Distal ischemia adaptive hypertension treatment and risk assessment system based on artificial intelligence

Abstract

The invention discloses a remote ischemia adaptive hypertension treatment and risk assessment system based on artificial intelligence. The system comprises a physiological signal acquisition module, a preprocessing module, an artery function assessment module, a heart rhythm analysis module, a molecular biology assessment module, a user state association module and an assessment module. The physiological signal acquisition module is responsible for executing standard distant ischemia adaptation training and acquiring original data including systolic pressure, diastolic pressure, heart rate and pulse waveform data of upper limbs on two sides in real time in the execution process; the preprocessing module can sequentially carry out band-pass filtering, waveform segmentation and feature extraction processing on original data, then feature values extracted from single waveforms and the waveforms are combined to form a data set, and the data set is output. According to the method, arteriosclerosis can be evaluated, subclavian artery stenosis can be judged, and the nitric oxide amount can be calculated synchronously, so that fusion of multiple physiological data can be ensured in an adaptive treatment process, and meanwhile, multi-data joint evaluation is realized in combination with artificial intelligence.

Description

Technical Field

[0001] This invention relates to the field of blood pressure monitoring and measurement technology, specifically to an artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system. Background Technology

[0002] Remote ischemic adaptation, as a physical therapy method, revolves around the application of brief, non-invasive pressure and release to the limbs. This precisely stimulates the body's endogenous protective mechanisms. During pressure application, local blood flow to the limb is temporarily obstructed, triggering a series of physiological changes; upon release, the body makes corresponding adjustments. This chain reaction lays the foundation for the protection of the cardiovascular system and is key to the efficacy of this treatment method.

[0003] From a therapeutic perspective, remote ischemic adaptation has a significant protective effect on the cardiovascular and cerebrovascular systems. Through the aforementioned manipulation of the limbs, a series of changes beneficial to cardiovascular and cerebrovascular health are generated within the body. These changes encompass multiple aspects, such as improving vascular endothelial function and regulating inflammatory responses. This approach, based on stimulating endogenous protective mechanisms, avoids the potential side effects of exogenous drugs, providing a safe and effective new pathway for the prevention and treatment of cardiovascular and cerebrovascular diseases, and has demonstrated significant value in clinical applications.

[0004] For example, a distal ischemic preconditioning cardiovascular therapy device and blood pressure monitoring method, authorized by China with publication number CN 118058727 B, discloses a distal ischemic preconditioning cardiovascular therapy method that can effectively train patients for ischemic preconditioning and avoid problems such as limb numbness caused by unsuitable pressure values.

[0005] However, the inventors discovered that the aforementioned distal ischemic preconditioning cardiovascular therapy device has a limited functionality in its adaptation training equipment. It can typically only execute preset pressurization / release cycles, lacks real-time and quantitative assessment of training effects, and cannot deeply integrate the training process with multi-data physiological monitoring and artificial intelligence analysis, thus reducing the overall treatment and assessment effectiveness. Summary of the Invention

[0006] To address these issues, this invention provides an artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] The AI-based remote ischemic hypertension treatment and risk assessment system includes a physiological signal acquisition module, a preprocessing module, an arterial function assessment module, a heart rhythm analysis module, a molecular assessment module, a user status association module, and an assessment module.

[0009] The physiological signal acquisition module is responsible for performing standard remote ischemia adaptation training and, during the process, real-time acquisition of raw data including systolic blood pressure, diastolic blood pressure, heart rate, and pulse wave data of both upper limbs;

[0010] The preprocessing module can sequentially perform bandpass filtering, waveform segmentation, and feature extraction on the raw data, and then combine the feature values ​​extracted from a single waveform with the waveform itself to form a dataset and output it.

[0011] The arterial function assessment module can calculate the arteriosclerosis index based on the dataset. Then, the stenosis probability value P is calculated based on the bilateral systolic blood pressure difference.

[0012] The heart rhythm analysis module can quickly identify possible abnormal heart rhythm signals based on artificial intelligence algorithm models, determine the pattern of heart rhythm changes, and determine whether there is arrhythmia.

[0013] The molecular assessment module can generate the total nitric oxide value during the calculation process based on the AI ​​computational model. ;

[0014] The user status association module can receive and store data streams output from the arterial function assessment module, heart rhythm analysis module, and molecular assessment module. Then, through timestamp alignment technology, it associates data streams from different sources on a unified timeline to form a structured health data sequence with the timeline as the axis.

[0015] The assessment module can integrate the multi-dimensional data stream output by the user status correlation module through a built-in mathematical model, and dynamically calculate the comprehensive risk value of multi-dimensional indicators. It also outputs the corresponding level response mechanism.

[0016] Furthermore, the physiological signal acquisition module also includes an execution unit, which includes a pair of independently pressurizable upper limb cuffs, a high-precision pressure sensor, a pneumatic pump, and a vent valve, for conducting remote ischemia adaptation training.

[0017] Furthermore, the bandpass filtering is used to extract pure physiological signals from noise; waveform segmentation is used to identify the start point, peak point and end point of the waveform segment corresponding to each independent heartbeat from the continuous, filtered pulse wave signal; feature extraction is used to convert the waveform into digital features.

[0018] Furthermore, the arterial function assessment module includes a first calculation unit and a screening unit;

[0019] The first calculation unit can construct a time series of pulse pressure, observe its dynamic changes during the ischemic period and the initial stage of reperfusion, and then calculate the arteriosclerosis index based on the pressure value. Arteriosclerosis Index The calculation formula is as follows:

[0020]

[0021] in, and All are calibration coefficients. Mean arterial pressure, This represents the pulse wave conduction time.

[0022] Furthermore, the screening unit can calculate the stenosis probability value P based on the bilateral systolic blood pressure difference, and determine whether the stenosis probability value P poses a risk by comparing it with a preset high-risk threshold; the formula for calculating the stenosis probability value P is as follows:

[0023] in, , and All are weighting coefficients. This represents the difference in systolic blood pressure between the left and right upper limbs. For about The function, It is about The function, It is a combined function of the percentage decrease in the slope of the rising branch (S) and the percentage increase in the roundness of the peak (R).

[0024] Furthermore, the heart rhythm analysis module includes a second calculation unit and a feature fusion unit;

[0025] The second calculation unit is used to calculate the maximum slope based on the waveform and to initially identify abnormal heart rhythm signals. The specific content is as follows:

[0026] 1) Determine the heartbeat cycle based on the time interval between two consecutive pulse wave peaks;

[0027] 2) Normalize the amplitude of each wave and calculate the maximum slope to determine the absolute irregularity of the pulse wave amplitude and the slope of the rising branch;

[0028] The feature fusion unit can extract local spatial features of pulse wave morphology with abnormal signals based on artificial intelligence algorithm models, and finally realize the identification and judgment of various premature beats through feature fusion.

[0029] Furthermore, the normalized amplitude processing involves subtracting the minimum value from the original amplitude of each pulse wave, and then dividing by the difference between the maximum and minimum values ​​to linearly map the original amplitude to the [0, 1] interval.

[0030] To calculate the maximum slope: First, discretely sample the pulse wave signals in the dataset to determine the start and end points of the rising limb of each pulse wave. Then, calculate the slope using the ratio of the amplitude difference to the time difference between two adjacent sampling points. Finally, calculate the maximum slope of the rising limb of that pulse wave. Maximum slope The calculation formula is as follows:

[0031]

[0032] in, The slope This represents the local minimum of the pulse wave signal. This represents the local maximum value of the pulse wave signal. =1, 2, ..., N.

[0033] Furthermore, the artificial intelligence algorithm model is constructed by combining convolutional neural networks and long short-term memory networks.

[0034] Furthermore, the feature fusion unit also includes a Softmax classifier, which is capable of calculating the probability value of each heart rhythm category output by the artificial intelligence algorithm model. Probability value The calculation formula is as follows:

[0035]

[0036] in, According to heart rhythm category, For the first The normalized amplitude value corresponding to the class, where N is the total number of heart rhythm classes.

[0037] Furthermore, the evaluation module also includes a hierarchical unit, which can establish a multi-level response mechanism based on percentile distribution.

[0038] The present invention has the following advantages: The present invention performs physical therapy through a physiological signal acquisition module, and combines an arterial function assessment module, a heart rate analysis module, and a molecular assessment module to simultaneously assess arteriosclerosis, determine subclavian artery stenosis, and calculate nitric oxide levels, thereby ensuring the fusion of multiple physiological data during the adaptive treatment process and solving the problem of the single function of traditional adaptive training systems.

[0039] Meanwhile, the artificial intelligence algorithm model uses a hybrid architecture of convolutional neural networks and long short-term memory networks to simultaneously analyze pulse wave morphology characteristics and rhythm timing information, enabling accurate identification of complex arrhythmias such as atrial fibrillation and premature ventricular contractions. This facilitates subsequent multi-data joint evaluation and significantly improves the effectiveness of hypertension treatment and risk assessment.

[0040] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. Attached Figure Description

[0041] To more intuitively illustrate the prior art and this application, exemplary drawings are provided below. It should be understood that the specific shapes and structures shown in the drawings should not generally be regarded as limiting conditions for implementing this application; for example, based on the technical concept disclosed in this application and the exemplary drawings, those skilled in the art are capable of making conventional adjustments or further optimizations to the addition / reduction / classification, specific shapes, positional relationships, connection methods, size ratios, etc. of certain units (components).

[0042] Figure 1 This is a flowchart illustrating the implementation of the artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system of the present invention.

[0043] Figure 2 This is an implementation architecture diagram of the molecular assessment module in the artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system of the present invention. Detailed Implementation

[0044] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these embodiments are merely for further explanation of the present invention and should not be construed as limiting the scope of protection of the present invention. Technical engineers in the field can make some non-essential improvements and adjustments to the present invention based on the above-described content. 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.

[0045] Please see Figures 1-2 The AI-based remote ischemic hypertension treatment and risk assessment system includes: a physiological signal acquisition module, a preprocessing module, an arterial function assessment module, a heart rhythm analysis module, a molecular assessment module, a user status association module, and an assessment module.

[0046] The physiological signal acquisition module is responsible for performing standard remote ischemia adaptation training. During the training, it collects raw data in real time, including systolic blood pressure, diastolic blood pressure, heart rate, and pulse wave data of both upper limbs, providing basic data for all subsequent analyses.

[0047] The physiological signal acquisition module also includes an execution unit, which comprises a pair of independently pressurizable upper limb cuffs, a high-precision pressure sensor, a pneumatic pump, and a deflation valve. The upper limb cuffs can precisely control the pressure on the left and right upper limbs separately; the high-precision pressure sensor monitors and provides feedback on pressure changes within the cuff in real time, ensuring the accuracy and safety of the pressurization process; the pneumatic pump, as the power source, provides a stable and adjustable air pressure to the cuffs to achieve the desired pressurization effect; and the deflation valve is responsible for rapidly releasing the gas from the cuffs after a preset time or pressure condition is reached, restoring blood flow to the limbs. The classic working principle of the execution unit is as follows:

[0048] Apply pressure to one or more upper limbs (usually up to 200-250 mmHg, higher than systolic blood pressure to block arterial blood flow), maintain the pressure (usually for 5 minutes), and then release the pressure (completely release and restore blood flow for 5 minutes). Repeat this cycle 4-5 times.

[0049] The preprocessing module sequentially performs bandpass filtering, waveform segmentation, and feature extraction on the raw data. It then combines the extracted feature values ​​from individual waveforms with the waveforms themselves to form a dataset and outputs it. This process removes interference from heartbeat pulse waves, slow fluctuations caused by respiration, and limb movements.

[0050] Bandpass filtering is used to extract pure physiological signals from noise; waveform segmentation is used to identify the start, peak, and end points of the waveform segment corresponding to each independent heartbeat from a continuous, filtered pulse wave signal; feature extraction is used to convert the waveform into computable digital features. The specific details are as follows:

[0051] 1) Bandpass filtering

[0052] A Butterworth filter is used to perform frequency analysis and selection on the input raw data. Based on its preset lower cutoff frequency and upper cutoff frequency, the lower cutoff frequency allows signals above its cutoff frequency to pass through, while attenuating signals below its cutoff frequency; the upper cutoff frequency allows signals below its cutoff frequency to pass through, while attenuating signals above its cutoff frequency.

[0053] After these two steps, the signal between the lower and upper cutoff frequencies (i.e., the passband signal of the bandpass filter) is output. The signal in this passband range experiences minimal attenuation during transmission, while signals at other frequencies are effectively suppressed.

[0054] 2) Waveform segmentation

[0055] The envelope of the signal is extracted using the Hilbert transform, which converts a real signal into an analytic signal. The magnitude of the analytic signal is the signal envelope. Peak values ​​within the envelope are then detected using a dynamic adaptive threshold. A peak is considered detected when the value of the envelope signal exceeds this threshold. After a peak is detected, waveform segmentation is performed based on its location. A complete waveform range is defined by extending a certain length to both sides of the peak.

[0056] For example, if the average period of a normal heartbeat is known to be T, the range of the waveform is defined by extending 0.5T to the left and right of the peak value, thereby dividing the continuous signal into individual single heartbeat pulse waves.

[0057] The dynamic adaptive threshold is calculated by multiplying the average peak amplitude detected over a recent period (e.g., 10-20 seconds) by an empirical coefficient (e.g., 0.5). This avoids the impact of signal strength fluctuations caused by changes in cuff pressure or user status.

[0058] 3) Feature extraction

[0059] From each pulse wave, feature values ​​are extracted in real time. These feature values ​​include, but are not limited to, waveform peak value (corresponding to the systolic blood pressure oscillation amplitude of that heartbeat), waveform area, rising limb slope (reflecting the heart's ejection velocity), falling limb morphology, and the time of peak appearance.

[0060] The arterial function assessment module can calculate the arteriosclerosis index based on the dataset. Then, the stenosis probability value P is calculated based on the difference in systolic blood pressure between the two sides. Next, during system initialization or periodic screening, the difference in systolic blood pressure between the left and right upper limbs is accurately calculated and recorded.

[0061] The arterial function assessment module includes a first calculation unit and a screening unit. The first calculation unit can construct a time series of pulse pressure (PP) to observe its dynamic changes during ischemia and the initial reperfusion period, and then calculate the arteriosclerosis index based on the pressure value. In order to facilitate the analysis of the dynamic response mode of PP and The changing trends of the arteriosclerosis index are used to assess vascular elasticity and quantify the degree of arteriosclerosis, providing a basis for the early detection of cardiovascular disease risk. The calculation formula is as follows:

[0062]

[0063] in, and All are calibration coefficients. Mean arterial pressure, The pulse wave conduction time. Closely related to the stiffness of the arterial wall, the stiffer the blood vessel, the faster the wave velocity. The shorter the mean arterial pressure. The calculation formula is as follows:

[0064]

[0065] in, This is the diastolic blood pressure, which is the same as the systolic blood pressure.

[0066] The screening unit calculates the stenosis probability value P based on the bilateral systolic blood pressure difference and compares it to a preset high-risk threshold to determine if there is a risk associated with stenosis. If the probability value exceeds the high-risk threshold (e.g., 90%), a strong warning of severe stenosis is issued. The formula for calculating the stenosis probability value P is as follows:

[0067] in, , and All are weighting coefficients. , This represents the difference in systolic blood pressure between the left and right upper limbs. For about The function. It is about The function. It is a combined function of the percentage decrease in the slope of the rising branch (S) and the percentage increase in the roundness of the peak (R). ,in, and All are weighting coefficients. + =1.

[0068] For example, when When <20mmHg =0; when When >20mmHg according to A linear mapping is applied to the degree exceeding the threshold. , To adjust the data.

[0069] When the screening unit is in operation, the bilateral synchronous pressure measurement mode must be activated. Both cuffs will inflate and deflate simultaneously to ensure that the measured blood pressure comes from exactly the same moment, eliminating errors caused by heart rate fluctuations and natural blood pressure variations.

[0070] The heart rhythm analysis module can quickly identify possible abnormal heart rhythm signals based on artificial intelligence algorithm models, determine the pattern of heart rhythm changes, and determine whether there are arrhythmias such as atrial fibrillation, ventricular premature beats, and atrial premature beats. This facilitates subsequent ischemic adaptation training to improve myocardial blood supply and thus assists in the treatment of the above-mentioned arrhythmias.

[0071] The heart rhythm analysis module includes a second calculation unit and a feature fusion unit. The second calculation unit is used to calculate the maximum slope based on the waveform and to initially identify abnormal heart rhythm signals. Details are as follows:

[0072] 1) Determine the heartbeat cycle based on the time interval between two consecutive pulse wave peaks;

[0073] 2) Normalize the amplitude and calculate the maximum slope for each wave to determine the absolute irregularity of the pulse wave amplitude and the slope of the rising limb. In atrial fibrillation, the ventricles become unevenly filled due to the loss of effective atrium contraction, resulting in changes in the ejection volume and velocity with each heartbeat. This variation reflects the absolute irregularity of the pulse wave amplitude and the slope of the rising limb, exhibiting highly irregular fluctuations.

[0074] Normalized amplitude: Subtract the minimum value from the original amplitude of each pulse wave, and then divide by the difference between the maximum and minimum values ​​to linearly map the original amplitude to the interval [0, 1], so that the amplitudes of different pulse waves are comparable.

[0075] Calculating the maximum slope: First, discretely sample the pulse wave signal in the dataset to determine the start and end points of the rising limb of each pulse wave. Then, calculate the slope using the ratio of the amplitude difference to the time difference between two adjacent sampling points. The slope reflects the rising velocity of the pulse wave near that point. Finally, calculate the maximum slope of the rising limb of that pulse wave. If the maximum slope If the value exceeds the threshold, an abnormal signal is output. Maximum slope. The calculation formula is as follows:

[0076]

[0077] in, The slope This represents the local minimum of the pulse wave signal. This represents the local maximum value of the pulse wave signal. =1, 2, ..., N. Slope The calculation formula is as follows:

[0078]

[0079] in, Sampling time, This corresponds to the amplitude of the pulse wave signal. The sampling time for adjacent sampling points. This represents the amplitude of the pulse wave signal corresponding to adjacent sampling points.

[0080] The feature fusion unit can extract local spatial features of pulse wave morphology with abnormal signals based on artificial intelligence algorithm models. Finally, through feature fusion, it can identify and predict the probability of normal heart rhythm, atrial fibrillation and various premature beats.

[0081] The artificial intelligence algorithm model is constructed by combining convolutional neural networks (CNNs) and long short-term memory networks (LSTMs). The CNN branches are primarily responsible for mining the spatial morphological features of pulse wave patterns containing abnormal signals. Their convolutional kernels slide sequentially across the segments, extracting local morphological features through an automatic learning mechanism. Different convolutional kernels perform different tasks; some focus on identifying waveforms with steep rising branches, while others are more sensitive to waveforms with low amplitude. These features are crucial for distinguishing between normal waves, atrial fibrillation waves, and premature beats.

[0082] The Long Short-Term Memory (LSTM) branch is used to extract pulse interval sequences from pulse wave morphologies containing abnormal signals, learn contextual information about the heart rhythm, and thus identify patterns of heart rhythm changes. Specifically, LSTM can effectively detect the process of a premature beat gradually returning to a normal rhythm after a compensatory interval, and can also accurately depict the absolutely irregular heartbeat pattern in atrial fibrillation.

[0083] The feature fusion unit also includes a Softmax classifier, which is capable of calculating the probability value of each heart rhythm category output by the artificial intelligence algorithm model. To ensure diagnostic accuracy and reduce the risk of false positives, based on a medical diagnostic threshold (e.g., 85%), only when the probability of abnormal categories such as atrial fibrillation and premature beats exceeds this preset threshold will it be considered a positive event. This effectively controls the false alarm rate. Probability value. The calculation formula is as follows:

[0084]

[0085] in, According to heart rhythm category, For the first The normalized amplitude value corresponding to the class, where N is the total number of heart rhythm classes.

[0086] The aforementioned artificial intelligence algorithm model generally refers to a deep learning model that automatically learns complex features and patterns from data by constructing a multi-layer nonlinear transformation structure to complete a specific task. It is not limited to a specific combination of architectures. For example, it can be a hybrid model that integrates convolutional neural networks (CNN) and graph neural networks (GNN). Its working principle is as follows: the CNN branch uses convolutional kernels to slide on the pulse wave segment and extracts local spatial morphological features (such as the steepness of the rising branch and amplitude features) through an automatic learning mechanism. The GNN branch constructs the pulse wave data into a graph structure to capture the topological relationships and global dependency information between different pulse waves. Finally, it achieves accurate classification and probabilistic judgment of arrhythmia through feature fusion.

[0087] When vascular endothelial cells are stimulated by blood flow shear stress, they release nitric oxide (NO). Remote ischemia adaptation training artificially creates an ischemia-reperfusion process through the compression and release of a cuff. When the cuff is released instantaneously, blood flow is rapidly restored, generating a high-intensity instantaneous shear stress peak in the artery. This physical stimulus is the key mechanism triggering the massive release of NO.

[0088] The molecular assessment module can generate a total nitric oxide value during the calculation process based on the AI ​​computing model, enabling non-invasive quantitative assessment of the impact of remote ischemic adaptation training on vascular endothelial function, thereby facilitating subsequent evaluation of the protective effect of training on cardiovascular and cerebrovascular systems.

[0089] Among them, the AI ​​computing model can calculate the amount of nitric oxide generated in all cycles of a single training session based on the data from remote ischemia adaptation training. The calculation formula is as follows:

[0090]

[0091] Where k is the calibration coefficient, and n is the number of ischemia-reperfusion cycles during the process. For the first In each training cycle, the difference between the cuff compression pressure and the estimated diastolic pressure at the moment of cuff release is recorded. For the first Effective stimulus time in each training cycle.

[0092] The user status association module can receive and store data streams from the arterial function assessment module, heart rhythm analysis module, and molecular assessment module. Then, through timestamp alignment technology, it associates data streams from different sources on a unified timeline to form a structured health data sequence with the timeline as the axis.

[0093] The timestamp alignment technology allocates a unified master clock source to all input data streams. When asynchronous data packets from each module are received, their internal timestamps are extracted first. Then, data points at equal intervals are reconstructed on a common time axis through an interpolation algorithm, ultimately generating a structured healthy data sequence in which all data are strictly synchronized.

[0094] Based on the integrated health data sequence, a sliding time window mechanism is used to aggregate and calculate feature indicators within a specific period. Association rule mining algorithms are then used to identify potential coupling relationships between different physiological data (such as the accompanying patterns of training intensity and cardiac arrhythmias). These analytical results are continuously injected into the user status sequence, generating a comprehensive view that includes real-time status assessment, historical trend comparison, and correlation factor analysis, providing data support for the evaluation module's decision-making.

[0095] The data stream includes real-time heart rhythm classification from the heart rhythm module, instantaneous blood pressure and pulse pressure from the artery module, arteriosclerosis index and subclavian artery stenosis probability from the artery function assessment module, and NO equivalent from the molecular assessment module.

[0096] The assessment module can integrate the multi-dimensional data streams output by the user status correlation module through a built-in mathematical model, and dynamically calculate the comprehensive risk value of multi-dimensional indicators. It also outputs the corresponding response level. (Comprehensive risk value) The calculation formula is as follows:

[0097]

[0098] in, , , and All are weighting coefficients.

[0099] Multidimensional indicators include the arteriosclerosis index Narrow probability value P, maximum slope exceeding the threshold (If the maximum slope) If the threshold is not exceeded, then the maximum slope is... =0) and nitric oxide value Principal component analysis was used to reduce the dimensionality of the original indicators and eliminate collinearity interference. Then, the Cox proportional hazards model or random forest algorithm was used to determine the weight coefficients of each indicator.

[0100] The evaluation module also includes a hierarchical unit, which can establish a multi-level response mechanism based on percentile distribution.

[0101] For example, a score of 0-50 corresponds to a green safety zone, 51-80 triggers a yellow alert, and a score exceeding 80 activates a red alert. Each threshold range is equipped with a differentiated intervention strategy library; for instance, a red alert automatically triggers a stroke emergency plan, while a yellow alert activates an enhanced monitoring protocol.

[0102] This invention performs physical therapy through a physiological signal acquisition module, combined with an arterial function assessment module, a heart rate analysis module, and a molecular assessment module. It can simultaneously assess arteriosclerosis, determine subclavian artery stenosis, and calculate nitric oxide levels, thereby ensuring the fusion of multiple physiological data during the adaptive treatment process and solving the problem of single-function traditional adaptive training systems.

[0103] Meanwhile, the artificial intelligence algorithm model uses a hybrid architecture of convolutional neural networks and long short-term memory networks to simultaneously analyze pulse wave morphology characteristics and rhythm timing information, enabling accurate identification of complex arrhythmias such as atrial fibrillation and premature ventricular contractions. This facilitates subsequent multi-data joint evaluation and significantly improves the effectiveness of hypertension treatment and risk assessment.

[0104] This invention also non-invasively calculates nitric oxide equivalents through a molecular assessment module. For the first time, the system provides a molecular-level quantitative basis for the improvement of vascular endothelial function in remote ischemic adaptation training, realizing an upgrade in the comprehensive evaluation dimension of training effect from macroscopic physiological data to microscopic biomolecular science.

[0105] 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. An artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system, characterized in that, It includes a physiological signal acquisition module, a preprocessing module, an arterial function assessment module, a heart rhythm analysis module, a molecular assessment module, a user status association module, and an assessment module; The physiological signal acquisition module is responsible for performing standard remote ischemia adaptation training and, during the process, real-time acquisition of raw data including systolic blood pressure, diastolic blood pressure, heart rate, and pulse wave data of both upper limbs; The preprocessing module can sequentially perform bandpass filtering, waveform segmentation, and feature extraction on the raw data, and then combine the feature values ​​extracted from a single waveform with the waveform itself to form a dataset and output it. The arterial function assessment module can calculate the arteriosclerosis index based on the dataset. Then, the stenosis probability value P is calculated based on the bilateral systolic blood pressure difference. The heart rhythm analysis module can quickly identify possible abnormal heart rhythm signals based on artificial intelligence algorithm models, determine the pattern of heart rhythm changes, and determine whether there is arrhythmia. The molecular assessment module can generate the total nitric oxide value during the calculation process based on the AI ​​computational model. ; The user status association module can receive and store data streams output from the arterial function assessment module, heart rhythm analysis module, and molecular assessment module. Then, through timestamp alignment technology, it associates data streams from different sources on a unified timeline to form a structured health data sequence with the timeline as the axis. The assessment module can integrate the multi-dimensional data stream output by the user status correlation module through a built-in mathematical model, and dynamically calculate the comprehensive risk value of multi-dimensional indicators. It also outputs the corresponding level response mechanism.

2. The artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system according to claim 1, characterized in that, The physiological signal acquisition module also includes an execution unit, which includes a pair of independently pressurizable upper limb cuffs, a high-precision pressure sensor, a pneumatic pump, and a vent valve, for remote ischemia adaptation training.

3. The artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system according to claim 1, characterized in that, The bandpass filter is used to extract pure physiological signals from noise; waveform segmentation is used to identify the start point, peak point and end point of the waveform segment corresponding to each independent heartbeat from the continuous, filtered pulse wave signal; feature extraction is used to convert the waveform into digital features.

4. The artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system according to claim 1, characterized in that, The arterial function assessment module includes a first calculation unit and a screening unit; The first calculation unit can construct a time series of pulse pressure, observe its dynamic changes during the ischemic period and the initial stage of reperfusion, and then calculate the arteriosclerosis index based on the pressure value. Arteriosclerosis Index The calculation formula is as follows: in, and All are calibration coefficients. Mean arterial pressure, This represents the pulse wave conduction time.

5. The artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system according to claim 4, characterized in that, The screening unit can calculate the stenosis probability value P based on the bilateral systolic blood pressure difference and compare it with a preset high-risk threshold to determine whether the stenosis probability value P poses a risk. The formula for calculating the stenosis probability value P is as follows: in, , and All are weighting coefficients. This represents the difference in systolic blood pressure between the left and right upper limbs. For about The function, It is about The function, It is a combined function of the percentage decrease in the slope of the rising branch (S) and the percentage increase in the roundness of the peak (R).

6. The artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system according to claim 4, characterized in that, The heart rhythm analysis module includes a second calculation unit and a feature fusion unit; The second calculation unit is used to calculate the maximum slope based on the waveform and to initially identify abnormal heart rhythm signals. The specific content is as follows: 1) Determine the heartbeat cycle based on the time interval between two consecutive pulse wave peaks; 2) Normalize the amplitude of each wave and calculate the maximum slope to determine the absolute irregularity of the pulse wave amplitude and the slope of the rising branch; The feature fusion unit can extract local spatial features of pulse wave morphology with abnormal signals based on artificial intelligence algorithm models, and finally realize the identification and judgment of various premature beats through feature fusion.

7. The artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system according to claim 6, characterized in that, The normalized amplitude processing involves subtracting the minimum value from the original amplitude of each pulse wave, then dividing by the difference between the maximum and minimum values, and linearly mapping the original amplitude to the [0, 1] interval. To calculate the maximum slope: First, discretely sample the pulse wave signals in the dataset to determine the start and end points of the rising limb of each pulse wave. Then, calculate the slope using the ratio of the amplitude difference to the time difference between two adjacent sampling points. Finally, calculate the maximum slope of the rising limb of that pulse wave. Maximum slope The calculation formula is as follows: in, The slope This represents the local minimum of the pulse wave signal. This represents the local maximum value of the pulse wave signal. =1, 2, ..., N.

8. The artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system according to claim 6, characterized in that, The artificial intelligence algorithm model is constructed by combining convolutional neural networks and long short-term memory networks.

9. The artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system according to claim 6, characterized in that, The feature fusion unit also includes a Softmax classifier, which can calculate the probability value of each heart rhythm category output by the artificial intelligence algorithm model. ; probability value The calculation formula is as follows: in, According to heart rhythm category, For the first The normalized amplitude value corresponding to the class, where N is the total number of heart rhythm classes.

10. The artificial intelligence-based remote ischemic adaptive hypertension treatment and risk assessment system according to claim 1, characterized in that, The evaluation module also includes a hierarchical unit, which can establish a multi-level response mechanism based on percentile distribution.

Images