Method and system for risk assessment and path planning of anesthetic recovery elderly patient transfer

Abstract

This invention relates to the field of medical rehabilitation technology, and in particular to a method and system for risk assessment and route planning during the transport of elderly patients undergoing anesthesia and resuscitation. The method involves acquiring physiological monitoring, anesthetic drug administration, and transport environment data; extracting and concatenating features to generate a risk feature vector; constructing a dynamic risk propagation map that integrates time-varying characteristics and spatial constraints using spatial topological information; performing constrained path search based on risk accumulation limits and transport duration limits; selecting a target transport route for transport; and collecting real-time data combined with early warning rules for risk warning. This invention improves the safety and timeliness of the transport process for elderly patients.

Description

Technical Field

[0001] This invention relates to the field of medical rehabilitation technology, and in particular to a method and system for risk assessment and pathway planning for the transfer of elderly patients undergoing anesthesia and resuscitation. Background Technology

[0002] In the intra-hospital transport of elderly patients recovering from anesthesia, risk assessment and route planning are crucial for ensuring patient safety. Current practices primarily rely on the subjective judgment of medical staff based on clinical experience, combined with the patient's basic vital signs parameters, such as heart rate, blood pressure, and blood oxygen saturation, using simple scoring tables or threshold judgments to assess transport risks. Meanwhile, route planning typically adopts the shortest path principle, selecting the route with the shortest geometric distance from the operating room or recovery room to the destination (such as the ICU or ward). In some scenarios, the smoothness of vertical transport between floors is considered, but insufficient attention is paid to dynamic factors in the transport environment.

[0003] Furthermore, risk assessments often lack consideration for the time-varying nature of patient conditions and pharmacokinetics. The physiological state of elderly patients recovering from anesthesia may fluctuate rapidly during transport due to drug residues, changes in position, or environmental stress. Relying solely on a single static assessment before transport cannot capture these changing trends, easily leading to an underestimation of high-risk conditions. Route planning does not adequately consider the spatial topological constraints and emergency response accessibility of the transport environment. When using the shortest path strategy, spatial factors such as corridor congestion, elevator waiting times, and the location of emergency rescue equipment may be overlooked, resulting in longer actual travel times than expected or the inability to obtain timely medical intervention in emergencies, thus increasing safety risks during transport. Therefore, existing technologies urgently require an integrated risk assessment and route optimization method that can incorporate dynamic patient physiological information and spatial environmental constraints. Summary of the Invention

[0004] This invention provides a method and system for risk assessment and route planning for the transfer of elderly patients undergoing anesthesia and resuscitation, which can solve the problems in the prior art.

[0005] A first aspect of this invention provides a method for risk assessment and pathway planning for the transport of elderly patients undergoing anesthesia recovery, comprising:

[0006] Acquire physiological monitoring data, anesthetic drug data, and transport environment data of patients awaiting transfer;

[0007] Feature extraction and feature concatenation are performed on the physiological monitoring data and the anesthetic drug data to generate a risk feature vector;

[0008] Based on the risk feature vector and the spatial topology information in the transit environment data, a dynamic risk propagation graph that integrates the time-varying characteristics of risk and spatial constraints is constructed, and the edge weights are defined by the path accessibility conditions and emergency response accessibility constraints.

[0009] Based on the dynamic risk propagation map, a constrained path search is performed. By setting a risk accumulation limit and a transfer time limit, a set of feasible paths that meet the dual constraints is generated. The risk fluctuation variance and emergency intervention coverage of each feasible path are calculated. A path transfer scoring function is constructed and a target transfer path is selected.

[0010] Based on the target transport route, the patient to be transported is transported and real-time physiological monitoring data is collected. Risk warnings are then issued in conjunction with preset early warning triggering rules.

[0011] Feature extraction and feature concatenation are performed on the physiological monitoring data and the anesthetic drug data to generate a risk feature vector, including:

[0012] The physiological monitoring data is sampled hierarchically according to a preset time scale sequence to obtain physiological signal sequences at multiple resolution levels. Frequency domain decomposition and time domain statistical analysis are performed on each sequence to extract spectral energy distribution features and statistical moment features, forming a scale feature subset corresponding to each resolution level.

[0013] Calculate the mutual information between scale feature subsets of adjacent resolution levels, construct inter-level feature transfer weights based on the mutual information, and perform weighted summation of scale feature subsets of each resolution level according to the feature transfer weights to generate cross-scale fused feature vectors.

[0014] Pharmacokinetic modeling is performed on the anesthetic drug data to calculate the drug elimination half-life and current drug residual concentration. Consciousness recovery index and muscle strength recovery index are calculated respectively by combining the preset pharmacological action function to generate an anesthesia recovery vector.

[0015] The risk feature vector is obtained by concatenating the cross-scale fusion feature vector with the anesthesia recovery vector.

[0016] Pharmacokinetic modeling is performed on the anesthetic drug data to calculate the drug elimination half-life and current residual drug concentration. Based on a preset pharmacological function, the consciousness recovery index and muscle strength recovery index are calculated respectively, generating an anesthesia recovery vector, including:

[0017] Extract the dosing time sequence, dosing dose sequence, and anesthetic drug type from the anesthetic drug data, and obtain the standard clearance rate and standard distribution volume from the preset pharmacokinetic parameter library according to the anesthetic drug type;

[0018] The patient's weight, age, sex, and serum creatinine level are obtained. The glomerular filtration rate is calculated, and the standard clearance rate is individually corrected to obtain an individualized clearance rate. The drug elimination half-life is calculated based on the individualized clearance rate and the standard volume of distribution.

[0019] For each dosing time in the dosing time sequence, the time interval from the dosing time to the current time is calculated. The initial blood drug concentration is calculated based on the dosing dose and the standard distribution volume. The residual concentration component is calculated using a first-order kinetic equation in combination with the time interval and the drug elimination half-life. The current drug residual concentration is obtained by summing the residual concentration components.

[0020] An S-shaped curve function containing the half-maximal effect concentration is established as a preset pharmacological action function. The current drug residual concentration is substituted into the S-shaped curve function, and the half-maximal effect concentrations related to consciousness and muscle strength are set respectively. The consciousness recovery index and muscle strength recovery index are calculated and combined with the current drug residual concentration and the drug elimination half-life to form an anesthesia recovery vector.

[0021] Based on the risk feature vector and the spatial topology information in the transit environment data, a risk propagation graph integrating the time-varying characteristics of risk and spatial constraints is constructed. The edge weights are defined by path accessibility conditions and emergency response accessibility constraints, including:

[0022] Spatial node coordinates, node types and connection relationships are extracted from the transport environment data to construct an initial topology map. The risk feature vector is expanded in time series to obtain a risk evolution trajectory sequence. The risk decay coefficient corresponding to each moment is calculated based on the risk time-varying decay function.

[0023] The initial risk state of the risk evolution trajectory sequence is injected into the starting node. The spatial length and estimated passage time are calculated for each edge. The risk state value at the corresponding moment is selected from the risk evolution trajectory sequence, and the risk arrival value of the edge endpoint is calculated in combination with the corresponding risk decay coefficient.

[0024] Obtain the clearance dimensions, turning radius, and ground friction coefficient of each side's corresponding channel. Combine the minimum passage section requirements and maximum climbing capacity threshold of the transfer equipment to calculate the passage feasibility index. When the passage feasibility index is lower than the preset passage lower limit, a penalty weight is assigned to the side.

[0025] Extract the location of the emergency response unit and the standard response delay, calculate the shortest path length from the midpoint of each side to the nearest emergency response unit, calculate the emergency intervention time based on the shortest path length and the standard response delay, and normalize it to obtain the emergency accessibility coefficient.

[0026] The risk arrival value, the travel feasibility index, and the emergency accessibility coefficient are weighted and fused to generate edge weights, and a dynamic risk propagation graph is constructed.

[0027] The risk feature vector is expanded over time to obtain a risk evolution trajectory sequence, and the risk decay coefficient at each time point is calculated based on the time-varying risk decay function, including:

[0028] A time-dimensional projection transformation is performed on the risk feature vector, and the start and end times of the risk observation window are set. Dynamic sampling is performed between the start and end times with an adaptive time step, and the adaptive time step adjusts the sampling density according to the fluctuation amplitude of each component of the risk feature vector.

[0029] At each sampling time, the current component value of the risk feature vector is extracted and the risk change rate relative to the starting time is calculated. The instantaneous risk intensity at each sampling time is calculated by combining the current component value and the instantaneous risk intensity at each sampling time is connected in time sequence to form a risk evolution trajectory sequence.

[0030] The risk evolution trajectory sequence is piecewise linearly fitted to obtain multiple risk decay curves. The slope of each curve is calculated and the inflection point is identified. Based on the inflection point, the risk time-varying decay function is divided into multiple decay stages, and differentiated decay index parameters are set.

[0031] For any sampling moment in the risk evolution trajectory sequence, determine the decay stage and substitute the corresponding decay index parameter into the risk time-varying decay function to obtain the risk decay coefficient for each sampling moment.

[0032] Based on the dynamic risk propagation map, a constrained path search is performed. A set of feasible paths satisfying these dual constraints is generated by setting a risk accumulation cap and a transfer time cap. The risk volatility variance and emergency intervention coverage of each feasible path are calculated. A path transfer scoring function is constructed, and target transfer paths are selected, including:

[0033] Extract node risk propagation values ​​and edge connectivity from the dynamic risk propagation graph, calculate the difference in risk propagation values ​​between adjacent nodes as the risk gradient, and generate a weighted adjacency matrix by weighting the edge connectivity. Based on the risk accumulation limit and the transit time limit, determine the maximum number of high-risk nodes and the maximum number of path hops, and use these as constraints to perform a breadth-first search on the weighted adjacency matrix to generate a set of feasible paths.

[0034] For each feasible path, the risk propagation value of each node is extracted and organized into a risk sequence. The risk sequence is differentially processed and the standard deviation of the differential result is calculated as the risk fluctuation variance. Nodes on the path where emergency equipment or emergency personnel are configured are identified and the maximum interval distance between them is calculated. The emergency intervention coverage rate is calculated based on the ratio of the maximum interval distance to the total path length.

[0035] The cumulative risk value, total transit time, risk volatility variance, and emergency intervention coverage rate are used as scoring indicators. Weighting coefficients are assigned and a path transit scoring function is constructed. The path transit score of each feasible path is calculated, and the feasible path with the highest path transit score is selected as the target transit path.

[0036] Based on the risk accumulation limit and the transit time limit, the maximum number of high-risk nodes and the maximum path hop count are determined, and used as constraints to perform a breadth-first search on the weighted adjacency matrix to generate a set of feasible paths; including:

[0037] Extract node risk propagation values ​​from the weighted adjacency matrix and sort them in descending order. Accumulate the sorted risk propagation values ​​until the accumulated value reaches the risk accumulation limit. Use the number of nodes involved in the accumulation process as the maximum number of high-risk nodes. Calculate the expected path jump time based on the edge weight distribution characteristics of the weighted adjacency matrix. Determine the maximum number of path jumps based on the ratio of the expected time to the upper limit of the transit time.

[0038] During the breadth-first search process, a constraint status identifier is established for each path. The constraint status identifier records the number of high-risk nodes currently included in the path and the number of path hops already executed. When expanding the path, the constraint status identifier is dynamically updated according to the risk propagation value of the candidate node. Only the paths whose constraint status identifiers meet the dual constraint conditions are expanded. All paths that reach the target node and whose constraint status identifiers are compliant are collected to form a feasible path set.

[0039] A second aspect of this invention provides a system for risk assessment and route planning for the transport of elderly patients undergoing anesthesia recovery, comprising:

[0040] The data acquisition unit is used to acquire physiological monitoring data, anesthetic drug data, and transport environment data of the patient to be transported.

[0041] The feature generation unit is used to extract and concatenate features from the physiological monitoring data and the anesthetic drug data to generate a risk feature vector.

[0042] The risk propagation graph construction unit is used to construct a dynamic risk propagation graph that integrates the time-varying characteristics of risk and spatial constraints based on the risk feature vector and the spatial topology information in the transit environment data. The edge weights are defined by the path accessibility conditions and emergency response accessibility constraints.

[0043] The path selection unit is used to perform constrained path search based on the dynamic risk propagation map, generate a set of feasible paths that meet the dual constraints by setting a risk accumulation limit and a transfer time limit, calculate the risk fluctuation variance and emergency intervention coverage of each feasible path, construct a path transfer scoring function, and select the target transfer path.

[0044] The transfer early warning unit is used to transfer the patient to be transferred based on the target transfer route and collect real-time physiological monitoring data, and to issue risk warnings in combination with preset early warning triggering rules.

[0045] A third aspect of the present invention provides an electronic device, comprising:

[0046] processor;

[0047] Memory used to store processor-executable instructions;

[0048] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0049] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0050] This method for assessing the risks and planning the transport routes of elderly patients after anesthesia resuscitation significantly improves the safety and timeliness of the transport process. By integrating physiological monitoring, anesthetic drug administration, and environmental data to generate risk feature vectors, it can accurately capture the postoperative physiological vulnerability and residual effects of anesthesia in elderly patients, effectively avoiding transport risks caused by misjudgment based on a single indicator. The dynamic risk propagation map integrates the time-varying characteristics of risk with spatial topological constraints, enabling route planning to adapt to fluctuations in the patient's condition and the hospital's architectural structure in real time, significantly reducing the probability of complications caused by prolonged transport time or spatial obstacles.

[0051] By combining risk accumulation caps and transport time caps with dual constraints, the route search ensures that the selected route minimizes transport time within a controllable risk range, making it particularly suitable for emergency scenarios involving elderly patients with insufficient respiratory and circulatory function compensation. Introducing risk fluctuation variance and emergency intervention coverage as key parameters in the scoring function allows for the selection of routes with minimal peak risk fluctuations and optimal accessibility to emergency resources from multiple feasible options. This significantly improves rescue efficiency in emergencies and avoids delays in the rescue window due to inappropriate route selection.

[0052] Based on real-time physiological monitoring data and preset early warning rules, this risk warning mechanism can dynamically assess the evolution of the patient's condition during transport. Once a threshold is triggered, an alarm is immediately triggered, guiding medical staff to intervene in advance. This method significantly reduces the incidence of high-risk events in elderly patients, such as hypoxemia and arrhythmia, during transport through a closed-loop data system, while optimizing the allocation of medical resources. It is suitable for practical deployment in complex transport scenarios in large hospitals. Attached Figure Description

[0053] Figure 1 This is a flowchart illustrating the method for risk assessment and route planning for the transfer of elderly patients undergoing anesthesia resuscitation according to an embodiment of the present invention.

[0054] Figure 2 This is a flowchart illustrating the method for constructing risk feature vectors in an embodiment of the present invention. Detailed Implementation

[0055] 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, and 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.

[0056] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0057] Figure 1 This is a flowchart illustrating the method for risk assessment and route planning for the transfer of elderly patients undergoing anesthesia resuscitation, as described in an embodiment of the present invention.

[0058] Methods for risk assessment and pathway planning for the transfer of elderly patients undergoing anesthesia resuscitation include:

[0059] Acquire physiological monitoring data, anesthetic drug data, and transport environment data of patients awaiting transfer;

[0060] Feature extraction and feature concatenation are performed on the physiological monitoring data and the anesthetic drug data to generate a risk feature vector;

[0061] Based on the risk feature vector and the spatial topology information in the transit environment data, a dynamic risk propagation graph that integrates the time-varying characteristics of risk and spatial constraints is constructed, and the edge weights are defined by the path accessibility conditions and emergency response accessibility constraints.

[0062] Based on the dynamic risk propagation map, a constrained path search is performed. By setting a risk accumulation limit and a transfer time limit, a set of feasible paths that meet the dual constraints is generated. The risk fluctuation variance and emergency intervention coverage of each feasible path are calculated. A path transfer scoring function is constructed and a target transfer path is selected.

[0063] Based on the target transport route, the patient to be transported is transported and real-time physiological monitoring data is collected. Risk warnings are then issued in conjunction with preset early warning triggering rules.

[0064] Figure 2 This is a flowchart illustrating the method for constructing a risk feature vector according to an embodiment of the present invention. The method involves feature extraction and concatenation of the physiological monitoring data and the anesthetic drug data to generate a risk feature vector, including:

[0065] The physiological monitoring data is sampled hierarchically according to a preset time scale sequence to obtain physiological signal sequences at multiple resolution levels. Frequency domain decomposition and time domain statistical analysis are performed on each sequence to extract spectral energy distribution features and statistical moment features, forming a scale feature subset corresponding to each resolution level.

[0066] Calculate the mutual information between scale feature subsets of adjacent resolution levels, construct inter-level feature transfer weights based on the mutual information, and perform weighted summation of scale feature subsets of each resolution level according to the feature transfer weights to generate cross-scale fused feature vectors.

[0067] Pharmacokinetic modeling is performed on the anesthetic drug data to calculate the drug elimination half-life and current drug residual concentration. Consciousness recovery index and muscle strength recovery index are calculated respectively by combining the preset pharmacological action function to generate an anesthesia recovery vector.

[0068] The risk feature vector is obtained by concatenating the cross-scale fusion feature vector with the anesthesia recovery vector.

[0069] After acquiring physiological monitoring data and anesthetic medication data of patients awaiting transport, deep feature extraction and fusion processing are required to generate feature vectors that comprehensively reflect the patient's transport risk status. Physiological monitoring data typically includes multimodal time series data such as electrocardiogram signals, blood pressure waveforms, blood oxygen saturation curves, and respiratory rate sequences. These data exhibit different physiological meanings and risk indication characteristics at different time scales.

[0070] To address the multi-scale characteristics of physiological monitoring data, a hierarchical sampling strategy is employed for feature extraction. Specifically, the preset timescale sequences can be set to five levels: 1 second, 5 seconds, 15 seconds, 30 seconds, and 60 seconds, corresponding to ultra-short-term fluctuations, short-term trends, medium-term changes, metastable states, and steady-state characteristics, respectively. Taking electrocardiogram (ECG) signals as an example, the 1-second scale can capture the high-frequency components of heart rate variability and the morphological features of the QRS complex; the 5-second scale can reflect the rapid regulation capability of heart rate; and the 15- to 60-second scales can reflect the regulatory state of the autonomic nervous system. For each timescale, the original signal is resampled using a sliding window method to obtain the physiological signal sequence at the corresponding resolution level.

[0071] Physiological signal sequences at each resolution level were subjected to frequency domain decomposition and time domain statistical analysis. Frequency domain decomposition employed Fast Fourier Transform (FFT) to convert the signal to the frequency domain, calculating the power spectral density of each frequency band and extracting the energy proportions of the extremely low-frequency, low-frequency, and high-frequency bands as spectral energy distribution characteristics. Taking heart rate variability analysis as an example, the energy in the extremely low-frequency band reflects the influence of humoral and thermoregulation, the energy in the low-frequency band mainly reflects sympathetic nerve activity, and the energy in the high-frequency band corresponds to parasympathetic nerve activity; the proportional relationship among these three bands can indicate the autonomic nervous system's balance. Time domain statistical analysis calculated the mean, standard deviation, skewness coefficient, and kurtosis coefficient of the signal sequence as statistical moment characteristics. The mean reflects the baseline level of physiological parameters, the standard deviation reflects the fluctuation amplitude, the skewness coefficient describes the symmetry of the distribution, and the kurtosis coefficient characterizes the frequency of extreme values. Combining the spectral energy distribution characteristics with the statistical moment characteristics forms a subset of scale features corresponding to each resolution level.

[0072] There are information redundancy and complementarity among feature subsets at different resolution levels, requiring quantitative analysis to determine the feature transfer weights between levels. The mutual information between feature subsets at adjacent resolution levels is calculated; mutual information measures the statistical dependence between two feature sets. Let the... The hierarchical scale feature subset is , No. The hierarchical scale feature subset is The mutual information between the two The mutual information is calculated using the joint probability distribution and marginal probability distribution. Higher mutual information indicates greater information overlap between adjacent layers; therefore, the weight of layers with high mutual information should be reduced during feature fusion to avoid information redundancy. Inter-layer feature transfer weights are constructed based on mutual information. Specifically, the mutual information is normalized and its reciprocal is taken, resulting in lower weights for layers with high information redundancy and higher weights for layers with strong information independence. The scale feature subsets of each resolution layer are weighted and summed according to the feature transfer weights to achieve effective integration of cross-scale information, generating a cross-scale fusion feature vector. This vector integrates multi-level physiological state information from instantaneous fluctuations to long-term trends, providing a more comprehensive reflection of the patient's physiological stability.

[0073] Processing anesthetic drug data requires modeling and analysis based on pharmacodynamic principles. Anesthetic drug data includes information such as the type of anesthetic drug, dosage, administration time, and route of administration. For commonly used intravenous anesthetics such as propofol and remifentanil, their metabolism in vivo follows first-order kinetics or a multi-compartment model. The elimination half-life is a key parameter describing the rate of drug metabolism, representing the time required for the drug's plasma concentration to decrease to half of its initial concentration. Based on the dosage, administration time, and elimination half-life, the residual drug concentration at the current moment can be calculated. Taking a single-compartment model as an example, the residual drug concentration... It can be calculated using an exponential decay function, which involves the initial concentration, elimination rate constant, and time interval from the dosing time. For multiple dosings, the residual concentrations from each dosing need to be summed.

[0074] Residual drug concentration is closely related to a patient's anesthesia recovery status, but different drugs have different mechanisms of influence on consciousness and muscle strength, requiring conversion through a pharmacological action function. The pre-defined pharmacological action function, established based on clinical pharmacology studies, describes the dose-response relationship between drug concentration and physiological effects. The consciousness recovery index reflects the degree to which a patient recovers from anesthesia to wakefulness, and is generally negatively correlated with the residual concentration of sedative drugs. When the sedative drug concentration is higher than the consciousness recovery threshold, the consciousness recovery index is low, indicating that the patient is still in a state of deep sedation; when the drug concentration drops below the threshold, the consciousness recovery index gradually increases, and the patient begins to exhibit spontaneous consciousness. The muscle strength recovery index reflects the degree of relief from neuromuscular blockade and is related to the residual concentration of muscle relaxants. Muscle relaxants competitively block acetylcholine receptors at the neuromuscular junction, leading to a decrease in muscle contractility. The muscle strength recovery index can be quantitatively assessed using the ratio of four cascaded stimuli; when the ratio recovers to above 0.9, muscle strength is considered to have essentially returned to normal. Combining the calculated consciousness recovery index and the muscle strength recovery index forms an anesthesia recovery vector, which quantifies the patient's current anesthesia recovery status.

[0075] The cross-scale fusion feature vector and the anesthesia recovery vector are concatenated to obtain the final risk feature vector. This concatenation operation links the two vectors dimensionally, forming a higher-dimensional comprehensive feature representation. The cross-scale fusion feature vector primarily reflects the patient's physiological stability and stress response, while the anesthesia recovery vector reflects the degree to which drug residues inhibit the patient's protective reflexes and autonomic regulation. The combination of these two vectors comprehensively characterizes the intrinsic risk factors faced by the patient during transport. The dimensionality of the risk feature vector is typically between 50 and 100, preserving rich risk indication information while avoiding the computational complexity issues caused by excessively high dimensionality. This risk feature vector will serve as the core input for subsequent dynamic risk propagation map construction and route planning, directly impacting the safety assessment and optimization decisions of the transport route.

[0076] In practical applications, the feature extraction process needs to consider data quality and real-time requirements. Physiological monitoring data suffers from sensor noise, motion artifacts, and signal loss, necessitating preprocessing before feature extraction, including filtering and denoising, artifact detection, and missing value imputation. The accuracy of anesthetic medication data depends on the integrity of the electronic anesthesia recording system, requiring accurate recording of key information such as administration time and dosage. The computational efficiency of the feature extraction algorithm is also a crucial consideration, requiring a balance between feature representation capability and computation time to ensure that risk feature vectors can be generated during the transport preparation phase, providing timely decision support for subsequent route planning.

[0077] Pharmacokinetic modeling is performed on the anesthetic drug data to calculate the drug elimination half-life and current residual drug concentration. Based on a preset pharmacological function, the consciousness recovery index and muscle strength recovery index are calculated respectively, generating an anesthesia recovery vector, including:

[0078] Extract the dosing time sequence, dosing dose sequence, and anesthetic drug type from the anesthetic drug data, and obtain the standard clearance rate and standard distribution volume from the preset pharmacokinetic parameter library according to the anesthetic drug type;

[0079] The patient's weight, age, sex, and serum creatinine level are obtained. The glomerular filtration rate is calculated, and the standard clearance rate is individually corrected to obtain an individualized clearance rate. The drug elimination half-life is calculated based on the individualized clearance rate and the standard volume of distribution.

[0080] For each dosing time in the dosing time sequence, the time interval from the dosing time to the current time is calculated. The initial blood drug concentration is calculated based on the dosing dose and the standard distribution volume. The residual concentration component is calculated using a first-order kinetic equation in combination with the time interval and the drug elimination half-life. The current drug residual concentration is obtained by summing the residual concentration components.

[0081] An S-shaped curve function containing the half-maximal effect concentration is established as a preset pharmacological action function. The current drug residual concentration is substituted into the S-shaped curve function, and the half-maximal effect concentrations related to consciousness and muscle strength are set respectively. The consciousness recovery index and muscle strength recovery index are calculated and combined with the current drug residual concentration and the drug elimination half-life to form an anesthesia recovery vector.

[0082] In the risk assessment of transporting elderly patients recovering from anesthesia, pharmacokinetic modeling is a core component in evaluating the patient's recovery status. Anesthetic drug data typically includes multiple dosing records, with the time, dosage, and type of anesthetic drug used for each administration accurately recorded. The dosing time sequence is extracted from the anesthetic drug data; this sequence is arranged chronologically, recording all dosing points from the start of anesthesia induction to the end of the maintenance phase. The dosage sequence corresponds one-to-one with the dosing time sequence, recording the specific dosage for each administration, typically in milligrams or micrograms. Identification of the anesthetic drug type is crucial because different anesthetic drugs have significantly different pharmacokinetic characteristics. Common anesthetic drugs include intravenous anesthetics such as propofol, remifentanil, and sufentanil, as well as inhaled anesthetics such as sevoflurane and isoflurane.

[0083] The pre-defined pharmacokinetic parameter library stores the pharmacokinetic parameters of various anesthetic drugs under standard physiological conditions. Standard clearance reflects the rate at which the drug is eliminated from the body, usually measured in liters per hour or milliliters per minute. Standard volume of distribution describes the distribution range of the drug in the body, reflecting the degree of distribution of the drug from the blood to tissues, measured in liters or liters per kilogram of body weight. These standard parameters are usually established based on pharmacokinetic study data from healthy adults. However, due to the decline in physiological function, the actual pharmacokinetic parameters of elderly patients often deviate from the standard values, thus requiring individualized adjustments.

[0084] A patient's weight directly affects the volume of distribution of drugs; the greater the weight, the wider the distribution area of ​​the drug. Age is an important factor affecting drug metabolism; elderly patients generally have decreased liver and kidney function, leading to reduced drug clearance. Gender differences are reflected in differences in body fat percentage and metabolic enzyme activity; female patients typically have a higher body fat percentage, resulting in a relatively larger volume of distribution for lipid-soluble anesthetic drugs. Serum creatinine is a key indicator for assessing renal function, and glomerular filtration rate (GFR) can be calculated from serum creatinine levels. GFR is calculated using the CKD-EPI formula or the Cockcroft-Gault formula, which comprehensively consider age, gender, weight, and serum creatinine levels. The calculated GFR is expressed in milliliters per minute per 1.73 square meters. A normal adult's GFR is usually above 90, while elderly patients often have a GFR below 60, indicating impaired renal function.

[0085] The correction for individualized clearance is based on the ratio of glomerular filtration rate (GFR) to standard clearance. For anesthetic drugs primarily excreted by the kidneys, individualized clearance is positively correlated with GFR. A correction factor is introduced during the correction process, determined by the ratio of the patient's GFR to the standard GFR. The standard GFR is typically taken as 100 ml / min / 1.73 m² as a reference value. Individualized clearance equals the standard clearance multiplied by the correction factor. When the patient's GFR is 50%, the correction factor is 0.5, and the individualized clearance is correspondingly reduced to half of the standard clearance.

[0086] The elimination half-life of a drug is an important parameter describing the rate at which a drug is eliminated from the body. It is defined as the time required for the blood drug concentration to decrease to half of its initial concentration. The calculation of the elimination half-life is based on first-order kinetics, and its value is related to individualized clearance rate and standard volume of distribution. Specifically, the elimination half-life... With individualized clearance rate and standard distribution volume The relationship between them can be represented as The formula 0.693 is an approximation of the natural logarithm 2. This formula indicates that the lower the clearance rate or the larger the volume of distribution, the longer the drug's elimination half-life and the longer it remains in the body.

[0087] For each dosing time in the dosing time sequence, its contribution to the drug residual concentration at the current time needs to be calculated, with time intervals between each dosing time. Defined as the first The time difference between the time of the first dose and the current time, expressed in hours or minutes. Initial blood drug concentration. The calculation is based on the dosage. and standard distribution volume The calculation formula is: This formula assumes that the drug is rapidly distributed throughout the body after administration, and the blood drug concentration reaches the initial peak.

[0088] The first-order kinetic equation describes the exponential decay of drug concentration over time. For the second-order kinetic equation... The residual concentration at the current time after the first dose. Through formula The calculation yielded the following result. The exponential term in the formula reflects the rate of drug concentration decay; the longer the time interval or the shorter the elimination half-life, the lower the residual concentration component. When the time interval equals one elimination half-life, the residual concentration component decreases to 50% of the initial blood drug concentration; when the time interval equals two elimination half-lives, the residual concentration component decreases to 25% of the initial blood drug concentration.

[0089] Current drug residue concentration It is the sum of the residual concentrations at all administration times, calculated using the following formula: ,in This represents the number of administrations. This cumulative process reflects the additive effect of multiple administrations; even if the residual concentration of a single administration is already low, the accumulation of multiple administrations can still lead to a high total residual concentration. For continuously infused anesthetic drugs, the administration time sequence can be discretized into multiple short-interval administration events. The dose administered in each event is equal to the infusion rate multiplied by the time interval, and the residual concentration is calculated using the same method.

[0090] The preset pharmacological action function adopts an S-curve function, also known as the Emax model or Hill equation. This function can describe the nonlinear relationship between drug concentration and pharmacological effect. The mathematical expression of the S-curve function is as follows: ,in For pharmacological effects, For maximum pharmacological effect, The half-number effect concentration, Hill coefficient describes the steepness of the curve and the half-maximal effect concentration. This refers to the drug concentration at which 50% of the maximum effect is produced; different pharmacological effects correspond to different half-maximum concentration values.

[0091] The calculation of the consciousness recovery index requires setting the consciousness-related half-maximum effect concentration. Consciousness recovery index It reflects the degree of recovery of the patient's consciousness, and its calculation is based on the current residual drug concentration. Substitute the S-curve function and the pharmacological effect Converted to a recovery index. The conversion method is as follows: When the drug concentration is much higher than the half-maximal effect concentration, the pharmacological effect is close to the maximum value and the consciousness recovery index is close to 0, indicating that the patient is still in a state of deep sedation; when the drug concentration is much lower than the half-maximal effect concentration, the pharmacological effect is close to 0 and the consciousness recovery index is close to 1, indicating that the patient's consciousness has been basically recovered.

[0092] The calculation of the muscle strength recovery index is similar, but it requires setting the muscle strength-related half-maximal effect concentration. The half-maximal effective concentration (WIC) of muscle relaxants is generally lower than that of sedatives because muscle strength recovery is more sensitive to drug concentration. Muscle strength recovery index. The calculation is also based on the S-curve function, through... The patient's muscle strength recovery index is close to 1, indicating that the patient's muscle strength has recovered to the level where they can breathe independently and maintain an open airway. This is an important basis for determining whether the patient is suitable for transport.

[0093] The anesthesia recovery vector combines multiple key parameters into a multidimensional vector, typically with four dimensions, including the current residual drug concentration, drug elimination half-life, recovery of consciousness index, and muscle strength recovery index. The vector is represented as follows: superscript This indicates a transpose. This vector comprehensively reflects the patient's anesthesia recovery status. The current residual drug concentration and elimination half-life describe the dynamic process of drug metabolism, while the consciousness recovery index and muscle strength recovery index directly reflect the degree of physiological function recovery. As an important component of the risk feature vector, the anesthesia recovery vector, together with features extracted from physiological monitoring data, is used for subsequent risk assessment and pathway planning.

[0094] Based on the risk feature vector and the spatial topology information in the transit environment data, a risk propagation graph integrating the time-varying characteristics of risk and spatial constraints is constructed. The edge weights are defined by path accessibility conditions and emergency response accessibility constraints, including:

[0095] Spatial node coordinates, node types and connection relationships are extracted from the transport environment data to construct an initial topology map. The risk feature vector is expanded in time series to obtain a risk evolution trajectory sequence. The risk decay coefficient corresponding to each moment is calculated based on the risk time-varying decay function.

[0096] The initial risk state of the risk evolution trajectory sequence is injected into the starting node. The spatial length and estimated passage time are calculated for each edge. The risk state value at the corresponding moment is selected from the risk evolution trajectory sequence, and the risk arrival value of the edge endpoint is calculated in combination with the corresponding risk decay coefficient.

[0097] Obtain the clearance dimensions, turning radius, and ground friction coefficient of each side's corresponding channel. Combine the minimum passage section requirements and maximum climbing capacity threshold of the transfer equipment to calculate the passage feasibility index. When the passage feasibility index is lower than the preset passage lower limit, a penalty weight is assigned to the side.

[0098] Extract the location of the emergency response unit and the standard response delay, calculate the shortest path length from the midpoint of each side to the nearest emergency response unit, calculate the emergency intervention time based on the shortest path length and the standard response delay, and normalize it to obtain the emergency accessibility coefficient.

[0099] The risk arrival value, the travel feasibility index, and the emergency accessibility coefficient are weighted and fused to generate edge weights, and a dynamic risk propagation graph is constructed.

[0100] After obtaining the risk feature vector, it is necessary to combine the patient's physiological risk status with the physical spatial structure of the hospital to form a graphical model that reflects the propagation pattern of risk in space. Transfer environment data is typically stored in Building Information Modeling (BIM) or Geographic Information System (GIS) formats, containing the three-dimensional coordinates of key locations such as corridor nodes, elevator lobbies, ward entrances, and nurse stations, as well as the connectivity between these locations. The coordinate information of these spatial nodes is extracted, and the type label of each node is recorded (e.g., "elevator waiting area," "corner," "ward entrance / exit"). An initial topology graph is then constructed based on the passageway connections in the architectural drawings. This topology graph is represented as an undirected or directed graph, where nodes represent key locations in the space, and edges represent passable path segments.

[0101] The risk feature vector reflects the patient's risk status at the start of transport. However, during transport, the patient's physiological state changes over time. Factors such as anesthetic drug metabolism, fluctuations in physiological indicators, and environmental stress responses all contribute to the dynamic evolution of risk levels. To capture this time-varying characteristic, the risk feature vector is temporally unfolded. Specifically, based on the estimated total transport duration, the time axis is divided into several time steps, each corresponding to a snapshot of the risk status. Through physiological model predictions or historical data statistics, the patient's risk status value at each time step is calculated, forming a risk evolution trajectory sequence. This sequence records the trend of the patient's risk level changes from the start of transport to the expected end.

[0102] In actual transport, the impact of the initial risk gradually weakens over time with continuous monitoring by medical staff. A time-varying risk decay function is introduced to quantify this process. This function typically uses an exponential decay form, with the time interval as input and the risk decay coefficient as output. The decay coefficient ranges from 0 to 1; the longer the time interval, the smaller the decay coefficient, indicating a weaker residual impact of the initial risk. This function can be used to calculate the risk decay coefficient at each moment in the risk evolution trajectory sequence, providing a time-dimensional adjustment parameter for subsequent risk propagation calculations.

[0103] When constructing a dynamic risk propagation map, the initial risk state of the risk evolution trajectory sequence needs to be injected into the starting node. The starting node is usually the exit of the anesthesia recovery room or operating room where the patient is currently located. The risk value of this node is set as the first element of the risk evolution trajectory sequence, representing the risk level at the start of the transfer. For each edge in the topology map, its spatial length is calculated. The spatial length is calculated using the Euclidean distance formula based on the three-dimensional coordinates of the nodes at both ends of the edge. Considering that the moving speed of the transfer equipment is affected by factors such as the width of the passage, ground conditions, and turning angle, it is necessary to further estimate the time required to pass through the edge. The estimated passage time can be obtained through statistical analysis of historical transfer data, or calculated based on the standard moving speed of the transfer equipment and the path complexity coefficient.

[0104] After obtaining the estimated transit time, the risk state value at the corresponding moment is selected from the risk evolution trajectory sequence. For example, if the estimated transit time for a certain edge is 2 minutes, and the starting point of this edge corresponds to the 5th minute after the start of transport, then the endpoint of this edge corresponds to the 7th minute, and the risk state value at the 7th minute is extracted from the risk evolution trajectory sequence. Simultaneously, based on the time interval between this moment and the start of transport, the corresponding risk decay coefficient is calculated using a time-varying risk decay function. Multiplying the extracted risk state value by the risk decay coefficient yields the risk arrival value at the edge endpoint. This risk arrival value reflects the patient's expected risk level upon arrival at the edge endpoint, comprehensively considering the time evolution and decay effect of risk.

[0105] The physical feasibility of the transfer route is a fundamental constraint for route planning. Hospital internal corridors are complex and diverse, with some corridors having limited width, some areas containing steps or ramps, and some corners having small turning radii. Obtaining the clear dimensions of each side of the corridor, i.e., the effective width and height of the corridor, ensures that transfer equipment can pass smoothly. The turning radius reflects the curvature of the corridor at the corner; a turning radius that is too small will prevent the transfer bed or wheelchair from turning smoothly. The coefficient of friction reflects the resistance of the ground material to the movement of the transfer equipment; smooth tile floors and rough, non-slip floors have different coefficients of friction.

[0106] Based on the technical parameters of the transfer equipment, a feasibility index is calculated. The minimum passage cross-section requirement for the transfer equipment is determined by its external dimensions, including length, width, and height. If the clearance dimensions of the passage are smaller than the minimum passage cross-section requirement, the passage is impassable. The maximum climbing capacity threshold reflects the maximum climbing angle of the transfer equipment on a ramp. If the slope of the passage exceeds this threshold, the equipment will be unable to pass. The feasibility index comprehensively considers the clearance dimension matching degree, turning radius adequacy, ground friction coefficient suitability, and slope feasibility, and is calculated as a weighted value between 0 and 1. When the feasibility index is lower than the preset lower limit, it indicates that there is a significant physical obstacle on the path segment, and a penalty weight is assigned to that side. The penalty weight is usually set to a large positive value, causing the path search algorithm to tend to avoid these unfavorable paths during the optimization process.

[0107] Emergency response capability is crucial for the safe transport of elderly patients. During transport, if a patient experiences a sudden physiological abnormality, rapid intervention by medical staff or an emergency medical team is necessary. Hospitals typically have multiple emergency response units, such as emergency medical team outposts, nurses' stations, and doctors' on-call rooms. The coordinates of these outpost locations and the standard response time from receiving a call to arriving at the scene are extracted. The standard response time includes personnel reaction time, equipment preparation time, and movement time.

[0108] For each edge in the topology graph, the shortest path length from its midpoint to the nearest emergency response unit is calculated. The midpoint represents the patient's average position on that path segment. The shortest path length is calculated on the topology graph using a graph search algorithm (such as Dijkstra's algorithm). Based on the shortest path length, the movement speed of the emergency response unit, and the standard response delay, the emergency intervention time is calculated. The emergency intervention time reflects the total time required from the occurrence of a sudden situation on the patient's path segment to the arrival of the emergency team on-site. To facilitate subsequent weight fusion, the emergency intervention time is normalized to obtain an emergency accessibility coefficient. The normalization method can use a linear mapping or a sigmoid function to map the emergency intervention time to the interval between 0 and 1. The shorter the emergency intervention time, the higher the emergency accessibility coefficient, indicating a stronger emergency response capability for that path segment.

[0109] After obtaining the risk arrival value, traversability feasibility index, and emergency accessibility coefficient, edge weights are generated through weighted fusion. The fusion formula adopts a linear weighting form, where each of the three indicators is multiplied by its corresponding weight coefficient and then summed. The weight coefficients need to be adjusted based on clinical experience and actual needs. Generally, the weight coefficient for the risk arrival value is larger because the patient's physiological safety is the primary consideration; the weight coefficient for the traversability feasibility index is secondary, ensuring the physical feasibility of the route; and the weight coefficient for the emergency accessibility coefficient is relatively smaller, serving as a supplementary guarantee for safety redundancy. The generated edge weights comprehensively reflect the integrated characteristics of the route segment in the three dimensions of risk propagation, physical traversability, and emergency support. After assigning weights to all edges, a dynamic risk propagation graph is formed. This graph not only includes the spatial topology but also integrates the time-varying characteristics of patient risk, physical environmental constraints, and emergency response capabilities, providing a comprehensive decision-making basis for subsequent route search and optimization.

[0110] The risk feature vector is expanded over time to obtain a risk evolution trajectory sequence, and the risk decay coefficient at each time point is calculated based on the time-varying risk decay function, including:

[0111] A time-dimensional projection transformation is performed on the risk feature vector, and the start and end times of the risk observation window are set. Dynamic sampling is performed between the start and end times with an adaptive time step, and the adaptive time step adjusts the sampling density according to the fluctuation amplitude of each component of the risk feature vector.

[0112] At each sampling time, the current component value of the risk feature vector is extracted and the risk change rate relative to the starting time is calculated. The instantaneous risk intensity at each sampling time is calculated by combining the current component value and the instantaneous risk intensity at each sampling time is connected in time sequence to form a risk evolution trajectory sequence.

[0113] The risk evolution trajectory sequence is piecewise linearly fitted to obtain multiple risk decay curves. The slope of each curve is calculated and the inflection point is identified. Based on the inflection point, the risk time-varying decay function is divided into multiple decay stages, and differentiated decay index parameters are set.

[0114] For any sampling moment in the risk evolution trajectory sequence, determine the decay stage and substitute the corresponding decay index parameter into the risk time-varying decay function to obtain the risk decay coefficient for each sampling moment.

[0115] After obtaining the risk feature vector of the patient to be transferred, it is necessary to perform a temporal unfolding of the vector to capture the dynamic evolution of risk over time. The risk feature vector typically contains a comprehensive representation of multiple physiological indicators and pharmacological parameters, which exhibit significant time-varying characteristics during anesthesia recovery. To accurately characterize the evolution of risk, a time-dimensional projection transformation is performed on the risk feature vector, mapping the originally static feature representation onto the time axis to form a traceable dynamic sequence.

[0116] When performing time-dimensional projection transformation, the time range of the risk observation window needs to be clearly defined. The start time is usually set at the moment when the anesthesia and surgery end and the patient enters the recovery phase. At this time, the patient's anesthetic drug concentration is at a high level, and their physiological state has not yet fully stabilized. The end time is set based on clinical experience at the expected completion time of transfer, generally covering the entire time from the recovery room to the target ward, usually between 30 and 120 minutes. After determining the observation window, dynamic sampling using an adaptive time step is adopted, rather than fixed-interval sampling. The core idea of ​​the adaptive time step is to dynamically adjust the sampling density according to the fluctuation amplitude of each component of the risk feature vector, increasing the sampling frequency during periods of drastic risk changes and decreasing the sampling frequency during periods of relatively stable risk, thereby ensuring the capture of key risk changes while avoiding redundant data.

[0117] In practice, it is necessary to calculate the difference between each component of the risk feature vector at adjacent time points. This difference reflects the instantaneous rate of change of each physiological or pharmacological parameter. For physiological indicators such as blood pressure and heart rate, the fluctuation amplitude is usually quantified by standard deviation or coefficient of variation; for pharmacological parameters such as drug residue concentration, the fluctuation amplitude is characterized by concentration gradient. The fluctuation amplitudes of each component are weighted and summed to obtain a comprehensive fluctuation index. When the comprehensive fluctuation index exceeds a preset threshold, the current time step is shortened to 50% to 70% of the baseline step size to increase the sampling density; when the comprehensive fluctuation index is below the threshold, the time step is extended to 150% to 200% of the baseline step size to reduce the sampling frequency. The baseline step size is usually set to 2 to 5 minutes, with the specific value determined based on the patient's risk level and transport distance.

[0118] At each sampling time, the current component value of the risk feature vector is extracted. These component values ​​include real-time monitored physiological parameters such as systolic blood pressure, diastolic blood pressure, heart rate, blood oxygen saturation, and respiratory rate, as well as drug residual concentration and anesthesia recovery index calculated through pharmacokinetic models. The current component value is compared with the baseline value at the start time, and the rate of change of risk relative to the start time is calculated. The rate of change of risk is defined as the difference between the current component value and the baseline value, divided by the baseline value, and then multiplied by the risk weight coefficient corresponding to that component. The risk weight coefficient is determined based on clinical experience and statistical data; for example, the weight coefficient for blood oxygen saturation is usually higher than that for heart rate because a decrease in blood oxygen saturation poses a more direct threat to patient safety.

[0119] After obtaining the risk change rate for each component, the instantaneous risk intensity at each sampling time is calculated by combining the current component value. The instantaneous risk intensity comprehensively considers both the absolute level and relative trend of the current physiological state, providing a more complete reflection of the patient's risk status at that moment. During calculation, the current values ​​of each component are normalized, mapping physiological parameters of different dimensions to a unified numerical range, typically between 0 and 1. The normalized component values ​​are then weighted and fused with the corresponding risk change rate, with the fusion weight dynamically adjusted based on the clinical importance of the component. For example, when the change rate of a certain physiological parameter increases significantly, the weight of that parameter in the instantaneous risk intensity calculation is correspondingly increased to highlight its contribution to the overall risk. The instantaneous risk intensities calculated at each sampling time are connected chronologically to form a risk evolution trajectory sequence. This sequence visually demonstrates the dynamic change process of the patient's risk level over time, providing fundamental data for subsequent risk decay analysis.

[0120] Piecewise linear fitting of the risk evolution trajectory sequence aims to identify the phased characteristics of risk changes. The piecewise linear fitting employs the least squares method, decomposing the continuous risk evolution trajectory into multiple straight line segments while minimizing the fitting error. Each straight line segment represents a relatively stable stage of risk change, and its slope reflects the rate of risk decay or growth during that stage. During the fitting process, a minimum segment length constraint is set to avoid overfitting caused by excessive segmentation. The minimum segment length is typically set to 10% to 15% of the total observation window duration to ensure that each stage has sufficient time span and data support.

[0121] After completing piecewise linear fitting, the slope of each curve segment is calculated, and inflection points are identified. An inflection point is defined as the moment when the slope difference between two adjacent straight line segments becomes significant, typically determined by the rate of change of the slope exceeding a preset threshold. Inflection points signify a shift in risk evolution patterns, such as transitioning from a rapid decay phase to a slow decay phase, or from a stable phase to a risk rebound phase. The identified inflection points divide the time-varying risk decay function into multiple decay phases, each corresponding to different risk evolution characteristics.

[0122] Based on the slope characteristics of each decay stage, differentiated decay index parameters are set. These parameters control the rate and pattern of risk decay; a larger decay index corresponds to a faster rate of risk reduction, while a smaller index corresponds to a slower rate. In the early stages of anesthesia recovery, the patient's drug concentration is high, and physiological state fluctuates significantly, resulting in a relatively rapid rate of risk decay. At this time, a larger decay index is set, typically between 1.5 and 2.5. As recovery progresses, the drug is gradually metabolized, and the physiological state stabilizes, slowing the rate of risk decay. At this time, a smaller decay index is set, typically between 0.8 and 1.2. For stages where risk rebound occurs, the decay index is negative, indicating that the risk is increasing during this stage.

[0123] For any sampling moment in the risk evolution trajectory sequence, determine the decay stage to which that moment belongs. The method involves comparing the current moment with each inflection point moment to determine which two adjacent inflection points the current moment falls between, thus identifying the stage. After determining the stage, substitute the decay exponent parameter corresponding to that stage into the time-varying risk decay function. The time-varying risk decay function typically uses exponential decay, taking the time difference between the current moment and the start moment of the stage as the independent variable and the decay exponent parameter as the exponent coefficient to calculate the risk decay coefficient for that moment. The risk decay coefficient ranges from 0 to 1; a smaller value indicates a higher degree of risk decay, and a larger value indicates a lower degree of risk decay.

[0124] Through the above process, a corresponding risk decay coefficient is calculated for each sampling moment in the risk evolution trajectory sequence. These risk decay coefficients not only reflect the overall decay trend of risk over time but also capture the differential characteristics of risk evolution at different stages. In the subsequent construction of the dynamic risk propagation map and route planning, the risk decay coefficient serves as an important time-varying parameter, used to dynamically adjust the risk weights of each route node. This ensures that the transport route planning fully considers the time-varying characteristics of risk, thereby providing a safer and more reliable transport solution for elderly patients.

[0125] Based on the dynamic risk propagation map, a constrained path search is performed. A set of feasible paths satisfying these dual constraints is generated by setting a risk accumulation cap and a transfer time cap. The risk volatility variance and emergency intervention coverage of each feasible path are calculated. A path transfer scoring function is constructed, and target transfer paths are selected, including:

[0126] Extract the node risk propagation value and edge connection relationship from the dynamic risk propagation graph, calculate the difference between the risk propagation values ​​of adjacent nodes as the risk gradient, and generate a weighted adjacency matrix by weighting the edge connection relationship.

[0127] Based on the risk accumulation limit and the transit time limit, the maximum number of high-risk nodes and the maximum number of path hops are determined, and used as constraints to perform a breadth-first search on the weighted adjacency matrix to generate a set of feasible paths.

[0128] For each feasible path, the risk propagation value of each node is extracted and organized into a risk sequence. The risk sequence is differentially processed and the standard deviation of the differential result is calculated as the risk fluctuation variance. Nodes on the path where emergency equipment or emergency personnel are configured are identified and the maximum interval distance between them is calculated. The emergency intervention coverage rate is calculated based on the ratio of the maximum interval distance to the total path length.

[0129] The cumulative risk value, total transit time, risk volatility variance, and emergency intervention coverage rate are used as scoring indicators. Weighting coefficients are assigned and a path transit scoring function is constructed. The path transit score of each feasible path is calculated, and the feasible path with the highest path transit score is selected as the target transit path.

[0130] After obtaining the dynamic risk propagation map, key information needs to be extracted to support subsequent pathfinding. The dynamic risk propagation map represents the risk status and propagation relationships of each node in the transport space in a graph structure. Each node corresponds to a spatial location on the transport path, and its attributes include the risk propagation value for that location. This value comprehensively reflects the physiological risks, environmental risks, and risks of delayed emergency response faced by the patient at that location. Edge connections describe the accessibility and traversal conditions between adjacent nodes, and the edge weights are jointly determined by the path traversal conditions and emergency response accessibility.

[0131] When extracting node risk propagation values ​​from a dynamic risk propagation graph, all nodes in the graph are traversed and their risk propagation value attributes are read. These values ​​are then organized into node risk vectors. For adjacent nodes, the difference in their risk propagation values ​​is calculated as the risk gradient, which reflects the rate of change of risk level as the node moves along the path. If a node... Risk propagation value Adjacent nodes Risk propagation value Then the risk gradient between the two nodes is The larger the absolute value of the risk gradient, the more drastic the risk changes along that path segment, and the more attention it needs to be paid to it during path planning.

[0132] The weighted processing of edge connections requires comprehensive consideration of both accessibility and emergency response accessibility. Accessibility includes physical constraints such as passage width, slope, and obstacle distribution, while emergency response accessibility involves factors such as whether the path segment is easily accessible to emergency equipment and whether emergency personnel are stationed there. For connecting nodes... and edge Its weight This can be represented as a weighted combination of traffic resistance and emergency response delay. Traffic resistance is quantified based on factors such as channel width, turning radius, and ground material, while emergency response delay is estimated based on the distance of the path segment to the nearest emergency resource point and the response time. All edge weights are organized into a weighted adjacency matrix. Matrix elements Represents a node To the node The edge weight is such that if two nodes are not directly connected, the corresponding element is either infinity or zero.

[0133] Before performing a path search, constraints need to be determined based on the risk accumulation limit and the transit time limit. Risk accumulation limit The maximum allowed cumulative risk value along the entire route is defined, based on the patient's risk tolerance and safe transport standards. Maximum transport time is also specified. This limits the maximum allowable time for the transfer process, preventing patients from spending excessively long periods in transit due to excessively long routes. Based on these two limits, the maximum number of high-risk nodes and the maximum number of route hops can be derived. Maximum number of high-risk nodes This indicates the maximum number of high-risk nodes allowed to be traversed on the path. High-risk nodes are defined as nodes whose risk propagation value exceeds a preset threshold. The node. Maximum path hop count. This indicates the maximum number of nodes allowed to be traversed from the origin to the destination. This value is estimated based on the upper limit of transit time and the average travel time between nodes.

[0134] Based on the above constraints, a breadth-first search is performed on the weighted adjacency matrix to generate a set of feasible paths. The breadth-first search starts from the starting node and expands neighboring nodes layer by layer, checking whether the current path satisfies the constraints at each expansion. Specifically, a queue of nodes to be expanded is maintained, initially adding the starting node to the queue. Each time, a node is taken from the queue, and all its neighboring nodes are traversed. For each neighboring node, the cumulative risk value, path hop count, and number of high-risk nodes after adding it to the current path are calculated. If the cumulative risk value does not exceed... The number of hops in the path did not exceed And the number of high-risk nodes did not exceed If a path reaches its destination node, it is added to the queue and its path information is recorded. When a path reaches its destination node, it is added to the set of feasible paths. The search process continues until the queue is empty or a sufficient number of feasible paths have been found.

[0135] For each path in the set of feasible paths, risk volatility variance and emergency intervention coverage need to be calculated to assess path quality. For each path... ,in Given the path length, the risk propagation values ​​of each node are extracted and organized into a risk sequence. By performing a difference operation on the risk sequence, we obtain the sequence of risk changes between adjacent nodes. ,in The standard deviation of the differenced results is used as the variance of risk volatility. This indicator reflects the degree of fluctuation in risk levels along the route. The smaller the variance in risk fluctuation, the more stable the changes in risk levels along the route, and the smaller the risk shocks faced by patients during transport.

[0136] Calculating emergency intervention coverage requires identifying nodes along the path that are equipped with emergency equipment or personnel. All nodes along the path are traversed, and each node is checked to see if it is equipped with emergency equipment such as defibrillators, first-aid supplies, and oxygen supply equipment, or if it has emergency personnel such as nurses or doctors stationed there. Nodes with emergency resources are marked as emergency nodes. The path segment lengths between adjacent emergency nodes are calculated, and the maximum value is taken as the maximum interval distance. Total path length This is the sum of distances between all adjacent nodes. Emergency intervention coverage. Defined as the complement of the ratio of the maximum interval distance to the total path length, i.e. A higher emergency intervention coverage rate indicates a more even distribution of emergency resources along the route, meaning that patients can quickly receive emergency intervention if an emergency occurs at any point during transport.

[0137] When constructing a route transfer scoring function, four scoring indicators need to be comprehensively considered: cumulative risk value, total transfer time, risk volatility variance, and emergency intervention coverage. (Cumulative risk value) This is the sum of risk propagation values ​​at all nodes along the path, reflecting the overall risk level of the entire path. Total transit time. This represents the sum of travel times between all adjacent nodes along the path, reflecting transfer efficiency. To ensure comparability of the indicators, each indicator is normalized, mapping it to the range of 0 to 1. For cumulative risk value and risk volatility variance, inverse normalization is used, meaning the smaller the value, the closer it is to 1 after normalization. For emergency intervention coverage rate, its original value is used directly as the normalization result. For total transfer time, inverse normalization is used, with the upper limit of transfer time as a reference.

[0138] Path transfer scoring function Defined as the weighted sum of all normalized indicators, with weighting coefficients set according to clinical priority and safety requirements. The weighting coefficient for the cumulative risk value. The weighting coefficients for risk volatility variance are typically set relatively high because controlling the overall risk level is the primary objective of path planning. Secondly, stable risk changes help reduce patients' stress responses. Weighting coefficients for emergency intervention coverage. This reflects the level of importance attached to emergency response capabilities. The weighting coefficient for total transfer time. Relatively small, balancing efficiency with safety. Each weighting coefficient satisfies... Calculate the transit score for each path in the feasible path set, and select the path with the highest score as the target transit path. If multiple paths have the same score, prioritize the path with the lower cumulative risk value. If the cumulative risk values ​​are still the same, select the path with the shorter total transit time.

[0139] Once the target transfer route is determined, the route information is output to the transfer execution system, including the location coordinates of each node on the route, estimated arrival time, risk propagation value, and emergency resource allocation. Transfer personnel then perform the actual transfer operation according to the target route, continuously collecting real-time physiological monitoring data of the patient during the transfer process and comparing it with the risk level predicted during route planning. If the actual risk is significantly higher than the predicted value, a route replanning process is triggered, re-searching for the optimal route based on the current location and real-time risk status.

[0140] Based on the risk accumulation limit and the transit time limit, the maximum number of high-risk nodes and the maximum path hop count are determined, and used as constraints to perform a breadth-first search on the weighted adjacency matrix to generate a set of feasible paths, including:

[0141] Extract node risk propagation values ​​from the weighted adjacency matrix and sort them in descending order. Accumulate the sorted risk propagation values ​​until the accumulated value reaches the risk accumulation limit. Use the number of nodes involved in the accumulation process as the maximum number of high-risk nodes. Calculate the expected path jump time based on the edge weight distribution characteristics of the weighted adjacency matrix. Determine the maximum number of path jumps based on the ratio of the expected time to the upper limit of the transit time.

[0142] During the breadth-first search process, a constraint status identifier is established for each path. The constraint status identifier records the number of high-risk nodes currently included in the path and the number of path hops already executed. When expanding the path, the constraint status identifier is dynamically updated according to the risk propagation value of the candidate node. Only the paths whose constraint status identifiers meet the dual constraint conditions are expanded. All paths that reach the target node and whose constraint status identifiers are compliant are collected to form a feasible path set.

[0143] When determining the constraint parameters for path search, risk propagation values ​​for all nodes are extracted from the weighted adjacency matrix. These risk propagation values ​​reflect the potential risk level of each spatial node under the current patient's physiological state. The extracted risk propagation values ​​are sorted in descending order of numerical value, placing the node with the highest risk at the front. A step-by-step accumulation operation is performed on the sorted risk propagation value sequence, with real-time monitoring to ensure that the accumulated value reaches the preset risk accumulation limit. When the accumulated value first reaches or exceeds the upper limit, the accumulation operation stops, and the number of nodes participating in the accumulation at this point is the maximum number of high-risk nodes. This calculation process ensures that, even when traversing multiple high-risk areas during route planning, the overall risk accumulation will not exceed the safety threshold.

[0144] Determining the maximum path hop count requires statistical analysis based on the edge weight distribution characteristics of the weighted adjacency matrix. Edge weights comprehensively reflect path traversability and emergency response accessibility, and their numerical distribution characteristics can be used to estimate the expected duration of path hops. Specifically, the average value is calculated by statistically analyzing all non-zero edge weights in the weighted adjacency matrix. This represents the expected duration of a single path jump. This mean reflects the average time cost required to move from one node to an adjacent node under the current transit environment. It sets an upper limit on transit time. Divide by expected duration This yields the theoretically maximum number of jumps that can be executed. Considering the delays and uncertainties inherent in actual transport processes, this ratio is rounded down and multiplied by a safety factor. (Typically, the value is between 0.8 and 0.9), which ultimately determines the maximum number of path hops. The formula for calculating this parameter is: ,in For safety reasons, The average edge weights This indicates the floor function.

[0145] When performing a breadth-first search, a constraint state identifier structure is established for each candidate path. This identifier structure contains two key fields: the current high-risk node count. hop count of the executed path In the initial state, the constraint status identifier of the path corresponding to the starting node is... and All are set to 0. During path expansion, when considering a candidate node... When adding a node to the current path, determine whether the risk propagation value of that node exceeds the high-risk node threshold. If the threshold is exceeded, then... Increase by 1; regardless of whether the threshold is exceeded, Increment all by 1 to record the increase in path hop count. The updated constraint status flags must simultaneously satisfy... and Two conditions must be met before a path can be allowed to continue expanding.

[0146] Breadth-first search is implemented using a queue data structure. Initially, the starting node and its corresponding constraint status identifier are added to the queue. In each iteration, a path and its constraint status identifier are retrieved from the head of the queue, and the endpoint of that path is traversed through all its adjacent nodes in the weighted adjacency matrix. For each adjacent node, it is checked whether the node already appears in the current path to avoid forming a loop. If the node does not appear in the path, its constraint status identifier is updated according to its risk propagation value. The specific update rule is: if the risk propagation value of the adjacent node is less than or equal to the constraint status identifier, the constraint status identifier is updated accordingly. Greater than Then the new high-risk node count ,otherwise New path hop count .

[0147] Immediately after updating the constraint status flags, perform a check on the dual constraints. The check conditions are as follows: and If both conditions are met, the expanded path and its updated constraint status are added to the tail of the queue for further processing. If either condition is not met, the expanded path is discarded and no further exploration is performed. This immediate constraint checking mechanism effectively avoids the deep expansion of invalid paths and significantly improves search efficiency.

[0148] During the search process, when the end node of a path reaches the preset target node, a final compliance check needs to be performed on the path. The check includes: whether the path's constraint status indicators are satisfied. and The search process checks for duplicate nodes in the path and whether the total edge weights of the paths are within a reasonable range. Only paths that pass all checks are collected into the feasible path set. To improve search efficiency, an upper limit can be set on the number of paths collected. When the number of feasible paths collected reaches this limit, the search process can be terminated early.

[0149] In the specific implementation of path expansion, the dynamic characteristics of edge weights also need to be considered. Since the transit environment may change over time, edge weights are not fixed. During the search process, edge weights can be dynamically adjusted based on the time difference between the current moment and the search start moment. The adjustment method can employ a time decay function, making the estimated values ​​of edge weights farther from the current moment more uncertain. This dynamic adjustment mechanism allows path search to better adapt to changes in the actual transit environment.

[0150] For the storage and management of constraint state identifiers, a hash table structure can be used to achieve efficient query and update operations. The key of the hash table is a unique identifier for the path, and the value is the corresponding constraint state identifier structure. When expanding a path, the constraint state identifier is quickly retrieved through the path identifier, updated, and then stored back into the hash table. This data structure design keeps the maintenance overhead of constraint states at a low level, ensuring real-time search even in large-scale spatial topology networks.

[0151] After generating a set of feasible paths, the paths in the set undergo initial screening. Screening criteria include: whether the path length is within a reasonable range, whether the distribution of high-risk nodes in the path is too concentrated, and whether the average edge weight of the path significantly deviates from the global mean. These screening conditions further eliminate paths that meet the constraints but have poor practical feasibility, providing a higher-quality candidate set for subsequent path scoring and selection.

[0152] The entire constraint path search process effectively prunes the search space by transforming risk accumulation constraints and time constraints into operable node number limits and hop count limits. The dynamic maintenance mechanism of constraint status identifiers ensures that each path always satisfies the dual constraints during expansion, preventing the generation and propagation of invalid paths. The breadth-first search strategy guarantees that paths with fewer hops are explored first, resulting in a set of feasible paths with good timeliness while satisfying the constraints. This systematic constraint search method can quickly generate a set of feasible paths that meet both safety and timeliness requirements in complex transportation environments, providing a reliable foundation for subsequent path scoring and optimization.

[0153] A second aspect of this invention provides a system for risk assessment and route planning for the transport of elderly patients undergoing anesthesia recovery, comprising:

[0154] The data acquisition unit is used to acquire physiological monitoring data, anesthetic drug data, and transport environment data of the patient to be transported.

[0155] The feature generation unit is used to extract and concatenate features from the physiological monitoring data and the anesthetic drug data to generate a risk feature vector.

[0156] The risk propagation graph construction unit is used to construct a dynamic risk propagation graph that integrates the time-varying characteristics of risk and spatial constraints based on the risk feature vector and the spatial topology information in the transit environment data. The edge weights are defined by the path accessibility conditions and emergency response accessibility constraints.

[0157] The path selection unit is used to perform constrained path search based on the dynamic risk propagation map, generate a set of feasible paths that meet the dual constraints by setting a risk accumulation limit and a transfer time limit, calculate the risk fluctuation variance and emergency intervention coverage of each feasible path, construct a path transfer scoring function, and select the target transfer path.

[0158] The transfer early warning unit is used to transfer the patient to be transferred based on the target transfer route and collect real-time physiological monitoring data, and to issue risk warnings in combination with preset early warning triggering rules.

[0159] A third aspect of the present invention provides an electronic device, comprising:

[0160] processor;

[0161] Memory used to store processor-executable instructions;

[0162] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0163] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0164] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0165] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for anesthesia recovery elderly patient transport risk assessment and path planning, characterized in that, include: Acquire physiological monitoring data, anesthetic drug data, and transport environment data of patients awaiting transfer; Feature extraction and feature concatenation are performed on the physiological monitoring data and the anesthetic drug data to generate a risk feature vector; Based on the risk feature vector and the spatial topology information in the transit environment data, a dynamic risk propagation graph that integrates the time-varying characteristics of risk and spatial constraints is constructed, and the edge weights are defined by the path accessibility conditions and emergency response accessibility constraints. Based on the dynamic risk propagation map, a constrained path search is performed. By setting a risk accumulation limit and a transfer time limit, a set of feasible paths that meet the dual constraints is generated. The risk fluctuation variance and emergency intervention coverage of each feasible path are calculated. A path transfer scoring function is constructed and a target transfer path is selected. Based on the target transport route, the patient to be transported is transported and real-time physiological monitoring data is collected. Risk warnings are then issued in conjunction with preset early warning triggering rules.

2. The method of claim 1, wherein, Feature extraction and feature concatenation are performed on the physiological monitoring data and the anesthetic drug data to generate a risk feature vector, including: The physiological monitoring data is sampled hierarchically according to a preset time scale sequence to obtain physiological signal sequences at multiple resolution levels. Frequency domain decomposition and time domain statistical analysis are performed on each sequence to extract spectral energy distribution features and statistical moment features, forming a scale feature subset corresponding to each resolution level. Calculate the mutual information between scale feature subsets of adjacent resolution levels, construct inter-level feature transfer weights based on the mutual information, and perform weighted summation of scale feature subsets of each resolution level according to the feature transfer weights to generate cross-scale fused feature vectors. Pharmacokinetic modeling is performed on the anesthetic drug data to calculate the drug elimination half-life and current drug residual concentration. Consciousness recovery index and muscle strength recovery index are calculated respectively by combining the preset pharmacological action function to generate an anesthesia recovery vector. The risk feature vector is obtained by concatenating the cross-scale fusion feature vector with the anesthesia recovery vector.

3. The method of claim 2, wherein, Pharmacokinetic modeling is performed on the anesthetic drug data to calculate the drug elimination half-life and current residual drug concentration. Based on a preset pharmacological function, the consciousness recovery index and muscle strength recovery index are calculated respectively, generating an anesthesia recovery vector, including: Extract the dosing time sequence, dosing dose sequence, and anesthetic drug type from the anesthetic drug data, and obtain the standard clearance rate and standard distribution volume from the preset pharmacokinetic parameter library according to the anesthetic drug type; The patient's weight, age, sex, and serum creatinine level are obtained. The glomerular filtration rate is calculated, and the standard clearance rate is individually corrected to obtain an individualized clearance rate. The drug elimination half-life is calculated based on the individualized clearance rate and the standard volume of distribution. For each dosing time in the dosing time sequence, the time interval from the dosing time to the current time is calculated. The initial blood drug concentration is calculated based on the dosing dose and the standard distribution volume. The residual concentration component is calculated using a first-order kinetic equation in combination with the time interval and the drug elimination half-life. The current drug residual concentration is obtained by summing the residual concentration components. An S-shaped curve function containing the half-maximal effect concentration is established as a preset pharmacological action function. The current drug residual concentration is substituted into the S-shaped curve function, and the half-maximal effect concentrations related to consciousness and muscle strength are set respectively. The consciousness recovery index and muscle strength recovery index are calculated and combined with the current drug residual concentration and the drug elimination half-life to form an anesthesia recovery vector.

4. The method of claim 1, wherein, Based on the risk feature vector and the spatial topology information in the transit environment data, a risk propagation graph integrating the time-varying characteristics of risk and spatial constraints is constructed. The edge weights are defined by path accessibility conditions and emergency response accessibility constraints, including: Spatial node coordinates, node types and connection relationships are extracted from the transport environment data to construct an initial topology map. The risk feature vector is expanded in time series to obtain a risk evolution trajectory sequence. The risk decay coefficient corresponding to each moment is calculated based on the risk time-varying decay function. The initial risk state of the risk evolution trajectory sequence is injected into the starting node. The spatial length and estimated passage time are calculated for each edge. The risk state value at the corresponding moment is selected from the risk evolution trajectory sequence, and the risk arrival value of the edge endpoint is calculated in combination with the corresponding risk decay coefficient. Obtain the clearance dimensions, turning radius, and ground friction coefficient of each side's corresponding channel. Combine the minimum passage section requirements and maximum climbing capacity threshold of the transfer equipment to calculate the passage feasibility index. When the passage feasibility index is lower than the preset passage lower limit, a penalty weight is assigned to the side. Extract the location of the emergency response unit and the standard response delay, calculate the shortest path length from the midpoint of each side to the nearest emergency response unit, calculate the emergency intervention time based on the shortest path length and the standard response delay, and normalize it to obtain the emergency accessibility coefficient. The risk arrival value, the travel feasibility index, and the emergency accessibility coefficient are weighted and fused to generate edge weights, and a dynamic risk propagation graph is constructed.

5. The method according to claim 4, characterized in that, The risk feature vector is expanded over time to obtain a risk evolution trajectory sequence, and the risk decay coefficient at each time point is calculated based on the time-varying risk decay function, including: A time-dimensional projection transformation is performed on the risk feature vector, and the start and end times of the risk observation window are set. Dynamic sampling is performed between the start and end times with an adaptive time step, and the adaptive time step adjusts the sampling density according to the fluctuation amplitude of each component of the risk feature vector. At each sampling time, the current component value of the risk feature vector is extracted and the risk change rate relative to the starting time is calculated. The instantaneous risk intensity at each sampling time is calculated by combining the current component value and the instantaneous risk intensity at each sampling time is connected in time sequence to form a risk evolution trajectory sequence. The risk evolution trajectory sequence is piecewise linearly fitted to obtain multiple risk decay curves. The slope of each curve is calculated and the inflection point is identified. Based on the inflection point, the risk time-varying decay function is divided into multiple decay stages, and differentiated decay index parameters are set. For any sampling moment in the risk evolution trajectory sequence, determine the decay stage and substitute the corresponding decay index parameter into the risk time-varying decay function to obtain the risk decay coefficient for each sampling moment.

6. The method according to claim 1, characterized in that, Based on the dynamic risk propagation map, a constrained path search is performed. A set of feasible paths satisfying these dual constraints is generated by setting a risk accumulation cap and a transfer time cap. The risk volatility variance and emergency intervention coverage of each feasible path are calculated. A path transfer scoring function is constructed, and target transfer paths are selected, including: Extract node risk propagation values ​​and edge connectivity from the dynamic risk propagation graph, calculate the difference in risk propagation values ​​between adjacent nodes as the risk gradient, and generate a weighted adjacency matrix by weighting the edge connectivity. Based on the risk accumulation limit and the transit time limit, determine the maximum number of high-risk nodes and the maximum number of path hops, and use these as constraints to perform a breadth-first search on the weighted adjacency matrix to generate a set of feasible paths. For each feasible path, the risk propagation value of each node is extracted and organized into a risk sequence. The risk sequence is differentially processed and the standard deviation of the differential result is calculated as the risk fluctuation variance. Nodes on the path where emergency equipment or emergency personnel are configured are identified and the maximum interval distance between them is calculated. The emergency intervention coverage rate is calculated based on the ratio of the maximum interval distance to the total path length. The cumulative risk value, total transit time, risk volatility variance, and emergency intervention coverage rate are used as scoring indicators. Weighting coefficients are assigned and a path transit scoring function is constructed. The path transit score of each feasible path is calculated, and the feasible path with the highest path transit score is selected as the target transit path.

7. The method of claim 6, wherein, Based on the risk accumulation limit and the transit time limit, the maximum number of high-risk nodes and the maximum path hop count are determined, and used as constraints to perform a breadth-first search on the weighted adjacency matrix to generate a set of feasible paths; including: Extract node risk propagation values ​​from the weighted adjacency matrix and sort them in descending order. Accumulate the sorted risk propagation values ​​until the accumulated value reaches the risk accumulation limit. Use the number of nodes involved in the accumulation process as the maximum number of high-risk nodes. Calculate the expected path jump time based on the edge weight distribution characteristics of the weighted adjacency matrix. Determine the maximum number of path jumps based on the ratio of the expected time to the upper limit of the transit time. During the breadth-first search process, a constraint status identifier is established for each path. The constraint status identifier records the number of high-risk nodes currently included in the path and the number of path hops already executed. When expanding the path, the constraint status identifier is dynamically updated according to the risk propagation value of the candidate node. Only the paths whose constraint status identifiers meet the dual constraint conditions are expanded. All paths that reach the target node and whose constraint status identifiers are compliant are collected to form a feasible path set.

8. Anesthesia recovery elderly patient transport risk assessment and pathway planning system for implementing the method of any one of claims 1-7, characterized in that, include: The data acquisition unit is used to acquire physiological monitoring data, anesthetic drug data, and transport environment data of the patient to be transported. The feature generation unit is used to extract and concatenate features from the physiological monitoring data and the anesthetic drug data to generate a risk feature vector. The risk propagation graph construction unit is used to construct a dynamic risk propagation graph that integrates the time-varying characteristics of risk and spatial constraints based on the risk feature vector and the spatial topology information in the transit environment data. The edge weights are defined by the path accessibility conditions and emergency response accessibility constraints. The path selection unit is used to perform constrained path search based on the dynamic risk propagation map, generate a set of feasible paths that meet the dual constraints by setting a risk accumulation limit and a transfer time limit, calculate the risk fluctuation variance and emergency intervention coverage of each feasible path, construct a path transfer scoring function, and select the target transfer path. The transfer early warning unit is used to transfer the patient to be transferred based on the target transfer route and collect real-time physiological monitoring data, and to issue risk warnings in combination with preset early warning triggering rules.

9. An electronic device, comprising: include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon computer program instructions, wherein, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images