Infusion control method and device with disinfection function

Abstract

The invention discloses an infusion control method and device with a disinfection function, and belongs to the technical field of medical patient nursing equipment.The infusion control method comprises the steps that physiological recognition information of a patient is obtained, a steady-state physiological rhythm is collected, and a physiological baseline fingerprint is generated; a multi-modal sensor array is used for dynamic exploration, exploration data are obtained, calibration and interpretation are carried out, microenvironment dynamic data are generated and analyzed, a risk evolution track is generated, and then a dynamic disinfection strategy is obtained; the infusion pulse and the physiological baseline fingerprint are synchronized, when the deviation value of the feedback signal and the physiological baseline fingerprint exceeds a deviation threshold value, an intervention instruction is generated, infusion parameters of the infusion device are adjusted, linkage is conducted based on a dynamic disinfection strategy, and supplementary disinfection operation is triggered. Personalized risk prediction and adaptive infusion based on physiological baseline fingerprints are adopted, infusion parameters and supplementary disinfection operation are controlled in a linkage mode, and active, accurate and collaborative safety management of the infusion process can be achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical patient care equipment, in particular to an infusion control method and device with disinfection function. BACKGROUND

[0002] As one of the most basic and most extensive treatment and support means in clinical practice, the safety and effectiveness of intravenous infusion are directly related to the treatment effect and rehabilitation process of patients. The infusion stand, infusion pump and other equipment used to achieve this process are indispensable patient care auxiliary devices in hospital wards and nursing environments, and their core function is to stably deliver liquid drugs, nutrients and other substances to patients according to preset requirements. Therefore, precise control and safety monitoring of the infusion process is an important issue in the field of modern medical care.

[0003] In the related art, the Chinese utility model patent with publication number CN120459444A discloses a multifunctional intelligent mobile infusion instrument and a control method thereof, which includes: a real-time monitoring system including a pulse wave sensor, an oxygen saturation sensor, and a heart rate sensor for real-time collection of vital sign data of a patient; an accurate infusion control pump for accurate control of infusion volume; a liquid heating module for adjusting the temperature of infusion liquid according to patient needs; a nanometer needle dermal puncture device and its control system for ensuring painless infusion; a spectral blood vessel recognition system for automatic recognition of blood vessel position; a wireless communication module for supporting remote data transmission and device control, so that medical staff can monitor the infusion condition of the patient at different locations; and a user interaction interface including a touch screen and voice recognition control for device operation and state display, facilitating user operation and information acquisition.

[0004] For the related art in the above, the inventors believe that although a variety of advanced functional modules are integrated to build a framework of a multifunctional infusion instrument, there are significant limitations in the core infusion safety control logic. First, its monitoring is limited to whole-body vital signs, and lacks deep monitoring of the key local microenvironment of infusion, resulting in a lag in complication discovery. Second, its infusion control is a static "one-size-fits-all" mode, which cannot be adjusted adaptively according to individuals and real-time conditions. Finally, its safety guarantee mechanism is passive, focusing on post-alarm, and lacks proactive risk prediction and full-process dynamic infection protection capabilities. SUMMARY

[0005] To solve the above problems, the present application provides an infusion control method and device with disinfection function, which adopts individualized risk prediction and adaptive infusion based on physiological baseline fingerprint, and controls infusion parameters and supplementary disinfection operations in parallel, to achieve active, accurate and coordinated safety management of the infusion process.

[0006] The above-mentioned object can be achieved by the following scheme: An infusion control method and device with disinfection function, comprising obtaining physiological identification information of a patient and collecting at least one steady physiological rhythm, generating a physiological baseline fingerprint; using a multi-modal sensor array to dynamically probe the target infusion site, obtaining probe data including tissue impedance spectrum, local micro blood perfusion map and surface biophysical feature image, and referring to the physiological baseline fingerprint, the probe data is personalized calibrated and interpreted, and microenvironment dynamic data is generated; based on the microenvironment dynamic data, real-time analysis is carried out, the risk evolution in the infusion process is dynamically predicted, the risk evolution trajectory is generated, and the dynamic disinfection strategy is generated according to the risk evolution trajectory; synchronize the infusion pulse with the physiological baseline fingerprint, control the infusion device to start adaptive infusion, and real-time feedback signal from the multi-modal sensor array is obtained; based on the feedback signal and the physiological baseline fingerprint, the deviation value is compared, when the deviation value exceeds the deviation threshold value determined by the statistical characteristics of the physiological baseline fingerprint, the intervention instruction is generated; according to the intervention instruction, the infusion parameters of the infusion device are adjusted, and the dynamic disinfection strategy is linked to trigger supplementary disinfection operation.

[0007] Optionally, the generating physiological baseline fingerprint comprises: obtaining pulse waveform, skin electrical response and finger temperature fluctuation of the patient as the steady physiological rhythm; coupling the steady physiological rhythm in time domain and frequency domain to extract rhythm morphological parameters representing individual specificity; mapping the rhythm morphological parameters and the physiological identification information to construct a multi-dimensional feature vector and generate a physiological baseline fingerprint.

[0008] Optionally, the obtaining probe data including tissue impedance spectrum, local micro blood perfusion map and surface biophysical feature image comprises: deploying a multi-modal sensor array with high-frequency electrical induction unit, laser Doppler imaging unit and thermal infrared spectrum detection unit around the target infusion area; frequency scanning of subcutaneous tissue of infusion site is carried out through the high-frequency electrical induction unit, impedance response under different frequencies is recorded, and tissue impedance spectrum is constructed; using the laser Doppler imaging unit, real-time capture of blood flow hemodynamic changes of subcutaneous microvessel network of infusion site is carried out, and local micro blood perfusion map is drawn; based on the thermal infrared spectrum detection unit, surface temperature distribution, skin color gradient and micro texture changes of infusion site are captured to generate surface biophysical feature image; time sequence association and spatial registration of the above tissue impedance spectrum, local micro blood perfusion map and surface biophysical feature image are carried out to obtain probe data.

[0009] Optionally, the generating the microenvironment dynamic data comprises: taking the physiological state reflected in the physiological baseline fingerprint as a reference, performing dynamic normalization processing on each dimension feature in the exploration data to obtain normalized multi-dimensional features; and fusing the normalized multi-dimensional features in time series to construct a spatiotemporal evolution graph and generate the microenvironment dynamic data.

[0010] Optionally, the generating the dynamic disinfection strategy according to the risk evolution trajectory comprises: identifying a key feature change vector in the microenvironment dynamic data, the key feature change vector comprising a tissue conductivity gradient, microcirculation perfusion heterogeneity, and entropy increase of a surface temperature field; constructing a trend of a probability of occurrence of fluid extravasation, inflammatory reaction or microbial colonization at the infusion site changing over time by forward extrapolation on the feature change vector to generate the risk evolution trajectory; and positioning the risk evolution trajectory to a corresponding region in a cooperative response structure to generate the dynamic disinfection strategy.

[0011] Optionally, the constructing the cooperative response structure comprises: dividing a plurality of risk response levels according to a dynamic coupling relationship between tissue tolerance and microbial load at the infusion site, and configuring a corresponding disinfectant release mode; combining different disinfectant release modes according to a cooperative relationship between tissue repair rate and microbial clearance efficiency to form a multi-stage linked disinfection strategy sequence; and performing real-time screening and adaptation on the disinfection strategy sequence based on a probability change of fluid extravasation, inflammatory reaction or microbial colonization in the risk evolution trajectory to obtain the cooperative response structure.

[0012] Optionally, the acquiring the feedback signal from the multi-modal sensor array in real time comprises: identifying a current physiological cycle phase of the patient according to a rhythm morphology parameter extracted from the physiological baseline fingerprint, and dynamically matching a timing sequence of infusion pulses based on the physiological cycle phase to obtain an infusion timing framework; taking the infusion timing framework as a basic rhythm, and combining instantaneous changes in tissue compliance and local blood perfusion levels reflected in the microenvironment dynamic data to modulate the intensity and interval of infusion pulses in real time to form an adaptive infusion coordinated with the physiological state; and based on the adaptive infusion process, continuously capturing tissue impedance changes, micro-blood flow fluctuations and dynamic responses of surface biophysical characteristics of the infusion site by the multi-modal sensor array to obtain the feedback signal.

[0013] Optionally, the generating intervention instruction comprises: extracting dynamic features of physiological rhythm from the physiological baseline fingerprint, constructing an expected response template representing a healthy response of the infusion site to external stimulation; performing a quantitative comparison of morphological difference and energy difference between the real-time acquired feedback signal and the expected response template, calculating a deviation value; when the deviation value exceeds a deviation threshold determined by a coefficient of variation of the physiological baseline fingerprint, performing attribution analysis according to specific feature dimensions of the deviation value, and generating an intervention instruction.

[0014] Optionally, the triggering of the supplementary disinfection operation comprises: based on the intervention instruction, performing analysis, converting into specific adjustment parameters of flow rate, pulse frequency or infusion pressure of the infusion device and performing adjustment; matching the risk type implied by the intervention instruction with the dynamic disinfection strategy, and selecting a suitable disinfection mode; based on the disinfection mode, triggering the supplementary disinfection operation by activating photosensitive disinfection medium through light of specific wavelength or promoting targeted penetration of disinfection ions through micro-current.

[0015] Based on the same inventive concept, the present application also provides an infusion control device with disinfection function, comprising: a physiological fingerprint generation module for acquiring physiological identification information of a patient and collecting at least one steady physiological rhythm, and generating a physiological baseline fingerprint; a probe data calibration module for using a multi-modal sensor array to dynamically probe a target infusion site, acquiring probe data including tissue impedance spectrum, local micro blood perfusion map and surface biophysical feature image, and performing individualized calibration and interpretation of the probe data with reference to the physiological baseline fingerprint, and generating microenvironment dynamic data; a risk prediction and disinfection module for performing real-time analysis based on the microenvironment dynamic data, dynamically predicting risk evolution in the infusion process, generating a risk evolution trajectory, and generating a dynamic disinfection strategy according to the risk evolution trajectory; a synchronous infusion control module for synchronizing infusion pulses with the physiological baseline fingerprint, controlling the infusion device to start adaptive infusion, and acquiring real-time feedback signals from the multi-modal sensor array; a real-time intervention judgment module for comparing deviation values based on the feedback signals and the physiological baseline fingerprint, generating an intervention instruction when the deviation value exceeds a deviation threshold determined by statistical features of the physiological baseline fingerprint; and an infusion and disinfection linkage module for adjusting infusion parameters of the infusion device according to the intervention instruction, and performing linkage based on the dynamic disinfection strategy, triggering a supplementary disinfection operation.

[0016] Compared with the prior art, the present application has the following advantages: 1、The present application changes the infusion control from a general, standardized mode to a highly personalized mode by constructing a patient's personalized physiological baseline fingerprint. This method can identify the physiological homeostasis specific to a particular patient and use it as a benchmark to interpret various monitoring data during the infusion process, effectively reducing the interference of individual differences on risk assessment, so that the identification of infusion abnormalities is no longer dependent on universal thresholds, but is based on the individual's own response pattern, thereby improving the sensitivity and accuracy of risk monitoring.

[0017] 2、The present application realizes the prospective prediction and active intervention of infusion risk. This method is not satisfied with passive alarm of the occurred complications, but dynamically predicts the evolution trend of liquid extravasation, inflammatory reaction and other risks through real-time analysis of the dynamic data of the infusion site microenvironment. Based on this prediction, dynamic disinfection strategies can be generated and deployed in advance, moving the safety management checkpoint forward, realizing the change from passive response to active prevention, and winning valuable disposal time for clinical intervention, thereby effectively reducing the incidence and severity of complications.

[0018] 3、The present application constructs an intelligent closed-loop control system that coordinates infusion adjustment and disinfection operation. When it is judged that the infusion process deviates from the healthy response track, the intervention instruction triggered is not a single action, but drives the synchronous linkage of infusion parameter adjustment and supplementary disinfection operation. This cooperative mechanism can simultaneously process the problem from two dimensions of physical source and biological risk, form a comprehensive effect of treating the symptoms and the root cause, realize dynamic, accurate and all-round safety protection of the infusion process, and improve the safety and comfort of patients during infusion treatment.

[0019] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application will be realized and attained by the structure particularly pointed out in the description and claims hereof as well as the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0021] Figure 1 is a flowchart of a infusion control method with disinfection function according to an embodiment of the present application.

[0022] Figure 2 is a deviation comparison diagram of expected response template and real-time feedback signal according to an embodiment of the present application.

[0023] Figure 3 This is a comparison chart of infusion parameter adjustments under the intervention command of this invention embodiment.

[0024] Figure 4 This is a schematic diagram of the structure of an infusion control device with disinfection function according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Reference Figure 1 One embodiment of the present invention proposes an infusion control method with disinfection function, which adopts personalized risk prediction and adaptive infusion based on physiological baseline fingerprint, and links the control of infusion parameters and supplementary disinfection operations to achieve proactive, precise and collaborative safety management of the infusion process.

[0027] The method described in this embodiment specifically includes: Acquire the patient's physiological identification information and collect at least one steady-state physiological rhythm to generate a physiological baseline fingerprint; Optionally, generating a physiological baseline fingerprint includes: The patient's pulse waveform, skin conductance response, and fingertip temperature fluctuations were acquired as the steady-state physiological rhythm. The steady-state physiological rhythms are coupled in the time and frequency domains to extract rhythmic morphological parameters that characterize individual specificity; The rhythmic morphological parameters are associated and mapped with the physiological identification information to construct a multi-dimensional feature vector and generate a physiological baseline fingerprint.

[0028] Specifically, the process starts with the acquisition of physiological signals from the patient. By deploying non-invasive physiological sensors such as photoplethysmography (PPG) sensor, electrodermal activity (EDA) sensor, and high-precision temperature sensor at appropriate locations such as the patient's fingertip or wrist, three steady-state physiological rhythms of the patient in a resting state are synchronously acquired. Specifically, the pulse waveform reflecting the dynamics of the cardiovascular system, the electrodermal response representing the activity of the autonomic nervous system, and the fingertip temperature fluctuation revealing the local metabolic and circulatory state. After obtaining these continuous time series signals, the feature extraction and coupling stage is entered. The goal of this stage is to transform the raw waveform data into quantified parameters with high individual specificity, i.e., rhythm morphology parameters. For the acquired pulse waveform, analysis is performed in both time and frequency domains. Time-domain analysis extracts relevant indicators of heart rate variability, such as the standard deviation of adjacent heart cycle intervals, while frequency-domain analysis calculates the energy ratio of low-frequency components to high-frequency components through Fourier transform or wavelet transform to assess the balance state of sympathetic and parasympathetic nerves. Similarly, for the EDA signal, its baseline level, fluctuation amplitude, and frequency are analyzed, and for the fingertip temperature fluctuation signal, its mean value and rate of change are analyzed. The key is that this method does not analyze each signal in isolation, but rather couples the features to explore the change patterns of EDA amplitude and fingertip temperature when the low-frequency component of the pulse wave increases. This cross-modal correlation constitutes a unique physiological rhythm pattern. Finally, these multi-dimensional rhythm morphology parameters extracted through coupled analysis are associated with the patient's unique physiological identification information. This is accomplished by constructing a multi-dimensional feature vector that contains the time-domain and frequency-domain rhythm morphology parameter sets extracted from the pulse waveform, the rhythm morphology parameter set extracted from the EDA response, and the rhythm morphology parameter set extracted from the fingertip temperature fluctuation. By combining these three parameter set vectors into a higher-dimensional feature vector, a physiological baseline fingerprint that can describe the patient's physiological characteristics in a resting state is generated.

[0029] Exemplarily, a physiological baseline fingerprint is established for a male adult patient A who is ready to receive intravenous infusion. First, when the patient is in a quiet and relaxed resting state, a wearable device integrated with a photoplethysmography sensor, a galvanic skin sensor and a high-precision temperature sensor is worn on his left wrist. Five minutes of physiological signals are continuously collected to obtain three sets of steady-state physiological rhythm data. Specifically, the acquired pulse waveform data shows that his average heart rate is 65 beats per minute, and the standard deviation of the interval between adjacent heart cycles is 55 milliseconds; the acquired galvanic skin response data shows that his skin conductance baseline level is 4.5 microsiemens; the acquired finger temperature fluctuation data shows that his average finger temperature is 35.2 degrees Celsius. Subsequently, time-domain and frequency-domain feature coupling analysis is performed on these data to identify the specific pattern of patient A, i.e., when the proportion of low-frequency component energy of his pulse wave increases to 0.4, the amplitude of his galvanic skin response will increase by 0.3 microsiemens within 2 seconds, and the finger temperature will decrease by 0.1 degrees Celsius. Finally, these quantified rhythm morphological parameters, including average heart rate, heart rate variability indicators, galvanic skin baseline and response amplitude, finger temperature mean value, and parameters of the above coupling pattern, are associated and mapped with the unique identity information of patient A to construct a multi-dimensional feature vector, i.e., the physiological baseline fingerprint exclusive to patient A. By collecting and coupling analyzing multiple steady-state physiological rhythms, a solid and high-precision foundation is provided for subsequent personalized calibration and abnormality judgment.

[0030] Using a multi-modal sensor array to dynamically explore the target infusion site, acquiring exploration data including tissue impedance spectrum, local micro-blood flow perfusion map and surface biophysical feature image, and referring to the physiological baseline fingerprint to perform personalized calibration and interpretation on the exploration data, generating micro-environment dynamic data; Optionally, the acquisition of exploration data including tissue impedance spectrum, local micro-blood flow perfusion map and surface biophysical feature image comprises: Around the target infusion area, a multi-modal sensor array with high-frequency electrical induction unit, laser Doppler imaging unit and thermal infrared spectral detection unit is deployed; By the high-frequency electrical induction unit, the subcutaneous tissue of the infusion site is frequency-scanned to record the impedance response at different frequencies, and the tissue impedance spectrum is constructed; Using the laser Doppler imaging unit, the hemodynamic changes of the subcutaneous microvascular network of the infusion site are captured in real time to draw a local micro-blood flow perfusion map; Based on the thermal infrared spectral detection unit, the surface temperature distribution, skin color gradient and micro-texture changes of the infusion site are captured to generate a surface biophysical feature image; The above tissue impedance spectrum, local micro-blood flow perfusion map and surface biophysical feature image are time-series associated and spatially registered to obtain exploration data.

[0031] Specifically, first, a flexible patch-type multi-modal sensor array integrated with multiple sensing units is deployed on the skin surface around the target infusion area, such as the venipuncture point on the patient's forearm. After deployment, each functional unit in the array starts working synchronously. First, the high-frequency electrical sensing unit built-in the array starts to probe the subcutaneous tissue of the infusion site. It usually contains at least four groups of microelectrodes, two of which are used to apply a low-intensity, frequency-in-the-kilohertz-to-megahertz range scanning sinusoidal wave excitation current, and the other two are used to measure the voltage response across the tissue. By calculating the voltage-current relationship at different frequencies, the complex impedance value is obtained. Plotting a series of complex impedance values recorded in the entire frequency scanning range will build an impedance spectrum that reflects the composition of the tissue, the integrity of the cell structure, and the content of the interstitial fluid. Second, the laser Doppler imaging unit in the array emits a low-power monochromatic coherent laser beam to the same target area. When the laser photons collide with the red blood cells flowing in the subcutaneous capillary network, the scattered light will produce a Doppler shift, and the size of the shift is proportional to the velocity of the red blood cells. The photosensitive element array in the laser Doppler imaging unit is responsible for receiving these scattered lights and calculating the average frequency shift of each pixel point in real time, and then converting it into blood perfusion information. By two-dimensional scanning imaging of the entire field of view, a local micro blood perfusion map reflecting the local capillary blood flow velocity and density can be drawn in real time. Third, the thermal infrared spectral detection unit integrated in the array performs non-contact scanning on the target area. By capturing the infrared radiation generated by the skin surface due to metabolism and blood flow changes, a high-resolution surface temperature distribution map is generated. At the same time, by analyzing the reflection and absorption characteristics of specific waveband infrared spectra, subtle changes in skin color, i.e., skin color gradient, can be analyzed, which is related to hemoglobin concentration and oxygen saturation. The high-resolution imaging mode can also capture the changes in skin microtexture caused by tissue edema or dehydration. These three pieces of information together constitute the surface biophysical feature image. Finally, the three sets of heterogeneous data collected in parallel are time-series correlated and spatially registered. Time-series correlation ensures that at any time, the impedance spectrum, the local micro blood perfusion map, and the surface biophysical feature image all reflect the same physiological state at that instant. Spatial registration, through a pre-set calibration or image registration algorithm, maps the three sets of data into the same physical coordinate system, so that each point in the data can accurately correspond to the same micro-unit in the infusion area. Finally, a structured exploration data containing electrical, hemodynamic, and surface physical multi-dimensional information is output.

[0032] Exemplarily, around the selected forearm cephalic vein puncture point of the patient A, a flexible patch-like multi-modal sensor array is deployed. After the array is closely attached to the skin, each sensing unit is synchronously activated to acquire the exploration data. Firstly, the high-frequency electrical sensing unit in the array starts to work, and through its microelectrode array, a sinusoidal wave excitation current with a frequency sweeping from 1 kHz to 1 MHz is applied to the subcutaneous tissue, and the voltage response is measured. It is recorded that at a frequency of 50 kHz, the tissue impedance value is 520 ohms, and at a frequency of 500 kHz, the impedance value is 280 ohms. The response data of the entire frequency band is plotted into a curve to form the tissue impedance spectrum. Secondly, the laser Doppler imaging unit in the array emits a laser with a wavelength of 780 nm to the target area, and captures the Doppler frequency shift generated by the movement of red blood cells in the subcutaneous microvascular network in real time. After calculation, a local micro blood perfusion map with a resolution of 128x128 pixels is generated, and the average blood perfusion value in the map is 75 perfusion units. Thirdly, the thermal infrared spectrum detection unit integrated in the array scans the same area and generates a surface temperature distribution map, which shows that the average skin temperature is 34.1 degrees Celsius, and the skin micro-texture state is captured, which together constitute the surface biophysical feature image. Finally, the three groups of data collected at the same time are time-series correlated and spatially registered to ensure that each point in the tissue impedance spectrum, each pixel in the perfusion map, and each region in the feature image can be accurately corresponded to the same physical location of the infusion site, thereby fusing into a structured comprehensive exploration data. Through the collaborative work of the multi-modal sensor array, the initial microenvironment of the infusion site can be comprehensively and finely scanned from multiple dimensions such as electricity, hemodynamics, and surface physics.

[0033] Optionally, the generating the microenvironment dynamic data comprises: based on the physiological state reflected in the physiological baseline fingerprint as a reference, dynamically normalizing each dimension feature in the exploration data to obtain normalized multi-dimensional features; fusing the normalized multi-dimensional features in time series to construct a spatiotemporal evolution map and generate the microenvironment dynamic data.

[0034] Specifically, the above-mentioned physiological baseline fingerprint is taken as the "gold standard" for judging whether the patient is in a physiological steady state. Only when the real-time physiological rhythm of the patient matches the physiological baseline fingerprint, the exploration data collected at this moment is established as the personalized initial reference state of the infusion site of the patient. Subsequently, during the entire infusion process, the exploration data is continuously acquired, and each dimension feature in the exploration data, such as the tissue impedance value at a specific frequency, the micro blood perfusion value at a specific position, and the surface temperature value, is normalized based on the initial reference state. For the normalized value of the feature at the time t , there is: ; wherein, is the real-time measurement raw value of the feature at this moment; is the initial baseline value of the aforementioned feature established under the physiological steady state. Through this ratio operation, the absolute dimension of the raw measurement value is eliminated. This effectively reduces the interference caused by inherent differences in individual body mass, skin thickness, basal metabolic level, etc., so that the subsequent analysis can focus on the real physiological changes caused by the infusion event, thereby obtaining the normalized multi-dimensional features. Next, these normalized multi-dimensional features in time series are fused to construct the spatiotemporal evolution map. This fusion is a structured integration of information representing different physiological dimensions within a shared spatiotemporal coordinate system. For example, a data matrix is constructed, with the spatial dimension corresponding to the two-dimensional area covered by the sensor array, the third dimension being the feature vector containing multiple features such as normalized tissue impedance, normalized micro blood perfusion, and normalized surface temperature at the same position point, and the fourth dimension being time. This dynamically updated matrix is the spatiotemporal evolution map. By analyzing the synergistic change patterns of various features in the spatiotemporal evolution map, higher-level physiological events can be analyzed. For example, a significant decrease in tissue impedance combined with a slight change in micro blood perfusion indicates fluid penetration; while a sharp increase in micro blood perfusion accompanied by an increase in surface temperature strongly indicates the occurrence of an inflammatory stress response. Finally, this spatiotemporal evolution map, which dynamically displays the tissue activity, fluid penetration degree, and stress response intensity at the infusion site, as well as their spatial distribution and temporal evolution rules, constitutes the microenvironment dynamic data required by the method.

[0035] For example, after the infusion process of patient A begins, the probe data is continuously acquired and personalized calibration is performed with reference to the physiological baseline fingerprint to generate microenvironment dynamic data. At the moment when the infusion begins, it is confirmed that the real-time physiological rhythm of patient A completely matches the previously generated physiological baseline fingerprint, and the probe data collected at this moment, i.e., tissue impedance of 520 ohms, blood perfusion of 75 perfusion units, etc., is established as the personalized initial baseline state. When the infusion is carried out to the 15th minute, a new set of probe data is collected, in which the tissue impedance measurement raw value at 50 kilohertz frequency decreases to 494 ohms. At this time, dynamic normalization processing is performed, and the normalized multi-dimensional features are calculated. Taking tissue impedance as an example, the calculation process is Similarly, the blood perfusion 80 measured at this moment is also normalized to 1 perfusion unit, and so on for all other features. These continuously acquired and calculated normalized feature values are fused in the time series in the coordinate system corresponding to the spatial position of the sensor, and a dynamically updated spatiotemporal evolution map is constructed. This map intuitively shows that after 15 minutes of infusion, the tissue impedance has decreased relative to the initial state, while the blood perfusion has increased, and the spatial distribution and evolution trend of these changes constitute the microenvironment dynamic data. By using the physiological baseline fingerprint for personalized calibration and dynamic normalization, the noise caused by individual inherent physiological differences is successfully reduced.

[0036] Based on the microenvironment dynamic data, real-time analysis is performed to dynamically predict the risk evolution during infusion, generate a risk evolution trajectory, and generate a dynamic disinfection strategy based on the risk evolution trajectory; Optionally, the dynamic disinfection strategy based on the risk evolution trajectory comprises: Identifying the key feature change vector in the microenvironment dynamic data, the key feature change vector including tissue conductivity gradient, microcirculation perfusion heterogeneity, and entropy increase of surface temperature field; By forward extrapolation of the feature change vector, the trend of the probability of liquid extravasation, inflammatory reaction or microbial colonization at the infusion site changing over time is constructed, and a risk evolution trajectory is generated; Constructing a cooperative response structure, positioning the risk evolution trajectory to the corresponding area in the cooperative response structure, and generating a dynamic disinfection strategy.

[0037] Specifically, first, the key feature change vector that can predict the occurrence of adverse events needs to be identified and quantified from the high-dimensional data in real time. Specifically, the tissue conductivity gradient will be calculated, the resistance component at a specific frequency will be extracted and converted into a conductivity value, a conductivity spatial distribution map will be formed, then through spatial differential operation, the conductivity change rate of each detection point and its adjacent points in the horizontal and vertical directions will be calculated, the vector of the conductivity gradient of the point will be synthesized, and the gradient amplitude will be calculated to quantify the degree of local tissue conductivity spatial change. At the same time, the microcirculation perfusion heterogeneity will be evaluated, which is usually quantified by calculating the spatial distribution variance or Gini coefficient of the perfusion value of each pixel point on the local micro blood flow perfusion map. The increase in heterogeneity means that the blood flow distribution is out of order, which is an important precursor of inflammatory response or microthrombosis. In addition, the entropy increase of the surface temperature field will also be calculated by applying the information entropy formula to the temperature distribution map in the surface biophysical feature image. A rapidly rising entropy value represents the trend of temperature distribution from order to chaos, reflecting the trend of local metabolic abnormalities or inflammatory point spread. The above-mentioned quantified key features are taken as inputs to form a key feature change vector that changes over time. Then a prediction model based on time series analysis, such as Kalman filter or long short-term memory network, is used to forward extrapolate the key feature change vector. This model is based on historical data and current state to predict the evolution trend of these key features in a short period of time in the future. By inputting these predicted feature trends into a pre-trained risk assessment function, the function behaves as a mathematical mapping model based on multivariate logistic regression and time series analysis. In the training stage, a large amount of clinically labeled data is used to learn the correlation strength between each feature and the complication. First, the input feature vector As is weighted and fused for tissue conductivity gradient, etc. to calculate the linear weighted sum There is: ; Among them, the weight vector and the bias term are determined in training. Then, a nonlinear transformation is applied to the linear combination result to calculate the prediction probability of the risk event There is: ; For multi-risk prediction, this process is implemented by setting up multiple logistic regression units in parallel at the output layer, ultimately generating probability curves that change over time. These curves together form a visualized risk evolution trajectory. Finally, a dynamic disinfection strategy is generated based on this risk evolution trajectory. This process involves automatically selecting the corresponding disinfection parameter combination for the region when the real-time generated risk evolution trajectory is located in a specific area of ​​the collaborative response structure, thereby generating a dynamic disinfection strategy that precisely matches the current and future predicted risk levels.

[0038] For example, at the 30-minute mark of infusion in patient A, significant changes in key feature vectors were identified. By extracting tissue resistance components at specific frequencies and converting them to conductivity values, a conductivity spatial distribution map was formed. The tissue conductivity gradient was then calculated using spatial differentiation, showing an increase of 30% in amplitude compared to the initial state. Calculation of the spatial variance of the local microcirculation perfusion map revealed a 25% increase in microcirculation perfusion heterogeneity. Applying the information entropy formula to the surface temperature distribution map showed a 15% increase in surface temperature field entropy. This set of key feature vectors, composed of conductivity gradient, perfusion heterogeneity, and temperature entropy increase, was input into a pre-trained long short-term memory network model for forward extrapolation. For example, for the risk of inflammatory response, the feature vector at the current time step is... The weight vector obtained through training Bias term The sum is -1.2. First, calculate the linear weighted sum. Then calculate the risk probability. The results were kept to three significant figures. Predictions showed that within the next 20 minutes, the probability of extravasation at patient A's infusion site would linearly increase to 40%, while the probability of an inflammatory response would climb exponentially to 70%. These probability trends over time collectively constituted a visualized risk evolution trajectory. Finally, this risk evolution trajectory was mapped to a pre-defined collaborative response structure, which automatically matched and generated a dynamic disinfection strategy called "Secondary Anti-inflammatory Priority" based on the characteristic that the inflammatory response probability exceeded 60% and the growth rate was high. By identifying deep patterns in the micro-data, the probability and trend of adverse events were predicted in advance, while the automatically generated dynamic disinfection strategy improved the ability of interventions to accurately target the major risks that were about to occur.

[0039] Optionally, the construction of the collaborative response structure includes: Based on the dynamic coupling relationship between tissue tolerance at the infusion site and microbial load, multiple risk response levels are defined, and corresponding disinfectant release methods are configured. The different disinfectant release modes are combined according to the synergistic relationship between the tissue repair rate and the microorganism removal efficiency to form a multi-stage linked disinfection strategy sequence. Based on the probability changes of fluid extravasation, inflammatory response or microorganism colonization in the risk evolution trajectory, the disinfection strategy sequence is screened and adapted in real time to obtain a synergistic response structure.

[0040] Specifically, a two-dimensional evaluation matrix is first established. One dimension represents the predicted level of microbial load, which is derived from the probability of microorganism colonization in the risk evolution trajectory; the other dimension represents the tissue tolerance of the infusion site. The tissue tolerance is a comprehensive evaluation index, which is calculated based on parameters such as tissue impedance, micro-blood flow perfusion stability and surface temperature field entropy in the microenvironment dynamic data, and reflects the current tissue's ability to withstand external stimuli such as disinfectants. Based on this two-dimensional matrix, the risk response level is divided into multiple regions, such as high-risk and high-tolerance region, high-risk and low-tolerance region, medium-risk and high-tolerance region, etc. For each region, one or more basic disinfectant release modes are pre-configured, such as high-strength light-sensitive activation for high-tolerance regions, and mild micro-current ion permeation for low-tolerance regions. Secondly, based on the above basic configuration, different disinfectant release modes are further combined to form a multi-stage linked disinfection strategy sequence with time sequence logic. The core of this combination is the synergistic relationship between the tissue repair rate and the microorganism removal efficiency. For example, a sequence can be designed. The first stage of the sequence uses a high-intensity disinfection method to maximize the microorganism removal efficiency in a short time and rapidly reduce the risk of infection. The second stage then switches to a low-intensity, long-acting disinfection method, which has less stimulation to the tissue and is beneficial to the recovery of the tissue repair rate, while effectively inhibiting the proliferation of residual microorganisms. A library containing multiple such disinfection strategy sequences will be pre-constructed, and each sequence will be optimized for a specific risk evolution pattern and tissue state transition path to achieve the best synergistic treatment effect. Finally, the dynamic core of this construction process is real-time screening and adaptation. After the risk evolution trajectory reflecting the future risk trend is generated, it is input into the strategy library for matching. The matching algorithm analyzes the morphological characteristics of the probability curves of fluid extravasation, inflammatory response and microorganism colonization in the risk evolution trajectory, such as peak value, rising rate and duration. For example, a trajectory that predicts a sharp rise in the probability of microorganism colonization but a stable inflammatory response probability will trigger the screening of a disinfection strategy sequence that prioritizes high-efficiency microorganism removal. Conversely, if the growth of the inflammatory response probability is much faster than that of the microorganism colonization, the sequence with anti-inflammatory effect or minimal tissue stimulation will be selected first. Through this real-time screening and adaptation based on prediction, the static strategy library is dynamically applied, and finally a synergistic response structure is obtained.

[0041] Exemplarily, first, according to clinical data analysis, the microbial load prediction level is divided into three levels of low, medium and high, and the tissue tolerance comprehensive evaluation value is divided into three levels of fragile, normal and strong, to form a nine-square risk response level matrix. For example, for the area of high microbial load-strong tissue tolerance, a high-intensity pulsed light-sensitive disinfectant release mode is configured; and for the area of medium microbial load-fragile tissue tolerance, a low-intensity micro-current ion permeation mode is configured. Secondly, a plurality of multi-stage linked disinfection strategy sequences are preset, such as a sequence named inflammatory inhibition and repair promotion, the first stage of which adopts 660 nm red light irradiation to inhibit inflammation, and the second stage switches to micro-current stimulation to promote tissue repair, which aims to optimize the synergistic relationship between tissue repair rate and microbial clearance efficiency. Finally, when the risk evolution trajectory of the patient A's real-time generated predicted inflammatory response probability sharply rising is input into this structure for real-time screening and adaptation. Analyzing the characteristics of the risk evolution trajectory, matching to the medium microbial load-normal tissue tolerance area in the risk response level matrix, and screening the most suitable inflammatory inhibition and repair promotion sequence from the strategy sequence library. This decision mechanism that can dynamically match the optimal disinfection scheme according to the risk prediction is the synergistic response structure. Through fine classification of risk and synergistic combination of strategies, not only is it possible to realize automatic intervention, but also it is more scientific and fine.

[0042] Synchronizing the infusion pulses with the physiological baseline fingerprint, controlling the infusion device to start adaptive infusion, and acquiring feedback signals from the multi-modal sensor array in real time; Optionally, the acquiring feedback signals from the multi-modal sensor array in real time comprises: According to the rhythm morphology parameters extracted from the physiological baseline fingerprint, identifying the current physiological cycle phase of the patient, and dynamically matching the timing of the infusion pulses based on the physiological cycle phase to obtain an infusion timing framework; Taking the infusion timing framework as the basic rhythm, combining the instantaneous changes of tissue compliance and local blood perfusion level reflected in the microenvironment dynamic data, and real-time modulating the intensity and interval of the infusion pulses to form adaptive infusion coordinated with the physiological state; Based on the adaptive infusion process, continuously capturing the dynamic responses of tissue impedance changes, micro-blood flow fluctuations and surface biophysical characteristics of the infusion site through the multi-modal sensor array to obtain feedback signals.

[0043] Specifically, first, the physiological baseline fingerprint is called and key rhythmic morphological parameters are extracted from it, especially those related to the cardiac and respiratory cycles, such as the R-wave peak time points identified from the pulse waveform. These time points represent the periodic peaks or troughs of the patient's physiological activities. Based on this, the current physiological cycle phase of the patient is identified, such as whether it is in the systolic or diastolic phase. By dynamically matching the pulse delivery initiation of the infusion drops with these specific physiological cycle phases, such as choosing to deliver in the diastolic phase when the vascular pressure is lower, a timing framework for infusion is established that is synchronized with the patient's intrinsic physiological rhythm. This timing framework for infusion ensures that the basic rhythm of the infusion is coordinated with the patient's physiological state. Second, further real-time modulation of the infusion pulses is performed to form adaptive infusion. This process is not just a temporal synchronization, but also a dynamic adjustment of infusion intensity and interval. The timing framework for infusion is used as the basic rhythm, and real-time introduction of tissue state information reflected in the microenvironment dynamic data. Specifically, the tissue compliance of the infusion site is concerned, which is usually derived from the low-frequency characteristics of the tissue impedance spectrum, reflecting the difficulty of the tissue to accommodate liquid; at the same time, the local blood perfusion level is concerned, which is directly obtained from the local micro-perfusion map. Through a control algorithm, the two instantaneous changes are normalized as modulation factors to obtain their two instantaneous change rates relative to the individual baseline state, and then, according to the preset proportional-integral-derivative control rule, the two instantaneous change rates are mapped to the adjustment amount of the basic infusion pulse intensity and the basic infusion interval; when the tissue compliance decreases or the blood perfusion decreases, the control output will proportionally reduce the pulse intensity and correspondingly prolong the pulse interval, thereby reducing the hydrostatic pressure and metabolic load of the local tissue caused by liquid infusion per unit time. This infusion method that combines physiological rhythm synchronization and local tissue state feedback is adaptive infusion. Finally, during the entire adaptive infusion process, uninterrupted monitoring is continuously performed through the multi-modal sensor array to capture the immediate impact of infusion behavior on the local microenvironment, thereby obtaining feedback signals. This means that when each modulated infusion pulse acts on the tissue, the transient changes in tissue impedance, the pulsatile response of micro-blood flow, and the slight fluctuations in surface biophysical characteristics caused by the pulse are simultaneously recorded. These continuous acquisition, closely related to infusion events, high temporal resolution dynamic response data streams collectively constitute the feedback signal for subsequent analysis and intervention decision-making. This feedback signal completely records the real-time response of the tissue to each adaptive infusion pulse.

[0044] Exemplarily, first, the rhythm morphology parameters of the cardiac cycle of patient A are extracted from the physiological baseline fingerprint of patient A, and it is identified that the heart rate of patient A is 65 beats per minute, i.e. about 0.92 seconds per cardiac cycle. Based on this, it is identified that the patient is currently in the diastolic phase of the physiological cycle, and the timing of the infusion pulse is dynamically matched, and the infusion device is set to start a micro-infusion every 50 milliseconds after detecting a pulse trough each time, thereby establishing an infusion timing framework synchronized with the cardiac cycle of the patient. Second, this framework is used as the basic rhythm, and real-time modulation is performed in combination with micro-environment dynamic data. When it is monitored that the tissue compliance of the infusion site decreases by 10% due to liquid accumulation, the control algorithm immediately reduces the intensity of the next infusion pulse, i.e. the single infusion volume, by 15%, and extends the infusion interval from 0.92 seconds to 1.1 seconds, forming an adaptive infusion. Finally, during this adaptive infusion process, the multi-modal sensor array continuously captures the real-time response of the tissue triggered by each micro-adjusted infusion pulse, including the transient decrease and rebound waveform of the tissue impedance after the pulse action, the instantaneous pulsatile enhancement of the micro-blood perfusion, and the slight fluctuation of the surface temperature. These continuously recorded dynamic response data streams synchronized with the infusion event form a feedback signal together. By synchronizing with the physiological rhythm and adjusting in real time according to the tissue feedback, the mechanical and fluid pressure impact of the infusion on the blood vessels and surrounding tissues is reduced, and the safety of the infusion and the comfort of the patient are improved.

[0045] Based on the deviation value comparison between the feedback signal and the physiological baseline fingerprint, when the deviation value exceeds the deviation threshold value determined by the statistical characteristics of the physiological baseline fingerprint, an intervention instruction is generated; Optionally, the generation of the intervention instruction comprises: extracting the dynamic characteristics of the physiological rhythm from the physiological baseline fingerprint, and constructing an expected response template representing the healthy response of the infusion site to external stimulation; quantitatively comparing the feedback signal obtained in real time with the expected response template in terms of morphology difference and energy difference, and calculating a deviation value; When the deviation value exceeds the deviation threshold value determined by the coefficient of variation of the physiological baseline fingerprint, an attribution analysis is performed according to the specific characteristic dimension of the deviation value, and an intervention instruction is generated.

[0046] Specifically, first from the patient's physiological baseline fingerprint, the dynamic characteristics of its physiological rhythm in the resting state are extracted, such as the typical response recovery time determined by heart rate variability, the adjustment mode of the autonomic nervous system to small stimuli reflected by the skin electric response, etc. Combined with the parameters of each infusion pulse in adaptive infusion, through a pre-established biophysical model, the ideal response waveform of the tissue at the infusion site to such an external stimulus under healthy conditions is simulated and constructed. The training set of the biophysical model should come from a large number of infusion pulse stimulation-tissue response paired data collected by a multi-modal sensor array under the healthy state of healthy subjects or patients. This waveform, i.e. the expected response template, contains the change amplitude and recovery pattern of tissue impedance, micro blood flow and surface temperature after pulse infusion. Next, enter the real-time comparison and deviation calculation stage. The feedback signal obtained in real time through the multi-modal sensor array is accurately aligned and quantitatively compared with the expected response template generated in the previous step. This comparison is divided into two dimensions. The first is the morphological difference comparison, such as using the dynamic time warping algorithm to calculate the shape distance between the feedback signal waveform and the expected response template waveform, to quantify the morphological distortion of the response process on the time axis, such as tailing or advancing. The second is the energy difference comparison, by calculating the difference between the integral value or root mean square value of the two signals within the response period, to quantify whether the overall intensity of the response is too strong or too weak. The difference values of the two dimensions are fused by weighting to calculate a comprehensive deviation value. For the calculation of the deviation value , there is: ; wherein, and are preset weight coefficients for adjusting the importance of morphological difference and energy difference in comprehensive evaluation; represents the waveform of the feedback signal obtained in real time; represents the waveform of the expected response template; indicates the dynamic time warping distance between the two waveforms; is the signal length, i.e. the number of sampling points; is the root mean square value of the real-time feedback signal; is the root mean square value of the expected response template; The value is a very small positive number. Finally, a decision is made. When the calculated deviation value exceeds a dynamically set deviation threshold, a significant deviation from the healthy response is considered abnormal. This deviation threshold is not a fixed value, but is determined by the statistical characteristics of the physiological baseline fingerprint, especially its inherent coefficient of variation. For an individual whose physiological rhythm fluctuates significantly, the deviation threshold will be relaxed accordingly, and vice versa, thus achieving personalized thresholding. Once the deviation value exceeds the limit, a general alarm is not immediately issued, but attribution analysis is performed based on the specific characteristics of the deviation value. For example, if the deviation mainly comes from the tailing of morphological differences, it may be determined that there is poor fluid penetration; if it mainly comes from a surge in energy differences, it may be determined that there is an overreaction to the early inflammatory response. Based on this attribution analysis, a specific intervention instruction is generated, such as reducing the infusion pulse intensity if extravasation is suspected, or preparing to initiate an anti-inflammatory and disinfection mode if inflammation is suspected. Figure 2 As shown, when the response value of the real-time feedback signal deviates significantly from the expected template (i.e., the shaded area in the figure), the deviation value will be calculated. Once this deviation exceeds the preset personalized deviation threshold, an intervention command will be triggered.

[0047] For example, firstly, based on patient A's physiological baseline fingerprint and the pulse parameters of the current adaptive infusion, a predictive response template was constructed using a biophysical model, depicting the ideal recovery curve of tissue impedance after a normal infusion pulse. At 45 minutes of infusion, the acquired real-time feedback signal waveform showed that the tissue impedance recovery process was significantly prolonged, and the overall fluctuation intensity exceeded expectations. Subsequently, deviation values ​​were compared, a comprehensive deviation value was calculated, and a setting was established. It is 0.6. The value is 0.4. The calculated dynamic time warping distance between waveforms is 50, and the signal length is 150 sampling points. Combining the values ​​of the 150 sampling points of the real-time feedback signal, the root mean square value of the real-time feedback signal is calculated. Similarly, the root mean square value of the expected response template. Set a minimum number for Then the deviation value , the result is reserved to two decimal places. At the same time, according to the statistical characteristics reflecting the more stable physiological rhythm in the physiological baseline fingerprint of the patient's nail, the personalized deviation threshold is determined to be 2.00. Since the calculated deviation value 2.65 exceeds the deviation threshold 2.00, it is determined that a significant abnormality has occurred. Through attribution analysis, it is found that the deviation is mainly due to the tailing of morphological differences and the significant enhancement of energy differences, and it is determined that this abnormal mode is poor liquid permeability accompanied by excessive early inflammatory response, and an intervention instruction is generated accordingly. Suspected exosmosis with inflammation, reduce the infusion pulse intensity and start the first-level anti-inflammatory disinfection mode. By quantitatively comparing real-time feedback with personalized expected templates, a sensitive and specific abnormality detection mechanism is established, so that subsequent intervention measures can directly hit the core of the problem.

[0048] According to the intervention instruction, the infusion parameters of the infusion device are adjusted, and the supplementary disinfection operation is triggered based on the dynamic disinfection strategy.

[0049] Optionally, the triggering of the supplementary disinfection operation comprises: Based on the analysis of the intervention instruction, the specific adjustment parameters of the flow rate, pulse frequency or infusion pressure of the infusion device are converted and executed; The risk type implied by the intervention instruction is matched with the dynamic disinfection strategy, and the appropriate disinfection mode is selected; Based on the disinfection mode, the light-sensitive disinfection medium is activated by specific wavelength light or the targeted penetration of disinfection ions is promoted by micro-current, triggering the supplementary disinfection operation.

[0050] Specifically, the intervention instruction is first analyzed. Since the intervention instruction itself already contains a preliminary attribution to the risk type, such as suspected exosmosis or suspected inflammation, this information can be used to find and match the specific execution scheme from the pre-set parameter adjustment rule library. For example, for the instruction of suspected exosmosis, the rule library may indicate that the flow rate of the infusion device should be reduced, or the infusion pulse intensity in adaptive infusion should be reduced, and the interval of the infusion pulse should be increased. These abstract instructions are converted into specific adjustment parameters that can be executed by the infusion device hardware, such as reducing the flow rate by a certain percentage from the current value, or setting the pulse frequency to a new value, and then sending the instruction to the controller of the infusion device and executing the adjustment. For example, Figure 3As shown, after the abnormality is detected, the infusion flow rate and pulse frequency are reduced, while the infusion pressure is adjusted appropriately. At the same time, the type of risk implied by the intervention instruction is matched with the dynamic disinfection strategy previously generated based on the risk evolution trajectory. The dynamic disinfection strategy itself is a collection of multiple disinfection protocols, each of which corresponds to one or more risk combinations. For example, a disinfection mode labeled as anti-inflammatory priority can be associated with a risk evolution trajectory with a significant increase in the probability of inflammatory response in the dynamic disinfection strategy. When the received intervention instruction is suspected inflammation, the anti-inflammatory priority disinfection mode is accurately selected in the dynamic disinfection strategy. This matching process ensures that the upcoming supplementary disinfection operation is the most suitable for the current emergency situation. After selecting the appropriate disinfection mode, the trigger phase of the supplementary disinfection operation is entered. The specific implementation of this phase depends on the definition of the selected disinfection mode. If the selected disinfection mode is based on photosensitive disinfection medium, such as pre-coating the surface of the infusion patch or infusion catheter with photosensitive agents such as methylene blue, the specific wavelength LED light source integrated in the multi-modal sensor array is activated. The wavelength, intensity and irradiation time of the light source are all preset by the disinfection mode, and the photosensitive medium generates active oxygen species such as singlet oxygen by light excitation, thereby locally and targetedly killing microorganisms at the infusion site. If the selected disinfection mode is based on micro-current to promote disinfection ion penetration, such as through the electroosmotic ion introduction technology, the microelectrode in the array is controlled to apply a weak, safe direct current or pulse current on the skin surface. The current forms an electric field that drives effective disinfection ions such as silver ions or chlorine ions in the disinfection solution to penetrate the stratum corneum and reach the deep layer where microorganisms may colonize. Through this precisely controlled physical method, a targeted supplementary disinfection operation synchronized with infusion parameter adjustment is triggered.

[0051] Exemplarily, after generating the intervention instruction of suspected extravasation with inflammation for the patient A, and reducing the infusion pulse intensity and starting the first-level anti-inflammatory disinfection mode, the linkage operation is immediately executed. First, the instruction is parsed and converted into specific adjustment parameters for the infusion device, that is, sending an instruction to the controller of the infusion device to further reduce the average pulse intensity of adaptive infusion by 20% on the current basis. The infusion device executes the adjustment immediately after receiving the instruction. At the same time, the type of inflammation risk implied in the intervention instruction is matched with the dynamic disinfection strategy generated previously, and the disinfection mode marked as first-level anti-inflammatory is accurately selected. The mode is defined as activating the photosensitive disinfection medium by light of a specific wavelength. The light-emitting unit deployed on the multi-modal sensor array is triggered to emit red light with a wavelength of 660 nanometers at a preset power for 60 seconds of continuous irradiation on the infusion site. This red light irradiation can effectively inhibit local inflammatory response while being non-invasive to the tissue. Thus, an accurate linkage of infusion parameter adjustment and supplementary disinfection operation driven by the intervention instruction is completed. By adjusting the infusion parameters, the continuous stimulation to the tissue is reduced, and by triggering the supplementary disinfection operation, the pathological reaction that has already occurred is actively handled. The damage can be effectively controlled and the recovery can be promoted.

[0052] Based on the same inventive concept, as Figure 4 shown, the present application also provides an infusion control device with disinfection function, which comprises: a physiological fingerprint generation module for acquiring physiological identification information of a patient and collecting at least one steady physiological rhythm to generate a physiological baseline fingerprint; a probe data calibration module for dynamically probing a target infusion site using a multi-modal sensor array to obtain probe data including tissue impedance spectrum, local micro blood perfusion map and surface biophysical feature image, and performing individualized calibration and interpretation on the probe data with reference to the physiological baseline fingerprint to generate microenvironment dynamic data; a risk prediction disinfection module for real-time analysis based on the microenvironment dynamic data to dynamically predict the risk evolution in the infusion process, generate a risk evolution trajectory, and generate a dynamic disinfection strategy according to the risk evolution trajectory; a synchronous infusion control module for synchronizing the infusion pulse with the physiological baseline fingerprint, controlling the infusion device to start adaptive infusion, and obtaining feedback signals from the multi-modal sensor array in real time; a real-time intervention judgment module for comparing the deviation value based on the feedback signal and the physiological baseline fingerprint, and generating an intervention instruction when the deviation value exceeds the deviation threshold determined by the statistical characteristics of the physiological baseline fingerprint; an infusion disinfection linkage module for adjusting the infusion parameters of the infusion device according to the intervention instruction, and triggering supplementary disinfection operation based on the dynamic disinfection strategy.

Claims

1. An infusion control method with a disinfection function, characterized by, The method comprises: acquiring physiological identification information of a patient and collecting at least one steady physiological rhythm to generate a physiological baseline fingerprint; using a multi-modal sensor array to dynamically explore a target infusion site, acquiring exploration data including tissue impedance spectrum, local micro blood perfusion map and surface biophysical feature image, and referring to the physiological baseline fingerprint to perform individualized calibration and interpretation on the exploration data to generate microenvironment dynamic data; based on the microenvironment dynamic data, performing real-time analysis to dynamically predict the risk evolution in the infusion process, generating a risk evolution trajectory, and generating a dynamic disinfection strategy according to the risk evolution trajectory; synchronizing the infusion pulse with the physiological baseline fingerprint, controlling the infusion device to start adaptive infusion, and acquiring feedback signals from the multi-modal sensor array in real time; based on the feedback signal and the physiological baseline fingerprint, comparing the deviation value, when the deviation value exceeds the deviation threshold determined by the statistical characteristics of the physiological baseline fingerprint, generating an intervention instruction; according to the intervention instruction, adjusting the infusion parameters of the infusion device, and based on the dynamic disinfection strategy, triggering a supplementary disinfection operation.

2. The infusion control method with sterilization function according to claim 1, characterized in that, The generation of the physiological baseline fingerprint comprises: acquiring pulse waveform, skin electrical response and fingertip temperature fluctuation of the patient as the steady physiological rhythm; coupling the features in time domain and frequency domain of the steady physiological rhythm to extract rhythm morphological parameters representing individual specificity; associating and mapping the rhythm morphological parameters with the physiological identification information to construct a multi-dimensional feature vector and generate a physiological baseline fingerprint.

3. The infusion control method with sterilization function according to claim 1, characterized in that, The acquisition of exploration data including tissue impedance spectrum, local micro blood perfusion map and surface biophysical feature image comprises: deploying a multi-modal sensor array with high-frequency electrical sensing unit, laser Doppler imaging unit and thermal infrared spectral detection unit around the target infusion area; conducting frequency scanning on the subcutaneous tissue of the infusion site through the high-frequency electrical sensing unit, recording the impedance response at different frequencies, and constructing the tissue impedance spectrum; using the laser Doppler imaging unit to capture the real-time changes of the blood flow hemodynamics of the subcutaneous microvascular network of the infusion site, and drawing the local micro blood perfusion map; based on the thermal infrared spectral detection unit, capturing the surface temperature distribution, skin color gradient and micro texture changes of the infusion site to generate the surface biophysical feature image; time series association and spatial registration of the above tissue impedance spectrum, local micro blood perfusion map and surface biophysical feature image to obtain the exploration data.

4. The infusion control method with sterilization function according to claim 1, characterized in that, The generation of microenvironment dynamic data comprises: based on the physiological state reflected in the physiological baseline fingerprint as a reference, dynamically normalizing each dimension feature in the exploration data to obtain normalized multi-dimensional features; fusing the normalized multi-dimensional features in time series to construct a spatiotemporal evolution map and generate microenvironment dynamic data.

5. The infusion control method with sterilization function according to claim 1, characterized in that, The generation of a dynamic disinfection strategy according to the risk evolution trajectory comprises: identifying a key feature change vector in the microenvironment dynamic data, the key feature change vector including a tissue conductivity gradient, a microcirculation perfusion heterogeneity, and an entropy increase of a surface temperature field; constructing a trend of a probability of the infusion site developing a fluid extravasation, an inflammatory reaction, or a microbial colonization changing over time by forward extrapolating the feature change vector, and generating a risk evolution trajectory; constructing a cooperative response structure, positioning the risk evolution trajectory to a corresponding region in the cooperative response structure, and generating a dynamic disinfection strategy.

6. The infusion control method with sterilization function according to claim 5, characterized in that, The constructing the cooperative response structure includes: dividing a plurality of risk response levels according to a dynamic coupling relationship between tissue tolerance and microbial load of the infusion site, and configuring a corresponding disinfectant release mode; combining different disinfectant release modes according to a cooperative relationship between tissue repair rate and microbial clearance efficiency, forming a multi-stage linked disinfection strategy sequence; based on the probability change of the fluid extravasation, the inflammatory reaction, or the microbial colonization in the risk evolution trajectory, the disinfection strategy sequence is screened and adapted in real time to obtain the cooperative response structure.

7. The infusion control method with sterilization function according to claim 2, characterized in that, The real-time feedback signal from the multi-modal sensor array includes: identifying a current physiological cycle phase of the patient according to a rhythm morphology parameter extracted from the physiological baseline fingerprint, and dynamically matching a timing of an infusion pulse based on the physiological cycle phase to obtain an infusion timing framework; using the infusion timing framework as a basic rhythm, combining instantaneous changes of tissue compliance and local blood perfusion levels reflected in the microenvironment dynamic data, and real-time modulating an intensity and interval of the infusion pulse to form an adaptive infusion coordinated with the physiological state; based on the adaptive infusion process, continuously capturing tissue impedance changes, microcirculation fluctuations, and dynamic responses of surface biophysical characteristics of the infusion site through the multi-modal sensor array to obtain the feedback signal.

8. The infusion control method with sterilization function according to claim 1, characterized in that, The generating the intervention instruction includes: extracting a dynamic feature of a physiological rhythm from the physiological baseline fingerprint to construct an expected response template representing a healthy response of the infusion site to external stimulation; quantitatively comparing a morphological difference and an energy difference between the real-time feedback signal and the expected response template to calculate a deviation value; when the deviation value exceeds a deviation threshold value determined by a coefficient of variation of the physiological baseline fingerprint, performing attribution analysis according to specific feature dimensions of the deviation value to generate the intervention instruction.

9. The infusion control method with sterilization function according to claim 1, characterized in that, The triggering a supplementary disinfection operation includes: based on the intervention instruction, converting it into specific adjustment parameters of flow rate, pulse frequency, or infusion pressure of the infusion device and performing the adjustment; matching a risk type implied by the intervention instruction with the dynamic disinfection strategy to select an appropriate disinfection mode; based on the disinfection mode, activating a photosensitive disinfection medium or promoting targeted penetration of disinfection ions through a specific wavelength of light or a micro-current to trigger the supplementary disinfection operation.

10. The infusion control device with disinfection function, applied to the infusion control method with disinfection function according to any one of claims 1-9, characterized in that, The device includes: a physiological fingerprint generation module for acquiring physiological identification information of a patient and collecting at least one steady-state physiological rhythm to generate a physiological baseline fingerprint; An exploration data calibration module is configured to dynamically explore a target infusion site using a multi-modal sensor array, to obtain exploration data including tissue impedance spectrum, local micro blood perfusion map, and surface biophysical feature image, and to perform personalized calibration and interpretation of the exploration data with reference to the physiological baseline fingerprint, to generate microenvironment dynamic data; A risk prediction disinfection module is configured to perform real-time analysis based on the microenvironment dynamic data, to dynamically predict risk evolution during infusion, to generate a risk evolution trajectory, and to generate a dynamic disinfection strategy according to the risk evolution trajectory; A synchronous infusion control module is configured to synchronize infusion pulses with the physiological baseline fingerprint, to control an infusion device to start adaptive infusion, and to obtain feedback signals from the multi-modal sensor array in real time; A real-time intervention judgment module is configured to compare a deviation value based on the feedback signals and the physiological baseline fingerprint, to generate an intervention instruction when the deviation value exceeds a deviation threshold determined by statistical characteristics of the physiological baseline fingerprint; An infusion disinfection linkage module is configured to adjust infusion parameters of the infusion device according to the intervention instruction, and to perform linkage based on the dynamic disinfection strategy, to trigger a supplementary disinfection operation.

Images