A data big data analysis method and system for non-invasive cardiac output monitoring

Abstract

The application discloses a data big data analysis method and system for noninvasive cardiac output monitoring, which comprises the following steps: collecting radial artery multi-channel signals through a sensor array signal processing platform, processing and screening effective signals, extracting features through a pulse wave transmission time analysis algorithm, and constructing a time series data set; then, a radial artery tension waveform analysis model is used to decompose characteristic wave groups to obtain parameters, blood vessel elasticity dynamic compensation model is combined to adjust compensation coefficients to obtain blood vessel elasticity related parameters; data is integrated and input into a big data analysis module, multi-feature fusion processing is performed to establish correlation mapping and iterative optimization, and monitoring results are output, and each key step comprises detailed sub-steps to ensure processing precision. The system corresponds to the method and is provided with six units. The method and the system solve the problems of loose connection between signal processing and analysis in the prior art, dependence on a single parameter and lack of dynamic compensation, improve monitoring precision, and meet clinical requirements.

Description

Technical Field

[0001] This invention relates to the field of cardiac output monitoring and analysis technology, and in particular to a big data analysis method and system for non-invasive cardiac output monitoring. Background Technology

[0002] In medical monitoring, cardiac output monitoring is a crucial means of assessing cardiac function and guiding clinical treatment, and the safety and convenience of its monitoring methods are of great concern. Traditional invasive cardiac output monitoring requires invasive procedures to obtain physiological data, which may cause complications such as infection and bleeding, and is not suitable for long-term continuous monitoring or vulnerable populations. With the development of sensor technology and big data analysis technology, non-invasive cardiac output monitoring has become a research hotspot. Collecting peripheral vascular physiological signals through sensors and combining them with algorithm analysis to achieve cardiac output monitoring has become an important direction. Currently, radial artery signals are often used as a signal source for non-invasive cardiac output monitoring because they are easy to collect and closely related to cardiac function. At the same time, the analysis of parameters such as pulse wave transit time and vascular elasticity, as well as the application of sensor array signal processing platforms, provide technical support for improving the accuracy of non-invasive cardiac output monitoring, and promote the development of non-invasive cardiac output monitoring towards a more efficient and safer direction to meet the clinical needs for real-time and accurate cardiac function assessment.

[0003] Current non-invasive cardiac output monitoring technologies still have significant shortcomings in practical applications. On the one hand, existing technologies for analyzing radial artery signals are mostly limited to single parameters or simple models, failing to fully integrate multi-dimensional parameters such as pulse wave propagation time and dynamic changes in vascular elasticity. Furthermore, they lack dynamic compensation mechanisms tailored to different individuals' basic physiological information, leading to potential biases in monitoring results when individual physiological states change, making it difficult to adapt to the monitoring needs of different populations. On the other hand, the signal processing and big data analysis stages of existing technologies are not tightly integrated. The filtering and feature extraction processes for multi-channel signals acquired by the sensor array fail to effectively address the specific needs of non-invasive cardiac output monitoring. Simultaneously, the big data analysis module lacks sufficient ability to fuse multi-source data, failing to fully explore the correlations between data points. This results in poor stability of the established mapping relationship between data and cardiac output monitoring indicators, making it difficult to consistently output high-precision monitoring results. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of existing technologies, this invention provides a data big data analysis method and system for non-invasive cardiac output monitoring.

[0005] The technical solution adopted in this invention is a big data analysis method for non-invasive cardiac output monitoring, comprising the following steps: S1, acquiring multi-channel physiological signals of the radial artery region through a sensor array signal processing platform, which performs multi-dimensional filtering and signal enhancement processing on the acquired signals to screen out effective signals that meet the requirements of non-invasive cardiac output monitoring data; S2, extracting features from the effective signals acquired in S1 based on a pulse wave conduction time analysis algorithm to determine the conduction time difference of the pulse wave between different monitoring points and constructing a pulse wave conduction time series dataset; S3, using a radial artery tension waveform analytical model to perform waveform decomposition on the pulse wave conduction time series dataset obtained in S2, separating the systolic wave, diastolic wave, and dicrotic wave features in the radial artery tension waveform. S3 involves obtaining morphological and temporal parameters of different characteristic wave components; S4 uses a vascular elasticity dynamic compensation model to dynamically compensate the parameters of different characteristic wave components extracted in S3, and adjusts the compensation coefficients based on the basic physiological information of the monitored subjects to obtain the compensated vascular elasticity correlation parameters; S5 integrates the compensated vascular elasticity correlation parameters from S4 with the pulse wave conduction time series data from S2, and inputs them into the big data analysis module for non-invasive cardiac output monitoring. This module uses a multi-feature fusion algorithm to perform hierarchical processing on the integrated data and establish a correlation mapping between the data and non-invasive cardiac output monitoring indicators; S6 iteratively optimizes the correlation mapping established in S5 through the big data analysis module, and outputs the final monitoring result data that meets the accuracy requirements of non-invasive cardiac output monitoring.

[0006] Furthermore, the expression used in the pulse wave propagation time analysis algorithm is as follows: ,in, The pulse wave conduction time. The time when the pulse wave feature point is collected at the i-th monitoring point. For the first The time when pulse wave characteristic points were collected at each monitoring point The total number of monitoring points. Let be the signal strength weighting coefficient for the i-th monitoring point. Let be the position correction coefficient for the i-th monitoring point. Let be the signal stability coefficient of the i-th monitoring point. Let be the environmental interference compensation coefficient for the i-th monitoring point.

[0007] Furthermore, the expression used in the analytical model of the radial artery tension waveform is: ,in, The tension value of the radial artery at position r and time t. The amplitude coefficient of the contraction wave. The position attenuation coefficient, Radial position coordinates, The coordinates of the radial artery center are: The pulse wave angular frequency, For time, This represents the phase shift of the contraction phase wave. The amplitude coefficient of the diastolic wave. The time decay coefficient, The start of diastole Radial phase coefficient, This represents the phase shift of the diastolic wave.

[0008] Furthermore, the expression used in the vascular elasticity dynamic compensation model is: ,in, The compensated vascular elastic modulus, The original value of vascular elastic modulus before compensation. The number of samples is calculated for dynamic compensation. Let be the compensation coefficient for the j-th sampling. The blood pressure value is from the j-th sample. To set the average blood pressure value over a given time period, Let j be the heart rate value from the j-th sample. To set the average heart rate value over a given time period, is the time decay factor for the j-th sampling.

[0009] Furthermore, the signal processing expression of the sensor array signal processing platform is as follows: ,in, For the platform to output valid signals, This refers to the number of channels in the sensor array. Let p be the original input signal of the p-th channel. Let p be the signal gain coefficient of the p-th channel. Let be the interference intensity coefficient of the k-th type of interference signal in the p-th channel. This is the suppression coefficient for the k-th type of interference in the p-th channel.

[0010] Furthermore, the multi-feature fusion algorithm expression of the data big data analysis module for non-invasive cardiac output monitoring is as follows: ,in, This is the result of non-invasive cardiac output monitoring. The number of feature categories, The number of parameters for each type of feature. Let be the value of the y-th parameter in the x-th feature class. Let y be the weight coefficient of the y-th parameter of the x-th feature. Let y be the normalized coefficient of the y-th parameter of the x-th feature. Let y be the error correction coefficient for the y-th parameter of the x-th feature. Let y be the stability coefficient of the y-th parameter of the x-th feature.

[0011] Further, step S3 includes the following sub-steps: S31, performing signal segmentation processing on the pulse wave conduction time series dataset obtained in S2, dividing the entire time series into multiple independent pulse wave period segments according to the periodic characteristics of the pulse wave, with each period segment containing a complete pulse wave signal; S32, for each divided pulse wave period segment, using a wavelet transform algorithm to decompose the signal into multiple scales, obtaining wavelet coefficients at different scales, and filtering out signal components containing radial artery tension waveform calibration features through wavelet coefficients; S33, reconstructing the filtered signal components to remove noise interference introduced during the decomposition process, obtaining a relatively smooth original contour of the radial artery tension waveform; S34, setting a characteristic wave recognition threshold based on the physiological characteristics of the radial artery tension waveform, separating the systolic wave, diastolic wave, and dicrotic wave through threshold comparison and waveform slope analysis, while recording the start time, peak time, and trough time time parameters of different characteristic waves, and measuring the peak amplitude and wave width morphological parameters of different characteristic waves.

[0012] Further, S4 includes the following sub-steps: S41, collecting basic physiological information of the monitored subject, including age, height, weight, gender, and history of underlying diseases, and inputting the data into the parameter configuration module of the vascular elasticity dynamic compensation model. This module assigns corresponding initial compensation coefficients to different basic physiological information according to preset mapping rules; S42, extracting calibration parameters related to vascular elasticity from the different characteristic wave component parameters obtained in S3, such as the vascular wall stress value corresponding to the peak value during systole and the vascular wall strain value corresponding to the trough value during diastole, and establishing the correlation between the calibration parameters and the initial compensation coefficients; S43, dynamically adjusting the initial compensation coefficients based on the real-time physiological data changes of the monitored subject during the monitoring process, such as heart rate fluctuations and blood pressure changes, calculating the compensation coefficient deviation value after each adjustment, and re-optimizing the adjustment strategy if the deviation value exceeds the set range; S44, substituting the adjusted compensation coefficients into the vascular elasticity dynamic compensation model, performing compensation calculations on the extracted vascular elasticity-related calibration parameters, generating the compensated vascular elasticity modulus and vascular compliance parameters, and storing them in the data cache module.

[0013] Further, S5 includes the following sub-steps: S51, retrieving the compensated vascular elasticity correlation parameters output by S4 and the pulse wave conduction time series data generated by S2 from the data storage unit, and performing time axis alignment processing on the two types of data to ensure that each set of data corresponds one-to-one in the time dimension and eliminate time synchronization deviation; S52, performing feature standardization processing on the aligned dataset, converting the values ​​of different types of parameters into a unified data range to avoid interference to subsequent analysis due to differences in parameter magnitudes, and marking outliers in the dataset and storing them separately; S53, inputting the standardized dataset into a multi-feature fusion algorithm, which first performs feature layering on the data, dividing the data into a time feature layer, a morphological feature layer, and a physiological correlation feature layer, and then performs feature filtering within different layers to retain feature parameters that have a significant impact on non-invasive cardiac output monitoring indicators; S54, establishing a mathematical mapping relationship between different feature parameters and non-invasive cardiac output monitoring indicators through cross-correlation calculation between feature layers, generating a preliminary correlation mapping model, and performing a preliminary evaluation of the model's fit.

[0014] A big data analysis system for non-invasive cardiac output monitoring, comprising: a multi-channel radial artery signal acquisition unit connected to a sensor array signal processing platform, for acquiring multi-channel physiological signals from the radial artery region and transmitting the acquired signals to the sensor array signal processing platform; a signal filtering and enhancement unit integrated within the sensor array signal processing platform, receiving signals transmitted from the multi-channel radial artery signal acquisition unit, performing multi-dimensional filtering and enhancement processing on the signals, filtering out effective signals, and transmitting the effective signals to a pulse wave transit time feature extraction unit; and a pulse wave transit time feature extraction unit connected to both the signal filtering and enhancement unit and the radial artery tension waveform analysis unit, for extracting features from the effective signals based on a pulse wave transit time analysis algorithm. The system obtains a pulse wave conduction time series dataset and transmits it to the radial artery tension waveform analysis unit. The radial artery tension waveform analysis unit, connected to the vascular elasticity dynamic compensation unit, uses the radial artery tension waveform analysis model to decompose the pulse wave conduction time series dataset, obtains parameters of different characteristic wave components, and transmits them to the vascular elasticity dynamic compensation unit. The vascular elasticity dynamic compensation unit, connected to the big data analysis and integration unit, uses the vascular elasticity dynamic compensation model to dynamically compensate the characteristic wave component parameters, obtains the compensated vascular elasticity correlation parameters, and transmits them to the big data analysis and integration unit. The analysis and result output unit, connected to the vascular elasticity dynamic compensation unit, receives the compensated vascular elasticity correlation parameters and pulse wave conduction time series data, establishes a correlation mapping through a multi-feature fusion algorithm, iteratively optimizes it, and outputs the final monitoring result data.

[0015] Beneficial Effects: This invention proposes a data big data analysis method and system for non-invasive cardiac output monitoring. It acquires multi-channel radial artery signals through a sensor array signal processing platform and performs targeted filtering and enhancement, solving the problems of loose integration between signal processing and big data analysis in existing technologies, and the lack of integration of multi-channel signal processing with monitoring requirements. This provides high-quality and effective signals for subsequent analysis. Features are extracted and a time-series dataset is constructed using a pulse wave transit time analysis algorithm. Multi-dimensional parameters are obtained by decomposing characteristic wave components using a radial artery tension waveform analytical model. The compensation coefficient is then adjusted using a vascular elasticity dynamic compensation model combined with individual baseline physiological information. This breaks through the limitations of existing technologies, which are limited to single parameters or simple models and lack dynamic compensation mechanisms. It can adapt to changes in the physiological state of different populations and reduce monitoring result deviations. The big data analysis module integrates multi-source data, and a multi-feature fusion algorithm is used to establish correlation mappings and iteratively optimize them, improving the multi-source data fusion capability, fully exploring data correlations, enhancing the stability of the mapping relationship between data and cardiac output monitoring indicators, and achieving continuous output of high-precision monitoring results. Simultaneously, the entire process requires no invasive operation, balancing safety and convenience, and meeting the clinical needs for real-time and accurate cardiac function assessment. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method steps of the present invention;

[0017] Figure 2 This is a diagram showing the system unit composition of the present invention. Detailed Implementation

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] like Figure 1 As shown, a big data analysis method for non-invasive cardiac output monitoring includes the following steps:

[0020] S1, multi-channel physiological signals of the radial artery region are collected through a sensor array signal processing platform. The platform performs multi-dimensional filtering and signal enhancement processing on the collected signals to select effective signals that meet the requirements of non-invasive cardiac output monitoring data.

[0021] Specifically, step S1 is implemented as follows: First, the sensor array signal processing platform is started. This platform is equipped with an array of 16 pressure sensors with a sensor spacing of 0.5 mm, covering a 2 cm × 2 cm area where the radial artery pulsation is most pronounced. The sampling frequency is set to 1000 Hz to ensure that high-frequency pulse wave details can be captured. After the platform is started, the sensor array is placed close to the radial artery at the subject's wrist, and multi-channel physiological signals are collected through pressure sensing. The initial collection time is set to 30 seconds to obtain a basic signal library. Subsequently, the platform performs multi-dimensional filtering processing on the collected raw signals. The low-pass filter cutoff frequency is set to 50 Hz to filter out high-frequency electromagnetic interference; the high-pass filter cutoff frequency is set to 0.5 Hz to remove low-frequency baseline drift; at the same time, an adaptive filtering algorithm is enabled to adjust the filtering parameters in real time to cope with the interference caused by the subject's limb micro-movements. After filtering, the signal-to-noise ratio of the effective signal is improved through a signal enhancement algorithm. Channel signals with a signal-to-noise ratio below 20 dB are marked as invalid and removed. Finally, at least 8 effective channels are selected. These effective signals must meet the requirements of no significant signal interruption within 5 consecutive seconds and signal amplitude fluctuation within 5%. This provides a high-quality data foundation for subsequent analysis. This step, through precise signal acquisition and processing, avoids interference in the original signal from misleading subsequent feature extraction and ensures the reliability of the data source for the entire monitoring process.

[0022] S2, based on the pulse wave propagation time analysis algorithm, the effective signal obtained in S1 is feature extracted to determine the propagation time difference of the pulse wave between different monitoring points and to construct a pulse wave propagation time series dataset;

[0023] Specifically, step S2 is implemented as follows: Based on a preset pulse wave propagation time analysis algorithm, feature point identification is first performed on the effective signals selected in S1. The algorithm first locates the rising edge inflection point (i.e., the start of systole) and falling edge inflection point (i.e., the start of diastole) of the pulse wave in each effective channel signal, with the identification accuracy controlled within 1 millisecond. For each effective channel, the timestamps of the rising edge inflection points within 10 consecutive pulse cycles are calculated. Then, two adjacent effective channels (such as channel 1 and channel 2, channel 3 and channel 4, etc.) are selected, and the difference between the corresponding rising edge inflection point timestamps is calculated to obtain preliminary pulse wave propagation time data. Subsequently, statistical analysis is performed on the preliminary propagation time data of each channel combination, removing outliers exceeding the average ± 3 times the standard deviation, and then averaging the remaining data to obtain the pulse wave propagation time value of that channel combination. This process is repeated to obtain the propagation time values ​​of all effective channel combinations, constructing a pulse wave propagation time series dataset containing channel number, propagation time value, and corresponding pulse cycle number. The dataset must contain at least 20 consecutive pulse cycles of effective data. This step involves accurately extracting pulse wave conduction time features to establish a quantified time series dataset, providing a crucial time dimension reference for subsequent analysis of radial artery tension waveforms. Simultaneously, cross-validation of multi-channel data enhances the accuracy of conduction time data and avoids analytical errors caused by single-channel signal bias.

[0024] S3 uses the radial artery tension waveform analytical model to decompose the pulse wave conduction time series dataset obtained in S2, and separates the systolic wave, diastolic wave and dicrotic wave characteristic wave components in the radial artery tension waveform, and obtains the morphological parameters and time parameters of different characteristic wave components.

[0025] Specifically, step S3 is implemented as follows: Using a radial artery tension waveform analysis model, the pulse wave conduction time series dataset constructed in S2 is first split according to the pulse cycle. Each pulse cycle corresponds to a set of conduction time data and the corresponding original signal segment. The model first performs waveform smoothing on the original signal segment of each pulse cycle using a moving average method, with a window size of 5 sampling points to reduce the impact of signal noise on waveform decomposition. Then, according to waveform characteristics, each smoothed signal segment is decomposed into three characteristic wave components: systolic wave, diastolic wave, and dicrotic wave. The systolic wave is defined as the waveform portion from the rising edge inflection point to the signal peak. Peak identification is determined by finding the time point corresponding to the maximum signal amplitude, with an accuracy controlled within 0.5 milliseconds. The diastolic wave is defined as the waveform portion from the peak to the next rising edge inflection point. The dicrotic wave is defined as a secondary small peak waveform appearing in the diastolic wave, which must meet the condition that its amplitude reaches more than 30% of the systolic wave peak value. After decomposition, morphological and temporal parameters of each characteristic wave component are extracted. Morphological parameters include the peak amplitude of the systolic wave, the trough amplitude of the diastolic wave, and the peak amplitude of the dicrotic wave, with measurement precision controlled to 0.1 mV. Temporal parameters include the duration of the systolic wave (time from the rising edge inflection point to the peak), the duration of the diastolic wave (time from the peak to the next rising edge inflection point), and the time to the appearance of the dicrotic wave (time from the peak to the peak of the dicrotic wave), with measurement precision controlled to 1 ms. This step, through refined waveform decomposition and parameter extraction, transforms the complex radial artery tension waveform into quantifiable characteristic parameters, providing direct waveform feature basis for subsequent analysis of vascular elasticity. Simultaneously, by clearly defining the characteristic waves and measurement standards, the consistency and comparability of the parameters are ensured.

[0026] S4. The vascular elasticity dynamic compensation model is used to dynamically compensate the different characteristic wave component parameters extracted in S3. The compensation coefficient is adjusted in combination with the basic physiological information of the monitored object to obtain the compensated vascular elasticity correlation parameters.

[0027] Specifically, step S4 is implemented as follows: Using a dynamic compensation model for vascular elasticity, the basic physiological information of the subject is first collected, including age (accurate to 1 year), height (accurate to 1 cm), weight (accurate to 0.5 kg), gender, and whether the subject has underlying diseases such as hypertension or diabetes. This information is then input into the model's parameter configuration module. The module assigns initial compensation coefficients to different basic physiological information according to preset mapping rules. For example, the initial compensation coefficient for healthy men aged 20-30 is set to 1.0, and the initial compensation coefficient for hypertensive patients over 60 years old is set to 1.2. The adjustment step size for the compensation coefficient is set to 0.05. Subsequently, from the characteristic wave component parameters extracted in S3, three key parameters directly related to vascular elasticity are selected: the peak amplitude of the systolic wave, the trough amplitude of the diastolic wave, and the duration of the systolic wave. The correlation between these parameters and the initial compensation coefficient is established, meaning that the change in the value of each key parameter corresponds to a specific adjustment ratio of the compensation coefficient. During monitoring, the subject's heart rate (sampling frequency set to 1 Hz) and non-invasive blood pressure (measured every 30 seconds, with an accuracy of ±2 mmHg) are collected in real time. When heart rate fluctuations exceed 5 beats / minute or blood pressure fluctuations exceed 10 mmHg, the model automatically adjusts the compensation coefficient according to the fluctuation amplitude. For example, when the heart rate increases by 10 beats / minute, the compensation coefficient increases by 0.1 from the initial value; when the blood pressure decreases by 10 mmHg, the compensation coefficient decreases by 0.08. After adjustment, the compensation coefficient is substituted into the model calculation formula to calculate the compensation for the selected key vascular elasticity parameters, obtaining the compensated vascular elastic modulus, vascular compliance, and other parameters. These parameters need to be stored in the data cache and updated every 10 seconds to ensure the dynamic timeliness of the parameters. This step, by combining individual differences of the subject with real-time physiological state to adjust the compensation coefficient, eliminates the influence of individual physiological characteristics and real-time state changes on vascular elasticity parameters, improves the accuracy of the parameters, and provides reliable vascular status data support for the accurate calculation of subsequent cardiac output monitoring results.

[0028] S5 integrates the compensated vascular elasticity correlation parameters in S4 with the pulse wave conduction time series data in S2, and inputs them into the big data analysis module for non-invasive cardiac output monitoring. This module uses a multi-feature fusion algorithm to perform hierarchical processing on the integrated data and establishes a correlation mapping between the data and non-invasive cardiac output monitoring indicators.

[0029] Specifically, step S5 is implemented as follows: First, the latest compensated vascular elasticity-related parameters (including vascular elastic modulus and vascular compliance) output from S4 and the pulse wave conduction time series data generated in S2 are retrieved from the data storage unit. The time dimension of the two types of data is unified through a timestamp alignment algorithm to ensure that each set of vascular elasticity parameters can correspond to the pulse wave conduction time data within a specific time period, with the time synchronization error controlled within 50 milliseconds. After alignment, the dataset is subjected to feature standardization processing. The Min-Max standardization method is used to uniformly transform the values ​​of vascular elastic modulus (original range is usually 1000-5000 kPa) and pulse wave conduction time (original range is usually 100-300 milliseconds) to the range of 0-1. At the same time, outliers in the dataset are identified through the 3σ criterion, that is, data exceeding the mean ± 3 times the standard deviation are marked as outliers and stored separately in the outlier data area, and do not participate in subsequent fusion analysis. The standardized dataset is input into a multi-feature fusion algorithm. The algorithm first divides the data into three layers: a time feature layer (containing pulse wave transit time and time parameters of various feature waves), a morphological feature layer (containing amplitude parameters of various feature waves), and a physiological correlation feature layer (containing vascular elasticity parameters, heart rate, and blood pressure). A feature importance assessment threshold (set to 0.6) is defined for each layer. The algorithm uses a random forest to calculate the feature importance score of each parameter, discarding parameters with scores below the threshold and retaining key parameters such as pulse wave transit time, systolic peak amplitude, and vascular elastic modulus. Subsequently, cross-correlation calculations are performed between layers. For example, the Pearson correlation coefficient between pulse wave transit time in the time feature layer and vascular elastic modulus in the physiological correlation feature layer is calculated. Based on the correlation coefficient, a mathematical mapping relationship is established between each feature parameter and non-invasive cardiac output monitoring indicators (cardiac output, cardiac index), generating a preliminary correlation mapping model. The model fit must reach 0.85 or higher; if it does not, the feature selection threshold and correlation calculation method are readjusted. This step establishes a quantitative correlation between feature parameters and cardiac output monitoring indicators through data integration, standardization, and multi-feature fusion, laying the model foundation for subsequent output of accurate monitoring results. At the same time, outlier handling and feature screening improve the stability and reliability of the model.

[0030] S6 iteratively optimizes the correlation mapping established in S5 through the big data analysis module, and outputs the final monitoring result data that meets the accuracy requirements of non-invasive cardiac output monitoring.

[0031] Specifically, the implementation process of step S6 is as follows: The big data analysis module calls the preliminary correlation mapping model established in S5. First, it inputs the integrated data of the latest 10 consecutive pulse cycles (including compensated vascular elasticity parameters and pulse wave conduction time data). The model outputs the corresponding non-invasive cardiac output monitoring results (cardiac output value). This result is compared with the preset reference range (the normal range of cardiac output in adults at rest is 4-8 liters / minute) to determine whether it is within a reasonable range. If the result exceeds the reference range, the module automatically starts the iterative optimization process. First, it analyzes the reason for the excess. If it is due to abnormal feature parameters (such as excessive fluctuation in pulse wave conduction time), it returns to S2 to re-extract the conduction time data for that time period. If it is due to model parameter deviation, it adjusts the feature weight coefficient in the multi-feature fusion algorithm by 10% of the initial weight and recalculates the correlation mapping model. After optimization, the same data is input again for verification. If the output result still exceeds the reference range, the iteration process is repeated. The upper limit of the number of iterations is set to 5. If the result is still abnormal after 5 iterations, the module issues a data abnormality prompt. If the results are within a reasonable range, the module continues to input the integrated data for the next 10 pulse cycles, repeating the above verification and optimization process. Simultaneously, the coefficient of variation (COP) of 20 consecutive monitoring results is calculated. When the COP is below 5%, the model is considered to have reached a stable state, and the final non-invasive cardiac output monitoring results are output. The results must include cardiac output, cardiac index, and the COP of the monitoring results, stored in the results database and simultaneously displayed on the monitoring interface. This step ensures the stability and accuracy of the correlation mapping model through iterative optimization, avoiding errors in monitoring results caused by single data bias or model parameter errors, ultimately outputting high-precision and stable monitoring results that meet clinical requirements for the reliability of cardiac output monitoring results.

[0032] Preferably, the expression used in the pulse wave propagation time analysis algorithm is: ,in, The pulse wave conduction time. The time when the pulse wave feature point is collected at the i-th monitoring point. For the first The time when pulse wave characteristic points were collected at each monitoring point The total number of monitoring points. Let be the signal strength weighting coefficient for the i-th monitoring point. Let be the position correction coefficient for the i-th monitoring point. Let be the signal stability coefficient of the i-th monitoring point. Let be the environmental interference compensation coefficient for the i-th monitoring point.

[0033] Specifically, the implementation process of the pulse wave propagation time analysis algorithm is as follows: First, determine the total number of monitoring points. Based on the number of channels in the sensor array, typically 8-12 effective monitoring points are selected to ensure data representativeness. When collecting pulse wave characteristic point times at each monitoring point, signal feature recognition technology is used to accurately capture the data, with the time recording accuracy controlled within 0.1 milliseconds. The signal strength weighting coefficient is determined based on the signal-to-noise ratio (SNR) of each monitoring point. The coefficient is set to 0.9-1.0 for monitoring points with a high SNR (above 30 dB) and 0.6-0.8 for those with a low SNR (20-30 dB). The position correction coefficient is adjusted based on the distance between the monitoring point and the center of the radial artery. The coefficient is set to 1.0 for distances within 0.5 mm from the center, and decreases by 0.1 for every 0.5 mm increase in distance, with the range controlled between 0.7 and 1.0. The signal stability coefficient is calculated based on the fluctuation amplitude of 10 consecutive pulse cycles. A fluctuation amplitude of less than 3% is set to 0.9-1.0, and a fluctuation amplitude of 3%-5% is set to 0.7-0.8. The environmental interference compensation coefficient is set according to the electromagnetic interference intensity in the monitoring environment. For environments with low interference intensity, a coefficient of 0.95-1.0 is set, and for environments with moderate interference intensity, a coefficient of 0.85-0.95 is set. The calculation involves first multiplying the time difference of each monitoring point by its corresponding coefficient and then summing the results. Then, the sum of the products of the stability coefficient and the interference compensation coefficient for each monitoring point is calculated. Finally, the pulse wave conduction time is obtained by dividing the sum of the two products. Through the precise setting and calculation of multi-dimensional coefficients, the influence of monitoring point location differences, signal strength variations, and environmental interference on the conduction time calculation is eliminated, improving the accuracy of pulse wave conduction time data and providing reliable time dimension parameters for subsequent cardiac output monitoring.

[0034] Preferably, the expression used in the analytical model of the radial artery tension waveform is: ,in, The tension value of the radial artery at position r and time t. The amplitude coefficient of the contraction wave. The position attenuation coefficient, Radial position coordinates, The coordinates of the radial artery center are: The pulse wave angular frequency, For time, This represents the phase shift of the contraction phase wave. The amplitude coefficient of the diastolic wave. The time decay coefficient, The start of diastole Radial phase coefficient, This represents the phase shift of the diastolic wave.

[0035] Specifically, the implementation process of the radial artery tension waveform analytical model is as follows: The systolic wave amplitude coefficient is determined based on the pressure change amplitude during radial artery systole. In healthy adults at rest, it is typically set to 0.8-1.2, while in children or the elderly, it is adjusted to 0.6-1.0 based on vascular elasticity. The positional attenuation coefficient is set based on the distance between the radial position and the center of the radial artery. For every 0.1 mm increase in distance from the center, the coefficient increases by 0.05, with a range controlled between 0.1-0.5 to ensure accurate reflection of the tension attenuation pattern at different positions. The pulse wave angular frequency is calculated based on the subject's heart rate. For a heart rate of 60-80 beats / minute, it is set to 6.28-8.37 radians / second. For every 10 beats / minute deviation in heart rate, the angular frequency increases or decreases by 1.05 radians / second. The systolic wave phase offset is adjusted based on the time difference between the signal acquisition start time and the systolic start point. A time difference of 0-50 milliseconds corresponds to an offset of 0-0.314 radians. The diastolic wave amplitude coefficient is typically 0.4-0.6 times that of the systolic wave amplitude coefficient, adjusted according to changes in vascular elasticity. The time decay coefficient is set based on the duration of diastole, set to 0.8-1.2 when the diastolic duration is 0.4-0.6 seconds. The radial phase coefficient is set based on the elastic characteristics of the radial artery wall, set to 0.5-0.8 for vessels with good elasticity and 1.0-1.3 for vessels with poor elasticity. The diastolic wave phase offset and the systolic wave phase offset are maintained at a difference of 0.523-0.785 radians. During calculation, the synergistic effect of various parameters accurately simulates the changes in radial artery tension at different locations and times. Through dynamic adjustment of multiple parameters, a refined analysis of the radial artery tension waveform is achieved, accurately separating each characteristic wave component, providing detailed waveform data support for subsequent extraction of vascular elasticity-related parameters.

[0036] Preferably, the expression used in the vascular elasticity dynamic compensation model is: ,in, The compensated vascular elastic modulus, The original value of vascular elastic modulus before compensation. The number of samples is calculated for dynamic compensation. Let be the compensation coefficient for the j-th sampling. The blood pressure value is from the j-th sample. To set the average blood pressure value over a given time period, Let j be the heart rate value from the j-th sample. To set the average heart rate value over a given time period, is the time decay factor for the j-th sampling.

[0037] Specifically, the implementation process of the vascular elasticity dynamic compensation model is as follows: The original value of the vascular elastic modulus before compensation is calculated using the characteristic wave parameters extracted by S3. In healthy adults at rest, this is typically 1500-3000 kPa, and needs to be updated in real time according to the actual waveform parameters of the subject. The sampling number for dynamic compensation calculation is set to 10-20 times to ensure the statistical representativeness of the sampled data. The compensation coefficient is determined based on the subject's basic physiological information: 0.9-1.1 for healthy individuals aged 20-40, 1.1-1.3 for individuals aged 40-60, and 1.3-1.5 for individuals over 60 or with underlying diseases, with an adjustment of 0.05 every 5 years. Blood pressure values ​​are collected every 30 seconds using a non-invasive blood pressure monitoring module. The average blood pressure value is the average of 5 consecutive collections, and the calculation must ensure that the accuracy of each collection is within ±2 mmHg. Heart rate values ​​are collected in real time using a heart rate monitoring module at a sampling frequency of 1 Hz. The average heart rate value is the average of 1 minute of consecutive collections. The time decay factor is set according to the sampling interval. It is set to 0.98-1.0 for a 5-second sampling interval, and decreases by 0.02 for every 5-second increase in the interval, with the range controlled between 0.9 and 1.0. The calculation first calculates the relative rate of change between the blood pressure difference and the heart rate difference for each sampling, then multiplies this by the corresponding compensation coefficient and the time decay factor, sums the results, and finally multiplies this by the original elastic modulus to obtain the compensated elastic modulus. By combining the subject's real-time physiological state and baseline information to adjust the compensation parameters, the influence of individual differences and real-time physiological fluctuations on vascular elasticity parameters is effectively eliminated, improving the accuracy of vascular elasticity data and laying the foundation for accurate calculation of subsequent cardiac output monitoring results.

[0038] Preferably, the signal processing expression of the sensor array signal processing platform is: ,in, For the platform to output valid signals, This refers to the number of channels in the sensor array. Let p be the original input signal of the p-th channel. Let p be the signal gain coefficient of the p-th channel. Let be the interference intensity coefficient of the k-th type of interference signal in the p-th channel. This is the suppression coefficient for the k-th type of interference in the p-th channel.

[0039] Specifically, the signal processing implementation process of the sensor array signal processing platform is as follows: The number of channels in the sensor array is set to 16-32 according to monitoring requirements to ensure coverage of the key area of ​​the radial artery and avoid signal redundancy; the raw input signal of each channel is acquired through a pressure sensor, with the signal amplitude controlled within the range of 0.1-5 mV. During acquisition, it is necessary to ensure good contact between the sensor and the skin to avoid signal distortion caused by poor contact. The signal gain coefficient is adjusted according to the raw signal amplitude: channels with amplitudes less than 0.5 mV are set to 1.5-2.0, those with amplitudes of 0.5-2 mV are set to 1.0-1.5, and those with amplitudes greater than 2 mV are set to 0.8-1.0, ensuring that the signal amplitude of each channel is within the ideal analysis range of 0.5-3 mV after adjustment. Interference signals are categorized into three types: electromagnetic interference, limb micro-motion interference, and power frequency interference. The interference intensity coefficient for each type is detected in real time by the interference monitoring module. The coefficients are set to 0.1-0.3 for low interference intensity, 0.3-0.5 for medium intensity, and 0.5-0.8 for high intensity. The corresponding suppression coefficients are adjusted according to the interference type: 0.8-1.0 for electromagnetic interference, 0.7-0.9 for limb micro-motion interference, and 0.9-1.0 for power frequency interference. The calculation first multiplies the original signal of each channel by the gain coefficient, then multiplies by the product of each type of interference (1 - interference intensity coefficient × suppression coefficient). Finally, the results for all channels are summed to obtain the effective output signal. Through the collaborative processing of multi-channel signals and targeted interference suppression, the signal-to-noise ratio and quality of the output signal are significantly improved, effectively filtering out various interference signals and providing a high-quality signal source for subsequent feature extraction and analysis, thus avoiding the misleading influence of interference signals on subsequent processes.

[0040] Preferably, the multi-feature fusion algorithm expression of the data big data analysis module for non-invasive cardiac output monitoring is as follows: ,in, This is the result of non-invasive cardiac output monitoring. The number of feature categories, The number of parameters for each type of feature. Let be the value of the y-th parameter in the x-th feature class. Let y be the weight coefficient of the y-th parameter of the x-th feature. Let y be the normalized coefficient of the y-th parameter of the x-th feature. Let y be the error correction coefficient for the y-th parameter of the x-th feature. Let y be the stability coefficient of the y-th parameter of the x-th feature.

[0041] Specifically, the implementation process of the multi-feature fusion algorithm in the big data analysis module of non-invasive cardiac output monitoring is as follows: Feature categories are divided into three types: time features, morphological features, and physiologically related features. The number of parameters for each feature type is set according to actual analysis needs. Time features include 3-5 parameters (such as pulse wave conduction time, systolic duration, etc.), morphological features include 4-6 parameters (such as systolic peak amplitude, dicrotic wave amplitude, etc.), and physiologically related features include 2-4 parameters (such as vascular elastic modulus, heart rate, etc.). The value of each feature parameter is obtained through extraction and processing in the previous steps. The accuracy and timeliness of the parameters must be ensured, with a real-time update frequency of 10 seconds per update. The weighting coefficients are set according to the degree of influence of the features on the cardiac output monitoring results. Key parameters such as pulse wave conduction time and vascular elastic modulus are set to 0.2-0.3, while secondary parameters are set to 0.05-0.15, and the sum of all parameter weighting coefficients is 1.0. The normalization coefficient is set according to the parameter value range, uniformly converting parameters of different magnitudes to the 0-1 range to ensure balanced weighting of each parameter during the fusion process. Error correction coefficients are set according to the measurement accuracy of the parameters: 0.95-1.0 for parameters with high measurement accuracy (error less than 5%), 0.9-0.95 for parameters with medium accuracy (error 5%-10%), and 0.85-0.9 for parameters with low accuracy (error 10%-15%). Stability coefficients are set according to the fluctuation range of the parameters: 0.95-1.0 for fluctuation ranges less than 3%, and 0.9-0.95 for fluctuation ranges of 3%-5%. The calculation first multiplies each parameter of each feature category with its corresponding weight coefficient and normalization coefficient, then sums the results. Next, the sum of the products of all parameter error correction coefficients and stability coefficients is calculated. Finally, the two products are divided to obtain the cardiac output monitoring result. Through weighted fusion and error correction of multiple features, the information value of parameters in each dimension is fully utilized, effectively reducing the impact of single parameter errors on the monitoring results, improving the accuracy and stability of cardiac output monitoring results, and meeting clinical requirements for the reliability of cardiac output monitoring results.

[0042] Preferably, step S3 includes the following sub-steps: S31, performing signal segmentation processing on the pulse wave conduction time series dataset obtained in S2, dividing the entire time series into multiple independent pulse wave period segments according to the periodic characteristics of the pulse wave, with each period segment containing a complete pulse wave signal; S32, for each divided pulse wave period segment, using a wavelet transform algorithm to decompose the signal into multiple scales, obtaining wavelet coefficients at different scales, and filtering out signal components containing radial artery tension waveform calibration features through wavelet coefficients; S33, reconstructing the filtered signal components to remove noise interference introduced during the decomposition process, obtaining a relatively smooth original contour of the radial artery tension waveform; S34, setting a characteristic wave recognition threshold based on the physiological characteristics of the radial artery tension waveform, separating the systolic wave, diastolic wave, and dicrotic wave through threshold comparison and waveform slope analysis, and simultaneously recording the start time, peak time, and trough time time parameters of different characteristic waves, and measuring the peak amplitude and wave width morphology parameters of different characteristic waves.

[0043] Specifically, step S3 includes four sub-steps: In S31, when segmenting the pulse wave propagation time series dataset obtained in S2, the start and end points of each cycle are first determined using a pulse wave cycle recognition algorithm, with the cycle recognition accuracy controlled within 1 millisecond. Then, according to the standard that each cycle segment contains a complete pulse wave signal, the entire time series is divided into independent cycle segments. After segmentation, it is necessary to ensure that the signal length of each cycle segment is consistent with the actual duration of the cycle to avoid signal truncation or redundancy. In S32, when using the wavelet transform algorithm to perform multi-scale decomposition of the signal in each cycle segment, the decomposition scale is set to 5-8 levels. The signal components containing key features are screened out through the energy distribution of wavelet coefficients at each level, with the screening criterion being that the coefficient energy accounts for more than 30% of the total energy. The components are preserved; when reconstructing the selected components in S33, the inverse wavelet transform algorithm is used. During the reconstruction process, the noise introduced by the decomposition needs to be removed so that the smoothness of the original contour of the reconstructed radial artery tension waveform meets the requirement that the signal fluctuation amplitude is less than 5%; when setting the feature wave identification threshold in S34, the threshold is determined according to the peak amplitude of the reconstructed waveform, usually set to 10%-15% of the peak amplitude. Each feature wave is separated by threshold comparison and waveform slope analysis (the slope change rate is determined to be a feature point if it exceeds 50% / millisecond). At the same time, the accuracy of recording time parameters is controlled within 0.5 milliseconds, and the accuracy of measuring morphological parameters is controlled within 0.1 millivolts. By refining the steps, the accuracy of radial artery tension waveform decomposition and parameter extraction is ensured, providing reliable feature data for subsequent vascular elasticity analysis.

[0044] Preferably, step S4 includes the following sub-steps: S41, collecting basic physiological information of the monitored subject, including age, height, weight, gender, and history of underlying diseases, and inputting the data into the parameter configuration module of the vascular elasticity dynamic compensation model. This module assigns corresponding initial compensation coefficients to different basic physiological information according to preset mapping rules; S42, extracting calibration parameters related to vascular elasticity from the different characteristic wave component parameters obtained in S3, such as the vascular wall stress value corresponding to the peak value during systole and the vascular wall strain value corresponding to the trough value during diastole, and establishing the correlation between the calibration parameters and the initial compensation coefficients; S43, dynamically adjusting the initial compensation coefficients based on the real-time physiological data changes of the monitored subject during the monitoring process, such as heart rate fluctuations and blood pressure changes, calculating the deviation value of the compensation coefficient after each adjustment, and re-optimizing the adjustment strategy if the deviation value exceeds the set range; S44, substituting the adjusted compensation coefficients into the vascular elasticity dynamic compensation model, performing compensation calculations on the extracted vascular elasticity-related calibration parameters, generating the compensated vascular elasticity modulus and vascular compliance parameters, and storing them in the data cache module.

[0045] Specifically, step S4 includes four sub-steps: S41: When collecting the subject's basic physiological information, age is recorded to the nearest 1 year, height to the nearest 1 cm, and weight to the nearest 0.5 kg. The history of underlying diseases must clearly state the type and duration of the disease. After inputting this information into the parameter configuration module, the module allocates initial compensation coefficients according to preset mapping rules. The difference in coefficients between healthy individuals and patients with underlying diseases is controlled within 0.2-0.3. S42: When extracting key parameters related to vascular elasticity from the characteristic wave parameters in S3, the focus is on selecting the vascular wall stress value corresponding to the peak systolic wave (measurement accuracy 0.1 kPa) and the vascular wall strain value corresponding to the trough diastolic wave (measurement accuracy 0.001). The correlation between these parameters and the initial compensation coefficients is established. The correlation is determined using a linear fitting algorithm, and the goodness of fit must reach 0.8. 5 or higher; When adjusting the initial compensation coefficient based on real-time physiological data in S43, the heart rate fluctuation monitoring frequency is 1 Hz, and blood pressure changes are measured every 30 seconds. When the fluctuation exceeds the set range (heart rate ±5 beats / minute, blood pressure ±10 mmHg), the adjustment is initiated with an adjustment step size of 0.05. At the same time, the deviation value is calculated. If the deviation value exceeds 0.1, the adjustment strategy is re-optimized; When substituting the adjusted compensation coefficient into the model calculation in S44, the number of iterations of the compensation calculation is set to 3-5 times to ensure that the error of the compensated vascular elasticity parameters (elastic modulus, compliance) is less than 5%, and it is stored in the data cache module. The cache update frequency is consistent with the parameter calculation frequency (once every 10 seconds). By controlling the compensation process step by step, the accuracy of vascular elasticity parameters is improved, adapting to the differences in the physiological state of different subjects.

[0046] Preferably, step S5 includes the following sub-steps: S51, retrieving the compensated vascular elasticity correlation parameters output by S4 and the pulse wave conduction time series data generated by S2 from the data storage unit, and performing time axis alignment processing on the two types of data to ensure that each set of data corresponds one-to-one in the time dimension and eliminate time synchronization deviation; S52, performing feature standardization processing on the aligned dataset, converting the values ​​of different types of parameters into a unified data range to avoid interference to subsequent analysis due to differences in parameter magnitudes, and marking outliers in the dataset and storing them separately; S53, inputting the standardized dataset into a multi-feature fusion algorithm, which first performs feature layering on the data, dividing the data into a time feature layer, a morphological feature layer, and a physiological correlation feature layer, and then performs feature filtering within different layers to retain feature parameters that have a significant impact on non-invasive cardiac output monitoring indicators; S54, establishing a mathematical mapping relationship between different feature parameters and non-invasive cardiac output monitoring indicators through cross-correlation calculation between feature layers, generating a preliminary correlation mapping model, and performing a preliminary evaluation of the model's fit.

[0047] Specifically, step S4 includes four sub-steps: S41: When collecting the subject's basic physiological information, age is recorded to the nearest 1 year, height to the nearest 1 cm, and weight to the nearest 0.5 kg. The history of underlying diseases must clearly state the type and duration of the disease. After inputting this information into the parameter configuration module, the module allocates initial compensation coefficients according to preset mapping rules. The difference in coefficients between healthy individuals and patients with underlying diseases is controlled within 0.2-0.3. S42: When extracting key parameters related to vascular elasticity from the characteristic wave parameters in S3, the focus is on selecting the vascular wall stress value corresponding to the peak systolic wave (measurement accuracy 0.1 kPa) and the vascular wall strain value corresponding to the trough diastolic wave (measurement accuracy 0.001). The correlation between these parameters and the initial compensation coefficients is established. The correlation is determined using a linear fitting algorithm, and the goodness of fit must reach 0.8. 5 or higher; When adjusting the initial compensation coefficient based on real-time physiological data in S43, the heart rate fluctuation monitoring frequency is 1 Hz, and blood pressure changes are measured every 30 seconds. When the fluctuation exceeds the set range (heart rate ±5 beats / minute, blood pressure ±10 mmHg), the adjustment is initiated with an adjustment step size of 0.05. At the same time, the deviation value is calculated. If the deviation value exceeds 0.1, the adjustment strategy is re-optimized; When substituting the adjusted compensation coefficient into the model calculation in S44, the number of iterations of the compensation calculation is set to 3-5 times to ensure that the error of the compensated vascular elasticity parameters (elastic modulus, compliance) is less than 5%, and it is stored in the data cache module. The cache update frequency is consistent with the parameter calculation frequency (once every 10 seconds). By controlling the compensation process step by step, the accuracy of vascular elasticity parameters is improved, adapting to the differences in the physiological state of different subjects.

[0048] The pulse wave conduction time analysis algorithm in this invention is a technical means to extract the conduction time characteristics of pulse waves between different monitoring points from radial artery signals. Its implementation process is as follows: First, effective multi-channel signals are acquired using a sensor array signal processing platform. Key feature points, such as the systolic initiation point, of the pulse wave in each channel are accurately located using feature point recognition technology, and the time of each feature point is recorded (accuracy within 0.1 milliseconds). Then, adjacent effective monitoring points (usually 8-12) are selected, and the time difference between the corresponding feature points is calculated. After removing outliers, the average is taken to obtain the conduction time value of each channel combination. Finally, a time series dataset containing channel number, conduction time value, and pulse cycle sequence number is constructed. The calculation also incorporates multi-dimensional coefficient optimization results, such as signal strength weighting coefficient (0.6-1.0) and position correction coefficient (0.7-1.0). This algorithm provides quantified pulse wave time dimension data for subsequent analysis. Through multi-channel cross-validation and coefficient correction, the influence of signal strength differences, monitoring point position deviations, and environmental interference on conduction time calculation is eliminated, improving data accuracy and providing reliable time feature support for radial artery tension waveform analysis and cardiac output monitoring result calculation.

[0049] The radial artery tension waveform analytical model in this invention is an analytical model used to decompose radial artery signal waveforms and extract characteristic wave component parameters. Its implementation process is as follows: First, a pulse wave conduction time series dataset is received, the signal is split according to the pulse cycle and smoothed (using a moving average window of 5 sampling points); then, the smoothed signal is decomposed into systolic waves, diastolic waves, and dicrotic waves using the model. During decomposition, the systolic wave amplitude coefficient (0.6-1.2), positional attenuation coefficient (0.1-0.5), and pulse wave angular frequency (dynamically adjusted parameters such as 6.28-8.37 radians / second) are combined to accurately simulate tension changes at different positions and times; subsequently, the morphological parameters (peak amplitude accuracy 0.1 mV) and time parameters (duration accuracy 1 ms) of each characteristic wave are extracted, where the dicrotic wave must meet the recognition condition that its amplitude reaches more than 30% of the systolic wave peak value. This model transforms complex radial artery tension waveforms into quantifiable feature parameters. Through refined waveform decomposition and parameter definition, it avoids subjective errors caused by manual identification, ensuring the consistency and comparability of feature parameters. This provides direct waveform feature basis for the dynamic compensation model of vascular elasticity, and helps to correlate subsequent vascular elasticity status analysis with cardiac output monitoring indicators.

[0050] The vascular elasticity dynamic compensation model in this invention is a compensation model used to correct vascular elasticity-related parameters and eliminate the influence of individual differences and real-time physiological fluctuations. Its implementation process is as follows: First, basic physiological information of the subjects (age, height, weight, etc.) is collected. Initial compensation coefficients are assigned through the parameter configuration module (0.9-1.1 for healthy individuals, 1.3-1.5 for patients with underlying diseases, etc.). Then, key parameters such as vascular wall stress (accuracy 0.1 kPa) and strain (accuracy 0.001) are extracted from the radial artery tension waveform analysis results, and a correlation is established with the initial coefficients (fit ≥ 0.85). Subsequently, the coefficients are adjusted according to real-time physiological data (heart rate sampling frequency 1 Hz, blood pressure measured every 30 seconds). When fluctuations exceed the threshold (heart rate ± 5 beats / minute, blood pressure ± 10 mmHg), adjustments are made in 0.05 steps, and the deviation value is calculated (if it exceeds 0.1, an optimization strategy is employed). Finally, the coefficients are substituted to calculate the compensated vascular elastic modulus and compliance (error < 5%), and the data is updated every 10 seconds. This model dynamically corrects vascular elasticity parameters. By combining individual baseline information with real-time status, it breaks through the limitations of traditional fixed parameter analysis, adapts to the physiological differences of different populations, improves the accuracy of vascular elasticity data, and lays the foundation for establishing a precise correlation between vascular status and cardiac output indicators in the big data analysis module.

[0051] The sensor array signal processing platform in this invention is a hardware and algorithm integration platform for acquiring and processing multi-channel radial artery signals. Its implementation process is as follows: First, signals are acquired using an array of 16-32 pressure sensors (0.5 mm spacing, covering a 2 cm × 2 cm area) at a sampling frequency of 1000 Hz for 30 seconds to establish a basic library. Then, the raw signals undergo multi-dimensional filtering (50 Hz low-pass, 0.5 Hz high-pass, and adaptive filtering) to remove electromagnetic interference, limb micro-movements, and other interference. Subsequently, a signal enhancement algorithm is used to improve the signal-to-noise ratio (SNR). Combined with a signal gain coefficient (dynamically adjusted from 0.8 to 2.0), the amplitude of each channel signal is normalized to 0.5-3 mV. Simultaneously, the interference intensity coefficient (0.1-0.8) and suppression coefficient (0.7-1.0) are calculated. Multi-channel signals are then processed collaboratively using formulas to eliminate invalid channels with an SNR < 20 dB, ultimately outputting signals from at least 8 effective channels. This platform provides a high-quality, low-interference radial artery signal source, solving the problems of poor signal quality and excessive interference in traditional single-channel acquisition. Through multi-channel collaboration and targeted processing, it ensures that subsequent algorithms and models have reliable data source support, avoiding subsequent analysis deviations caused by interference from the original signal. It is the fundamental guarantee for the entire non-invasive cardiac output monitoring process.

[0052] like Figure 2As shown, a big data analysis system for non-invasive cardiac output monitoring is disclosed. This system is applied to a big data analysis method for non-invasive cardiac output monitoring and includes: a multi-channel radial artery signal acquisition unit connected to a sensor array signal processing platform, used to acquire multi-channel physiological signals from the radial artery region and transmit the acquired signals to the sensor array signal processing platform; a signal filtering and enhancement unit integrated within the sensor array signal processing platform, receiving signals transmitted from the multi-channel radial artery signal acquisition unit, performing multi-dimensional filtering and enhancement processing on the signals, filtering out effective signals, and transmitting the effective signals to a pulse wave transit time feature extraction unit; and a pulse wave transit time feature extraction unit connected to both the signal filtering and enhancement unit and the radial artery tension waveform analysis unit, performing feature extraction on the effective signals based on a pulse wave transit time analysis algorithm. The process involves extracting and transmitting a pulse wave conduction time series dataset to the radial artery tension waveform analysis unit. This unit, connected to the vascular elasticity dynamic compensation unit, uses a radial artery tension waveform analysis model to decompose the pulse wave conduction time series dataset, obtaining parameters of different characteristic wave components, which are then transmitted to the vascular elasticity dynamic compensation unit. The vascular elasticity dynamic compensation unit, connected to the big data analysis and integration unit, uses a vascular elasticity dynamic compensation model to dynamically compensate the characteristic wave component parameters, obtaining compensated vascular elasticity correlation parameters, which are then transmitted to the big data analysis and integration unit. Finally, the analysis and result output unit, connected to the vascular elasticity dynamic compensation unit, receives the compensated vascular elasticity correlation parameters and pulse wave conduction time series data, establishes a correlation mapping through a multi-feature fusion algorithm, iteratively optimizes it, and outputs the final monitoring result data.

[0053] A data big data analysis method and system for non-invasive cardiac output monitoring is proposed. This method utilizes a sensor array signal processing platform to perform multi-dimensional filtering and enhancement of multi-channel radial artery signals, specifically targeting the needs of non-invasive cardiac output monitoring to select effective signals. This addresses the problems of loose integration between signal processing and big data analysis in existing technologies, and the failure to process multi-channel signals as needed, laying a high-quality data foundation for subsequent analysis. By extracting features and constructing a time-series dataset using a pulse wave conduction time analysis algorithm, and combining this with a radial artery tension waveform analytical model to separate systolic, diastolic, and dicrotic waves to obtain multi-dimensional parameters, a vascular elasticity dynamic compensation model is used to adjust the compensation coefficient based on the basic physiological information of the monitored subjects. This overcomes the limitations of existing technologies that rely on single parameters or simple models and lack dynamic compensation, allowing for flexible adaptation to changes in the physiological state of different populations and significantly reducing monitoring result bias.

[0054] This method and system integrate compensated vascular elasticity parameters and pulse wave conduction time data through a big data analysis module. It employs a multi-feature fusion algorithm to process the data hierarchically and establish correlation mappings. Furthermore, it can improve the stability of the mappings through iterative optimization. This effectively solves the problems of weak multi-source data fusion capabilities and insufficient data correlation mining in existing technologies, ensuring the continuous output of high-precision monitoring results. At the same time, the entire process relies on non-invasive methods to collect signals and analyze data without any invasive operations. While ensuring monitoring accuracy, it also takes into account safety and convenience. This avoids the complication risks of traditional invasive monitoring and meets the clinical needs for long-term continuous monitoring and real-time assessment of cardiac function, making it suitable for monitoring applications in different scenarios.

[0055] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0056] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A big data analysis method for non-invasive cardiac output monitoring, characterized in that, Includes the following steps: S1. Multi-channel physiological signals from the radial artery region are acquired using a sensor array signal processing platform. This platform performs multi-dimensional filtering and signal enhancement on the acquired signals to select valid signals that meet the requirements for non-invasive cardiac output monitoring data. S2. Based on a pulse wave transit time analysis algorithm, features are extracted from the valid signals acquired in S1 to determine the transit time difference between different monitoring points, thus constructing a pulse wave transit time series dataset. S3. A radial artery tension waveform analytical model is used to decompose the pulse wave transit time series dataset obtained in S2, separating the systolic, diastolic, and dicrotic wave characteristic wave components from the radial artery tension waveform, and obtaining the morphological parameters and temporal parameters of different characteristic wave components. S4: Dynamic compensation calculation is performed on the different characteristic wave component parameters extracted in S3 using the vascular elasticity dynamic compensation model. The compensation coefficient is adjusted in combination with the basic physiological information of the monitored object to obtain the compensated vascular elasticity correlation parameters. S5: The compensated vascular elasticity correlation parameters in S4 are integrated with the pulse wave conduction time series data in S2 and input into the big data analysis module of non-invasive cardiac output monitoring. This module uses a multi-feature fusion algorithm to perform hierarchical processing on the integrated data and establish a correlation mapping between the data and non-invasive cardiac output monitoring indicators. S6: The correlation mapping established in S5 is iteratively optimized through the big data analysis module to output the final monitoring result data that meets the accuracy requirements of non-invasive cardiac output monitoring. The expression used in the pulse wave propagation time analysis algorithm is: ,in, The pulse wave conduction time. The time when the pulse wave feature point is collected at the i-th monitoring point. For the first The time when pulse wave characteristic points were collected at each monitoring point The total number of monitoring points. Let be the signal strength weighting coefficient for the i-th monitoring point. Let be the position correction coefficient for the i-th monitoring point. Let be the signal stability coefficient of the i-th monitoring point. Let be the environmental interference compensation coefficient for the i-th monitoring point; The expression used in the analytical model of radial artery tension waveform is: ,in, The tension value of the radial artery at position r and time t. The amplitude coefficient of the contraction wave. The position attenuation coefficient, Radial position coordinates, The coordinates of the radial artery center are: The pulse wave angular frequency, For time, This represents the phase shift of the contraction phase wave. The amplitude coefficient of the diastolic wave. The time decay coefficient, The start of diastole Radial phase coefficient, This refers to the phase shift of the diastolic wave. The expression used in the dynamic compensation model for vascular elasticity is: ,in, The compensated vascular elastic modulus, The original value of vascular elastic modulus before compensation. The number of samples is calculated for dynamic compensation. Let be the compensation coefficient for the j-th sampling. The blood pressure value is from the j-th sample. To set the average blood pressure value over a given time period, Let j be the heart rate value from the j-th sample. To set the average heart rate value over a given time period, Let j be the time decay factor for the j-th sample; The signal processing expression of the sensor array signal processing platform is: ,in, For the platform to output valid signals, This refers to the number of channels in the sensor array. Let p be the original input signal of the p-th channel. Let p be the signal gain coefficient of the p-th channel. Let be the interference intensity coefficient of the k-th type of interference signal in the p-th channel. Let be the suppression coefficient of the k-th type of interference in the p-th channel; The multi-feature fusion algorithm expression of the data big data analysis module for non-invasive cardiac output monitoring is as follows: ,in, This is the result of non-invasive cardiac output monitoring. The number of feature categories, The number of parameters for each type of feature. Let be the value of the y-th parameter in the x-th feature class. Let y be the weight coefficient of the y-th parameter of the x-th feature. Let y be the normalized coefficient of the y-th parameter of the x-th feature. Let y be the error correction coefficient for the y-th parameter of the x-th feature. Let y be the stability coefficient of the y-th parameter of the x-th feature; S5 includes the following sub-steps: S51, retrieve the compensated vascular elasticity correlation parameters output from S4 and the pulse wave conduction time series data generated in S2 from the data storage unit, and perform time axis alignment processing on the two types of data to ensure that each set of data corresponds one-to-one in the time dimension and eliminate time synchronization deviation; S52, perform feature standardization processing on the aligned dataset, convert the values ​​of different types of parameters into a unified data range to avoid interference to subsequent analysis due to differences in parameter magnitudes, and mark outliers in the dataset and store them separately; S53, input the standardized dataset into a multi-feature fusion algorithm, which first performs feature layering on the data, dividing the data into a time feature layer, a morphological feature layer, and a physiological correlation feature layer, and then performs feature filtering within different layers to retain feature parameters that have a significant impact on non-invasive cardiac output monitoring indicators; S54, establish a mathematical mapping relationship between different feature parameters and non-invasive cardiac output monitoring indicators through cross-correlation calculation between feature layers, generate a preliminary correlation mapping model, and perform a preliminary evaluation of the model's fit.

2. The big data analysis method for non-invasive cardiac output monitoring according to claim 1, characterized in that, S3 includes the following sub-steps: S31, performing signal segmentation processing on the pulse wave conduction time series dataset obtained in S2, dividing the entire time series into multiple independent pulse wave period segments according to the periodic characteristics of the pulse wave, with each period segment containing a complete pulse wave signal; S32, for each segmented pulse wave period, using a wavelet transform algorithm to decompose the signal into multiple scales, obtaining wavelet coefficients at different scales, and filtering out signal components containing radial artery tension waveform calibration features through wavelet coefficients; S33, reconstructing the filtered signal components to remove noise interference introduced during the decomposition process, obtaining a relatively smooth original contour of the radial artery tension waveform; S34, setting a characteristic wave recognition threshold based on the physiological characteristics of the radial artery tension waveform, separating the systolic wave, diastolic wave, and dicrotic wave through threshold comparison and waveform slope analysis, while recording the start time, peak time, and trough time time parameters of different characteristic waves, and measuring the peak amplitude and wave width morphology parameters of different characteristic waves.

3. The big data analysis method for non-invasive cardiac output monitoring according to claim 1, characterized in that, S4 includes the following sub-steps: S41, collecting basic physiological information of the monitored subjects, including age, height, weight, gender, and history of underlying diseases, and inputting the data into the parameter configuration module of the vascular elasticity dynamic compensation model. This module assigns corresponding initial compensation coefficients to different basic physiological information according to preset mapping rules; S42, extracting calibration parameters related to vascular elasticity from the different characteristic wave component parameters obtained in S3, such as the vascular wall stress value corresponding to the peak value during systole and the vascular wall strain value corresponding to the trough value during diastole, and establishing the correlation between the calibration parameters and the initial compensation coefficients; S43, Based on the real-time physiological data changes of the monitored object during the monitoring process, such as heart rate fluctuations and blood pressure changes, the initial compensation coefficient is dynamically adjusted, and the deviation value of the compensation coefficient after each adjustment is calculated. If the deviation value exceeds the set range, the adjustment strategy is re-optimized. S44, The adjusted compensation coefficient is substituted into the vascular elasticity dynamic compensation model, and the extracted vascular elasticity correlation calibration parameters are compensated to generate the compensated vascular elasticity modulus and vascular compliance parameters, and stored in the data cache module.

4. A big data analysis system for non-invasive cardiac output monitoring, characterized in that, This system is applied to a data big data analysis method for non-invasive cardiac output monitoring as described in claim 1, comprising: a multi-channel radial artery signal acquisition unit, which is connected to a sensor array signal processing platform for acquiring multi-channel physiological signals from the radial artery region and transmitting the acquired signals to the sensor array signal processing platform; a signal filtering and enhancement processing unit, which is integrated within the sensor array signal processing platform, receives the signals transmitted from the multi-channel radial artery signal acquisition unit, performs multi-dimensional filtering and enhancement processing on the signals, filters out effective signals, and transmits the effective signals to a pulse wave conduction time feature extraction unit; and a pulse wave conduction time feature extraction unit, which is connected to both the signal filtering and enhancement processing unit and the radial artery tension waveform analysis unit, and extracts features from the effective signals based on a pulse wave conduction time analysis algorithm to obtain the pulse wave. The pulse wave conduction time series dataset is transmitted to the radial artery tension waveform analysis unit. This unit, connected to the vascular elasticity dynamic compensation unit, uses the radial artery tension waveform analysis model to decompose the pulse wave conduction time series dataset, obtaining parameters of different characteristic wave components, which are then transmitted to the vascular elasticity dynamic compensation unit. The vascular elasticity dynamic compensation unit, connected to the big data analysis and integration unit, uses the vascular elasticity dynamic compensation model to dynamically compensate the characteristic wave component parameters, obtaining compensated vascular elasticity correlation parameters, which are then transmitted to the big data analysis and integration unit. The analysis and result output unit, connected to the vascular elasticity dynamic compensation unit, receives the compensated vascular elasticity correlation parameters and pulse wave conduction time series data, establishes a correlation mapping through a multi-feature fusion algorithm, iteratively optimizes it, and outputs the final monitoring result data.

Images