Bronchial asthma symptom intelligent monitoring method and system based on multi-source data

Abstract

The invention discloses a bronchial asthma symptom intelligent monitoring method and system based on multi-source data, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining multi-mode asthma data after time-space alignment; performing weighted fusion on the western medicine index, the environment index, the tongue condition index and the pulse condition index by using a preset pathogenesis mapping rule to generate an asthma pathology state vector; inputting the asthma pathology state vector into an emergency response channel, analyzing abnormal change of short-term peak flow velocity, synchronously inputting into a chronic evolution channel to track long-term change of tongue condition and pulse condition, and generating emergency response integral and syndrome integral; and superposing the emergency response integral and the syndrome integral on a time axis, and triggering risk assessment when the emergency response integral and the syndrome integral meet the time condition coincidence and the peak flow rate declines to reach an early warning threshold value. According to the method, asthma pathology state vectors are respectively input into an emergency response channel and a chronic evolution channel, so that synchronous monitoring and dynamic modeling of acute attack and chronic disease courses are realized.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and system for intelligent monitoring of bronchial asthma symptoms based on multi-source data. Background Technology

[0002] Bronchial asthma is a chronic inflammatory airway disease involving multiple cells and cellular components, clinically manifested as recurrent episodes of wheezing, coughing, chest tightness, and shortness of breath. In recent years, due to multiple factors including environmental changes, lifestyle modifications, and genetic factors, the incidence of asthma has been rising globally. Traditional asthma monitoring methods primarily rely on patients' subjective symptom descriptions, pulmonary function tests (such as peak flow meters), and physicians' clinical experience. However, with the development of IoT technology, wearable devices, and artificial intelligence, automated monitoring systems based on single-source physiological signals (such as respiratory rate and blood oxygen saturation) or single environmental parameters (such as air quality index) are increasingly being applied to asthma management. These systems continuously collect relevant patient data, perform preliminary analysis using machine learning algorithms, and issue warnings when abnormalities occur, thereby achieving dynamic monitoring of the asthma condition.

[0003] However, existing technologies generally suffer from problems such as single data sources and insufficient integration of multi-dimensional pathological features, resulting in low accuracy in predicting acute asthma attacks. Specifically, traditional methods typically rely on Western medical physiological indicators (such as PEF and SpO2) while neglecting the potential correlation between macroscopic diagnostic information such as tongue and pulse findings in Traditional Chinese Medicine and environmental exposure factors, lacking the ability to comprehensively model the overall disease state. Although existing technologies attempt to extract tongue and pulse features through multimodal data acquisition, image recognition, and signal processing techniques, and combine some environmental parameters for preliminary correlation analysis, they still fail to achieve deep integration and dynamic modeling of Western and Traditional Chinese Medicine indicators and environmental factors at the pathological mechanism level. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an intelligent monitoring method for bronchial asthma symptoms based on multi-source data to solve the problem of insufficient fusion of multi-dimensional pathological features.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides an intelligent monitoring method for bronchial asthma symptoms based on multi-source data, which includes acquiring spatiotemporally aligned multimodal asthma data; the multimodal asthma data includes respiratory rate data, blood oxygen saturation data, allergen concentration, pollutant index data, tongue image and pulse signal;

[0008] Based on the respiratory rate and blood oxygen saturation data, time windows are segmented, and the respiratory rate fluctuation intensity, peak flow rate, and high-risk periods within the time windows are extracted as Western medicine indicators. Based on the allergen concentration and pollutant index data, the cumulative number of minutes of allergen exceedance and high-intensity exposure time are recorded as environmental indicators. The tongue image is input into a lightweight segmentation model, and the tongue coating thickness index and sublingual ecchymosis coverage are extracted as tongue image indicators. The pulse signal is analyzed through wavelet transform to eliminate motion artifact noise and form a four-dimensional pulse mechanical spectrum as a pulse indicator. The Western medicine indicators, environmental indicators, tongue image indicators, and pulse indicators are weighted and fused using preset pathogenesis mapping rules to generate an asthma pathological state vector.

[0009] The asthma pathological state vector is input into the emergency response channel to analyze short-term abnormal changes in peak flow rate, and simultaneously input into the chronic evolution channel to track long-term changes in tongue and pulse appearance, generating emergency response score and syndrome score; the emergency response score and syndrome score are superimposed on the time axis, and when the emergency response score and syndrome score meet the time conditions of overlap and the peak flow rate decreases to the warning threshold, risk assessment is triggered to generate acute risk level, dominant trigger type, weight contribution value and symptom severity prediction value;

[0010] Based on the risk assessment results, implement multi-level early warning responses and personalized feedback pushes.

[0011] As a preferred embodiment of the intelligent monitoring method for bronchial asthma symptoms based on multi-source data described in this invention, wherein: inputting the tongue image into a lightweight segmentation model and extracting the tongue coating thickness index and sublingual ecchymosis coverage as tongue image indicators refers to processing the tongue image using a model based on MobileNetV2 combined with U-Net architecture and extracting the tongue coating thickness index and sublingual ecchymosis coverage as tongue image indicators.

[0012] As a preferred embodiment of the intelligent monitoring method for bronchial asthma symptoms based on multi-source data described in this invention, the tongue image processing based on the MobileNetV2 combined with the U-Net architecture is as follows:

[0013] By analyzing the pixel features of tongue images through depthwise separable convolution, the tongue surface region and the sublingual region are extracted to generate multi-scale feature maps.

[0014] The multi-scale feature maps are upsampled by the U-Net decoder to assign classification labels to the tongue surface region and the sublingual region, thus distinguishing the tongue surface region, the sublingual region and the background region.

[0015] Extract the L-channel brightness value of each pixel in the tongue surface area, and map the L-channel brightness value to the tongue coating thickness level to form a tongue coating thickness distribution value matrix;

[0016] Based on the HSV saturation values ​​of pixels in the sublingual region, the boundaries of consecutive high-saturation pixels are recorded to form the boundaries of the ecchymosis connected domain.

[0017] By fusing the tongue coating thickness distribution matrix with the boundary of the ecchymosis connected domain, a two-dimensional feature map of tongue coating and ecchymosis is generated.

[0018] Extract the tongue coating thickness index and sublingual ecchymosis coverage rate from the tongue coating-ecchymosis dual-dimensional feature map;

[0019] The background area is the part of the tongue image excluding the tongue surface area and the sublingual area, distinguishing it from non-tongue areas.

[0020] As a preferred embodiment of the intelligent monitoring method for bronchial asthma symptoms based on multi-source data described in this invention, the pulse signal is analyzed using wavelet transform to eliminate motion artifact noise and form a four-dimensional pulse biomechanical spectrum as a pulse indicator, as detailed below.

[0021] The pulse signal is decomposed using discrete wavelet transform and Daubechies wavelet basis, preserving low-frequency components and removing motion artifact noise from high-frequency components.

[0022] A clean pulse signal is generated by reconstructing the time series of the pulse signal using inverse wavelet transform.

[0023] The frequency distribution, fluctuation intensity, time offset, and time evolution trend of low-frequency components are extracted from the clean pulse signal to form a four-dimensional pulse mechanical spectrum as a pulse indicator.

[0024] As a preferred embodiment of the intelligent monitoring method for bronchial asthma symptoms based on multi-source data described in this invention, the pathogenesis mapping rule is based on the TCM pathogenesis theory of asthma and the Western medicine clinical standards, mapping Western medicine indicators to airway status, environmental indicators to environmental triggers, tongue indicators to constitution imbalance, and pulse indicators to pulse characteristics.

[0025] As a preferred embodiment of the intelligent monitoring method for bronchial asthma symptoms based on multi-source data described in this invention, the step of inputting the asthma pathological state vector into the emergency response channel and analyzing short-term abnormal changes in peak flow rate refers to extracting the peak flow rate value from the asthma pathological state vector, inputting it into the emergency response channel, analyzing the short-term abnormal decrease in the peak flow rate value through a time window, and marking the emergency abnormal event to form an emergency response score.

[0026] As a preferred embodiment of the intelligent monitoring method for bronchial asthma symptoms based on multi-source data described in this invention, the synchronous input of the chronic evolution channel to track long-term changes in tongue and pulse characteristics refers to simultaneously extracting tongue and pulse indicators from the asthma pathological state vector, inputting them into the chronic evolution channel, tracking the daily variation gradient of the tongue coating thickness index and the four-dimensional pulse mechanical spectrum, and marking the onset date of deterioration to form a syndrome score.

[0027] As a preferred embodiment of the intelligent monitoring method for bronchial asthma symptoms based on multi-source data described in this invention, the step of superimposing the emergency response score and the syndrome score on the time axis means constructing a time axis with the onset date of deterioration as the origin, superimposing the emergency response score and the syndrome score on the time axis according to the time sequence, detecting the overlap of the time interval and spatial location of the emergency response score and the syndrome score, and triggering risk assessment.

[0028] As a preferred embodiment of the intelligent monitoring method for bronchial asthma symptoms based on multi-source data described in this invention, the risk assessment includes determining the dominant trigger type based on weighted contribution values ​​and determining the acute risk level based on symptom severity prediction values.

[0029] The dominant precipitating factors include emergency response-driven, syndrome-driven, and mixed-driven.

[0030] The acute risk levels are categorized as low risk, medium risk, and high risk.

[0031] Secondly, the present invention provides an intelligent monitoring system for bronchial asthma symptoms based on multi-source data, including a data acquisition module for acquiring spatiotemporally aligned multimodal asthma data; the multimodal asthma data includes respiratory rate data, blood oxygen saturation data, allergen concentration, pollutant index data, tongue image, and pulse signal;

[0032] The fusion module is used to segment time windows based on the respiratory rate data and blood oxygen saturation data, and extract the respiratory rate fluctuation intensity, peak flow rate value, and high-risk periods within the time windows as Western medicine indicators; based on the allergen concentration and pollutant index data, it counts the cumulative minutes of allergen exceedance and high-intensity exposure time records as environmental indicators; it inputs the tongue image into a lightweight segmentation model to extract the tongue coating thickness index and sublingual ecchymosis coverage as tongue image indicators; it analyzes the pulse signal through wavelet transform to eliminate motion artifact noise and form a four-dimensional pulse mechanical spectrum as a pulse image indicator; and it uses preset pathogenesis mapping rules to weightedly fuse the Western medicine indicators, environmental indicators, tongue image indicators, and pulse image indicators to generate an asthma pathological state vector.

[0033] The assessment module is used to input the asthma pathological state vector into the emergency response channel to analyze short-term abnormal changes in peak flow rate, and simultaneously input it into the chronic evolution channel to track long-term changes in tongue and pulse appearance, generating emergency response score and syndrome score; the emergency response score and syndrome score are superimposed on the time axis, and when the emergency response score and syndrome score meet the time conditions of overlap and the peak flow rate decreases to the warning threshold, risk assessment is triggered to generate acute risk level, dominant trigger type, weight contribution value and symptom severity prediction value;

[0034] The early warning module is used to execute multi-level early warning responses and personalized feedback pushes based on the risk assessment results.

[0035] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the intelligent monitoring method for bronchial asthma symptoms based on multi-source data as described in the first aspect of the present invention.

[0036] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the intelligent monitoring method for bronchial asthma symptoms based on multi-source data as described in the first aspect of the present invention.

[0037] The beneficial effects of this invention are as follows: By inputting the asthma pathological state vector into the emergency response channel and the chronic evolution channel respectively, simultaneous monitoring and dynamic modeling of acute exacerbations and chronic disease courses are achieved. The emergency response channel accurately identifies short-term airway obstruction events by analyzing the time window of peak flow rate and calculating the maximum decrease, thus improving the response speed to sudden risks. The chronic evolution channel effectively captures the long-term deterioration trend of asthma by analyzing the gradient changes of daily tongue and pulse indicators, enhancing the quantifiability and predictability of syndrome evolution. The superposition and fusion of the integrals of the two channels on the time axis makes the interaction between emergency response and syndrome evolution explicit, thereby achieving dual determination of the timing and triggering type of acute asthma exacerbation. Attached Figure Description

[0038] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart of an intelligent monitoring method for bronchial asthma symptoms based on multi-source data.

[0040] Figure 2This is a schematic diagram of an intelligent monitoring system for bronchial asthma symptoms based on multi-source data.

[0041] Figure 3 This is a flowchart for processing tongue image.

[0042] Figure 4 This is a flowchart of pulse signal processing. Detailed Implementation

[0043] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0044] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0045] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0046] Reference Figures 1-4 As an embodiment of the present invention, this embodiment provides an intelligent monitoring method for bronchial asthma symptoms based on multi-source data, comprising the following steps:

[0047] S1. Obtain spatiotemporally aligned multimodal asthma data.

[0048] Furthermore, respiratory rate data and blood oxygen saturation data are collected through medical wearable devices (multi-parameter smart wristbands), and millisecond-level timestamps are added to the respiratory rate data and blood oxygen saturation data;

[0049] Allergen concentration and pollutant index data are collected using environmental sensors, and the geographic location labels (latitude and longitude coordinates) of the allergen concentration and pollutant index data are recorded.

[0050] The tongue image is captured using a mobile application (software on a smartphone), and pulse signals are collected using a wristband sensor.

[0051] Respiratory rate data, blood oxygen saturation data, allergen concentration, pollutant index data, tongue image, and pulse signal are aligned to a unified timeline via Network Time Protocol (NTP).

[0052] A unique identifier is assigned to the aligned respiratory rate data, blood oxygen saturation data, allergen concentration, pollutant index data, tongue image, and pulse signal. Combining millisecond-level timestamps and geographic location labels (latitude and longitude coordinates), the data is sorted in ascending order by timestamp and latitude and longitude coordinates to generate a spatiotemporal correlation index. This index is stored as a key-value pair database containing timestamps, latitude and longitude coordinates, and data content. By calculating the hash values ​​of the respiratory rate data, blood oxygen saturation data, allergen concentration, pollutant index data, tongue image, and pulse signal in the key-value pair database, an integrity check tag is generated, and multimodal asthma data with integrity check tags is output.

[0053] The hash value expression is:

[0054] H = SHA-256(D);

[0055] Where H is a 256-bit hash value, used as an integrity verification mark, D is the input data, namely the respiratory rate data, blood oxygen saturation data, allergen concentration, pollutant index data, tongue image and pulse signal in the key-value pair database, and SHA-256 is the hash function.

[0056] S2. Based on respiratory rate data and blood oxygen saturation data, time windows are segmented, and the intensity of respiratory rate fluctuations, peak flow rate, and high-risk periods within the time windows are extracted as Western medicine indicators. Based on allergen concentration and pollutant index data, the cumulative number of minutes of allergen exceedance and high-intensity exposure time are recorded as environmental indicators. Tongue images are input into a lightweight segmentation model, and the tongue coating thickness index and sublingual ecchymosis coverage are extracted as tongue image indicators. Wavelet transform is used to analyze pulse signals, eliminate motion artifact noise, and form a four-dimensional pulse mechanical spectrum as a pulse image indicator. Using preset pathogenesis mapping rules, Western medicine indicators, environmental indicators, tongue image indicators, and pulse image indicators are weighted and fused to generate an asthma pathological state vector.

[0057] Furthermore, respiratory rate data is extracted from multimodal asthma data and segmented into consecutive overlapping time window sequences at fixed durations;

[0058] Waveform variability quantification is performed on respiratory rate data within a time window sequence to generate respiratory rate fluctuation intensity. Waveform variability quantification refers to calculating the standard deviation of respiratory rate data within each time window sequence as the result of waveform variability quantification. First, the mean is calculated, expressed as:

[0059]

[0060] Where μ is the mean of the respiratory rate data within the time window, n is the number of sampling points for the respiratory rate data within the time window, i is the index of the sampling point for the respiratory rate data within the time window, and x iThe respiratory rate is the respiratory rate at the i-th respiratory rate data sampling point within the time window;

[0061] Based on the mean, the standard deviation (waveform variability) is calculated using the following expression:

[0062]

[0063] Where σ is the standard deviation of respiratory rate data within the time window, i.e., the intensity of respiratory rate fluctuation;

[0064] Based on blood oxygen saturation data from multimodal asthma data, the data is divided into consecutive overlapping time windows of fixed duration. The decrease in blood oxygen saturation data within each time window is checked to see if it exceeds a preset clinical threshold. The start and end timestamps of the time windows that exceed the preset clinical threshold are recorded and marked as potential airway obstruction. The preset clinical threshold is set based on medical standards related to the decrease in blood oxygen saturation and airway obstruction in asthma patients. The preset clinical threshold is set to a decrease in blood oxygen saturation of 5% (i.e., a 5% decrease from the initial value of the time window, for example, from 98% to 93%).

[0065] Based on respiratory rate and blood oxygen saturation data, the data is divided into consecutive overlapping time windows of fixed duration. Within each time window, the respiratory rate is checked to see if it exceeds a threshold of 20 breaths per minute and the decrease in blood oxygen saturation exceeds a clinical threshold of 5%. Combinations meeting these conditions are recorded. Peak flow rate (a clinical parameter reflecting respiratory function, measured in liters per minute) is determined according to clinical mapping rules. The respiratory rate threshold is set based on asthma clinical standards, with 20 breaths per minute as the upper limit of the normal range. Values ​​above 20 breaths per minute (e.g., 21 breaths per minute or higher) are considered abnormal, while values ​​below or equal to 20 breaths per minute are considered normal. This range is applicable to monitoring asthma patients at rest. The clinical mapping rules, based on asthma clinical standards, determine the corresponding peak flow rate value by identifying combinations of respiratory rate exceeding the threshold of 20 breaths per minute and a decrease in blood oxygen saturation exceeding a clinical threshold of 5% within the time window sequence.

[0066] When the intensity of respiratory rate fluctuations within a time window exceeds the respiratory fluctuation threshold and the time window has been marked as a potential airway obstruction state, the start and end timestamps of the marked time window are recorded, and the marked time window is determined to be a high-risk period. The respiratory fluctuation threshold is set based on the standard deviation of the patient's historical respiratory rate data. The potential airway obstruction state is a time window in which the decrease in blood oxygen saturation data exceeds a preset clinical threshold by 5%. The start and end timestamps of the time window exceeding the preset clinical threshold are marked to determine the state.

[0067] The intensity of respiratory rate fluctuations, peak flow rate, and high-risk periods are directly integrated into Western medicine indicators.

[0068] Based on Western medicine indicators, the precise geographical location data of patients is recorded synchronously. Using the latitude and longitude coordinates of each geographical location data as the geometric center, a square geographical grid with a side length of 100 meters is constructed. The timestamps and data content of Western medicine indicators (including respiratory rate fluctuation intensity, peak flow rate value and high-risk periods) within the square geographical grid are recorded.

[0069] Access pollutant index data within the coverage area of ​​a square geographic grid through the cloud platform (AWS), examine the time series of pollutant index data within each square geographic grid, record the time windows in which pollutant concentrations exceed the daily average limit of 25 micrograms per cubic meter of air quality standards within multiple consecutive time windows, and mark the start and end timestamps of the time windows as high-intensity exposure time records.

[0070] Allergen concentrations within a square geographic grid coverage area are analyzed using AWS: The time series of allergen concentrations within the square geographic grid is read from the AWS database. Allergen concentration data is located based on the latitude and longitude coordinates of the square geographic grid. It is checked whether the allergen concentration in each time window exceeds the pollen warning threshold for the region. The duration of the exceeding time window is recorded, the total duration is calculated, and the cumulative number of minutes of allergen exceeding the threshold is output. The pollen warning threshold is set based on the air quality standards of the local environmental department and asthma clinical guidelines. The pollen warning threshold range is: an allergen concentration exceeding 50 grains / cubic meter is considered to trigger a pollen warning.

[0071] The high-intensity exposure time record is directly combined with the cumulative minutes of allergen exceedance to generate environmental indicators.

[0072] The tongue image is processed using a lightweight segmentation model, as follows:

[0073] First, the RGB pixel matrix of the tongue image is obtained from the mobile application. The RGB pixel matrix has a resolution of 512x512 and contains red, green, and blue channel values ​​and a timestamp. Then, the resolution of the tongue image is adjusted to 256x256. The integrity of pixel color and texture is maintained by using bilinear interpolation. Bilinear interpolation determines the new pixel value by weighted averaging the color values ​​of the four neighboring pixels around each pixel, and the RGB pixel values ​​are normalized to the range of 0-1 to adapt to the input of the lightweight segmentation model.

[0074] Then, a lightweight segmentation model is run. The lightweight segmentation model is based on MobileNetV2 combined with U-Net architecture. It processes the pixels of the tongue image through depthwise separable convolution, extracts the red tone and smooth texture features of the tongue surface region and the purple tone and vascular texture features of the sublingual region, and generates multi-scale feature maps to capture local and global information. At the same time, it records the texture features of the pixels in the tongue image, including the gray-level gradient direction and intensity difference. It compares the gray-level gradient direction and intensity difference of adjacent pixels to determine the pixel-level texture similarity and records additional information on pixel-level texture similarity.

[0075] Subsequently, the multi-scale feature map is upsampled through the U-Net decoder to restore the resolution to 256x256. A classification label is assigned to each pixel of the tongue image to distinguish the tongue surface region, the sublingual region and the background region. The pixel-level classification result is output and the classification probability of each pixel is recorded.

[0076] Based on the classification results, a mask is generated to separate the tongue surface and sublingual regions. Pixels in the tongue surface region are marked as 1, pixels in the sublingual region are marked as 0, and pixels in the background region are marked as 0, forming a 256x256 resolution binary mask matrix. Then, the pixels of the tongue image are converted from the RGB color space to the Lab color space. First, the RGB values ​​are converted to the XYZ color space, and then to the Lab color space, separating the luminance L channel and the chroma a and b channels. Finally, the luminance value of the L channel of the tongue image is adjusted, and the average luminance of the channel is standardized to the range of 50-100 through linear transformation, while keeping the a and b channels unchanged, outputting a standardized tongue image, which provides a basis for the subsequent generation of tongue coating thickness distribution values.

[0077] Based on the standardized tongue image, the masks of the tongue surface and sublingual regions, and the texture features of multi-scale feature maps, the tongue surface region is defined as the first image node, and the sublingual region as the second image node. Adjacency weights between nodes are established based on pixel-level texture similarity. Specifically: First, the tongue surface region (pixel value 1) and the sublingual region (pixel value 0) are extracted from a 256x256 resolution binary mask matrix and labeled as the first and second image nodes, respectively. Next, the grayscale gradient direction and intensity difference of the boundary pixels of the tongue surface and sublingual regions are extracted from the texture features of the multi-scale feature maps to determine the texture similarity of boundary pixel pairs. The similarity of boundary pixel pairs between the first and second image nodes is recorded as the adjacency weight. Finally, based on the Lab brightness values ​​of the tongue surface region pixels, a... Tongue coating thickness distribution matrix: In the Lab color space of the standardized tongue image, the L channel brightness value of each pixel in the tongue surface area (first image node) is extracted, and the L channel brightness value is mapped to the tongue coating thickness level (the higher the L channel brightness value, the thinner the tongue coating), forming a 256x256 resolution tongue coating thickness distribution matrix; At the same time, the boundary of the ecchymosis connected domain is located based on the HSV saturation value of the sublingual region pixels: The standardized tongue image is converted to the HSV color space (including H hue channel, S saturation channel and V brightness channel), the S channel saturation value of each pixel in the sublingual region (second image node) is extracted, and pixels with S channel saturation values ​​higher than the preset ecchymosis threshold (based on the clinical tongue image standard for asthma) are checked, and the boundary of continuous high saturation pixels is recorded to form the boundary of the ecchymosis connected domain;

[0078] The tongue coating thickness distribution matrix and the ecchymosis connected domain boundary are integrated; specifically, the tongue coating thickness distribution matrix of the tongue surface region and the ecchymosis connected domain boundary of the sublingual region are integrated into the same 256x256 resolution grid by direct superposition. The tongue coating thickness distribution matrix fills the pixels of the tongue surface region, and the ecchymosis connected domain boundary fills the pixels of the sublingual region, generating a tongue coating-ecchymosis two-dimensional feature map.

[0079] Finally, the tongue coating thickness index and sublingual ecchymosis coverage rate were extracted from the tongue coating-ecchymosis dual-dimensional feature map, specifically:

[0080] First, access the 256x256 resolution tongue coating-ecchymosis two-dimensional feature map, extract the tongue coating thickness distribution value matrix of the tongue surface region and the boundary of the ecchymosis connected domain in the sublingual region. Based on the tongue coating thickness distribution value matrix, check the tongue coating thickness level of each pixel in the tongue surface region, record the number of high thickness level pixels (thickness level greater than 3, determined according to the clinical tongue image standard for asthma), and determine the proportion of high thickness level pixels to the total number of pixels in the tongue surface region as the tongue coating thickness index. Based on the boundary of the ecchymosis connected domain, check the pixels covered by the ecchymosis boundary in the sublingual region, and record the proportion of ecchymosis boundary pixels to the total number of pixels in the sublingual region as the sublingual ecchymosis coverage rate.

[0081] Subsequently, the integrity of the tongue coating-ecchymosis two-dimensional feature map was verified through generative adversarial reconstruction: the tongue coating-ecchymosis two-dimensional feature map was reconstructed using a generative adversarial network (GAN), and the consistency of the tongue coating thickness distribution values ​​and ecchymosis connected region boundaries between the reconstructed feature map and the original feature map was compared. This confirmed that the pixel value differences in the reconstructed feature map were within a preset verification threshold (based on asthma clinical tongue image standards, such as a difference of less than 5%), thus verifying the integrity of the tongue coating-ecchymosis two-dimensional feature map. Then, the tongue coating thickness index and sublingual ecchymosis coverage were re-extracted from the reconstructed feature map, and the differences between the reconstructed feature map's tongue coating thickness index and sublingual ecchymosis coverage and the original feature map were checked. These differences were confirmed to be within a preset verification threshold, ensuring the accuracy of the original tongue coating thickness index and sublingual ecchymosis coverage. Finally, tongue image indicators, including the tongue coating thickness index and sublingual ecchymosis coverage, were output. Generative adversarial reconstruction is based on existing technologies such as generative adversarial networks (GANs), which are widely used in the field of medical imaging.

[0082] The pulse signal is analyzed using wavelet transform to eliminate motion artifact noise and generate a four-dimensional pulse dynamic spectrum as a pulse indicator. Specifically: First, the pulse signal is obtained from a mobile application; then, discrete wavelet transform is applied to decompose the pulse signal, using the Daubechies wavelet basis to decompose it into four layers of different frequency components. The pulse signal is divided into low-frequency components (e.g., 0.5-2Hz, containing the main fluctuation characteristics of the pulse signal) and high-frequency components (e.g., greater than 2Hz, containing potential motion artifact noise). The low-frequency components are retained to extract the main fluctuation characteristics of the pulse signal; then, the high-frequency components are examined, and motion artifact noise is identified based on clinical pulse criteria for asthma, marking frequencies greater than 5Hz and amplitudes exceeding the normal range. Abnormal peak values ​​that are twice the normal range (e.g., average amplitude of 2-5Hz) are identified. Using the identified motion artifact noise, the corresponding wavelet coefficients (e.g., wavelet coefficients with frequencies greater than 5Hz and amplitudes exceeding twice the average amplitude of the high-frequency components in the 2-5Hz range) are removed, retaining low-frequency components and noise-free high-frequency components as the basis for the clean pulse signal. Subsequently, the low-frequency components and noise-free high-frequency components in the clean pulse signal basis are integrated through inverse wavelet transform to reconstruct the pulse signal time series and generate the clean pulse signal. Wavelet transform is an existing technology well-known in the field of noise reduction processing. Wavelet transform includes discrete wavelet transform and inverse wavelet transform, used to denoise data and form clean data.

[0083] The frequency distribution, fluctuation intensity, time offset, and time evolution trend of low-frequency components are directly extracted from the clean pulse signal to form a 256x256 resolution four-dimensional pulse dynamic spectrum, which is then used as a pulse indicator.

[0084] Using pre-defined pathogenesis mapping rules, Western medicine indicators, environmental indicators, tongue appearance indicators, and pulse appearance indicators are weighted and fused to generate an asthma pathological state vector. The pathogenesis mapping rules refer to the rules that map Western medicine indicators to airway status, environmental indicators to environmental triggers, tongue appearance indicators to constitution imbalance, and pulse appearance indicators to pulse characteristics. These rules are based on traditional Chinese medicine pathogenesis theory of asthma and Western medical clinical standards.

[0085] S3. Input the asthma pathological state vector into the emergency response channel to analyze short-term abnormal changes in peak flow rate, and simultaneously input it into the chronic evolution channel to track long-term changes in tongue and pulse appearance, generating emergency response score and syndrome score; superimpose the emergency response score and syndrome score on the time axis. When the emergency response score and syndrome score meet the time conditions of overlap and the peak flow rate decreases to the warning threshold, trigger risk assessment to generate acute risk level, dominant trigger type, weight contribution value and symptom severity prediction value.

[0086] Furthermore, peak flow rate values ​​are extracted from the asthma pathological state vector and input into the emergency response channel, while tongue and pulse indicators are extracted and input into the chronic evolution channel.

[0087] Within the emergency response channel: the peak flow rate value is divided into continuous time windows with a duration of 5 minutes; the maximum decrease in peak flow rate value within each window is calculated; when the maximum decrease exceeds the dynamic emergency threshold set based on the patient's baseline (a decrease of 15% to 30% of the patient's baseline peak flow rate value, in liters per minute), the start and end times of the time window exceeding the dynamic emergency threshold are marked as an emergency abnormal event.

[0088] The expression for the maximum decrease is:

[0089]

[0090] Among them, A y F represents the maximum decrease within the time window W. h L represents the high-frequency component coefficients (e.g., 2-5 Hz) of the peak velocity value within the time window W after wavelet transform decomposition, where T is the duration of the time window W (e.g., 5 minutes). p I represents the initial value (liters / minute) of the peak flow rate within the time window W. p y is the endpoint value of the peak flow rate (liters / minute) within the time window W, h is the high-frequency component, p is the peak flow rate value, W is the time window, and inwindowW indicates within the time window;

[0091] It should be noted that this formula is based on signal processing techniques such as wavelet transform and time window analysis. It is used to calculate the maximum decrease in peak velocity within a time window, complete the detection of sudden abnormal events, and assess the acute risk of asthma.

[0092] Within the chronic evolution pathway: At a fixed time each day (e.g., 8 AM), the tongue coating thickness index and the four-dimensional pulse dynamic spectrum of the pulse index are extracted from the asthma pathological state vector to form a daily data sequence; the absolute value of the difference between the current day's tongue coating thickness index and the four-dimensional pulse dynamic spectrum and the previous day's tongue coating thickness index and the four-dimensional pulse dynamic spectrum is calculated as the daily variation gradient; when the daily variation gradient exceeds the preset chronic deterioration threshold (set based on the patient's baseline daily variation gradient historical data, ranging from 0.1 to 0.3, for example, 0.1 is mild deterioration, and 0.3 is significant deterioration) for three consecutive days, the current day is marked as the deterioration start day, and the three-day average daily variation gradient is recorded;

[0093] The expression for the absolute value of the difference is:

[0094] U d =|C d -C d-1 |+|B d -B d-1 |;

[0095] Among them, U d Let C be the diurnal variation gradient on day d. d C represents the tongue coating thickness index on day d. d-1 B represents the tongue coating thickness index on day d-1. d B represents the four-dimensional pulse dynamics spectrum on day d. d-1 This is the four-dimensional pulse dynamics spectrum for day d-1, where d is the index of the current day;

[0096] The duration of the start and end times (e.g., millisecond-level timestamp difference) of the sudden abnormal event in the patient's emergency response channel is extracted from the time window to form an emergency response score. The emergency response score is adjusted according to the time period in which the start and end times of the time window are located. During the daytime (e.g., 06:00 to 22:00), the emergency response score is adjusted using an adjustment factor (e.g., 1.2), and during the nighttime (e.g., 22:00 to 06:00), the emergency response score is adjusted using an adjustment factor (e.g., 1.5) to reflect the diurnal difference in asthma risk.

[0097] The following methods are used to extract the diurnal variation gradients of the onset date of deterioration and the three-day average daily variation gradient, as well as the diurnal variation gradient of the tongue coating thickness index, from the patient's chronic evolution pathway. The continuous trend of the diurnal variation gradient of the tongue coating thickness index (e.g., exceeding the chronic deterioration threshold for three consecutive days, e.g., 0.1) is determined, generating a tongue coating deterioration score. The diurnal variation gradient of the four-dimensional pulse dynamics spectrum is also extracted, and its continuous trend (e.g., exceeding the chronic deterioration threshold for three consecutive days) is determined, generating a pulse deterioration score. An adjustment factor (e.g., 1.0) is used to adjust the tongue coating deterioration score and the pulse deterioration score to form a syndrome score, reflecting the long-term deterioration trend of asthma. The adjustment factor is set based on the patient's peak flow rate, tongue coating thickness index, historical data of the diurnal variation gradient of the four-dimensional pulse dynamics spectrum (e.g., stable values ​​within three months), and the TCM pathogenesis theory of asthma. It is used to adjust the weights of the emergency response score, the tongue coating deterioration score, and the pulse deterioration score, generating a syndrome score and adjusting the emergency response score, reflecting the contribution of peak flow rate, tongue coating thickness index, and four-dimensional pulse dynamics spectrum to asthma risk.

[0098] A timeline with the onset of deterioration as the origin is constructed. The syndrome score and emergency response score are superimposed on the timeline. When the time interval between the emergency response score and the syndrome score is less than or equal to 1 hour and they coincide in spatial position on the timeline, the acute risk level, dominant cause type, weighted contribution value, and symptom severity prediction value are generated. The weighted contribution value is calculated based on the relative contribution ratio of the emergency response score and the syndrome score, and the symptom severity prediction value is obtained by comprehensively evaluating the peak flow rate decrease and daily variation gradient.

[0099] The weighted contribution value is expressed as:

[0100]

[0101] Among them, O e O is the weighted contribution value of the emergency response score. w E represents the weighted contribution value of the syndrome integral. s S represents the integral value of the emergency response (e.g., a normalized value between 0 and 1). w Here, is the syndrome score (e.g., normalized value from 0 to 1), e is the emergency weight, w is the syndrome weight, and s is the emergency quantification value.

[0102] The predictor of symptom severity is expressed as:

[0103] V r =kP a +mG g ;

[0104] Among them, V r P is a predictor of the severity of a patient's symptoms (e.g., a normalized value of 0 to 1). a G represents the maximum decrease in peak flow rate (e.g., liters per minute).g The three-day average daily variation gradient is represented by k (e.g., normalized value 0.1 to 0.3), the maximum decrease in peak flow rate is represented by k (e.g., 0.6), the three-day average daily variation gradient is represented by m (e.g., 0.4), the severity is represented by r (e.g., high risk), and the maximum decrease in peak flow rate is represented by a.

[0105] The dominant cause type is determined based on the relative magnitude of the weighted contribution value, including:

[0106] O e >O w The dominant triggering factor was emergency response.

[0107] O e <O w The dominant precipitating factor is the syndrome-dominant type;

[0108] O e =O w The dominant cause type is mixed (emergency response + syndrome);

[0109] Emergency response refers to a rapid response to sudden and urgent situations, usually involving acute events such as a sudden deterioration of a patient's symptoms (e.g., airway obstruction); syndrome refers to chronic, persistent symptoms or states, usually reflecting long-term trends or gradually changing characteristics, such as the continued deterioration of a patient's chronic disease (e.g., inflammation).

[0110] Acute risk levels are determined based on the severity of symptom severity predictive values, including:

[0111] V r ≤0.3 indicates low risk;

[0112] 0.3 <V r ≤0.6 is considered medium risk;

[0113] V r A value >0.6 indicates high risk.

[0114] Based on patients' peak flow rate values ​​and acute exacerbation records over the past three months, an individualized early warning baseline is established: For patients with no acute exacerbation records in the past three months, a yellow warning threshold is set at a maximum decrease in peak flow rate of 25% of the peak flow rate value, indicating a potential risk of symptom exacerbation; for patients with one or more acute exacerbation records, a red warning threshold is set at a maximum decrease in peak flow rate of 18%, indicating a higher risk of symptom exacerbation; On a timeline with the exacerbation onset date as the origin, emergency response scores and syndrome scores are monitored in real time to quantify the intensity of sudden acute events and chronic symptoms, respectively; When the time interval between the emergency response score and syndrome score is less than or equal to 1 hour and their timeline positions coincide, and the maximum decrease in peak flow rate reaches the yellow or red warning threshold, a risk assessment is triggered (including calculating the weighted contribution value to determine the dominant trigger type and calculating the symptom severity predictor value to determine the acute risk level).

[0115] It should be noted that the emergency response channel and the chronic evolution channel are established based on signal processing and feature extraction techniques using wavelet transform, time window analysis, and traditional Chinese medicine pathogenesis theory. By generating emergency response integrals and syndrome integrals and superimposing them on the time axis, risk prediction is driven to assess the acute risk and trigger type of asthma in real time.

[0116] S4. Based on the risk assessment results, implement multi-level early warning response and personalized feedback push.

[0117] Furthermore, based on the acute risk level, the type of dominant trigger, the weighted contribution value, and the predictive value of symptom severity, combined with an individualized early warning baseline, multi-level early warning implementation and personalized feedback push are carried out;

[0118] The multi-level early warning response is as follows:

[0119] For acute cases classified as low risk, a low-level warning signal is issued, prompting patients to maintain their existing management strategies (including daily monitoring, medication maintenance, environmental management, lifestyle adjustments, and traditional Chinese medicine treatment).

[0120] For acute cases classified as medium risk, a medium-level warning signal will be issued, and patients are advised to increase the frequency of monitoring and consult with healthcare professionals.

[0121] For acute cases classified as high risk, a high-level early warning signal will be issued, and patients and medical staff will be urgently notified to take intervention measures (including emergency medication, medical assistance, environmental isolation, symptom monitoring, and traditional Chinese medicine emergency care).

[0122] On a timeline with the onset of exacerbation as the origin, emergency response scores and syndrome scores are continuously tracked, and the timestamp, dominant trigger type, and acute risk level of each risk assessment are recorded to form a time series of asthma risk for patients.

[0123] Based on the dominant trigger type, personalized feedback is generated: if it is the emergency response, patients are advised to avoid environmental triggers and use fast-relieving medications; if it is the syndrome, patients are advised to adjust long-term control medications and optimize their lifestyle; if it is a mixed trigger, comprehensive intervention strategies (including drug combination, environment and lifestyle, high-frequency monitoring, medical consultation and TCM integration) are provided by combining emergency and syndrome recommendations.

[0124] The system pushes warning signals and feedback to patients via mobile applications, while uploading asthma risk time series data to a cloud platform for remote access and analysis by healthcare professionals.

[0125] This embodiment also provides an intelligent monitoring system for bronchial asthma symptoms based on multi-source data, including:

[0126] The acquisition module is used to acquire spatiotemporally aligned multimodal asthma data; the multimodal asthma data includes respiratory rate data, blood oxygen saturation data, allergen concentration, pollutant index data, tongue image and pulse signal;

[0127] The fusion module is used to segment time windows based on respiratory rate data and blood oxygen saturation data, and extract the respiratory rate fluctuation intensity, peak flow rate value, and high-risk periods within the time window as Western medicine indicators; based on allergen concentration and pollutant index data, it counts the cumulative minutes of allergen exceedance and high-intensity exposure time records as environmental indicators; it inputs tongue image images into a lightweight segmentation model to extract the tongue coating thickness index and sublingual ecchymosis coverage as tongue image indicators; it analyzes pulse signals through wavelet transform to eliminate motion artifact noise and form a four-dimensional pulse mechanical spectrum as a pulse indicator; and it uses preset pathogenesis mapping rules to weightedly fuse Western medicine indicators, environmental indicators, tongue image indicators, and pulse indicators to generate an asthma pathological state vector.

[0128] The assessment module is used to input the asthma pathological state vector into the emergency response channel to analyze short-term abnormal changes in peak flow rate, and simultaneously input it into the chronic evolution channel to track long-term changes in tongue and pulse appearance, generating emergency response score and syndrome score; the emergency response score and syndrome score are superimposed on the time axis. When the emergency response score and syndrome score meet the time conditions of overlap and the peak flow rate decreases to the warning threshold, risk assessment is triggered, generating acute risk level, dominant trigger type, weight contribution value and symptom severity prediction value;

[0129] The early warning module is used to execute multi-level early warning responses and personalized feedback pushes based on risk assessment results.

[0130] This embodiment also provides a computer device applicable to the intelligent monitoring method for bronchial asthma symptoms based on multi-source data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the intelligent monitoring method for bronchial asthma symptoms based on multi-source data as proposed in the above embodiment.

[0131] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0132] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the intelligent monitoring method for bronchial asthma symptoms based on multi-source data as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0133] In summary, this invention achieves simultaneous monitoring and dynamic modeling of acute exacerbations and chronic disease progression by inputting the asthma pathological state vector into the emergency response channel and the chronic evolution channel, respectively. The emergency response channel accurately identifies short-term airway obstruction events and improves the response speed to sudden risks by analyzing the time window of peak flow rate and calculating the maximum decrease. The chronic evolution channel effectively captures the long-term deterioration trend of asthma by analyzing the gradient changes of daily tongue and pulse indicators, enhancing the quantifiability and predictability of syndrome evolution. The superposition and fusion of the integrals of the two channels on the time axis makes the interaction between emergency response and syndrome evolution explicit, thereby achieving dual determination of the timing and triggering type of acute asthma exacerbations.

[0134] 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, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent monitoring of bronchial asthma symptoms based on multi-source data, characterized in that: include, Acquire spatiotemporally aligned multimodal asthma data; the multimodal asthma data includes respiratory rate data, blood oxygen saturation data, allergen concentration, pollutant index data, tongue image, and pulse signal; Based on the respiratory rate and blood oxygen saturation data, time windows are segmented, and the respiratory rate fluctuation intensity, peak flow rate, and high-risk periods within the time windows are extracted as Western medicine indicators. Based on the allergen concentration and pollutant index data, the cumulative number of minutes of allergen exceedance and high-intensity exposure time are recorded as environmental indicators. The tongue image is input into a lightweight segmentation model, and the tongue coating thickness index and sublingual ecchymosis coverage are extracted as tongue image indicators. The pulse signal is analyzed through wavelet transform to eliminate motion artifact noise and form a four-dimensional pulse mechanical spectrum as a pulse indicator. The Western medicine indicators, environmental indicators, tongue image indicators, and pulse indicators are weighted and fused using preset pathogenesis mapping rules to generate an asthma pathological state vector. The asthma pathological state vector is input into the emergency response channel to analyze short-term abnormal changes in peak flow rate, and simultaneously input into the chronic evolution channel to track long-term changes in tongue and pulse appearance, generating emergency response score and syndrome score; the emergency response score and syndrome score are superimposed on the time axis, and when the emergency response score and syndrome score meet the time conditions of overlap and the peak flow rate decreases to the warning threshold, risk assessment is triggered to generate acute risk level, dominant trigger type, weight contribution value and symptom severity prediction value; Based on the risk assessment results, implement multi-level early warning responses and personalized feedback pushes.

2. The intelligent monitoring method for bronchial asthma symptoms based on multi-source data as described in claim 1, characterized in that: The step of inputting the tongue image into a lightweight segmentation model and extracting the tongue coating thickness index and sublingual ecchymosis coverage as tongue image indicators refers to processing the tongue image based on a model using MobileNetV2 combined with the U-Net architecture to extract the tongue coating thickness index and sublingual ecchymosis coverage as tongue image indicators.

3. The intelligent monitoring method for bronchial asthma symptoms based on multi-source data as described in claim 2, characterized in that: The model based on MobileNetV2 combined with the U-Net architecture processes tongue images as follows: By analyzing the pixel features of tongue images through depthwise separable convolution, the tongue surface region and the sublingual region are extracted to generate multi-scale feature maps. The multi-scale feature maps are upsampled by the U-Net decoder to assign classification labels to the tongue surface region and the sublingual region, thus distinguishing the tongue surface region, the sublingual region and the background region. Extract the L-channel brightness value of each pixel in the tongue surface area, and map the L-channel brightness value to the tongue coating thickness level to form a tongue coating thickness distribution value matrix; Based on the HSV saturation values ​​of pixels in the sublingual region, the boundaries of consecutive high-saturation pixels are recorded to form the boundaries of the ecchymosis connected domain. By fusing the tongue coating thickness distribution matrix with the boundary of the ecchymosis connected domain, a two-dimensional feature map of tongue coating and ecchymosis is generated. Extract the tongue coating thickness index and sublingual ecchymosis coverage rate from the tongue coating-ecchymosis dual-dimensional feature map; The background area is the part of the tongue image excluding the tongue surface area and the sublingual area, distinguishing it from non-tongue areas.

4. The intelligent monitoring method for bronchial asthma symptoms based on multi-source data as described in claim 1, characterized in that: The pulse signal is analyzed using wavelet transform to eliminate motion artifact noise and form a four-dimensional pulse dynamic spectrum as a pulse indicator, as detailed below. The pulse signal is decomposed using discrete wavelet transform and Daubechies wavelet basis, preserving low-frequency components and removing motion artifact noise from high-frequency components. A clean pulse signal is generated by reconstructing the time series of the pulse signal using inverse wavelet transform. The frequency distribution, fluctuation intensity, time offset, and time evolution trend of low-frequency components are extracted from the clean pulse signal to form a four-dimensional pulse mechanical spectrum as a pulse indicator.

5. The intelligent monitoring method for bronchial asthma symptoms based on multi-source data as described in claim 1, characterized in that: The pathogenesis mapping rule is based on the TCM pathogenesis theory of asthma and the clinical standards of Western medicine, mapping Western medicine indicators to airway status, environmental indicators to environmental triggers, tongue indicators to constitution imbalance, and pulse indicators to pulse characteristics.

6. The intelligent monitoring method for bronchial asthma symptoms based on multi-source data as described in claim 1, characterized in that: The process of inputting the asthma pathological state vector into the emergency response channel and analyzing short-term abnormal changes in peak flow rate refers to extracting the peak flow rate value from the asthma pathological state vector, inputting it into the emergency response channel, analyzing the short-term abnormal decrease in the peak flow rate value through a time window, and marking the emergency abnormal event to form an emergency response score.

7. The intelligent monitoring method for bronchial asthma symptoms based on multi-source data as described in claim 1, characterized in that: The synchronous input chronic evolution channel for tracking long-term changes in tongue and pulse characteristics refers to simultaneously extracting tongue and pulse indicators from the asthma pathological state vector, inputting them into the chronic evolution channel, tracking the daily variation gradient of the tongue coating thickness index and the four-dimensional pulse mechanical spectrum, and marking the onset date of deterioration to form the syndrome score.

8. The intelligent monitoring method for bronchial asthma symptoms based on multi-source data as described in claim 1, characterized in that: The superposition of the emergency response score and the syndrome score on the time axis refers to constructing a time axis with the onset date of deterioration as the origin, superimposing the emergency response score and the syndrome score on the time axis according to the time sequence, detecting the overlap of the time interval and spatial location of the emergency response score and the syndrome score, and triggering risk assessment.

9. The intelligent monitoring method for bronchial asthma symptoms based on multi-source data as described in claim 1, characterized in that: The risk assessment includes determining the dominant trigger type based on weighted contribution values; and determining the acute risk level based on symptom severity predictive values. The dominant precipitating factors include emergency response-driven, syndrome-driven, and mixed-driven. The acute risk levels are categorized as low risk, medium risk, and high risk.

10. A multi-source data-based intelligent monitoring system for bronchial asthma symptoms, based on the multi-source data-based intelligent monitoring method for bronchial asthma symptoms according to any one of claims 1 to 9, characterized in that: include, The acquisition module is used to acquire spatiotemporally aligned multimodal asthma data; the multimodal asthma data includes respiratory rate data, blood oxygen saturation data, allergen concentration, pollutant index data, tongue image, and pulse signal; The fusion module is used to segment time windows based on the respiratory rate data and blood oxygen saturation data, and extract the respiratory rate fluctuation intensity, peak flow rate value, and high-risk periods within the time windows as Western medicine indicators; based on the allergen concentration and pollutant index data, it counts the cumulative minutes of allergen exceedance and high-intensity exposure time records as environmental indicators; it inputs the tongue image into a lightweight segmentation model to extract the tongue coating thickness index and sublingual ecchymosis coverage as tongue image indicators; it analyzes the pulse signal through wavelet transform to eliminate motion artifact noise and form a four-dimensional pulse mechanical spectrum as a pulse image indicator; and it uses preset pathogenesis mapping rules to weightedly fuse the Western medicine indicators, environmental indicators, tongue image indicators, and pulse image indicators to generate an asthma pathological state vector. The assessment module is used to input the asthma pathological state vector into the emergency response channel to analyze short-term abnormal changes in peak flow rate, and simultaneously input it into the chronic evolution channel to track long-term changes in tongue and pulse appearance, generating emergency response score and syndrome score; the emergency response score and syndrome score are superimposed on the time axis, and when the emergency response score and syndrome score meet the time conditions of overlap and the peak flow rate decreases to the warning threshold, risk assessment is triggered to generate acute risk level, dominant trigger type, weight contribution value and symptom severity prediction value; The early warning module is used to execute multi-level early warning responses and personalized feedback pushes based on the risk assessment results.

Images