Intestinal obstruction catheter real-time pressure regulating system and intestinal obstruction catheter

Abstract

The present application relates to the technical field of medical devices, and discloses a real-time pressure regulating system for an intestinal obstruction catheter and the intestinal obstruction catheter. The system comprises a data acquisition module that generates a dynamic three-dimensional intestinal pressure field model, synchronously monitors intestinal PH value, oxidation-reduction potential and gas composition; an intestinal obstruction type identification module that analyzes pressure waveform characteristics and identifies the obstruction type in combination with knowledge graph reasoning; a comprehensive obstruction index output module that fuses multiple physiological data to obtain a comprehensive obstruction index; a cross-patient scheme adaptation module that adapts a cross-patient treatment scheme to obtain an initial airbag adjustment strategy; a scheme optimization and closed-loop fuzzy control module that constructs a personalized intestinal data coupling model, combines an MPC and a double closed-loop fuzzy PID controller to optimize a strategy output control instruction; and through model prediction and closed-loop control, the airbag pressure can be adjusted in real time and accurately according to the intestinal condition of the patient, dynamic and accurate control of the intestinal obstruction treatment process is realized, and the treatment effect is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical devices, in particular to a real-time pressure regulation system for an intestinal obstruction catheter and the intestinal obstruction catheter. BACKGROUND

[0002] In modern medicine, intestinal obstruction is a common and intractable digestive system disease that seriously affects the physical health and quality of life of patients. Intestinal obstruction not only causes typical symptoms such as abdominal pain, abdominal distension, vomiting, and cessation of defecation and flatus, but also can lead to intestinal perforation, septic shock and other life-threatening complications if not treated in time and effectively. As an important medical device for treating intestinal obstruction, the precise regulation of intestinal pressure by the intestinal obstruction catheter plays a key role in relieving obstruction symptoms and promoting the recovery of intestinal function. With the rapid development of technology, artificial intelligence, big data, sensors and other technologies are increasingly widely used in the medical field, bringing new opportunities for the innovation of intestinal obstruction treatment technology. Therefore, the development of an efficient and accurate real-time pressure regulation system for an intestinal obstruction catheter is of great significance, which can significantly improve the treatment effect of intestinal obstruction and bring better hope for recovery to patients.

[0003] In the development process of medical technology, traditional intestinal obstruction treatment and pressure regulation methods have gradually shown limitations. Due to one-sided understanding of the intestinal condition, there is a lack of multi-dimensional data sources in the identification of intestinal obstruction types. And due to the lack of high-precision identification mechanism, it mostly relies on the experience of physicians, so that the judgment of obstruction type and severity is prone to deviation, resulting in uncontrollable overall judgment error of the disease condition, and the determination of subsequent treatment plan mostly adopts the universal treatment plan adaptation mechanism within the physician's cognitive. Therefore, the regulation decision of the air bag or water bag pressure of the intestinal obstruction mostly depends on the experience and cognition of the physician, so that the error of the treatment effect cannot be controlled.

[0004] Therefore, the present application provides a real-time pressure regulation system for an intestinal obstruction catheter and the intestinal obstruction catheter. SUMMARY

[0005] The application provides an intestinal obstruction catheter real-time pressure regulation system and an intestinal obstruction catheter, and the dynamic three-dimensional model can intuitively and accurately show the pressure change in the intestinal tract. The synchronous monitoring of various parameters helps analyze the intestinal obstruction from different angles, so that the doctor can more accurately and comprehensively judge the condition. The deep learning and knowledge reasoning technology are combined to give full play to the advantages of both, which can extract effective features from complex pressure data and use prior knowledge in the knowledge graph to assist in judgment, greatly improving the accuracy and reliability of the obstruction type identification, and providing key basis for subsequent targeted treatment. The intestinal obstruction index comprehensively reflects the physiological indicators of the intestinal tract, and compared with single parameter evaluation, can more comprehensively and accurately measure the severity of the obstruction, providing a quantitative and more valuable index for the doctor to develop a treatment plan. The successful experience of previous cases is used for reference, and the specific situation of the current patient is combined to quickly generate a targeted treatment plan, improve the treatment efficiency and applicability of the plan, and avoid the omissions that may occur when the doctor develops a plan from scratch. The individual differences of patients are fully considered, and through model prediction and closed-loop control, the airbag pressure can be adjusted in real time and accurately according to the intestinal condition of the patient, so that the dynamic and accurate control of the intestinal obstruction treatment process is realized, and the treatment effect is improved.

[0006] The application provides an intestinal obstruction catheter real-time pressure regulation system, which comprises:

[0007] A data acquisition module is configured to generate a dynamic three-dimensional intestinal pressure field model based on real-time high-density acquisition of pressure distribution data in the intestinal tract of a patient, and synchronously monitor the pH value, oxidation-reduction potential and gas composition of the intestinal tract of the patient.

[0008] An intestinal obstruction type identification module is configured to analyze the pressure waveform features of the dynamic three-dimensional intestinal pressure field model in real time based on a convolutional long short-term memory network, and identify the intestinal obstruction type of the patient in combination with a knowledge graph reasoning mechanism.

[0009] A comprehensive obstruction index output module is configured to fuse the pH value, oxidation-reduction potential and gas composition of the intestinal tract of the patient based on an improved conflict evidence combination method, and obtain the comprehensive obstruction index of the patient.

[0010] A cross-patient scheme adaptation module is configured to adapt a treatment scheme across patients based on the intestinal obstruction type and the comprehensive obstruction index of the patient and an obstruction case knowledge base, and obtain an initial airbag adjustment strategy for the patient.

[0011] A scheme optimization and closed-loop fuzzy control module is configured to construct a personalized intestinal tract data coupling model for the patient, and optimize the initial airbag adjustment strategy in combination with a model prediction control method and a double-closed-loop fuzzy PID controller, and output corresponding control instructions.

[0012] Preferably, the intestinal obstruction type identification module comprises:

[0013] a pressure waveform feature analysis submodule, configured to analyze, in real time, pressure waveform features of the dynamic three-dimensional intestinal pressure field model based on a convolutional long short-term memory network, and construct a pressure waveform feature three-dimensional tensor of the dynamic three-dimensional intestinal pressure field model;

[0014] a message transmission strength analysis submodule, configured to analyze, based on the pressure waveform feature three-dimensional tensor, message transmission strength between two adjacent intestinal pressure monitoring points of the patient;

[0015] a message transmission formula construction submodule, configured to construct, based on the pressure waveform feature vector of each intestinal pressure monitoring point of the patient and the message transmission strength between two adjacent intestinal pressure monitoring points of the patient, a message transmission formula of an intestinal pressure distribution graph neural network of the patient;

[0016] a graph neural network model establishment submodule, configured to construct, based on the message transmission formula of the intestinal pressure distribution graph neural network of the patient, physiological correlations between different intestinal pressure monitoring points of the patient, and obtain a graph neural network model;

[0017] a multi-classifier prediction submodule, configured to classify, based on a plurality of different classifiers, physiological correlation features of the graph neural network model, and obtain intestinal obstruction type prediction probability distributions of each classifier;

[0018] an ensemble learning submodule, configured to output an intestinal obstruction type of the patient based on a weighted average method and the intestinal obstruction type prediction probability distributions of all the classifiers.

[0019] Preferably, the message transmission strength analysis submodule comprises:

[0020] a local fluctuation range calculation unit, configured to calculate, based on the feature values of each pressure waveform feature dimension of each intestinal pressure monitoring point of the patient obtained from the pressure waveform feature three-dimensional tensor, a local fluctuation feature range value of each pressure waveform feature dimension of each intestinal pressure monitoring point of the patient;

[0021] a fluctuation frequency calculation unit, configured to calculate a fluctuation frequency of each pressure waveform feature dimension of each intestinal pressure monitoring point of the patient;

[0022] a pressure change activity calculation unit, configured to calculate, based on the local fluctuation feature range value and the fluctuation frequency of each pressure waveform feature dimension of each intestinal pressure monitoring point of the patient, a pressure change activity of each pressure waveform feature dimension of each intestinal pressure monitoring point of the patient;

[0023] a message transmission tendency calculation unit, configured to calculate, based on the pressure change activities of two adjacent intestinal pressure monitoring points of the patient and average feature values of each pressure waveform feature dimension, a message transmission tendency of each pressure waveform feature dimension of the two adjacent intestinal pressure monitoring points of the patient;

[0024] a message transmission strength calculation unit configured to calculate a message transmission strength of the patient based on message transmission tendencies of each pair of adjacent intestinal pressure monitoring points of the patient in each pressure waveform feature dimension.

[0025] Preferably, the fluctuation frequency calculation unit comprises:

[0026] calculating an average feature value of each intestinal pressure monitoring point of the patient in each pressure waveform feature dimension;

[0027] calculating a fluctuation frequency of each intestinal pressure monitoring point of the patient in each pressure waveform feature dimension based on the average feature value of each intestinal pressure monitoring point in each pressure waveform feature dimension and the feature value of each intestinal pressure monitoring point at each time point in each pressure waveform feature dimension:

[0028]

[0029] wherein, is the fluctuation frequency of the intestinal pressure monitoring point with coordinates (x, y, z) in the cth pressure waveform feature dimension, t0 is a starting monitoring time point, t1 is a terminal monitoring time point, sgn() is a sign function, is the feature value of the intestinal pressure monitoring point with coordinates (x, y, z) at time t in the cth pressure waveform feature dimension, is the average feature value of the intestinal pressure monitoring point with coordinates (x, y, z) in the cth pressure waveform feature dimension, is the feature value of the intestinal pressure monitoring point with coordinates (x, y, z) at time t+1 in the cth pressure waveform feature dimension.

[0030] Preferably, the message transmission tendency calculation unit calculates the message transmission tendency of each pair of adjacent intestinal pressure monitoring points of the patient in each pressure waveform feature dimension based on pressure variation activities of the each pair of adjacent intestinal pressure monitoring points of the patient and average feature values in each pressure waveform feature dimension, and the method comprises:

[0031]

[0032] wherein, is the message transmission tendency of the two intestinal pressure monitoring points with coordinates (i, j, z) and (m, n, o) in the cth pressure waveform feature dimension, min() is a minimum value, max() is a maximum value, is the pressure variation activity of the intestinal pressure monitoring point with coordinates (i, j, z) in the cth pressure waveform feature dimension, is the pressure variation activity of the intestinal pressure monitoring point with coordinates (m, n, o) in the cth pressure waveform feature dimension, an average feature value of the intestinal pressure monitoring point with coordinates (i, j, z) in the cth pressure waveform feature dimension, an average feature value of the intestinal pressure monitoring point with coordinates (m, n, o) in the cth pressure waveform feature dimension, a global maximum value of the feature values of all monitoring points (x ′ , y ′ , z ′ ) and all pressure waveform feature dimensions.

[0033] Preferably, the message passing formula construction submodule comprises:

[0034] a weight matrix determination unit configured to determine a weight matrix of each layer of the graph neural network based on the message passing strength between each two adjacent intestinal pressure monitoring points of the patient;

[0035] a message passing formula construction unit configured to construct a message passing formula of the intestinal pressure distribution graph neural network of the patient based on the weight matrix of each layer of the graph neural network and the pressure waveform feature vector of each intestinal pressure monitoring point of the patient:

[0036]

[0037] wherein, is a feature representation of a node p of the graph neural network in the l+1th layer, σ() is an activation function, N(p) is a neighbor node set of the node p, d p is a degree of the node p, d k is a degree of a node k in the neighbor node set of the node p, W (l) is a weight matrix of the lth layer of the graph neural network, is a pressure waveform feature vector of the node k in the graph neural network in the lth layer, b (l) is a bias term of the lth layer of the graph neural network.

[0038] Preferably, the scheme optimization and closed-loop fuzzy control module comprises:

[0039] a peristalsis frequency acquisition module configured to acquire a peristalsis frequency of the intestinal tract of the patient based on the optical fiber sensor monitoring.

[0040] Preferably, the scheme optimization and closed-loop fuzzy control module comprises:

[0041] a personalized intestinal data coupling model establishment submodule configured to train an intestinal dynamics model represented by a nonlinear state space equation based on historical pressure data and historical peristalsis frequency of the patient, to obtain a pressure-peristalsis frequency coupling relationship of the intestinal tract of the patient as the personalized intestinal data coupling model of the patient.

[0042] The closed-loop fuzzy control submodule is configured to couple a model predictive control method and personalized intestinal data of a patient to predict a change rate of intestinal peristalsis frequency of the patient in a future period as a peristalsis disturbance intensity, and combine a double closed-loop fuzzy PID controller to optimize an initial airbag adjustment strategy and output a corresponding control instruction.

[0043] Preferably, an input variable of an outer loop of the double closed-loop fuzzy PID controller is a pressure deviation rate, which is used to adjust the airbag pressure.

[0044] An input variable of an inner loop of the double closed-loop fuzzy PID controller is the peristalsis disturbance intensity, which is used to compensate for intestinal peristalsis disturbance.

[0045] An output variable of the double closed-loop fuzzy PID controller is an airbag inflation and deflation rate.

[0046] An intestinal obstruction catheter of an intestinal obstruction catheter real-time pressure adjustment system, comprising:

[0047] A flexible catheter body provided with a treatment airbag capable of being inflated and deflated at a distal end;

[0048] A nano piezoresistive sensor array close to the treatment airbag, which is used to collect intestinal pressure distribution data in real time;

[0049] A micro electrochemical sensor group, which is used to synchronously monitor intestinal pH value, oxidation-reduction potential and gas composition;

[0050] A wireless transmission module, which is used to instantaneously transmit the intestinal pressure distribution data, the intestinal pH value, the oxidation-reduction potential and the gas composition to an intestinal obstruction type identification module and a comprehensive obstruction index output module of the intestinal obstruction catheter real-time pressure adjustment system, respectively;

[0051] A bidirectional inflation and deflation device, which is used to adjust the pressure of the treatment airbag based on the control instruction.

[0052] The beneficial effects generated by the present application relative to the prior art are: the data acquisition module generates a dynamic three-dimensional intestinal pressure field model by real-time high-density collection of pressure distribution data in the patient's intestinal tract, while monitoring the intestinal tract PH value, oxidation-reduction potential and gas composition, providing rich and real-time data basis for comprehensively understanding the patient's intestinal condition. The dynamic three-dimensional model can intuitively and accurately show the pressure changes in the intestinal tract, and the synchronous monitoring of multiple parameters helps to analyze the intestinal obstruction from different angles, making the doctor's judgment of the disease more accurate and comprehensive. The intestinal obstruction type recognition module uses a convolutional long short-term memory network to analyze the pressure waveform characteristics, and combines a knowledge graph reasoning mechanism to identify the type of intestinal obstruction. This method combines the advantages of deep learning and knowledge reasoning technology, fully utilizes the advantages of both, can extract effective features from complex pressure data, and can use prior knowledge in the knowledge graph to assist in judgment, greatly improving the accuracy and reliability of obstruction type identification, providing key basis for subsequent targeted treatment. The comprehensive obstruction index output module uses an improved conflict evidence combination method to fuse multiple monitoring parameters to obtain a comprehensive obstruction index. This index comprehensively reflects the physiological indicators of the intestinal tract, and compared with single parameter evaluation, can more comprehensively and accurately measure the severity of obstruction, providing a quantitative and more valuable index for doctors to develop treatment plans. The cross-patient treatment plan adaptation module adapts the treatment plan across patients according to the patient's obstruction type, comprehensive obstruction index and obstruction case knowledge base to obtain an initial airbag adjustment strategy. This strategy draws on the successful experience of previous cases, combined with the specific circumstances of the current patient, quickly generates a targeted treatment plan, improves treatment efficiency and the applicability of the plan, and avoids the omissions that may occur when doctors develop a plan from scratch. The scheme optimization and closed-loop fuzzy control module constructs a personalized intestinal data coupling model, and uses model predictive control method and double closed-loop fuzzy PID controller to optimize the initial airbag adjustment strategy and output control instructions. This method fully considers the individual differences of patients, and through model prediction and closed-loop control, it can adjust the airbag pressure in real time and accurately according to the patient's intestinal condition, realize dynamic and accurate control of the intestinal obstruction treatment process, improve the treatment effect, reduce the occurrence of complications, and provide safer and more effective treatment for patients.

[0053] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the application. The objects and other advantages of the present application can be realized and attained by particularly pointed out in the specification.

[0054] The technical solutions of the present application will be further described in detail below with the help of the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0055] The accompanying drawings are included to provide a further understanding of the application, and are incorporated in and constitute a part of the specification, illustrate embodiments of the application, and are used to explain the present application, but are not intended to limit the present application. In the drawings:

[0056] Figure 1 The information transmission execution schematic diagram between the built-in function modules of the intestinal obstruction catheter real-time pressure regulation system in an embodiment of the present application;

[0057] Figure 2 The input and output quantity flow conversion schematic diagram of the double closed loop fuzzy PID controller in an embodiment of the present application;

[0058] Figure 3 The information transmission execution schematic diagram between the built-in structure and function modules of the intestinal obstruction catheter in an embodiment of the present application. DETAILED DESCRIPTION

[0059] The preferred embodiments of the present application are described below in conjunction with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and are not intended to limit the present application.

[0060] Embodiment 1:

[0061] The present application provides an intestinal obstruction catheter real-time pressure regulation system, referring to Figure 1 , comprising:

[0062] The data acquisition module is used to collect the pressure distribution data in the intestinal tract of the patient in real time at high density, and the data changes over time to generate a dynamic three-dimensional intestinal tract pressure field model, which is a dynamic three-dimensional model reflecting the pressure in the intestinal tract at different positions and time changes. The model can intuitively and comprehensively show the distribution and change of the pressure in the intestinal tract in space and time, and synchronously monitor the PH value (for example, the normal PH value in the intestinal tract is about 7.0-7.5, if the PH value deviates obviously, it may indicate that the acid-base balance in the intestinal tract is out of balance, such as when some inflammation occurs, the PH value may be reduced), oxidation-reduction potential (such as the metabolic activity of some bacteria in the intestinal tract can change the oxidation-reduction potential, which can reflect the oxidation-reduction state in the intestinal tract) and gas composition (such as the content of gases such as methane and hydrogen, some intestinal diseases may cause abnormal gas composition, such as when intestinal obstruction occurs, the production and discharge of gas may be blocked, resulting in a change in the proportion of gas composition) of the patient's intestinal tract. These parameters reflect the chemical environment and physiological state in the intestinal tract.

[0063] The intestinal obstruction type identification module uses a convolutional long short-term memory network to analyze the pressure waveform features of a dynamic three-dimensional intestinal pressure field model in real time, and combines this with a knowledge graph reasoning mechanism to identify the type of intestinal obstruction in patients. The pressure waveform features are extracted from key characteristics such as the shape, fluctuations, and periodicity of the pressure waveform. For example, different types of intestinal obstruction may cause different characteristics in the pressure waveform; mechanical intestinal obstruction may cause periodic, dramatic fluctuations in the pressure waveform, similar to an electrocardiogram. Figure 1 The pressure waveform of a normal bowel obstruction exhibits a specific pattern of fluctuation; however, the pressure waveform of a mechanical bowel obstruction may be relatively flat, but the frequency and amplitude of fluctuations will differ from those of a normal obstruction. The knowledge graph reasoning mechanism records the typical characteristics, related symptoms, and correlations among various types of bowel obstruction. When the convolutional long short-term memory network extracts pressure waveform features, the knowledge graph reasoning mechanism matches and infers these features against existing knowledge. For example, if the convolutional long short-term memory network detects a specific high-frequency, violent fluctuation characteristic in the pressure waveform, the knowledge graph reasoning mechanism, based on its stored knowledge, determines that this characteristic may be related to mechanical bowel obstruction, thereby helping to identify the type of bowel obstruction in the patient.

[0064] The comprehensive obstruction index output module is used to fuse intestinal pH, redox potential, and gas composition based on an improved conflict evidence combination method to derive a comprehensive obstruction index for patients; it can more comprehensively and accurately reflect the severity of intestinal obstruction. For example, using a micro-electrochemical sensor array, the patient's intestinal pH is obtained as 6.0 (normal range assumed to be 7.0-7.5), redox potential as 200mV (normal range assumed to be 100-300mV), and methane content as 5% (normal range assumed to be 1%-3%) and hydrogen content as 3% (normal range assumed to be 0.5%-2%). Based on medical knowledge and past experience, initial weights are assigned to pH, redox potential, methane content, and hydrogen content. Assuming a weight of 0.3 for pH, 0.2 for redox potential, 0.25 for methane content, and 0.25 for hydrogen content, these weights reflect the relative importance of each parameter in assessing the severity of intestinal obstruction. Different conditions and studies may lead to variations in weight settings. Because different parameters have different dimensions and value ranges, the data is standardized to facilitate data fusion. For example, pH value, redox potential, methane content, and hydrogen content are all converted to a range of 0-1. Assuming that after standardization, the pH value becomes 0.8 (significantly deviating from the normal range; the closer the value is to 1, the greater the deviation), the redox potential becomes 0.5 (in the middle of the normal range), the methane content becomes 0.75 (above the normal range), and the hydrogen content becomes 0.8 (above the normal range).

[0065] In actual cases, there can be conflicts between these data. For example, sometimes the PH value shows that the intestinal condition is poor, but the oxidation-reduction potential is relatively normal, which is a contradiction. The improved conflict evidence combination method will analyze and process this conflict. For example, through learning and analysis of a large number of case data, it is found that when the PH value and the oxidation-reduction potential conflict, in some specific intestinal disease conditions, the PH value is more reliable, so the weight of the PH value is appropriately increased in this case. Assuming that after the conflict analysis, the PH value weight is adjusted to 0.4, the oxidation-reduction potential weight is adjusted to 0.15, the methane content weight is still 0.25, and the hydrogen content weight is still 0.25.

[0066] Based on the adjusted weights, the comprehensive obstruction index is calculated. Comprehensive obstruction index = 0.4 x 0.8 + 0.15 x 0.5 + 0.25 x 0.75 + 0.25 x 0.8 = 0.32 + 0.075 + 0.1875 + 0.2 = 0.7825. The higher this value, the more likely the patient's intestinal obstruction is more severe. For example, set the comprehensive obstruction index 0-0.4 as mild obstruction, 0.4-0.7 as moderate obstruction, and 0.7 or more as severe obstruction, then according to the calculation result, this patient may be in a severe obstruction state.

[0067] The cross-patient scheme adaptation module is used for cross-patient treatment scheme adaptation based on the intestinal obstruction type and the comprehensive obstruction index of the patient and the obstruction case knowledge base, to obtain the initial airbag adjustment strategy of the patient. The obstruction case knowledge base stores a large number of treatment cases of past intestinal obstruction patients, including treatment schemes under different obstruction types and different severity (measured by the comprehensive obstruction index) and the implementation effects of these schemes, etc.

[0068] When a new patient is diagnosed with an intestinal obstruction type and a comprehensive obstruction index is obtained, the cross-patient scheme adaptation module acts as a scheme search matcher and searches for similar cases in the obstruction case knowledge base. For example, it is found that a patient with similar intestinal obstruction type and comprehensive obstruction index to the current patient was found previously, and the airbag adjustment strategy used for that patient at the time achieved good treatment results. Then, the cross-patient scheme adaptation module will draw on this successful case to develop a similar initial airbag adjustment strategy for the current patient, such as determining the initial airbag pressure, the time interval for inflation and deflation, etc.

[0069] The scheme optimization and closed-loop fuzzy control module is used to construct a personalized intestinal data coupling model of the patient, and combine a model predictive control method and a double closed-loop fuzzy PID controller to optimize the initial airbag adjustment strategy and output corresponding control instructions. The personalized intestinal data coupling model is constructed based on the historical data of the patient's own intestinal pressure, peristalsis frequency, etc., and it can simulate the patient's intestinal response under different conditions. The model predictive control method uses this personalized model to predict the changes of the patient's intestinal tract in the future period of time under different airbag adjustment strategies, such as predicting the change of intestinal peristalsis frequency, the trend of pressure change, etc. The double closed-loop fuzzy PID controller has two control loops, the outer loop adjusts the airbag pressure according to the pressure deviation rate (the deviation degree of the actual pressure and the expected pressure), and the inner loop compensates the intestinal peristalsis disturbance according to the peristalsis disturbance intensity (the predicted intestinal peristalsis frequency change rate). For example, when the model predictive control method predicts that the intestinal peristalsis will accelerate and may cause the pressure to rise, the inner loop of the double closed-loop fuzzy PID controller will adjust the control strategy according to the peristalsis disturbance intensity, and perform appropriate deflation operation on the airbag in advance to avoid high pressure; at the same time, the outer loop will further fine-tune the airbag pressure according to the deviation of the actual pressure and the expected pressure, to ensure that the pressure is always maintained within a suitable range. In this way, the initial airbag adjustment strategy is continuously optimized, and accurate control instructions are output, realizing precise control of the intestinal obstruction treatment process.

[0070] Embodiment 2:

[0071] On the basis of embodiment 1, the intestinal obstruction type recognition module comprises:

[0072] The pressure waveform feature analysis submodule is used to analyze the pressure waveform features of the dynamic three-dimensional intestinal pressure field model in real time based on a convolutional long short-term memory network, that is, to extract key features from the fluctuations, periods, amplitudes, etc. of the pressure waveform, for example, it can identify whether the pressure waveform is regularly fluctuating like a sine wave, or chaotic fluctuation, or sudden pressure rise or fall, etc.

[0073] And a pressure waveform feature three-dimensional tensor of the dynamic three-dimensional intestinal pressure field model is constructed. For example, one dimension of the three-dimensional tensor represents the height layout of the intestinal pressure monitoring points, one dimension represents the width layout of the intestinal pressure monitoring points, and the last dimension represents the feature dimension of the intestinal pressure monitoring points. The three dimensions of the three-dimensional tensor may also correspond to the intestinal position, time and feature dimension respectively. For example, the first dimension represents the position of different pressure monitoring points in the intestine, the second dimension represents the monitoring time point, and the third dimension contains feature values such as pressure change rate and amplitude. In this way, the complex pressure waveform information is presented in a structured tensor form, which is convenient for subsequent analysis.

[0074] The message transmission strength analysis submodule is configured to analyze the message transmission strength between each pair of adjacent intestinal pressure monitoring points of the patient based on the pressure waveform feature three-dimensional tensor. The message transmission strength means that the pressure change of one intestinal pressure monitoring point affects another intestinal pressure monitoring point. The message transmission strength determines the strength of information transmission between adjacent monitoring points.

[0075] The message transmission formula construction submodule is configured to construct a message transmission formula of the intestinal pressure distribution graph neural network of the patient based on the pressure waveform feature vector of each intestinal pressure monitoring point of the patient and the message transmission strength between each pair of adjacent intestinal pressure monitoring points. The pressure waveform feature vector of each pressure monitoring point is composed of the feature values of the point in each pressure waveform feature dimension, which represents the unique pressure change characteristics of the monitoring point. The formula defines how an intestinal pressure monitoring point updates its feature representation according to the information of adjacent points in the intestinal pressure distribution graph neural network. For example, the formula can specify that the new feature representation of a certain monitoring point at a certain time is the operation of its own original feature vector and the feature vector of the adjacent monitoring point according to the message transmission strength, which clearly defines the transmission and update rules of information between different monitoring points and provides a basis for constructing the graph neural network model.

[0076] The graph neural network model establishment submodule is configured to construct the physiological correlation between different intestinal pressure monitoring points of the patient based on the message transmission formula of the intestinal pressure distribution graph neural network of the patient, and obtain a graph neural network model. The graph neural network constructs a model with intestinal pressure monitoring points as nodes and information transmission relationships defined by the message transmission formula as edges. Each node (pressure monitoring point) receives information from adjacent nodes and updates its state according to the message transmission formula. Through continuous iteration of this process, the model can learn the complex interaction relationships between different intestinal pressure monitoring points. For example, the model can find that some pressure monitoring points will quickly affect multiple intestinal pressure monitoring points around them when their pressure changes, while the influence between some monitoring points is relatively weak. Such a model can present the propagation mechanism and physiological correlation of pressure changes in the intestine as a whole, providing strong support for subsequent analysis of intestinal obstruction types.

[0077] The multi-classifier prediction submodule is configured to classify the physiological correlation features of the graph neural network model based on multiple different classifiers to obtain a prediction probability distribution of the intestinal obstruction type of each classifier. Different classifiers analyze the physiological correlation features of the intestinal pressure monitoring points learned by the graph neural network model based on their respective algorithms and principles. For example, a decision tree classifier can classify features based on their importance and threshold values, and a support vector machine classifier can distinguish different categories by finding the optimal classification hyperplane. Each classifier gives a prediction probability distribution for the intestinal obstruction type (such as mechanical intestinal obstruction, dynamic intestinal obstruction, etc.). For example, a certain classifier can predict that the probability of a patient suffering from mechanical intestinal obstruction is 0.7, the probability of suffering from dynamic intestinal obstruction is 0.2, and the probability of suffering from other types of intestinal obstruction is 0.1. Through the prediction of multiple classifiers, the possibility information about the intestinal obstruction type can be obtained from different angles.

[0078] The ensemble learning submodule is configured to output the intestinal obstruction type of the patient based on the weighted average method and the prediction probability distribution of the intestinal obstruction type of all classifiers. Since different classifiers have different accuracy and reliability in different situations, the weighted average method assigns a weight to each classifier, which reflects the credibility of the classifier. For example, through preliminary experimental verification, it is found that a certain classifier has a high accuracy in identifying mechanical intestinal obstruction, so it is given a high weight in the prediction of mechanical intestinal obstruction. Then, for each type of intestinal obstruction, multiply the probability of each classifier predicting this type by its corresponding weight, and then sum up to obtain the comprehensive prediction probability. Finally, according to the type with the highest comprehensive prediction probability, the intestinal obstruction type of the patient is determined. For example, for mechanical intestinal obstruction, classifier A has a prediction probability of 0.6 and a weight of 0.4, and classifier B has a prediction probability of 0.8 and a weight of 0.6. The comprehensive prediction probability is 0.6x0.4+0.8x0.6=0.72, and if this probability is the highest among the comprehensive prediction probabilities of all obstruction types, the patient is determined to have mechanical intestinal obstruction.

[0079] Embodiment 3:

[0080] On the basis of embodiment 2, the message passing intensity analysis submodule includes:

[0081] The local fluctuation range calculation unit is configured to calculate a local fluctuation feature range value of each intestinal pressure monitoring point of the patient in each pressure waveform feature dimension based on the feature value of each intestinal pressure monitoring point of the patient at each time point in each pressure waveform feature dimension obtained from the pressure waveform feature three-dimensional tensor. The pressure waveform feature three-dimensional tensor contains rich information, and each element corresponds to a feature value of a specific monitoring point at a time point and in a pressure waveform feature dimension. For example, assuming that the pressure change rate is the pressure waveform feature dimension of interest, for a pressure monitoring point on the intestinal tract, there will be a series of pressure change rate feature values that change over time within a time period. The local fluctuation range calculation unit finds the maximum value and the minimum value within a local time period from the values, and the difference between the two is the local fluctuation feature range value of the monitoring point in the pressure waveform feature dimension within the local time period. For example, within the local time period of 10 seconds to 20 seconds, the maximum value of the pressure change rate of the monitoring point is 0.5, and the minimum value is -0.3, and then the local fluctuation feature range value is 0.5-(-0.3) = 0.8. This value reflects the magnitude of the pressure change of the monitoring point in the specific pressure waveform feature dimension within the local time period.

[0082] The fluctuation frequency calculation unit is configured to calculate the fluctuation frequency of each intestinal pressure monitoring point of the patient in each pressure waveform feature dimension, which reflects the frequency of the change of the pressure in the feature dimension.

[0083] The pressure change activity calculation unit is configured to calculate the pressure change activity of each intestinal pressure monitoring point of the patient in each pressure waveform feature dimension based on the local fluctuation feature range value and the fluctuation frequency of each intestinal pressure monitoring point of the patient in each pressure waveform feature dimension. The combination of the two can more comprehensively describe the activity of the pressure change. For example, for a certain intestinal pressure monitoring point in the pressure change rate, if the local fluctuation feature range value is large, it indicates that the magnitude of the change of the pressure in the dimension is large; at the same time, the fluctuation frequency is also high, i.e., the pressure changes frequently, and then the pressure change activity obtained by multiplying the two or through other calculation methods (according to specific settings) will be high. Assuming that through a certain calculation method, the local fluctuation feature range value of the monitoring point in the pressure change rate dimension is 0.8, and the fluctuation frequency is 4 times per minute, and the pressure change activity is calculated to be 3.2. The higher the pressure change activity, the more active the pressure change of the monitoring point in the pressure waveform feature dimension.

[0084] a message passing tendency calculation unit configured to calculate, based on the pressure change activity of each pair of adjacent intestinal pressure monitoring points of the patient and the average feature value of each pressure waveform feature dimension, a message passing tendency of each pair of adjacent intestinal pressure monitoring points of the patient in each pressure waveform feature dimension, the value reflecting the degree of influence of the pressure change of one intestinal pressure monitoring point on another intestinal pressure monitoring point in the feature dimension.

[0085] a message passing strength calculation unit configured to calculate, based on the message passing tendency of each pair of adjacent intestinal pressure monitoring points of the patient in each pressure waveform feature dimension, a message passing strength of the patient. The message passing strength calculation unit performs weighted summation or other operations on the message passing tendency values of the different dimensions (weights are set according to experimental verification, for example, certain feature dimensions are more important for determining the type of intestinal obstruction, and the weights are higher). This provides key parameters for the message passing formula of the subsequent intestinal pressure distribution map neural network.

[0086] Embodiment 4:

[0087] On the basis of Embodiment 3, the fluctuation frequency calculation unit comprises:

[0088] an average feature value calculation subunit configured to calculate the average feature value of each intestinal pressure monitoring point of the patient in each pressure waveform feature dimension; in a period of time, the pressure change rate feature values of the monitoring point at different times t are obtained, the feature values are added and divided by the total number of monitoring times, to obtain the average feature value of the monitoring point in the pressure change rate feature dimension, which represents an average level of pressure change of the monitoring point in the feature dimension.

[0089] a fluctuation frequency calculation subunit configured to calculate, based on the average feature value of each intestinal pressure monitoring point in each pressure waveform feature dimension and the feature value of each intestinal pressure monitoring point at each time in each pressure waveform feature dimension, the fluctuation frequency of each intestinal pressure monitoring point in each pressure waveform feature dimension:

[0090]

[0091] wherein, is the fluctuation frequency of the intestinal pressure monitoring point with coordinates (x, y, z) in the cth pressure waveform feature dimension, t0 is the starting monitoring time, t1 is the ending monitoring time, sgn() is a sign function, is the feature value of the intestinal pressure monitoring point with coordinates (x, y, z) at time t in the cth pressure waveform feature dimension, is the average feature value of the intestinal pressure monitoring point with coordinates (x, y, z) in the cth pressure waveform feature dimension, is the feature value of the intestinal pressure monitoring point with coordinates (x, y, z) at time t + 1 in the c-th pressure waveform feature dimension.

[0092] wherein the sign of a number is determined. Here, The feature value at time t is The difference between the average feature value and the feature value at time t is given as 1 (positive difference), 0 (zero difference) or -1 (negative difference). Likewise, is the determination of the difference between the average feature value and the feature value at the next time t + 1.

[0093] The accumulation process is actually a count of the number of times the pressure feature value fluctuates around the average feature value. For example, if the pressure feature value frequently changes from one side of the average feature value to the other side over a certain period of time, the accumulation value will be larger, indicating that the fluctuations are more frequent.

[0094] The fluctuation frequency of each intestinal pressure monitoring point in each pressure waveform feature dimension is calculated through two main steps, which provides key data for subsequent analysis of the activity level of intestinal pressure changes.

[0095] Example 5:

[0096] Based on example 3, the message passing tendency calculation unit calculates the message passing tendency of each pair of adjacent intestinal pressure monitoring points of the patient in each pressure waveform feature dimension based on the pressure change activity of each pair of adjacent intestinal pressure monitoring points of the patient and the average feature value in each pressure waveform feature dimension. The method comprises:

[0097]

[0098] wherein, is the message passing tendency of the two intestinal pressure monitoring points with coordinates (i, j, z) and (m, n, o) in the c-th pressure waveform feature dimension, min() is the minimum value, max() is the maximum value, is the pressure change activity of the intestinal pressure monitoring point with coordinates (i, j, z) in the c-th pressure waveform feature dimension, is the pressure change activity of the intestinal pressure monitoring point with coordinates (m, n, o) in the c-th pressure waveform feature dimension, is the average feature value of the intestinal pressure monitoring point with coordinates (i, j, z) in the c-th pressure waveform feature dimension, is the average feature value of the intestinal pressure monitoring point with coordinates (m, n, o) in the c-th pressure waveform feature dimension, is the determination of the difference between the average feature value and the feature value at the next time t + 1. ′ ′ z​′ ) and the characteristic value of all pressure waveform feature dimensions takes the global maximum value, which plays a role of normalization in the formula, ensuring that the calculated message passing tendency value is within a reasonable range, facilitating the comparison and analysis of message passing tendency between different monitoring points.

[0099] This calculated value ultimately reflects the degree of tendency of the mutual influence of pressure changes between the two adjacent monitoring points in a specific pressure waveform feature dimension. The larger the value, the more obvious the influence of the pressure change of one monitoring point on the other monitoring point.

[0100] This step is of great significance for understanding the mutual influence trend of pressure changes at different positions in the intestinal tract and is a key link in constructing the message passing relationship of the intestinal pressure distribution map neural network.

[0101] Example 6:

[0102] Based on Example 2, the message passing formula construction submodule includes:

[0103] The weight matrix determination unit is used to determine the weight matrix of each layer of the graph neural network based on the message passing strength between each pair of adjacent intestinal pressure monitoring points of the patient. For each layer in the graph neural network, the message passing strength between each pair of adjacent monitoring points is mapped to the corresponding element in the weight matrix. Assuming that there are n nodes (corresponding to n intestinal pressure monitoring points) in the graph neural network, the weight matrix ω p,k is an n × n matrix (l represents the number of layers). If node p and node k are adjacent intestinal pressure monitoring points, and the message passing strength between them is ω p,k , then in the weight matrix ω p,k , the elements at the corresponding positions (p, k) and (k, p) (depending on the type of graph neural network, it may be symmetrically set or set according to a specific rule) will be set to ω p,kRelated values. Generally, the greater the message passing strength, the greater the weight matrix element value, so that in the information transmission process, the connection between adjacent nodes based on greater message passing strength will play a more important role in feature updating. In addition to simply mapping the message passing strength to the weight matrix element, the overall structure and hierarchical characteristics of the graph neural network also need to be considered. The weight matrix of different layers may need to be adjusted according to the learning goals and data processing needs of the network. For example, in the shallower network layers, the weight matrix may focus more on capturing local, direct stress information transmission relationships, so the mapping of message passing strength to weight matrix elements may be more direct; while in deeper network layers, the weight matrix needs to consider more complex global information, at which time the message passing strength may need to be weighted or transformed to better integrate stress information from different regions. At the same time, in order to enable the graph neural network to effectively learn, the weight matrix may also need to satisfy some mathematical properties, such as normalization operations, to ensure stable propagation and learning of information in the network.

[0104] The message passing formula construction unit is configured to construct a message passing formula of the intestinal pressure distribution graph neural network of the patient based on the weight matrix of each layer of the graph neural network and the pressure waveform feature vector of each intestinal pressure monitoring point of the patient.

[0105]

[0106] In the formula, is the feature representation of node p at the l+1 layer of the graph neural network, which integrates the information transmitted by the node itself and its neighbor nodes. As the network level deepens, this feature representation will be constantly updated, becoming richer and more representative, thus better reflecting the state of the intestinal pressure monitoring point in the entire network structure.

[0107] σ() is an activation function that introduces nonlinearity in the formula. It transforms the linearly combined input so that the graph neural network can learn complex nonlinear relationships. Common activation functions such as ReLU (Rectified Linear Unit) output the input value when the input is greater than 0 and output 0 when the input is less than 0. Through the activation function, the graph neural network can simulate the complex pressure changes and interaction relationships in the intestine, avoiding the model becoming a simple linear combination and improving the model's expressive power.

[0108] N(p) is the neighbor node set of node p, i.e., other intestinal pressure monitoring points directly connected to node p. In the intestinal pressure distribution graph neural network, the information of these neighbor nodes will be transmitted to node p, jointly affecting the update of the feature representation of node p. For example, in the actual intestine, adjacent pressure monitoring points will influence each other's pressure changes, and N(p) simulates this adjacent relationship.

[0109] d p is the degree of node p, d k is the degree of node k in the neighbor set of node p, i.e., the number of edges connected to the corresponding node. During the message passing process, considering the degree of a node helps to normalize the information passed by the neighbor nodes. For example, if a node has a large degree, it means that it receives more information. By dividing by the degree value, it can avoid some nodes with too many connections from having too much influence on the target node's feature representation, making the information passing more balanced and reasonable.

[0110] W (l) is the weight matrix of the l-th layer of the graph neural network, as described earlier, which is determined based on the message passing strength between each pair of adjacent intestinal pressure monitoring points. The weight matrix determines the importance of the information passed by the neighbor nodes to the current node p. Different weight matrix settings will affect the learning method and effect of the graph neural network on the intestinal pressure data.

[0111] is the pressure waveform feature vector of node k in the l-th layer of the graph neural network, which contains the feature information learned by node k about the intestinal pressure waveform at the l-th layer. As the network level advances, this feature information will be passed to the neighbor nodes through the message passing formula, helping the neighbor nodes update their own feature representation.

[0112] b (l) is the bias term of the l-th layer of the graph neural network, which provides an additional learnable parameter for the l-th layer of the graph neural network. It can be regarded as a base value that does not depend on the input data and can help the model better fit the data. In the intestinal pressure distribution graph neural network, the bias term helps to adjust the node feature representation, making it more accurately reflect the complex relationship between the intestinal pressure monitoring points. Usually at the beginning of model training, the bias term bl will be set to an initial value with random initialization. For example, the initial bias value can be randomly sampled from a normal distribution with a mean of 0 and a small standard deviation (such as 0.01). The reason for this is that in the early stage of training, the model has not fully learned the distribution and features of the data, and random initialization can make the model start exploring in different parameter spaces, avoiding all nodes starting from the same initial state, which helps the model converge to a better solution more quickly. For example, assuming that there are m nodes in a certain layer of the graph neural network, then the bias term b (l)is a vector of length m, where each element in the vector is randomly drawn from the normal distribution above. During the model training process, the bias term b l is updated together with other parameters such as the weight matrix, based on the gradient of the loss function using the backpropagation algorithm. The loss function measures the difference between the model's prediction and the true value. Taking the intestinal obstruction type identification as an example, if there is a deviation between the model's predicted intestinal obstruction type and the actual type, the contribution of each parameter (including the bias term b (l) ) to the loss function can be calculated through the backpropagation algorithm, and then the gradient of each parameter is obtained. According to the gradient information, the value of the bias term is adjusted using an optimizer (such as stochastic gradient descent, SGD, etc.) to gradually reduce the loss function and make the model's prediction more accurate. For example, when using the stochastic gradient descent optimizer, the update formula for the bias term b (l) may be where α is the learning rate, which controls the step size of each parameter update, is the gradient of the loss function L with respect to the bias term b (l) .

[0113] The value range of the bias term will change during the training process. A smaller bias value may result in a relatively gentle change in node feature representation, which may slow down the model's learning speed but can avoid the model fitting noise data too aggressively; a larger bias value may cause the node feature representation to change significantly, which may speed up the model's learning speed but may also cause the model to converge to a local optimal solution too early, resulting in overfitting. In practical applications, the appropriate bias term value needs to be determined through multiple experiments and parameter tuning to balance the model's learning speed and generalization ability, so that it performs best in tasks such as identifying intestinal obstruction types. For example, different combinations of initial bias values and learning rates are tried in different experiments, and the performance of the model on the training set and validation set is observed, and the parameter settings that can make the model achieve the best accuracy or other evaluation indicators on the validation set are selected.

[0114] This formula is crucial for understanding how information is transmitted and updated between different pressure monitoring points in the intestinal tract, and then establishing a graph neural network model to analyze the type of intestinal obstruction.

[0115] Embodiment 7:

[0116] On the basis of embodiment 1, further comprising:

[0117] A peristalsis frequency acquisition module is configured to acquire the peristalsis frequency of the patient's intestinal tract based on the monitoring obtained by the optical fiber sensor.

[0118] When monitoring intestinal peristalsis frequency, fiber optic sensors are placed in appropriate locations to sense the physical changes caused by intestinal peristalsis, such as pressure changes and minute displacements in surrounding tissues. These changes alter the propagation characteristics of light in the optical fiber (such as light intensity and phase). The fiber optic sensor detects these changes in optical signals and converts them into electrical or digital signals to obtain relevant data on intestinal peristalsis. After processing, the intestinal peristalsis frequency is determined.

[0119] Intestinal peristalsis frequency is an important indicator reflecting the functional state of the intestines. During intestinal obstruction, the peristalsis frequency may become abnormal, such as being too fast or too slow. After acquiring intestinal peristalsis frequency information, the system can combine it with other data (such as intestinal pressure distribution, obstruction type, etc.) for comprehensive analysis. For example, if the intestinal peristalsis frequency is significantly lower than the normal range, and intestinal pressure is elevated, combined with other parameters indicating mechanical intestinal obstruction, then the treatment plan may need to focus more on relieving the obstruction by adjusting the balloon pressure, while also considering how to promote the restoration of normal intestinal peristalsis. Furthermore, the rate of change of peristalsis frequency can also provide valuable information. If the peristalsis frequency drops rapidly in a short period, it may indicate a worsening of the condition or a need for timely adjustment of the treatment plan. Therefore, the peristalsis frequency and its rate of change acquisition module provides the system with dynamic, real-time intestinal peristalsis information, which helps to achieve more precise monitoring and adjustment of the intestinal obstruction treatment process. The intestinal peristalsis frequency data acquired by this module can provide important references for treatment plan optimization and closed-loop fuzzy control modules. The personalized intestinal data coupling model establishment submodule in the scheme optimization and closed-loop fuzzy control module can combine intestinal peristalsis frequency data with the patient's historical pressure data to train an intestinal dynamics model that better reflects the patient's actual situation, thereby more accurately describing the patient's intestinal pressure-peristalsis frequency coupling relationship.

[0120] Example 8:

[0121] Based on Example 1, the scheme optimization and closed-loop fuzzy control module includes:

[0122] A personalized intestinal data coupling model establishment submodule is used to train an intestinal dynamics model represented by a nonlinear state-space equation based on the patient's historical pressure data and historical peristaltic frequency. This model obtains the pressure-peristaltic frequency coupling relationship of the patient's intestines as a personalized intestinal data coupling model. Historical pressure data records the pressure values ​​of the patient's intestines at different points in the past, reflecting changes in intestinal pressure over time, such as pressure fluctuations before and after intestinal obstruction and during treatment. Historical peristaltic frequency data reflects the frequency of intestinal peristaltic activity within a corresponding time period. Changes in intestinal peristaltic frequency are closely related to intestinal health; different conditions may lead to changes in peristaltic frequency.

[0123] Nonlinear state-space equations are a mathematical model used to describe the dynamic behavior of a system in state space and can handle nonlinear relationships in the system. In the intestinal dynamics model, it can be used to characterize the complex interaction between intestinal pressure and peristalsis frequency. Since the physiological process of the intestine is not a simple linear relationship, nonlinear state-space equations can more accurately reflect the true dynamics of the intestine.

[0124] Using historical pressure data and historical peristalsis frequency as input, the parameters in the nonlinear state-space equations are continuously adjusted so that the output of the model matches the actual observed data as much as possible. This process can accurately simulate the relationship between intestinal pressure and peristalsis frequency based on the input historical data. For example, using some optimization algorithms (such as variants of gradient descent algorithm), the model parameters are gradually optimized by minimizing the error (such as mean square error) between the model prediction and the actual observation.

[0125] Each patient's intestinal physiological characteristics are different, and this personalized model can accurately simulate the patient's intestinal condition, providing strong support for subsequent control strategy optimization. For example, different patients may have different coupling relationships between pressure and peristalsis frequency due to different types and severity of intestinal diseases, and the personalized model can capture these differences.

[0126] The closed-loop fuzzy control submodule is used to predict the patient's intestinal peristalsis frequency change rate in the future period as the peristalsis disturbance intensity based on the model predictive control method and the patient's personalized intestinal data coupling model. This method is based on the model's ability to predict the future behavior of the system, and can predict the possible trend of intestinal peristalsis frequency in advance. For example, based on the current patient's intestinal pressure, peristalsis frequency, and the coupling relationship reflected by the personalized model, the change of intestinal peristalsis frequency in the next few minutes is predicted. The change rate of intestinal peristalsis frequency will affect the balloon adjustment strategy during the treatment of intestinal obstruction, and the larger the change rate, the stronger the peristaltic disturbance. For example, if it is predicted that the intestinal peristalsis frequency will increase rapidly in the future, then this rapid change rate indicates a strong peristaltic disturbance intensity.

[0127] And combine with the double closed-loop fuzzy PID controller to optimize the initial balloon adjustment strategy and output the corresponding control command. According to the peristaltic disturbance intensity and the pressure deviation rate, the double closed-loop fuzzy PID controller calculates the appropriate balloon inflation and deflation rate and other control parameters, and then outputs the corresponding control command to adjust the balloon pressure of the intestinal obstruction catheter, achieving precise control of the intestinal obstruction treatment process. For example, if the predicted peristaltic disturbance intensity increases and the pressure deviation rate is also large, the controller may output instructions to speed up the deflation rate of the balloon to relieve intestinal pressure and adapt to changes in intestinal peristalsis, improving treatment effectiveness.

[0128] Example 9:

[0129] Based on Example 1, referring to Figure 2 The input variable of the outer loop of the double-loop fuzzy PID controller is the pressure deviation rate, which is used to adjust the airbag pressure. The pressure deviation rate reflects the degree of difference between the current intestinal pressure and the expected pressure, which is usually obtained by calculating (actual pressure - expected pressure) / expected pressure. The outer loop adjusts the airbag pressure according to this deviation rate to make the intestinal pressure tend to the expected pressure value. For example, when the pressure deviation rate is positive, it means that the actual pressure is higher than the expected pressure, and the outer loop will output the corresponding control signal through the algorithm of the double-loop fuzzy PID controller to make the airbag deflate appropriately to reduce the intestinal pressure, making it tend to the expected pressure value; when the pressure deviation rate is negative, that is, the actual pressure is lower than the expected pressure, the outer loop will control the airbag to inflate to increase the intestinal pressure. The fuzzy PID control algorithm will dynamically adjust the control strength according to the size and trend of the pressure deviation rate. For example, when the pressure deviation rate is large, the controller will adjust the airbag inflation and deflation operation with a large amplitude to quickly reduce the pressure deviation; when the pressure deviation rate is small, the adjustment amplitude is correspondingly reduced to avoid over-adjustment.

[0130] The input variable of the inner loop of the double-loop fuzzy PID controller is the peristalsis disturbance intensity, which is used to compensate for the intestinal peristalsis disturbance; the change of intestinal peristalsis frequency will interfere with the airbag adjustment strategy, for example, sudden acceleration of peristalsis frequency may cause instability of intestinal pressure, affecting the treatment effect. The peristalsis disturbance intensity quantifies the degree of this disturbance, the greater the frequency change rate, the higher the peristalsis disturbance intensity.

[0131] When a strong peristalsis disturbance intensity is detected, the inner loop will adjust the control strategy to offset the interference of peristalsis on the airbag adjustment effect. For example, if the peristalsis disturbance intensity increases, the inner loop may adjust the airbag inflation and deflation speed appropriately to offset the influence of pressure fluctuation caused by peristalsis. Specifically, when the peristalsis frequency accelerates, it may cause the intestinal pressure to rise, and the inner loop will advance or accelerate the airbag deflation operation to prevent the pressure from being too high; conversely, when the peristalsis frequency slows down, it may cause the intestinal pressure to drop, and the inner loop will control the airbag inflation appropriately to maintain the pressure stability. In this way, the inner loop can compensate for the intestinal peristalsis disturbance in real time, improving the stability and accuracy of the airbag adjustment strategy.

[0132] The output variable of the double-loop fuzzy PID controller is the inflation and deflation rate of the balloon. By integrating the processing of the outer loop on the pressure deviation rate and the compensation of the inner loop on the peristalsis disturbance intensity, the controller calculates the appropriate inflation and deflation rate of the balloon. This rate determines the speed at which the balloon inflates or deflates. For example, if the pressure deviation rate is large and the peristalsis disturbance intensity is also high, the controller may output a larger balloon deflation rate to quickly reduce the pressure in the intestinal tract while adapting to the disturbance caused by peristalsis; if the pressure deviation rate is small and the peristalsis disturbance intensity is low, the inflation and deflation rate of the balloon will be reduced accordingly, allowing the pressure in the intestinal tract to smoothly approach the desired pressure value. By precisely controlling the inflation and deflation rate of the balloon, precise regulation of the intestinal pressure during the treatment of intestinal obstruction is achieved, improving the treatment effect.

[0133] The double-loop fuzzy PID controller processes the uncertainty of the input variables through fuzzy logic, dynamically adjusts the parameters (proportion, integral, and derivative coefficients) of the PID controller according to different input conditions, and thus achieves more precise control.

[0134] Through two interrelated control loops, different input variables are processed to achieve precise control of the inflation and deflation rate of the balloon, thereby optimizing the treatment process of intestinal obstruction. This double-loop structure can more effectively cope with complex changes in intestinal physiology, improving the control accuracy and stability of the system.

[0135] Embodiment 10:

[0136] Based on Embodiments 1-9, an intestinal obstruction catheter for an intestinal obstruction catheter real-time pressure regulation system, according to Figure 3 , comprising:

[0137] A flexible catheter body with an inflatable and deflatable treatment balloon at the distal end; this design allows the catheter to be comfortably placed in the intestinal tract, and the treatment of intestinal obstruction-related symptoms is achieved by inflating and deflating the treatment balloon. For example, at the site of intestinal obstruction, a properly inflated balloon can exert pressure on the intestinal tract to help alleviate the obstruction; deflation can adjust the pressure according to the treatment progress. The treatment balloon is a key component that directly acts on the intestinal tract to assist in the treatment of intestinal obstruction. By adjusting the amount of gas in the balloon, the pressure on the intestinal wall is changed, thereby improving the obstruction of the intestinal tract and promoting the flow of intestinal contents.

[0138] A nano-mechanical pressure sensor array close to the treatment balloon for real-time acquisition of intestinal pressure distribution data; it is arranged close to the treatment balloon. Such a layout can more accurately collect pressure distribution data in the intestinal tract near the treatment balloon, as the pressure changes in this area are closely related to the action of the balloon on the intestinal tract.

[0139] Real-time acquisition of intestinal pressure distribution data. These data record the pressure values and changes at different positions in detail. For example, when the pressure in a certain part of the intestine rises due to obstruction, the nano-piezoresistive sensor array can quickly perceive and record it, providing accurate basis for subsequent analysis of intestinal obstruction and adjustment of balloon pressure.

[0140] Micro-electrochemical sensor group for simultaneous monitoring of intestinal pH, redox potential and gas composition;

[0141] Wireless transmission module for real-time transmission of intestinal pressure distribution data, intestinal pH, redox potential and gas composition to the intestinal obstruction type identification module and comprehensive obstruction index output module of the intestinal obstruction catheter real-time pressure regulation system, respectively; for example, after receiving the pressure distribution data, the intestinal obstruction type identification module can analyze the pressure waveform characteristics to identify the obstruction type; the comprehensive obstruction index output module calculates the comprehensive obstruction index in combination with these data.

[0142] Bidirectional inflation and deflation device for pressure regulation of the treatment balloon based on control instructions. Based on the control instructions issued by the intestinal obstruction catheter real-time pressure regulation system, the pressure of the treatment balloon is regulated. When the system analyzes and judges that the pressure of the balloon needs to be increased to better relieve the obstruction, it will send an inflation instruction to the bidirectional inflation and deflation device; on the contrary, when the pressure is too high or the treatment situation changes and the pressure of the balloon needs to be reduced, the bidirectional inflation and deflation device will execute a deflation instruction. Through this precise inflation and deflation control, dynamic regulation of the pressure of the balloon during the treatment of intestinal obstruction is achieved, so as to achieve the best treatment effect.

[0143] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technology, the present application also intends to include these modifications and variations.

Claims

1. An ileus catheter real-time pressure regulation system, comprising: The method comprises the following steps: a data acquisition module for generating a dynamic three-dimensional intestinal pressure field model based on real-time high-density collection of pressure distribution data in the intestinal tract of a patient, and simultaneously monitoring the pH value, oxidation-reduction potential and gas composition in the intestinal tract of the patient; an intestinal obstruction type identification module for real-time analysis of the pressure waveform characteristics of the dynamic three-dimensional intestinal pressure field model based on a convolutional long short-term memory network, and identifying the intestinal obstruction type of the patient by combining a knowledge graph reasoning mechanism; a comprehensive obstruction index output module for obtaining the comprehensive obstruction index of the patient by fusing the pH value, oxidation-reduction potential and gas composition in the intestinal tract based on an improved conflict evidence combination method; a cross-patient scheme adaptation module for adapting the treatment scheme of the patient based on the intestinal obstruction type and the comprehensive obstruction index of the patient and an obstruction case knowledge base, and obtaining the initial air sac adjustment strategy of the patient; a scheme optimization and closed-loop fuzzy control module for constructing a personalized intestinal data coupling model of the patient, optimizing the initial air sac adjustment strategy by combining a model predictive control method and a double-closed-loop fuzzy PID controller, and outputting the corresponding control instructions; wherein the intestinal obstruction type identification module comprises: a pressure waveform feature analysis submodule for real-time analysis of the pressure waveform characteristics of the dynamic three-dimensional intestinal pressure field model based on a convolutional long short-term memory network, and constructing a pressure waveform feature three-dimensional tensor of the dynamic three-dimensional intestinal pressure field model; a message passing strength analysis submodule for analyzing the message passing strength between each pair of adjacent intestinal pressure monitoring points of the patient based on the pressure waveform feature three-dimensional tensor; a message passing formula construction submodule for constructing a message passing formula of the intestinal pressure distribution graph neural network of the patient based on the pressure waveform feature vector of each intestinal pressure monitoring point of the patient and the message passing strength between each pair of adjacent intestinal pressure monitoring points of the patient; a graph neural network model establishment submodule for constructing the physiological correlation between different intestinal pressure monitoring points of the patient based on the message passing formula of the intestinal pressure distribution graph neural network of the patient, and obtaining a graph neural network model; a multi-classifier prediction submodule for classifying the physiological correlation features of the graph neural network model based on multiple different classifiers, and obtaining the intestinal obstruction type prediction probability distribution of each classifier; an ensemble learning submodule for outputting the intestinal obstruction type of the patient based on the intestinal obstruction type prediction probability distribution of all classifiers and the weighted average method.

2. The ileus catheter real-time pressure regulation system of claim 1, wherein, The message passing strength analysis submodule comprises: a local fluctuation range calculation unit for calculating the local fluctuation feature range value of each pressure waveform feature dimension of each intestinal pressure monitoring point of the patient based on the feature values of each pressure waveform feature dimension of each intestinal pressure monitoring point of the patient at each time point obtained from the pressure waveform feature three-dimensional tensor; a fluctuation frequency calculation unit for calculating the fluctuation frequency of each pressure waveform feature dimension of each intestinal pressure monitoring point of the patient; a pressure change activity calculation unit for calculating the pressure change activity of each pressure waveform feature dimension of each intestinal pressure monitoring point of the patient based on the local fluctuation feature range value and the fluctuation frequency of each pressure waveform feature dimension of each intestinal pressure monitoring point of the patient. The message transmission tendency calculation unit is configured to calculate, based on the pressure change activity of the two adjacent intestinal pressure monitoring points of the patient and the average feature value in each pressure waveform feature dimension, the message transmission tendency of the two adjacent intestinal pressure monitoring points of the patient in each pressure waveform feature dimension. The message transmission force calculation unit is configured to calculate, based on the message transmission tendency of the two adjacent intestinal pressure monitoring points of the patient in each pressure waveform feature dimension, the message transmission force of the patient.

3. The ileus catheter real-time pressure regulation system of claim 2, wherein, The fluctuation frequency calculation unit comprises: calculating the average feature value of each intestinal pressure monitoring point of the patient in each pressure waveform feature dimension; calculating the fluctuation frequency of each intestinal pressure monitoring point of the patient in each pressure waveform feature dimension based on the average feature value of each intestinal pressure monitoring point of the patient in each pressure waveform feature dimension and the feature value of each intestinal pressure monitoring point of the patient in each pressure waveform feature dimension at each time point: In the formula, is the fluctuation frequency of the intestinal pressure monitoring point with coordinates (x, y, z) in the cth pressure waveform feature dimension, t0is the starting monitoring time, t1is the termination monitoring time, sgn() is the sign function, is the feature value of the intestinal pressure monitoring point with coordinates (x, y, z) at time t in the cth pressure waveform feature dimension, is the average feature value of the intestinal pressure monitoring point with coordinates (x, y, z) in the cth pressure waveform feature dimension, is the feature value of the intestinal pressure monitoring point with coordinates (x, y, z) at time t+1in the cth pressure waveform feature dimension.

4. The ileus catheter real-time pressure regulation system of claim 2, wherein, The method for calculating, by the message transmission tendency calculation unit, the message transmission tendency of the two adjacent intestinal pressure monitoring points of the patient in each pressure waveform feature dimension based on the pressure change activity of the two adjacent intestinal pressure monitoring points of the patient and the average feature value in each pressure waveform feature dimension comprises: where, is the message passing tendency of two intestinal pressure monitoring points with coordinates (i,j,z) and (m,n,o) in the c-th pressure waveform feature dimension, min() is the minimum value, max() is the maximum value, is the pressure change activity of the intestinal pressure monitoring point with coordinates (i,j,z) in the c-th pressure waveform feature dimension, is the pressure change activity of the intestinal pressure monitoring point with coordinates (m,n,o) in the c-th pressure waveform feature dimension, is the average feature value of the intestinal pressure monitoring point with coordinates (i,j,z) in the c-th pressure waveform feature dimension, is the average feature value of the intestinal pressure monitoring point with coordinates (m,n,o) in the c-th pressure waveform feature dimension, is the global maximum value of the feature value for all monitoring points (x ' ,y ' ,z ' ) and all pressure waveform feature dimensions.

5. The ileus catheter real-time pressure regulation system of claim 1, wherein, The message transmission formula construction sub-module comprises: The weight matrix determination unit is configured to determine the weight matrix of each layer of the graph neural network based on the message transmission force between the two adjacent intestinal pressure monitoring points of the patient. The message transmission formula construction unit is configured to construct the message transmission formula of the intestinal pressure distribution graph neural network of the patient based on the weight matrix of each layer of the graph neural network and the pressure waveform feature vector of each intestinal pressure monitoring point of the patient. wherein, is the feature representation of node p at the (l+1)th layer of the graph neural network, σ() is an activation function, N(p) is the set of neighboring nodes of node p, d p is the degree of node p, d k is the degree of node k in the set of neighboring nodes of node p, W (l) is the weight matrix of the lth layer of the graph neural network, is the stress waveform feature vector of node k at the lth layer in the graph neural network, b (l) is the bias term of the lth layer of the graph neural network.

6. The ileus catheter real-time pressure regulation system of claim 1, wherein, Further comprising: The peristalsis frequency acquisition module is configured to acquire the peristalsis frequency of the intestinal tract of the patient based on the monitoring of the optical fiber sensor.

7. The ileus catheter real-time pressure regulation system of claim 1, wherein, The scheme optimization and closed-loop fuzzy control module comprises: The personalized intestinal data coupling model establishment sub-module is configured to train the intestinal dynamics model represented by the nonlinear state space equation based on the historical pressure data and the historical peristalsis frequency of the patient, to obtain the pressure-peristalsis frequency coupling relationship of the intestinal tract of the patient as the personalized intestinal data coupling model of the patient. The closed-loop fuzzy control sub-module is configured to predict the peristalsis frequency change rate of the intestinal tract of the patient in the future period as the peristaltic disturbance intensity based on the model predictive control method and the personalized intestinal data coupling model of the patient, to combine the double closed-loop fuzzy PID controller to optimize the initial air bag adjustment strategy, and to output the corresponding control instruction.

8. The real-time pressure adjustment system of the intestinal obstruction catheter according to claim 1, wherein: The input variable of the outer ring of the double closed-loop fuzzy PID controller is the pressure deviation rate, which is used to adjust the air bag pressure. The input variable of the inner ring of the double closed-loop fuzzy PID controller is the peristaltic disturbance intensity, which is used to compensate for the peristaltic disturbance of the intestinal tract. The output variable of the double closed-loop fuzzy PID controller is the air bag inflation and deflation rate.

9. An ileus catheter having a real-time pressure regulation system according to any one of claims 1 to 8, characterized in that, The flexible catheter body provided with the inflatable and deflatable treatment air bag at the distal end; The nano piezoresistive sensor array close to the treatment air bag is configured to collect the intestinal pressure distribution data in real time. The micro electrochemical sensor group is configured to synchronously monitor the pH value, oxidation-reduction potential and gas composition of the intestinal tract. ​ The wireless transmission module is used for transmitting the intestinal pressure distribution data, the intestinal PH value, the oxidation-reduction potential and the gas composition to the intestinal obstruction type identification module and the comprehensive obstruction index output module of the real-time pressure regulation system of the intestinal obstruction catheter in real time, respectively. The bidirectional inflation and deflation device is used for regulating the pressure of the treatment air bag based on the control instruction.

Images