A method and system for dynamic monitoring of magnesium nanoparticles for sepsis

By constructing a dynamic pathological simulation liquid medium for sepsis, and collecting and weighting the parameters of magnesium nanoparticles in real time, the shortcomings of existing technologies in monitoring the stability of magnesium nanoparticles in sepsis patients are solved, enabling accurate stability assessment and early warning in complex pathological environments.

CN123111899APending Publication Date: 2026-11-17THE FIRST AFFILIATED HOSPITAL OF JINAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611618487.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing nano-magnesium particle stability monitoring technology has significant limitations in its application in sepsis patients, failing to meet the needs of precision research and development and clinical quality control. In particular, under pathological microenvironments characterized by inflammatory factor overload, plasma ion disorder, and acid-base dynamic imbalance, nano-magnesium particles are prone to non-specific aggregation and activity decay.

Method used

By constructing a dynamic pathological simulation liquid medium for sepsis, the hydration particle size, polydispersity coefficient, and Zeta potential of magnesium nanoparticles are collected in real time. Combined with pH value, inflammatory factor concentration, and ionic strength, the core dominant factors are determined, weighted gain is applied, and multi-parameter weighted fusion is performed to obtain a comprehensive stability index, thereby achieving dynamic monitoring.

Benefits of technology

It improves the accuracy of assessing the colloidal stability of magnesium nanoparticles in complex pathological environments, can respond to dynamic disturbances in the pathological environment, accurately capture subtle instability features, and improve the accuracy and sensitivity of monitoring.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123111899A_ABST
    Figure CN123111899A_ABST
Patent Text Reader

Abstract

This invention relates to the field of biomedical monitoring technology, and provides a method and system for dynamic monitoring of magnesium nanoparticles in sepsis. The method includes: injecting a sample of magnesium nanoparticles into a liquid medium; obtaining a hydration particle size temporal fluctuation coefficient based on the hydration particle size; obtaining the mean Zeta potential and potential fluctuation characteristics based on the Zeta potential value; obtaining an acid-base regulating factor positively correlated with real-time pH, an inflammation regulating factor positively correlated with real-time inflammatory factor concentration, and an ion regulating factor positively correlated with real-time ionic strength; determining the core dominant factor and applying weighted gains to the corresponding characteristic parameters; and weightedly fusing the polydispersity coefficient, hydration particle size temporal fluctuation coefficient, mean Zeta potential, and potential fluctuation characteristics to obtain a comprehensive stability index. This invention can respond to dynamic disturbances in the pathological environment, accurately capture subtle instability characteristics of magnesium nanoparticles, thereby improving the accuracy of colloidal stability assessment under complex pathological conditions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of biomedical monitoring technology, and more specifically, to a method and system for dynamic monitoring of magnesium nanoparticles for sepsis. Background Technology

[0002] Existing nanoparticle stability monitoring technologies are mostly based on conventional buffer static detection, relying solely on three basic parameters—hydrated particle size, polydispersity index, and zeta potential—for single-trend prediction, enabling only passive over-limit warnings in general scenarios. For sepsis treatment, existing technologies have significant limitations. Specifically, sepsis patients have a unique pathological microenvironment characterized by inflammatory factor overload, plasma ion disorder, and dynamic acid-base imbalance. Nanoparticles are highly susceptible to non-specific aggregation and activity decay, resulting in a significant discrepancy between conventional static monitoring data and actual in vivo stability. This makes it difficult to meet the precise development and clinical quality control needs of sepsis-specific nanoparticle formulations. Summary of the Invention

[0003] To achieve dynamic monitoring of magnesium nanoparticles, this invention provides a method and system for dynamic monitoring of magnesium nanoparticles in sepsis, the specific technical solution of which is as follows: A method for dynamic monitoring of magnesium nanoparticles for sepsis includes the following steps: S1. Construct a dynamic pathological simulation liquid medium for sepsis and inject the sample of magnesium nanoparticles into the liquid medium. Collect the hydration particle size, polydispersity coefficient and Zeta potential value of the magnesium nanoparticles. Obtain the hydration particle size time-series fluctuation coefficient based on the hydration particle size and obtain the mean Zeta potential and potential fluctuation characteristics based on the Zeta potential value. S2, obtain the real-time pH value, real-time inflammatory factor concentration and real-time ionic strength of the liquid medium, and obtain the acid-base regulating factor positively correlated with the real-time pH value, the inflammatory regulating factor positively correlated with the real-time inflammatory factor concentration and the ionic regulating factor positively correlated with the real-time ionic strength according to the real-time pH value, the real-time inflammatory factor concentration and the real-time ionic strength respectively. S3, determine the core dominant factor. If the acid-base regulation factor is the core dominant factor, apply weighted gain to the potential fluctuation characteristics. If the ion regulation factor is the core dominant factor, apply weighted gain to the hydration particle size temporal fluctuation coefficient and polydispersity coefficient. If the inflammation regulation factor is the core dominant factor, apply weighted gain to the hydration particle size temporal fluctuation coefficient and the mean Zeta potential. S4. The polydispersity index, the temporal fluctuation coefficient of hydrated particle size, the mean value of Zeta potential, and the potential fluctuation characteristics are weighted and fused to obtain a comprehensive stability index for representing the colloidal stability of magnesium nanoparticles. The stability of magnesium nanoparticles is monitored based on the comprehensive stability index.

[0004] The proposed method for dynamic monitoring of magnesium nanoparticles for sepsis constructs a dynamic pathological simulation liquid medium for sepsis, collects characteristic parameters of magnesium nanoparticles and the liquid medium in real time, and assigns weighted gains to regulatory factors based on the determination of core dominant factors. On this basis, the polydispersity coefficient, the temporal fluctuation coefficient of hydrated particle size, the mean Zeta potential, and the potential fluctuation characteristics are weighted and fused to obtain an accurate comprehensive stability index for representing the colloidal stability of magnesium nanoparticles. This index can respond to dynamic disturbances in the pathological environment, accurately capture subtle instability characteristics of magnesium nanoparticles, and thus improve the accuracy of colloidal stability assessment under complex pathological conditions.

[0005] Preferably, step S1 specifically includes: S11, construct a dynamic pathological simulation liquid medium for sepsis and inject the sample of magnesium nanoparticles into the liquid medium, and dynamically adjust the pH value, electrolyte concentration and inflammatory factor concentration parameters of the liquid medium according to the progression of sepsis. S12. After each parameter adjustment, the hydration particle size, polydispersity coefficient and Zeta potential value of the magnesium nanoparticles are collected synchronously. The standard deviation and average value of the hydration particle size within the preset window are obtained. The standard deviation is normalized according to the average value to obtain the time-series fluctuation coefficient of the hydration particle size. The mean value of the Zeta potential and the potential fluctuation characteristics are obtained according to the Zeta potential value. Among them, the characteristics of potential fluctuation include the amplitude of potential fluctuation and the frequency of potential fluctuation.

[0006] Preferably, step S2 specifically includes: S21, based on the normal steady-state pH value of the human body, obtain the absolute value of the acid-base deviation between the real-time acid-base value and the normal steady-state pH value of the human body, and normalize and fit the absolute value of acid-base based on the maximum pathological acid-base deviation in clinical sepsis to obtain the acid-base regulation factor that is positively correlated with the acid-base value. S22, obtain the real-time concentrations of TNF-α and IL-6 inflammatory factors and normalize them. Then, perform weighted fusion on the normalized real-time concentrations of TNF-α and IL-6 inflammatory factors to obtain inflammatory regulatory factors that are positively correlated with the concentrations of inflammatory factors. S23. Obtain the standard ion intensity of plasma. Based on the ion intensity deviation between the real-time ion intensity and the standard ion intensity of plasma, obtain the relative deviation value of ion disorder. Normalize and fit the relative deviation value of ion disorder to obtain the ion regulation factor that is positively correlated with the real-time ion intensity and is used to quantify the stress intensity of sepsis ion disorder on particle aggregation.

[0007] Preferably, step S4 specifically includes: S41, the polydispersity index, the temporal fluctuation coefficient of hydrated particle size, the mean value of Zeta potential, and the potential fluctuation characteristics are weighted and fused to obtain the initial stability index; S42, obtain the temporal fluctuation amplitude and average particle size of hydrated particles within a preset sliding time window, and obtain the fluctuation amplitude correction coefficient based on the temporal fluctuation amplitude and average particle size. S43. The initial stability index is corrected according to the fluctuation amplitude correction coefficient and then normalized to obtain the comprehensive stability index used to represent the stability of the nano-magnesium particle colloid.

[0008] Preferably, the method for dynamic monitoring of magnesium nanoparticles for sepsis further includes the following steps: S5. Obtain historical parameters of liquid media, historical parameters of magnesium nanoparticles, and corresponding historical comprehensive stability indices under different sepsis disease progression, and construct a sample dataset using historical parameters of liquid media, historical parameters of magnesium nanoparticles, and corresponding historical comprehensive stability indices. S6. Label each sample in the sample dataset with an instability cause label to obtain the sample training set. The instability cause labels include charge imbalance type aggregation label, ion-induced aggregation label and inflammation adsorption type instability label. S7. Construct a neural network model. Train the neural network model using a sample training set to obtain the current parameters of the liquid medium, the current parameters of the nano-magnesium particles, and the corresponding current comprehensive stability index. Input these parameters into the trained neural network model to obtain the current dominant instability cause.

[0009] Preferably, step S3 specifically includes: S31, respectively obtain the absolute deviation of acid-base regulation factor, inflammation regulation factor and ion regulation factor from their respective steady-state benchmark values, and take the regulation factor corresponding to the largest absolute deviation value as the core dominant factor.

[0010] Preferably, step S3 further includes: S32, if the acid-base regulation factor is the core dominant factor, normalize the mapping based on the deviation of the acid-base regulation factor from the corresponding steady-state reference value to obtain the acid-base weight gain coefficient, and perform gain correction on the basic weight of the potential fluctuation characteristics according to the acid-base weight gain coefficient. Then, perform global normalization on the weights of the hydration particle size time-series fluctuation coefficient, polydispersity coefficient and the mean of the Zeta potential. S33, if the ion regulation factor is the core dominant factor, normalize the mapping based on the absolute value of the deviation of the ion regulation factor from the corresponding steady-state reference value to obtain the ion weight gain coefficient, and then perform gain correction on the basic weights of the hydration particle size time-series fluctuation coefficient and the polydispersity coefficient according to the ion weight gain coefficient. Then, perform global normalization on the weights of the potential fluctuation characteristics and the mean of the Zeta potential. S34. If the inflammatory regulatory factor is the core dominant factor, normalize the mapping based on the deviation of the inflammatory regulatory factor from the corresponding steady-state benchmark value to obtain the inflammatory weight gain coefficient. Then, according to the inflammatory weight gain coefficient, perform gain correction on the basic weights of the hydration particle size time-series fluctuation coefficient and the mean of the Zeta potential. Finally, perform global normalization on the weights of the potential fluctuation characteristics and the polydispersity coefficient.

[0011] A dynamic monitoring system for magnesium nanoparticles in sepsis, used to implement the aforementioned dynamic monitoring method for magnesium nanoparticles, comprising: The magnesium particle parameter acquisition module is used to collect the hydration particle size, polydispersity coefficient and zeta potential value of nano-magnesium particles in the liquid medium of sepsis dynamic pathological simulation. The module obtains the hydration particle size temporal fluctuation coefficient based on the hydration particle size and the mean zeta potential and potential fluctuation characteristics based on the zeta potential value. The medium parameter acquisition module is used to acquire the real-time pH value, real-time inflammatory factor concentration, and real-time ionic strength of the liquid medium. Based on the real-time pH value, real-time inflammatory factor concentration, and real-time ionic strength, it acquires acid-base regulating factors positively correlated with the real-time pH value, inflammatory regulating factors positively correlated with the real-time inflammatory factor concentration, and ionic regulating factors positively correlated with the real-time ionic strength, respectively. The feature weight gain module is used to determine the core dominant factor. If the acid-base regulation factor is the core dominant factor, the potential fluctuation feature is weighted and gained. If the ion regulation factor is the core dominant factor, the hydration particle size temporal fluctuation coefficient and polydispersity coefficient are weighted and gained. If the inflammation regulation factor is the core dominant factor, the hydration particle size temporal fluctuation coefficient and the mean Zeta potential are weighted and gained. The stability monitoring module is used to perform weighted fusion of polydispersity coefficient, hydration particle size temporal fluctuation coefficient, mean Zeta potential, and potential fluctuation characteristics to obtain a comprehensive stability index that represents the colloidal stability of magnesium nanoparticles. The stability of magnesium nanoparticles is monitored based on the comprehensive stability index.

[0012] Preferably, the nano-magnesium particle dynamic monitoring system for sepsis further includes: The sample training set acquisition module is used to acquire historical parameters of liquid media, historical parameters of magnesium nanoparticles, and corresponding historical comprehensive stability indices under different sepsis disease progression. The sample dataset is constructed using the historical parameters of liquid media, historical parameters of magnesium nanoparticles, and corresponding historical comprehensive stability indices. Each sample in the sample dataset is labeled with an instability inducing factor label to obtain the sample training set. The instability inducing factor labels include charge imbalance type aggregation label, ion-induced aggregation label, and inflammation adsorption type instability label. The neural network model is trained using a sample training set and takes the current parameters of the liquid medium, the current parameters of the magnesium nanoparticles, and the corresponding current comprehensive stability index as inputs, and outputs the current dominant instability cause.

[0013] Preferably, the feature weight gain module includes: The dominant factor determination unit is used to obtain the absolute value of the deviation of acid-base regulation factor, inflammation regulation factor and ion regulation factor from their respective steady-state benchmark values, and the regulation factor corresponding to the largest absolute value of deviation is the core dominant factor. The acid-base weight gain unit is used to normalize the acid-base weight gain coefficient when the acid-base regulation factor is the core dominant factor. It is then used to normalize the deviation of the acid-base regulation factor from the corresponding steady-state reference value, and to correct the basic weight of the potential fluctuation characteristics based on the acid-base weight gain coefficient. Finally, the weights of the hydration particle size time-series fluctuation coefficient, polydispersity coefficient and the mean of the Zeta potential are global normalized. The ion weight gain unit is used to perform normalization mapping based on the deviation of the ion regulation factor from the corresponding steady-state reference value when the ion regulation factor is the core dominant factor, to obtain the ion weight gain coefficient, and to perform gain correction on the basic weights of the hydration particle size time-series fluctuation coefficient and the polydispersity coefficient according to the ion weight gain coefficient. Then, the weights of the potential fluctuation characteristics and the mean of the Zeta potential are globally normalized. The inflammation weight gain unit is used to normalize the deviation of the inflammation regulatory factor from the corresponding steady-state benchmark value when the inflammation regulatory factor is the core dominant factor, to obtain the inflammation weight gain coefficient. Then, the basic weights of the hydration particle size temporal fluctuation coefficient and the mean of the Zeta potential are adjusted according to the inflammation weight gain coefficient. Finally, the weights of the potential fluctuation characteristics and the polydispersity coefficient are globally normalized. Attached Figure Description

[0014] The invention will be further understood from the following description taken in conjunction with the accompanying drawings. The components in the drawings are not necessarily drawn to scale, but rather the emphasis is on illustrating the principles of the embodiments. In different views, the same reference numerals designate corresponding parts.

[0015] Figure 1 This is a schematic diagram of the overall process of a method for dynamic monitoring of magnesium nanoparticles for sepsis in one embodiment of the present invention. Figure 2 This is a flowchart illustrating step S1 in one embodiment of the present invention; Figure 3 This is a flowchart illustrating step S2 in one embodiment of the present invention; Figure 4 This is a flowchart illustrating step S3 in one embodiment of the present invention; Figure 5 This is a flowchart illustrating step S4 in one embodiment of the present invention; Figure 6 This is a schematic diagram of the overall process of a method for dynamic monitoring of magnesium nanoparticles for sepsis in another embodiment of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to its embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the scope of protection of the invention.

[0017] like Figure 1 As shown, an embodiment of the present invention provides a method for dynamic monitoring of magnesium nanoparticles for sepsis, comprising the following steps: S1. Construct a dynamic pathological simulation liquid medium for sepsis and inject the sample of magnesium nanoparticles into the liquid medium. Collect the hydration particle size, polydispersity coefficient and Zeta potential value of the magnesium nanoparticles. Obtain the hydration particle size temporal fluctuation coefficient based on the hydration particle size and obtain the mean Zeta potential and potential fluctuation characteristics based on the Zeta potential value.

[0018] For the construction of a liquid medium for simulating the dynamic pathology of sepsis, a simulated plasma base solution can be prepared using sterile ultrapure water. The base solutes include sodium chloride, potassium chloride, calcium chloride, sodium bicarbonate, etc., to simulate the basic ionic environment of human plasma. The base osmotic pressure is controlled and the solution is degassed at room temperature to remove the interference of solution bubbles on light scattering detection, thus ensuring the stability and accuracy of the detection of magnesium nanoparticles.

[0019] As a preferred technical solution, such as Figure 2 As shown, step S1 specifically includes: S11, construct a dynamic pathological simulation liquid medium for sepsis and inject the sample of magnesium nanoparticles into the liquid medium. Adjust the pH value, electrolyte concentration and inflammatory factor concentration parameters of the liquid medium dynamically according to the progression of sepsis.

[0020] Here, the pH value, electrolyte concentration, and inflammatory factor concentration parameters of the liquid medium are dynamically adjusted according to the progression of sepsis. The main purpose is to simulate the dynamically fluctuating pathological microenvironment within the human body. Based on the pathophysiological characteristics of early, middle, and late stages of sepsis, dynamic simulation of the disease progression can be achieved. For example, the pH value, electrolyte solution addition, and inflammatory factor addition can be uniformly fine-tuned in 10-minute cycles. After each parameter adjustment, the mixture is homogenized at a constant temperature of 37°C.

[0021] Specifically, in the early stages of sepsis, the pH is maintained between 7.30 and 7.35, with simultaneous adjustments to sodium and calcium ion concentrations, and the addition of low concentrations of inflammatory factors such as TNF-α (tumor necrosis factor-α) and IL-6 (interleukin-6). In the middle stages of sepsis, the pH, sodium, and calcium ion concentrations are appropriately lowered, while the concentrations of inflammatory factors are increased. In the late stages of sepsis, with significant acidosis, the pH, sodium, and calcium ion concentrations are further lowered, and high concentrations of inflammatory factors are added for infiltration.

[0022] S12. After each parameter adjustment, the hydrated particle size, polydispersity coefficient and Zeta potential value of the magnesium nanoparticles are collected synchronously. The standard deviation and average value of the hydrated particle size within the preset window are obtained. The standard deviation is normalized according to the average value to obtain the time-series fluctuation coefficient of the hydrated particle size. The mean value of the Zeta potential and the potential fluctuation characteristics are obtained according to the Zeta potential value.

[0023] After each round of parameter adjustment, dynamic light scattering and electrophoretic light scattering data are collected using an integrated dynamic light scattering / electrophoretic light scattering particle size potentiometer. Dynamic light scattering data includes, but is not limited to, hydrated particle size and polydispersity index. Electrophoretic light scattering data includes, but is not limited to, Zeta potential values, and potential fluctuation characteristics include potential fluctuation amplitude and potential fluctuation frequency.

[0024] The Zeta potential value at each time point can be calculated in real time based on the particle electrophoretic mobility. At the same time, the potential fluctuation amplitude and fluctuation frequency are continuously recorded to extract the potential temporal fluctuation characteristics, so as to obtain the dynamic change state of charge adsorption, charge neutralization and charge imbalance on the surface of magnesium nanoparticles, and reflect the continuous disturbance effect of ions and inflammatory factors on the particle surface charge in the liquid medium.

[0025] It can continuously collect at least 20 sets of valid hydrated particle size data within a sliding time window, such as 10 minutes, and obtain the hydrated particle size temporal fluctuation coefficient and temporal fluctuation amplitude based on the hydrated particle size data. The hydrated particle size temporal fluctuation amplitude is used to characterize the degree of absolute fluctuation change in the particle size of magnesium nanoparticles within the window, and can intuitively reflect the intensity of instantaneous particle aggregation and dispersion. Generally, the absolute value of the difference between the maximum hydrated particle size value and the minimum hydrated particle size value within the window can be taken. By using the temporal fluctuation amplitude, short-term particle size mutations that cannot be identified by the conventional hydrated particle size mean can be captured, adapting to the particle micro-aggregation behavior caused by instantaneous disturbances in the pathological environment of sepsis.

[0026] The hydration particle size time-series fluctuation coefficient is mainly used to eliminate the evaluation error caused by the difference in particle size baseline, so as to realize the horizontal comparison of the stability of magnesium nanoparticles with different particle size specifications. It can be defined as: standard deviation of hydration particle size within the window / average value of hydration particle size within the window × 100%.

[0027] S2, obtain the real-time pH value, real-time inflammatory factor concentration and real-time ionic strength of the liquid medium, and obtain the acid-base regulating factor positively correlated with the real-time pH value, the inflammatory regulating factor positively correlated with the real-time inflammatory factor concentration and the ionic regulating factor positively correlated with the real-time ionic strength based on the real-time pH value, the real-time inflammatory factor concentration and the real-time ionic strength respectively.

[0028] As a preferred technical solution, such as Figure 3 As shown, step S2 specifically includes: S21, using the normal steady-state pH value of the human body as a benchmark, obtains the absolute value of the acid-base deviation between the real-time acid-base value and the normal steady-state pH value of the human body. Based on the maximum pathological acid-base deviation in clinical sepsis, the absolute value of acid-base is normalized and fitted to obtain an acid-base regulation factor that is positively correlated with the acid-base value.

[0029] The absolute value of the acid-base deviation between the real-time pH value and the normal steady-state pH value of the human body can be obtained using 7.35 as a benchmark. The maximum pathological acid-base deviation in clinical sepsis can be set empirically, such as 0.25. For example, the acid-base regulation factor = 1.0 + 0.4 × absolute value of acid-base deviation / maximum pathological acid-base deviation in clinical sepsis. When the absolute value of the acid-base deviation is 0, the acid-base regulation factor equals 1.0, representing no acid-base disturbance; the greater the deviation of the real-time pH value from the normal steady-state pH value of the human body, the closer the acid-base regulation factor is to 1.4, the stronger the acidosis disturbance, and correspondingly, the higher the weighting gain of the Zeta potential fluctuation characteristics.

[0030] S22, obtain the real-time concentrations of TNF-α and IL-6 inflammatory factors and normalize them. Then, perform weighted fusion on the normalized real-time concentrations of TNF-α and IL-6 inflammatory factors to obtain inflammatory regulatory factors that are positively correlated with the concentrations of inflammatory factors.

[0031] The upper limit concentration of TNF-α in late-stage sepsis pathology was used to normalize the real-time TNF-α inflammatory factor concentration, and the upper limit concentration of IL-6 in late-stage sepsis pathology was used to normalize the real-time IL-6 inflammatory factor concentration. The normalized real-time TNF-α and IL-6 inflammatory factor concentrations were then weighted and fused to obtain a comprehensive inflammatory stress index. Based on this, the comprehensive inflammatory stress index was normalized and fitted to obtain the inflammatory regulatory factor. For example, the inflammatory regulatory factor = 1 + 0.4 × comprehensive inflammatory stress index.

[0032] Generally speaking, the higher the degree of inflammatory factor overload, the more severe the overall inflammation, and the higher the level of inflammatory regulatory factors. Correspondingly, the weight ratio of particle size fluctuation and charge shift features that are sensitive to inflammation can be adaptively enhanced.

[0033] S23. Obtain the standard ion intensity of plasma. Based on the ion intensity deviation between the real-time ion intensity and the standard ion intensity of plasma, obtain the relative deviation value of ion disorder. Normalize and fit the relative deviation value of ion disorder to obtain the ion regulation factor that is positively correlated with the real-time ion intensity and is used to quantify the stress intensity of sepsis ion disorder on particle aggregation.

[0034] Real-time ionic strength can be calculated based on the molar concentrations of key electrolytes such as sodium and calcium and their corresponding ionic valence states. For example, the ion regulation factor = 1.0 + 0.4 × |(Real-time ionic strength - Plasma standard ionic strength) / Plasma standard ionic strength|. The plasma standard ionic strength can be taken as 0.15 mol / L. Generally, the more significant the ionic disturbance and the more severe the deviation of the ionic strength from the plasma standard ionic strength, the higher the corresponding ion regulation factor. In this case, the weighting of hydrated particle size fluctuation characteristics and PDI dispersion characteristics can be specifically amplified to accurately capture ion bridging-induced particle micro-aggregation behavior.

[0035] For acid-base regulators, inflammation regulators, and ion regulators, a range was defined, such as between 1.0 and 1.4. The larger the regulator, the stronger the corresponding pathological perturbation, and the higher the regulatory weight gain on the stability parameters of the magnesium nanoparticles.

[0036] Thus, this invention obtains the corresponding regulatory factors for each dimension by measuring the real-time pH value, real-time inflammatory factor concentration, and real-time ionic strength of the liquid medium, and performs adaptive weight ratio adjustment, which is beneficial to the accurate acquisition of the subsequent comprehensive stability index.

[0037] S3. Determine the core dominant factor. If the acid-base regulation factor is the core dominant factor, apply weighted gain to the potential fluctuation characteristics. If the ion regulation factor is the core dominant factor, apply weighted gain to the hydration particle size temporal fluctuation coefficient and polydispersity coefficient. If the inflammation regulation factor is the core dominant factor, apply weighted gain to the hydration particle size temporal fluctuation coefficient and the mean Zeta potential.

[0038] As a preferred technical solution, such as Figure 4 As shown, step S3 specifically includes: S31, respectively obtain the absolute deviation of acid-base regulation factor, inflammation regulation factor and ion regulation factor from their respective steady-state benchmark values, and take the regulation factor corresponding to the largest absolute deviation value as the core dominant factor.

[0039] Because the pathological microenvironment of sepsis is characterized by multiple coupled disturbances such as acid-base imbalance, ion imbalance, and inflammatory stress, in order to avoid the confusion of weight gain and ambiguity of operating condition identification caused by the simultaneous action of different regulatory factors, the core dominant factor is first determined based on the absolute value of the deviation of the regulatory factor from its corresponding steady-state benchmark value, and the current dominant operating condition is determined based on the core dominant factor.

[0040] The steady-state baseline value is a preset value, typically set to 1.0. If the absolute deviation of all adjustment factors is less than the effective threshold, such as 0.1, the pathological microenvironment is considered to be in a steady state, and its basic weight ratio is not adjusted.

[0041] S32, if the acid-base regulation factor is the core dominant factor, normalize the mapping based on the deviation of the acid-base regulation factor from the corresponding steady-state reference value to obtain the acid-base weight gain coefficient, and perform gain correction on the basic weight of the potential fluctuation characteristics according to the acid-base weight gain coefficient. Then, perform global normalization on the weights of the hydration particle size time-series fluctuation coefficient, polydispersity coefficient and the mean of the Zeta potential.

[0042] For example, the acid-base weighting gain coefficient can be defined as: 1 + 0.35 × ((acid-base adjustment factor - 1.0) / 0.4) 2This approach aims to achieve a pathological fit where mild acidosis shows a small gain and severe acidosis shows a significant enhancement. The weighting of potential fluctuation characteristics is the product of the corresponding base weight and the acid-base weight gain coefficient. When there is no acid-base disturbance, the acid-base adjustment factor = 1.0, the acid-base weight gain coefficient = 1.0, and there is no weight gain. As acidosis worsens, the acid-base adjustment factor increases, and the acid-base weight gain coefficient exhibits a quadratic nonlinear growth. This enhances the evaluation priority of charge fluctuation characteristics under severe acid-base imbalance to match the rapid instability of particle charge in late-stage sepsis.

[0043] The basic weight for the potential fluctuation characteristics can be set to 0.15. After adjusting the basic weight for the potential fluctuation characteristics according to the acid-base weight gain coefficient, the weight ratios of the hydration particle size temporal fluctuation coefficient, polydispersity coefficient, and mean Zeta potential are compressed and normalized accordingly to ensure that the sum of the global weights is always 1.0.

[0044] Generally, septic acidosis exhibits threshold-like and nonlinear destructive effects on magnesium nanoparticles. Mild acid-base shifts only induce weak charge fluctuations, while severe acidosis rapidly induces particle charge neutralization, interfacial instability, and irreversible aggregation. Here, a quadratic nonlinear gain is used to retain multi-parameter comprehensive assessment capabilities in mild cases and actively increase the weight of Zeta potential fluctuation characteristics and suppress interference from particle size steady-state parameters in severe cases, thereby improving the predictive sensitivity of acid-base pathological instability.

[0045] S33, if the ion regulation factor is the core dominant factor, normalize the mapping based on the absolute value of the deviation of the ion regulation factor from the corresponding steady-state reference value to obtain the ion weight gain coefficient, and then perform gain correction on the basic weights of the hydration particle size temporal fluctuation coefficient and the polydispersity coefficient according to the ion weight gain coefficient. Then, perform global normalization on the weights of the potential fluctuation characteristics and the mean of the Zeta potential.

[0046] For example, the ion weight gain coefficient = 1 + 0.38 × ((ion adjustment factor - 1.0) / 0.4) 2 The weighting percentage of the hydration particle size temporal fluctuation coefficient = corresponding basic weight × ion weight gain coefficient × 1.05, and the weighting percentage of the polydispersity coefficient = corresponding basic weight × ion weight gain coefficient. When there is no ion disturbance, the ion adjustment factor = 1.0, the ion weight gain coefficient = 1.0, and there is no weight gain. As electrolyte disturbance intensifies and the ion bridging effect strengthens, the ion weight gain coefficient increases rapidly in a quadratic nonlinear manner, which can enhance the evaluation priority of ion perturbation sensitive parameters.

[0047] The basic weights for the hydration particle size temporal fluctuation coefficient and the polydispersity index can be set to 0.35 and 0.25, respectively. The parameter 1.05 in the weight ratio of the hydration particle size temporal fluctuation coefficient is used to highlight the high sensitivity of the hydration particle size temporal fluctuation coefficient to transient ion perturbations, so as to capture early ionic micro-agglomeration signals that are difficult to identify by traditional models and avoid the lag of evaluation by a single PDI (Polydispersity Index) parameter.

[0048] The instability of magnesium nanoparticles induced by ion disorder in sepsis exhibits a temporal sequence of initial fluctuations, followed by aggregation, and then dispersion and failure. This invention utilizes a quadratic nonlinear gain ion weighting gain coefficient to not only enhance the ability to identify aggregation risks under severe ion disorder but also to capture early latent instability through highly sensitive particle size fluctuation parameters.

[0049] S34. If the inflammatory regulatory factor is the core dominant factor, normalize the mapping based on the deviation of the inflammatory regulatory factor from the corresponding steady-state benchmark value to obtain the inflammatory weight gain coefficient. Then, according to the inflammatory weight gain coefficient, perform gain correction on the basic weights of the hydration particle size time-series fluctuation coefficient and the mean of the Zeta potential. Finally, perform global normalization on the weights of the potential fluctuation characteristics and the polydispersity coefficient.

[0050] For example, the inflammation weight gain coefficient = 1 + 0.32 × ((ion regulation factor - 1.0) / 0.4) 2 The weighting percentage of the temporal fluctuation coefficient of hydrated particle size = corresponding base weight × inflammation weight gain coefficient, and the weighting percentage of the mean Zeta potential = corresponding base weight × inflammation weight gain coefficient × 1.08. When there is no inflammatory compression, the inflammation regulatory factor = 1.0, the inflammation weight gain coefficient = 1.0, and there is no weight gain. As the concentration of inflammatory factors increases and the adsorption effect at the particle interface accumulates and strengthens, the inflammation weight gain coefficient increases non-linearly in a quadratic manner. This can specifically enhance the evaluation priority of the two inflammation-sensitive parameters and adapt to the identification needs of chronic persistent instability.

[0051] The time-series fluctuation coefficient of hydrated particle size and the basic weights corresponding to the mean Zeta potential can be set to 0.35 and 0.25, respectively. The parameter 1.08 in the weight ratio of the mean Zeta potential is used to highlight the high sensitivity of the mean Zeta potential to the early charge neutralization behavior of inflammatory adsorption, thereby achieving early identification and early warning of inflammatory instability.

[0052] S4. The polydispersity index, the temporal fluctuation coefficient of hydrated particle size, the mean value of Zeta potential, and the potential fluctuation characteristics are weighted and fused to obtain a comprehensive stability index for representing the colloidal stability of magnesium nanoparticles. The stability of magnesium nanoparticles is monitored based on the comprehensive stability index.

[0053] As a preferred technical solution, such as Figure 5 As shown, step S4 specifically includes: S41, the polydispersity index, the temporal fluctuation coefficient of hydrated particle size, the mean Zeta potential, and the potential fluctuation characteristics are weighted and fused to obtain the initial stability index.

[0054] Here, the initial stability index characterizes the overall steady-state colloidal properties of the magnesium nanoparticles.

[0055] S42, obtain the temporal fluctuation amplitude and average particle size of hydrated particles within a preset sliding time window, and obtain the fluctuation amplitude correction coefficient based on the temporal fluctuation amplitude and average particle size.

[0056] The fluctuation amplitude correction coefficient is mainly used to quantify the degree of disturbance caused by instantaneous aggregation and sudden instability. It can be defined as: 1 - √(time-series fluctuation amplitude / average particle size). The larger the instantaneous fluctuation amplitude of particle size and the more intense the micro-aggregation, the smaller the fluctuation amplitude correction coefficient, and the stronger its negative correction effect on the overall colloidal stability.

[0057] S43. The initial stability index is corrected according to the fluctuation amplitude correction coefficient and then normalized to obtain the comprehensive stability index used to represent the stability of the nano-magnesium particle colloid.

[0058] When the particles are in a stable state, the fluctuation amplitude is extremely small, and the fluctuation amplitude correction coefficient approaches 1. At this time, the steady-state parameter evaluation is the main method to ensure the evaluation accuracy. When the pathological environment is suddenly disturbed by acid, alkali, ion, or inflammation, and the particles undergo instantaneous micro-aggregation or sudden changes in particle size, the temporal fluctuation amplitude will increase significantly. At this time, the fluctuation amplitude correction coefficient will decrease rapidly, which can automatically lower the overall stability index and improve the sensitivity and anticipation of the instability prediction of nano-magnesium particles in the pathological scenario of sepsis.

[0059] The modified initial stability index (i.e., the product of the fluctuation amplitude correction coefficient and the initial stability index) can be normalized based on the maximum and minimum value normalization method, so that the larger the value of the comprehensive stability index, the better the stability of the nano-magnesium particle colloid.

[0060] In summary, the proposed method for dynamic monitoring of magnesium nanoparticles for sepsis constructs a dynamic pathological simulation liquid medium for sepsis, collects characteristic parameters of magnesium nanoparticles and the liquid medium in real time, and weights and gains regulatory factors based on the determination of core dominant factors. On this basis, the polydispersity coefficient, the temporal fluctuation coefficient of hydrated particle size, the mean Zeta potential, and the potential fluctuation characteristics are weighted and fused to obtain an accurate comprehensive stability index for representing the colloidal stability of magnesium nanoparticles. This index can respond to dynamic disturbances in the pathological environment, accurately capture subtle instability characteristics of magnesium nanoparticles, and thus improve the accuracy of colloidal stability assessment under complex pathological conditions.

[0061] In one embodiment, such as Figure 6 As shown, the method for dynamic monitoring of magnesium nanoparticles for sepsis further includes the following steps: S5. Obtain historical parameters of the liquid medium, historical parameters of the magnesium nanoparticles, and the corresponding historical comprehensive stability index under different sepsis disease progressions, and construct a sample dataset using the historical parameters of the liquid medium, historical parameters of the magnesium nanoparticles, and the corresponding historical comprehensive stability index.

[0062] The sample dataset includes operational scenarios depicting the progression of sepsis at three stages: early, middle, and late. In practical applications, continuous, multi-hour time-series monitoring of the liquid medium and the characteristic parameters of the magnesium nanoparticles under each operational condition can be performed to obtain historical parameters of the liquid medium, historical parameters of the magnesium nanoparticles, and the corresponding historical comprehensive stability index.

[0063] S6. Label each sample in the sample dataset with an instability cause label to obtain the sample training set. The instability cause labels include charge imbalance type aggregation label, ion-induced aggregation label, and inflammation adsorption type instability label.

[0064] Manual labeling can be used to label samples based on the dominant type of working condition disturbance and the microscopic evolution characteristics of particles. These labels can be divided into three categories: charge imbalance type aggregation labels, ion-induced aggregation labels, and inflammation adsorption type instability labels, in order to construct a training set of samples with pathological cause labels.

[0065] S7. Construct a neural network model. Train the neural network model using a sample training set to obtain the current parameters of the liquid medium, the current parameters of the nano-magnesium particles, and the corresponding current comprehensive stability index. Input these parameters into the trained neural network model to obtain the current dominant instability cause.

[0066] The neural network model consists of, in sequence, an LSTM temporal coding layer, a pathological differential feature focusing mapping layer, a feature concatenation and dimensionality compression layer, a feature normalization and discrimination weight matrix layer, and a Softmax multi-class output layer.

[0067] First, the parameters of the input LSTM temporal coding layer are concatenated to obtain multi-dimensional temporal input features. Then, based on the LSTM temporal coding layer, long-short-term dependency feature encoding is performed on the input features to extract the slow drift features and transient mutation features of the stability of magnesium nanoparticles. The pathological differential feature focusing mapping layer has three sets of dedicated pathological feature convolutional mapping kernels, which respectively strengthen the sensitive features corresponding to three types of instability inducing factors. These kernels are mainly used to assign differentiated attention weights to different input features, adaptively strengthening the weights of potential fluctuation features sensitive to acid-base perturbations, particle size fluctuation features sensitive to ion perturbations, and mean Zeta potential features sensitive to inflammation perturbations, thereby suppressing ineffective noise features and achieving feature decoupling in multi-perturbation coupled scenarios. This pathological differential feature focusing mapping layer mainly uses 1×1 lightweight convolution to redistribute channel dimension weights, outputting three decoupled feature vectors, solving the feature aliasing problem caused by the superposition of multiple pathological perturbations.

[0068] The feature concatenation and dimensionality compression layer concatenates the three decoupled features through channels, and then passes them through two fully connected layers to complete dimensionality compression and feature purification, outputting refined features: the first FC1 layer uses ReLU activation and Dropout to filter out temporal noise and redundant features, and the second FC2 layer uses Tanh activation to enhance the ability to represent positive and negative shifts in pathological features, adapting to the distribution of pathological perturbation-shifted data. The feature normalization and discrimination weight matrix layer assigns independent feature discrimination thresholds and response weights to the three types of instability causes, performing specific scoring mapping on the refined features output by the feature concatenation and dimensionality compression layer, outputting a 3-dimensional original logical vector, corresponding to the matching scores of the three types of instability causes. The Softmax multi-classification output layer is mainly used to normalize the probability of the 3-dimensional original logical vector, outputting a confidence probability distribution with a sum of 1: P=[P1,P2,P3], which correspond to the confidence levels of charge imbalance-type aggregation, ion-induced aggregation, and inflammation adsorption-type instability, respectively. The category with the highest probability is taken as the dominant instability cause in the current frame.

[0069] After obtaining the dominant instability cause through forward inference using a neural network model, the Grad-weighted Class Activation Map (Grad-CAM) can be used for backward backtracking to obtain the contribution weights of each dimension of the input features to the final classification result, and output the real-time feature thermal weight distribution: if the potential fluctuation feature dimension has the highest weight, it is determined that acid and base are dominant; if the hydration particle size fluctuation amplitude and fluctuation coefficient dimensions have the highest weight, it is determined that ion is dominant; if the Zeta potential mean feature dimension has the highest weight, it is determined that inflammation is dominant.

[0070] Thus, the neural network model, by setting up a pathological differential feature focusing mapping layer, a feature normalization and discrimination weight matrix layer, and a gradient-weighted class activation mapping (Grad-CAM), has specifically designed differentiated temporal features for the three types of sepsis instability induction mechanisms, which can improve the accuracy of instability induction judgment.

[0071] For different instability causes, corresponding optimization strategies for the surface modification of magnesium nanoparticles can be matched. For example, for charge imbalance-type aggregation instability causes, zwitterionic surface modification schemes can be provided; for high salt ion-induced aggregation instability causes, hydrophilic polymer grafting modification schemes can be provided.

[0072] Existing nanoparticle stability monitoring models often only predict particle size change trends and cannot identify instability triggers. The surface modification strategies for nanoparticle samples they output have weak pathological adaptability, making it difficult to meet the precise development and clinical quality control needs of sepsis-specific nanoparticle formulations. Compared to existing technologies, this invention, by tracing the instability triggers and core dominant operating conditions, outputs corresponding nanoparticle surface modification optimization strategies based on different pathological instability trigger types. This can improve the efficiency of stability optimization for nanoparticles in sepsis scenarios.

[0073] An embodiment of the present invention also provides a dynamic monitoring system for magnesium nanoparticles in sepsis, for implementing the aforementioned dynamic monitoring method for magnesium nanoparticles, comprising a magnesium particle parameter acquisition module, a medium parameter acquisition module, a feature weight gain module, and a stability monitoring module.

[0074] The magnesium particle parameter acquisition module is used to collect the hydration particle size, polydispersity coefficient, and Zeta potential value of nano-magnesium particles in the liquid medium of sepsis dynamic pathological simulation. The hydration particle size time-series fluctuation coefficient is obtained based on the hydration particle size, and the mean Zeta potential and potential fluctuation characteristics are obtained based on the Zeta potential value.

[0075] The construction of a liquid medium simulating the dynamic pathology of sepsis can be determined by adding lipopolysaccharide (LPS) to a culture medium containing a simulated liquid medium or by purifying inflammatory mediators to establish an in vitro sepsis model. For example, treating HK-2 cells or H9C2 cells (i.e., rat cardiomyocytes) with LPS can induce activation of inflammatory pathways and apoptosis.

[0076] The medium parameter acquisition module is used to acquire the real-time pH value, real-time inflammatory factor concentration, and real-time ionic strength of the liquid medium. Based on the real-time pH value, real-time inflammatory factor concentration, and real-time ionic strength, it acquires acid-base regulating factors positively correlated with the real-time pH value, inflammatory regulating factors positively correlated with the real-time inflammatory factor concentration, and ionic regulating factors positively correlated with the real-time ionic strength, respectively.

[0077] The feature weighting and gain module is used to determine the core dominant factor. If the acid-base regulation factor is the core dominant factor, the potential fluctuation feature is weighted and gained. If the ion regulation factor is the core dominant factor, the hydration particle size temporal fluctuation coefficient and polydispersity coefficient are weighted and gained. If the inflammation regulation factor is the core dominant factor, the hydration particle size temporal fluctuation coefficient and the mean Zeta potential are weighted and gained. The stability monitoring module is used to perform weighted fusion of the polydispersity coefficient, the hydration particle size temporal fluctuation coefficient, the mean Zeta potential, and the potential fluctuation feature to obtain a comprehensive stability index that represents the colloidal stability of the magnesium nanoparticles. The stability of the magnesium nanoparticles is monitored based on the comprehensive stability index.

[0078] Specifically, the feature weight gain module includes a dominant factor determination unit, an acid-base weight gain unit, an inflammation weight gain unit, and an ion weight gain unit.

[0079] The dominant factor determination unit is used to obtain the absolute deviation of acid-base regulation factor, inflammation regulation factor and ion regulation factor from their respective steady-state reference values, and the regulation factor corresponding to the largest absolute deviation value is taken as the core dominant factor; the acid-base weight gain unit is used to perform normalization mapping based on the deviation of acid-base regulation factor from its corresponding steady-state reference value when acid-base regulation factor is the core dominant factor, to obtain acid-base weight gain coefficient, and to perform gain correction on the basic weight of potential fluctuation characteristics based on acid-base weight gain coefficient, and then to perform global normalization processing on the weights of hydrated particle size temporal fluctuation coefficient, polydispersity coefficient and mean Zeta potential.

[0080] The ion weight gain unit is used to perform normalization mapping based on the deviation of the ion regulation factor from the corresponding steady-state reference value when the ion regulation factor is the core dominant factor, to obtain the ion weight gain coefficient, and then to perform gain correction on the basic weights of the hydration particle size temporal fluctuation coefficient and the polydispersity coefficient according to the ion weight gain coefficient. Finally, the weights of the potential fluctuation characteristics and the mean of the Zeta potential are globally normalized.

[0081] The inflammation weight gain unit is used to normalize the deviation of the inflammation regulatory factor from the corresponding steady-state benchmark value when the inflammation regulatory factor is the core dominant factor, to obtain the inflammation weight gain coefficient. Then, the basic weights of the hydration particle size temporal fluctuation coefficient and the mean of the Zeta potential are adjusted according to the inflammation weight gain coefficient. Finally, the weights of the potential fluctuation characteristics and the polydispersity coefficient are globally normalized.

[0082] The proposed nano-magnesium particle dynamic monitoring system for sepsis constructs a dynamic pathological simulation liquid medium for sepsis, collects characteristic parameters of nano-magnesium particles and the liquid medium in real time, and weights and gains regulatory factors based on the determination of core dominant factors. On this basis, it performs weighted fusion of polydispersity coefficient, hydration particle size temporal fluctuation coefficient, mean Zeta potential, and potential fluctuation characteristics to obtain an accurate comprehensive stability index for representing the colloidal stability of nano-magnesium particles. It can respond to dynamic disturbances in the pathological environment, accurately capture subtle instability characteristics of nano-magnesium particles, and thus improve the accuracy of colloidal stability assessment under complex pathological conditions.

[0083] In one embodiment, the nanomagnesium particle dynamic monitoring system for sepsis further includes a sample training set acquisition module and a neural network model.

[0084] The sample training set acquisition module is used to acquire historical parameters of the liquid medium, historical parameters of the magnesium nanoparticles, and the corresponding historical comprehensive stability index under different sepsis disease progressions. A sample dataset is constructed using the historical parameters of the liquid medium, historical parameters of the magnesium nanoparticles, and the corresponding historical comprehensive stability index. Each sample in the sample dataset is labeled with an instability inducing factor label to obtain the sample training set. The instability inducing factor labels include charge imbalance type aggregation label, ion-induced aggregation label, and inflammation adsorption type instability label. The neural network model is used to train using the sample training set. It takes the current parameters of the liquid medium, the current parameters of the magnesium nanoparticles, and the corresponding current comprehensive stability index as input and outputs the currently dominant instability inducing factor.

[0085] The neural network model consists of, in sequence, an LSTM temporal coding layer, a pathological differential feature focusing mapping layer, a feature concatenation and dimensionality compression layer, a feature normalization and discrimination weight matrix layer, and a Softmax multi-class output layer.

[0086] First, the parameters of the input LSTM temporal coding layer are concatenated to obtain multi-dimensional temporal input features. Then, based on the LSTM temporal coding layer, long-short-term dependency feature encoding is performed on the input features to extract the slow drift features and transient mutation features of the stability of magnesium nanoparticles. The pathological differential feature focusing mapping layer has three sets of dedicated pathological feature convolutional mapping kernels, which respectively strengthen the sensitive features corresponding to three types of instability inducing factors. These kernels are mainly used to assign differentiated attention weights to different input features, adaptively strengthening the weights of potential fluctuation features sensitive to acid-base perturbations, particle size fluctuation features sensitive to ion perturbations, and mean Zeta potential features sensitive to inflammation perturbations, thereby suppressing ineffective noise features and achieving feature decoupling in multi-perturbation coupled scenarios. This pathological differential feature focusing mapping layer mainly uses 1×1 lightweight convolution to redistribute channel dimension weights, outputting three decoupled feature vectors, solving the feature aliasing problem caused by the superposition of multiple pathological perturbations.

[0087] The feature concatenation and dimensionality compression layer concatenates the three decoupled features through channels, and then passes them through two fully connected layers to complete dimensionality compression and feature purification, outputting refined features: the first FC1 layer uses ReLU activation and Dropout to filter out temporal noise and redundant features, and the second FC2 layer uses Tanh activation to enhance the ability to represent positive and negative shifts in pathological features, adapting to the distribution of pathological perturbation-shifted data. The feature normalization and discrimination weight matrix layer assigns independent feature discrimination thresholds and response weights to the three types of instability causes, performing specific scoring mapping on the refined features output by the feature concatenation and dimensionality compression layer, outputting a 3-dimensional original logical vector, corresponding to the matching scores of the three types of instability causes. The Softmax multi-classification output layer is mainly used to normalize the probability of the 3-dimensional original logical vector, outputting a confidence probability distribution with a sum of 1: P=[P1,P2,P3], which correspond to the confidence levels of charge imbalance-type aggregation, ion-induced aggregation, and inflammation adsorption-type instability, respectively. The category with the highest probability is taken as the dominant instability cause in the current frame.

[0088] After obtaining the dominant instability cause through forward inference using a neural network model, the Grad-weighted Class Activation Map (Grad-CAM) can be used for backward backtracking to obtain the contribution weights of each dimension of the input features to the final classification result, and output the real-time feature thermal weight distribution: if the potential fluctuation feature dimension has the highest weight, it is determined that acid and base are dominant; if the hydration particle size fluctuation amplitude and fluctuation coefficient dimensions have the highest weight, it is determined that ion is dominant; if the Zeta potential mean feature dimension has the highest weight, it is determined that inflammation is dominant.

[0089] Thus, the neural network model, by setting up a pathological differential feature focusing mapping layer, a feature normalization and discrimination weight matrix layer, and a gradient-weighted class activation mapping (Grad-CAM), has specifically designed differentiated temporal features for the three types of sepsis instability induction mechanisms, which can improve the accuracy of instability induction judgment.

[0090] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for dynamic monitoring of magnesium nanoparticles for sepsis, characterized in that, Includes the following steps: S1. Construct a dynamic pathological simulation liquid medium for sepsis and inject the sample of magnesium nanoparticles into the liquid medium. Collect the hydration particle size, polydispersity coefficient and Zeta potential value of the magnesium nanoparticles. Obtain the hydration particle size time-series fluctuation coefficient based on the hydration particle size and obtain the mean Zeta potential and potential fluctuation characteristics based on the Zeta potential value. S2, obtain the real-time pH value, real-time inflammatory factor concentration and real-time ionic strength of the liquid medium, and obtain the acid-base regulating factor positively correlated with the real-time pH value, the inflammatory regulating factor positively correlated with the real-time inflammatory factor concentration and the ionic regulating factor positively correlated with the real-time ionic strength according to the real-time pH value, the real-time inflammatory factor concentration and the real-time ionic strength respectively. S3, determine the core dominant factor. If the acid-base regulation factor is the core dominant factor, apply weighted gain to the potential fluctuation characteristics. If the ion regulation factor is the core dominant factor, apply weighted gain to the hydration particle size temporal fluctuation coefficient and polydispersity coefficient. If the inflammation regulation factor is the core dominant factor, apply weighted gain to the hydration particle size temporal fluctuation coefficient and the mean Zeta potential. S4. The polydispersity index, the temporal fluctuation coefficient of hydrated particle size, the mean value of Zeta potential, and the potential fluctuation characteristics are weighted and fused to obtain a comprehensive stability index for representing the colloidal stability of magnesium nanoparticles. The stability of magnesium nanoparticles is monitored based on the comprehensive stability index.

2. The method for dynamic monitoring of magnesium nanoparticles for sepsis as described in claim 1, characterized in that, Step S1 specifically includes: S11, construct a dynamic pathological simulation liquid medium for sepsis and inject the sample of magnesium nanoparticles into the liquid medium, and dynamically adjust the pH value, electrolyte concentration and inflammatory factor concentration parameters of the liquid medium according to the progression of sepsis. S12. After each parameter adjustment, the hydration particle size, polydispersity coefficient and Zeta potential value of the magnesium nanoparticles are collected synchronously. The standard deviation and average value of the hydration particle size within the preset window are obtained. The standard deviation is normalized according to the average value to obtain the time-series fluctuation coefficient of the hydration particle size. The mean value of the Zeta potential and the potential fluctuation characteristics are obtained according to the Zeta potential value. Among them, the characteristics of potential fluctuation include the amplitude of potential fluctuation and the frequency of potential fluctuation.

3. The method for dynamic monitoring of magnesium nanoparticles for sepsis as described in claim 2, characterized in that, Step S2 specifically includes: S21, based on the normal steady-state pH value of the human body, obtain the absolute value of the acid-base deviation between the real-time acid-base value and the normal steady-state pH value of the human body, and normalize and fit the absolute value of acid-base based on the maximum pathological acid-base deviation in clinical sepsis to obtain the acid-base regulation factor that is positively correlated with the acid-base value. S22, obtain the real-time concentrations of TNF-α and IL-6 inflammatory factors and normalize them. Then, perform weighted fusion on the normalized real-time concentrations of TNF-α and IL-6 inflammatory factors to obtain inflammatory regulatory factors that are positively correlated with the concentrations of inflammatory factors. S23. Obtain the standard ion intensity of plasma. Based on the ion intensity deviation between the real-time ion intensity and the standard ion intensity of plasma, obtain the relative deviation value of ion disorder. Normalize and fit the relative deviation value of ion disorder to obtain the ion regulation factor that is positively correlated with the real-time ion intensity and is used to quantify the stress intensity of sepsis ion disorder on particle aggregation.

4. The method for dynamic monitoring of magnesium nanoparticles for sepsis as described in claim 3, characterized in that, Step S4 specifically includes: S41, the polydispersity index, the temporal fluctuation coefficient of hydrated particle size, the mean value of Zeta potential, and the potential fluctuation characteristics are weighted and fused to obtain the initial stability index; S42, obtain the temporal fluctuation amplitude and average particle size of hydrated particles within a preset sliding time window, and obtain the fluctuation amplitude correction coefficient based on the temporal fluctuation amplitude and average particle size. S43. The initial stability index is corrected according to the fluctuation amplitude correction coefficient and then normalized to obtain the comprehensive stability index used to represent the stability of the nano-magnesium particle colloid.

5. The method for dynamic monitoring of magnesium nanoparticles for sepsis as described in claim 4, characterized in that, It also includes the following steps: S5. Obtain historical parameters of liquid media, historical parameters of magnesium nanoparticles, and corresponding historical comprehensive stability indices under different sepsis disease progression, and construct a sample dataset using historical parameters of liquid media, historical parameters of magnesium nanoparticles, and corresponding historical comprehensive stability indices. S6. Label each sample in the sample dataset with an instability cause label to obtain the sample training set. The instability cause labels include charge imbalance type aggregation label, ion-induced aggregation label and inflammation adsorption type instability label. S7. Construct a neural network model. Train the neural network model using a sample training set to obtain the current parameters of the liquid medium, the current parameters of the nano-magnesium particles, and the corresponding current comprehensive stability index. Input these parameters into the trained neural network model to obtain the current dominant instability cause.

6. The method for dynamic monitoring of magnesium nanoparticles for sepsis as described in claim 1, characterized in that, Step S3 specifically includes: S31, respectively obtain the absolute deviation of acid-base regulation factor, inflammation regulation factor and ion regulation factor from their respective steady-state benchmark values, and take the regulation factor corresponding to the largest absolute deviation value as the core dominant factor.

7. The method for dynamic monitoring of magnesium nanoparticles for sepsis as described in claim 6, characterized in that, Step S3 specifically also includes: S32, if the acid-base regulation factor is the core dominant factor, normalize the mapping based on the deviation of the acid-base regulation factor from the corresponding steady-state reference value to obtain the acid-base weight gain coefficient, and perform gain correction on the basic weight of the potential fluctuation characteristics according to the acid-base weight gain coefficient. Then, perform global normalization on the weights of the hydration particle size time-series fluctuation coefficient, polydispersity coefficient and the mean of the Zeta potential. S33, if the ion regulation factor is the core dominant factor, normalize the mapping based on the absolute value of the deviation of the ion regulation factor from the corresponding steady-state reference value to obtain the ion weight gain coefficient, and then perform gain correction on the basic weights of the hydration particle size time-series fluctuation coefficient and the polydispersity coefficient according to the ion weight gain coefficient. Then, perform global normalization on the weights of the potential fluctuation characteristics and the mean of the Zeta potential. S34. If the inflammatory regulatory factor is the core dominant factor, normalize the mapping based on the deviation of the inflammatory regulatory factor from the corresponding steady-state benchmark value to obtain the inflammatory weight gain coefficient. Then, according to the inflammatory weight gain coefficient, perform gain correction on the basic weights of the hydration particle size time-series fluctuation coefficient and the mean of the Zeta potential. Finally, perform global normalization on the weights of the potential fluctuation characteristics and the polydispersity coefficient.

8. A dynamic monitoring system for magnesium nanoparticles in sepsis, used to implement the dynamic monitoring method for magnesium nanoparticles as described in any one of claims 1-7, characterized in that, include: The magnesium particle parameter acquisition module is used to collect the hydration particle size, polydispersity coefficient and zeta potential value of nano-magnesium particles in the liquid medium of sepsis dynamic pathological simulation. The module obtains the hydration particle size temporal fluctuation coefficient based on the hydration particle size and the mean zeta potential and potential fluctuation characteristics based on the zeta potential value. The medium parameter acquisition module is used to acquire the real-time pH value, real-time inflammatory factor concentration, and real-time ionic strength of the liquid medium. Based on the real-time pH value, real-time inflammatory factor concentration, and real-time ionic strength, it acquires acid-base regulating factors positively correlated with the real-time pH value, inflammatory regulating factors positively correlated with the real-time inflammatory factor concentration, and ionic regulating factors positively correlated with the real-time ionic strength, respectively. The feature weight gain module is used to determine the core dominant factor. If the acid-base regulation factor is the core dominant factor, the potential fluctuation feature is weighted and gained. If the ion regulation factor is the core dominant factor, the hydration particle size temporal fluctuation coefficient and polydispersity coefficient are weighted and gained. If the inflammation regulation factor is the core dominant factor, the hydration particle size temporal fluctuation coefficient and the mean Zeta potential are weighted and gained. The stability monitoring module is used to perform weighted fusion of polydispersity coefficient, hydration particle size temporal fluctuation coefficient, mean Zeta potential, and potential fluctuation characteristics to obtain a comprehensive stability index that represents the colloidal stability of magnesium nanoparticles. The stability of magnesium nanoparticles is monitored based on the comprehensive stability index.

9. A dynamic monitoring system for magnesium nanoparticles for sepsis as described in claim 8, characterized in that, Also includes: The sample training set acquisition module is used to acquire historical parameters of liquid media, historical parameters of magnesium nanoparticles, and corresponding historical comprehensive stability indices under different sepsis disease progression. The sample dataset is constructed using the historical parameters of liquid media, historical parameters of magnesium nanoparticles, and corresponding historical comprehensive stability indices. Each sample in the sample dataset is labeled with an instability inducing factor label to obtain the sample training set. The instability inducing factor labels include charge imbalance type aggregation label, ion-induced aggregation label, and inflammation adsorption type instability label. The neural network model is trained using a sample training set and takes the current parameters of the liquid medium, the current parameters of the magnesium nanoparticles, and the corresponding current comprehensive stability index as inputs, and outputs the current dominant instability cause.

10. A dynamic monitoring system for magnesium nanoparticles for sepsis as described in claim 9, characterized in that, The feature weight gain module includes: The dominant factor determination unit is used to obtain the absolute value of the deviation of acid-base regulation factor, inflammation regulation factor and ion regulation factor from their respective steady-state benchmark values, and the regulation factor corresponding to the largest absolute value of deviation is the core dominant factor. The acid-base weight gain unit is used to normalize the acid-base weight gain coefficient when the acid-base regulation factor is the core dominant factor. It is then used to normalize the deviation of the acid-base regulation factor from the corresponding steady-state reference value, and to correct the basic weight of the potential fluctuation characteristics based on the acid-base weight gain coefficient. Finally, the weights of the hydration particle size time-series fluctuation coefficient, polydispersity coefficient and the mean of the Zeta potential are global normalized. The ion weight gain unit is used to perform normalization mapping based on the deviation of the ion regulation factor from the corresponding steady-state reference value when the ion regulation factor is the core dominant factor, to obtain the ion weight gain coefficient, and to perform gain correction on the basic weights of the hydration particle size time-series fluctuation coefficient and the polydispersity coefficient according to the ion weight gain coefficient. Then, the weights of the potential fluctuation characteristics and the mean of the Zeta potential are globally normalized. The inflammation weight gain unit is used to normalize the deviation of the inflammation regulatory factor from the corresponding steady-state benchmark value when the inflammation regulatory factor is the core dominant factor, to obtain the inflammation weight gain coefficient. Then, the basic weights of the hydration particle size temporal fluctuation coefficient and the mean of the Zeta potential are adjusted according to the inflammation weight gain coefficient. Finally, the weights of the potential fluctuation characteristics and the polydispersity coefficient are globally normalized.