that the moment vector is inverted; The feature matrix is ​​weighted and fused based on the attention weights to obtain the fused feature vector of the throat region.

7. The intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition as described in claim 1, characterized in that, When the inflammation prediction module performs inflammation prediction on the patient based on the fused feature vector to obtain the patient's pharyngeal inflammation level, it is specifically used for: Extract the high-level feature representation of the fused feature vector to obtain the deep feature vector of the pharyngeal region; The inflammation category probability of the patient is calculated based on the deep feature vector to obtain the inflammation probability distribution of the patient; The patient's throat inflammation level is generated based on the inflammation probability distribution.

8. The intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition as described in claim 7, characterized in that, The inflammation prediction module calculates the inflammation probability distribution using the following formula: ； In the formula, For the first The i-th deep feature vector is predicted as the i-th The probability of inflammation-like symptoms. It is a natural exponential function. For the first Inflammation-like score, For the first Inflammation-like score, The ordinal number representing the degree of inflammation. This represents the total number of inflammation severity categories.

9. The intelligent monitoring system for pharyngeal inflammation based on multi-parameter acquisition as described in claim 1, characterized in that, When the health report generation module generates a health report for the patient based on the level of throat inflammation, it is specifically used for: Based on the corresponding medical explanation text matched with the level of throat inflammation, the description of the inflammation level of the throat inflammation is obtained; Based on the description of the inflammation level, a medical knowledge base was searched to obtain the precautions for the pharyngeal inflammation level. The description of the inflammation level and the precautions are processed to optimize the language, resulting in the patient's health report.

10. A method for intelligent monitoring of pharyngeal inflammation based on multi-parameter acquisition, characterized in that, The method includes: S1. Simultaneously collect multi-source physiological data of the pharynx in patients; and perform noise removal on the multi-source physiological data to obtain standardized multimodal data of the multi-source physiological data; S2. Extract the temperature anomaly region distribution features, acoustic spectrum resonance frequency band features, and mucosal texture anomaly features from the standardized multimodal data; S3. Perform feature-level fusion on the temperature anomaly region distribution characteristics, the acoustic spectrum resonance frequency band characteristics, and the mucosal texture anomaly characteristics to obtain the fused feature vector of the pharyngeal region; S4. Based on the fused feature vector, predict the inflammation of the patient to obtain the pharyngeal inflammation level of the patient; S5. Generate a health report for the patient based on the level of throat inflammation.

Images