Gastrointestinal tract peristalsis space-time simulation method based on digital twinning

Abstract

The invention relates to a gastrointestinal tract peristalsis space-time simulation method based on digital twinning, which comprises the following steps: abstracting a neuromuscular activation process of a gastrointestinal tract into a physiological excitation field, constructing a variable topology strain network of a gastrointestinal wall, and forming a functional contraction path; introducing a control equation set to enable the excitation field to reflect fluctuation evolution in time, reflect tension distribution in space and cause geometric deformation in structure; analyzing the behavior difference between the twinborn body and the real gastrointestinal system by using the error field; error feedback is conducted to excitation field parameters, and closed-loop adjustment is achieved; historical state data and current feedback are fused, and a time sequence diagram network is adopted for modeling a creeping trend; the process from real fitting to future behavioral ability evolution prediction is realized; mapping the continuous creeping process into a space-time activity density map; extracting activity mutation, wriggling stagnation and reverse propagation characteristics of the key area by using topological convolution; the semantic indexes easy to understand are output, a predictable peristalsis behavior model is constructed, and clinical visibility and auxiliary diagnosis value are improved.

Description

Technical Field

[0001] This invention relates to a method for simulating gastrointestinal peristalsis, specifically a spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins. Background Technology

[0002] Currently, in the field of gastrointestinal motility simulation methods, such as the Chinese patent CN110585597A intelligent intestinal motility regulation system, there are still many shortcomings and technical bottlenecks. In particular, it exposes obvious limitations in simulation accuracy, regulation dimensions, feedback mechanisms, and model intelligence, making it difficult to meet the current needs of personalized medicine, dynamic prediction, and high-resolution spatiotemporal simulation. Existing technologies mainly focus on building systems around the goal of intelligent regulation. The core of this system is a closed-loop neural regulation system composed of an implantable intestinal motility detection module and an external control unit. By real-time acquisition of heart rate signals and physiological signals of intestinal motility, it achieves regulation and feedback control of intestinal nerve stimulation. Although this solution shows good practicality and engineering implementation advantages in the treatment of functional digestive disorders, from the perspective of simulation, model mechanism construction, and system intelligent evolution capabilities, it still cannot achieve high-fidelity, high-resolution, and personalized dynamic modeling of gastrointestinal motility behavior, and it does not have predictable and interpretable digital twin capabilities. First, its essence is a neurostimulation regulation system driven by physiological parameters, lacking a complete framework for gastrointestinal structural modeling and dynamic analysis. The system relies more on hardware interconnection of detection and regulation links rather than behavioral modeling at the mechanism level. Its implanted detection module cannot provide structural analysis of multi-dimensional physiological behaviors such as complete tension field, motion morphology, and muscle wall strain during gastrointestinal peristalsis. This is fundamentally different from the structure-function-behavior integrated modeling emphasized by digital twin technology. Second, although the system can perform rhythmic feedback regulation to a certain extent, its regulation logic mainly relies on threshold judgment, single-point stimulation control, and correlation analysis of some physiological parameters. It lacks a full-process simulation and data closed-loop update mechanism with time-space-structure as the core dimension. In particular, it does not provide algorithmic mechanisms or mathematical modeling support for predicting spatiotemporal evolution trends and generating peristaltic paths, and cannot support process understanding and feedforward intervention of complex peristaltic behavior.

[0003] Third, while traditional implantable systems can continuously monitor signal changes, they cannot achieve virtual-real synchronization. This means they cannot map the collected physiological behaviors into a digital model in real time, nor can they perform intervention simulations and early warning tests using a digital twin. Fourth, existing systems lack a biomimetic mechanism for multi-source data fusion. Their regulation is based on a linear correlation between heart rate and gut nerve signals, and cannot embed multi-source heterogeneous data such as electromyography, MRI images, pressure distribution, and fluid velocity. This invention, however, integrates multimodal sensor information to form an excitation field-driven model, truly establishing a full-chain digital twin system from neural activation to structural response to behavioral simulation. Fifth, regarding intelligence, existing regulatory models lack self-learning and error memory capabilities, and do not possess the function of trend prediction and active control based on historical behavioral deviations. Their regulatory mechanism is a passive, reactive feedback driven by fixed logic. Sixth, in terms of system scalability, existing technologies, due to their reliance on implantable control and interaction with external physical hardware, have weak scalability and cannot easily adapt to the functional modeling needs of individuals with different anatomical structures, age groups, and disease types.

[0004] In summary, while existing technologies possess certain hardware integration innovations and engineering practice foundations in the regulation of gastrointestinal function, they still suffer from numerous shortcomings in core areas such as the construction of truly meaningful digital twin models, spatiotemporal simulation of gastrointestinal behavior, virtual-real linkage feedback, error prediction and correction, peristaltic behavior evolution analysis, and visualization semantic output. These shortcomings include static structure, delayed feedback, simplified mechanisms, weak intelligence, and poor scalability, making it impossible to support the upgrade needs of precision medicine and future intelligent digestive systems. Summary of the Invention

[0005] The purpose of this invention is to provide a spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins, thereby addressing some of the drawbacks and shortcomings pointed out in the background art.

[0006] The present invention addresses the aforementioned technical problems by employing the following technical solution: a spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins, comprising: abstracting the neuromuscular activation process of the gastrointestinal tract into a physiological excitation field, wherein the physiological excitation field determines key parameters such as the initiation point, wave velocity, and rhythm of muscle peristalsis; constructing a variable topological strain network of the gastrointestinal wall based on the physiological excitation field, wherein the structure of the variable topological strain network adapts to neural activation, forming a functional contraction path; and introducing a set of control equations to make the excitation field exhibit wave evolution in time, tension distribution in space, and geometric deformation in structure. The study utilizes error field analysis to identify behavioral differences between twins and the real gastrointestinal system; it transmits error feedback to excitation field parameters to achieve closed-loop regulation; it integrates historical state data with current feedback and uses a time-series graph network to model peristaltic trends; and it evolves from fitting reality to predicting future behavior. The continuous peristaltic process is mapped as a spatiotemporal activity density map; topological convolution is used to extract activity mutations, peristaltic stagnation, and backpropagation features in key regions; the output includes regions with low transmission efficiency, stress concentration areas, and suspected spasm sources, which are reconstructed into semantic indicators that are easy for doctors to understand.

[0007] Furthermore, the method for constructing a variable topology strain network for the gastrointestinal wall using the physiological excitation field includes: The gastrointestinal wall region is discretized into multiple muscle nodes, and an initial network topology consisting of multiple nodes and connecting edges is constructed. The nodes represent muscle units, and the edges represent muscle synergy or stress propagation paths. Gastrointestinal nerve signals are collected, and a physiological excitation field reflecting the spatial distribution of nerve activity is constructed. The excitation field is used to control the activation state of network nodes. Based on the spatial intensity and propagation direction of the physiological excitation field, the connection relationship between nodes in the network is dynamically adjusted to form a variable topology, wherein the connection of nodes in the activated region is enhanced, and the connection in the non-activated region is weakened or broken; based on the change of the network topology, the strain state and muscle tension of each node are dynamically calculated to form a muscle contraction path consistent with the actual neural activation behavior. By combining contraction pathways with time-series driving, the propagation and evolution of gastrointestinal peristalsis behavior in the spatiotemporal domain are simulated.

[0008] Furthermore, the physiological excitation field is constructed based on the neuronal group signals of the gastrointestinal autonomic nervous system, and the signal sources include electromyography, bioelectric field distribution map or neural stimulation data; the connection relationship change mechanism of the network structure is based on dynamic connection weight adjustment rules, which combine the current excitation intensity and historical activation trend to determine the increase, decrease or breakage of connection edges.

[0009] Furthermore, the contraction path is naturally formed by the stress propagation path during the propagation of the excitation field within the network; the variable topology strain network and the actual gastrointestinal physical sensing data, including local tension, content flow velocity, and pressure feedback, form a virtual-real synchronous feedback mechanism to adjust the excitation field parameters and enable the adaptive correction of the twin model. The gastrointestinal wall is discretized into a multi-node strain network, where each node represents a muscle unit and each edge represents a potential muscle synergy. The excitation field originates from one or more neural activation sources and propagates dynamically within this network, its path influenced by local muscle state, structural coupling strength, and physical feedback. An adaptive propagation stress function is used to model this propagation path, as follows: in: Indicates the location and time The stress-driven excitation propagation potential energy function is the basis for the formation of the contraction path; The elastic response coefficient of muscle structure represents the spatial distribution and reflects the physical deformation characteristics of different tissue parts; This represents the strain error factor acquired by the local sensor, which comes from the tension or pressure deviation feedback of the real gastrointestinal system; This represents the global time evolution function of the current excitation field, reflecting the fluctuation trend of the neural signal throughout the network; This represents the muscle synergy strength field calculated after virtual-real fusion, and represents the functional connectivity between nodes; This represents the sensitivity parameter of virtual-real coupling, which can adjust the weighting of the feedback response under different gastrointestinal states; This scheme constructs an error-driven excitation and regulation mechanism, using sensor data as input, to analyze the deviation between the twin model and the real gastrointestinal tract in real time, and through... The above formula introduces feedback weights to regulate the excitation propagation process, causing it to converge to a physical state consistency; this evolves a potential energy minimum propagation trajectory that more closely approximates the actual physiological peristaltic path; this mechanism can be updated in real time within each simulation cycle, and the stress propagation path... The system will adjust based on feedback, gradually optimizing the location, intensity, and propagation direction of the excitation source to form a closed-loop regulation system.

[0010] Furthermore, the method by which the governing equations enable the excitation field to exert temporal, spatial, and structural effects includes: A digital twin structure of the gastrointestinal wall region is constructed, including multiple muscle unit nodes and their connections; a physiological excitation field is initialized, which reflects the activation state of neural signals in the gastrointestinal wall space; The excitation field is driven by the control equations to evolve continuously in the time dimension, forming periodic, delayed or staggered excitation patterns; the excitation field diffuses in the spatial dimension and forms a tension gradient, driving muscle units in different regions to produce differentiated stress states. Based on the distribution of excitation tension, the muscle unit is induced to have a geometric response in its structure, including shortening of length, increase of thickness or distortion of shape, which causes geometric deformation on the peristaltic path; excitation propagation, tension formation and structural deformation behavior are used as the basis for peristalsis driving, and the digital twin is used to simulate the gastrointestinal peristalsis process in the spatiotemporal range.

[0011] Furthermore, the set of control equations includes multiple control mechanisms for describing excitation initiation, propagation speed, energy decay, signal interference, and reflex behavior, to simulate the natural propagation process of actual neural signals in the time dimension; the propagation of the excitation field in the spatial dimension can form tension regions of different intensities, constituting muscle synergistic contraction regions, tension release buffer regions, and content propulsion channels, enabling regional functional division of labor.

[0012] Furthermore, the response to the structural deformation is dynamically controlled by the spatial intensity and propagation direction of the excitation field, forming a traceable and predictable peristaltic path and dynamic morphological evolution of the gastrointestinal wall; the propagation behavior of the excitation field and the structural response are coupled, and the integrated dynamic linkage of control source, tissue function and structural geometry is realized through the control equation set; Specifically: The local structural deformation of the gastrointestinal wall muscle tissue is controlled based on the spatial intensity and propagation direction of the excitation field, and a high degree of coupling and linkage between excitation behavior and structural response is achieved through a set of governing equations; by constructing a traceable and predictable peristaltic path generation mechanism, neural excitation is used as the control source and muscle tissue deformation as the response terminal, and spatiotemporal continuous transmission is achieved in the digital space through nonlinear functional relationships. in: In position and time Local structural deformation of the gastrointestinal wall; After scaling factor The weighted excitation field spatial intensity function represents the degree of direct influence of the excitation source on the local region; The gradient of the propagation direction of the excitation field in space; Custom coupling transformation functions are used to map directional propagation information to organizational deformation response (e.g., convolutional or parabolic transformations). Tissue sensitivity coefficient reflects the ability of different tissue structures to respond to the direction of excitation (including the density of muscle fiber arrangement). The structural nonlinear deformation gain factor controls the explosiveness or smoothness of the response, adapting to the muscle contraction characteristics at different stages of peristalsis.

[0013] Furthermore, the process of constructing the closed-loop regulation includes: Data on the peristaltic behavior of a gastrointestinal twin during a simulated period are obtained, including muscle tension distribution, peristaltic path, and contents flow direction; actual physiological data within the corresponding time and space range of the real gastrointestinal system are collected simultaneously; the twin data and the real data are compared in spatial and temporal dimensions to generate an error field map, wherein the error field is a continuous multidimensional distributed data structure used to describe the behavioral deviation between the simulated state and the real state; Based on the high deviation region in the error field, identify abnormal features such as activation path mismatch, fluctuation delay, or creep decay, and establish a mapping relationship between error and excitation source; apply error feedback to the physiological excitation field of the twin model to adjust the excitation parameters such as neural activation intensity, diffusion rate, activation frequency, or spatial distribution range.

[0014] Furthermore, the error field is a continuous difference map constructed in the space-time dimension to identify regional deviation trends, error aggregation points and their propagation paths; the adjustment object of the error feedback is the excitation field source parameters of the twin, so as to modify behavior at the level of neural activation mechanism.

[0015] Furthermore, during the feedback adjustment process of the excitation field, predictive adjustment is performed based on the evolution trend of the historical error field to correct parameters in the potential deviation region in advance, thereby achieving feedforward control; the error evolution trajectory is recorded in multiple feedback cycles to form an error memory mechanism, which is used to improve the adaptability and stability of the twin model to individual physiological characteristics.

[0016] The beneficial effects of this invention are as follows: By constructing a physiological excitation field based on neuromuscular drive and combining it with a spatial topological strain network and a set of control equations, the system can simulate the entire process of peristalsis, including initiation, propagation, tension response, and structural deformation, with high fidelity, thus restoring the dynamic movement characteristics of the gastrointestinal tract under natural conditions. Furthermore, by integrating multimodal data from the patient (such as electromyography, MRI, and pressure sensors), a one-to-one mapping between the model and the patient is achieved, supporting personalized customization and evolution of peristalsis patterns, and ensuring that the simulation results closely match the individual's physiological state.

[0017] By introducing the concept of an error field and a closed-loop regulation mechanism, the differences between the twin model and real gastrointestinal behavior can be continuously compared, and the excitation source parameters can be adjusted in real time to form a virtual-real feedback closed loop of perception-decision-control, significantly improving simulation accuracy and stability. Through error evolution trend learning and memory mechanisms, it can not only fit the current state but also predict future behavioral deviations, achieving feedforward regulation and effectively intervening in potential dysfunctional areas in advance, possessing a kind of intelligent evolutionary capability. The constructed spatiotemporal peristalsis visualization atlas and semantic diagnostic indicators (such as tension accumulation areas, peristalsis stagnation areas, and spasm sources) present complex physiological processes in an intuitive form, greatly facilitating doctors to conduct dynamic observation, etiological analysis, and intervention strategy formulation. Attached Figure Description

[0018] Figure 1 This is a flowchart of the spatiotemporal simulation of gastrointestinal peristalsis based on digital twins, as described in this invention.

[0019] Figure 2 This is a diagram showing the closed-loop regulatory function relationship of the gastrointestinal twin model based on the error field according to the present invention.

[0020] Figure 3 This is a diagram showing the relationship between the visualization of complex peristaltic behavior and the medical semantic transformation function of this invention.

[0021] Figure 4 This is a simplified flowchart of the implementation process of digital twin modeling of gastrointestinal motility disorder in Wang, according to Embodiment 1 of the present invention.

[0022] Figure 5This is a flowchart illustrating the implementation of the digital twin model control and multi-layer response for Wang's gastrointestinal tract in Embodiment 2 of the present invention.

[0023] Figure 6 This is a flowchart illustrating the implementation of closed-loop regulation and predictive control of Wang's gastrointestinal twin model in Embodiment 3 of the present invention.

[0024] Figure 7 This is Example 4 of the present invention, showing ultrasound images of a 49-year-old male patient with acute heart failure. Figure 7 A) Maximum internal diameter of the hepatic vein; Figure 7 B) Smooth muscle actin (SMA) inner diameter; Figure 7 C) Smooth muscle actin (SMV) inner diameter; Figure 7 D) Thickened gastric antrum wall; Figure 7 E) Thick jejunal wall; Figure 7 F) Thickness of the ascending colon wall; Figure 7 G) Maximum diastolic area of ​​the gastric antrum; Figure 7 H) Minimum contraction area of ​​the gastric antrum. Detailed Implementation

[0025] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0026] Combined with appendix Figure 1In a digital twin-based spatiotemporal simulation method for gastrointestinal peristalsis, the neuromuscular activation process of the gastrointestinal tract is first modeled and abstracted to construct a physiological excitation field with physiological driving characteristics. This field simulates the regulatory state of the gastrointestinal autonomic nervous system on muscle tissue. Spatially, this field covers the gastrointestinal wall region, and temporally, it exhibits dynamic propagation characteristics. Its internal parameters include activation initiation point, excitation intensity distribution, propagation speed, rhythmic control frequency, and energy attenuation coefficient. This field, as the driving source, determines key control factors such as the starting position, propagation path, rhythm waveform, and cycle length of gastrointestinal peristalsis. Based on this, a variable topological strain network for the gastrointestinal wall muscle tissue is constructed. This network consists of multiple nodes representing local muscle groups, with potential cooperative relationships established between nodes through connecting edges. The network's topology has adaptive adjustment capabilities; when the physiological excitation field intensifies in a certain region, the corresponding region's network... Network nodes will form higher connection density and stronger strain synergy, thereby constructing a local contraction core area. Conversely, the connection relationship in inactive areas will be weakened or broken, indicating a state of muscle relaxation or functional isolation. Therefore, the network can adjust its topological state in real time according to the changes in the excitation field during the simulation period, forming a functional peristaltic contraction path. Furthermore, a set of control equations is introduced to incorporate the evolution behavior of the excitation field into the dynamic control framework. This set of control equations considers the rate of change of time, spatial diffusion trend and local tissue properties, so that the excitation field manifests as a periodic evolution similar to a peristaltic wave in the time dimension, a directional diffusion propagation of the tension gradient in the spatial dimension, and directly acts on the muscle nodes in the variable topology strain network in the structural dimension, causing traceable deformation of the geometric structure, such as muscle shortening, thickening and torsion. This completes the full-link dynamic modeling process from neural signal modeling, stress propagation simulation, muscle network response to structural morphological changes.

[0027] Combined with appendix Figure 2To achieve high consistency and adaptive regulation between the twin model and the real gastrointestinal system at the dynamic behavioral level, an error field analysis mechanism is introduced. This mechanism synchronously maps the output data of the twin simulation with physiological data collected from the real gastrointestinal tract (including muscle tension distribution, peristaltic path, pressure changes, and contents flow velocity) in both spatial and temporal dimensions, constructing a multidimensional continuous error field map to describe the degree of inconsistency between the two under the same conditions. This error field not only captures single-point deviations but also identifies error clusters, evolutionary paths, and trend changes. Based on this error field, the system maps the identified high-deviation regions back to their control source, i.e., the input parameters of the physiological excitation field, including neural activation intensity, propagation direction, rhythm frequency, and diffusion range, and then converts the error data back into the input parameters of the physiological excitation field. The signal is transmitted as a feedback regulation signal to the excitation field control equations, realizing dynamic correction of the excitation field source, thus forming a closed-loop regulation mechanism of perception difference-transmission feedback-update control. On this basis, historical state data and current periodic feedback data are further integrated to construct the time series feature trajectory of peristaltic behavior. A time series graph neural network is used to learn and model the spatial-temporal co-evolution process of peristaltic state. This enables the system to not only gradually correct the current twin model behavior to fit the real state, but also to extract patterns from the existing historical and current states, identify behavioral trends, and predict future evolution paths. This achieves a leap from static fitting to dynamic prediction, enabling the digital twin to have the learning and forward-looking behavioral evolution capabilities of physiological peristaltic patterns.

[0028] Combined with appendix Figure 3This method achieves the visualization and medical semantic representation of complex peristaltic behavior. It jointly maps the continuous peristaltic process simulated by twins across time and space, generating a spatiotemporal activity density map with dynamic distribution characteristics. This map, based on spatial coordinates, numerically encodes information such as activation intensity, peristaltic amplitude, and tension changes of gastrointestinal wall muscle units at each moment, dynamically stacking them on the time axis to form a three-dimensional spatiotemporal dataset. This map accurately reflects the spatial propagation efficiency and temporal continuity of peristaltic activity. Through this density map visualization method, it achieves full-cycle coverage of peristaltic behavior and focused identification of key abnormal patterns. Furthermore, a topological convolutional network is introduced to extract multi-scale features of the spatiotemporal activity map, both local and global. The topological convolution is used to capture the changing trends of connections between local muscle groups and the propagation paths of abnormal signals, accurately identifying peristaltic mutation points, stagnant segments, and propagation points. The system infers functional structural abnormalities by analyzing complex behavioral characteristics such as directional reversal and combining the evolution trend of the topological structure between nodes with the signal distribution status. Based on the above high-dimensional feature analysis results, the system performs semantic reconstruction on the activity abnormalities in key areas, generating a set of clinically interpretable indicator types, including but not limited to areas with low transmission efficiency (indicating abnormally slow propulsion of contents or local dynamic disorders), stress concentration areas (indicating areas of excessive force on muscle tissue and high potential risk of damage), and suspected spasticity sources (representing the starting point of abnormal excitation intensity or frequency concentration, triggering forced peristalsis or pain). These semantic indicators are presented in the form of structured labels on the digital twin visual interface and have the ability to register and interact with actual images. This facilitates the rapid acquisition of the spatial location, change trend, and clinical suggestions of key functional abnormality areas without the need for in-depth understanding of complex model data, thereby improving the interpretability, diagnostic assistance value, and operability of the model in real medical scenarios.

[0029] Example 1: Combined with appendix Figure 4In this embodiment, a 55-year-old male patient, Mr. Wang, presented with intermittent abdominal pain, bloating, and constipation for three months. Preliminary clinical diagnosis indicated gastrointestinal motility dysfunction. A gastrointestinal motility simulation analysis was performed to aid in diagnosis and intervention planning. During implementation, the system monitored Mr. Wang's non-invasive neuroelectrical activity, using magnetic resonance imaging (MRI) and high-resolution manometry (HRM) to obtain three-dimensional structural information of his gastrointestinal tract and the electromyographic signal change patterns in different regions of the gastrointestinal wall. The system discretized the entire gastrointestinal wall in a digital twin model, dividing it into approximately 320 muscle nodes, each representing a 1 square centimeter muscle unit. The connecting edges between nodes were based on… A basic network topology was established based on anatomical adjacency and the synergistic relationship of electrical signals between muscles. Then, based on the collected neural signal activity data, the neural activity distribution map, i.e. the physiological excitation field, was reconstructed in space. In Wang's specific data, it can be observed that there are concentrated areas with significantly higher neural activation intensity in the lower part of the stomach body, the beginning of the small intestine, and the proximal transverse colon. The system identifies these strongly activated areas as excitation source points and begins to map the excitation intensity to the corresponding node activity. The connection edge weight of the muscle nodes in the activated area increases significantly, forming a stronger contraction synergistic response. In contrast, the area far from the excitation center shows sparse or broken connections, forming obvious functional partition boundaries. Based on this, the system further applies tension calculations to each node. According to the actual physiological strain properties of the muscle tissue and the modulation intensity from nerve signals, the local strain rate is dynamically calculated. The results show that Wang formed a significant contractile tension depression in the gastric antrum, meaning the excitation field caused synchronous muscle contraction, forming a functional peristaltic wavehead. Subsequently, the system embeds this contraction path into the time-series control module, extrapolating the propagation path and speed of the peristaltic wave in a 5-second cycle within the current simulation period. During the simulation, the propagation speed of the contraction path from the lower part of the stomach body to the pylorus was observed to be 3.2 seconds per second. The simulated wave velocity was measured in centimeters and accompanied by slight wave velocity swing. At the 4.5-second mark of the simulation, a wave head stall phenomenon was formed in the proximal duodenum. This phenomenon was also reflected in the real pressure data, verifying the high degree of consistency between the twin behavior and the patient's real physiological state. After the model ran for a total of 45 seconds, three suspicious areas of low transmission efficiency and one suspected reverse peristalsis starting point were marked. Subsequently, by comparing the simulated output map with physiological images, it was confirmed that Mr. Wang had early gastric antral nerve function regulation disorder, not organic obstruction. Finally, it was recommended to use mild prokinetic drugs in combination with biofeedback training to avoid unnecessary surgical intervention.

[0030] After completing the initial peristaltic pathway modeling, to further improve the dynamic accuracy of the simulation model, a neural source-driven physiological excitation field construction and network connection dynamic adjustment mechanism were introduced. During implementation, Wang underwent 48 hours of continuous gastrointestinal neuromonitoring, including multimodal neuromyograph data acquisition in both resting and postprandial states. The equipment used included a multi-channel surface electromyography (sEMG) system, an electrogastrography array electrode system, and a novel electric field spatial sensing array. A total of 32,000 signal samples were obtained. In the data processing stage, the system first normalized and temporally aligned the neuronal signals from different sources, constructing a dynamic neural activity atlas covering the gastric body to the sigmoid colon. This atlas was refreshed every 1 second. Each frame of the atlas corresponds to a reconstruction of the physiological excitation field in the spatial dimension. The neural signals are further mapped to the muscle nodes on the digital twin to obtain the current neural activation value, fluctuation frequency and historical trend weight of each node. In the upper part of Wang's small intestine, nodes 16 to 25 showed abnormally high frequency sudden neural peaks for three consecutive cycles. Based on this, the system identified the region as a high fluctuation source point and assigned it a high activation factor in the network. Next, the system updates the weight of each connection edge in real time according to the proposed connection weight dynamic adjustment rule. If the current neural excitation intensity of two nodes is highly synchronized, the excitation frequency is similar, and they have maintained a continuous activation state in the past three cycles, the edge weight is increased and the cooperative tension propagation coefficient is improved. Conversely, if there is a disconnect in excitation, frequency misalignment, or conduction lag between nodes in the current or previous period, the connection edge is weakened or disconnected. For example, in Wang's model, the neural excitation values ​​of nodes 87 and 88 fluctuated dramatically in opposite directions over the past 4 seconds. Ultimately, the system reduced the weight of this connection edge from 0.65 to 0.12 and disconnected it in the next cycle to reflect functional discoordination. As time progresses, the model network topology gradually evolves from an initial equal-weighted grid-like connection state to a highly partitioned self-organizing map. Its structure highly matches the asymmetric distribution characteristics of Wang's gastrointestinal neural regulation pattern. In the 12th simulation cycle, the system identified 4 tension aggregation clusters and 2 isolated excitation patches, which were marked as uncoordinated muscle groups in the model. These regions precisely correspond to the focal tension abnormalities in the gastric wall observed by Mr. Wang on MRI images, verifying the performance of the connection weight reconstruction mechanism driven by the physiological excitation field in the actual physiological level. In addition, by introducing historical neural activation trajectories into the connection edge decision factor, the system no longer judges the topology solely based on real-time data, but integrates the current state and trend prediction to construct a contraction coordination relationship network with time memory. For example, in the middle section of Mr. Wang's large intestine, due to the oscillating pattern of delayed activation-premature fallback in local excitation in three consecutive rounds of simulation, the system inferred that there was short-cycle coupling disorder in this region. Finally, the model adjusted the network in this region to an unstable topology and suggested considering it as a risk area for functional colonic spasm.

[0031] Given that Wang's previous digital model had already completed muscle node discretization, neural excitation field construction, and adaptive adjustment of connection weights, virtual and real data fusion was introduced, and a stress propagation function was constructed. The study simulated the formation of actual contraction paths during peristalsis and used gastrointestinal physical sensor data for feedback correction. Wang's gastrointestinal wall was modeled as a two-dimensional strain network containing 256 muscle nodes, each node representing a 1cm² muscle unit. The initial connection strength between adjacent nodes was determined based on prior neural coordination signals and structural coupling coefficients. Real-time tension data acquired from actual equipment included high-frequency dynamic gastric electrical signals (sampling frequency 100Hz), local gastric wall pressure values ​​(range 12–45 mmHg), and changes in the rate of contents transport (range 0.3–4.5 cm / s). This information was fused and projected as a strain error factor. For example, a persistent tension delay was detected in the gastric antrum region (node ​​numbers 85–92), which the system identified as an error enhancement area. The values ​​at these nodes range from 0.2 to 0.4, while the neural activation evolution function at synchronization time... The value is 0.65 (unitized excitation intensity), and its time derivative is 0.18 / s, indicating that the neural excitation is in a slow enhancement phase at this moment; simultaneously, the corresponding spatial elastic response coefficient... The tissue modeling module identified a value between 0.55 and 0.70, indicating that the muscles in this area are relatively soft and easily deformable; synergistic strength field The measured values ​​ranged from 0.6 to 0.9, indicating a certain degree of interaction between muscle groups in this area; the virtual-real coupling sensitive factor... In this test, a value of 1.5 is set at the midpoint of the recommended adjustment range of 1.2–2.0, enhancing the feedback's influence on the propagation path; substituting the above values ​​into the formula: At node 89, the spatial elastic gradient change is set to 0.12 / cm. , , The calculation result is: This stress potential energy value is used by the system to evaluate the preferred path selection for excitation field propagation. The system considers all nodes... After normalization, a propagation potential energy heatmap was plotted. The results showed that the peak potential energy in the gastric antrum region was between nodes 89 and 90, which became one of the starting points for the systolic wave propagation. During the synchronous simulation cycle, the system recorded multiple consecutive moments... The trajectory change identified a contraction path advancing from nodes 89→90→94→98, while at node 92, a sudden increase in pressure feedback (from 24 to 39 mmHg) caused... The value changed drastically, jumping from 0.32 to 0.51. The system identified this as a high-error disturbance region, adjusted the propagation path to the direction of nodes 91–95, and adjusted it again in the next cycle. The excitation intensity was fine-tuned to 0.68 to smooth the propagation impact and prevent the model from deviating further from reality. Ultimately, in the third simulation cycle, the fit between the propagation path and the actual electrogastrogram activity trend improved from 82.3% in the first round to 94.1%, with the closed-loop feedback mechanism successfully correcting the propagation direction. The system then continuously optimized the stress path, marking two newly formed structural blockage points after the sixth cycle, indicating muscle fiber rigidity or focal neural modulation failure in this area. It was recommended to combine this with endoscopic examination to confirm the presence of early smooth muscle degeneration or ganglion conduction disorders.

[0032] Example 2: Combined with appendix Figure 5 Based on Example 1, in the construction of Wang's digital twin gastrointestinal system, the model control core stage is entered. Based on the mechanism of the control equation set driving the excitation field to play a role in the three dimensions of time, space and structure, the model is further refined to simulate its complete peristaltic process. Based on the structural modeling results of Wang's gastrointestinal region, the system first constructs a digital twin structural framework containing 312 muscle unit nodes. 568 connecting edges are determined between the nodes according to anatomical level, nerve innervation area division and muscle synergy to form a dynamic graph network. This structure covers all the main peristaltic units from the stomach body to the middle of the colon in the spatial dimension. Based on this, the system initializes the physiological excitation field, using the time-series fluctuation curve extracted from Wang's electromyography data as the activation source function. Multiple neural excitation origins are set in the gastric antrum and proximal jejunum regions with high node density, with an initial excitation intensity of 0.65 (normalized scale). Simultaneously, a time-recursive control equation is introduced to continuously diffuse the excitation signal outwards at 0.2-second intervals, forming a periodic and interlaced propagation trajectory with realistic physiological characteristics. In the third simulation cycle, the system simulates a significant rhythmic drift between the excitation bands in the gastric body-antrum region, specifically manifested as a leading wave... Arriving at the small intestine initiation zone 0.6 seconds earlier than the main wave, this behavioral characteristic was reproduced by adjusting the time evolution delay term in the control parameter set. Simultaneously, the excitation field generated a tension gradient in the spatial dimension, forming a high-tension boundary region on the left side wall of the stomach body. The local node tension values ​​were between 4.1 and 5.6 N / cm², significantly higher than the average muscle wall tension (2.3 N / cm²). The system thus identified this as a stress accumulation zone and activated the structural response module. Based on the excitation tension distribution and the material properties of the muscle nodes (such as elastic modulus, thickness sensitivity factor, etc.), the morphology of the local muscle unit was calculated. The changes, particularly at nodes 145–149, showed a 2.8% shortening in length and a 0.6 mm increase in thickness in the model. Comparison with MRI images revealed a slight inward arching deformation of the gastric wall in the corresponding areas in real-world images, further demonstrating the model's structural response accuracy. Furthermore, the governing equations nested morphological coupling functions at the structural layer, enabling the geometric deformation between adjacent nodes to propagate, thus realistically reproducing the dynamic process of muscle chain contraction-relaxation during peristalsis. This mechanism demonstrated high fidelity in the 6th cycle simulated by Wang, within 3.4 seconds. Between 5.1 seconds, the contraction wave successfully advanced from the pylorus to the middle of the duodenum, with an overall propagation rate of 2.7 cm / s. The system simultaneously outputs a three-layer superimposed map of excitation value, tension distribution, and structural deformation for each time slice, which is used to retrospectively analyze the evolutionary causal chain of the model's behavior. The final simulation output results show that Wang's gastric antral peristaltic thrust was significantly weak and the frequency was unstable, while the jejunal initiation segment showed premature excitation accompanied by forced structural deformation. Based on this, the system proposed that there is a physiological mechanism of delayed autonomic nerve feedback or abnormal local stress sensitivity in this area, and suggested further regulation of its excitation stability through drugs.

[0033] Based on the completion of structural construction, initial establishment of excitation pathways and peristalsis simulation, in order to further conform to the real gastrointestinal neurophysiological characteristics and identify potential functional area disorders, a set of control equations driven by multiple control mechanisms is introduced to comprehensively simulate the natural propagation process of neural signals in the time dimension and their response expression in the spatial structure. In the control mechanism setting, the system sets the excitation initiation rule based on the earliest activation time period in Wang's electromyography recording (i.e., continuous enhanced discharge signal in the prepyloric region within 2 minutes after a meal), determining the initial nerve excitation point as node 143, and defining the initial excitation intensity as 0.78 (normalized units). The propagation velocity parameter is calculated from the local muscle fiber density and tissue elasticity factor, and is 3.1 cm / s. The propagation rate term in the control equation is dynamically updated, so that the signal exhibits non-uniform contraction fluctuation characteristics in different regions. In terms of the energy attenuation mechanism, combined with the obvious signal amplitude reduction observed in Wang's data during the propagation of the excitation in the anterior wall region of the stomach, the system introduces an attenuation function based on the propagation distance and muscle viscosity coefficient. In the simulation, the excitation field energy naturally attenuates from the initial value of 0.78 to 0.43 when it propagates to the node 137–140 region, effectively reproducing the signal dissipation characteristics. Simultaneously, signal interference and reflection rules are introduced. When the excitation wave encounters a muscle node that has not fully recovered from the previous cycle during propagation, the system triggers phase interference judgment in the control equation and fine-tunes the amplitude and propagation direction of the current waveform. For example, in the Wang's colon simulation, nodes 230 and 234 have residual excitation from the previous cycle, causing subsequent excitation waves to partially overlap and interfere. The system judges this as non-destructive overlap and reduces the amplitude of the wave in this area by 12% and deflects the propagation direction by 7 degrees according to the set mechanism, reflecting the local interaction phenomenon that is common in the propagation of real gastrointestinal signals. In addition, in the region of nodes 195–198, due to the presence of a significant muscle wall hardening area (extracted from MRI and elastomer imaging), the system simulates the reflection of the excitation wave at this location. Its propagation direction is reversed and it is transmitted back to nodes 180–183 with 40% energy. This mechanism realistically reproduces the influence of gastrointestinal lesions on the reflection and disturbance of peristaltic waves in the simulation. At the spatial propagation level, the excitation field, controlled by the governing equations, gradually delineates various functional tension regions in Wang's gastrointestinal model. Simulation data shows that in the region at the junction of the gastric antrum and jejunum, the tension value continuously reaches above 4.2–5.1 N / cm². The system marks this region as a muscle-coordinated contraction zone, which is the main force source for effectively propelling chyme transport. Behind this region, at nodes 112–117, the tension value rapidly decreases to 1.6 N / cm². The system designates this region as a tension release buffer, used to receive upstream peristaltic propulsion waves and reduce local rebound energy. Further downstream, at nodes 130–138, a channel with a continuously decreasing tension gradient is formed, with the tension change rate maintained at -0.22 N / cm². cm was defined by the system as the content propulsion path. The chyme propulsion speed was tracked along this path and increased by 34%, which was basically consistent with the actual tracking speed of the sensor's internal markers, verifying the physical correspondence of the functional division of the simulated area by the digital model. The system then annotated these areas with heat maps and output tension zoning maps and dynamic evolution curves of functional areas. Through the visualization interface, it quickly identified that the tension release function of the posterior gastric antrum of Mr. Wang was too strong, forming a dynamic break point. Based on this, the system suggested that the area be subjected to nerve stimulation modulation in the intervention plan to rebuild the continuity of tension waves. Finally, after two weeks of follow-up intervention, the patient's symptoms were significantly relieved, the reflected wave amplitude in the simulated data decreased by 21%, and the co-contraction area expanded by 6.8%.

[0034] Building upon the gradual realization of structural topology reconstruction, neural excitation modeling, stress path propagation, and tension feedback closed-loop regulation in Wang's digital twin gastrointestinal peristalsis simulation system, a higher-order coupling mechanism is further introduced, dynamically controlled by the spatial intensity and propagation direction of the excitation field, to achieve precise simulation and dynamic prediction of peristaltic behavior at the physical structural layer. Specifically addressing Wang's issues of insufficient contraction coordination and weakened propulsive force in the gastric body to antrum region, the system models the local geometric deformation process of the muscle layer by constructing a nonlinear coupled structural deformation function. The core control mechanism is expressed by the following formula: All parameters in the formula are calibrated based on Wang's integrated parameters and data collected by actual sensors. Among them, the excitation field spatial intensity function... High-density electromyography potential values ​​from the prepyloric region were obtained, and then... After weighting with a region magnification factor of 1.2, the peak value is 0.92, and the mean value ranges from 0.58 to 0.76; excitation directional gradient Modeling was performed by altering the main direction of the peristaltic propagation path over multiple cycles, specifically measuring the instantaneous rate of change of direction at nodes numbered 145–152 on the posterior wall of the gastric antrum. This reflects poor stability of neural signal direction; the system sets a tissue sensitivity coefficient to enhance coupling strength. =1.6, this value is in the upper-middle range of the suggested range of 1.2–2.0, indicating that the direction of muscle fiber arrangement in this region is extremely sensitive to the direction of nerve activation; coupling function An improved parabolic response mapping function is employed, with the output range set to 0.4–1.0, to smooth out stress response differences caused by different propagation directions. Furthermore, to simulate the asymmetric dynamic characteristics of rapid muscle contraction and slow recovery during peristalsis, a nonlinear gain exponent is used. The value is set to 1.8, which is suitable for waveform enhancement scenarios with high-intensity mutations and low-intensity recovery, reflecting the phased burst contraction characteristics of the gastric antrum.

[0035] Substituting specific numerical values ​​into the formula for an example calculation, taking the actual input data of node 148 on the posterior wall of Wang's stomach body at the 3.2-second simulation cycle as an example, we know: , , ,but: This result indicates that the local muscle deformation at node 148 at that moment was approximately 0.209 units (e.g., normalized thickness increment, local contraction rate, etc.), which manifests as slight shortening and lateral expansion deformation in the simulation atlas. Comparison with MRI structural images revealed a 3.1 mm thickness change in the muscle wall morphology within the cavity, validating the simulation accuracy. Subsequently, the system continuously... Spatiotemporal trajectory backtracking revealed a clear propagation deformation waveband in the middle segment of the gastric antrum. The path, from node 145→149→153, propagated forward in an elliptical shape with a total propagation rate of 2.2 cm / s and a corresponding time delay of approximately 1.3 seconds. The system labeled this path as the "main propagation deformation chain." Comparison with Wang's content logistics tracking data showed that the propagation efficiency of this path region was improved by approximately 31%. This indicates that the constructed structural response mechanism not only accurately matches the actual dynamic deformation of the gastrointestinal tract at the morphological level but also effectively supports the evolution prediction of peristaltic dynamics at the behavioral level.

[0036] Furthermore, in the simulation of the region from node 138 to 142, the elastic parameters of the muscle wall in this segment are relatively high ( The mean value is only 0.48, and the directional response is ambiguous. Volatility is as high as ), resulting in the calculated When the value is below 0.08, the system identifies it as a deformation and passivation area, meaning that although the excitation propagates to this area, it fails to form an effective structural response. This area is the area where Mr. Wang complained of the most obvious abdominal distension. The corresponding endoscopic examination results also show that the area of ​​gastric wall peristalsis is not obvious, which further verifies the clinical consistency of the system in terms of the traceable and predictable characteristics of peristaltic deformation path.

[0037] Example 3: Combined with appendix Figure 6 Based on Example 2, in the mid-to-late stages of Wang's individualized digital twin gastrointestinal motility simulation project, a closed-loop regulation mechanism was initiated. This mechanism relies on high-frequency synchronization between real-time physiological data and the virtual twin model, and achieves excitation field regulation and peristaltic behavior correction based on error field construction and dynamic feedback control. During implementation, Wang underwent continuous 72-hour dynamic gastrointestinal monitoring, including simultaneous recordings of high-density electromyography, intraluminal pressure sensors, and content logistics tracking markers (capsule-type wireless tracking system). Based on this data, the system constructed a dynamic database of actual peristaltic behavior from the stomach body to the anterior colon, including the trend of tension changes per second, the velocity of contents movement, and the frequency of local muscle wall activation. During the same time period, Wang's digital twin model also ran synchronous simulation cycles with a step size of 0.1 seconds, recording virtual peristaltic behavior data. The system extracted three core indicators: 1) node-level tension distribution map; 2) structural propagation path vector map; and 3) content propagation timeline.

[0038] Subsequently, the system activated the error field generation module, which fused the output of the twin model with the real monitoring data in two dimensions: space (node ​​position matching) and time (synchronization time alignment) to generate a three-dimensional error field map. The map showed in Wang's model that the tension peak deviation in the region from node 46 to 61 was higher than 2.1 N / cm², the content advance delay exceeded 1.8 seconds, and the propagation direction drift angle reached more than 21 degrees. Based on this, the system marked this region as a high error accumulation area and determined it to be a mismatch type deviation of the excitation source path. Next, the system uses the gradient distribution of error values ​​to infer the trajectory of corresponding excitation parameter changes and establishes a mapping relationship between error and excitation source. For example, after identifying a continuous propagation delay phenomenon in the middle of the gastric antrum, the system found that the original activation intensity of this area was set to 0.68 and the diffusion rate was 2.3 cm / s, but the actual observed peristaltic wave propagation only reached 1.5 cm / s. Therefore, the system adjusts the excitation intensity to 0.74 according to the error spectrum trend, adjusts the diffusion coefficient to 2.8 cm / s, and increases the local activation frequency by 10% and expands its influence radius by 0.9 cm in the ganglion simulation module to enhance the regional excitation coverage.

[0039] These adjustments will take effect in the next simulation cycle. In the second round of comparison, the system found that the difference in the peak tension in the original high error region was reduced to within 0.6 N / cm², the propagation direction shift was reduced to within 6 degrees, and the propagation delay was shortened to 0.4 seconds, achieving an error decay rate of over 80%. More importantly, the model tracked the trend of error field changes and identified that the region had changed from significant mismatch to marginal convergence. The system reduced the adjustment amplitude and locked the excitation update frequency to avoid over-adjustment causing system oscillations. Throughout Wang's treatment cycle, this closed-loop adjustment mechanism was repeatedly applied 18 times, with an average convergence time of 7.4 seconds for the error field, ultimately improving the model's behavioral accuracy across the entire gastric segment to over 95.3%. The adjustment path, convergence trend, and current excitation state of each region can be viewed in real time on the system interface through the error evolution graph, assisting in the identification of functional abnormality sources and intervention nodes. After the treatment plan was adjusted, Wang's symptoms improved significantly, and the patient's abdominal distension and delayed emptying sensation were greatly reduced.

[0040] After Wang's digital twin gastrointestinal motility simulation entered a stable closed-loop regulation phase, the model was further optimized at the microstructural level using an error field mechanism constructed in a spatiotemporal dimension. The focus was on analyzing the root causes of regional functional abnormalities and attempting to achieve long-term dynamic stability through precise correction at the source level of the excitation field. Based on a comparison of Wang's simulation results from the previous cycle with actual physiological data, the system constructed an error field map covering the gastric body to the jejunum. This map was expressed in three dimensions, with the horizontal axis representing spatially distributed nodes (312 in total) and the vertical axis representing the time axis (total length 30 seconds, step size 0.1 seconds). Error values ​​were encoded with color intensity, representing the deviation between the tension response, peristaltic wave propagation path, and contents flow velocity of each node at each moment and the actual monitored values. After the map is generated, the system uses an image convolution scanning algorithm to identify regions where the error value is consistently greater than the threshold of 0.25 and marks them as error cluster points. Subsequently, by calculating the time series error change rate, the system found that a high error path is formed between node numbers 96 and 103, which is continuously propagating downstream with an average propagation rate of 2.1 cm / s. The direction is basically consistent with the main channel of the gastric antrum to duodenum. However, an error peak of 0.44 appears at node 100, indicating that there is a break in the neural activation-structural response chain in this region.

[0041] After performing a reverse matching between this error path and the spatial intensity and excitation source location of the current excitation field, the system found that the original excitation source was set at node 94, with a fixed neural activation intensity of 0.72 and a diffusion range of 1.8 cm. However, this configuration could not completely cover the extended area of ​​the error path in the last three cycles. Therefore, the system constructed an error-excitation response mapping function based on the error gradient, taking node 100, the point of maximum error, as the derivation center point, and assigned its inverse distance weight back to the source region. The result suggested shifting the excitation source location from node 94 to near node 96, expanding the diffusion radius to 2.4 cm, and increasing the excitation intensity to 0.79 to enhance the effective coverage of the distant region. In addition, the system detected a delayed creeping offset in this path based on the error evolution trend, that is, the model waveform arrives at the corresponding structural node 1.2 seconds before the actual behavior. The system determined that the excitation rhythm was mismatched with the physiological rhythm, and therefore lowered the frequency control parameter of the excitation source from once every 10 seconds to once every 12.5 seconds. After adjustment, the model regenerates the error field map in the next cycle. The system compares the changes before and after and shows that the average error has been reduced to below 0.09, the peak error on the propagation path has decreased from 0.44 to 0.13, the overlap between the creep path and the actual waveform has increased to 92.4%, the content logistics delay has been shortened by nearly 0.9 seconds, and the overall convergence cycle of the feedback mechanism has been shortened from 7.8 seconds to 4.6 seconds.

[0042] In the next round of observation, the system determined, based on the gradient flow field showing the error value changing over time, that some nodes were still accumulating hysteretic errors even without direct control from the excitation source, indicating a hierarchical delay in the neural control chain. Accordingly, the system further shifted control adjustment from the muscle node layer to the neural activation source configuration layer, fine-tuning the diffusion partial derivative weights of the waveform propagation function to create a smoother forward propagation of the excitation signal over time, reducing excitation offset at the mechanistic level. After the third adjustment, no further error accumulation or propagation anomalies were observed in this region during a continuous 10-second simulation cycle. The system determined that this region had entered a three-level convergence state of neural-structural-functionality and set this segment as the model reference segment to drive self-learning optimization in other regions. Ultimately, through multiple rounds of error field identification, path tracing, and control source adjustment, the overall peristaltic behavior simulation accuracy of Wang reached 96.7%, with the error drift rate approaching zero. The behavioral deviation prediction graph output by the system confirmed the absence of any active error bands, and the visualized dynamic peristaltic evolution trend was stable and consistent with the patient's clinical presentation.

[0043] After Wang's digital twin gastrointestinal motility system ran continuously for over 72 hours and underwent multiple rounds of error feedback closed-loop regulation, a predictive regulation and error memory mechanism module based on the historical error field evolution trend was introduced to achieve feedforward control of future behavioral deviations, further enhancing the model's adaptability and system stability. Initially, in the conventional closed-loop regulation mode, the system generated a current error field map by comparing it with actual physiological data after each simulation cycle, and then adjusted the parameters of the excitation field. However, in Wang's simulation cycles 8 to 11, the system found that although the error values ​​in some areas had been significantly reduced in the current cycle, these errors still recurred periodically in the following cycles, especially in the jejunal segment at nodes 173 to 180. Whenever a strong contraction occurred in the gastric antrum, a delayed error peak appeared in this area after 3.5 to 5.2 seconds, with the highest error value reaching 0.39. Content flow tracking showed that the propulsion speed in this area decreased by more than 25%. This phenomenon suggests that traditional hysteresis feedback can no longer meet the dynamic evolution requirements of complex neuromuscular coupled systems.

[0044] To address this issue, the system employs an error evolution trajectory recording mechanism. This mechanism serializes and stores the error field map for each simulation cycle and constructs an error evolution vector cluster, recording characteristic parameters such as the spatiotemporal location, duration, propagation direction, and propagation speed of error peaks. Based on the error trajectories of the past five cycles, the system performs trend modeling and identifies a latent deviation outbreak pattern at node 178, characterized by no error in the early stages followed by a sudden shift in the later stages. Based on this, the system predicts that this node will experience another sudden deviation around the 4th second of the 12th cycle, triggering a feedforward adjustment mechanism. This mechanism no longer waits for the deviation to occur before correction; instead, it proactively adjusts preventative parameters in the excitation source settings based on historical trends. The system increases the propagation speed between nodes 170 and 177 in the excitation field from 2.8 cm / s to 3.4 cm / s and fine-tunes the diffusion range to cover high-risk areas in advance. Simultaneously, a control function is introduced to regulate the activation frequency, causing the excitation rhythm to start 0.8 seconds earlier in this segment, thereby preemptively capturing the error outbreak window. After the 12th cycle was completed, the error value that should have appeared at the system verification node 178 had decreased from the predicted 0.38 to only 0.06. The propagation path remained stable, the content flow was smooth, the system prediction error was suppressed in advance, and the feedforward mechanism effectively completed the early warning intervention.

[0045] More importantly, during operation, the system marks the aforementioned nodes as recurrent deviation zones and establishes a deviation evolution profile for them in the error memory mechanism. This profile, indexed by time, records the error characteristic waveforms, excitation-response relationships, peristaltic output results, and intervention and adjustment history of this region at different cycles. When the model subsequently re-enters the behavioral simulation of this region, the system calls upon this error memory data and initializes and fine-tunes the excitation field parameters to improve the model's pre-adaptability to the individual characteristics of this node. For example, in cycle 15, when abnormal waveform interference occurs in the gastric body segment, the error memory mechanism triggers the system to adjust the excitation threshold of node 175–180 in advance at 2.2 seconds, successfully preventing the formation of potential delayed peristaltic path deviation.

[0046] Through continuous operation, the error memory bank gradually expanded to cover 96 high-frequency deviation nodes. The overall prediction and regulation success rate of the system reached 92.6%, and the average error response time was shortened from the initial 3.2 seconds to 1.4 seconds. The long-term stability of the model was significantly improved. Wang's gastrointestinal motility simulation model can now achieve advance behavioral regulation without relying on real-time error feedback. It has neural system-like learning and behavioral prediction capabilities and can use the error evolution prediction map and regulation suggestion timetable provided by the system to formulate targeted drug or neuromodulation intervention rhythms.

[0047] Example 4: like Figure 7 As shown, this embodiment selected a 49-year-old male who developed intractable abdominal distension and delayed gastric emptying 36 hours after being admitted to the ICU for acute heart failure (NYHA class IV). After obtaining ethical and family consent, the team... The prototype of the "Digital Twin-Gastrointestinal Motility Monitoring System" is started synchronously at all times.

[0048] 1. Bedside ultrasound acquisition: Hepatic veins with a maximum internal diameter of 0.98 cm were sequentially acquired. Figure 7 A), SMA inner diameter 0.47cm ( Figure 7 B), SMV inner diameter 0.52cm ( Figure 7 C) Gastric antrum wall thickness 0.77cm ( Figure 7 D) Jejunal wall thickness 0.22cm ( Figure 7 E), the wall thickness of the ascending colon is 0.26 cm ( Figure 7 F), the maximum diastolic area of ​​the gastric antrum is 15.40 cm², and the minimum contraction area is 9.09 cm². Figure 7 G, Figure 7 H). 2. Multimodal synchronous signal acquisition: gastrointestinal electrogastrography (4 leads, 200Hz sampling), focal pressure probe (8 points, 1kPa range), and food intake record (600mL liquid). 3. Image registration: the immediate CT-gastrointestinal scan was reconstructed into a 0.5mm resolution 3D surface mesh, which was used as the static geometric baseline of the twin model.

[0049] The CT mesh was simplified to 318 muscle unit nodes and 594 potential cooperative edges; the node density was increased in the gastric antrum, proximal jejunum, and ascending colon (edge ​​length 2–4 mm) to capture stress concentration areas. Then, a "material property tensor" was generated based on the local wall thickness and elastic modulus (24–36 kPa) and bound to the nodes to prepare for subsequent tension-deformation coupling. This discretization strategy is completely consistent with the "discretizing the gastrointestinal wall into a multi-node variable topology strain network" in the application, ensuring that the subsequent network weights can be dynamically adjusted.

[0050] Based on the dominant frequency of 3.2 cpm on gastric electrogastrography and the P wave frequency of 11.5 cpm in the small intestine, two levels of activation sources were set: a low-frequency source in the gastric antrum-pyloric region and a high-frequency source in the proximal small intestine. The activation intensity κ0 = 0.62 (normalized), and the diffusion coefficient... Attenuation constant The three-dimensional coupled PDE described in the application documents is adopted: Where Φ is the excitation potential. The source function is the neuromuscular coupling function; the potential gradient directly maps to nodal tension. The system dynamically adjusts the weights of the connecting edges based on the intensity and direction of the excitation field. Local authorities The algorithm enhances coordination when it is active and weakens or breaks it when it is not; the algorithm implementation is completely consistent with the "dynamic connection weight adjustment rule". By the end of the 5th simulation cycle, the network topology has self-organized into two high-weight main channels: the gastric antrum-jejunum propulsion chain and the ascending colon reverse peristaltic buffer chain.

[0051] Running for 30 seconds on a GPU cluster (16 TFLOPS) with a step size of 0.05 s: the peak of gastric antral contraction occurred at t=9.3 s, predicting the emptying rate. Jejunal propulsion wave velocity The pressure peak in the ascending colon was 5.8 kPa.

[0052] The simulation results were registered with bedside pressure-ultrasound synchronous data to generate a three-dimensional error field. The tension deviation is greatest in the region of nodes 87–103 (gastric antrum near the pylorus), with an average of Content delivery was delayed by 1.9 seconds; the direction of propagation shifted by 17°.

[0053] The error field generation and mapping mechanism fully conforms to the description of "spatial-temporal dimension synchronous mapping" in the application document.

[0054] Based on the high deviation region of the error field, the system will activate the intensity Increased to 0.71, diffusion coefficient D increased to An auxiliary activation source with a radius of 1.0 cm was set at the pylorus. A time-series graph neural network was introduced to learn the evolution trajectory of Δ; if Δ > threshold 0.3 within the predicted t+3s, κ and D were adjusted in advance to avoid error rebound. After the second round of simulation, the tension error in this region decreased to The directional drift is less than 6°; after the third round, the overall model accuracy improved to 96.1%, consistent with the "error decay rate exceeding 80%" stated in the application document. The system outputs three layers of interpretable information: a spatiotemporal density map: tension waves from the gastric antrum to the jejunum are displayed in thermal form, with high error areas flashing in red; topological convolution index: automatically marking one "low propulsion efficiency area" and one "stress concentration area"; and a adjustment log: a time-parameter-error tripartite table for ICU physicians to review at any time. This semantic output paragraph is consistent with the functional description in the application document.

[0055] Based on the system's recommendations, the doctor in Administer a low dose of prokinetic agent (itopride 50mg) and adjust hemodynamic parameters (target MAP 70mmHg). Follow up in 4 hours. Gastric antral emptying time decreased from 83 min to 42 min; abdominal circumference decreased by 3.1 cm; and the patient's complaint of abdominal distension was significantly relieved. After synchronous updating of the digital twin model, the activation frequency of the gastric antrum-jejunal channel automatically decreased by 6%, indicating that the system has captured the drug-neural coupling effect and achieved drug efficacy-model linkage. No data loss was observed after 48 hours of continuous operation; the average single-cycle calculation delay was 0.32 s, meeting the requirements for real-time monitoring. The overall MAE of the model-measured difference was 0.14; the error memory module automatically avoided over-adjustment (oscillation amplitude <3%), meeting the "avoid system oscillation" requirement of the application documents.

[0056] This embodiment maps bedside ultrasound and multimodal physiological signals in real time to a digital twin framework of "physiological excitation field - variable topology strain network - closed-loop error feedback": 1. Highly individualized: Node-level geometry, materials, and neural parameters are all derived from the patient's own body; 2. Real-time closed loop: Error field-excitation field mapping enables the system to complete self-adaptation within <10s; 3. Feedforward control: Introducing error memory and trend prediction, proactively correcting 3-5 seconds in advance; 4. Interpretable output: Spatiotemporal density-topological convolution dual-view visually locates lesions; 5. Clinical benefits: Significantly shortens gastric emptying time, guiding precise administration of prokinetic drugs.

[0057] In summary, this embodiment fully verifies the feasibility, real-time performance, and clinical value of the invention in complex critical care scenarios, laying a technical foundation for its subsequent promotion to multi-center ICU monitoring and intelligent drug infusion.

Claims

1. A spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins, characterized in that... Includes the following steps: The neuromuscular activation process of the gastrointestinal tract is abstracted as a physiological excitation field, which determines the key parameters of muscle peristalsis initiation, wave velocity, and rhythm. A variable topological strain network for the gastrointestinal wall is constructed based on a physiological excitation field. The structure of the variable topological strain network adapts to neural activation, forming a functional contraction path. A set of control equations is introduced to make the excitation field exhibit wave evolution in time, tension distribution in space, and geometric deformation in structure. The behavioral differences between twins and the real gastrointestinal system are analyzed using the error field; the error feedback is transmitted to the excitation field parameters to achieve closed-loop regulation; By integrating historical state data with current feedback, a time-series graphical network is used to model the creeping trend. The evolution of behavioral abilities from fitting reality to predicting the future; The continuous peristaltic process is mapped as a spatiotemporal activity density map; topological convolution is used to extract activity mutations, peristaltic stagnation, and backpropagation features in key regions; the output includes regions with low transmission efficiency, stress concentration areas, and suspected spasm sources, which are reconstructed into semantic indicators that are easy for doctors to understand.

2. The spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins according to claim 1, characterized in that... The method for constructing a variable topological strain network for the gastrointestinal wall using the physiological excitation field includes: The gastrointestinal wall region is discretized into multiple muscle nodes, and an initial network topology consisting of multiple nodes and connecting edges is constructed. The nodes represent muscle units, and the edges represent muscle synergy or stress propagation paths. Gastrointestinal nerve signals are collected, and a physiological excitation field reflecting the spatial distribution of nerve activity is constructed. The excitation field is used to control the activation state of network nodes. Based on the spatial intensity and propagation direction of the physiological excitation field, the connection relationship between nodes in the network is dynamically adjusted to form a variable topology, wherein the connection of nodes in the activated region is enhanced, and the connection in the non-activated region is weakened or broken; based on the change of the network topology, the strain state and muscle tension of each node are dynamically calculated to form a muscle contraction path consistent with the actual neural activation behavior. By combining contraction pathways with time-series driving, the propagation and evolution of gastrointestinal peristalsis behavior in the spatiotemporal domain are simulated.

3. The spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins according to claim 2, characterized in that... The physiological excitation field is constructed based on the neuronal group signals of the gastrointestinal autonomic nervous system. The signal sources include electromyography, bioelectric field distribution maps, or neural stimulation data. The connection relationship change mechanism of the network structure is based on dynamic connection weight adjustment rules. These rules combine the current excitation intensity with historical activation trends to determine the increase, decrease, or breakage of connection edges.

4. The spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins according to claim 3, characterized in that... The contraction path is a stress propagation path naturally formed by the excitation field during its propagation within the network; the variable topology strain network and the actual gastrointestinal physical sensing data, including local tension, content flow velocity, and pressure feedback, form a virtual-real synchronous feedback mechanism to adjust the excitation field parameters and enable adaptive correction of the twin model.

5. The spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins according to claim 1, characterized in that... The methods by which the governing equations enable the excitation field to exert temporal, spatial, and structural effects include: A digital twin structure of the gastrointestinal wall region is constructed, including multiple muscle unit nodes and their connections; a physiological excitation field is initialized, which reflects the activation state of neural signals in the gastrointestinal wall space; The excitation field is driven by the control equations to evolve continuously in the time dimension, forming periodic, delayed or staggered excitation patterns; the excitation field diffuses in the spatial dimension and forms a tension gradient, driving muscle units in different regions to produce differentiated stress states. Based on the distribution of excitation tension, the muscle unit is induced to have a geometric response in its structure, including shortening of length, increase of thickness or distortion of shape, which causes geometric deformation on the peristaltic path; excitation propagation, tension formation and structural deformation behavior are used as the basis for peristalsis driving, and the digital twin is used to simulate the gastrointestinal peristalsis process in the spatiotemporal range.

6. The spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins according to claim 5, characterized in that... The set of control equations includes multiple control mechanisms to describe excitation initiation, propagation speed, energy decay, signal interference, and reflex behavior, in order to simulate the natural propagation process of actual neural signals in the time dimension; the propagation of the excitation field in the spatial dimension can form tension regions of different intensities, constituting muscle synergistic contraction regions, tension release buffer regions, and content propulsion channels, enabling regional functional division of labor.

7. The spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins according to claim 6, characterized in that... The response of the structural deformation is dynamically controlled by the spatial intensity and propagation direction of the excitation field, forming a traceable and predictable peristaltic path and dynamic morphological evolution of the gastrointestinal wall; the propagation behavior of the excitation field and the structural response are coupled, and the integrated dynamic linkage of control source, tissue function and structural geometry is realized through the control equation set.

8. The spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins according to claim 1, characterized in that... The process of constructing the closed-loop regulation includes: Data on the peristaltic behavior of a gastrointestinal twin during a simulated period are obtained, including muscle tension distribution, peristaltic path, and contents flow direction; actual physiological data within the corresponding time and space range of the real gastrointestinal system are collected simultaneously; the twin data and the real data are compared in spatial and temporal dimensions to generate an error field map, wherein the error field is a continuous multidimensional distributed data structure used to describe the behavioral deviation between the simulated state and the real state; Based on the high deviation region in the error field, identify abnormal features such as activation path mismatch, fluctuation delay, or creep decay, and establish a mapping relationship between error and excitation source; apply error feedback to the physiological excitation field of the twin model to adjust the excitation parameters such as neural activation intensity, diffusion rate, activation frequency, or spatial distribution range.

9. The spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins according to claim 8, characterized in that... The error field is a continuous difference map constructed in the space-time dimension, which identifies regional deviation trends, error aggregation points and their propagation paths; the adjustment object of the error feedback is the excitation field source parameters of the twin, so as to correct behavior at the level of neural activation mechanism.

10. The spatiotemporal simulation method for gastrointestinal peristalsis based on digital twins according to claim 9, characterized in that... During the feedback adjustment process of the excitation field, predictive adjustment is performed based on the evolution trend of the historical error field to correct parameters in potential deviation areas in advance and achieve feedforward control; the error evolution trajectory is recorded in multiple feedback cycles to form an error memory mechanism, which is used to improve the adaptability and stability of the twin model to individual physiological characteristics.

Images