Multi-element fusion carotid artery blood vessel monitoring method, system, electronic device and storage medium

Abstract

The application discloses a kind of multi-element fusion carotid vascular monitoring method, system, electronic equipment and storage medium, it is related to medical monitoring field.The method comprises: obtaining the time-domain acoustic signal of carotid artery of the individual to be measured, and the time-domain acoustic signal is carried out frequency domain transformation, and frequency-domain acoustic signal is obtained;Nonlinear flow state feature and topological form feature are extracted based on the frequency-domain acoustic signal;The nonlinear flow state feature and the topological form feature are compared with the corresponding baseline feature in the historical acoustic baseline of the individual to be measured, and feature deviation result is obtained;Risk assessment is carried out based on the feature deviation result, when the feature deviation result exceeds preset range, risk early warning is triggered.The application suppresses the interference such as power frequency, baseline drift, respiratory body movement and environmental noise mixed in time-domain waveform, thereby reducing the false positive rate of carotid vascular acoustic monitoring, and realizing early risk early warning of vascular lesion.

Description

Technical Field

[0001] This invention relates to the field of medical monitoring, specifically to a multi-integrated carotid artery vascular monitoring method, system, electronic device, and storage medium. Background Technology

[0002] Carotid artery diseases, such as carotid artery stenosis and atherosclerotic plaques, are one of the main causes of ischemic stroke. Currently, clinical screening mainly relies on imaging techniques such as Doppler ultrasound, CT angiography, or magnetic resonance angiography. These methods require specialized medical equipment and operators, making it difficult to achieve routine, long-term monitoring in home or community settings.

[0003] Acoustic auscultation, as a non-invasive and portable method, has the potential to achieve long-term monitoring of carotid arteries. However, existing acoustic monitoring methods typically use single time-domain or frequency-domain features, such as signal amplitude or spectral energy, and rely on population-wide thresholds as judgment criteria. This fails to consider the inherent differences between individuals in terms of body size, vascular anatomy, and basal metabolic status, leading to a high false alarm rate. Furthermore, a single feature is insufficient to fully characterize the blood flow changes caused by early vascular stenosis.

[0004] Patent document CN121313112A discloses an intelligent monitoring system for arteriovenous fistulas (AVFs). This system integrates photoplethysmography (PPG) signals and vascular sound signals, and combines an attention mechanism and a bidirectional long short-term memory (LSTM) temporal prediction model to dynamically predict the risk of AVF stenosis. However, this approach targets the specific anatomical location of AVFs, employs dual-modal acquisition of PPG and acoustic signals, and its core algorithm relies on training with a large number of labeled samples.

[0005] Therefore, how to reduce the false alarm rate of carotid artery acoustic monitoring and achieve early risk warning of vascular lesions is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] The technical problem to be solved by this invention is to reduce the false alarm rate of carotid artery acoustic monitoring and to achieve early risk warning of vascular lesions.

[0007] To address the aforementioned technical problems, in a first aspect, the present invention provides a multi-element fusion method for carotid artery vascular monitoring, comprising: acquiring the time-domain acoustic signal of the carotid artery of an individual to be tested, and performing a frequency-domain transformation on the time-domain acoustic signal to obtain a frequency-domain acoustic signal; extracting nonlinear flow characteristics and topological morphological features based on the frequency-domain acoustic signal; comparing the nonlinear flow characteristics and the topological morphological features with the corresponding baseline features in the historical acoustic baseline of the individual to be tested to obtain feature deviation results; performing a risk assessment based on the feature deviation results, and triggering a risk warning if the feature deviation results exceed a preset range.

[0008] Optionally, the method further includes: acquiring the cardiac time-domain acoustic signal of the individual to be tested; using the cardiac time-domain acoustic signal as a baseline, removing interference from the cardiac acoustic signal in the carotid artery acoustic signal of the individual to be tested.

[0009] Optionally, comparing the nonlinear flow characteristics and the topological morphology characteristics with the corresponding baseline characteristics in the historical acoustic baseline of the individual under test to obtain the feature deviation result includes: performing feature space splicing on the nonlinear flow characteristics and the topological morphology characteristics to generate a multi-dimensional fusion feature vector; calculating the deviation of the multi-dimensional fusion feature vector from the historical acoustic baseline to obtain the feature deviation result.

[0010] Optionally, frequency domain physical features are also extracted based on the frequency domain acoustic signal. The frequency domain physical features include at least one of skewness, kurtosis, zero-crossing rate, and high-low frequency peak ratio. The skewness, kurtosis, zero-crossing rate, and high-low frequency peak ratio are all calculated based on the frequency domain acoustic signal.

[0011] Optionally, the nonlinear flow characteristics include a fractal dimension, which is a numerical feature calculated based on the frequency domain acoustic signal and is used to characterize the acoustic signal complexity corresponding to changes in carotid artery flow patterns; the topological features include a symbolic distance field, which is a distance field calculated based on the frequency domain acoustic signal and characterizing the deviation distance of the frequency domain acoustic signal relative to a reference spectrum distribution.

[0012] Optionally, the method for generating the historical acoustic baseline includes: acquiring multiple sets of historical acoustic signals of the individual under stable physiological conditions, removing abnormal samples exceeding a preset signal interval, and obtaining valid historical acoustic signals; extracting historical nonlinear flow characteristics and historical topological features from the valid historical acoustic signals; and performing mean-averaging processing on the historical nonlinear flow characteristics and the historical topological features to construct the historical acoustic baseline of the individual under test.

[0013] Optionally, the step of performing a frequency domain transformation on the time-domain acoustic signal to obtain a frequency-domain acoustic signal includes: estimating the power spectral density of the time-domain acoustic signal to obtain the power spectral density as the frequency-domain acoustic signal.

[0014] Secondly, the present invention provides a multi-element fusion carotid artery vascular monitoring system, comprising: a signal acquisition module for acquiring time-domain acoustic signals of the carotid artery and heart of the individual under test, and performing frequency-domain transformation on the time-domain acoustic signals to obtain frequency-domain acoustic signals; a feature extraction module for extracting nonlinear flow characteristics and topological morphological characteristics based on the frequency-domain acoustic signals, and performing feature space splicing on the extracted features to generate a multi-element fusion feature vector; a baseline management module for constructing and storing the historical acoustic baseline of the individual under test; and a risk assessment module for calculating the deviation of the multi-element fusion feature vector from the historical acoustic baseline, performing risk assessment based on the deviation, and triggering a risk warning if the deviation exceeds a preset range.

[0015] Thirdly, the present invention provides an electronic device including a processor and a memory, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the steps of the method described in the first aspect.

[0016] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described in the first aspect above.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] 1. Since the nonlinear flow characteristics and topological features are extracted based on the frequency domain acoustic signal after frequency domain transformation, compared with extraction based directly on the time domain waveform, the influence of power frequency interference, baseline drift, respiratory and body motion artifacts and environmental noise components mixed in the time domain waveform can be suppressed, thereby more stably and accurately characterizing blood flow changes and improving the detection sensitivity of early stenosis.

[0019] 2. By comprehensively employing two complementary features—nonlinear flow characteristics and topological morphology characteristics—the fractal dimension quantifies the complexity of the flow state, and the symbolic distance field characterizes the deviation of the spectrum from the reference distribution. Both reflect the acoustic signal changes caused by early vascular stenosis from different dimensions, further improving detection sensitivity.

[0020] 3. By constructing an individual-specific historical acoustic baseline based on the individual's own historical acoustic data, and using the deviation from this baseline instead of the general threshold for risk assessment, the false alarm rate caused by differences in individual background signals is reduced.

[0021] 4. Since the time-domain acoustic signal of the heart is acquired simultaneously with the time-domain acoustic signal of the carotid artery, and the interference of the cardiac acoustic signal is removed from the carotid artery acoustic signal using the cardiac time-domain acoustic signal as a baseline, the confusion of cardiac acoustic components such as cardiac pulsation with the carotid artery blood flow characteristics is avoided, thereby further improving the accuracy and reliability of the detection. Attached Figure Description

[0022] Figure 1 A flowchart illustrating the overall process of the multi-element fusion carotid artery vascular monitoring method provided in this embodiment of the invention;

[0023] Figure 2 A flowchart for constructing an individual health acoustic baseline provided in an embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram of the multi-dimensional feature parallel extraction architecture provided in an embodiment of the present invention;

[0025] Figure 4 This is a schematic diagram of the structure of the multi-element fusion carotid artery vascular monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0026] 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 are only some embodiments of the present invention, and not all embodiments. 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.

[0027] It should be noted that in the description of this invention, the terms "first," "second," etc., are used for descriptive purposes only to distinguish similar objects, and do not imply any order of precedence, nor should they be construed as indicating or implying relative importance. Furthermore, unless otherwise stated, "a plurality of" means two or more.

[0028] Example 1: A multi-dimensional fusion method for carotid artery vascular monitoring. (See also...) Figure 1 This embodiment takes the carotid artery as the monitoring object and provides a multi-dimensional fusion carotid artery vascular monitoring method, which includes the following steps.

[0029] Step S110: Acquire the time-domain acoustic signal of the carotid artery of the individual to be tested, and perform a frequency-domain transformation on the time-domain acoustic signal to obtain a frequency-domain acoustic signal. Specifically, a digital stethoscope can be used to acquire the time-domain acoustic signals of both carotid arteries of the individual to be tested. During acquisition, the subject is in a seated position, the probe is gently placed on the carotid artery pulsation site in the neck, and the raw acoustic waveform is continuously acquired for no less than 60 seconds on each side, with a sampling rate of no less than 2000Hz. Subsequently, the acquired time-domain acoustic signal is preprocessed by filtering out 50Hz power frequency interference and high-frequency noise above 1000Hz through a bandpass filter; then, the filtered time-domain acoustic signal is transformed in the frequency domain to obtain a frequency-domain acoustic signal. The frequency-domain transformation can be a short-time Fourier transform or a power spectral density estimation; in this embodiment, the power spectral density is estimated on the preprocessed time-domain acoustic signal to obtain the power spectral density as the frequency-domain acoustic signal.

[0030] In one possible embodiment, the data collection method is as follows: the subject sits quietly for no less than 15 minutes before the data collection, and after moderate-intensity activity, the time is extended to no less than 30 minutes, and the data collection is carried out at a fixed time every day; the resting heart rate is controlled between 50 and 110 beats per minute, and the data collection time is extended if arrhythmia occurs; the subject maintains an upright sitting position, a neutral head, and natural breathing during the data collection, and the ambient noise is no higher than 40 dB.

[0031] In one possible embodiment, a standard acquisition point is located one fingerbreadth below the upper edge of the thyroid cartilage and 0.5 to 1 cm medial to the anterior border of the sternocleidomastoid muscle, i.e., 10 to 15 mm below the bifurcation of the common carotid artery. A medical marker pen can be used to mark the reference point to ensure that the acquisition across days during long-term home monitoring corresponds to the same carotid artery segment. The raw waveform is continuously acquired for no less than 120 seconds on each side, with a sampling rate of no less than 2000 Hz.

[0032] In one possible embodiment, while acquiring the time-domain acoustic signal of the carotid artery, a simultaneous time-domain acoustic signal of the heart is acquired in the precordial region of the individual under test. Since the carotid artery acoustic signal often contains cardiac acoustic components such as cardiac pulsations transmitted through the thoracic cavity and blood flow, and these cardiac acoustic components are not caused by changes in the flow regime of the carotid artery itself, the cardiac time-domain acoustic signal is used as a baseline to remove interference from the carotid artery acoustic signal of the individual under test. This yields a signal that better reflects the changes in the flow regime of the carotid artery itself, improving the accuracy of detection. It should be noted that the above-mentioned interference removal using the cardiac time-domain acoustic signal as a baseline can employ methods known in the art, such as signal separation, adaptive filtering, or spectral subtraction; this embodiment does not limit the specific implementation of these methods.

[0033] Step S120: Extract nonlinear flow characteristics and topological features based on the frequency domain acoustic signal. (See also...) Figure 3It should be noted that the original time-domain acoustic signal of the tested individual is usually superimposed with various interference components such as power frequency interference, baseline drift, respiratory and body motion artifacts, and environmental noise. These interferences intertwine with the effective vascular sound signal in the time-domain waveform. If nonlinear flow characteristics and topological morphology features are directly calculated based on the time-domain waveform, the obtained features are easily dominated by the above interferences, making it difficult to accurately characterize the nonlinear flow and topological morphology changes of blood flow. Therefore, this embodiment first performs frequency domain transformation on the time-domain acoustic signal to obtain a frequency-domain acoustic signal, and extracts nonlinear flow characteristics and topological morphology features based on the frequency-domain acoustic signal, thereby suppressing the influence of time-domain interference and more stably and accurately characterizing the early changes in blood flow.

[0034] In one possible embodiment, the nonlinear flow characteristics include a fractal dimension. The fractal dimension is calculated based on the frequency-domain acoustic signal and is used to characterize the acoustic signal complexity corresponding to changes in the carotid artery flow regime.

[0035] Studies have found that carotid artery stenosis leads to a decrease in the fractal dimension of ipsilateral blood flow acoustic signals. The fractal dimension shows a gradient decreasing trend between the healthy state, the non-stenotic side, and the stenotic side; that is, the fractal dimension in the healthy state is higher than that in the non-stenotic side, and the non-stenotic side is higher than that in the stenotic side. The more severe the stenosis, the further the fractal dimension deviates from the healthy baseline. Therefore, the fractal dimension and its deviation from an individual's healthy baseline can serve as a quantitative indicator reflecting hemodynamic disturbances caused by carotid artery stenosis.

[0036] In an exemplary embodiment, the Higuchi fractal dimension algorithm is used to analyze the frequency domain acoustic signal, i.e., the power spectral density signal, to quantify the chaotic complexity of the blood flow state. To eliminate the influence of signal length on the fractal dimension calculation, the analysis window length is standardized for each frequency domain acoustic signal to be analyzed before calculating the fractal dimension. For example, the power spectral density signal is standardized to have a length of N = 3001 sampling points.

[0037] The maximum scale parameter of the Higuchi fractal dimension algorithm can be determined through iteration: Higuchi fractal dimensions are calculated by taking values ​​one by one within a preset range, for example, k=2 to k=150. The optimal maximum scale parameter, kmax, is selected based on the stability of the fractal dimension value, ensuring that the coefficient of variation of the fractal dimension value is less than 5%. For example, kmax=100. Under this parameter, the minimum segment length is N divided by kmax and rounded up, i.e., 31 sampling points, and the maximum segment length is N minus 1, i.e., 3000 sampling points. An overlapping sliding window with a step size of 1 is used to generate a total of N minus kmax plus 1, i.e., 2902 subsequences, to ensure the completeness and stability of feature extraction. For each scale k, k subsequences are constructed, and the curve length of each subsequence is calculated and averaged to obtain the mean curve length. The least-squares linear fit is performed on the logarithm of the mean curve length and the logarithm of the reciprocal of the scale. Points with k=1 are excluded during the fitting to reduce low-frequency noise interference. The resulting slope is the Higuchi fractal dimension value.

[0038] The parameters mentioned above, such as maximum scale kmax=100, signal length N=3001, segment length range 31 to 3000, sliding step size 1, and exclusion k=1, can be preset.

[0039] The linear fitting determination coefficients for the cardiac channel, the left carotid artery channel, and the right carotid artery channel were all verified to be greater than 0.99.

[0040] In an exemplary embodiment, to verify the robustness of the Higuchi fractal dimension calculation results, box counting and detrended fluctuation analysis were further used to calculate the fractal dimension of the same frequency domain acoustic signal, and the result was compared with that obtained by the Higuchi fractal dimension algorithm. When the fractal dimensions obtained by different algorithms show a consistent trend, it indicates that the fractal dimension feature is not sensitive to algorithm selection and has good robustness.

[0041] In one exemplary embodiment, confirmatory analysis of the fractal dimension was performed based on the power spectral density signal. The sample included a healthy control group (n=10 sides), the stenotic side of a patient with carotid artery stenosis (n=14 sides), and the non-stenotic side (n=10 sides). The results showed that the Higuchi fractal dimension of the three groups exhibited a clear decreasing trend: the healthy control group (0.9831±0.0116) was higher than the non-stenotic side (0.9714±0.0184), and the non-stenotic side was higher than the stenotic side (0.9646±0.0179).

[0042] In the overall group comparison, the mean Higuchi fractal dimension of the healthy control group was 0.9831±0.0116, and its 95% reference range (mean ± 1.96 times standard deviation) was [0.9604, 1.0058].

[0043] The mean value on the stenotic side was 0.9646±0.0179, which was significantly lower than that on the healthy control group by independent samples t test (p=0.0091). The mean value on the non-stenotic side, i.e. the side opposite to the affected side, was 0.9714±0.0184, which was slightly lower than that on the healthy control group, but the difference was not statistically significant (p>0.05), and it was not statistically significant compared with the stenotic side (paired t test, p=0.4307).

[0044] To control for physiological differences in anatomical position between the left and right sides, further matching comparisons were performed by side. The left stenosis group (0.9623±0.0199) was lower than the left side of healthy individuals (0.9794±0.0159), with the difference being close but not statistically significant (p=0.1719), which may be related to the smaller sample size (n=5) in the left stenosis group. The right stenosis group (0.9583±0.0194) was significantly lower than the right side of healthy individuals (0.9868±0.0041), p=0.0124, indicating that the weakening effect of right stenosis on ipsilateral blood flow complexity was more prominent. Within patients with unilateral stenosis, the fractal dimension of the stenotic side (0.9646) was lower than that of the non-stenotic side (0.9714), but the difference was not statistically significant (p=0.4307).

[0045] The above results collectively indicate that carotid artery stenosis leads to a significant decrease in the fractal dimension of ipsilateral blood flow acoustic signals, and the magnitude of the decrease is related to the side of stenosis, with the right side showing a more significant decrease. Overall, a gradient change is observed, with the healthy control group showing a higher fractal dimension than the non-stenotic side, and the non-stenotic side showing a higher fractal dimension than the stenotic side, indirectly suggesting that the more severe the stenosis, the further the fractal dimension deviates from the healthy baseline. Therefore, fractal dimension can serve as a quantitative indicator reflecting hemodynamic disturbances caused by carotid artery stenosis.

[0046] It should be noted that Higuchi fractal dimension, box counting, detrended fluctuation analysis and its parameter selection, parameter traversal, least squares fitting, etc. are all known techniques in the field, and this embodiment does not limit their specific implementation.

[0047] In one possible embodiment, the topological morphology feature includes a symbolic distance field. This symbolic distance field is calculated based on the frequency-domain acoustic signal and characterizes the deviation of the frequency-domain acoustic signal relative to a reference spectral distribution. When the vascular morphology undergoes a physical change, the frequency-domain acoustic signal shifts relative to the reference spectral distribution, and the symbolic distance field changes accordingly, thereby capturing early, subtle changes that are difficult to detect with traditional time-frequency features.

[0048] In one exemplary embodiment, a spectral distance field analysis method based on power spectral density envelope constraint is used to evaluate the deviation between the detected carotid murmur and a reference spectral distribution, including the following processing:

[0049] The audio signal of the carotid artery murmur can be read in WAV format; the dual-channel signal is converted into a single-channel signal and the DC component is removed; the power spectral density of the preprocessed time-domain signal is estimated using the Welch method, and segmented windowing and overlapping processing are performed using the Hanning window to reduce the variance of the spectral estimation; the power spectral density curve is smoothed and the spectral information in the frequency range of 1 to 600 Hz is extracted.

[0050] A reference spectrum model is established using carotid artery sample data from the reference population. For each frequency point, the upper and lower boundaries of the power spectral density of the reference sample are statistically analyzed to construct a reference spectrum envelope, which describes the spectral distribution range of the reference population and serves as a reference benchmark for deviation detection.

[0051] The power spectral density to be analyzed is compared with the reference spectral envelope at each frequency point. When the power spectral density is above or within the boundary of the reference envelope, the distance at that frequency point is recorded as 0. When the power spectral density exceeds the upper boundary or is below the lower boundary, the distance between it and the nearest boundary is calculated as the local spectral distance at that frequency point. The spectral distance distribution along the frequency dimension is obtained by calculating all frequency points, i.e., the symbol distance field.

[0052] The symbolic distance field score is obtained by calculating the Euclidean norm of the local spectral distances at all frequency points. This score is equal to the square root of the sum of the squares of the local spectral distances at each frequency point. The power spectral density estimation, envelope construction, and distance calculation described above can all be performed using methods known in the art or their open-source implementations; this embodiment does not limit the specific implementation.

[0053] In some embodiments, frequency domain physical features can also be extracted based on the frequency domain acoustic signal as additional features for comparison. The frequency domain physical features include at least one of skewness, kurtosis, zero-crossing rate, and high-low frequency peak ratio, all calculated based on the frequency domain acoustic signal. Skewness and kurtosis reflect the degree of asymmetry and sharpness of the frequency domain acoustic signal distribution; the zero-crossing rate reflects the frequency of the frequency domain acoustic signal crossing a reference value; and the high-low frequency peak ratio is the ratio of a preset high-frequency energy peak to a preset low-frequency energy peak, reflecting the proportion of high-frequency eddy current disturbance components.

[0054] In an exemplary embodiment, the low frequency band is 50 to 200 Hz, corresponding to the physiological laminar flow frequency band; the mid frequency band is 200 to 300 Hz, corresponding to the moderate stenosis turbulence frequency band; and the high frequency band is 300 to 600 Hz, corresponding to the severe stenosis pathological frequency band; wherein 340 Hz can be used as a reference critical frequency for severe stenosis, i.e., stenosis of not less than 60%.

[0055] Step S130: Compare the nonlinear flow characteristics and the topological morphology characteristics with the corresponding baseline characteristics in the historical acoustic baseline of the individual under test to obtain the feature deviation results.

[0056] Specifically, the extracted features, namely the fractal dimension values ​​and the symbolic distance field, are concatenated in the feature space to generate a multivariate fusion feature vector Vcur. The symbolic distance field can participate in the concatenation in the form of a spectral distance distribution vector along the frequency dimension, or it can participate in the concatenation in the form of a symbolic distance field score.

[0057] In embodiments that include frequency domain physical features, these features are concatenated into the multi-element fusion feature vector. Subsequently, Vcur is compared with a pre-stored historical acoustic baseline of the individual under test, and the feature deviation D is calculated. The deviation D is equal to Vcur minus the baseline feature mean vector Vmean, then divided by the baseline feature standard deviation vector Vstd, where Vmean is the baseline feature mean vector and Vstd is the baseline feature standard deviation vector.

[0058] In some embodiments, for a symbolic distance field characterized in matrix or high-dimensional form, it can be flattened into a one-dimensional vector before participating in the above splicing and deviation calculation.

[0059] Step S140: Perform a risk assessment based on the feature deviation results. If the feature deviation results exceed a preset range, trigger a risk warning.

[0060] In an exemplary embodiment, graded warnings can be issued based on the magnitude of the deviation D according to the following rules:

[0061] When D is less than 1.0, the current acoustic characteristics are determined to be within the normal fluctuation range of the individual, and no warning is required;

[0062] When D is not less than 1.0 and less than 1.5, a Level 1 observation warning is triggered, prompting a periodic review.

[0063] When D is not less than 1.5 and less than 2.0, a level 2 warning is triggered, and it is recommended to seek medical attention for a follow-up examination in the near future.

[0064] When D is not less than 2.0 and less than 2.5, a Level 3 warning is triggered, and it is recommended to seek medical attention as soon as possible;

[0065] When D is not less than 2.5, a Level 4 high-risk warning is triggered, and immediate medical attention is recommended. The above-mentioned grading thresholds can be set based on multiples of the individual's baseline standard deviation and can be adjusted in conjunction with calibration data.

[0066] The warning level corresponding to the deviation can be set and adjusted manually based on relevant experience, standards, manuals, etc. in this field.

[0067] In other embodiments, the risk assessment may further input the multivariate fusion feature vector or the feature deviation into a pre-trained classification or regression model, which then outputs a risk level or risk probability. The classification or regression model may employ conventional machine learning models in the art, such as support vector machines, random forests, or gradient boosting trees, and may be trained and inferred using corresponding open-source implementations. This invention does not limit the specific type of the model or its training method.

[0068] See Figure 2 The process of constructing the historical acoustic baseline includes: In the initial stage of use, guiding the user to complete multiple sets of acoustic signal acquisitions under stable physiological conditions, such as sitting still, without strenuous exercise, and without acute stress. Features are extracted from each set of historical acoustic signals, and abnormal samples exceeding a preset confidence interval (e.g., exceeding the 95% confidence interval) are removed. The mean (Vmean) and standard deviation (Vstd) of the remaining valid samples are calculated to construct and store the historical acoustic baseline for the individual being tested. The baseline can be reconstructed and updated at preset intervals or after significant changes in the user's physiological state to maintain individual representativeness.

[0069] In one exemplary embodiment, the number of groups n is not less than 20, and the interval between each acquisition is not less than 24 hours. It should be noted that this invention does not limit the number of groups of historical acoustic signals. Where other methods can be used to construct an effective individual baseline with fewer groups, the number of groups can also be less than 20, which is also within the scope of this invention.

[0070] In some embodiments, the system can automatically determine the stable physiological state. Specifically, for individuals who have been sitting still for at least 15 minutes, if they simultaneously meet the following conditions: resting heart rate between 50 and 110 beats per minute, regular heart rhythm, and frequency domain acoustic signal similarity to the baseline spectral fingerprint of at least 85%, they are included in the baseline construction sample pool.

[0071] Example 2: A multi-element fusion carotid artery vascular monitoring system. (See also...) Figure 4 This embodiment provides a multi-dimensional fusion carotid artery vascular monitoring system, including a signal acquisition module, a feature extraction module, a baseline management module, and a risk assessment module.

[0072] The signal acquisition module uses a digital stethoscope as the acoustic sensor, with a built-in analog-to-digital converter and bandpass filter circuit. The passband is 20Hz to 1000Hz. The acquired time-domain acoustic signal is transmitted to the processing terminal in real time via Bluetooth or USB interface, and the time-domain acoustic signal is transformed into a frequency-domain acoustic signal. In a preferred embodiment, the stethoscope probe of the signal acquisition module uses a high-sensitivity microelectromechanical system (MEMS) microphone. The signal acquisition module is also used to acquire the cardiac time-domain acoustic signal of the individual being tested, and using the cardiac time-domain acoustic signal as a baseline, interference from the cardiac acoustic signal is removed from the carotid artery acoustic signal.

[0073] The feature extraction module runs on a processing terminal, which can be a personal computer, a smart terminal, or an embedded processor. It is used to perform the feature extraction described in step S120 of Embodiment 1, extract nonlinear flow features and topological morphology features based on the frequency domain acoustic signal, and stitch the extraction results together in the feature space to generate a multi-element fusion feature vector.

[0074] The baseline management module is used to build, store, and update individual health acoustic baselines. The baseline database supports local storage and cloud synchronization, and provides baseline quality assessment interfaces such as effective sample size and baseline variance.

[0075] The risk assessment module is used to calculate the deviation degree according to the logic described in steps S130 to S140 of Embodiment 1 and output the graded early warning result. At the same time, it can generate a multi-dimensional graphical trend report. The trend report uses the monitoring time axis as the horizontal axis and the normalized deviation degree as the vertical axis to show the fluctuation curve of the fractal dimension and the historical evolution of the sign distance field to help judge the vascular function status.

[0076] It should be understood that the method and system described in this invention are not limited to the carotid artery, and can also be applied to the functional monitoring of other arteries such as the femoral artery and brachial artery in other embodiments. This invention does not limit these applications.

[0077] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-source fusion method for carotid artery vascular monitoring, characterized in that, Includes the following steps: Acquire the time-domain acoustic signal of the carotid artery of the individual to be tested, and perform frequency domain transformation on the time-domain acoustic signal to obtain the frequency domain acoustic signal; Nonlinear flow characteristics and topological features are extracted based on the frequency domain acoustic signals; The nonlinear flow characteristics and topological morphology characteristics are compared with the corresponding baseline characteristics in the historical acoustic baseline of the individual under test to obtain the feature deviation results; Risk assessment is performed based on the deviation results of the features. If the deviation results exceed a preset range, a risk warning is triggered.

2. The method according to claim 1, characterized in that, The step of acquiring the time-domain acoustic signal of the carotid artery of the individual under test and performing a frequency-domain transformation on the time-domain acoustic signal to obtain a frequency-domain acoustic signal further includes: Acquire the time-domain acoustic signal of the heart of the individual to be tested; Using the cardiac time-domain acoustic signal as a baseline, interference from the cardiac acoustic signal in the carotid artery acoustic signal of the individual under test was eliminated.

3. The method according to claim 1, characterized in that, The step of comparing the nonlinear flow characteristics and the topological morphology characteristics with the corresponding baseline characteristics in the historical acoustic baseline of the individual under test to obtain the feature deviation results includes: The nonlinear flow characteristics and the topological morphology characteristics are spliced ​​together in feature space to generate a multi-dimensional fusion feature vector; The deviation of the multi-element fusion feature vector from the historical acoustic baseline is calculated to obtain the feature deviation result.

4. The method according to claim 1, characterized in that, Furthermore, frequency domain physical features are extracted based on the frequency domain acoustic signal. The frequency domain physical features include at least one of skewness, kurtosis, zero-crossing rate, and high-low frequency peak ratio. The skewness, kurtosis, zero-crossing rate, and high-low frequency peak ratio are all calculated based on the frequency domain acoustic signal.

5. The method according to claim 1, characterized in that, The nonlinear flow characteristics include fractal dimension, which is a numerical feature calculated based on the frequency domain acoustic signal and is used to characterize the acoustic signal complexity corresponding to changes in carotid artery flow.

6. The method according to claim 1, characterized in that, The topological features include a symbolic distance field, which is a distance field calculated based on the frequency-domain acoustic signal and characterizing the deviation of the frequency-domain acoustic signal from the reference spectrum distribution.

7. The method according to claim 1, characterized in that, The method for generating the historical acoustic baseline includes: Multiple sets of historical acoustic signals of the individual under stable physiological conditions are acquired, and abnormal samples exceeding the preset signal interval are removed to obtain the effective historical acoustic signals. Using the effective historical acoustic signals as the object, historical nonlinear flow characteristics and historical topological features are extracted; the historical nonlinear flow characteristics and historical topological features are averaged to construct the historical acoustic baseline of the individual to be tested.

8. The method according to claim 1, characterized in that, The step of performing a frequency domain transformation on the time-domain acoustic signal to obtain a frequency-domain acoustic signal includes: estimating the power spectral density of the time-domain acoustic signal to obtain the power spectral density of the frequency-domain acoustic signal.

9. A multi-source fusion carotid artery vascular monitoring system, characterized in that, include: The signal acquisition module is used to acquire time-domain acoustic signals of the carotid artery and heart of the individual under test, and to perform frequency-domain transformation on the time-domain acoustic signals to obtain frequency-domain acoustic signals. The feature extraction module is used to extract nonlinear flow features and topological features based on the frequency domain acoustic signal, and to perform feature space splicing on the extracted features to generate a multi-dimensional fusion feature vector; The baseline management module is used to build and store the historical acoustic baseline of the individual being tested; The risk assessment module is used to calculate the deviation of the multi-element fusion feature vector from the historical acoustic baseline, perform risk assessment based on the deviation, and trigger a risk warning if the deviation exceeds a preset range.

10. An electronic device comprising a processor and a memory, characterized in that, The memory stores a computer program that, when executed by the processor, implements the steps of the method as described in any one of claims 1 to 8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 8.

Images