Data Fusion Processing Method and System for MEMS-Based Intelligent Sensors

Abstract

The present invention relates to a data fusion processing method and system for an MEMS-based intelligent sensor, including the following steps: decomposing the multi-modal physiological parameter time series data stream to obtain a key vital sign feature sequence; calculating the phase coupling degree of the key vital sign feature sequence to obtain an organ-interfunction association strength matrix; performing high-dimensional feature space projection on the critically ill patient based on the organ-interfunction association strength matrix to obtain a patient pathological state feature map; performing non-linear dynamics analysis on the patient pathological state feature map to obtain a disease progression dynamic trajectory; and performing clinical decision support reasoning based on the disease progression dynamic trajectory to obtain personalized treatment recommendations, solving the technical problem that existing analysis methods are often limited to the observation of a single physiological parameter or simple statistical analysis and lack the ability to understand the disease mechanism from an overall perspective.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent sensors, and particularly relates to a data fusion processing method and system for an MEMS-based intelligent sensor. Background Art

[0002] In a modern medical environment, for critically ill patients, real-time monitoring and accurate assessment of their physiological states are of crucial importance. However, traditional monitoring methods based on a single sensor or a few types of sensors are difficult to comprehensively and accurately capture the complex physiological information of patients. With the development of microelectromechanical system (MEMS) technology, the application of intelligent sensors provides a new idea for solving this problem. By deploying multiple MEMS sensors in the intensive care unit, continuous monitoring of multiple vital sign parameters of patients can be achieved. However, how to effectively fuse this multi-modal data and extract valuable medical information from it has become an urgent problem to be solved.

[0003] In addition, in clinical practice, doctors are faced with a highly complex dynamic system - the human body. The interactions between different organs and their functional correlations not only affect the progression of diseases but also determine the effectiveness of treatment strategies. Existing analysis methods are often limited to the observation of a single physiological parameter or simple statistical analysis, lacking the ability to understand the disease mechanism from an overall perspective. Therefore, developing a method that can reflect the strength of functional associations between organs is of great significance for in-depth understanding of the pathological process and formulating personalized treatment plans. This need has promoted the research and development of data fusion processing methods for MEMS-based intelligent sensors.

[0004] Finally, although there have been many studies on data fusion and signal processing technologies, there are still many challenges in applying them to clinical decision support. For example, how to ensure the authenticity and reliability of the extracted vital sign feature sequences, and how to identify patterns closely related to disease progression from a high-dimensional feature space. The existence of these problems not only limits the wide application of existing technologies in clinical practice but also highlights the importance of further exploring data fusion processing methods for MEMS-based intelligent sensors. By overcoming these challenges, the quality and efficiency of critically ill patient management can be significantly improved, thereby improving the treatment outcomes of patients. Summary of the Invention

[0005] The main object of the present invention is to provide a data fusion processing method and system for an MEMS-based intelligent sensor, which solves the technical problem that existing analysis methods are often limited to the observation of a single physiological parameter or simple statistical analysis and lack the ability to understand the disease mechanism from an overall perspective.

[0006] To achieve the above object, the present invention provides a data fusion processing method for an MEMS-based intelligent sensor, comprising the following steps:

[0007] Perform multi-channel synchronous sampling and fusion processing on the physiological signals of critically ill patients collected by the MEMS sensor to obtain a multi-modal physiological parameter time-series data stream;

[0008] Decompose the multi-modal physiological parameter time-series data stream into non-stationary signals to obtain a key vital sign feature sequence;

[0009] Calculate the phase coupling degree of the key vital sign feature sequence to obtain an inter-organ functional association strength matrix;

[0010] Perform high-dimensional feature space projection on the critically ill patient based on the inter-organ functional association strength matrix to obtain a patient pathological state feature map;

[0011] Perform non-linear dynamics analysis on the patient pathological state feature map to obtain a disease progression dynamic trajectory;

[0012] Perform clinical decision support reasoning based on the disease progression dynamic trajectory to obtain personalized treatment recommendations.

[0013] Further, the decomposing the multi-modal physiological parameter time-series data stream into non-stationary signals to obtain a key vital sign feature sequence includes:

[0014] Perform adaptive bi-orthogonal filtering separation on the multi-modal physiological parameter time-series data stream to obtain a non-linear physiological fluctuation component, and perform multi-level harmonic analysis on the non-linear physiological fluctuation component through an adaptive oscillation recognition mechanism to obtain a physiological system phase transition feature set;

[0015] Perform statistical significance analysis on the physiological system phase transition feature set based on a multi-scale covariance structure method to obtain a key physiological coupling network, and perform topological stability evaluation on the key physiological coupling network to obtain multi-organ function linkage data;

[0016] Perform phase synchronization metric calculation and extraction on the multi-organ function linkage data to obtain a key vital sign feature sequence; wherein, the key vital sign feature sequence includes cardiovascular system regulation ability, respiratory compensation reserve index, circulatory system steady-state maintenance parameter, and multi-system coordination function feature.

[0017] Further, the calculating the phase coupling degree based on the key vital sign feature sequence to obtain an inter-organ functional association strength matrix includes:

[0018] Perform bispectral coherence analysis on the key vital sign feature sequence to obtain a non-linear phase synchronization feature spectrum, and perform time-varying causality analysis on the non-linear phase synchronization feature spectrum based on wavelet bidirectional transformation to obtain a functional network connection strength map;

[0019] Perform directed information transfer analysis on the functional network connection strength map through transfer entropy calculation method to obtain a multi-organ interaction matrix, and perform resonance frequency band screening on the multi-organ interaction matrix to obtain a key physiological system collaborative working mode;

[0020] Calculate the phase locking value for the key physiological system collaborative working mode to obtain organ function synchronization degree data, and perform graph theory network characteristic analysis based on the organ function synchronization degree data to obtain an organ-to-organ functional association strength matrix; wherein, the organ-to-organ functional association strength matrix includes a cardio-pulmonary function collaboration efficiency coefficient, a cerebrovascular autoregulation ability parameter, a renal blood perfusion state, and a multi-system metabolic balance network characteristic.

[0021] Furthermore, perform high-dimensional feature space projection based on the organ-to-organ functional association strength matrix to obtain a patient pathological state feature map, including:

[0022] Perform tensor decomposition operation on the organ-to-organ functional association strength matrix to obtain a multi-dimensional physiological system interaction feature space, perform topological structure extraction on the multi-dimensional physiological system interaction feature space to obtain a disease state topological skeleton map, and perform multi-scale persistence calculation on the disease state topological skeleton map to obtain a set of key pathological state branch points;

[0023] Perform non-linear boundary mapping on the set of key pathological state branch points through kernel surface reconstruction technology to obtain an organ function degradation partition map, and perform spectral clustering segmentation processing on the organ function degradation partition map to obtain a pathological state subspace structure;

[0024] Extract geometric depth features from the pathological state subspace structure to obtain a disease progression path network, and perform Ricci curvature calculation based on the disease progression path network to obtain a pathological state transformation risk surface;

[0025] Perform curvature flow evolution analysis on the pathological state transformation risk surface based on differential geometric mapping to obtain a disease state manifold map, and construct a feature vector field based on the disease state manifold map to obtain a patient pathological state feature map, wherein the patient pathological state feature map includes a multi-organ function collaborative disorder mode, a critical disease development stage marker, a vital sign instability risk index, and an organ function recovery potential assessment feature.

[0026] Further, performing non-linear dynamics analysis on the pathological state feature map of the patient to obtain the disease progression dynamic trajectory, including:

[0027] Calculating the multi-scale complexity of the pathological state feature map of the patient to obtain the disease evolution complexity feature sequence, and performing recursive quantitative analysis on the disease evolution complexity feature sequence to obtain the state transition probability matrix;

[0028] Extracting the dominant dynamic mode of the state transition probability matrix through singular value decomposition to obtain the disease state attractor structure, and performing bifurcation theory analysis on the disease state attractor structure to obtain the set of clinical state transition critical points;

[0029] Performing reachability analysis on the set of clinical state transition critical points to obtain the disease state transition network diagram, and calculating the minimum action path based on the disease state transition network diagram to obtain the key regulatory nodes of disease progression;

[0030] Performing dynamic system evolution simulation on the key regulatory nodes of disease progression based on the stochastic differential equation to obtain the time-varying disease state manifold, and performing vector field integral calculation based on the time-varying disease state manifold to obtain the disease progression dynamic trajectory.

[0031] Further, performing vector field integral calculation based on the time-varying disease state manifold to obtain the disease progression dynamic trajectory, including:

[0032] Performing Stokes' theorem transformation on the time-varying disease state manifold to obtain the disease state boundary flux distribution diagram, and performing tangential field decomposition on the disease state boundary flux distribution diagram to obtain the pathological state migration velocity vector field;

[0033] Performing geometric property analysis on the pathological state migration velocity vector field through the Riemann curvature tensor to obtain the set of geodesics of the disease progression path, and performing manifold parallel transport calculation based on the set of geodesics of the disease progression path to obtain the minimum action path of clinical state transition;

[0034] Solving the Hamilton-Jacobi equation for the minimum action path of clinical state transition to obtain the disease progression time-optimal control surface, and constructing a Poincaré section based on the disease progression time-optimal control surface to obtain the disease progression dynamic trajectory.

[0035] Further, performing clinical decision support reasoning based on the disease progression dynamic trajectory to obtain personalized treatment recommendations, including:

[0036] Perform trajectory segmentation on the dynamic trajectory of the disease progression to obtain a sequence of key pathological stage conversion points, and perform multi-dimensional critical state analysis on the sequence of key pathological stage conversion points to obtain a set of treatment intervention time windows;

[0037] Based on a Bayesian causal network, infer the inter-organ action mechanism for the set of treatment intervention time windows to obtain a multi-level pathological causal chain structure, and perform key node sensitivity analysis on the multi-level pathological causal chain structure to obtain a treatment target priority matrix;

[0038] Through multi-objective constraint optimization, perform treatment resource allocation calculation on the treatment target priority matrix to obtain a multi-dimensional pharmacokinetic parameter space, and construct a non-linear response surface for the multi-dimensional pharmacokinetic parameter space to obtain a treatment plan - physiological response mapping diagram;

[0039] Perform Pareto optimal solution analysis on the treatment plan - physiological response mapping diagram to obtain a set of multi-objective balanced treatment strategies, and based on the set of multi-objective balanced treatment strategies, perform clinical risk - benefit quantitative assessment to obtain the safety margin of the individualized intervention plan;

[0040] Based on a temporal Bayesian decision network, perform dynamic treatment path planning on the safety margin of the individualized intervention plan to obtain a sequence of phased treatment goals, and based on the sequence of phased treatment goals, perform treatment plan combination optimization calculation to obtain personalized treatment recommendations.

[0041] The present invention also provides a data fusion processing system based on a MEMS intelligent sensor, including:

[0042] A sampling module, configured to perform multi-channel synchronous sampling and fusion processing on the physiological signals of critically ill patients collected by the MEMS sensor to obtain a multi-modal physiological parameter time series data stream;

[0043] A decomposition module, configured to perform non-stationary signal decomposition on the multi-modal physiological parameter time series data stream to obtain a sequence of key vital sign features;

[0044] A calculation module, configured to calculate the phase coupling degree for the sequence of key vital sign features to obtain an inter-organ functional association strength matrix;

[0045] A projection module, configured to project the critically ill patient into a high-dimensional feature space based on the inter-organ functional association strength matrix to obtain a patient pathological state feature map;

[0046] An analysis module, configured to perform non-linear dynamics analysis on the patient pathological state feature map to obtain a dynamic trajectory of disease progression;

[0047] An inference module, configured to perform clinical decision support inference based on the dynamic trajectory of disease progression to obtain personalized treatment suggestions.

[0048] The present invention also provides a computer device, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps of the method described in any one of the above are implemented.

[0049] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described in any one of the above are implemented.

[0050] A data fusion processing method for a MEMS-based intelligent sensor provided by the present invention includes the following steps: performing multi-channel synchronous sampling and fusion processing on physiological signals of critically ill patients collected by MEMS sensors to obtain a multi-modal physiological parameter time series data stream; decomposing the multi-modal physiological parameter time series data stream into non-stationary signals to obtain a key vital sign feature sequence; calculating the phase coupling degree of the key vital sign feature sequence to obtain an organ-interfunction association strength matrix; performing high-dimensional feature space projection on the critically ill patient based on the organ-interfunction association strength matrix to obtain a patient pathological state feature map; performing non-linear dynamics analysis on the patient pathological state feature map to obtain a disease progression dynamic trajectory; performing clinical decision support inference based on the disease progression dynamic trajectory to obtain personalized treatment suggestions, solving the technical problem that existing analysis methods are often limited to the observation of single physiological parameters or simple statistical analysis and lack the ability to understand the disease mechanism from an overall perspective, and realizing non-linear dynamics analysis of the patient pathological state feature map, which can track the dynamic trajectory of disease progression. This method not only helps to predict the development trend of the condition, but also can timely adjust the treatment strategy to cope with possible changes, thereby improving the treatment effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a schematic diagram of the steps of a data fusion processing method for a MEMS-based intelligent sensor according to an embodiment of the present invention;

[0052] Figure 2 is a structural block diagram of a data fusion processing system for a MEMS-based intelligent sensor according to an embodiment of the present invention;

[0053] Figure 3 is a schematic structural block diagram of a computer device according to an embodiment of the present invention.

[0054] The implementation, functional features and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0056] As Figure 1 shown, Figure 1 is a schematic diagram of the steps of a data fusion processing method for a MEMS-based intelligent sensor in an embodiment of the present invention;

[0057] An embodiment of the present invention provides a data fusion processing method for a MEMS-based intelligent sensor, including the following steps:

[0058] Step S1, perform multi-channel synchronous sampling and fusion processing on the physiological signals of critically ill patients collected by MEMS sensors to obtain a multi-modal physiological parameter time-series data stream.

[0059] Specifically, in the data fusion processing method for a MEMS-based intelligent sensor, performing multi-channel synchronous sampling and fusion processing on the physiological signals of critically ill patients is a crucial first step. This process first relies on multiple MEMS sensors working simultaneously, which can accurately capture different types of physiological signals, such as electrocardiogram, respiratory rate, and blood oxygen saturation. Through multi-channel synchronous sampling, it is ensured that all the collected data not only has temporal synchronization but also can reflect the multiple physiological states of the patient at the same moment, thus providing a solid foundation for subsequent data analysis. Then, the original data streams from different sensors are subjected to fusion processing, and advanced algorithms are used to integrate them into a multi-modal physiological parameter time-series data stream. This step not only needs to consider the problem of data time synchronization but also needs to solve the compatibility and consistency between different types of data. For example, in an intensive care environment, a critically ill patient with severe heart disease may need continuous monitoring of their heart activity, respiratory condition, and oxygen content in the blood. Suppose three different MEMS sensors are used to measure the above three physiological parameters respectively. Then, in actual operation, these three sensors will collect data synchronously at a preset time interval (such as once per second). Subsequently, the system will use specific data fusion algorithms, such as Kalman filtering or wavelet transform, to process these original data. In this way, not only can the noise interference caused by sensor individual differences or environmental factors be eliminated, but also the originally isolated signals can be transformed into a unified time-series data stream containing multiple pieces of information. In this way, the medical team can comprehensively understand the patient's real-time health status on one platform and make more accurate diagnosis and treatment decisions based on this, significantly improving the management efficiency and quality of critically ill patients. This process fully reflects the key role of the multi-channel synchronous sampling and fusion processing step in improving the performance of the medical monitoring system.

[0060] Step S2: Perform non-stationary signal decomposition on the multi-modal physiological parameter time-series data stream to obtain a key vital sign feature sequence.

[0061] Specifically, in the data fusion processing method based on MEMS intelligent sensors, non-stationary signal decomposition of the multi-modal physiological parameter time-series data stream is one of the key steps, aiming to extract clinically significant key vital sign feature sequences from complex physiological data. This process first requires identifying and separating those signal components that vary over time rather than being constant, because these components often contain important information about the patient's health status. Specifically, by applying advanced signal processing techniques such as empirical mode decomposition (EMD) or wavelet transform, etc., the fluctuating components in different frequency ranges can be extracted from the original data stream, thereby revealing the underlying physiological dynamic changes. These decomposed signal components can better reflect the changing trends of the patient's vital signs and provide a more accurate basis for subsequent analysis. For example, in an intensive care environment, assume we are monitoring a patient with severe respiratory failure. The multi-modal physiological parameter time-series data stream of this patient includes various signals such as electrocardiogram, respiratory rate, and blood oxygen saturation. Due to the impact of the disease, these signals exhibit significant non-stationary characteristics, that is, they are not constant but change over time. To extract useful information from these complex and variable signals, we can use wavelet transform technology to decompose the original data stream into sub-signals at multiple scales. Each sub-signal represents the fluctuating components in different frequency ranges, and some of them may be closely related to the patient's breathing pattern or cardiac activity. In this way, we can obtain a series of key vital sign feature sequences, such as the breathing fluctuation characteristics in a specific frequency band or the changing trend of heart rate variability. These feature sequences not only help to more accurately evaluate the patient's current health status but also provide important clues for predicting the development of the disease. Therefore, by performing non-stationary signal decomposition on the multi-modal physiological parameter time-series data stream, the medical team can obtain more refined and in-depth patient physiological information, and then formulate more personalized and effective treatment plans. This process fully demonstrates the important role of non-stationary signal decomposition in enhancing the management of critically ill patients.

[0062] Step S3: Calculate the phase coupling degree of the key vital sign feature sequence to obtain an organ-interfunction association strength matrix.

[0063] Specifically, in the data fusion processing method of MEMS-based intelligent sensors, calculating the phase coupling degree of the key vital sign feature sequences to obtain the inter-organ functional association strength matrix is a crucial step. This process first starts with the key vital sign feature sequences extracted from the previous steps, which contain the time-dynamic change information of different physiological parameters. By applying the phase coupling degree analysis technique, the synchronization and interaction degree between different physiological signals can be quantified, thereby revealing the functional correlation between organs. Specifically, the phase coupling degree calculation usually involves transforming each feature sequence into the frequency domain or time-frequency domain, and then evaluating the phase relationship between them through methods such as cross-spectrum density or phase difference distribution. This analysis can help identify which physiological signals have significant phase synchronization phenomena, and then infer the functional association strength between the corresponding organs. For example, in an intensive care environment, assume we are monitoring a patient with severe heart disease and accompanying respiratory failure. After the non-stationary signal decomposition of the multi-modal physiological parameter time-series data stream of this patient, key vital sign feature sequences including heart rate variability, respiratory rate fluctuations, and blood oxygen saturation changes are obtained. To further understand the interaction between these physiological signals, we can adopt the phase coupling degree calculation method. First, transform these feature sequences into the frequency domain and calculate the cross-spectrum density and phase difference distribution between them. Then, by analyzing these results, we can find that there is a significant phase synchronization phenomenon between heart rate variability and respiratory rate fluctuations, indicating a close functional association between the heart and lungs. In addition, the phase relationship between blood oxygen saturation changes and other physiological parameters can also be observed, thereby inferring the coordinated working conditions of multiple organ systems throughout the body. Finally, by integrating all these phase coupling degree results, we can construct an inter-organ functional association strength matrix, which not only reflects the interaction strength between organ systems but also provides in-depth insights into the overall health status of the patient for clinicians. Based on this matrix, the medical team can more accurately evaluate the condition and develop personalized treatment plans, improving the management effect and treatment success rate of critically ill patients. This process demonstrates the powerful ability of phase coupling degree calculation in revealing the functional correlation between organs.

[0064] Step S4, project the critically ill patient onto a high-dimensional feature space based on the inter-organ functional association strength matrix to obtain a patient pathological state feature map.

[0065] Specifically, in the data fusion processing method of MEMS-based intelligent sensors, projecting a critically ill patient onto a high-dimensional feature space based on the inter-organ functional association strength matrix to obtain the patient's pathological state feature map is a crucial step. This process first depends on the inter-organ functional association strength matrix obtained in the previous step, which contains information on the interaction strengths between different physiological signals. By mapping these complex interaction relationships into a high-dimensional feature space, the patient's pathological state and its dynamic changes can be more intuitively displayed. Specifically, this process usually involves applying dimensionality reduction techniques such as principal component analysis (PCA) or t-distributed stochastic neighbor embedding (t-SNE) to convert the original high-dimensional data into a low-dimensional visualization chart, enabling doctors to more easily identify patterns and features related to disease progression. For example, in an intensive care environment, assume we are monitoring a patient with severe heart disease accompanied by respiratory failure. After calculating the phase coupling degree, we obtain a detailed inter-organ functional association strength matrix, which contains the interaction strengths between key vital sign features such as heart rate variability, respiratory rate fluctuations, and blood oxygen saturation changes. To further understand these complex relationships and convert them into actionable information, we map this matrix into a high-dimensional feature space. Using principal component analysis (PCA), we can compress the multi-dimensional data into a two-dimensional or three-dimensional space, thereby generating the patient's pathological state feature map. In this process, each point represents the overall health state of the patient at a certain moment, and the distance between points reflects the degree of change in the health state between different time points. For example, if the pathological state map of the patient shows an obvious clustering trend during a certain period, this may indicate that the condition is relatively stable; on the contrary, if the points in the map are scattered, it may indicate the deterioration of the condition or the emergence of new complications. In addition, by observing the distribution of different regions in the map, doctors can also discover potential pathological patterns, such as signs of dysfunction of certain specific organs, and then formulate more precise treatment plans. This method of high-dimensional feature space projection not only improves the understanding of the patient's pathological state but also provides strong support for clinical decision-making, enhancing the management efficiency and treatment effect of critically ill patients. The entire process demonstrates how to extract valuable information from complex physiological data and convert it into an intuitive visualization chart to better guide medical practice.

[0066] Step S5: Perform non-linear dynamics analysis on the patient's pathological state feature map to obtain the dynamic trajectory of disease progression.

[0067] Specifically, in the data fusion processing method of MEMS-based intelligent sensors, performing nonlinear dynamics analysis on the patient's pathological state feature map to obtain the disease progression dynamic trajectory is a crucial step. This process first depends on the patient's pathological state feature map generated in the previous step, which converts complex physiological data into an intuitive visualization chart through high-dimensional feature space projection technology. To further understand the patient's pathological state and its changing trend over time, it is necessary to apply nonlinear dynamics analysis methods to reveal the internal relationships and dynamic change patterns among these features. Specifically, this method usually includes calculating nonlinear dynamics indicators such as Lyapunov exponents and fractal dimensions to capture the complexity and instability in the system. For example, in an intensive care environment, assume we are monitoring a patient with severe heart disease and accompanying respiratory failure. After high-dimensional feature space projection, we obtain a detailed patient pathological state feature map, where each point represents the overall health state of the patient at a certain moment. To extract the dynamic information related to disease progression from these data, we can adopt nonlinear dynamics analysis methods. First, calculate the Lyapunov exponent to evaluate the chaos degree of the system, which can help us understand the complexity and unpredictability of the condition. Then, using fractal dimension analysis can reveal the self-similarity and complexity of different regions in the pathological state map, thereby identifying potential pathological patterns. In addition, phase space reconstruction technology can also be applied to map time series data into the phase space and observe the changes in its trajectory. Through these analysis methods, we can construct a disease progression dynamic trajectory, which not only reflects the changing trend of the patient's condition over time but also helps doctors predict the future development of the condition. For example, if the trajectory shows obvious fluctuations or accelerating changes, it may indicate the deterioration of the condition or the emergence of new complications; on the contrary, if the trajectory tends to be stable, it indicates that the condition is relatively stable. This nonlinear dynamics analysis method not only improves the understanding of disease progression but also provides strong support for clinical decision-making, enabling the medical team to adjust the treatment plan more timely and accurately, and enhancing the management efficiency and treatment effect of critically ill patients. The whole process demonstrates how to extract valuable dynamic information from complex pathological state feature maps and convert it into key clues for guiding clinical practice.

[0068] Step S6, perform clinical decision support reasoning based on the disease progression dynamic trajectory to obtain personalized treatment recommendations.

[0069] Specifically, in the data fusion processing method of MEMS-based intelligent sensors, clinical decision support reasoning based on the dynamic trajectory of disease progression to obtain personalized treatment recommendations is the final and crucial step. This process first relies on the dynamic trajectory of disease progression generated in the previous steps, which reveals the changing trends and complex patterns of the patient's condition over time through nonlinear dynamics analysis. To convert this complex dynamic information into actionable treatment recommendations, an advanced clinical decision support system (CDSS) is needed, which combines medical knowledge bases and machine learning algorithms to comprehensively evaluate the patient's pathological state and its changing trends. Specifically, this method typically includes identifying key time points and turning points, predicting potential risk factors, and simulating the effects of different treatment options. For example, in an intensive care environment, assume we are monitoring a patient with severe heart disease and accompanying respiratory failure. After nonlinear dynamics analysis, we obtain a detailed dynamic trajectory of disease progression, which not only shows the fluctuations of the patient's condition but also reveals significant changes at certain specific time points. To extract personalized treatment recommendations from this dynamic information, we can use a clinical decision support system (CDSS). First, the system will identify early warning signals of deteriorating conditions based on the key turning points in the trajectory and propose preliminary intervention measures in combination with relevant guidelines and best practices in the medical knowledge base. Then, machine learning algorithms are used to train historical case data to simulate the effects of different treatment options, so as to recommend the most suitable treatment method for the current patient. For example, if the trajectory shows that the patient's heart rate variability significantly decreases over a certain period of time while the respiratory rate fluctuations increase, this may indicate further deterioration of cardiac and pulmonary functions. In this case, the CDSS may recommend adjusting the drug dosage or introducing new treatment means, such as mechanical ventilation support. In addition, the system can continuously update its recommendations based on the patient's real-time physiological data to ensure that the treatment plan always keeps in line with the patient's latest condition. This clinical decision support reasoning based on the dynamic trajectory of disease progression not only improves the accuracy and effectiveness of treatment but also provides a powerful tool for the medical team to help them formulate more personalized and timely treatment strategies, thus significantly improving the treatment outcomes and quality of life of critically ill patients. The whole process demonstrates how to extract valuable clinical information from complex dynamic trajectories and convert it into specific treatment recommendations to better guide medical practice.

[0070] In a specific embodiment, the non-stationary signal decomposition of the multi-modal physiological parameter time series data stream to obtain the key vital sign feature sequence includes:

[0071] Adaptive bi-orthogonal filtering separation is performed on the multi-modal physiological parameter time series data stream to obtain non-linear physiological fluctuation components, and multi-level harmonic analysis is performed on the non-linear physiological fluctuation components through an adaptive oscillation recognition mechanism to obtain a physiological system phase transition feature set;

[0072] Based on the multi-scale covariance structure method, statistical significance analysis is performed on the physiological system phase transition feature set to obtain a key physiological coupling network, and topological stability evaluation is performed on the key physiological coupling network to obtain multi-organ function linkage data;

[0073] Phase synchronization measurement and extraction are performed on the multi-organ function linkage data to obtain a key vital sign feature sequence; wherein, the key vital sign feature sequence includes cardiovascular system regulation ability, respiratory compensation reserve index, circulatory system steady-state maintenance parameters, and multi-system cooperation function features.

[0074] Specifically, in the data fusion processing method of MEMS-based intelligent sensors, the process of decomposing non-stationary signals from multi-modal physiological parameter time-series data streams to obtain key vital sign feature sequences is extremely complex and delicate. This process not only requires the application of advanced signal processing techniques but also the combination of means such as multi-scale data analysis and statistical significance analysis to comprehensively reveal the patient's pathological state and its dynamic changes. First, perform adaptive bi-orthogonal filtering separation on the multi-modal physiological parameter time-series data stream, which is the first step of the whole process. Through this method, non-linear physiological fluctuation components can be separated from the original data stream. These components contain the complex fluctuation patterns exhibited by the patient in different physiological states. For example, in an intensive care environment, a patient with severe heart disease and respiratory failure will show complex fluctuations in various physiological parameters such as electrocardiogram, respiratory rate, and blood oxygen saturation. Through adaptive bi-orthogonal filtering separation, these fluctuation components can be extracted from the mixed signal and further subjected to joint time-frequency analysis to generate a dynamic spectrum feature map. This map includes not only the cardiac electrophysiological activity pattern but also important information such as vascular resistance change characteristics, pulmonary ventilation function parameters, and body temperature regulation loop signals. This information provides rich basic data for subsequent analysis. Next, perform multi-level harmonic analysis on the dynamic spectrum feature map through an adaptive oscillation recognition mechanism, which is the key link of the second step. This mechanism can identify the multi-level harmonic components in the dynamic spectrum feature map and extract the parameter resonance boundary sequence from them. These boundary sequences reflect the resonance characteristics between different physiological signals and lay the foundation for further understanding the interaction between organs. For example, in the above patient example, through multi-level harmonic analysis, a specific resonance relationship between cardiac electrophysiological activity and respiratory rate can be found, which helps to reveal the changing trend of cardio-pulmonary coupling strength. Subsequently, perform non-stationary endpoint detection on the parameter resonance boundary sequence to obtain the physiological system phase transition feature set. This detection method can capture the transition points of the physiological system in different states, thereby helping to identify potential pathological changes. Perform statistical significance analysis on the physiological system phase transition feature set based on multi-scale covariance structure, which is the core content of the third step. Through this method, the correlation and synchronization between different physiological signals can be evaluated, and then a key physiological coupling network can be constructed. For example, in the case of the aforementioned patient, through multi-scale covariance analysis of features such as heart rate variability, respiratory rate fluctuation, and blood oxygen saturation change, the complex relationships between the cardio-pulmonary system, neuro-circulatory system feedback efficiency, and endocrine-metabolic axis regulation parameters can be revealed. Then, evaluate the topological stability of the key physiological coupling network to obtain multi-organ function linkage data. These data include not only the cardio-pulmonary coupling strength but also important contents such as neuro-circulatory system feedback efficiency, endocrine-metabolic axis regulation parameters, and organ-to-organ information transmission characteristics.These coupled data provide strong support for comprehensively understanding the patient's pathological state and its progression. Finally, phase synchronization metric calculations are performed on the multi-organ function coupled data to obtain a functional system integration matrix, and non-linear principal component extraction is performed on the functional system integration matrix to obtain a key vital sign feature sequence. This process quantifies the synchronization and coordination between different physiological systems through phase synchronization metric calculations. For example, in the case of the aforementioned patient, by calculating the phase synchronization degree of the cardiovascular system, respiratory system, and other important physiological systems, an integration matrix can be obtained, which reflects the collaborative work of each system. Then, using the non-linear principal component extraction method, the most representative feature sequences are extracted from the integration matrix, and these feature sequences include cardiovascular system regulation ability, respiratory compensation reserve index, circulatory system steady-state maintenance parameters, and multi-system collaborative function characteristics, etc. These key vital sign feature sequences not only provide doctors with a comprehensive patient health status assessment tool but also provide a scientific basis for formulating personalized treatment plans. To better understand the practical application of this complex process, we can continue to use the previous case of critically ill patients. Suppose this patient is experiencing a deteriorating stage of acute heart failure. Through the detailed analysis of the above steps, the medical team can not only monitor the patient's cardiac electrophysiological activity pattern, respiratory rate fluctuation, and blood oxygen saturation change in real time but also deeply understand the interaction between these signals. For example, through the dynamic spectrum feature mapping diagram, it is found that there is a significant resonance phenomenon between the cardiac electrophysiological activity and the respiratory rate, indicating that the cardio-pulmonary coupling strength may be affected. Then, through multi-level harmonic analysis and non-stationary endpoint detection, it is identified that a significant phase transition has occurred in the patient's physiological system at a certain moment, which may be an early warning signal of the deterioration of the condition. Further, through the analysis of the multi-organ function coupled data, it is found that the feedback efficiency of the neuro-circulatory system decreases and the regulation parameters of the endocrine-metabolic axis are abnormal, and these information provide important references for adjusting the treatment strategy. Finally, through phase synchronization metric calculations and non-linear principal component extraction, a series of key vital sign feature sequences are obtained, which not only reflect the patient's current health status but also provide strong support for predicting the future development of the condition. The whole process demonstrates how to extract valuable information from complex physiological data and transform it into key clues for guiding clinical practice, thereby improving the management efficiency and treatment effect of critically ill patients.

[0075] In a specific embodiment, performing phase coupling degree calculation based on the key vital sign feature sequence to obtain an inter-organ functional association strength matrix includes:

[0076] Performing bispectral coherence analysis on the key vital sign feature sequence to obtain a non-linear phase synchronization feature spectrum, and performing time-varying causality analysis on the non-linear phase synchronization feature spectrum based on wavelet bidirectional transformation to obtain a functional network connection strength map;

[0077] Perform directional information transfer analysis on the functional network connection strength map through the transfer entropy calculation method to obtain a multi-organ interaction matrix, and perform resonance frequency band screening on the multi-organ interaction matrix to obtain the collaborative working mode of key physiological systems;

[0078] Calculate the phase locking value for the collaborative working mode of the key physiological systems to obtain organ function synchronization degree data, and perform graph theory network characteristic analysis based on the organ function synchronization degree data to obtain the functional association strength matrix between organs; wherein, the functional association strength matrix between organs includes the cardiopulmonary function collaboration efficiency coefficient, the cerebrovascular autoregulation ability parameter, the renal blood perfusion state, and the multi-system metabolic balance network characteristics.

[0079] Specifically, the process of calculating the phase coupling degree based on the key vital sign feature sequence to obtain the organ - to - organ functional association strength matrix is an important step in deeply understanding the complex interactions between multi - organ systems. This process, through a series of advanced signal processing and data analysis techniques, can not only reveal the synchronization and coordination of different physiological systems but also provide strong support for personalized medicine. First, perform bispectral coherence analysis on the key vital sign feature sequence, which is the first step of the whole process. Through this method, non - linear phase synchronization feature spectra can be extracted from complex physiological signals. These feature spectra contain the internal connections and interaction patterns between various physiological systems. For example, in a patient with severe cardio - pulmonary disease, the non - linear phase synchronization feature between cardiac electrical activity and respiratory frequency is particularly important. Further, by performing high - order cross - frequency decomposition on the non - linear phase synchronization feature spectra, multi - organ signal coupling metric values can be obtained. These metric values reflect the coupling strength of different organ signals within a specific frequency range and are of great significance for evaluating the functional association between organs. For instance, in the example of the aforementioned patient, through high - order cross - frequency decomposition, it is found that there is significant coupling between cardiac electrical activity and respiratory frequency in certain high - frequency bands, which indicates the close interaction between the cardio - pulmonary systems. Next, perform time - varying causal relationship analysis on the multi - organ signal coupling metric values based on wavelet bidirectional transformation, which is the key link of the second step. This method can capture the dynamic causal relationships between different physiological signals and generate an organ - to - organ information flow vector field. This vector field not only shows how information is transmitted between different organs but also reveals the directionality and intensity of information transmission. For example, in the case of the aforementioned patient, through wavelet bidirectional transformation, it is found that cardiac electrical activity has a significant impact on respiratory frequency, and the change in respiratory frequency in turn feedback - regulates the function of the heart. Then, by calculating the topological complexity of the organ - to - organ information flow vector field, a functional network connection strength map can be obtained. This map includes important information such as the two - way regulation strength of the cardiovascular - respiratory system, the information transmission efficiency of the brain - kidney axis, the liver - pancreas metabolic synergy index, and the coupling markers of the circulatory - endocrine system. These data provide valuable clues for comprehensively understanding the patient's pathological state. Subsequently, perform directed information transfer analysis on the functional network connection strength map through the transfer entropy calculation method, which is the core content of the third step. Transfer entropy is an effective tool for quantifying the direction and intensity of information flow. Through this method, a multi - organ interaction matrix can be obtained. For example, in the example of the aforementioned patient, through transfer entropy calculation, it is found that the information transfer strength from the heart to the kidney is significantly higher than the feedback from the kidney to the heart, which may imply a certain potential pathological mechanism. Then, by screening the resonance frequency bands of the multi - organ interaction matrix, the collaborative working mode of key physiological systems can be obtained.These patterns include important elements such as cardiopulmonary gas exchange synchronization features, brain-heart bidirectional regulation pathway parameters, kidney-vascular pressure feedback loops, and liver-pancreas metabolic regulation networks. These collaborative working patterns not only reveal the interaction laws between different physiological systems but also provide a scientific basis for formulating personalized treatment plans. Finally, the calculation of the phase locking value for the collaborative working patterns of the key physiological systems is the last step in the whole process. Through this method, data on the degree of organ function synchronization can be obtained. For example, in the case of the aforementioned patient, through the calculation of the phase locking value, it was found that the heart and lungs showed a high degree of synchronization during the gas exchange process, while the bidirectional regulation between the heart and the brain showed a certain degree of dysregulation. Then, based on the data on the degree of organ function synchronization, the analysis of the graph theory network characteristics was carried out, and finally, the matrix of the functional correlation strength between organs was obtained. This matrix includes important information such as the cardiopulmonary function cooperation efficiency coefficient, the cerebrovascular autoregulation ability parameter, the renal blood perfusion state, and the characteristics of the multi-system metabolic balance network. These data not only provide doctors with a comprehensive tool for evaluating the patient's health status but also provide strong support for adjusting treatment strategies. To better understand the practical application of this complex process, we can continue to use the previous case of a critically ill patient. Suppose this patient is experiencing a deteriorating stage of acute heart failure. Through the detailed analysis of the above steps, the medical team can not only monitor the patient's cardiac electrophysiological activity patterns, respiratory rate fluctuations, and blood oxygen saturation changes in real time but also deeply understand the interactions between these signals. For example, through bispectral coherence analysis, it was found that there were significant non-linear phase synchronization features between the cardiac electrical activity and the respiratory rate, indicating a close interaction between the cardiopulmonary systems. Further, through high-order cross-frequency decomposition, it was identified that there was a significant coupling strength between the patient's heart and lungs in certain specific frequency ranges, which might be an early warning signal for the deterioration of the condition. Then, based on wavelet bidirectionality transformation, the causal influence of cardiac electrical activity on the respiratory rate was analyzed, and it was found that the abnormal cardiac function did affect the stability of the respiratory system. Through transfer entropy calculation, it was revealed that the information transfer intensity from the heart to other organs (such as the kidneys) increased significantly, indicating that the abnormal cardiac function had spread to other organ systems. Finally, through the calculation of the phase locking value for the collaborative working patterns of the key physiological systems and the analysis of the graph theory network characteristics, a detailed matrix of the functional correlation strength between organs was obtained. This matrix not only reflects the current pathological state of the patient but also provides an important reference basis for predicting the future development of the condition and adjusting treatment strategies. The whole process demonstrates how to extract valuable information from complex physiological data and transform it into key clues for guiding clinical practice, thereby improving the management efficiency and treatment effect of critically ill patients.

[0080] In a specific embodiment, the high-dimensional feature space projection is performed based on the matrix of the functional correlation strength between organs to obtain a characteristic map of the patient's pathological state, including:

[0081] Perform a tensor decomposition operation on the matrix of the functional association strength between organs to obtain a multi-dimensional physiological system interaction feature space, extract the topological structure of the multi-dimensional physiological system interaction feature space to obtain a topological skeleton graph of the disease state, and perform multi-scale persistence calculation on the topological skeleton graph of the disease state to obtain a set of key pathological state branch points;

[0082] Perform a non-linear boundary mapping on the set of key pathological state branch points through kernel surface reconstruction technology to obtain an organ function degradation partition map, and perform spectral clustering segmentation processing on the organ function degradation partition map to obtain a pathological state subspace structure;

[0083] Extract the geometric depth features of the pathological state subspace structure to obtain a disease progression path network, and perform Ricci curvature calculation based on the disease progression path network to obtain a pathological state transformation risk surface;

[0084] Perform curvature flow evolution analysis on the pathological state transformation risk surface based on differential geometric mapping to obtain a disease state manifold graph, and construct a feature vector field based on the disease state manifold graph to obtain a patient pathological state feature map, where the patient pathological state feature map includes multi-organ function coordination disorder patterns, critical disease development stage markers, vital sign instability risk indices, and organ function recovery potential assessment features.

[0085] Specifically, the process of projecting into a high-dimensional feature space based on the matrix of organ-interfunction association strengths to obtain a patient's pathological state feature map is a crucial step in deeply understanding and quantifying the interactions between complex physiological systems. This process, through a series of advanced data processing and analysis techniques, can not only reveal the collaborative disorder patterns of different physiological systems but also provide strong support for personalized medicine. First, a tensor decomposition operation is performed on the matrix of organ-interfunction association strengths, which is the first step of the whole process. Through this method, a multi-dimensional physiological system interaction feature space can be extracted from complex multi-dimensional data. These feature spaces contain the internal connections and interaction patterns between various physiological systems. For example, in a patient with severe cardio-pulmonary diseases, the interaction features between cardiac electrical activity and vital signs such as respiratory rate and blood oxygen saturation are particularly important. Further, based on homotopy persistence analysis, the topological structure of the multi-dimensional physiological system interaction feature space is extracted, and a disease state topological skeleton map can be obtained. This skeleton map not only shows the connection relationships between different physiological systems but also reveals the stability of these relationships. For instance, in the example of the aforementioned patient, it is found through homotopy persistence analysis that the coupling between cardiac electrical activity and respiratory rate has a high persistence, indicating a close interaction between the cardio-pulmonary systems. Then, by performing multi-scale persistence calculations on the disease state topological skeleton map, a set of key pathological state branch points can be obtained. These branch points reflect the important connection points and potential risk areas between physiological systems at different scales. Next, a non-linear boundary mapping is performed on the set of key pathological state branch points through kernel surface reconstruction technology, which is the core part of the second step. This method can capture the non-linear dynamic changes between different physiological systems and generate an organ function degradation partition map. This partition map not only shows the degree and scope of the degradation of different organ functions but also reveals the non-linear characteristics during the degradation process. For example, in the case of the aforementioned patient, through kernel surface reconstruction technology, it is found that there is a significant non-linear boundary between cardiac function degradation and respiratory rate fluctuations, which may imply a certain potential pathological mechanism. Then, by performing spectral clustering segmentation on the organ function degradation partition map, a pathological state subspace structure can be obtained. These subspace structures provide more detailed pathological information, which helps to identify different pathological states and their evolution paths. For example, in the example of the aforementioned patient, through spectral clustering segmentation, a close connection between cardiac function degradation and abnormal renal blood perfusion is found, which provides an important clue for further formulating treatment plans. Next, geometric depth feature extraction is performed on the pathological state subspace structure, which is the key content of the third step. Through this method, a disease progression path network can be extracted from the complex pathological state subspace. These networks not only show the conversion paths between different pathological states but also reveal the stability and risks of these paths.For example, in the case of the aforementioned patient, through geometric depth feature extraction, a relatively stable progression path was found between cardiac function degradation and renal failure, indicating the need for urgent intervention to prevent the deterioration of the condition. Subsequently, Ricci curvature calculation was performed based on the disease progression path network, and a pathological state transformation risk surface could be obtained. This surface not only reflects the difficulty and risk of transformation between different pathological states but also provides an important reference basis for evaluating the treatment effect. Then, curvature flow evolution analysis was carried out on the pathological state transformation risk surface based on differential geometric mapping, which is the core content of the fourth step. This method can capture the dynamic changes between different pathological states and generate a disease state manifold map. Such a manifold map not only shows the changing trend of pathological states over time but also reveals the geometric characteristics of these changes. For example, in the case of the above patient, through curvature flow evolution analysis, a significant dynamic association was found between cardiac function degradation and the decline in cerebrovascular autoregulation ability, indicating the need to comprehensively consider the functional states of multiple systems to adjust the treatment strategy. Then, based on the disease state manifold map, a feature vector field was constructed, and finally, a patient pathological state feature map was obtained. This map not only includes important contents such as multi-organ function dyssynergy patterns, markers of the development stage of severe diseases, risk indices of unstable vital signs, and features for evaluating the potential for organ function recovery but also provides a comprehensive tool for doctors to assess the patient's health status. To better understand the practical application of this complex process, we can continue to use the previous case of a critically ill patient. Suppose this patient is experiencing a deteriorating stage of acute heart failure. Through the detailed analysis of the above steps, the medical team can not only monitor the patient's cardiac electrophysiological activity patterns, respiratory rate fluctuations, and blood oxygen saturation changes in real time but also deeply understand the interactions between these signals. For example, through tensor decomposition operations, a multi-dimensional physiological system interaction feature space was extracted from the matrix of the strength of functional associations between organs, revealing the complex interaction patterns between cardiac electrical activity and vital signs such as respiratory rate and blood oxygen saturation. Further, based on homotopy persistence analysis, a disease state topological skeleton map was obtained, showing the tight coupling relationship between cardiac electrical activity and respiratory rate, and through multi-scale persistence calculation, a set of key pathological state bifurcation points was identified, which reflect the early warning signals of the deterioration of the condition. Then, through kernel surface reconstruction technology, the set of key pathological state bifurcation points was mapped into an organ function degradation partition map, revealing the non-linear boundary between cardiac function degradation and abnormal renal blood perfusion. Then, through spectral clustering segmentation processing of the organ function degradation partition map, a pathological state subspace structure was obtained, further refining the pathological information. For example, through geometric depth feature extraction, a stable progression path was found between cardiac function degradation and renal failure, indicating the need for urgent intervention measures. Subsequently, based on Ricci curvature calculation, a pathological state transformation risk surface was obtained, reflecting the difficulty and risk of transformation between different pathological states.Finally, through differential geometric mapping, curvature flow evolution analysis was carried out, a disease state manifold graph was generated, and based on this, an eigenvector field was constructed to obtain the patient's pathological state feature map. This map not only shows the multi-organ functional dyscoordination pattern of the patient, but also provides important information such as markers for the development stage of severe diseases, risk indices of unstable vital signs, and features for evaluating the potential for organ function recovery, providing a scientific basis for adjusting treatment strategies. The whole process demonstrates how to extract valuable information from complex physiological data and transform it into key clues for guiding clinical practice, thereby improving the management efficiency and treatment effect of critically ill patients.

[0086] In a specific embodiment, the non-linear dynamics analysis of the patient's pathological state feature map to obtain the disease progression dynamic trajectory includes:

[0087] Calculating the multi-scale complexity of the patient's pathological state feature map to obtain a disease evolution complexity feature sequence, and performing recursive quantitative analysis on the disease evolution complexity feature sequence to obtain a state transition probability matrix;

[0088] Extracting the dominant dynamic mode of the state transition probability matrix through singular value decomposition to obtain a disease state attractor structure, and performing bifurcation theory analysis on the disease state attractor structure to obtain a set of clinical state transition critical points, where the set of clinical state transition critical points includes precursors of cardiovascular function collapse, respiratory system compensation limit parameters, multi-organ functional dyscoordination markers, and irreversible points of metabolic disorders;

[0089] Performing reachability analysis on the set of clinical state transition critical points to obtain a disease state transition network graph, and calculating the minimum action path based on the disease state transition network graph to obtain key regulatory nodes for disease progression;

[0090] Based on a stochastic differential equation, dynamic system evolution simulation is carried out on the key regulatory nodes for disease progression to obtain a time-varying disease state manifold, and vector field integral calculation is carried out based on the time-varying disease state manifold to obtain a disease progression dynamic trajectory; wherein the disease progression dynamic trajectory includes a multi-organ function degradation time sequence pattern, a physiological system disorder diffusion path, a clinical state deterioration rate map, and an organ-interaction functional linkage collapse sequence.

[0091] Specifically, the process of performing nonlinear dynamics analysis on the patient's pathological state feature map to obtain the dynamic trajectory of disease progression is a key step in deeply exploring and understanding the dynamic changes of complex physiological systems. Through a series of advanced data processing and mathematical modeling techniques, this process can not only reveal the conversion rules between different pathological states but also provide strong support for formulating personalized treatment plans. First, multi-scale complexity calculation is performed on the patient's pathological state feature map, which is the first step in the whole process. This method aims to extract the disease evolution complexity feature sequences from the complex pathological state data, and these sequences contain various scale features of the disease condition changing over time. For example, in a patient with severe cardio-pulmonary diseases, the temporal evolution pattern between cardiac electrical activity and vital signs such as respiratory rate and blood oxygen saturation is particularly crucial. Further, by performing recurrence quantification analysis on the disease evolution complexity feature sequences, a state transition probability matrix can be obtained. This matrix not only shows the conversion possibilities between different pathological states but also reveals the probability distribution of these conversions. For instance, in the example of the aforementioned patient, it is found through recurrence quantification analysis that the coupling between cardiac electrical activity and respiratory rate has a relatively high conversion probability, indicating the close interaction between the cardio-pulmonary systems. Then, dominant dynamic mode extraction is performed on the state transition probability matrix through singular value decomposition, which is the core part of the second step. This method can capture the main conversion modes between different pathological states and generate the disease state attractor structure. This attractor structure not only shows the internal connections between different pathological states but also reveals the stability of these connections. For example, in the case of the aforementioned patient, through singular value decomposition, a significant attractor structure between cardiac function degradation and respiratory rate fluctuations is found, which may imply a certain underlying pathological mechanism. Next, by performing bifurcation theory analysis on the disease state attractor structure, a set of clinical state transition critical points can be obtained. These critical points reflect the risk areas where major pathological state changes occur under different conditions. For example, in the example of the aforementioned patient, through bifurcation theory analysis, important critical points such as the precursor of cardiovascular function collapse, the limit parameters of respiratory system compensation, the markers of multi-organ function coordination disorder, and the irreversible points of metabolic disorders are identified, providing a basis for timely intervention. Next, reachability analysis is performed on the set of clinical state transition critical points, which is the key content of the third step. Through this method, a disease state transition network diagram can be constructed starting from different critical points. This network diagram not only shows the conversion paths between different pathological states but also reveals the possibilities and risks of these paths. For example, in the case of the aforementioned patient, through reachability analysis, a close connection between cardiac function degradation and renal failure is found, which provides an important clue for further formulating treatment plans. Then, based on the disease state transition network diagram, the minimum action path calculation is performed to obtain the key regulatory nodes of disease progression.These nodes are not only the key turning points in the disease progression but also provide targets for formulating effective intervention measures. Subsequently, a dynamic system evolution simulation of the key regulatory nodes of the disease progression is carried out based on stochastic differential equations, which is the core content of the fourth step. This method can capture the dynamic changes between different pathological states and generate a time-varying disease state manifold. This manifold not only shows the changing trend of the pathological state over time but also reveals the geometric characteristics of these changes. For example, in the case of the above patient, through stochastic differential equations, the dynamic association between cardiac function degradation and renal failure was simulated, and it was found that the mutual influence between the two showed obvious non-linear characteristics. Then, based on the time-varying disease state manifold, vector field integral calculation is carried out, and finally, the dynamic trajectory of the disease progression is obtained. This trajectory includes not only important contents such as the sequential pattern of multi-organ function degradation, the diffusion path of physiological system disorder, the map of the deterioration rate of the clinical state, and the sequence of the collapse of the functional linkage between organs but also provides a comprehensive tool for doctors to evaluate the patient's health status. To better understand the practical application of this complex process, we can continue to use the previous case of a critically ill patient. Suppose this patient is experiencing a deteriorating stage of acute heart failure. Through the detailed analysis of the above steps, the medical team can not only monitor the patient's cardiac electrophysiological activity pattern, respiratory rate fluctuation, and blood oxygen saturation change in real time but also deeply understand the interaction between these signals. For example, through multi-scale complexity calculation, a disease evolution complexity feature sequence is extracted from the patient's pathological state feature map, revealing the complex interaction pattern between cardiac electrical activity and vital signs such as respiratory rate and blood oxygen saturation. Further, through recursive quantitative analysis of the disease evolution complexity feature sequence, a state transition probability matrix is obtained, showing the close coupling relationship between cardiac electrical activity and respiratory rate, and through singular value decomposition, the disease state attractor structure is identified. Then, through bifurcation theory analysis of the disease state attractor structure, important critical points such as the precursor of cardiovascular function collapse, the limit parameter of respiratory system compensation, the marker of multi-organ function coordination disorder, and the irreversible point of metabolic disorder are identified. Then, through the reachability analysis of these critical points, a disease state transition network diagram is constructed, further refining the pathological information. For example, through the calculation of the minimum action path, a stable progression path between cardiac function degradation and renal failure is found, indicating the need for emergency intervention measures. Then, based on stochastic differential equations, a dynamic system evolution simulation is carried out, generating a time-varying disease state manifold, and based on this, vector field integral calculation is carried out to obtain the dynamic trajectory of the disease progression. This trajectory not only shows the sequential pattern of multi-organ function degradation of the patient but also provides important information such as the diffusion path of physiological system disorder, the map of the deterioration rate of the clinical state, and the sequence of the collapse of the functional linkage between organs, providing a scientific basis for adjusting the treatment strategy.The whole process demonstrates how to extract valuable information from complex physiological data and convert it into key clues for guiding clinical practice, thereby improving the management efficiency and treatment effect of critically ill patients.

[0092] In a specific embodiment, the vector field integration calculation is performed based on the time-varying disease state manifold to obtain the dynamic trajectory of disease progression, including:

[0093] Perform a Stokes' theorem transformation on the time-varying disease state manifold to obtain a disease state boundary flux distribution map, and perform a tangential field decomposition on the disease state boundary flux distribution map to obtain a pathological state migration velocity vector field;

[0094] Analyze the geometric characteristics of the pathological state migration velocity vector field through the Riemann curvature tensor to obtain a set of geodesics of the disease progression path, and perform a manifold parallel transport calculation based on the set of geodesics of the disease progression path to obtain the minimum action path of clinical state conversion;

[0095] Solve the Hamilton-Jacobi equation for the minimum action path of clinical state conversion to obtain an optimal control surface of disease progression time, and construct a Poincaré section based on the optimal control surface of disease progression time to obtain the dynamic trajectory of disease progression.

[0096] Specifically, the process of calculating the vector field integral based on the time-varying disease state manifold to obtain the dynamic trajectory of disease progression is a key step in deeply understanding and quantifying the dynamic changes of complex physiological systems. Through a series of advanced mathematical and geometric analysis techniques, this process can not only reveal the conversion laws between different pathological states but also provide strong support for formulating personalized treatment plans. First, perform the Stokes' theorem transformation on the time-varying disease state manifold, which is the first step of the whole process. Stokes' theorem is an important mathematical tool that can transform a complex high-dimensional manifold into a flux distribution map on the boundary. By applying the Stokes' theorem transformation to the time-varying disease state manifold, a flux distribution map of the disease state boundary can be obtained. This distribution map not only shows the changes in the disease state at different time points but also reveals the boundary characteristics of these changes. For example, in a patient with severe cardio-pulmonary diseases, the dynamic change patterns between cardiac electrical activities and vital signs such as respiratory rate and blood oxygen saturation are particularly important. Further, by performing a tangential field decomposition on the flux distribution map of the disease state boundary, a velocity vector field of pathological state migration can be obtained. This vector field not only shows the migration direction and speed between different pathological states but also reveals the dynamic characteristics of these migrations. For instance, in the example of the aforementioned patient, through tangential field decomposition, it is found that there is a significant difference in the migration speed between cardiac function degradation and respiratory rate fluctuations, indicating the need to pay attention to the synchronous changes of the heart and respiratory systems. Next, perform a geometric property analysis on the velocity vector field of pathological state migration through the Riemann curvature tensor, which is the core part of the second step. The Riemann curvature tensor is a mathematical tool used to describe the geometric properties of a manifold. Through this method, a set of geodesics of the disease progression path can be extracted from the velocity vector field of pathological state migration. These geodesics not only show the shortest paths between different pathological states but also reveal the geometric properties of these paths. For example, in the case of the aforementioned patient, through Riemann curvature tensor analysis, a significant geodesic between cardiac function degradation and renal failure is found, which may imply a certain underlying pathological mechanism. Then, based on the set of geodesics of the disease progression path, perform a manifold parallel transport calculation to obtain the minimum-action path of clinical state conversion. These paths not only show the most effective conversion paths between different pathological states but also provide an important reference basis for evaluating the treatment effect. Then, solve the Hamilton-Jacobi equation for the minimum-action path of clinical state conversion, which is the key content of the third step. The Hamilton-Jacobi equation is a classical equation used to describe the optimal control problem of a dynamic system. Through this method, a time-optimal control surface of disease progression can be extracted from the minimum-action path. This surface not only shows the optimal conversion time between different pathological states but also reveals the time dependence of these conversions.For example, in the case of the aforementioned patient, through the solution of the Hamilton-Jacobi equation, it was found that the optimal transition time between cardiac function degradation and renal failure had obvious non-linear characteristics, suggesting that it was necessary to comprehensively consider the functional states of multiple systems to adjust the treatment strategy. Then, based on the optimal control surface of the disease progression time, a Poincaré section was constructed, and finally the dynamic trajectory of the disease progression was obtained. This trajectory not only includes important contents such as the temporal pattern of multi-organ function degradation, the diffusion path of physiological system disorders, the map of the deterioration rate of the clinical state, and the sequence of functional linkage collapse between organs, but also provides a comprehensive tool for doctors to evaluate the patient's health status. To better understand the practical application of this complex process, we can continue to use the previous case of a critically ill patient. Suppose this patient is experiencing a deteriorating stage of acute heart failure. Through the detailed analysis of the above steps, the medical team can not only monitor the patient's cardiac electrophysiological activity pattern, respiratory rate fluctuation, and blood oxygen saturation change in real time, but also deeply understand the interaction between these signals. For example, through the Stokes theorem transformation, the distribution map of the boundary flux of the disease state was extracted from the time-varying disease state manifold, revealing the dynamic change pattern between cardiac electrical activity and vital signs such as respiratory rate and blood oxygen saturation. Further, by performing a tangential field decomposition on the distribution map of the disease state boundary flux, the velocity vector field of the pathological state migration was obtained, showing the close coupling relationship between cardiac electrical activity and respiratory rate, and through the analysis of the Riemann curvature tensor, the set of geodesics of the disease progression path was identified. Then, by performing a manifold parallel transport calculation on these sets of geodesics, the minimum action path of the clinical state transition was obtained, providing a goal for further formulating the treatment plan. Then, by solving the Hamilton-Jacobi equation for the minimum action path, the optimal control surface of the disease progression time was obtained, revealing the optimal transition time between different pathological states. Finally, based on the optimal control surface of the disease progression time, a Poincaré section was constructed, generating the dynamic trajectory of the disease progression. This trajectory not only shows the temporal pattern of multi-organ function degradation of the patient, but also provides important information such as the diffusion path of physiological system disorders, the map of the deterioration rate of the clinical state, and the sequence of functional linkage collapse between organs, providing a scientific basis for adjusting the treatment strategy. Throughout the process, the Stokes theorem transformation helps us extract the distribution map of the flux on the boundary from complex high-dimensional data, thus revealing the dynamic change pattern of the disease; the tangential field decomposition further refines the speed and direction of the pathological state migration, providing detailed dynamic characteristics. The application of the Riemann curvature tensor enables us to capture the geometric characteristics of the disease progression path, while the manifold parallel transport calculation helps us find the most effective clinical state transition path.The solution of the Hamilton-Jacobi equation provides us with the optimal control surface, revealing the transition times between different pathological states, while the construction of the Poincaré section ultimately generates the dynamic trajectory of disease progression, providing a comprehensive tool for evaluating the patient's health status. The whole process demonstrates how to extract valuable information from complex physiological data and transform it into key clues for guiding clinical practice, thereby improving the management efficiency and treatment effect of critically ill patients.

[0097] In a specific embodiment, the clinical decision support reasoning based on the dynamic trajectory of disease progression to obtain personalized treatment recommendations includes:

[0098] Perform trajectory segmentation processing on the dynamic trajectory of disease progression to obtain a sequence of key pathological stage transition points, and perform multi-dimensional critical state analysis on the sequence of key pathological stage transition points to obtain a set of treatment intervention time windows;

[0099] Based on the Bayesian causal network, infer the inter-organ action mechanism for the set of treatment intervention time windows to obtain a multi-level pathological causal chain structure, and perform key node sensitivity analysis on the multi-level pathological causal chain structure to obtain a treatment target priority matrix;

[0100] Through multi-objective constraint optimization, calculate the treatment resource allocation for the treatment target priority matrix to obtain a multi-dimensional pharmacokinetic parameter space, and construct a non-linear response surface for the multi-dimensional pharmacokinetic parameter space to obtain a treatment plan - physiological response mapping relationship diagram;

[0101] Perform Pareto optimal solution analysis on the treatment plan - physiological response mapping relationship diagram to obtain a set of multi-objective balanced treatment strategies, and based on the set of multi-objective balanced treatment strategies, perform clinical risk - benefit quantitative assessment to obtain the safety margin of the individualized intervention plan;

[0102] Based on the temporal Bayesian decision network, perform dynamic treatment path planning for the safety margin of the individualized intervention plan to obtain a sequence of phased treatment goals, and based on the sequence of phased treatment goals, perform treatment plan combination optimization calculation to obtain personalized treatment recommendations.

[0103] Specifically, the process of clinical decision-making support reasoning based on the dynamic trajectory of disease progression to obtain personalized treatment recommendations is a complex and multi-level analysis process. It aims to formulate the most optimized treatment strategy for patients by deeply understanding the changes in the patient's pathological state and the underlying physiological mechanisms. This process not only requires the support of advanced mathematical and statistical methods but also a profound understanding of biology, medicine, and pharmacology knowledge. First, the dynamic trajectory of the disease progression is subjected to trajectory segmentation, which is the first step of the whole process. The purpose of trajectory segmentation is to identify a sequence of key pathological stage transition points, which mark the critical moments when the disease transitions from one stage to another. For example, in a patient with chronic kidney disease complicated with cardiovascular disease, a change in the correlation between the deterioration of kidney function and the decline in cardiac pumping ability may be observed at a certain time point, which is one of the key pathological stage transition points that we need to focus on. Then, multi-dimensional critical state analysis of these sequences of key pathological stage transition points can reveal the pathological state characteristics at each transition point and further determine the optimal timing for intervention, that is, the set of treatment intervention time windows. This analysis helps doctors understand when it is most effective to take action. For example, in the above case, it is found that the best time to initiate intensive treatment is before the significant decline in cardiac pumping ability. Next, based on the Bayesian causal network, the mechanism of action between organs is inferred for the set of treatment intervention time windows, which is the core part of the second step. The Bayesian causal network is a powerful tool that can help us understand the causal relationships between different variables. Through this method, a multi-level pathological causal chain structure can be inferred from the set of treatment intervention time windows, revealing the complex interaction patterns between different organs. For example, in the case of the aforementioned patient, through Bayesian causal network analysis, a potential causal link between cardiac function degradation and renal failure is found, indicating that the conditions of both systems should be considered simultaneously during treatment. Then, key node sensitivity analysis of these multi-level pathological causal chain structures can obtain a treatment target priority matrix, thereby clarifying which parts or processes should be the main targets of treatment. For example, in this case, it is found that improving the metabolic efficiency of cardiomyocytes is crucial for slowing down the disease progression, so it is taken as the primary treatment target. Subsequently, through multi-objective constraint optimization, the treatment resource allocation calculation is carried out for the treatment target priority matrix, which is an important part of the third step. Multi-objective constraint optimization is a method used to solve the resource allocation problem in complex systems. Through this method, the best combination of drug doses and treatment regimens can be calculated from the treatment target priority matrix, generating a multi-dimensional pharmacokinetic parameter space. This parameter space not only shows the effects of different drugs and treatment means but also reveals the interactions between them. For example, in the above case, through multi-objective constraint optimization, the best drug combination that can effectively improve cardiac function and protect renal function is found.Next, a non-linear response surface is constructed for this multi-dimensional pharmacokinetic parameter space, and a treatment plan - physiological response mapping diagram can be obtained, which shows the expected physiological responses under different treatment plans. This mapping diagram provides intuitive guidance for doctors to help them select the most suitable treatment plan for patients. Then, a Pareto optimal solution analysis is performed on the treatment plan - physiological response mapping diagram, which is a key step in the fourth step. Pareto optimal solution analysis is a method for evaluating the trade-offs between multiple options. Through this method, a set of treatment strategies that can achieve the best balance among multiple goals can be screened out from the treatment plan - physiological response mapping diagram, that is, the set of multi-objective balanced treatment strategies. For example, in the case of the aforementioned patient, through Pareto optimal solution analysis, a set of treatment plans that can significantly improve the heart's pumping ability without imposing an additional burden on the kidneys was found. Next, based on this set of multi-objective balanced treatment strategies, a clinical risk - benefit quantification assessment is carried out to obtain the safety margin of the individualized intervention plan, clarifying the risks and benefits of each treatment strategy. For example, in this case, it was found that although some high-intensity treatment plans can rapidly improve heart function, in the long run, they may increase the risk of kidney injury, so careful selection is required. Finally, based on the time-series Bayesian decision network, dynamic treatment path planning is carried out for the safety margin of the individualized intervention plan, which is the core task of the fifth step. The time-series Bayesian decision network is a powerful tool for decision-making analysis that combines the time dimension. Through this method, a sequence of phased treatment goals can be planned from the safety margin of the individualized intervention plan, setting specific treatment goals for each stage. For example, in the case of the above patient, through the time-series Bayesian decision network, a comprehensive treatment plan including stabilizing heart function in the short term, restoring partial kidney function in the medium term, and maintaining both in the long term was formulated. Then, based on this sequence of phased treatment goals, optimization calculations of treatment plan combinations are carried out, and finally personalized treatment recommendations are obtained. These recommendations not only consider the patient's current pathological state but also predict possible future changes, ensuring the flexibility and adaptability of the treatment plan. In summary, through a detailed analysis of the dynamic trajectory of disease progression, combined with advanced technologies such as Bayesian causal networks, multi-objective constrained optimization, Pareto optimal solution analysis, and time-series Bayesian decision networks, we can customize a set of scientific and reasonable personalized treatment recommendations for each patient. These technologies work together not only to improve the effectiveness and safety of treatment but also to provide strong support for clinical decision-making. For example, in the case of the aforementioned patient with chronic kidney disease complicated with cardiovascular disease, through the implementation of the above series of steps, the medical team can not only accurately locate the treatment target but also adjust the drug dosage and treatment plan according to the patient's specific situation, ultimately achieving the best treatment effect. Throughout the process, various advanced analysis methods complement each other, making personalized treatment possible and greatly improving the management efficiency and treatment quality of critically ill patients.

[0104] The data fusion processing method of the MEMS-based intelligent sensor in the embodiments of the present invention has been described above. Next, the data fusion processing system of the MEMS-based intelligent sensor in the embodiments of the present invention will be described. Please refer to Figure 2 An embodiment of the data fusion processing system of the MEMS-based intelligent sensor in the embodiments of the present invention includes:

[0105] A sampling module 21, configured to perform multi-channel synchronous sampling on the physiological signals of critically ill patients collected by the MEMS sensor to obtain a multi-modal physiological parameter time series data stream;

[0106] A decomposition module 22, configured to decompose the multi-modal physiological parameter time series data stream into non-stationary signals to obtain a key vital sign feature sequence;

[0107] A calculation module 23, configured to calculate the phase coupling degree of the key vital sign feature sequence to obtain an organ-interfunction correlation strength matrix;

[0108] A projection module 24, configured to project the critically ill patient into a high-dimensional feature space based on the organ-interfunction correlation strength matrix to obtain a patient pathological state feature map;

[0109] An analysis module 25, configured to perform non-linear dynamics analysis on the patient pathological state feature map to obtain a disease progression dynamic trajectory;

[0110] An inference module 26, configured to perform clinical decision support inference based on the disease progression dynamic trajectory to obtain personalized treatment suggestions.

[0111] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the description in the above method embodiment, and details will not be repeated here.

[0112] Refer to Figure 3 In addition, an embodiment of the present invention also provides a computer device, and its internal structure can be as Figure 3 shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements the above method.

[0113] Those skilled in the art can understand that Figure 3 The structure shown in Figure 3 is only a block diagram of some structures related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0114] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0115] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above method embodiments. Among them, any reference to a memory, storage, database or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be obtained in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0116] It should be noted that in this article, the terms "including", "comprising" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, device, article or method. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, device, article or method including that element.

[0117] The above are only the preferred embodiments of the present invention, and do not thereby limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be similarly included in the patent protection scope of the present invention.

Claims

1. A data fusion processing method based on MEMS intelligent sensor, characterized in that: The following steps are involved: Perform multi-channel synchronous sampling and fusion processing on the physiological signals of critically ill patients collected by MEMS sensors to obtain multi-modal physiological parameter time series data stream; Performing non-stationary signal decomposition on the multimodal physiological parameter time series data stream to obtain a key vital sign feature sequence; Calculating the phase coupling degree of the key vital sign characteristic sequence to obtain an inter-organ functional association strength matrix; Performing high-dimensional feature space projection on the critically ill patient based on the inter-organ functional association strength matrix to obtain a feature map of the patient's pathological state; Performing nonlinear dynamic analysis on the characteristic graph of the patient's pathological state to obtain a dynamic trajectory of disease progression; Perform clinical decision support reasoning based on the dynamic trajectory of disease progression to obtain personalized treatment recommendations; The non-stationary signal decomposition of the multimodal physiological parameter time series data stream to obtain a key vital sign feature sequence includes: Adaptively performing biorthogonal filtering separation on the multimodal physiological parameter time series data stream to obtain a nonlinear physiological fluctuation component, and performing multi-level harmonic analysis on the nonlinear physiological fluctuation component through an adaptive oscillation identification mechanism to obtain a physiological system phase change feature set; Based on the multi-scale covariance structure method, a statistical significance analysis is performed on the phase change feature set of the physiological system to obtain a key physiological coupling network, and a topological stability assessment is performed on the key physiological coupling network to obtain multi-organ function linkage data; Phase synchronization measurement and extraction are performed on the multi-organ function linkage data to obtain a key vital sign feature sequence; wherein the key vital sign feature sequence includes the cardiovascular system regulation ability, respiratory compensation reserve index, circulatory system homeostasis maintenance parameters and multi-system collaborative function characteristics.

2. The data fusion processing method based on MEMS intelligent sensor according to claim 1 is characterized in that: The phase coupling degree is calculated based on the key vital sign characteristic sequence to obtain the inter-organ functional association strength matrix, including: Performing bispectral coherence analysis on the key vital sign characteristic sequence to obtain a nonlinear phase synchronization characteristic spectrum, and performing time-varying causal relationship analysis on the nonlinear phase synchronization characteristic spectrum based on wavelet bidirectional transformation to obtain a functional network connection strength map; Performing directional information transfer analysis on the functional network connection strength map by using a transfer entropy calculation method to obtain a multi-organ interaction matrix, and performing resonance frequency band screening on the multi-organ interaction matrix to obtain a collaborative working mode of key physiological systems; The phase locking value of the collaborative working mode of the key physiological systems is calculated to obtain the organ function synchronization degree data, and a graph theory network characteristic analysis is performed based on the organ function synchronization degree data to obtain the organ function association strength matrix; wherein the organ function association strength matrix includes the cardiopulmonary function collaborative efficiency coefficient, cerebrovascular autoregulation capacity parameters, renal blood perfusion status and multi-system metabolic balance network characteristics.

3. The data fusion processing method based on MEMS intelligent sensor according to claim 1 is characterized in that: The high-dimensional feature space projection is performed based on the inter-organ functional association strength matrix to obtain the patient's pathological state feature map, including: Performing a tensor decomposition operation on the inter-organ functional association intensity matrix to obtain a multi-dimensional physiological system interaction feature space, performing topological structure extraction on the multi-dimensional physiological system interaction feature space to obtain a disease state topological skeleton diagram, and performing multi-scale persistence calculation on the disease state topological skeleton diagram to obtain a set of key pathological state branch points; Performing nonlinear boundary mapping on the key pathological state branch point set by using kernel surface reconstruction technology to obtain an organ function degradation partition map, and performing spectral clustering segmentation processing on the organ function degradation partition map to obtain a pathological state subspace structure; Performing geometric depth feature extraction on the pathological state subspace structure to obtain a disease progression path network, and performing Ricci curvature calculation based on the disease progression path network to obtain a pathological state transformation risk surface; Based on differential geometry mapping, a curvature flow evolution analysis is performed on the pathological state transformation risk surface to obtain a disease state manifold graph, and a characteristic vector field is constructed based on the disease state manifold graph to obtain a patient pathological state characteristic map, wherein the patient pathological state characteristic map includes a multi-organ dysfunction pattern, a critical disease development stage marker, a vital sign instability risk index, and an organ function recovery potential assessment feature.

4. The data fusion processing method based on MEMS intelligent sensor according to claim 1 is characterized in that: The nonlinear dynamic analysis of the characteristic map of the patient's pathological state to obtain a dynamic trajectory of disease progression includes: Performing multi-scale complexity calculation on the patient's pathological state characteristic map to obtain a disease evolution complexity characteristic sequence, and performing recursive quantitative analysis on the disease evolution complexity characteristic sequence to obtain a state transition probability matrix; Extracting the dominant dynamical mode of the state transition probability matrix by singular value decomposition to obtain a disease state attractor structure, and performing bifurcation theory analysis on the disease state attractor structure to obtain a set of clinical state transition critical points; Performing reachability analysis on the set of critical points of clinical state transition to obtain a disease state transition network diagram, and performing minimum action path calculation based on the disease state transition network diagram to obtain key regulatory nodes for disease progression; Based on stochastic differential equations, a dynamic system evolution simulation is performed on the key regulatory nodes of the disease progression to obtain a time-varying disease state manifold, and a vector field integral calculation is performed based on the time-varying disease state manifold to obtain a dynamic trajectory of disease progression.

5. The data fusion processing method based on MEMS intelligent sensor according to claim 4 is characterized in that: The step of performing vector field integral calculation based on the time-varying disease state manifold to obtain a dynamic trajectory of disease progression includes: Performing Stokes theorem transformation on the time-varying disease state manifold to obtain a disease state boundary flux distribution map, and performing tangential field decomposition on the disease state boundary flux distribution map to obtain a pathological state migration velocity vector field; The geometric characteristics of the pathological state migration velocity vector field are analyzed by Riemann curvature tensor to obtain a set of disease progression path geodesics, and a manifold parallel transmission calculation is performed based on the set of disease progression path geodesics to obtain a minimum action path for clinical state conversion; The Hamilton-Jacobi equation is solved for the minimum action path of the clinical state transition to obtain the optimal control surface of the disease progression time, and the Poincaré section is constructed based on the optimal control surface of the disease progression time to obtain the dynamic trajectory of the disease progression.

6. The data fusion processing method based on MEMS intelligent sensor according to claim 1 is characterized in that: The clinical decision support reasoning based on the dynamic trajectory of disease progression to obtain personalized treatment recommendations includes: Performing trajectory segmentation processing on the dynamic trajectory of disease progression to obtain a sequence of key pathological stage transition points, and performing multidimensional critical state analysis on the sequence of key pathological stage transition points to obtain a set of treatment intervention time windows; Based on the Bayesian causal network, the inter-organ action mechanism of the treatment intervention time window set is inferred to obtain a multi-level pathological causal chain structure, and the key node sensitivity analysis of the multi-level pathological causal chain structure is performed to obtain a treatment target priority matrix; The treatment resource allocation calculation is performed on the treatment target priority matrix through multi-objective constraint optimization to obtain a multidimensional pharmacological kinetic parameter space, and a nonlinear response surface is constructed on the multidimensional pharmacological kinetic parameter space to obtain a treatment plan-physiological response mapping relationship diagram; Performing a Pareto optimal solution analysis on the treatment plan-physiological response mapping relationship diagram to obtain a multi-objective balanced treatment strategy set, and performing a clinical risk-benefit quantitative assessment based on the multi-objective balanced treatment strategy set to obtain a safety boundary of the individualized intervention plan; Based on the temporal Bayesian decision network, dynamic treatment path planning is performed on the safety boundary of the individualized intervention plan to obtain a staged treatment target sequence, and treatment plan combination optimization calculation is performed based on the staged treatment target sequence to obtain personalized treatment recommendations.

7. A data fusion processing system based on MEMS intelligent sensor, characterized in that: include: The sampling module is used to perform multi-channel synchronous sampling and fusion processing on the physiological signals of critically ill patients collected by MEMS sensors to obtain a multi-modal physiological parameter time series data stream; A decomposition module, used for performing non-stationary signal decomposition on the multimodal physiological parameter time series data stream to obtain a key vital sign feature sequence; A calculation module, used to calculate the phase coupling degree of the key vital sign characteristic sequence to obtain an inter-organ functional association strength matrix; A projection module, used for performing high-dimensional feature space projection on the critically ill patient based on the inter-organ functional association strength matrix to obtain a feature map of the patient's pathological state; An analysis module, used for performing nonlinear dynamic analysis on the characteristic map of the patient's pathological state to obtain a dynamic trajectory of disease progression; A reasoning module, used for performing clinical decision support reasoning based on the dynamic trajectory of disease progression to obtain personalized treatment recommendations; The non-stationary signal decomposition of the multimodal physiological parameter time series data stream to obtain a key vital sign feature sequence includes: Adaptively performing biorthogonal filtering separation on the multimodal physiological parameter time series data stream to obtain a nonlinear physiological fluctuation component, and performing multi-level harmonic analysis on the nonlinear physiological fluctuation component through an adaptive oscillation identification mechanism to obtain a physiological system phase change feature set; Based on the multi-scale covariance structure method, a statistical significance analysis is performed on the phase change feature set of the physiological system to obtain a key physiological coupling network, and a topological stability assessment is performed on the key physiological coupling network to obtain multi-organ function linkage data; Phase synchronization measurement and extraction are performed on the multi-organ function linkage data to obtain a key vital sign feature sequence; wherein the key vital sign feature sequence includes the cardiovascular system regulation ability, respiratory compensation reserve index, circulatory system homeostasis maintenance parameters and multi-system collaborative function characteristics.

8. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images