A method and system for regulating enteral nutrition delivery based on real-time feedback of intragastric pressure

Abstract

This disclosure belongs to the field of medical device and clinical monitoring technology. Specifically, it discloses a method and system for regulating enteral nutrition infusion based on real-time feedback of intragastric pressure. Specifically, it includes: dynamically adjusting the infusion rate to optimize the patient's individualized gastric motility response by real-time monitoring of the patient's intragastric pressure signal and infusion rate command signal, combined with waveform recognition, feature extraction, system identification, and predictive models. This disclosure achieves a technological leap from passive response to active regulation through deep identification and predictive utilization of intragastric pressure signals, improving the safety of enteral nutrition infusion; it realizes dynamic individualization and adaptive adjustment of the infusion strategy, improving the accuracy and effectiveness of nutritional support; and by introducing multi-objective collaborative optimization decision-making, it achieves optimal overall infusion control.

Description

Technical Field

[0001] This disclosure belongs to the field of medical device and clinical monitoring technology, and relates to a method and system for regulating enteral nutrition infusion based on real-time feedback of gastric pressure. Background Technology

[0002] Enteral nutrition is a crucial means of providing life support to patients unable to eat normally. Nutritional solutions are typically delivered to the patient's gastrointestinal tract via a nasogastric tube or other route using an enteral nutrition infusion pump. Precise control of the infusion rate is paramount during this process, directly impacting treatment efficacy and patient safety. Therefore, utilizing physiological parameter monitoring technologies to guide infusion regulation, particularly by measuring the key physiological signal of intragastric pressure, has become a key area of ​​technological development in this field.

[0003] In existing technologies, the main methods for adjusting the enteral nutrition infusion rate include: one is for healthcare professionals to set a fixed infusion rate based on clinical experience and general guidelines, and then perform intermittent manual assessment and adjustment at fixed time intervals by methods such as auscultating bowel sounds and aspirating gastric contents. Another method is to use simple automated feedback control, such as installing a pressure sensor on the nasogastric tube. When the monitored intragastric pressure exceeds a preset fixed alarm threshold, the infusion pump automatically pauses or slows down the infusion.

[0004] However, the above-mentioned existing technical means have certain technical defects: (1) The adjustment method based on human experience is highly subjective and the feedback is seriously delayed, and it cannot respond to the dynamic changes of the patient's gastrointestinal function in real time.

[0005] (2) Although feedback control based on simple pressure threshold has achieved automation, it is still a passive and reactive on / off control. This method only treats the pressure signal as a single alarm indicator and only reacts after the pressure accumulates and reaches a dangerous level. It fails to deeply analyze the rich gastric dynamic information contained in the temporal changes of the pressure signal, and does not consider the huge physiological differences between different patients and the same patient at different recovery stages, resulting in insufficient regulation accuracy and still having a high risk of over-infusion or under-infusion. Public content

[0006] The technical problem to be solved by this disclosure is to overcome the problems raised in the prior art and provide a method and system for regulating enteral nutrition delivery based on real-time feedback of gastric pressure.

[0007] This disclosure solves the above-mentioned technical problems through the following technical solution:

[0008] The first aspect of this disclosure provides a method for regulating enteral nutrition infusion based on real-time feedback of gastric pressure, comprising: S1, acquiring in real-time the raw signal of gastric pressure of the target patient and the real-time infusion rate command signal of the enteral nutrition infusion pump, and performing time alignment.

[0009] S2. Perform waveform recognition and feature extraction on the original gastric pressure signal to obtain gastric motility waveform features, and analyze the input-output induction relationship between the real-time infusion rate command signal and the gastric motility waveform features to generate a dynamic feature set of gastric motility.

[0010] S3. Based on the gastric motility dynamic feature set, the gastric motility system parameter vector characterizing the gastric motility response characteristics of the target patient is estimated online through an identification algorithm.

[0011] S4. Using the gastric motility system parameter vector and the gastric motility dynamic feature set, the gastric motility state prediction result is calculated by the prediction model, and the gastric motility state prediction result is combined with the multi-objective optimization function to generate an optimized infusion rate instruction.

[0012] S5. Control the operation of the infusion pump according to the optimized infusion rate command, monitor the actual intragastric pressure response index sequence, and adaptively update the gastric motility system parameter vector according to the deviation between the actual intragastric pressure response and the predicted response index sequence generated by the prediction model.

[0013] The second aspect of this disclosure provides an enteral nutrition infusion regulation system based on real-time feedback of intragastric pressure, comprising: a signal acquisition module, a dynamic feature extraction module, a system parameter identification module, a prediction and decision module, and an adaptive update module.

[0014] The signal acquisition module is connected to the dynamic feature extraction module, the dynamic feature extraction module is connected to the system parameter identification module, the system parameter identification module is connected to the prediction and decision module, and the prediction and decision module is connected to the adaptive update module.

[0015] The signal acquisition module acquires the raw gastric pressure signal of the target patient and the real-time infusion rate command signal of the enteral nutrition infusion pump in real time, and performs time alignment.

[0016] The dynamic feature extraction module performs waveform recognition and feature extraction on the raw gastric pressure signal to obtain gastric motility waveform features, and analyzes the input-output induction relationship between the real-time infusion rate command signal and the gastric motility waveform features to generate a dynamic feature set of gastric motility.

[0017] The system parameter identification module, based on the gastric motility dynamic feature set, estimates online the gastric motility system parameter vector that characterizes the gastric motility response characteristics of the target patient through an identification algorithm.

[0018] The prediction and decision module uses the gastric motility system parameter vector and the gastric motility dynamic feature set to calculate the gastric motility state prediction result through the prediction model, and combines the gastric motility state prediction result with a multi-objective optimization function to generate an optimized infusion rate instruction.

[0019] The adaptive update module controls the operation of the infusion pump according to the optimized infusion rate command, monitors the actual intragastric pressure response index sequence, and adaptively updates the gastric motility system parameter vector based on the deviation between the actual intragastric pressure response and the predicted response index sequence generated by the prediction model.

[0020] Based on common knowledge in the field, the above-mentioned preferred conditions can be combined arbitrarily to obtain various preferred embodiments of this disclosure.

[0021] The positive and progressive effects of this disclosure are as follows:

[0022] 1. This disclosure identifies individualized gastric motility models of patients online and predicts future pressure trends, allowing for advance adjustment of the infusion rate to avoid potential gastric retention and reflux risks. This shifts the safety intervention point forward, changing the traditional approach of only taking remedial measures after pressure exceeds the limit.

[0023] 2. This disclosure constructs a gastric motility system parameter vector that evolves in real time with the patient's condition, thus abandoning the traditional one-size-fits-all fixed rate or static threshold mode, and ensuring that the infusion rate always matches the patient's current actual physiological tolerance and emptying rhythm.

[0024] 3. This disclosure uses a second-order dynamic model to parametrically model the individual gastric motility response, and combines rolling prediction and multi-objective optimization to maximize the nutrient delivery per unit time while ensuring the safety boundary of intragastric pressure and maintaining the stability of gastric autonomous rhythm, thereby achieving the overall optimal balance of safety, effectiveness and physiological friendliness, and overcoming the lack of adaptability of constant rate strategies. Attached Figure Description

[0025] Figure 1 This is a flowchart of the enteral nutrition infusion regulation method based on real-time gastric pressure feedback provided in Embodiment 1 of this disclosure.

[0026] Figure 2 This is a schematic diagram of the module connection of the enteral nutrition infusion regulation system based on real-time gastric pressure feedback provided in Embodiment 2 of this disclosure. Detailed Implementation

[0027] The present disclosure is further illustrated below by way of embodiments, but the present disclosure is not limited to the scope of the embodiments described herein.

[0028] The prefixes such as "first" and "second" used in this disclosure are merely for distinguishing different descriptive objects and do not limit the position, order, priority, quantity, or content of the described objects. The use of ordinal numbers and other prefixes used to distinguish descriptive objects in this disclosure does not constitute a limitation on the described objects. The description of the described objects is given in the claims or the context of the embodiments, and should not be construed as an unnecessary limitation. Furthermore, in the description of this embodiment, unless otherwise stated, "multiple" means two or more.

[0029] In this embodiment of the disclosure, the collection, storage, use, processing, transmission, provision, and disclosure of user personal information comply with relevant laws and regulations and do not violate public order and good morals.

[0030] Example 1

[0031] Please see Figure 1 As shown, this disclosure provides a method for regulating enteral nutrition infusion based on real-time feedback of gastric pressure. The specific steps are as follows: S1, real-time synchronous acquisition of the original gastric pressure signal of the target patient and the real-time infusion rate command signal of the enteral nutrition infusion pump, and time alignment.

[0032] S2. Perform waveform recognition and feature extraction on the original gastric pressure signal to obtain gastric motility waveform features, and analyze the input-output induction relationship between the real-time infusion rate command signal and the gastric motility waveform features to generate a dynamic feature set of gastric motility.

[0033] As an exemplary embodiment of this disclosure, the specific process of generating the gastric motility dynamic feature set includes: filtering and baseline correction of the original gastric pressure signal to obtain a preprocessed gastric pressure signal.

[0034] Specifically, the acquired raw intragastric pressure signal is preprocessed to obtain a preprocessed intragastric pressure signal suitable for analysis. The specific implementation process includes: using a digital filter to remove physiological noise and equipment noise from the raw intragastric pressure signal. Subsequently, an asymmetric least squares smoothing algorithm is applied to the filtered raw intragastric pressure signal to automatically identify and subtract baseline drift caused by changes in patient position or slow changes in intra-abdominal pressure, generating a zero-baseline preprocessed intragastric pressure signal.

[0035] In one specific example, the digital filter may be a fourth-order Butterworth low-pass filter with a cutoff frequency set between 0.5 Hz and 2 Hz to effectively filter out high-frequency interference such as heartbeat and respiration, while fully preserving the lower-frequency gastric contraction wave.

[0036] The positive pressure waves in the preprocessed gastric pressure signal whose morphology and duration conform to the characteristics of gastric peristaltic waves are identified and marked as candidate transport waves.

[0037] Specifically, the amplitude of the preprocessed intragastric pressure signal monitored in real time is marked as the start point of the waveform when the amplitude continuously rises and exceeds a first preset threshold, for example, 2 mmHg to 5 mmHg; and as the end point of the waveform when the amplitude falls back and falls below the first preset threshold. The duration of the monitored preprocessed intragastric pressure signal amplitude is then calculated. Preprocessed intragastric pressure signals with a duration falling within the preset range of 15 to 60 seconds are considered valid pressure fluctuations, marked as candidate transport waves, and stored in a short-term buffer sequence.

[0038] Candidate transport waves that appear continuously within a specified sliding time window and whose amplitudes show a non-decreasing trend are recorded as valid transport waves.

[0039] Specifically, candidate transport waves in the buffer sequence are screened to extract effective transport waves that truly characterize effective gastric emptying activity. The aim is to exclude isolated and disordered gastric contractions and capture physiologically significant and coordinated peristaltic wave sequences.

[0040] The specific process for acquiring the effective transmission wave includes: using a sliding time window method, with the time window length set to 90 to 180 seconds, searching for candidate transmission waves within this sliding time window. Only when at least two candidate transmission waves appear consecutively within the time window, and the peak amplitude of the latter wave is not less than the peak amplitude of the former wave, is this set of consecutive waves determined to be an effective transmission wave. This determination logic can be expressed by the formula: If candidate transmission waves exist... and Their peak values ​​are respectively and Then it must satisfy ,in A negligible amplitude tolerance, such as 0.5 mmHg, is used to enhance the robustness of the algorithm.

[0041] The amplitude, timing, and morphological features of the effective delivery wave are extracted, and the delay and correlation features between the changes in the real-time infusion rate command signal and the appearance of the effective delivery wave are analyzed. All extracted features are integrated to form a dynamic feature set of gastric motility.

[0042] In one specific example, the amplitude feature is specifically the maximum amplitude; the temporal feature is specifically the rise time from the trough to the peak and the time interval between two consecutive effective delivery peak points; the morphological feature is specifically the area under the curve obtained by numerically integrating the waveform curve on the time axis, which represents the work done in a single contraction.

[0043] Specifically, the specific process for obtaining the delay and correlation features is as follows: analyze the delay time between the point where the infusion rate changes significantly and the appearance of the subsequent effective delivery wave sequence, and record it as the delay feature between the change of the real-time infusion rate command signal and the appearance of the effective delivery wave; analyze the ratio of the infusion rate increment to the average amplitude of the subsequent effective delivery wave, and record it as the correlation feature between the change of the real-time infusion rate command signal and the appearance of the effective delivery wave.

[0044] This disclosure identifies individualized gastric motility models of patients online and predicts future pressure trends, allowing for advance adjustment of the infusion rate to avoid potential gastric retention and reflux risks. This shifts the safety intervention point forward, changing the traditional approach of only taking remedial measures after pressure exceeds the limit.

[0045] S3. Based on the gastric motility dynamic feature set, the gastric motility system parameter vector characterizing the gastric motility response characteristics of the target patient is estimated online through an identification algorithm.

[0046] As an exemplary embodiment of this disclosure, the identification algorithm may specifically be a recursive least squares method. Before obtaining the gastric motility system parameter vector characterizing the gastric motility response characteristics of the target patient through online estimation by the identification algorithm, the initial working condition determination is further included. The specific implementation process includes: using a trial infusion rate curve for initial infusion and monitoring the gastric pressure response index during this stage.

[0047] Specifically, the infusion pump is controlled to infuse according to a preset, extremely conservative, tentative infusion rate curve, which is typically a low-rate, short-duration pulse or step signal.

[0048] For example, infuse at a rate of 10 ml / h for 5 minutes, then pause for 5 minutes, and then infuse at a rate of 20 ml / h for 5 minutes.

[0049] Based on the gastric intragastric pressure response index, the initial compliance characteristics characterizing the gastric wall's ability to expand and the basic rhythmic characteristics characterizing the gastric autonomous rhythm are calculated.

[0050] Specifically, during the exploratory infusion, the intragastric pressure response indices are continuously monitored and recorded, and analyzed to calculate two key baseline characteristics: initial compliance characteristics and baseline rhythm characteristics. Used to quantify the stomach's capacity to expand or its pressure-volume relationship when receiving a fluid load, it is calculated as the volume of nutrient solution injected during the infusion phase. Compared with the baseline increase in intragastric pressure caused by this stage The ratio of initial compliance eigenvalues. A higher initial compliance eigenvalue indicates good gastric compliance and insensitivity to volume increases; conversely, a lower initial compliance eigenvalue indicates that the stomach is relatively rigid or in a state of high tension.

[0051] The basic rhythmic features This is used to capture the frequency of spontaneous gastric contractions under low-interference conditions. Spectral analysis is performed on the entire intragastric pressure response signal during the exploratory infusion phase, including both the infusion and pause periods. For example, using Fast Fourier Transform (FFT), the most significant frequency peak within the standard frequency range of 0.02 Hz to 0.08 Hz (1 to 5 times per minute) of slow gastric waves is identified. The frequency of this peak is then determined as the fundamental rhythmic characteristic. .

[0052] Based on the initial compliance characteristics and the basic rhythm characteristics, the initial parameter matrix of the recursive least squares method is customized.

[0053] Specifically, the recursive least squares method requires an initial parameter covariance matrix. By default, It is typically set as an identity matrix with very large diagonal elements, representing a lack of knowledge about the initial parameters, and adjustment based on the values ​​of the initial compliance characteristics and basic rhythm characteristics. The diagonal elements.

[0054] For example, if the measured initial compliance characteristics are very low, it indicates that the system may be very sensitive to changes in input, and the initial compliance can be appropriately reduced. The initial variance of the elements related to the input gain guides the algorithm to converge to a smaller gain value more quickly. In this way, the identification algorithm starts iterating from a starting point that is closer to reality, thus significantly reducing the cold start time of the model.

[0055] It should be noted that during the initial condition determination phase, if the gastric pressure response exceeds the safety warning threshold, the exploratory infusion will be terminated immediately and the safety fault-tolerant control process will be initiated.

[0056] As an exemplary embodiment of this disclosure, the specific process of obtaining the gastric motility system parameter vector includes: based on the gastric motility dynamic feature set and the real-time infusion rate command signal, the model coefficients of the pre-constructed second-order dynamic model are identified online by recursive least squares method.

[0057] Specifically, recursive least squares is an online parameter estimation algorithm that can iteratively update model parameters as each new data point arrives, without having to repeatedly calculate all historical data.

[0058] The specific process of obtaining the parameters of the pre-constructed second-order dynamic model through online identification includes: constructing a second-order discrete-time autoregressive moving average model to describe the gastric motility response, i.e., the pre-constructed second-order dynamic model, whose general form is: ,in Representative at , The gastric intragastric pressure response index at any given time can be selected from the set of gastric motility dynamic characteristics, such as the average amplitude of the effective transport wave or the area under the curve. Representative at , The infusion rate input at any given time, i.e., the value of the real-time infusion rate command signal; These are the model coefficients to be identified.

[0059] At every moment The new observational data will be used to evaluate the current intragastric pressure response index. and infusion rate input Substituting into the update formula of the recursive least squares method, the recursive least squares method mainly includes two core steps: (1) Calculate the Kalman gain (2) Use this gain to update the parameter estimation vector. The specific formula for its implementation is as follows: ,in for The parameter vector estimate at time step, whose elements include ; for The observation vector at time step is composed of past pressure response parameters and infusion rate inputs. That is, based on the prediction error at the current moment (the part in parentheses), according to the Kalman gain. The weights they represent are for the parameter estimation at the previous time step. Make corrections to obtain new parameter estimates. .

[0060] Based on the model coefficients, the system gain characterizing the sensitivity of the stomach to the infusion rate, the time constant characterizing the speed of the gastric motility response, and the natural frequency characterizing the gastric autonomous rhythm are calculated.

[0061] It should be noted that the system gain... This characterizes the gastric response sensitivity to changes in infusion rate, i.e., the change in intragastric pressure caused by a unit change in infusion rate under steady state. The calculation formula is as follows:

[0062] .

[0063] The time constant This characterizes the speed of the gastric motility system's response. In a second-order system, there are two time constants, which are the reciprocals of the roots of the model's transfer function characteristic equation. This is achieved by solving... Obtain the characteristic roots time constant and It can be derived from the formula It is derived that, where It is the sampling time interval.

[0064] The inherent frequency Characterized by the frequency of spontaneous rhythmic activity of the stomach in the absence of external input, also derived from eigenvalues The location determines it.

[0065] The system gain, the time constant, and the natural frequency are combined to form a gastric motility system parameter vector.

[0066] This disclosure constructs a gastric motility system parameter vector that evolves in real time with the patient's condition, abandoning the traditional one-size-fits-all fixed rate or static threshold mode, and ensuring that the infusion rate always matches the patient's current actual physiological tolerance and emptying rhythm.

[0067] S4. Using the gastric motility system parameter vector and the gastric motility dynamic feature set, the gastric motility state prediction result is calculated by the prediction model, and the gastric motility state prediction result is combined with the multi-objective optimization function to generate an optimized infusion rate instruction.

[0068] As an exemplary embodiment of this disclosure, the specific process of obtaining the gastric motility state prediction result includes: converting the second-order dynamic model into a state-space model based on the model coefficients.

[0069] Specifically, the state-space model can transform the second-order dynamic model obtained in the previous stage into a mathematical form more suitable for multi-step prediction. Based on the second-order dynamic model... Construct an equivalent state space, the specific formula of which is: ,in yes The system state vector at time t is typically defined in a second-order dynamic model as follows: . It is the infusion rate input; It is the system's output, i.e., the gastric pressure response index. .matrix It is composed of model coefficients The system matrix, input matrix, and output matrix are uniquely determined.

[0070] It should be noted that the specific meaning of the state-space model mentioned above is: the next state of the system. It is based on the current state and current input It is jointly determined, and the system's current output Then it is entirely determined by the current state. Decide.

[0071] The dynamic feature set of gastric motility in the previous regulation cycle is used as the initial state of the state space model. Rolling deduction is performed in the state space model to calculate the prediction curve of gastric pressure change trend in the future prediction time domain.

[0072] Specifically, at the beginning of each regulation cycle, the matrix defined by the current gastric motility system parameter vector is... And the latest state in the current set of gastric motility dynamic features, i.e. This serves as the initial condition for prediction. An infusion rate sequence is assumed for the future, for example, assuming the current infusion rate remains constant over the subsequent prediction time domain.

[0073] Furthermore, iterative calculations are performed using the representation formula of the state-space model to gradually deduce the future. The state and output at each time step, i.e. ,in Starting from 0, the process iterates until the end of the prediction time domain. Through this process, a prediction curve of the future gastric pressure change trend can be obtained. .

[0074] In one specific example, the future prediction time domain can be within the next 5 to 15 minutes.

[0075] The gastric pressure change trend prediction curve is analyzed to obtain the maximum predicted pressure value, the probability of pressure exceeding the limit, and the effective transport wave frequency and probability of occurrence in the future prediction time domain, which are used as the prediction results of gastric motility status.

[0076] Specifically, based on the predicted intragastric pressure change trend curve, the system checks whether the predicted intragastric pressure curve exceeds a preset safety threshold to assess the risk of future gastric retention. Simultaneously, it analyzes the local extreme points and periodicity of the predicted curve to predict the probability and frequency of future effective transport waves. The final output of the gastric motility prediction result is structured data containing key information such as the maximum predicted pressure value in the future prediction time domain, the probability of pressure exceeding the limit, and the predicted frequency and probability of effective transport waves.

[0077] As an exemplary embodiment of this disclosure, the specific generation process of the optimized infusion rate instruction includes: constructing a multi-objective optimization function, which integrates a nutritional objective function for maximizing the nutritional infusion dose, a safety objective function for minimizing the risk of excessive gastric pressure, and a physiological objective function for stabilizing the gastric spontaneous peristalsis rhythm.

[0078] Specifically, the multi-objective optimization function It can be expressed in the form of the following formula: ,in This represents the nutritional objective function, the purpose of which is to maximize the infusion rate; This represents the safety objective function, the purpose of which is to minimize the risk of excessive intragastric pressure. This represents the physiological target function, the purpose of which is to promote regular peristalsis of the stomach; These are preset weighting coefficients, reflecting the trade-offs in importance between different clinical goals, and can be adjusted according to the patient's overall condition. For example, for high-risk patients, The weight will be significantly higher than .

[0079] It should be noted that the nutritional objective function... It is usually defined as the integral of the difference between the infusion rate to be minimized and the maximum permissible rate, and its formula is as follows: That is, to pursue infusion rate Maximize the sum.

[0080] The security objective function This is achieved through a penalty function, which combines the predicted pressure sequence from the gastric motility prediction results obtained in the previous stage with a safety threshold. Compare. When predicting stress. Exceed At this point, the value of the safety objective function will increase sharply. Its formula can be expressed as follows: .

[0081] The physiological target function The aim is to encourage infusion rates to match or guide the spontaneous rhythm of the stomach, based on the inherent frequencies in the gastric motility system parameter vector. The degree of deviation of the penalized infusion rate pattern from the rhythm is expressed by the following formula: ,in Infusion rate sequence The spectrum analysis of the main frequency.

[0082] The predicted gastric motility state is used as a dynamic constraint condition for the multi-objective optimization function.

[0083] Specifically, after the multi-objective optimization function is constructed, the gastric motility state prediction results from the previous stage are introduced into the optimization problem as dynamic constraints, i.e., the predicted pressure throughout the entire prediction time domain. It must always remain below a rigid, insurmountable safety limit. ,Right now And the infusion rate It is also constrained by the physical properties of the injection pump, namely... .

[0084] Find the infusion rate curve that enables the multi-objective optimization function to reach the optimal solution under the premise of satisfying the dynamic constraints, and use it as the instruction for optimizing the infusion rate.

[0085] Specifically, the infusion rate curve can be solved using a numerical optimization algorithm within the Model Predictive Control (MPC) framework, such as Sequential Quadratic Programming (SQP). The specific implementation process includes: at the beginning of each control cycle, the controller solves for the infusion rate curve that minimizes the joint multi-objective optimization function over the next N steps, for example, the next 5 minutes. Simultaneously satisfying all constraints, the infusion rate curve obtained by the solution is used as the optimized infusion rate command.

[0086] It should be noted that in actual solution execution, usually only the first rate value of the sequence is executed, and then in the next control cycle, based on the new system state, the entire prediction and optimization process is repeated to form a closed-loop, rolling optimization control.

[0087] As an exemplary embodiment of this disclosure, the step of generating the optimized infusion rate instruction further includes a feedback correction step, the specific implementation of which includes: calculating the current gastric emptying motility index, which characterizes the actual contractile capacity of the stomach, based on the latest dynamic gastric motility feature set.

[0088] Specifically, the current gastric emptying motility index aims to assess the actual current gastric emptying capacity in real time and rapidly. The specific calculation formula is as follows: ,in It is the average amplitude of all valid transmitted waves within the most recent time window. It is its average area under the curve. It is the average rise time.

[0089] The calculation formula means that gastric emptying motility is directly proportional to the intensity and work done by contraction, and inversely proportional to the speed of contraction. By nonlinearly combining these measured characteristics, the current gastric emptying motility index can reflect the current motility state more comprehensively than a single parameter.

[0090] The current gastric emptying motility index and the predicted gastric motility state are fused and analyzed to determine whether the gastric motility system is in an unstable transitional state.

[0091] Specifically, the core logic of the fusion analysis to determine whether the gastric motility system is in an unstable transitional state is to compare the consistency between the current reality and future predictions. The specific implementation process includes: comparing the current gastric emptying motility index with its historical moving average and comparing the qualitative judgment of the future motility trend in the prediction results.

[0092] For example, if the current gastric emptying motility index decreases significantly by more than 20% to 30% compared to the previous cycle, while the model's prediction remains stable or increasing, then there is a significant contradiction between the two, suggesting that gastric motility may be in a rapid decline state that the model has not yet fully captured. Conversely, if the current gastric emptying motility index rises sharply, while the model predicts stability, it may indicate that gastric motility has entered an unstable hyperactive state.

[0093] If the gastric motility system is in an unstable transitional state, the infusion rate curve obtained by the multi-objective optimization function is conservatively corrected by a correction function, and the corrected curve is used as the final optimized infusion rate command.

[0094] Specifically, when gastric motility is determined to be in an unstable transitional state, a conservative correction mechanism is triggered. This mechanism applies to the infusion rate curve obtained in the previous step through a multi-objective optimization function. The correction operation does not completely overturn the original decision but rather scales it proportionally. The final optimized infusion rate command after correction is then determined. Specifically, it can be obtained through the following calculation formula: ,in It is the first in the initially calculated infusion rate sequence. One value, It is a correction factor.

[0095] It should be noted that when gastric motility is judged to be in a state of rapid decline, The threshold will be set in a range less than 1, for example, 0.5 to 0.8, thereby reducing the overall infusion rate.

[0096] When gastric motility is determined to be in an unstable hyperactive state, in order to avoid overstimulation, It could also be set to a value less than 1 or remain at 1 but with increased monitoring frequency.

[0097] When gastric motility is judged to be stable The value is 1, meaning no correction is performed.

[0098] S5. Control the operation of the infusion pump according to the optimized infusion rate command, monitor the actual intragastric pressure response index sequence, and adaptively update the gastric motility system parameter vector according to the deviation between the actual intragastric pressure response and the predicted response index sequence generated by the prediction model.

[0099] As an exemplary embodiment of this disclosure, the specific implementation process of adaptively updating the gastric motility system parameter vector includes: continuously collecting the actual intragastric pressure response index sequence during the period of executing the optimized infusion rate command.

[0100] Specifically, within each control cycle of the infusion pump executing the optimized infusion rate command, the system performs the following operations in parallel: continuously acquiring and recording the actually measured intragastric pressure response index sequence, denoted as... Using the prediction model defined by the currently used gastric motility system parameter vector, and inputting the same optimized infusion rate command sequence as the actual execution, a corresponding, purely theoretical predictive response index sequence is generated, denoted as... .

[0101] The actual intragastric pressure response index sequence is compared with the predicted response index sequence generated by the prediction model based on the same infusion rate command, and the model prediction bias is calculated.

[0102] Specifically, the normalized root mean square error (RMSE) is used to quantify the model's prediction bias. The specific calculation formula includes: Where N is the number of sampling points within the current evaluation time window, This represents the normal fluctuation range of the intragastric pressure signal, used for normalization to eliminate the influence of dimensions.

[0103] The The value intuitively reflects the accuracy of the model's predictions under the current infusion strategy.

[0104] If the model prediction deviation continues to exceed the adaptive update threshold, the system parameter identification step is triggered, and the gastric motility system parameter vector is re-estimated using the latest data stream.

[0105] Specifically, the calculated model prediction bias is compared with a preset adaptive update threshold. The adaptive update threshold is typically set within a tolerable error range, such as 0.1 to 0.2. Setting it too low will lead to excessively frequent model updates, increasing the computational burden; setting it too high may cause the system to use an inaccurate model for an extended period.

[0106] If the model prediction bias is detected to persist for a period of time, such as exceeding the adaptive update threshold for more than 3 to 5 consecutive evaluation cycles, it is determined that the current model parameters can no longer accurately describe the patient's gastric motility system. At this point, the model update mechanism is automatically triggered. This mechanism restarts the system parameter identification step.

[0107] It should be noted that the system parameter identification step differs from the initial identification in that this identification uses the latest batch of observation data accumulated since the last update, namely the gastric motility dynamic feature set, to re-estimate the gastric motility system parameter vector. The newly identified parameter vector will replace the old parameters and be used for subsequent prediction and decision-making, thereby completing an adaptive update of the model.

[0108] As an exemplary embodiment of this disclosure, S5 further includes a parallel safety fault-tolerant control step, the specific implementation of which includes: real-time monitoring of whether instantaneous high pressure or continuous plateau pressure appears in the original gastric pressure signal, in order to determine whether a safety event is triggered.

[0109] Specifically, the safety-tolerance control steps monitor the raw intragastric pressure signal in real time at an extremely high frequency, for example, 10 times per second. The monitoring logic includes two parallel judgment conditions: First, judging instantaneous high pressure, i.e., checking whether the instantaneous value of the signal exceeds an absolute, hard safety threshold, such as 25 mmHg to 35 mmHg. Second, judging plateau pressure, i.e., calculating the moving average of intragastric pressure over a time window of 1 to 2 minutes, and checking whether this average value continuously exceeds a lower plateau pressure threshold, such as 15 mmHg to 20 mmHg, and for a duration exceeding a preset tolerance period, such as 3 to 5 minutes.

[0110] If a transient high pressure or a sustained plateau pressure appears in the raw gastric pressure signal, a safety event is triggered, and a highest-priority pause infusion command is generated and executed, while S3 and S4 are frozen.

[0111] Specifically, if any of the above judgment conditions is met, the highest priority fault-tolerant response will be triggered immediately. The specific contents include: (1) generating a pause infusion command, which bypasses all regular decision and scheduling queues and is sent directly to the underlying controller of the infusion pump to stop infusion immediately. (2) setting a global safety interruption flag bit inside the system. The S3, S4 and feedback correction steps will check this flag bit before each execution. When the interruption flag is activated, these intelligent computing modules will be frozen immediately, suspending all data processing and model calculations to prevent them from making incorrect analyses and decisions based on abnormal high-pressure data.

[0112] Once the intragastric pressure has returned to a safe range and remained stable, S4 is restarted using a conservative strategy.

[0113] Specifically, the gastric pressure is continuously monitored while the system is frozen. When the gastric pressure drops to a preset safe recovery range, such as below 10 mmHg, and remains stable within this range for a preset period of time, such as 10–15 minutes, to confirm that the gastric pressure has stabilized rather than just temporarily dropped, the recovery procedure is initiated. During the recovery process, the safety interruption flag is first released, allowing the frozen steps to resume operation.

[0114] It should be noted that after resuming operation, S3 is not immediately re-executed using the latest data. Instead, the last valid gastric motility system parameter vector before the safety pause was triggered is forcibly loaded and used. Based on this, S4 is restarted in a more conservative mode, temporarily increasing the weight coefficient of the safety objective function by 1.5 to 2 times in the multi-objective optimization function. This forces the optimizer to generate a more cautious and conservative infusion rate instruction for the next few control cycles until the system has been running smoothly for a period of time, at which point the weight coefficient gradually returns to its normal value.

[0115] This disclosure uses a second-order dynamic model to parametrically model the individual gastric motility response, and combines rolling prediction and multi-objective optimization to maximize the nutrient delivery per unit time while ensuring the safety boundary of intragastric pressure and maintaining the stability of gastric autonomous rhythm, thereby achieving the overall optimal balance of safety, effectiveness and physiological friendliness, and overcoming the lack of adaptability of constant rate strategies.

[0116] Example 2

[0117] Please see Figure 2 As shown, this disclosure provides an enteral nutrition infusion regulation system based on real-time feedback of intragastric pressure, with the following specific module distribution: signal acquisition module, dynamic feature extraction module, system parameter identification module, prediction and decision-making module, and adaptive update module.

[0118] The signal acquisition module is connected to the dynamic feature extraction module, the dynamic feature extraction module is connected to the system parameter identification module, the system parameter identification module is connected to the prediction and decision module, and the prediction and decision module is connected to the adaptive update module.

[0119] The signal acquisition module acquires the raw gastric pressure signal of the target patient and the real-time infusion rate command signal of the enteral nutrition infusion pump in real time, and performs time alignment.

[0120] The dynamic feature extraction module performs waveform recognition and feature extraction on the raw gastric pressure signal to obtain gastric motility waveform features, and analyzes the input-output induction relationship between the real-time infusion rate command signal and the gastric motility waveform features to generate a dynamic feature set of gastric motility.

[0121] The system parameter identification module, based on the gastric motility dynamic feature set, estimates online the gastric motility system parameter vector that characterizes the gastric motility response characteristics of the target patient through an identification algorithm.

[0122] The prediction and decision module uses the gastric motility system parameter vector and the gastric motility dynamic feature set to calculate the gastric motility state prediction result through the prediction model, and combines the gastric motility state prediction result with a multi-objective optimization function to generate an optimized infusion rate instruction.

[0123] The adaptive update module controls the operation of the infusion pump according to the optimized infusion rate command, monitors the actual intragastric pressure response index sequence, and adaptively updates the gastric motility system parameter vector based on the deviation between the actual intragastric pressure response and the predicted response index sequence generated by the prediction model.

[0124] The above content is merely an example and illustration of the concept of this disclosure. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the disclosed concept or exceed the scope defined by this disclosure, they should all fall within the protection scope of this disclosure.

Claims

1. Enteral nutrition infusion regulation method based on real-time feedback of gastric pressure, characterized in that: include: S1. Real-time synchronous acquisition of the raw gastric pressure signal of the target patient and the real-time infusion rate command signal of the enteral nutrition infusion pump, and time alignment. S2. Perform waveform recognition and feature extraction on the raw gastric pressure signal to obtain gastric motility waveform features, and analyze the input-output induction relationship between the real-time infusion rate command signal and the gastric motility waveform features to generate a dynamic feature set of gastric motility. S3. Based on the gastric motility dynamic feature set, the gastric motility system parameter vector characterizing the gastric motility response characteristics of the target patient is estimated online through an identification algorithm; S4. Using the gastric motility system parameter vector and the gastric motility dynamic feature set, the gastric motility state prediction result is calculated by the prediction model, and the gastric motility state prediction result is combined with a multi-objective optimization function to generate an optimized infusion rate instruction. S5. Control the operation of the infusion pump according to the optimized infusion rate command, monitor the actual intragastric pressure response index sequence, and adaptively update the gastric motility system parameter vector according to the deviation between the actual intragastric pressure response and the predicted response index sequence generated by the prediction model.

2. The enteral nutrition infusion regulation method based on real-time gastric pressure feedback according to claim 1, characterized in that: The specific process for generating the gastric motility dynamic feature set includes: The raw intragastric pressure signal is filtered and baseline corrected to obtain a preprocessed intragastric pressure signal; Identify positive pressure waves in the preprocessed intragastric pressure signal whose morphology and duration conform to the characteristics of gastric peristaltic waves, and mark them as candidate transport waves; Candidate transport waves that appear continuously within a specified sliding time window and whose amplitudes show a non-decreasing trend are recorded as valid transport waves. The amplitude, timing, and morphological features of the effective delivery wave are extracted, and the delay and correlation features between the changes in the real-time infusion rate command signal and the appearance of the effective delivery wave are analyzed. All extracted features are integrated to form a dynamic feature set of gastric motility.

3. The enteral nutrition infusion regulation method based on real-time gastric pressure feedback according to claim 1, characterized in that: The identification algorithm can specifically be a recursive least squares method. Before obtaining the gastric motility system parameter vector characterizing the gastric motility response characteristics of the target patient through online estimation using the identification algorithm, the process further includes initialization condition determination, the specific implementation of which includes: An exploratory infusion rate curve was used for initial infusion, and intragastric pressure response indices were monitored during this phase. Based on the gastric intragastric pressure response index, the initial compliance characteristics characterizing the gastric wall expansion capacity and the basic rhythmic characteristics characterizing the gastric autonomic rhythm are calculated. Based on the initial compliance characteristics and the basic rhythm characteristics, the initial parameter matrix of the recursive least squares method is customized.

4. The enteral nutrition infusion regulation method based on real-time gastric pressure feedback according to claim 3, characterized in that: The specific process for obtaining the gastric motility system parameter vector includes: Based on the gastric motility dynamic feature set and the real-time infusion rate command signal, the model coefficients of the pre-constructed second-order dynamic model are identified online by the recursive least squares method. Based on the model coefficients, the system gain characterizing the sensitivity of the stomach to the infusion rate, the time constant characterizing the speed of the gastric motility response, and the natural frequency characterizing the gastric autonomic rhythm are calculated. The system gain, the time constant, and the natural frequency are combined to form a gastric motility system parameter vector.

5. The enteral nutrition infusion regulation method based on real-time gastric pressure feedback according to claim 4, characterized in that: The specific process for obtaining the gastric motility prediction results includes: Based on the model coefficients, the second-order dynamic model is converted into a state-space model; The gastric motility dynamic feature set of the previous regulation cycle is used as the initial state of the state space model. Rolling deduction is performed in the state space model to calculate the predicted curve of gastric pressure change trend in the future prediction time domain. The gastric pressure change trend prediction curve is analyzed to obtain the maximum predicted pressure value, the probability of pressure exceeding the limit, and the effective transport wave frequency and probability of occurrence in the future prediction time domain, which are used as the prediction results of gastric motility status.

6. The enteral nutrition infusion regulation method based on real-time gastric pressure feedback according to claim 5, characterized in that: The specific process for generating the optimized infusion rate command includes: A multi-objective optimization function is constructed, which integrates a nutritional objective function for maximizing the nutrient infusion dose, a safety objective function for minimizing the risk of excessive gastric pressure, and a physiological objective function for stabilizing the gastric spontaneous peristalsis rhythm. The predicted gastric motility state is used as the dynamic constraint condition of the multi-objective optimization function; Find the infusion rate curve that enables the multi-objective optimization function to reach the optimal solution under the premise of satisfying the dynamic constraints, and use it as the instruction for optimizing the infusion rate.

7. The enteral nutrition infusion regulation method based on real-time gastric pressure feedback according to claim 6, characterized in that: The process of generating the optimized infusion rate command includes a feedback correction step, the specific implementation of which includes: The current gastric emptying motility index, which characterizes the actual contractile capacity of the stomach, is calculated based on the latest dynamic feature set of gastric motility. The current gastric emptying motility index and the gastric motility state prediction results are fused and analyzed to determine whether the gastric motility system is in an unstable transitional state. If the gastric motility system is in an unstable transitional state, the infusion rate curve obtained by the multi-objective optimization function is conservatively corrected by a correction function, and the corrected curve is used as the final optimized infusion rate command.

8. The enteral nutrition infusion regulation method based on real-time gastric pressure feedback according to claim 1, characterized in that: The specific implementation process for adaptively updating the gastric motility system parameter vector includes: During the cycle of executing the optimized infusion rate command, the actual intragastric pressure response index sequence is continuously collected; The actual intragastric pressure response index sequence is compared with the predicted response index sequence generated by the prediction model based on the same infusion rate command, and the model prediction bias is calculated. If the model prediction deviation continues to exceed the adaptive update threshold, the system parameter identification step is triggered, and the gastric motility system parameter vector is re-estimated using the latest data stream.

9. The enteral nutrition infusion regulation method based on real-time gastric pressure feedback according to claim 8, characterized in that: The S5 also includes a parallel safety fault-tolerant control step, the specific implementation of which includes: Real-time monitoring of the raw gastric pressure signal for transient high pressure or sustained plateau pressure is used to determine whether a safety event has been triggered. If a transient high pressure or a sustained plateau pressure appears in the raw intragastric pressure signal, a safety event is triggered, and a highest priority pause infusion command is generated and executed, while S3 and S4 are frozen. Once the intragastric pressure has returned to a safe range and remained stable, S4 is restarted using a conservative strategy.

10. A system for performing the enteral nutrition infusion regulation method based on real-time intragastric pressure feedback as described in claims 1-9, characterized in that: include: The signal acquisition module synchronously acquires the raw gastric pressure signal of the target patient and the real-time infusion rate command signal of the enteral nutrition infusion pump in real time, and performs time alignment. The dynamic feature extraction module performs waveform recognition and feature extraction on the raw gastric pressure signal to obtain gastric motility waveform features, and analyzes the input-output induction relationship between the real-time infusion rate command signal and the gastric motility waveform features to generate a dynamic feature set of gastric motility. The system parameter identification module, based on the gastric motility dynamic feature set, estimates the gastric motility system parameter vector representing the gastric motility response characteristics of the target patient online through an identification algorithm; The prediction and decision module uses the gastric motility system parameter vector and the gastric motility dynamic feature set to calculate the gastric motility state prediction result through the prediction model, and combines the gastric motility state prediction result with a multi-objective optimization function to generate an optimized infusion rate instruction. The adaptive update module controls the operation of the infusion pump according to the optimized infusion rate command, monitors the actual intragastric pressure response index sequence, and adaptively updates the gastric motility system parameter vector based on the deviation between the actual intragastric pressure response and the predicted response index sequence generated by the prediction model.

Images