Personalized nursing optimization method for gastrointestinal postoperative fast track recovery

Abstract

The application relates to the technical field of medical care, and discloses a personalized nursing optimization method for rapid recovery after gastrointestinal surgery, which takes an infusion pump driving reference signal as a noise source input, performs adaptive filtering on an original acoustic signal to output purified borborygmus; a physiological-acoustic coupling index is calculated in combination with body surface electrocardio characteristics to quantize the synchronization degree of mechanical peristalsis and nerve innervation; when the coupling index confidence is insufficient, a pseudo-random fluid pulse excitation is applied for system identification, nerve reflex sensitivity and reflection response time delay parameters are extracted through cross-correlation analysis; a multi-dimensional state vector containing the above index and parameters is constructed, and a gain scheduling control algorithm is used to dynamically adjust the infusion flow rate. Through source reference adaptive filtering and active fluid pulse identification, the application accurately evaluates the nerve innervation state, and realizes closed-loop flow rate control adapting to individual recovery processes by using a gain scheduling algorithm.

Description

Technical Field

[0001] This invention relates to the field of medical and nursing technology, specifically to a personalized nursing optimization method for rapid recovery after gastrointestinal surgery. Background Technology

[0002] Following gastrointestinal surgery, early enteral nutritional support is crucial for promoting bowel function recovery and improving patient prognosis. However, postoperative patients commonly experience varying degrees of gastrointestinal motility disorders (such as postoperative paralytic ileus). Too rapid a nutrient infusion rate can easily lead to abdominal distension, reflux, and even aspiration pneumonia, while too slow a rate cannot meet metabolic needs and delays recovery. Therefore, developing a nursing optimization system that can accurately sense the patient's real-time bowel function status and adaptively adjust the infusion rate in a closed-loop manner has become a key technological requirement for accelerating postoperative recovery.

[0003] Existing intelligent nutrient infusion systems typically employ a constant-rate infusion or a timed, incremental infusion pattern. Some advanced devices are beginning to incorporate non-invasive electronic bowel sound monitoring technology to assist in assessing intestinal motility. This technology generally uses piezoelectric ceramic or electret microphones attached to the abdominal wall to collect abdominal acoustic signals. Conventional signal processing techniques such as bandpass filters, threshold segmentation, or wavelet transforms are used to filter out ambient background noise. The number of bowel sound bursts or their energy values ​​per unit time are used as evaluation indicators, and the infusion pump's operating rate is adjusted accordingly.

[0004] While existing technologies have achieved some degree of automation infusion and preliminary monitoring, several shortcomings remain: First, the mechanical vibration noise and electromagnetic interference generated by the motor and its transmission mechanism during infusion pump operation often highly overlap with the bowel sound signals, which are mainly distributed between 100Hz and 500Hz. Conventional filtering methods based on frequency domain separation struggle to effectively eliminate near-field mechanical noise in the same frequency band while preserving the characteristics of weak bowel sounds, resulting in a low signal-to-noise ratio and making misjudgments highly susceptible to motor noise interference. Second, passive monitoring based solely on acoustic events lacks a deep assessment of the intestinal nerve innervation state, making it difficult to distinguish between mechanical noise caused by changes in body position or gas displacement. False bowel sounds are different from genuine physiological peristalsis controlled by the vagus nerve. Furthermore, passive monitoring cannot acquire data when the intestine is in a quiescent state during the early stages of functional recovery, preventing the system from actively detecting the intestine's potential stress response capabilities. Finally, existing flow rate control strategies mostly use fixed-gain PID algorithms or simple logic rules, ignoring the significant differences in nerve reflex sensitivity and reflex response delay between different patients and at different stages of recovery within the same patient. It is difficult to adapt to this nonlinear and time-varying physiological characteristic with a single fixed control parameter, and the mismatch between the control algorithm and the physiological rhythm can easily lead to infusion regulation lag or overshoot, thereby causing intolerance complications. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a personalized nursing optimization method for rapid postoperative recovery of gastrointestinal patients. This method solves the problems of existing technologies, such as the difficulty in accurately extracting bowel sound signals in the noisy environment of infusion pump operation, and the inability to actively identify the patient's state and adaptively adjust the flow rate based on the individualized neural reflex characteristics of the patient.

[0006] To achieve the above objectives, the present invention provides a personalized nursing optimization method for rapid recovery after gastrointestinal surgery, comprising the following steps: The original acoustic signal from the patient's abdomen, the electrocardiogram signal from the body surface, and the drive reference signal of the infusion pump motor are acquired simultaneously. The drive reference signal is used as the noise source reference input, and adaptive filtering is performed on the original acoustic signal to output the purified bowel sound signal. The acoustic energy characteristics of the purified bowel sound signal and the vagal tone characteristics of the surface electrocardiogram signal were extracted, and the physiological-acoustic coupling index that quantifies the synchronization between mechanical peristalsis and nerve innervation was calculated. Determine whether the physiological-acoustic coupling index meets the preset confidence condition. If not, generate a pseudo-random binary sequence to apply fluid pulse excitation to the patient's gastrointestinal tract, and calculate the cross-correlation function between the fluid pulse excitation and the vagal nerve tension response, and extract the nerve reflex sensitivity parameter and reflex response delay parameter. A multidimensional state vector is constructed, which includes the physiological-acoustic coupling index, the neural reflex sensitivity parameter, and the reflex response delay parameter. The flow rate adjustment is calculated using a gain scheduling control algorithm, and the target flow rate of the infusion pump is updated according to the flow rate adjustment.

[0007] Preferably, the specific steps of the adaptive filtering process include: The signal characterizing the changes in motor torque and speed is acquired as the drive reference signal, and the drive reference signal is downsampled and aligned with the time axis. An adaptive filter is constructed, and the driving reference signal is processed using the weight coefficient vector to obtain an estimate of the mechanically conducted noise. Calculate the residual signal after subtracting the estimated value from the original acoustic signal, and use the residual signal as the purified bowel sound signal; Based on the instantaneous value of the residual signal, the weight coefficient vector is iteratively updated in real time along the negative gradient direction of the performance surface using the least mean square algorithm.

[0008] Preferably, the calculation steps of the physiological-acoustic coupling index include: The purified bowel sound signal is subjected to a short-time Fourier transform, and the ratio of the energy integral value in the effective frequency band to the total energy integral value in the entire frequency band is calculated, which is defined as the acoustic energy characteristic. R-wave detection and power spectral density analysis were performed on the surface electrocardiogram signal to extract high-frequency power and low-frequency power. The proportion of high-frequency power in the sum of low-frequency power and high-frequency power was calculated and defined as the high-frequency component proportion characteristic of vagal tone. Aligning the acoustic energy characteristics with the high-frequency component proportion characteristics on the time axis is used to calculate the physiological-acoustic coupling index.

[0009] Preferably, the physiological-acoustic coupling index, which quantifies the synchronization between mechanical peristalsis and neural innervation, is calculated using a weighted fusion model: The acoustic energy characteristics are linearly weighted and summed with the high-frequency component proportion characteristics after nonlinear mapping. The nonlinear mapping employs a Sigmoid function, which includes a slope factor for adjusting the sensitivity of the neural feature mapping and a center offset for aligning the center symmetry points. If the weighted summation result exceeds the preset numerical range, boundary clamping is performed.

[0010] Preferably, determining whether the physiological-acoustic coupling index meets a preset confidence condition specifically includes: Set a coupling confidence threshold; If the physiological-acoustic coupling index is less than or equal to the coupling confidence threshold, the confidence level of the current passive monitoring data is determined to be insufficient, triggering the subsequent step of generating a pseudo-random binary sequence. If the physiological-acoustic coupling index is greater than the coupling confidence threshold, it is determined to be a valid physiological peristalsis, and the current multidimensional state vector is directly used to enter the step of calculating the flow rate adjustment using the gain scheduling control algorithm.

[0011] Preferably, the generation of the pseudo-random binary sequence to apply fluid pulse stimulation to the patient's gastrointestinal tract specifically includes: A pseudo-random binary sequence with white noise autocorrelation characteristics is generated using a linear feedback shift register; The pseudo-random binary sequence is mapped to the perturbation amplitude of the flow rate fluctuation and superimposed on the base flow rate of the infusion pump to generate a fluid pulse excitation signal to drive the infusion pump to run. The perturbation amplitude is set to a value within the sub-sensing threshold.

[0012] Preferably, the specific steps for extracting the neural reflex sensitivity parameters and reflex response delay parameters include: The proportion of high-frequency components characterizing vagal tone is continuously monitored as the system's response output variable; The DC component is eliminated by removing the time average values ​​of the fluid pulse excitation signal and the high-frequency component proportion characteristics, respectively. Calculate the cross-correlation function between the fluid pulse excitation signal after DC removal and the high-frequency component proportion characteristic; The maximum peak value of the cross-correlation function is searched within a preset physiological reasonable time delay interval. The amplitude of the maximum peak value is defined as the neural reflex sensitivity parameter, and the time lag corresponding to the occurrence of the maximum peak value is defined as the reflex response delay parameter.

[0013] Preferably, the construction of the multidimensional state vector including the physiological-acoustic coupling index, the neural reflex sensitivity parameter, and the reflex response delay parameter further includes: If the current cycle does not trigger the step of applying fluid pulse stimulation to the patient's gastrointestinal tract, or if the extracted neural reflex sensitivity parameter is lower than the preset validity threshold, the moving average of the historical valid identification parameters recorded in the memory is used to fill the multidimensional state vector. If the historical data becomes invalid for more than a preset time limit, the multidimensional state vector is filled with preset standard physiological model parameters.

[0014] Preferably, the specific steps for calculating the flow rate adjustment using the gain scheduling control algorithm include: A discrete-time proportional-derivative control algorithm is used to calculate the deviation between the physiological-acoustic coupling index and the preset ideal coupling target value. The variable-scalar gain coefficient is calculated in real time based on the multidimensional state vector. The deviation is weighted using the variable proportional gain coefficient, and the flow rate adjustment for the current control cycle is calculated by combining the rate of change of the physiological-acoustic coupling index with the differential gain coefficient.

[0015] Preferably, the real-time calculation of the variable-scalar gain coefficient based on the multidimensional state vector is achieved through a function mapping that includes a sensitivity enhancement term and a time delay penalty term: Based on the fundamental gain constant; Using the aforementioned sensitivity enhancement term, the variable scaling gain coefficient and the neural reflex sensitivity parameter are set to have a positive correlation. Using the time delay penalty term, the deviation between the reflection response time delay parameter and the ideal physiological reflection time delay constant is calculated, and the variable proportional gain coefficient is reduced as the deviation increases by using the exponential decay characteristic of the Gaussian function.

[0016] This invention provides a personalized and optimized nursing care method for rapid recovery after gastrointestinal surgery. It has the following beneficial effects: 1. This invention introduces the reference signal of the infusion pump motor drive as the noise source input for adaptive filtering, and utilizes the physical causality between the motor torque signal and mechanically conducted noise to iteratively update the filter weight coefficients in real time. This method can accurately subtract the dynamic noise component related to motor operation from the original acoustic signal, solving the problem that traditional bandpass filtering is difficult to separate mechanical noise in the same frequency band, and ensuring high signal-to-noise ratio extraction of weak intestinal humming signals under continuous operation of the infusion pump.

[0017] 2. This invention constructs a physiological-acoustic coupling index that integrates acoustic energy and vagal nerve tension, and actively applies pseudo-random fluid pulse excitation for system identification when passive monitoring confidence is insufficient. This strategy combining passive monitoring and active detection not only quantifies the synchronization between mechanical peristalsis and nerve innervation, but also extracts nerve reflex sensitivity and time delay parameters through cross-correlation analysis, effectively eliminating false peristaltic interference and achieving accurate grading and assessment of the patient's gastrointestinal functional status.

[0018] 3. This invention utilizes a gain scheduling control algorithm based on multidimensional state vectors to dynamically adjust the infusion rate according to the real-time monitored coupling index and identified neural reflex characteristics. This method enhances the control gain by leveraging neural reflex sensitivity and applies attenuation penalties based on reflex response delay, enabling the infusion rate adjustment logic to adaptively match the intestinal nerve recovery process of different patients. This maximizes nutrient absorption efficiency while ensuring infusion safety, achieving truly personalized closed-loop nursing care. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the signal purification process based on source-reference adaptive filtering according to the present invention; Figure 3 This is a schematic diagram of the physiological-acoustic coupling feature extraction process under passive state according to the present invention; Figure 4 This is a schematic diagram of the system identification and active detection process based on PRBS modulation according to the present invention; Figure 5 This is a schematic diagram of the adaptive flow rate control process based on multidimensional state space according to the present invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see the appendix Figure 1 , Figure 1 This is a schematic flowchart of a method according to an embodiment of the present invention. The present invention provides a personalized nursing optimization method for rapid recovery after gastrointestinal surgery, comprising the following steps: S100 synchronously acquires the patient's abdominal acoustic signal, surface electrocardiogram signal, and infusion pump motor drive reference signal. Using the drive reference signal as the noise source reference input, it performs adaptive filtering on the raw acoustic signal. Through a minimum error algorithm, it iterates the filter weight coefficients in real time, subtracts the component related to the drive reference signal from the raw acoustic signal, eliminates the mechanical noise generated by the motor operation, and outputs the purified bowel sound signal. S200: Perform short-time Fourier transform on the output purified bowel sound signal to extract acoustic energy features within a preset frequency band. Simultaneously, perform R-wave detection and power spectral density analysis on the surface electrocardiogram signal, calculate the proportion of high-frequency components characterizing vagal nerve tension, calculate the degree of synchronous coupling between acoustic energy features and the proportion of high-frequency components on the time axis, generate a physiological-acoustic coupling index, and determine whether the physiological-acoustic coupling index meets the preset confidence conditions. If it does, proceed directly to step S400; otherwise, trigger step S300. S300 generates a pseudo-random binary sequence with a preset period and maps the sequence to a flow rate fluctuation command superimposed on the base flow rate of the infusion pump to apply fluid pulse excitation to the patient's gastrointestinal tract. It continuously monitors the dynamic response of the proportion of high-frequency components to the fluid pulse excitation, calculates the cross-correlation function between the fluid pulse excitation sequence and the high-frequency component proportion sequence, obtains the system's pulse response characteristics, analyzes the waveform of the cross-correlation function, and extracts the nerve reflex sensitivity parameter and the reflex response delay parameter. The nerve reflex sensitivity parameter is defined as the peak amplitude of the cross-correlation function, and the reflex response delay parameter is defined as the time lag corresponding to the peak. S400 constructs a multidimensional state vector containing physiological-acoustic coupling index, neural reflex sensitivity parameter, and reflex response delay parameter. Based on this multidimensional state vector, the flow rate adjustment amount for the current control cycle is calculated using a gain scheduling control algorithm. The target flow rate of the infusion pump is updated according to the flow rate adjustment amount, thereby realizing closed-loop adaptive adjustment of the infusion rate.

[0022] Please see the appendix Figure 2 , Figure 2This is a schematic diagram of a signal purification process based on source-reference adaptive filtering according to an embodiment of the present invention. During enteral nutrition infusion, in order to accurately extract weak bowel sound signals in an environment where the infusion pump motor continuously operates, generating mechanical noise and electromagnetic interference, this embodiment employs full-duplex signal acquisition and adaptive noise cancellation technology. This process specifically includes the following sub-steps: The S110 simultaneously acquires raw acoustic signals and drive reference signals. A piezoelectric sensor attached to the patient's abdomen collects mixed acoustic data, including target bowel sounds, mechanical conduction noise, and ambient background noise, as the raw acoustic signal. Simultaneously, signals characterizing changes in motor torque and speed are acquired in real-time via the current shunt sampling resistor of the infusion pump motor drive circuit or the low-level instruction interface of the digital controller, serving as the drive reference signal.

[0023] Since mechanically conducted noise originates from the physical rotation of the motor, there is a physical causal relationship and linear transmission between the drive reference signal and the mechanically conducted noise mixed into the original acoustic signal, satisfying the theoretical premise of adaptive noise cancellation. If the sampling rate of the drive reference signal is higher than that of the original acoustic signal, it needs to be downsampled beforehand, and the time delay introduced by the physical acoustic conduction from the motor to the sensor should be corrected based on cross-correlation analysis to ensure that its time axis is strictly aligned with the original acoustic signal.

[0024] S120: Construct an adaptive filter to estimate mechanically conducted noise. This filter can employ a transverse finite impulse response structure or a nonlinear filter structure, such as a Volterra series filter. The filter order and weight coefficient vector are set. Using the current weight coefficient vector, convolution or nonlinear mapping operations are performed on the input driving reference signal sequence to calculate the filter output value. This filter output value, in a physical sense, is the estimate of the mechanically conducted noise component mixed into the original acoustic signal at the current moment.

[0025] S130, Calculate the error signal after adaptive cancellation. Subtract the filter output value obtained in step S120 from the synchronously acquired original acoustic signal to obtain the residual signal. In the source-reference noise cancellation architecture of this embodiment, this residual signal is the purified acoustic signal after removing the relevant mechanical noise component, which mainly retains the target intestinal sound component unrelated to motor operation and unrelated environmental background noise.

[0026] In step S140, the filter weight coefficients are iteratively updated using the least mean square algorithm. To ensure that the noise estimate output by the filter approximates the actual mechanically conducted noise as closely as possible, the system dynamically adjusts the filter weight coefficient vector along the negative gradient direction of the performance surface based on the instantaneous value of the residual signal calculated in step S130. Through iterative iteration, the filter can automatically track and adapt to changes in the noise characteristics of the injection pump motor caused by variations in flow rate settings or fluid load fluctuations, ensuring that the output purified acoustic signal maintains a high signal-to-noise ratio during full-speed operation of the injection pump. The computation process is a well-known technique in the field of adaptive signal processing and will not be elaborated further here.

[0027] Please see the appendix Figure 3 , Figure 3 This is a schematic diagram of the physiological-acoustic coupling feature extraction process under passive state according to an embodiment of the present invention. This embodiment constructs a physiological-acoustic coupling index that integrates mechanical peristalsis features and neural innervation features. Its calculation process specifically includes the following sub-steps: S210, Extract the frequency domain energy characteristics of the purified bowel sound signal. Perform a short-time Fourier transform on the purified bowel sound signal output from the previous step, using a Hamming or Hanning window as the truncation function, and setting the overlap rate of adjacent time windows to 50% to 75%, converting the one-dimensional time-domain signal into a two-dimensional time-frequency distribution matrix. Based on the physiological acoustic characteristics of bowel sounds, 100Hz to 500Hz is selected as the effective frequency band where the first formant is located. Calculate the ratio of the energy integral value within this effective frequency band to the total energy integral value across the entire frequency band, and define this ratio as the acoustic energy characteristic. The specific calculation process of the short-time Fourier transform and frequency domain energy integration is a well-known technique in the field of digital signal processing and will not be elaborated here.

[0028] S220 extracts vagal tone features from surface electrocardiogram (ECG) signals. First, anti-interference preprocessing is performed on the synchronously acquired surface ECG signals. A notch filter is used to remove power frequency interference, and morphological filtering or wavelet transform is employed to suppress electromagnetic pulse noise introduced by the infusion pump motor. Subsequently, the purified ECG signals are analyzed for R-wave peak detection to construct an RR interval time series.

[0029] Since the RR interval sequence is a non-uniformly sampled signal, it needs to be resampled into a uniform time series with a frequency of 2Hz to 4Hz using cubic spline interpolation, followed by power spectral density analysis. Based on the physiological principles of the autonomic nervous system, high-frequency power in the 0.15Hz to 0.40Hz band and low-frequency power in the 0.04Hz to 0.15Hz band are extracted. The proportion of high-frequency power in the sum of low-frequency and high-frequency power is calculated, and this proportion is defined as the high-frequency component proportion characteristic representing vagal tone.

[0030] S230, Calculate the physiological-acoustic coupling index. Align the acoustic energy features obtained in step S210 with the high-frequency component proportion features obtained in step S220 on the time axis. Construct a weighted fusion model to quantify the degree of synchronous coupling between mechanical peristalsis and neural innervation. The formula for calculating the physiological-acoustic coupling index is as follows: ; In the formula, For a moment The calculated physiological-acoustic coupling index is used to quantify the degree of synchronization between mechanical peristalsis and nerve innervation, with a value ranging from 0 to 1. It is a continuous-time variable; In this embodiment, the preset weighting coefficients are used. The value is 0.6, which is used to adjust the weight of acoustic energy characteristics in the overall evaluation system; For a moment The extracted acoustic energy characteristics are the percentage of energy of the purified bowel sound signal within the effective frequency band. In this embodiment, the complementary weight corresponding to the neural feature part is set to 0.4; is the base of the natural logarithm; β is the slope factor of the Sigmoid function, used to adjust the sensitivity of neural feature mapping. The larger the value, the greater the sensitivity. In this embodiment, β is set to 10. For a moment The calculated high-frequency component ratio characteristic, i.e. the ratio of high-frequency power in the total power of the electrocardiogram signal on the body surface, is used to characterize vagal tone. This is the center offset of the Sigmoid function, used to align the intermediate value of the high-frequency component proportion feature to the center symmetric point of the nonlinear mapping function.

[0031] This formula maps the intensity of acoustic events to the activity of neural innervation under the same dimension through nonlinear mapping. If the calculation result exceeds the defined range of 0 to 1 due to transient fluctuations in the signal, the system will perform boundary clamping processing to force its value to be limited to the range of 0 to 1.

[0032] S240, Perform confidence assessment based on the coupling index. Set a coupling confidence threshold and compare the calculated physiological-acoustic coupling index with this threshold. If the physiological-acoustic coupling index is greater than the threshold, it indicates that the currently detected acoustic signal is accompanied by significant vagal nerve excitation, and is judged as effective physiological peristalsis, directly proceeding to the flow rate control step; if the physiological-acoustic coupling index is less than or equal to the threshold, it indicates a lack of evidence of neural innervation, suggesting possible pseudo-peristalsis or an inhibited state of intestinal function, and the system determines that the current passive monitoring data has insufficient confidence, triggering the subsequent active identification step.

[0033] Please see the appendix Figure 4 , Figure 4 This is a schematic diagram of a system identification and active detection process based on PRBS modulation according to an embodiment of the present invention. When passive monitoring cannot clearly determine the intestinal functional state, this embodiment employs a system identification method to actively detect the neural conduction characteristics of the gastrointestinal tract by applying a fluid pulse sequence with specific statistical characteristics and analyzing the resulting neural reflex response. This process specifically includes the following sub-steps: S310 generates a pseudo-random binary sequence and maps it to fluid control commands. A linear feedback shift register is used to generate a maximum-length sequence as the pseudo-random binary sequence. During the generation process, primitive polynomials of order 5 to 7 are selected as feedback logic to produce a binary code stream with a moderate period length and white noise-like autocorrelation characteristics.

[0034] The binary sequence is mapped into the flow rate control domain of the injection pump to generate a fluid pulse excitation signal superimposed on the base flow rate. The calculation formula is as follows: ; In the formula, Represents the instantaneous value of the fluid pulse excitation signal; Indicates the baseline infusion rate; The perturbation amplitude representing the flow rate fluctuation is typically set to 3 ml / h to 5 ml / h; The value is the mapped pseudo-random sequence value, which can be either -1 or +1.

[0035] S320 performs synchronous monitoring under fluid pulse excitation. It controls the infusion pump to operate according to the generated fluid pulse excitation signal, applying transient volume expansion stimulation to the patient's gastrointestinal tract. Simultaneously, it continuously acquires and calculates the patient's vagal tone characteristics (i.e., the aforementioned high-frequency component proportion characteristics) as the system's response output variable.

[0036] S330 calculates the cross-correlation function between the excitation and response signals. To eliminate the influence of baseline drift on the correlation calculation, the time averages of the fluid impulse excitation signal and the vagal nerve tension characteristic sequence are first calculated separately, and then subtracted from each sequence to remove the DC component. Subsequently, a cross-correlation operation is performed on the two sets of sequences after DC removal, i.e., the integral value of the product of one sequence and the time-shifted other sequence within the observation window is calculated. This operation utilizes the impulse-like autocorrelation characteristics of pseudo-random sequences, allowing the calculated cross-correlation function waveform to directly approximate the impulse response characteristics of the system to a unit fluid impulse. The specific mathematical principles of the cross-correlation operation are well-known techniques in the field of signals and systems, and will not be elaborated here.

[0037] S340, extract system identification parameters from the cross-correlation function. Analyze the cross-correlation function curve obtained in step S330 to extract two key parameters characterizing the functional state of the gut nervous system: Firstly, neural reflex sensitivity, defined as the maximum peak amplitude of the cross-correlation function within a preset physiological time lag interval, quantifies the response gain of the enteric nervous system to mechanical fluid stimulation. The calculation formula is as follows: ; In the formula, Indicates the sensitivity of nerve reflexes; Indicates time delay as The cross-correlation function value at time; For time-lag variables in the calculation of cross-correlation functions; This indicates a preset, physiologically reasonable time delay range, for example, set to 2 to 60 seconds; The lower limit of the preset physiologically reasonable time delay interval is set to 2 seconds to filter out non-causal artifacts caused by transient electromagnetic interference; The upper limit of the preset physiological reasonable time delay interval is set to 60 seconds, which covers the maximum physiological time of slow neural reflex arc.

[0038] Secondly, the reflex response delay is defined as the time lag corresponding to the occurrence of the maximum peak value mentioned above. This parameter characterizes the time consumed in the conduction pathway from the occurrence of mechanical stimulation to the initiation of neural regulation. The calculation formula is as follows: ; In the formula, This is the reflection response delay; The maximum value operator represents selecting the value of the independent variable that maximizes the subsequent function value; For time-lag variables in the calculation of cross-correlation functions; The lower limit of the preset physiologically reasonable time delay interval is set to 2 seconds to filter out non-causal artifacts caused by transient electromagnetic interference; The upper limit of the preset physiological reasonable time delay interval is set to 60 seconds, which covers the maximum physiological time of slow neural reflex arc; For time delay The cross-correlation function value at that time. If the calculated neural reflex sensitivity is lower than the preset noise floor threshold, it is determined that there is no significant neural reflex at present, and the system outputs the corresponding status flag.

[0039] Please see the appendix Figure 5 , Figure 5This is a schematic diagram of an adaptive flow rate control process based on a multidimensional state space according to an embodiment of the present invention. This embodiment constructs a multidimensional state space including passive monitoring features and active identification features, and uses a gain scheduling strategy to dynamically adjust the infusion flow rate to adapt to the individual differences in the patient's intestinal recovery process. The process specifically includes the following sub-steps: S410, Construct a multidimensional state vector for feedback control. At the beginning of each control cycle, read the latest state data, which includes the physiological-acoustic coupling index calculated in the previous steps, and the neural reflex sensitivity parameter and reflex response delay parameter extracted during the most recent active identification process. Combine these three parameters to form the multidimensional state vector at the current moment. If active identification is not triggered in the current cycle, or if active identification is triggered but the extracted neural reflex sensitivity is lower than the preset validity threshold, then the moving average of the historical valid identification parameters recorded in the memory is called. If historical data is missing or expired for more than 30 minutes, then the preset standard physiological model parameters are forcibly used to fill the gaps: the standard neural reflex sensitivity is set to 0.5, and the standard reflex response delay is set to 20 seconds, to maintain the integrity and continuity of the multidimensional state vector.

[0040] S420 calculates the flow rate adjustment based on gain scheduling. A discrete-time proportional-derivative (PDD) control algorithm is used to calculate the change in infusion flow rate. The proportional gain coefficient is not a fixed constant, but rather a nonlinear function value calculated in real-time based on a multidimensional state vector. The formula for calculating the flow rate adjustment is as follows: ; In the formula, For the first The flow rate adjustment per control cycle, typically expressed in ml / h; This is the index of the current discrete control cycle; For dependent on multidimensional state vectors The variable proportional gain coefficient, which changes dynamically with the patient's condition; For the first Physiological-acoustic coupling index measured over a control cycle; The preset differential gain coefficient is used to suppress oscillations during the adjustment process. In this embodiment... The value is 2.0; For the first Physiological-acoustic coupling index measured in each control cycle (i.e., the previous cycle); This represents the preset ideal coupling target value, in this embodiment... The value is 0.8; This indicates the time interval of the control cycle, i.e., the time difference between two adjustments.

[0041] S430, solve for the function mapping of the variable-scalar gain coefficient. Construct a gain scheduling function that includes a sensitivity enhancement term and a time delay penalty term. Dynamically modulate the base gain according to the neural reflex sensitivity parameter and the reflex response time delay parameter. Its calculation formula is: ; In the formula, This is the currently calculated variable gain coefficient; Represents the fundamental gain constant; This indicates the current sensitivity parameter of the neural reflex; This represents the normalized reference sensitivity, which is the average reflex sensitivity of a standard healthy population under the same stimulus. It is an exponential function; This represents the current reflection response delay parameter; The ideal physiological reflex delay constant is typically set to 15 to 30 seconds. The time delay tolerance variance parameter controls the decay width of the Gaussian function, corresponding to a time deviation tolerance of 5 to 10 seconds. This formula, through a combination of linear product terms and Gaussian exponent terms, achieves the adjustment logic of increasing gain when neural reflex sensitivity increases, and rapidly reducing gain through exponential decay when the reflex delay deviates from the ideal range.

[0042] S440 performs flow rate update and safety limiting. The target flow rate of the infusion pump is updated using the calculated flow rate adjustment. Before outputting the final command, a safety limiting process is performed on the target flow rate, i.e., it is determined whether the updated target flow rate is within the preset range defined by the minimum sustaining flow rate and the maximum safe flow rate. If the calculation result exceeds the boundary of this range, the corresponding boundary value is directly used as the final execution command to prevent the infusion rate from becoming too fast or stopping completely due to control algorithm overshoot.

[0043] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various 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 personalized nursing optimization method for rapid recovery after gastrointestinal surgery, characterized in that, Includes the following steps: The original acoustic signal from the patient's abdomen, the electrocardiogram signal from the body surface, and the drive reference signal of the infusion pump motor are acquired simultaneously. The drive reference signal is used as the noise source reference input, and adaptive filtering is performed on the original acoustic signal to output the purified bowel sound signal. The acoustic energy characteristics of the purified bowel sound signal and the vagal tone characteristics of the surface electrocardiogram signal were extracted, and the physiological-acoustic coupling index that quantifies the synchronization between mechanical peristalsis and nerve innervation was calculated. Determine whether the physiological-acoustic coupling index meets the preset confidence condition. If not, generate a pseudo-random binary sequence to apply fluid pulse excitation to the patient's gastrointestinal tract, and calculate the cross-correlation function between the fluid pulse excitation and the vagal nerve tension response, and extract the nerve reflex sensitivity parameter and reflex response delay parameter. A multidimensional state vector is constructed, which includes the physiological-acoustic coupling index, the neural reflex sensitivity parameter, and the reflex response delay parameter. The flow rate adjustment is calculated using a gain scheduling control algorithm, and the target flow rate of the infusion pump is updated according to the flow rate adjustment.

2. The personalized nursing optimization method for rapid postoperative recovery of gastrointestinal patients according to claim 1, characterized in that, The specific steps of the adaptive filtering process include: The signal characterizing the changes in motor torque and speed is acquired as the drive reference signal, and the drive reference signal is downsampled and aligned with the time axis. An adaptive filter is constructed, and the driving reference signal is processed using the weight coefficient vector to obtain an estimate of the mechanically conducted noise. Calculate the residual signal after subtracting the estimated value from the original acoustic signal, and use the residual signal as the purified bowel sound signal; Based on the instantaneous value of the residual signal, the weight coefficient vector is iteratively updated in real time along the negative gradient direction of the performance surface using the least mean square algorithm.

3. The personalized nursing optimization method for rapid postoperative recovery of gastrointestinal patients according to claim 1, characterized in that, The calculation steps for the physiological-acoustic coupling index include: The purified bowel sound signal is subjected to a short-time Fourier transform, and the ratio of the energy integral value in the effective frequency band to the total energy integral value in the entire frequency band is calculated, which is defined as the acoustic energy characteristic. R-wave detection and power spectral density analysis were performed on the surface electrocardiogram signal to extract high-frequency power and low-frequency power. The proportion of high-frequency power in the sum of low-frequency power and high-frequency power was calculated and defined as the high-frequency component proportion characteristic of vagal tone. Aligning the acoustic energy characteristics with the high-frequency component proportion characteristics on the time axis is used to calculate the physiological-acoustic coupling index.

4. The personalized nursing optimization method for rapid postoperative recovery of gastrointestinal patients according to claim 3, characterized in that, The physiological-acoustic coupling index, which quantifies the synchronization between mechanical peristalsis and neural innervation, is calculated using a weighted fusion model: The acoustic energy characteristics are linearly weighted and summed with the high-frequency component proportion characteristics after nonlinear mapping. The nonlinear mapping employs a Sigmoid function, which includes a slope factor for adjusting the sensitivity of the neural feature mapping and a center offset for aligning the center symmetry points. If the weighted summation result exceeds the preset numerical range, boundary clamping is performed.

5. The personalized nursing optimization method for rapid postoperative recovery of gastrointestinal patients according to claim 1, characterized in that, The determination of whether the physiological-acoustic coupling index meets the preset confidence condition specifically includes: Set a coupling confidence threshold; If the physiological-acoustic coupling index is less than or equal to the coupling confidence threshold, the confidence level of the current passive monitoring data is determined to be insufficient, triggering the subsequent step of generating a pseudo-random binary sequence. If the physiological-acoustic coupling index is greater than the coupling confidence threshold, it is determined to be a valid physiological peristalsis, and the current multidimensional state vector is directly used to enter the step of calculating the flow rate adjustment using the gain scheduling control algorithm.

6. The personalized nursing optimization method for rapid postoperative recovery of gastrointestinal patients according to claim 1, characterized in that, The process of generating a pseudo-random binary sequence to apply fluid pulse stimulation to the patient's gastrointestinal tract specifically includes: A pseudo-random binary sequence with white noise autocorrelation characteristics is generated using a linear feedback shift register; The pseudo-random binary sequence is mapped to the perturbation amplitude of the flow rate fluctuation and superimposed on the base flow rate of the infusion pump to generate a fluid pulse excitation signal to drive the infusion pump to run. The perturbation amplitude is set to a value within the sub-sensing threshold.

7. The personalized nursing optimization method for rapid postoperative recovery of gastrointestinal patients according to claim 1, characterized in that, The specific steps for extracting the neural reflex sensitivity parameters and reflex response delay parameters include: The proportion of high-frequency components characterizing vagal tone is continuously monitored as the system's response output variable; The DC component is eliminated by removing the time average values ​​of the fluid pulse excitation signal and the high-frequency component proportion characteristics, respectively. Calculate the cross-correlation function between the fluid pulse excitation signal after DC removal and the high-frequency component proportion characteristic; The maximum peak value of the cross-correlation function is searched within a preset physiological reasonable time delay interval. The amplitude of the maximum peak value is defined as the neural reflex sensitivity parameter, and the time lag corresponding to the occurrence of the maximum peak value is defined as the reflex response delay parameter.

8. The personalized nursing optimization method for rapid postoperative recovery of gastrointestinal patients according to claim 1, characterized in that, The construction of the multidimensional state vector, which includes the physiological-acoustic coupling index, the neural reflex sensitivity parameter, and the reflex response delay parameter, further includes: If the current cycle does not trigger the step of applying fluid pulse stimulation to the patient's gastrointestinal tract, or if the extracted neural reflex sensitivity parameter is lower than the preset validity threshold, the moving average of the historical valid identification parameters recorded in the memory is used to fill the multidimensional state vector. If the historical data becomes invalid for more than a preset time limit, the multidimensional state vector is filled with preset standard physiological model parameters.

9. The personalized nursing optimization method for rapid postoperative recovery of gastrointestinal patients according to claim 1, characterized in that, The specific steps for calculating the flow rate adjustment using the gain scheduling control algorithm include: A discrete-time proportional-derivative control algorithm is used to calculate the deviation between the physiological-acoustic coupling index and the preset ideal coupling target value. The variable-scalar gain coefficient is calculated in real time based on the multidimensional state vector. The deviation is weighted using the variable proportional gain coefficient, and the flow rate adjustment for the current control cycle is calculated by combining the rate of change of the physiological-acoustic coupling index with the differential gain coefficient.

10. The personalized nursing optimization method for rapid postoperative recovery of gastrointestinal patients according to claim 9, characterized in that, The real-time calculation of the variable-scalar gain coefficient based on the multidimensional state vector is achieved through a function mapping that includes a sensitivity enhancement term and a time delay penalty term: Based on the fundamental gain constant; Using the aforementioned sensitivity enhancement term, the variable scaling gain coefficient and the neural reflex sensitivity parameter are set to have a positive correlation. Using the time delay penalty term, the deviation between the reflection response time delay parameter and the ideal physiological reflection time delay constant is calculated, and the variable proportional gain coefficient is reduced as the deviation increases by using the exponential decay characteristic of the Gaussian function.

Images