Hemodialysis circulating pump power self-adaptive adjusting method based on multi-source signals

Abstract

The invention discloses a hemodialysis circulating pump power self-adaptive adjusting method based on a multi-source signal, and relates to the technical field of medical instruments, and the hemodialysis circulating pump power self-adaptive adjusting method comprises the following steps: collecting and preprocessing multi-source signal data in real time, and generating a standardized multi-source signal vector; eliminating signal noise entanglement of the standardized multi-source signal vector, verifying a shear stress value, and generating a low-noise fusion state vector and a real-time phase parameter; constructing a power regulation instruction according to the prediction track and the risk index, executing the power regulation instruction, and collecting state feedback data after execution in real time; the adjustment efficiency is evaluated by calculating the recovery rate deviation ratio of the state feedback data after execution, the adjustment efficiency and the state feedback data after execution are integrated, and an adjustment report is obtained. According to the method, signal noise entanglement of a standardized multi-source signal vector is eliminated, and inter-signal interference minimization and phase locking integration under high-dimensional space mapping are achieved.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and in particular to a method for adaptive adjustment of the power of a hemodialysis circulation pump based on multi-source signals. Background Technology

[0002] In the field of hemodialysis, circulation pump power regulation is a core mechanism for maintaining hemodynamic stability and patient physiological balance. This therapy is widely used in the management of chronic kidney disease, achieving efficient extracorporeal blood circulation through pump rate control. Conventional methods primarily rely on single physiological signals or equipment monitoring data for power adjustment, such as blood pressure or blood flow feedback, to optimize pump output and ensure the continuity and safety of the dialysis process. This method has been thoroughly validated in clinical settings, supporting the reliable implementation of various dialysis modalities.

[0003] Although conventional methods perform well in basic power control, they have limited ability to fuse multi-source signals and are susceptible to noise entanglement, which reduces signal purity. In addition, they lack predictive assessment of blood flow dynamics and are difficult to capture potential hypotension fluctuations in real time, thus limiting the accuracy and foresight of regulation. Summary of the Invention

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

[0005] Therefore, this invention provides a method for adaptive adjustment of the power of a hemodialysis circulation pump based on multi-source signals to solve the problems of signal noise interference and risk prediction lag.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for adaptive adjustment of the power of a hemodialysis circulation pump based on multi-source signals, comprising: Real-time acquisition and preprocessing of multi-source signal data to generate standardized multi-source signal vectors; Eliminate signal-noise entanglement in standardized multi-source signal vectors, verify shear stress values, and generate low-noise fused state vectors and real-time phase parameters; Feature extraction and pulsating waveform optimization are performed on the low-noise fused state vector and real-time phase parameters to generate a pulsating waveform parameter set; Predict the trajectory of the rate of change in circulating blood volume based on the set of pulsating waveform parameters, assess the potential risk of hypotension, and obtain the predicted trajectory and risk indicators. Power regulation commands are constructed based on predicted trajectories and risk indicators, and the power regulation commands are executed while real-time status feedback data is collected after execution. The regulation effectiveness is assessed by calculating the recovery rate deviation ratio of the post-execution status feedback data. The regulation effectiveness and the post-execution status feedback data are then integrated to obtain a regulation report.

[0007] As a preferred embodiment of the adaptive adjustment method for the power of a hemodialysis circulation pump based on multi-source signals described in this invention, the multi-source signal data includes core blood concentration signals, auxiliary physiological signals, and equipment signals; the preprocessing includes data cleaning and standardization.

[0008] As a preferred embodiment of the adaptive power adjustment method for hemodialysis circulation pumps based on multi-source signals described in this invention, the specific steps for eliminating signal-noise entanglement of standardized multi-source signal vectors, verifying shear stress values, and generating low-noise fused state vectors and real-time phase parameters are as follows.

[0009] The standardized multi-source signal vectors are mapped into a high-dimensional space to generate a preliminary fusion state representation. Phase synchronization adjustment is performed on the initial fusion state representation, and blood flow velocity gradient is simulated to assess whether the shear stress is within the safe range, generating a verified fusion adjustment value; Using the verified fusion adjustment value as the phase reference benchmark, the final entanglement release and phase-locked integration of all signal channels are performed to obtain a low-noise fusion state vector and real-time phase parameters.

[0010] As a preferred embodiment of the hemodialysis circulation pump power adaptive adjustment method based on multi-source signals described in this invention, the specific steps for extracting features and optimizing pulsating waveforms from the low-noise fused state vector and real-time phase parameters to generate a pulsating waveform parameter set are as follows. Based on low-noise fused state vectors and real-time phase parameters, dynamic blood flow components are captured and physiological rhythm changes are highlighted by empirical mode decomposition to determine the intermediate pulsation feature representation. The adjustment amplitude and frequency distribution represented by the simulated intermediate pulsation characteristics are analyzed, and phase locking is enhanced to smooth potential oscillations, resulting in a preliminary waveform optimization draft. The performance was verified and locked using the preliminary waveform optimization draft to obtain the pulsating waveform parameter set.

[0011] As a preferred embodiment of the adaptive adjustment method for the power of the hemodialysis circulation pump based on multi-source signals described in this invention, the step of predicting the trajectory of the rate of change of circulating blood volume based on the pulse waveform parameter set refers to inferring the potential blood pressure fluctuation path by linearly extrapolating the dynamic trend of blood flow based on the pulse waveform parameter set, and obtaining a preliminary risk trajectory draft.

[0012] As a preferred embodiment of the hemodialysis circulation pump power adaptive adjustment method based on multi-source signals described in this invention, the specific steps for assessing potential hypotension risk and obtaining predicted trajectories and risk indicators are as follows: Based on the preliminary draft risk trajectory and historical adjustment reports, deviation integral adjustments are made, and the severity of the trajectory deviating from the safety boundary is assessed to generate a predicted trajectory; The predicted trajectory is risk-assessed and weighted to generate risk indicators.

[0013] As a preferred embodiment of the hemodialysis circulation pump power adaptive adjustment method based on multi-source signals described in this invention, the specific steps for constructing the power adjustment command based on the predicted trajectory and risk indicators are as follows: Trend fitting is performed on the predicted trajectory and risk indicators to generate temporary power offset values; Based on the temporary power offset value, phase and frequency tuning are performed in conjunction with risk indicators, and the overall balance is verified to generate an optimized adjustment framework. The optimized regulation framework is integrated and locked to construct power regulation commands.

[0014] As a preferred embodiment of the hemodialysis circulation pump power adaptive adjustment method based on multi-source signals described in this invention, the specific steps for executing the power adjustment command and collecting post-execution status feedback data are as follows: Adjust the pump output power based on the power regulation command and calculate the instantaneous execution response value; Real-time acquisition of subsequent patient blood flow and physiological changes, combined with immediate execution response values ​​to assess the power stability range, and generating preliminary status feedback aggregation; The initial status feedback is synchronously verified and deviations are corrected to obtain post-execution status feedback data.

[0015] As a preferred embodiment of the hemodialysis circulation pump power adaptive adjustment method based on multi-source signals described in this invention, the specific steps for evaluating the adjustment efficiency by calculating the recovery rate deviation ratio of the post-execution state feedback data are as follows: Based on the post-execution status feedback data, the recovery effect is quantified, and a preliminary performance evaluation value is determined. The preliminary performance evaluation value is weighted and integrated with the post-implementation status feedback data to assess the impact of deviations on long-term stability and generate a regulation performance index.

[0016] As a preferred embodiment of the adaptive adjustment method for the power of a hemodialysis circulation pump based on multi-source signals described in this invention, the adjustment report is obtained by integrating the adjustment efficiency index and the post-execution status feedback data according to the timestamp.

[0017] The beneficial effects of this invention are as follows: by eliminating signal noise entanglement in standardized multi-source signal vectors and verifying shear stress values ​​to generate low-noise fused state vectors and real-time phase parameters, the invention achieves the minimization of signal interference and phase-locked integration under high-dimensional spatial mapping. This ensures the pure characterization of blood flow dynamics and the assessment of safe zones during hemodialysis, and enables reliable fusion of multi-source signals, thereby improving the accuracy and physiological compatibility of basic data for pump power regulation. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a method for adaptive adjustment of the power of a hemodialysis circulation pump based on multi-source signals.

[0020] Figure 2 This is a flowchart for multi-source signal acquisition and fusion processing.

[0021] Figure 3 This is a flowchart for extracting and optimizing pulsating waveform features.

[0022] Figure 4 This is a flowchart for power regulation and feedback evaluation. Detailed Implementation

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0026] Reference Figures 1-4This is one embodiment of the present invention, which provides a method for adaptive adjustment of the power of a hemodialysis circulation pump based on multi-source signals, comprising the following steps: S1. Real-time acquisition and preprocessing of multi-source signal data to generate standardized multi-source signal vectors; Multi-source signal data includes core blood concentration signals, auxiliary physiological signals, and device signals; Furthermore, core blood concentration signals are continuously acquired through optical sensors deployed in the extracorporeal circulation loop, which reflect dynamic changes in blood concentration; auxiliary physiological signals are acquired using electrophysiological electrodes and pressure transducers attached to the patient's body surface, which include cardiac activity, respiratory rhythm, and arterial pressure waveforms; and device signals are read from the operating interface of the extracorporeal circulation device, which include pump speed, flow output, and tubing pressure parameters. Preprocessing includes data cleaning and standardization; Furthermore, missing values ​​are imputed using a linear interpolation method based on temporal proximity points on the real-time acquired core blood concentration signal. The expression is as follows: ; in, As the core blood concentration signal, To The linear interpolation result at time t. To focus on the missing time points Adjacent sampled time points The core blood concentration signal, To focus on the missing time points Adjacent sampled time points The core blood concentration signal, for Adjacent sampled time points , for Adjacent sampled time points ; For auxiliary physiological signals, spline interpolation was used to impute missing values, and for device signals, zero-order hold was used for imputation. A sliding window mean value filtering method was used to eliminate transient interference, thus completing data cleaning. The cleaned core blood concentration signal, auxiliary physiological signal, and device signal were normalized separately, with each signal distributed within a uniform numerical range. The normalization method used linear scaling of the maximum and minimum values ​​of each signal channel during the acquisition period. The normalized core blood concentration signal, auxiliary physiological signal, and device signal were aligned by time and concatenated into a vector form to form a standardized multi-source signal vector.

[0027] S2. Eliminate signal-noise entanglement in the standardized multi-source signal vector, verify the shear stress value, and generate a low-noise fused state vector and real-time phase parameters. The standardized multi-source signal vectors are mapped into a high-dimensional space to generate a preliminary fusion state representation. Furthermore, principal component analysis is applied to the standardized multi-source signal vector for high-dimensional spatial mapping, projecting the signal onto an orthogonal high-dimensional space. Independent component analysis is then used to decompose the mapped signal, separating statistically independent source signal components to generate a preliminary fusion state representation. It should be noted that the projection to the orthogonal high-dimensional space is obtained by calculating the eigenvectors of the covariance matrix of the standardized multi-source signal vectors using the singular value decomposition method to obtain the projection direction required by the principal component analysis method. The standardized multi-source signal vectors are then linearly transformed along the eigenvector direction to complete the mapping to the orthogonal high-dimensional space. Phase synchronization adjustment is performed on the initial fusion state representation, and blood flow velocity gradient is simulated to assess whether the shear stress is within the safe range, generating a verified fusion adjustment value; Furthermore, the independent component analysis method is used to decompose the independent components obtained from the preliminary fusion state representation, generating time series components. The instantaneous phase of the time series components is calculated using Hilbert transform, and the phase alignment of the time series components is performed using time warping method with the main period of the arterial pressure waveform as a reference, thus completing the phase synchronization adjustment. Based on the phase-synchronized time series components, the Poiseuille flow method is used to simulate the gradient distribution of blood flow velocity on the cross-section of the pipeline, and then the shear stress value at the pipeline wall is calculated according to Newton's fluid shear stress formula. The calculated shear stress value is compared with the preset safe range. If the shear stress value is within the safe range, the current phase synchronization result is retained; otherwise, the phase synchronization parameters are finely adjusted in the reverse direction according to the shear stress deviation direction until the shear stress value returns to the safe range. Finally, the phase synchronization result constrained by shear stress is output as the fusion adjustment value after verification. It should be noted that the main period of the arterial pressure waveform is obtained by detecting the time difference between consecutive systolic pressure peaks in the waveform. Specifically, a peak detection algorithm is used to calculate the time distance between adjacent peaks in the waveform, which is the main period of the arterial pressure waveform (exemplary value range: 0.6 seconds - 1.2 seconds). The preset safety range (exemplary value range: 1.5 Pascals - 7.0 Pascals) is determined based on fluid mechanics. The principle of calculating the shear stress value at the pipe wall according to Newton's fluid shear stress formula is based on the assumption that the fluid flow behavior conforms to Newton's laws, that is, there is a linear relationship between the fluid shear stress and the velocity gradient, thereby calculating the shear stress value. Using the verified fusion adjustment value as the phase reference benchmark, the final entanglement release and phase locking integration of all signal channels are performed to obtain a low-noise fusion state vector and real-time phase parameters. Furthermore, using the verified fusion adjustment value as a phase reference, a Hilbert transform is applied to each signal channel component to obtain the analytic signal and calculate the instantaneous phase sequence. A reference instantaneous phase sequence is extracted from the verified fusion adjustment value. The phase difference sequence between the instantaneous phase sequence of each signal channel component and the reference instantaneous phase sequence is calculated. A zero-phase digital filtering method is used to perform low-pass filtering on the phase difference sequence to remove high-frequency jitter, resulting in a smooth phase offset. The phase offset is removed from the analytic signal phase of the original signal channel component to reconstruct the phase-corrected analytic signal. The real part of the corrected analytic signal is taken as the phase-locked signal channel component, thus completing phase locking. An eigenvalue weighted average method is used to reconstruct the phase-locked signal channel components. The reconstructed signal channel components are combined into a vector form according to the original channel order to form a low-noise fusion state vector. Simultaneously, the unified instantaneous phase value of each signal channel component after phase locking is collected as a real-time phase parameter. It should be noted that the eigenvalue proportional weighting method refers to the normalization of the eigenvalues ​​corresponding to each principal component obtained when performing principal component analysis on a standardized multi-source signal vector, and using these eigenvalues ​​as weight coefficients for the corresponding signal channel components. The product of each phase-locked signal channel component and its corresponding weight coefficient is linearly superimposed to complete the weighted reconstruction. The zero-phase digital filtering method refers to first using a forward filtering method to process the phase difference sequence from the beginning of the signal to the end, and then using a reverse filtering method to process the phase difference sequence from the end of the signal to the beginning.

[0028] S3. Perform feature extraction and pulsating waveform optimization on the low-noise fused state vector and real-time phase parameters to generate a pulsating waveform parameter set; Based on low-noise fused state vectors and real-time phase parameters, dynamic blood flow components are captured and physiological rhythm changes are highlighted by empirical mode decomposition to determine the intermediate pulsation feature representation. Furthermore, empirical mode decomposition (EMD) is applied to each signal channel component in the low-noise fusion state vector to select intrinsic mode functions (IMFs) corresponding to the main period frequency range of the arterial pressure waveform (exemplary range: 0.6 seconds - 1.2 seconds) as dynamic blood flow components. Hilbert transform is applied to each IMF to obtain an instantaneous phase sequence. The phase difference between the instantaneous phase sequence and the real-time phase parameters is calculated by comparing the differences point-by-point at the same time point to obtain an undecoupled phase difference sequence. Phase decoupling of the phase difference sequence is performed by detecting phase jump points using the cumulative difference method to obtain the phase difference between the instantaneous phase sequence and the real-time phase parameters. The phase difference is used as a phase offset to perform phase rotation on the analytical signal of the IMF, resulting in a phase-aligned IMF. The phase-aligned dynamic blood flow components are then weighted and superimposed using a weighted summation method, and the superposition result is output as an intermediate pulsation feature representation. It should be noted that the main cycle frequency range of the arterial pressure waveform was determined using the empirical mode decomposition method; The adjustment amplitude and frequency distribution represented by the simulated intermediate pulsation characteristics are analyzed, and phase locking is enhanced to smooth potential oscillations, resulting in a preliminary waveform optimization draft. Furthermore, the frequency spectrum distribution is calculated by applying Fast Fourier Transform to the intermediate pulsation feature representation. The specific range of adjustment amplitude and frequency distribution is determined by extracting the peak positions and corresponding amplitude values ​​in the amplitude spectrum (exemplary value range: 0Hz-20Hz). The high-frequency components in the intermediate pulsation feature representation are low-pass filtered using the Butterworth low-pass filtering method. The phase timestamp is applied to the low-pass filtered components using the Hilbert transform method as a reference to achieve phase locking enhancement. The phase-locked enhanced components are then weighted and fused with the adjustment amplitude and frequency distribution range to generate a preliminary waveform optimization draft. The performance verification and locking fusion were performed using the preliminary waveform optimization draft to obtain the pulsating waveform parameter set; Furthermore, the minimum mean square error criterion is applied to the preliminary waveform optimization draft to calculate the fitting error between the intermediate pulsation feature representation and the fitting error is compared to determine the performance verification result. Based on the performance verification result, the components with fitting errors lower than the calculation error are phase-locked and adjusted. Specifically, the instantaneous phase sequence is extracted using Hilbert transform and phase rotation is applied using real-time phase parameters as a reference to achieve locking fusion. The locked and fused components are combined using a weighted summation method to obtain the pulsation waveform parameter set.

[0029] S4. Predict the trajectory of the rate of change of circulating blood volume based on the set of pulsating waveform parameters, assess the potential risk of hypotension, and obtain the predicted trajectory and risk indicators. Based on the set of pulsating waveform parameters, the potential blood pressure fluctuation path is inferred by linear extrapolation of blood flow dynamics, and a preliminary draft of the risk trajectory is obtained. Furthermore, the empirical mode decomposition method is applied to the pulsating waveform parameter set to decompose it into an intrinsic mode function sequence. The low-frequency intrinsic mode functions are selected as blood flow dynamic trend components using the fast Fourier transform method. The Hilbert transform is applied to the blood flow dynamic trend components to extract the instantaneous amplitude and instantaneous frequency sequences. The least squares method is used to perform linear extrapolation on the instantaneous amplitude and instantaneous frequency sequences to calculate the amplitude and frequency values ​​of future time steps. The extrapolated amplitude, frequency, and phase values ​​are weighted and summed to reconstruct the blood pressure fluctuation path and generate a preliminary risk trajectory draft. Based on the preliminary draft risk trajectory and historical adjustment reports, deviation integral adjustments are made, and the severity of the trajectory deviating from the safety boundary is assessed to generate a predicted trajectory; Furthermore, the minimum mean square error criterion is applied to calculate the point-by-point deviation sequence between the preliminary risk trajectory draft and the historical adjustment report. The integral value of the deviation is calculated by accumulating and summing the deviation sequences, and the expression is as follows: ; in, This is the integral value of the deviation. Time in the deviation sequence A preliminary draft of the risk trajectory. Time in the deviation sequence Blood pressure regulation pathway This represents the total time of the deviation sequence; Historical adjustment reports are obtained from blood pressure regulation paths extracted from previous extracorporeal circulation device operation logs. A proportional scaling adjustment is applied to the initial risk trajectory draft based on the deviation integral value, generating an adjusted trajectory draft. The Euclidean distance between the adjusted trajectory draft and the preset safe zone boundary is calculated, and the severity of trajectory deviation from the safe zone is assessed based on the magnitude of the Euclidean distance. The adjusted trajectory draft is then corrected using a weighted summation method based on the severity, where the low-frequency trend component is extracted from the adjusted trajectory draft using empirical mode decomposition. A weight coefficient inversely proportional to the severity is applied to the low-frequency trend component, and amplitude scaling is corrected using a weighted summation method. An attenuation coefficient proportional to the severity is applied to the high-frequency fluctuation component, and amplitude is suppressed through low-pass filtering. The corrected low-frequency trend component and the attenuated high-frequency fluctuation component are combined using a weighted summation method to generate a predicted trajectory. The predicted trajectory is used to predict blood pressure change trends and assess whether deviations from the safe zone are likely. It should be noted that the severity of trajectory deviation from the safety boundary is assessed by the magnitude of the Euclidean distance, which is divided into three levels: low, medium, and high. A Euclidean distance of less than 5 mmHg indicates low severity, meaning the trajectory is basically within the safety boundary; a distance of 5-15 mmHg indicates medium severity, meaning the trajectory has a moderate deviation and may require mild intervention; a distance of more than 15 mmHg indicates high severity, meaning the trajectory has significantly deviated, posing a high risk of hypotension and requiring immediate adjustment. An example range for the magnitude of the Euclidean distance is 0-30 mmHg. The predicted trajectory is risk-assessed and weighted to generate risk indicators. Furthermore, the predicted trajectory is decomposed into low-frequency trend components and high-frequency fluctuation components using the empirical mode decomposition method. The risk level of hypotension is assessed by calculating the Euclidean distance of the low-frequency trend components, while the high-frequency fluctuation components are extracted using Hilbert transform to extract the instantaneous amplitude sequence to quantify the fluctuation intensity. Based on the risk level of hypotension and the fluctuation intensity, an inverse weighting coefficient is applied to the low-frequency trend components, and a direct weighting coefficient is applied to the high-frequency fluctuation components. The components are then integrated using a weighted summation method to generate a risk index.

[0030] S5. Construct power adjustment instructions based on the predicted trajectory and risk indicators, execute the power adjustment instructions and collect status feedback data after execution in real time; Trend fitting is performed on the predicted trajectory and risk indicators to generate temporary power offset values; Furthermore, the predicted trajectory is compared point by point with the preset center line of the safety interval, and the deviation value of the predicted trajectory relative to the center line of the safety interval at each time point is calculated to form a deviation sequence. The least mean square error criterion is applied to the deviation sequence to perform trend fitting, and the overall deviation direction and magnitude are extracted. Based on the value of the risk indicator, the adjustment urgency weight is determined by the normalized exponential mapping method (exemplary value range: 0.2–1.0). The product of the overall deviation magnitude and the adjustment urgency weight is used as the basic power adjustment amount to generate the components of the control dimension. The components of all control dimensions are superimposed and output as the temporary power offset value. It should be noted that the preset safe zone centerline value range (exemplary value range: -100 mmHg -180 mmHg) was determined by statistical analysis of the patient's arterial pressure waveform data using the mean clustering method; Based on the temporary power offset value, phase and frequency tuning are performed in conjunction with risk indicators, and the overall balance is verified to generate an optimized adjustment framework. Furthermore, the temporary power offset value is decomposed into amplitude and time components. The time component is phase-aligned using a phase rotation alignment method based on real-time phase parameters, synchronizing the power adjustment action with the main cycle of the arterial pressure waveform. The frequency tuning intensity (exemplary range: 0.2-1) is determined using a linear proportional amplification method based on the risk index value. A fast Fourier transform is applied to the temporary power offset value to extract the dominant frequency component, which is then adjusted to be closer to the fundamental frequency corresponding to the main cycle of the arterial pressure waveform, completing the frequency tuning. The phase-aligned and frequency-tuned temporary power offset value is superimposed on the current pump output power reference to form candidate adjustment data. The expected perturbation amplitude in the core blood concentration signal, auxiliary physiological signal, and equipment signal dimensions is calculated using the least squares method. The balance of perturbations in the core blood concentration signal, auxiliary physiological signal, and equipment signal is evaluated using the minimum mean square error criterion, and the data is then spliced ​​and integrated to obtain the optimized adjustment framework. The optimized regulation framework is integrated and locked to construct power regulation commands; Furthermore, the optimized control framework is validated for consistency using a timing alignment correction method to obtain dynamically compatible device control dimension components. Using real-time phase parameters as the time reference, the device control dimension components are time-aligned within the phase period and fused into a unified timing control sequence. The timing control sequence is then smoothed using a zero-phase digital filtering method to eliminate high-frequency jitter introduced by tuning, resulting in a smoothed timing control sequence. Finally, an integer encoding method is used to map the smoothed timing control sequence into a power control command format executable by the extracorporeal circulation device, forming a power regulation instruction. It should be noted that the integer encoding method maps the continuous control value at each moment in the timing control sequence to a set of discrete integer values, which serve as the power control command for the extracorporeal circulation device. Adjust the pump output power based on the power regulation command and calculate the instantaneous execution response value; Furthermore, power adjustment commands are sequentially sent to the control interface of the extracorporeal circulation device to drive the pump to perform the corresponding power adjustment action; within the first sampling cycle after the power adjustment is executed, feedback values ​​of actual pump speed, flow output, and pipeline pressure parameters are read from the device signals; real-time changes in arterial pressure waveforms are obtained from auxiliary physiological signals, and instantaneous response feedback values ​​of component concentrations are extracted from the core blood concentration signal; the feedback values ​​are compared item by item with the target values ​​in the power adjustment commands, the minimum mean square error criterion is used to calculate the sum of squares of deviations in the signal dimension, and the weighted summation method is used to weight the sum of squares of deviations in the dimension, and the reciprocal of the output weighted deviation value is used as the instantaneous execution response value; Real-time acquisition of subsequent patient blood flow and physiological changes, combined with immediate execution response values ​​to assess the power stability range, and generating preliminary status feedback aggregation; Furthermore, within multiple consecutive sampling cycles after the power adjustment command is executed, the core blood concentration signal is continuously acquired through an optical sensor, and the arterial pressure waveform in the auxiliary physiological signal is continuously acquired through electrophysiological electrodes and a pressure transducer. The pump speed, flow output, and pipeline pressure parameters in the device signal are read from the extracorporeal circulation device operation interface. The real-time acquired core blood concentration signal, auxiliary physiological signal, and device signal are aligned by time to form a multi-source feedback sequence. The numerical change rate of each signal channel in the multi-source feedback sequence is jointly analyzed with the instantaneous execution response value. Specifically, if the change rate of each channel is less than the historical fluctuation standard deviation (exemplary value range: 0.1-2) within three consecutive sampling cycles, it is determined that the current pump output power is in the power stability range. The multi-source feedback sequence corresponding to all sampling points in the power stability range and the instantaneous execution response value are packaged by timestamp to form a preliminary state feedback aggregation. It should be noted that the range of historical fluctuation standard deviation is determined based on historical adjustment reports using the sliding window standard deviation statistical method; The initial status feedback collection is synchronously verified and deviations are corrected to obtain post-execution status feedback data. Furthermore, the core blood concentration signal, auxiliary physiological signal, and equipment signal in the initial state feedback aggregation are aligned point by point according to timestamps. Each signal channel is checked for time drift or sampling synchronization failure. If time drift is found, linear interpolation is used to realign them on the time axis. The mean and standard deviation of each signal channel within the power stability range are calculated, and sampling points deviating from the mean by more than a multiple of the historical fluctuation standard deviation are marked as abnormal. Abnormal points are replaced using a sliding window midpoint filtering method, with the window length determined based on the main period of the arterial pressure waveform. After abnormal value correction, the phase consistency of each signal channel is checked again using real-time phase parameters. If the deviation of a channel's phase from the unified instantaneous phase value exceeds the allowable range, Hilbert transform is applied to reconstruct the analytical signal and perform phase rotation correction. The multi-source signal set after time synchronization verification and deviation correction is output as the post-execution state feedback data. It should be noted that time offset refers to the inconsistent recording time of the same physical event or phenomenon in each signal; for example, when changes in blood flow occur, there is a time difference between the sampling time of the core blood concentration signal and the sampling time of the auxiliary physiological signal.

[0031] S6. Evaluate the regulation effectiveness by calculating the recovery rate deviation ratio of the post-execution status feedback data, and integrate the regulation effectiveness and the post-execution status feedback data to obtain a regulation report; Based on the post-execution status feedback data, the recovery effect is quantified, and a preliminary performance evaluation value is determined. Furthermore, the core blood concentration signal, arterial pressure waveform from the auxiliary physiological signals, and flow output parameters from the equipment signals are extracted from the post-execution status feedback data. The variation amplitudes of the core blood concentration signal, arterial pressure waveform from the auxiliary physiological signals, and equipment signals are compared with the baseline values ​​(exemplary value range: 0-1) in the corresponding stable state in the historical adjustment report. The minimum mean square error criterion is used to calculate the recovery deviation. The recovery deviations of the core blood concentration signal, arterial pressure waveform from the auxiliary physiological signals, and equipment signals are weighted and summed to obtain the comprehensive recovery deviation value. The ratio of the time required for each signal in the post-execution status feedback data to reach a stable state to the main period of the arterial pressure waveform is calculated as the recovery time factor. The reciprocal of the comprehensive recovery deviation value is multiplied by the recovery time factor, and the output is the preliminary performance evaluation value. The preliminary performance assessment value is weighted and integrated with the post-implementation status feedback data to assess the impact of deviations on long-term stability and generate adjustment performance indicators. Furthermore, the fluctuation sequences of core blood concentration signals, arterial pressure waveforms in auxiliary physiological signals, and flow output parameters in equipment signals within the power stability range are extracted from the post-execution status feedback data. Empirical mode decomposition (EMD) is applied to each fluctuation sequence to extract low-frequency trend components to characterize long-term drift characteristics (e.g., blood flow velocity and blood pressure fluctuations). The cumulative deviation of each low-frequency trend component relative to the corresponding benchmark value in historical regulation reports is calculated as the long-term stability deviation. The long-term stability deviation is weighted and fused with the preliminary performance evaluation value using a weighted average method. The weighted fusion result is normalized using the minimum mean square error criterion, and the output is the regulation performance index. It should be noted that the weights of long-term stability deviation and preliminary performance evaluation value are determined by entropy weighting based on historical adjustment reports, with the weight of preliminary performance evaluation value (exemplary range: 0.4-0.7) and the weight of long-term stability deviation (exemplary range: 0.3-0.6). The adjustment report is obtained by integrating adjustment effectiveness indicators with post-implementation status feedback data by timestamp; Furthermore, the regulatory efficacy index is aligned one-to-one with the core blood concentration signal, auxiliary physiological signal, and equipment signal in the post-execution status feedback data according to their respective acquisition timestamps. For each timestamp, the regulatory efficacy index is used as a regulatory quality label for that time and attached to the corresponding post-execution status feedback data record. All labeled data records are organized into a time-series structured dataset in chronological order. A sliding window aggregation method is applied to the time-series structured dataset, with the window length set according to the main period of the arterial pressure waveform. The mean of the regulatory efficacy index and the statistical characteristics of each signal channel within the window are calculated. The aggregated time-series structured dataset is output in a standard data format to form a regulation report.

[0032] In summary, this invention achieves signal interference minimization and phase-locked integration under high-dimensional spatial mapping by eliminating signal-noise entanglement in standardized multi-source signal vectors and verifying shear stress values ​​to generate low-noise fused state vectors and real-time phase parameters. This ensures the pure characterization of blood flow dynamics and the assessment of safe zones during hemodialysis, and contributes to the reliable fusion of multi-source signals, thereby improving the accuracy and physiological compatibility of basic data for pump power regulation. Furthermore, by predicting the trajectory of circulating blood volume change rate based on the pulsation waveform parameter set and assessing potential hypotension risk to obtain predicted trajectories and risk indicators, this invention achieves dynamic risk quantification through the integral adjustment of extrapolated blood flow trends and historical deviations. This supports the construction and execution feedback integration of prospective power commands throughout the adaptive regulation cycle, enabling early intervention for potential physiological fluctuations, and ultimately optimizing patient blood flow stability and the sustainability of dialysis efficacy.

[0033] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for adaptive adjustment of the power of a hemodialysis circulation pump based on multi-source signals, characterized in that: include, Real-time acquisition and preprocessing of multi-source signal data to generate standardized multi-source signal vectors; Eliminate signal-noise entanglement in standardized multi-source signal vectors, verify shear stress values, and generate low-noise fused state vectors and real-time phase parameters; Feature extraction and pulsating waveform optimization are performed on the low-noise fused state vector and real-time phase parameters to generate a pulsating waveform parameter set; Predict the trajectory of the rate of change in circulating blood volume based on the set of pulsating waveform parameters, assess the potential risk of hypotension, and obtain the predicted trajectory and risk indicators. Power regulation commands are constructed based on predicted trajectories and risk indicators, and the power regulation commands are executed while real-time status feedback data is collected after execution. The regulation effectiveness is assessed by calculating the recovery rate deviation ratio of the post-execution status feedback data. The regulation effectiveness and the post-execution status feedback data are then integrated to obtain a regulation report.

2. The method for adaptive adjustment of hemodialysis circulation pump power based on multi-source signals as described in claim 1, characterized in that: The multi-source signal data includes core blood concentration signals, auxiliary physiological signals, and device signals; the preprocessing includes data cleaning and standardization.

3. The method for adaptive adjustment of hemodialysis circulation pump power based on multi-source signals as described in claim 2, characterized in that: The specific steps for eliminating signal-noise entanglement in the standardized multi-source signal vector, verifying shear stress values, and generating a low-noise fused state vector and real-time phase parameters are as follows. A high-dimensional spatial mapping is performed on the standardized multi-source signal vectors to generate a preliminary fusion state representation; Phase synchronization adjustment is performed on the initial fusion state representation, and blood flow velocity gradient is simulated to assess whether the shear stress is within the safe range, generating a verified fusion adjustment value; Using the verified fusion adjustment value as the phase reference benchmark, the final entanglement release and phase-locked integration of all signal channels are performed to obtain a low-noise fusion state vector and real-time phase parameters.

4. The method for adaptive adjustment of hemodialysis circulation pump power based on multi-source signals as described in claim 3, characterized in that: The specific steps for feature extraction and pulsating waveform optimization of the low-noise fused state vector and real-time phase parameters to generate a pulsating waveform parameter set are as follows. Based on low-noise fused state vectors and real-time phase parameters, dynamic blood flow components are captured and physiological rhythm changes are highlighted by empirical mode decomposition to determine the intermediate pulsation feature representation. The adjustment amplitude and frequency distribution represented by the simulated intermediate pulsation characteristics are analyzed, and phase locking is enhanced to smooth potential oscillations, resulting in a preliminary waveform optimization draft. The performance was verified and locked using the preliminary waveform optimization draft to obtain the pulsating waveform parameter set.

5. The method for adaptive adjustment of hemodialysis circulation pump power based on multi-source signals as described in claim 4, characterized in that: The prediction of the circulating blood volume change rate trajectory based on the pulsating waveform parameter set refers to inferring the potential blood pressure fluctuation path by linearly extrapolating the dynamic trend of blood flow based on the pulsating waveform parameter set, and obtaining a preliminary risk trajectory draft.

6. The method for adaptive adjustment of hemodialysis circulation pump power based on multi-source signals as described in claim 5, characterized in that: The specific steps for assessing potential hypotension risk and obtaining predicted trajectories and risk indicators are as follows. Based on the preliminary draft risk trajectory and historical adjustment reports, deviation integral adjustments are made, and the severity of the trajectory deviating from the safety boundary is assessed to generate a predicted trajectory; The predicted trajectory is risk-assessed and weighted to generate risk indicators.

7. The method for adaptive adjustment of hemodialysis circulation pump power based on multi-source signals as described in claim 6, characterized in that: The specific steps for constructing power adjustment commands based on predicted trajectories and risk indicators are as follows: Trend fitting is performed on the predicted trajectory and risk indicators to generate temporary power offset values; Based on the temporary power offset value, phase and frequency tuning are performed in conjunction with risk indicators, and the overall balance is verified to generate an optimized adjustment framework. The optimized regulation framework is integrated and locked to construct power regulation commands.

8. The method for adaptive adjustment of hemodialysis circulation pump power based on multi-source signals as described in claim 7, characterized in that: The specific steps for executing the power adjustment command and collecting real-time status feedback data after execution are as follows: Adjust the pump output power based on the power regulation command and calculate the instantaneous execution response value; Real-time acquisition of subsequent patient blood flow and physiological changes, combined with immediate execution response values ​​to assess the power stability range, and generating preliminary status feedback aggregation; The initial status feedback is synchronously verified and deviations are corrected to obtain post-execution status feedback data.

9. The method for adaptive adjustment of hemodialysis circulation pump power based on multi-source signals as described in claim 8, characterized in that: The method of evaluating the regulatory effectiveness by calculating the recovery rate deviation ratio of the post-execution state feedback data involves the following specific steps. Based on the post-execution status feedback data, the recovery effect is quantified, and a preliminary performance evaluation value is determined. The preliminary performance evaluation value is weighted and integrated with the post-implementation status feedback data to assess the impact of deviations on long-term stability and generate a regulation performance index.

10. The method for adaptive adjustment of hemodialysis circulation pump power based on multi-source signals as described in claim 9, characterized in that: The adjustment report is obtained by integrating adjustment performance indicators with post-implementation status feedback data by timestamp.

Images