A perioperative brain-metabolic coupling risk assessment system and method

By processing multi-source data and integrating multi-dimensional features, the perioperative brain metabolic risk is quantified, which solves the problems of assessment lag and bias in existing assessment systems. It realizes a comprehensive assessment of brain tissue micro-metabolism and cellular energy supply, and improves the accuracy of risk prediction and system stability.

CN123120316APending Publication Date: 2026-11-17CANCER INST & HOSPITAL CHINESE ACADEMY OF MEDICAL SCI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611384324.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-08
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing perioperative brain metabolic risk assessment systems lack quantitative evaluation of the availability of brain microenvironment substrates and the mediating efficacy of cellular-level energy supply. They cannot effectively quantify the interactive damage effect of surges in neuronal energy demand and decline in energy supply networks, leading to assessment lag and prediction bias.

Method used

The data acquisition and processing module acquires multi-source data and performs time-base alignment. It uses neural activity demand, astrocyte energy supply assessment unit, and brain vulnerability assessment unit to quantify neuronal energy demand and astrocyte energy supply status. It combines multi-dimensional features for dimensionless processing and global weight integration to generate a comprehensive imbalance index and output a metabolic supply and demand imbalance warning status.

Benefits of technology

It enables a comprehensive assessment of the microscopic substrate metabolism and cellular energy supply mediators in brain tissue, improving the comprehensiveness and accuracy of risk assessment, identifying physiological compensation exhaustion nodes, reducing computational bias, and ensuring system stability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123120316A_ABST
    Figure CN123120316A_ABST
Patent Text Reader

Abstract

This invention relates to the field of medical data processing technology, and discloses a perioperative brain metabolism coupling risk assessment system and method. The system includes a data acquisition and processing module that acquires multi-source data from subjects and performs time alignment and normalization to generate an assessment dataset. A state assessment module receives the dataset and calculates and outputs scores for neural activity demand, glucose-lactate metabolic resource insufficiency, astrocyte-glial energy supply impairment, and basic brain vulnerability. A coupling integration and risk output module performs a combination operation on the neural activity demand score and the astrocyte-glial energy supply impairment score to generate an interactive risk amplification score. The system then globally weights and integrates the five features to construct a comprehensive imbalance index. Based on this index, the probability of occurrence is predicted, and the risk level is output after comparison with a threshold. This invention quantifies the mismatch state when neural energy demand surges and energy supply is impaired concurrently, improving the comprehensiveness and accuracy of risk prediction.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, specifically to a perioperative brain-metabolic coupling risk assessment system and method. Background Technology

[0002] Perioperative delayed neurocognitive recovery is a common neurological complication following surgery and anesthesia. Early and accurate risk assessment is crucial for proactive clinical intervention and improved patient outcomes. Currently, commonly used clinical risk monitoring and assessment methods primarily rely on peripheral vital signs or single electroencephalogram (EEG) indicators. These conventional methods typically only reflect macroscopic systemic hemodynamic changes or the electrophysiological state of the cortical surface, failing to reveal the microscopic energy metabolism within brain tissue. Physiologically, efficient brain function highly depends on the dynamic metabolic coupling between neurons and astrocytes; that is, the transient recovery of neural electrical activity requires timely glucose conversion and lactate output support from astrocytes. During surgical stress and anesthesia recovery, the metabolic energy demand of the patient's neurons increases significantly. If local cerebral microcirculation perfusion is insufficient or the physiological energy supply function of astrocytes is impaired at this time, an imbalance between the supply and demand of metabolic substrates will occur.

[0003] However, existing risk assessment systems lack quantitative evaluation indicators for the availability of brain microenvironment substrates and the effectiveness of cellular-level energy mediators. They often simply linearly superimpose various physiological parameters, failing to objectively quantify the nonlinear interactive damage effects that occur when neuronal energy demand surges concurrently with energy network decline. Due to the failure to effectively identify key nodes of brain metabolic compensation exhaustion, existing models have low sensitivity to complex supply-demand mismatch states, easily leading to assessment bias and delayed early warning. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a perioperative brain metabolic coupling risk assessment system and method, which solves the problems of existing assessment methods lacking quantitative monitoring of brain tissue metabolic resource supply and astrocyte energy supply status, and being unable to effectively quantify the interactive mismatch effect when neural demand and energy supply are impaired, resulting in lag and limitations in the prediction of perioperative delayed neurocognitive recovery.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of this invention provides a perioperative brain-metabolic coupling risk assessment system, comprising: The data acquisition and processing module is used to acquire multi-source data of subjects during the perioperative period and perform time-baseline alignment and normalization to generate a standardized perioperative metabolic coupling assessment dataset. The status assessment module is used to receive the perioperative metabolic coupling assessment dataset, output the neural activity demand score through the neural activity demand assessment unit, extract the absolute difference between the current monitoring value and the reference baseline value of specific biomarkers through the astrocyte-collagen energy supply assessment unit, assign positive or negative polarity judgment coefficients according to the physiological characteristics of the biomarkers, calculate the energy supply impairment sub-score by retaining the changing trend consistent with the physiological damage risk through a non-negative truncation function, output the astrocyte-collagen energy supply impairment score after weighted summation, and output the basic brain vulnerability score through the brain vulnerability assessment unit. The coupling integration and risk output module is used to perform dimensionless processing on the neural activity demand score and the astrocyte-collagen energy supply impairment score, arithmetically add the normalized scores of the two, and input the overflow difference after subtracting a constant 1 into a non-negative truncation function to generate an interaction risk amplification score. When the sum of the normalized scores is not greater than the constant 1, it is assigned a value of 0. The module performs global weighting integration on the neural activity demand score, the glucose-lactic acid metabolic resource deficiency score, the astrocyte-collagen energy supply impairment score, the interaction risk amplification score, and the basic brain vulnerability score to construct a comprehensive imbalance index. The module calculates the predicted probability of occurrence based on the comprehensive imbalance index, compares the predicted probability of occurrence with the preset risk classification judgment threshold obtained by fitting a clinical training cohort, and outputs a metabolic supply and demand imbalance warning status and auxiliary prompt information.

[0006] Furthermore, the perioperative multi-source data includes continuous physiological monitoring data, intermittent biochemical data, and specific biomarker data; The process of the data acquisition and processing module performing time reference alignment includes: dividing the monitoring span into continuous analysis time windows, extracting the average and maximum values ​​of the continuous physiological monitoring data within the same analysis time window, mapping the detection values ​​of the intermittent biochemical data and the specific biomarker data to the corresponding analysis time window on the time axis, and maintaining the detection values ​​as physiological state reference values ​​for subsequent adjacent analysis time windows.

[0007] This technical feature solves the problem of temporal misalignment caused by inconsistent sampling frequencies of heterogeneous data by unifying the analysis time window, and realizes the structured alignment of discrete biochemical test data and continuous physiological monitoring features on the time axis.

[0008] Furthermore, the neural activity demand assessment unit extracts the degree of brainwave energy recovery, the degree of brainwave complexity recovery, the degree of burst inhibition recovery, and the degree of awakening from the continuous brainwave time series features in the perioperative metabolic coupling assessment dataset; The neural activity demand assessment unit performs dimensionless scaling on the degree of brainwave energy recovery, the degree of brainwave complexity recovery, the degree of burst inhibition recovery, and the degree of awakening conversion, and outputs the neural activity demand score through weighted summation; when performing the division operation of the dimensionless scaling, a small positive real number is superimposed on the denominator.

[0009] This technology uses multidimensional EEG dynamics to characterize the energy demand level of neurons during the transition from a metabolically inhibited state to a waking state, and quantitatively assesses the degree of recovery of brain electrical activity.

[0010] Furthermore, the brain metabolic resource supply assessment unit extracts the change in total hemoglobin concentration, the rate of change in oxyhemoglobin concentration, and mean arterial pressure to construct brain perfusion proxy indicators, and calculates the estimated value of local brain glucose uptake rate by combining it with arterial blood glucose concentration. The brain metabolic resource supply assessment unit uses arterial blood oxygen content and tissue oxygen extraction rate to estimate the brain oxygen metabolism rate. It then divides the estimated brain oxygen metabolism rate with the estimated local brain glucose uptake rate to obtain the oxygen-glucose metabolism ratio. The unit calculates the deviation gap between the oxygen-glucose metabolism ratio and a preset reference ratio. It constructs an instantaneous anaerobic metabolic load by combining the estimated local brain glucose uptake rate with the instantaneous anaerobic metabolic load. Based on the estimated local brain glucose uptake rate and the instantaneous anaerobic metabolic load, it performs a weighted integration to output the glucose-lactic acid metabolic resource deficiency score.

[0011] This technology uses non-invasive monitoring parameters to deduce the availability of microscopic substrates in brain tissue and the degree of uncoupling of oxygen and glucose metabolism, thereby quantifying the scarcity of metabolic resources in brain tissue.

[0012] Furthermore, before calculating the energy supply barrier sub-scores corresponding to each specific biomarker, the method further includes: extracting the absolute difference between the current monitoring value of the specific biomarker and the reference baseline value.

[0013] This technical feature provides fundamental data support for subsequent assessment of the decline in the energy-mediating efficiency of astrocytes by quantifying the dynamic deviation of biomarkers from the baseline.

[0014] Furthermore, the brain vulnerability assessment unit extracts multidimensional clinical baseline characteristics, including age, preoperative cognitive function, diabetic status, history of cerebrovascular disease, history of hypertension, and surgical trauma load. The brain vulnerability assessment unit converts the multidimensional clinical baseline features into standardized sub-item scores ranging from 0 to 1 based on a preset segmented scoring rule obtained from fitting a clinical training cohort. A weighted linear combination of these standardized sub-item scores is then performed to output the baseline brain vulnerability score. This technical feature transforms the patient's susceptibility history information into standardized values, thereby quantifying the differences in tolerance to hypoxia and metabolic resource depletion among different individuals.

[0015] Furthermore, when the specific biomarker or the multidimensional clinical baseline characteristics are missing, the status assessment module removes the missing item from the weighted calculation, proportionally amplifies the weights of the remaining valid items, and normalizes the arithmetic sum of the weights of the remaining valid items involved in the calculation to a value of 1. This technical feature, through weight normalization, avoids calculation anomalies and assessment biases caused by missing test results or incomplete medical records in actual medical environments.

[0016] Furthermore, the process by which the coupling integration and risk output module constructs the comprehensive imbalance index and calculates the predicted probability of occurrence includes: The global weight integration is completed by multiplying the neural activity demand score, the glucose-lactic acid metabolic resource deficiency score, the astrocyte energy supply impairment score, the interaction risk amplification score, and the basic brain vulnerability score by a preset weight coefficient obtained by fitting a clinical training cohort, and performing a weighted summation to generate the comprehensive imbalance index. Substituting the comprehensive imbalance index into the logistic regression mapping function, the predicted probability of occurrence is output, with values ​​ranging from 0 to 1. This technical feature integrates global weights of multi-dimensional variables and employs a nonlinear function to adapt to the pathogenesis of physiological complications, thereby achieving quantitative assessment of clinical risk.

[0017] A second aspect of the present invention provides a method for assessing perioperative brain-metabolic coupling risk, comprising: We acquired multi-source perioperative data from subjects and performed time-baseline alignment and normalization to generate a standardized perioperative metabolic coupling assessment dataset. The system receives the perioperative metabolic coupling assessment dataset, outputs a neural activity demand score through the neural activity demand assessment unit, outputs a glucose-lactic acid metabolic resource deficiency score through the brain metabolic resource supply assessment unit, assigns positive or negative polarity determination coefficients based on the physiological characteristics of the biomarkers, calculates the energy supply impairment sub-scores corresponding to each specific biomarker by retaining the changing trend consistent with the physiological impairment risk through a non-negative truncation function, performs a weighted summation of the energy supply impairment sub-scores through the astrocyte-colloid energy supply assessment unit to output the astrocyte-colloid energy supply impairment score, and outputs a basic brain vulnerability score through the brain vulnerability assessment unit. The neural activity demand score and the astrocyte energy supply impairment score are dimensionlessly processed to obtain normalized scores. The normalized scores are arithmetically added together to obtain the sum of normalized scores. The difference between the sum of normalized scores and a constant 1 is input into a non-negative truncation function to generate an interaction risk amplification score. When the sum of normalized scores is not greater than the constant 1, the interaction risk amplification score is assigned a value of 0. The neural activity demand score, glucose-lactic acid metabolic resource deficiency score, astrocyte energy supply impairment score, interaction risk amplification score, and basic brain vulnerability score are globally weighted and integrated to construct a comprehensive imbalance index. The probability of occurrence is predicted based on the comprehensive imbalance index. The predicted probability of occurrence is compared with a preset risk grading threshold obtained by fitting a clinical training cohort. The metabolic supply and demand imbalance warning status and auxiliary prompt information are output. The auxiliary prompt information does not directly replace clinical diagnostic decisions.

[0018] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-described perioperative brain-metabolic coupling risk assessment method.

[0019] This invention provides a perioperative brain-metabolic coupling risk assessment system and method. It has the following beneficial effects: 1. This invention integrates continuous physiological monitoring data with specific biomarker data, and uses a brain metabolic resource supply assessment unit and an astrocyte energy supply assessment unit to deduce the anaerobic metabolic load of brain tissue and the functional status of astrocytes, respectively, and outputs scores for insufficient glucose-lactic acid metabolic resources and scores for impaired astrocyte energy supply. Compared with conventional methods that rely on single EEG or hemodynamic indicators, this scheme extends the assessment scope to the level of brain tissue microscopic substrate metabolism and cellular energy supply mediation, and eliminates the interference of compensatory protective changes during the assessment process through direction determination rules, thereby improving the comprehensiveness of risk assessment.

[0020] 2. This invention treats the surge in neuronal energy demand and impaired astrocyte energy supply as independent variables. By subtracting the baseline compensatory tolerance constant and inputting a non-negative truncation function, it extracts and generates an interactive risk amplification score. This mathematically separates the supply-demand mismatch that occurs when increased neuronal energy demand and impaired astrocyte energy support occur simultaneously. It overcomes the shortcomings of traditional linear weighting methods that cannot reflect nonlinear superimposed damage, and helps identify nodes where the body's physiological compensatory exhaustion occurs.

[0021] 3. This invention integrates the quantitative characteristics of five dimensions—neural activity demand, insufficient metabolic resources, astrocyte-colloid energy supply impairment, amplified interaction risks, and basic brain vulnerability—with global weighting to construct a comprehensive imbalance index. This index is then converted into a predicted probability by substituting it into a logistic regression mapping function. Simultaneously, it configures weight ratio amplification and normalization processing logic for missing features. This achieves unified quantitative assessment of cross-dimensional variables, avoiding calculation errors and result bias when faced with missing clinical laboratory data or medical history information, and ensuring the stability of system operation. Attached Figure Description

[0022] Figure 1 This is a system architecture diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention; Figure 3 This is a timing diagram of the multi-source data acquisition and standardized preprocessing of the present invention; Figure 4 This is a schematic diagram illustrating the multi-dimensional feature calculation principle of the state assessment module of the present invention. Figure 5 This is a flowchart illustrating the coupling, integration, and comprehensive risk output of the present invention. Figure 6 This is a dynamic evolution trend diagram of the risk assessment of the anesthesia recovery period for subjects in this invention. Detailed Implementation

[0023] The technical solutions in 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.

[0024] See attached document Figure 1 This invention provides a perioperative brain metabolic coupling risk assessment system, including a data acquisition and processing module, a state assessment module, and a coupling integration and risk output module. The state assessment module further includes a neural activity demand assessment unit, a brain metabolic resource supply assessment unit, astrocyte-colloid energy supply assessment unit, and a brain vulnerability assessment unit.

[0025] See attached document Figure 2 This invention provides a method for assessing perioperative brain-metabolic coupling risk, comprising the following steps: S10: The data acquisition and processing module acquires multi-source data of the subjects during the perioperative period, performs a unified time reference alignment operation on the multi-source data, extracts the time series features of continuous physiological monitoring data, maps the intermittent biochemical test results and body fluid marker data to the corresponding analysis time window, and then performs artifact removal, missing value processing, normalization processing and preoperative baseline correction to generate a standardized perioperative metabolic coupling assessment dataset. S20, the neural activity demand assessment unit in the state assessment module receives the standardized perioperative metabolic coupling assessment dataset, calls the EEG activity data in it to calculate the neural activity demand score, so as to characterize the metabolic energy demand level of perioperative neuronal activity. S30, the brain metabolic resource supply assessment unit in the state assessment module calls blood glucose data, near-infrared brain oxygen monitoring data and circulatory respiratory parameters in the standardized perioperative metabolic coupling assessment dataset to deduce the current brain glucose substrate availability and lactate-like metabolic pressure status, and performs comprehensive calculations to output the glucose-lactic acid metabolic resource insufficiency score. S40, the astrocyte-mediated energy supply assessment unit in the status assessment module calls the astrocyte-mediated energy supply information in the standardized perioperative metabolic coupling assessment dataset, quantifies the impairment of astrocyte physiological functions in energy supply mediators such as glucose uptake, glycolysis, lactate output and homeostasis maintenance, and outputs the astrocyte energy supply impairment score. S50, the brain vulnerability assessment unit in the status assessment module calls the patient's basic information in the standardized perioperative metabolic coupling assessment dataset, transforms the patient's basic information into numerical susceptibility features, and outputs a basic brain vulnerability score. S60, the coupling integration and risk output module simultaneously acquires the neural activity demand score and the astrocyte energy supply impairment score, calculates the combined characteristics of the two to quantify the supply-demand mismatch state formed when the neural activity energy demand surges while the astrocyte energy supply support capacity lags behind, and generates an interactive risk amplification score. S70, the coupling integration and risk output module integrates the neural activity demand score, glucose-lactic acid metabolic resource deficiency score, astrocyte energy supply disorder score, basic brain vulnerability score, and interactive risk amplification score with global weights to construct a perioperative astrocyte-mediated energy supply and neural activity demand imbalance index. Based on the imbalance index, the predicted probability of delayed neurocognitive recovery is calculated, and then compared with the preset threshold to output the corresponding metabolic supply and demand imbalance warning status and auxiliary prompt information.

[0026] See attached document Figure 3The data acquisition and processing module is responsible for acquiring multi-source physiological and biochemical information of subjects during the perioperative period and transforming heterogeneous raw data into standardized temporal features for use by downstream assessment units. This processing is based on the dynamic metabolic coupling mechanism between neurons and astrocytes, meaning that the transient recovery of neural electrical activity is highly dependent on the uptake and conversion of local cerebral blood flow and glucose by astrocytes. To effectively assess this mismatch, the system needs to synchronously and concurrently extract cross-dimensional physiological indicators to construct standardized basic data. The processing of this module includes: In step S101, the data acquisition and processing module continuously receives raw data transmitted from the multidimensional monitoring device through the communication interface and simultaneously enters the patient's basic information.

[0027] The multi-source data includes continuous physiological monitoring data, intermittent biochemical data, and specific biomarker data. Among them, the continuous physiological monitoring data aims to reflect transient neurophysiological and regional cerebral hemodynamic changes, covering scalp EEG activity data with sampling rates between 128Hz and 512Hz, changes in local brain tissue oxygen saturation, oxyhemoglobin concentration, and deoxyhemoglobin concentration acquired via near-infrared spectroscopy, and circulatory and respiratory parameters such as mean arterial pressure, heart rate, and pulse oxygen saturation accessed via conventional vital sign monitoring equipment.

[0028] Intermittent biochemical data included peripheral arterial blood gas analysis results, specifically arterial blood glucose and arterial blood lactate concentrations, used to assess the macroscopic substrate supply status throughout the body. Specific biomarker data covered values ​​from peripheral blood samples collected at different points in the perioperative period, specifically the relative abundance or concentration of glial fibrillary acidic protein, solute carrier family 16 member 3, aldolase C, lactate dehydrogenase A, aquaporin 4, and glucose transporter 1. Patient baseline information included the subject's age, preoperative cognitive function scale score, and history of comorbidities such as diabetes and hypertension.

[0029] In step S102, in order to solve the timing misalignment problem caused by inconsistent sampling frequencies of heterogeneous data, the data acquisition and processing module uses the set anesthesia induction time or the start time of surgery as a unified time reference 0 point to perform timestamp alignment and mapping operations on multi-source heterogeneous data.

[0030] In one preferred implementation, the data acquisition and processing module divides the entire perioperative monitoring period into multiple equal and continuous analysis time windows, typically set to a length of 10 seconds to 5 minutes. This range is determined primarily by considering both the routine data refresh cycle of the clinical monitor and the time delay characteristics of the cerebral blood flow's own physiological regulation.

[0031] For continuous physiological monitoring data, the data acquisition and processing module extracts the statistical features of the corresponding time series within each analysis time window. Specifically, it calculates the average and maximum values ​​of the continuous signal sequence within that time window, using these as representative feature values ​​for that time window. When multiple intermittent detection data are entered within the same analysis time window, the system extracts the detection value with the latest timestamp as the representative value for that time window.

[0032] For intermittent biochemical data and specific biomarker data, the data acquisition and processing module reads the physical timestamp of the actual collection of the corresponding body fluid sample and maps its detection value to the corresponding or adjacent subsequent analysis time window on the time axis. Before the next valid sample test data is entered, the system maintains this mapped value as the physiological state reference value of the current time window, thereby achieving structured alignment of discrete data and continuous features in the time dimension.

[0033] In step S103, after the time structure alignment is completed, the data acquisition and processing module performs artifact identification and removal and missing value interpolation on the aligned time series features to reduce errors introduced by physical interference from the electrosurgical unit or sensor detachment in the surgical environment.

[0034] For EEG activity data and near-infrared brain oxygenation monitoring data that are susceptible to physical interference, the data acquisition and processing module is equipped with physiological thresholds and mutation gradient thresholds. These thresholds are typically established based on conventional human physiological limits and statistical quantiles from historical clinical databases. When the signal characteristic value within a single analysis time window exceeds the set normal human physiological range, or when the data jump amplitude between adjacent analysis time windows exceeds the preset gradient upper limit, the data acquisition and processing module marks the data within that analysis time window as artifact interference and removes it.

[0035] For local data breaks in the time series resulting from anomaly removal, the system introduces missing value imputation. If the duration of the breakpoint is less than the set imputation tolerance time limit (e.g., 5 min to 15 min), the data acquisition and processing module uses effective time window data before and after the missing segment to derive and fill in the missing value through a linear interpolation algorithm. If the duration of the breakpoint is greater than or equal to the imputation tolerance time limit, the data acquisition and processing module maintains the missing state of that period and marks it as an invalid evaluation interval, thereby avoiding bias in long-term state prediction caused by over-extrapolation calculation.

[0036] In step S104, in order to eliminate the differences in physical dimensions between physiological and biochemical indicators of different dimensions, the data acquisition and processing module performs numerical normalization and preoperative baseline correction on the cleaned data matrix, and generates and outputs a standardized perioperative metabolic coupling assessment dataset.

[0037] For parameters with different physical dimensions, the data acquisition and processing module, based on the physical characteristics of various physiological indicators, utilizes preset historical statistical minimum and maximum values, and employs a linear transformation operation of maximum-minimum scaling for processing. To ensure the completeness of the algorithm logic and the stability of data matrix operations, when calculating the difference between the current value and the statistical minimum value, and dividing it by the variable range (i.e., the difference between the statistical maximum and minimum values), the system routinely adds a small positive real number to the denominator. (Usually 10) -6 ~10 -8 This is to avoid triggering a calculation overflow exception with a denominator of 0 due to the accidental equality of specific historical maximum and minimum values.

[0038] Through the above normalization process, all kinds of effective values ​​are smoothly mapped to the dimensionless range of 0 to 1.

[0039] Simultaneously, the data acquisition and processing module extracts the average values ​​of various monitoring characteristics of the subjects within 3 to 5 minutes during the ward's awake and quiet state or the stable phase before anesthesia induction, establishing these values ​​as the reference baseline. For all collected physiological characteristics and biomarker concentrations, the module calculates the difference or percentage change between the normalized current characteristic value and the reference baseline within each analysis time window. The numerical matrix formed after dimensional unification and baseline correction is encapsulated into a standardized perioperative metabolic coupling assessment dataset and stored in system memory to support subsequent status assessment modules in performing independent-dimensional feature quantification calculations.

[0040] See attached document Figure 4 The neural activity demand assessment unit in the state assessment module is responsible for receiving continuous EEG time-series features from the standardized perioperative metabolic coupling assessment dataset, extracting and analyzing EEG time-frequency features to quantitatively assess the immediate energy demand level generated by the subject during the perioperative period, especially during the recovery period from general anesthesia or when subjected to major surgical stress, as the cerebral cortex neurons transition from a metabolically inhibited state to a conscious and active state.

[0041] The general technical principle underlying this assessment process is that the firing of action potentials and synaptic transmission in cortical neurons are highly dependent on the continuous operation of cell membrane ion pumps, a process that consumes a significant amount of energy. Since non-invasive clinical monitoring methods cannot directly measure the rate of consumption of microscopic high-energy phosphate bonds in the brain at the bedside, this embodiment selects multidimensional electroencephalographic characteristics as a macroscopic proxy indicator of neuronal metabolic energy demand. The calculation process for this neural activity demand score specifically includes: In step S201, the neural activity demand assessment unit in the state assessment module retrieves continuous EEG time-series features from the standardized perioperative metabolic coupling assessment dataset and calculates and analyzes the degree of EEG energy recovery and the degree of EEG complexity recovery within the time window. As a preferred approach, before feature extraction, the system performs preprocessing on the raw EEG signal, including bandpass filtering and notch filtering.

[0042] To assess the degree of EEG energy recovery, this unit extracts the total power of the EEG signal within a set target frequency band, typically ranging from 0.5 Hz to 45 Hz, which comprehensively covers the slow-wave and fast-wave bands in routine clinical EEG monitoring. During the stable phase of anesthesia maintenance, high-frequency discharges of the subject's central neurons are suppressed, resulting in a relatively low total power. As anesthetic drugs are metabolized and the patient awakens, the electrical activity of central neurons recovers, and the total power shows an upward trend. This unit objectively quantifies the recovery level of overall neuronal discharge activity by calculating the percentage increase or absolute difference between the total power within the current analysis time window and the lowest baseline power recorded during the intraoperative anesthesia maintenance period.

[0043] To assess the degree of EEG complexity recovery, this unit employs nonlinear dynamic indices to calculate the microscopic complexity of EEG signals, using sample entropy or spectral entropy as quantification parameters for complexity. When patients are under deep anesthesia or deep sedation, the EEG waveform exhibits regular and periodic slow-wave oscillations, corresponding to low sample entropy values, indicating that the cerebral cortex network is in a state of low synaptic information processing load and low basal metabolic consumption. When the cortex is activated by noxious surgical stimulation, the neuronal firing pattern shifts to desynchronized and complex fast-wave oscillations, with a significant increase in sample entropy values. This change physically corresponds to an increase in the rate of local brain tissue uptake of metabolic substrates.

[0044] In step S202, in order to capture the immediate impact of the release of deep metabolic inhibition and the cortical arousal tendency on brain energy balance, the neural activity demand assessment unit simultaneously calculates the degree of burst inhibition recovery and the degree of arousal conversion.

[0045] Burst inhibition is a self-protective metabolic reduction state that neurons adopt when facing deep anesthesia or acute cerebral hypoxia. During this state, the EEG time series exhibits alternating flat isoelectric resting lines and brief periods of high-amplitude fast waves. This unit calculates the absolute proportion of the isoelectric resting time within the current analysis time window to the total monitoring time window duration, obtaining the burst inhibition ratio. As the recovery process progresses, the burst inhibition ratio decreases from its highest extreme value recorded during the intraoperative deep anesthesia phase to 0. This rapid release of inhibition signifies the recovery of cortical network electrophysiological activity, accompanied by the reconstruction of intracellular ion concentration gradients and a significant increase in basal energy metabolic demand. This unit uses the difference between the burst inhibition ratio within the current analysis time window and the highest intraoperative value to characterize the degree of this acute energy rebound effect in real time.

[0046] Simultaneously, this unit calculates the degree of awakening transition, using the edge frequency of the EEG spectrum as a key quantitative indicator as a preferred method. This indicator extracts the upper frequency limit corresponding to the point where the power accumulation in the overall power spectral density reaches 95%. When the edge frequency of the EEG spectrum rapidly migrates from the low-frequency range representing the depth of anesthetic inhibition to the high-frequency range above 14Hz to 30Hz, it indicates the activation of the central nervous system's arousal pathway. This indicator can effectively predict the significant increase in the patient's demand for systemic circulating metabolic substrates in the initial stage of neurocognitive function recovery.

[0047] In step S203, the neural activity demand assessment unit performs dimensionless scaling and weighted fusion on four physiological characteristic indicators: the degree of recovery of brain electrical energy, the degree of recovery of brain electrical complexity, the degree of recovery of burst inhibition, and the degree of awakening transition, and calculates and outputs the final neural activity demand score.

[0048] Considering the significant differences in the initial physical dimensions and absolute value ranges of the four sub-feature indicators mentioned above, this unit retrieves the minimum and maximum values ​​of each indicator established within the current subject's historical monitoring period, and uses a maximum-minimum scaling operation to map them respectively to dimensionless scores between 0 and 1. When performing range division calculations, this unit routinely adds a small positive real number to the denominator. (Usually 10) -6 ~10 -8 A small constant approaching 0 This avoids calculation overflow exceptions where the denominator is 0 due to the maximum and minimum values ​​being accidentally equal.

[0049] After completing the dimensionless scaling, the system calculates the neural activity demand score according to the set weighted summation formula, the logic of which is shown in the following formula: ; in, Indicates the current time window The neural activity demand score is used to characterize the perioperative neuronal activity and its energy demand status. This indicates the brainwave energy score; This indicates the EEG complexity recovery score; This indicates the burst suppression and recovery score; The score indicates the transition from anesthesia-induced inhibition to a state of awakening; ~ The preset weighting coefficients are configured with an arithmetic sum of 1.0.

[0050] As a preferred parameter configuration range, ~ The values ​​can be set to 0.30, 0.25, 0.20, and 0.25 respectively, and the specific values ​​of the above weight coefficients are obtained by fitting the clinical training cohort.

[0051] To ensure the integrity of the algorithm logic in a real-world clinical multi-source monitoring environment, the system introduces a dynamic weight normalization mechanism to address missing indicators. This mechanism is applied when the patient does not exhibit significant burst suppression during the perioperative period. If an item is not applicable, or if a sub-feature fails quality control due to local sensor interference, the system will not assign a score of 0 to that unusable sub-item and include it in the total score calculation to avoid biased low-risk values ​​in the output. The system will remove that unusable sub-item from the current score. In the calculation, the effective sub-items are removed, and the weights of the remaining applicable sub-items are proportionally amplified and distributed so that the sum of the weights of the effective sub-items participating in the calculation is normalized to 1. The dimensionless continuous value generated by the final weighted summation is established as the neural activity demand score, and is uniformly transmitted and summarized to the coupling integration and risk output module, which is used as the energy demand benchmark input variable for subsequent joint determination of the energy demand relationship balance of brain metabolism.

[0052] See attached document Figure 4 The brain metabolic resource supply assessment unit in the state assessment module is responsible for calling up blood glucose data, near-infrared brain oxygen monitoring data and circulatory respiratory parameters from the standardized perioperative metabolic coupling assessment dataset to deduce the current microscopic substrate availability and anaerobic metabolic load of brain tissue.

[0053] The general technical principle underlying this assessment process is that the brain's energy supply under such perioperative stress is not solely dependent on the absolute concentrations of blood glucose and oxygen in peripheral arteries, but is limited by the actual perfusion of local cerebral blood flow and the efficiency of the blood-brain barrier in transporting metabolic substrates. Peripheral blood glucose readings within the normal physiological range do not equate to the absence of a metabolic crisis in the cerebral microcirculation. Therefore, this embodiment abandons the traditional approach of relying solely on a single circulatory or biochemical indicator to determine risk, instead using non-invasive monitoring parameters to jointly deduce local cerebral glucose uptake rate and the uncoupling state of oxygen-glucose metabolism, thereby quantitatively assessing the true extent of metabolic resource deficiency in the brain region. The calculation process for this glucose-lactic acid metabolic resource deficiency score specifically includes: In step S301, the brain metabolic resource supply assessment unit jointly calculates the estimated value of local brain glucose uptake rate and generates a glucose substrate availability insufficiency score based on it.

[0054] To objectively reflect the true blood perfusion level of local brain tissue, the system constructs a surrogate index for brain perfusion using near-infrared brain oxygenation signals and mean arterial pressure. This unit extracts and analyzes the changes in total hemoglobin concentration, the rate of change in oxyhemoglobin concentration, and mean arterial pressure within the analysis time window. Dimensionless scaling is applied to these multidimensional physical quantities, and a small positive real number is superimposed on the total range in the denominator of the scaling formula. (Usually 10) -6 ~10 -8 To avoid computational overflow anomalies, a weighted summation is then performed. The physical basis of this weighting logic is as follows: the change in total hemoglobin concentration positively represents the fullness of local cerebral blood volume; the rate of change in oxyhemoglobin positively reflects the instantaneous dynamics of arterial oxygen-rich blood flow inflow; the rate of change in deoxyhemoglobin negatively suggests increased local oxygen replenishment compensation or slow perfusion; and mean arterial pressure represents the macroscopic positive driving pressure of blood pumping from the whole body's blood vessels to the brain.

[0055] Based on the constructed brain perfusion surrogate index, the system derives the first... Individual brain regions or monitoring channels in time The estimated local brain glucose uptake rate, as a preferred method, is calculated using the following formula: ; in, This represents an estimated local brain glucose uptake rate, used to characterize the macroscopic rate at which brain tissue actually acquires energy substrates within this time window. This is a dimensionless proxy index calibrated based on the subject's preoperative stable baseline, used to convert the dimensionless proxy index into a relative physiological uptake level. This indicates the arterial blood glucose concentration corresponding to the current time window; This represents the brain perfusion proxy indicator constructed above; This represents the glucose uptake efficiency coefficient, which is a dimensionless relative correction parameter.

[0056] Considering that various perioperative pathological factors can impair the blood-brain barrier's ability to transport glucose, the system targets... Dynamic degradation correction logic is configured. The system synchronously acquires the relative abundance decrease score of glucose transporter 1, the preoperative diabetes control score, the perioperative blood glucose fluctuation score, and the score of severe hyperglycemia / hypoglycemia events, which reflect the blood-brain barrier glucose transport status. The system presets a minimum correction limit (usually set to 0.40–0.60) and performs a deduction operation on a baseline value of 1.0 using the above scores and their corresponding weights.

[0057] When any of the above correction factors is missing data or fails the system quality control, the system does not set it as risk-free or high-risk, but removes it from the deduction factor set and performs weight normalization on the remaining valid correction factors to ensure that the total weight is normalized again.

[0058] After deriving the uptake rate, this unit calculates the gap between the current estimated local brain glucose uptake rate and the preoperative steady-state baseline, and divides this gap by the baseline to obtain the percentage decrease. The system extracts key features of this percentage decrease throughout the monitoring period, maps them to the 0-1 range after max-min scaling, and outputs a glucose substrate availability insufficiency score to quantify the relative severity of substrate supply shortage.

[0059] In step S302, the brain metabolic resource supply assessment unit further monitors the oxygen-glucose metabolism uncoupling phenomenon and calculates the lactate-like metabolic stress score. Under normal physiological conditions, brain tissue mainly consumes glucose through aerobic oxidation, and the oxygen-to-glucose consumption rate is maintained at a fixed steady-state ratio. When local microcirculatory disturbances or a surge in neuronal energy demand leads to a relative insufficiency of oxygen supply, brain tissue is forced to initiate anaerobic glycolysis, at which point a significant uncoupling occurs between the aerobic metabolic rate and the glucose uptake rate.

[0060] This unit uses cerebral perfusion proxy indicators, arterial blood oxygen content, and tissue oxygen extraction rate derived from near-infrared signals to estimate the current cerebral oxygen metabolism rate through product operations.

[0061] After obtaining the estimated brain oxygen metabolism rate, the system divides it by the estimated local brain glucose uptake rate to obtain the oxygen-glucose metabolism ratio. The system extracts a reference ratio under theoretically fully aerobic metabolic conditions (with a constant set approximately 6), and calculates the deviation gap between the current metabolic ratio and this reference ratio to characterize the proportion of anaerobic energy production. Multiplying this proportion coefficient by the estimated local brain glucose uptake rate and the current stage risk weighting function quantifies the instantaneous lactate-like anaerobic metabolic load, the calculation logic of which is shown in the following formula: ; in, Indicates the first Individual brain regions or monitoring channels in time The transient lactate-like anaerobic metabolic load, a physical quantity that directly maps the transient pressure level of anaerobic metabolic product accumulation in the brain. This represents the oxygen-to-glucose metabolism ratio; This represents the reference oxygen-glucose metabolism ratio. This represents the risk weighting function.

[0062] The system analyzes the patient's perioperative stage (e.g., anesthesia maintenance, surgical stimulation, and recovery), history of critical hemodynamic adverse events, and current abnormalities in other dimensions. Dynamic upregulation is performed to compensate for differences in susceptibility to anaerobic metabolism at different physiological stages.

[0063] This unit extracts four key temporal characteristics of transient anaerobic metabolic load on the time axis: peak load exceeding the safe physiological threshold, the area of ​​the load curve integral over time (representing cumulative exposure), the duration of continuous load exceeding the threshold, and the magnitude of load rebound during the recovery period compared to the maintenance anesthesia period. These four characteristics encompass the intensity, persistence, and suddenness of the effects of anaerobic metabolites on neurons.

[0064] This unit uses conventional scaling operations to convert the above four features into dimensionless scores, and then sums them using preset weights to output a lactate-like metabolic stress score in the range of 0 to 1.

[0065] In step S303, the brain metabolic resource supply assessment unit in the state assessment module performs a final weighted integration of the glucose substrate availability insufficiency score and the lactate-like metabolic pressure score to calculate and output a glucose-lactate metabolic resource insufficiency score. The system linearly weights these two scores based on preset resource allocation parameters. This weighted integration is primarily based on clinical pathophysiological consensus: simple substrate deficiency and anaerobic metabolism accompanied by acidosis constitute a synergistic damage mechanism for irreversible neuronal injury, with both having a risk-additive effect. The continuous dimensionless value obtained from the weighted combination is the final glucose-lactate metabolic resource insufficiency score. This score is aggregated and transmitted to the coupling integration and risk output module as a key energy supply baseline input variable for assessing the brain's supply-demand imbalance state.

[0066] See attached document Figure 4 The astrocyte-mediated energy supply assessment unit in the status assessment module is responsible for calling up astrocyte-mediated energy supply information from the standardized perioperative metabolic coupling assessment dataset to quantify the impairment of astrocyte physiological functions in energy supply mediators such as glucose uptake, glycolysis, lactate output, and homeostasis maintenance.

[0067] The general technical principle underlying this assessment process is the existence of a specific astrocyte-neuronal lactate shuttle mechanism within the central nervous system. When neurons are in a high-energy-demand state, such as under stress or during wakefulness, their own glycolysis capacity is often insufficient to meet immediate energy consumption, making them highly dependent on astrocytes, which are closely attached to microvessels, to take up glucose from the blood. Astrocytes convert glucose into lactate through aerobic glycolysis, which is then released into the extracellular space for neurons to take up as a direct oxidative substrate. Due to the bidirectional permeability of the blood-brain barrier, abnormal relative abundance of astrocyte-specific metabolic enzymes and transporters in peripheral blood can directly reflect a decline in the executive efficiency of this microscopic energy supply mediation system in the brain at the macroscopic level. The calculation process for the astrocyte energy supply impairment score specifically includes: In step S401, the astrocyte energy assessment unit retrieves the relative abundance or concentration characteristics of various specific biomarkers in the dataset within the current analysis time window. The biomarker combination selected by the system corresponds to five key links in the energy supply mediation chain: glial fibrillary acidic protein represents the overall structural integrity and reactive activation state of astrocytes; member 3 of solute carrier family 16, as a monocarboxylic acid transporter, directly mediates the export of lactate from astrocytes to the extracellular space; aldolase C represents a key glycolytic enzyme specific to astrocytes; lactate dehydrogenase A is responsible for catalyzing the conversion of pyruvate to lactate; and aquaporin 4 is responsible for maintaining the volume homeostasis of astrocyte vascular foot processes and water-salt exchange at the blood-brain barrier.

[0068] The system acquires the current monitoring values ​​of each of the above biomarkers and simultaneously retrieves their reference baseline values ​​established during the preoperative stable phase. To quantify the dynamic deviation of the biomarkers, this unit calculates the absolute difference between the current monitoring value and the reference baseline value.

[0069] In step S402, to uniformly map the concentration changes of biomarkers with different physiological significance to the risk of impaired energy supply, the astral gel energy supply assessment unit constructs a direction determination rule based on the physiological characteristics of the biomarkers. Since different biomarkers exhibit opposite trends when energy supply is impaired, the system configures a specific direction determination coefficient for each biomarker.

[0070] Specifically, an abnormally elevated concentration of glial fibrillary acidic protein in the blood suggests reactive damage to astrocytes or disruption of blood-brain barrier permeability; therefore, the system sets its orientation determination coefficient to positive +1. Conversely, a decrease in the relative abundance of solute carrier family 16 member 3, aldolase C, lactate dehydrogenase A, and aquaporin 4 reveals a decline in the corresponding glycolysis or transport functions; the system sets the orientation determination coefficient for these four markers to negative -1. Based on this orientation determination rule, this unit calculates the energy supply barrier sub-score corresponding to each marker, and the calculation logic is shown in the following formula: ; in, Indicates the first Such markers in time The corresponding energy supply barrier sub-score is used to independently quantify the degree of damage to a single intermediary link; It is a non-negative cutoff function. Its technical purpose is to retain only the trend of change consistent with the risk of physiological damage. When the direction of change of the marker is conducive to energy recovery or is in the compensatory period, its risk score is assigned to 0 to eliminate interference. Indicates the preset corresponding number Directional determination coefficient of a marker; This indicates the effective characteristic concentration of the biomarker within the current analysis time window; The reference baseline value represents the preoperative stable phase of this biomarker; The small positive real numbers superimposed in the denominator (Usually 10) -6 ~10 -8 The purpose is to avoid triggering a calculation overflow anomaly with a denominator of 0 when the preoperative baseline value of a specific biomarker is undetectable or close to absolute 0 concentration.

[0071] In step S403, the star glue power supply assessment unit in the state assessment module integrates the power supply barrier sub-scores of the above-mentioned markers and calculates and outputs the final star glue power supply barrier score. The calculation logic of the star glue power supply barrier score is shown in the following formula: ; in, Indicates the current time window The star glue energy supply barrier score; Indicates the first The configuration weights of the biomarkers are obtained by fitting the clinical training cohort; Indicates the first The sub-scores of energy supply barriers corresponding to each marker.

[0072] Considering the varying contributions of different energy-supplying intermediaries to the overall metabolic support of neurons, the system incorporates a biomarker fusion weight matrix. Solute carrier family 16 member 3 and aldolase C, responsible for direct energy production and output, are assigned relatively high weight ranges, typically between 0.20 and 0.35; the weight ranges for other biomarkers are typically between 0.10 and 0.20.

[0073] To ensure the algorithm's integrity in real-world clinical testing scenarios, and to address situations where certain biomarker data are missing in some peripheral blood samples due to hemolysis interference, detection limits below the limit, or other objective reasons, this unit activates a weight normalization mechanism. The system removes missing biomarker entries and proportionally amplifies the weights of the remaining biomarkers that can normally acquire valid features, thus normalizing the sum of the weights of the valid sub-items involved in the calculation back to 1.0. This mechanism prevents low-risk total scores from being biased due to data gaps in individual sub-items.

[0074] The system multiplies each valid energy supply barrier sub-score by its corresponding renormalized weight coefficient and sums them. After smoothing through standard extreme value scaling operations to prevent zeros, a dimensionless continuous value between 0 and 1 is generated, which is established as the astral gel energy supply barrier score. This score is aggregated and transmitted to the coupling integration and risk output module as a core energy supply capability input parameter for assessing the balance between brain metabolic supply and demand.

[0075] See attached document Figure 4 The brain vulnerability assessment unit in the status assessment module is responsible for calling the patient's basic information in the standardized perioperative metabolic coupling assessment dataset, converting discrete clinical history information into numerical susceptibility features, and outputting a basic brain vulnerability score.

[0076] The general technical principle underlying this assessment process is that, due to aging, decreased preoperative cognitive reserve, and long-term depletion caused by microvascular lesions from chronic underlying diseases, the inherent tolerance of the brain to perioperative stress, ischemia-hypoxia, and metabolic resource depletion varies significantly among patients. Faced with the same surge in neural activity demands and astrocyte energy imbalance, the probability of neurocognitive impairment in highly vulnerable brains is higher than in healthy brains. Therefore, this embodiment transforms the patient's inherent susceptibility factors into standardized risk assessment baselines, which serve as the basic risk reference values ​​for measuring postoperative cognitive function recovery impairment. The calculation process of this score specifically includes: In step S501, the brain vulnerability assessment unit extracts the subject's multidimensional clinical baseline characteristics and converts them into standardized sub-item scores ranging from 0 to 1 according to a preset mapping rule. The dimensions extracted by this unit cover age, preoperative cognitive function, diabetes status, history of cerebrovascular disease, history of hypertension, and surgical trauma burden.

[0077] For quantifying the age dimension, the system assesses the age-stratified scores of the subjects. The physiological basis for this is that with increasing age, neuronal synaptic plasticity and the ability to autonomously regulate cerebral blood flow gradually decline. The system pre-sets segmented scoring rules: those under 60 years old are assigned 0 points; those between 60 and 74 years old are assigned 0.50 points; and those 75 years and older are assigned 1.00 points. As a parallel implementation method, the system can also skip discrete segmented scoring, directly extract continuous actual age values, and perform a maximum-minimum scaling linear transformation operation using preset reference upper and lower limits to normalize them to the 0-1 interval.

[0078] When assessing preoperative cognitive reserve, the system evaluates the converted values ​​of preoperative cognitive function scores. The system retrieves the patient's clinical scores from the Montreal Cognitive Assessment Scale or other equivalent cognitive scales completed preoperatively. Scores greater than or equal to the normal reference baseline are mapped to 0; scores within the mild cognitive impairment range are mapped to 0.50; and scores below the severe impairment threshold are mapped to 1.00. To achieve more refined risk quantification, the system also supports extracting the gap between the patient's preoperative cognitive scale score and the normal reference value, and dividing it by a preset scale reference range to calculate the continuous normalized cognitive impairment score.

[0079] Regarding the risk assessment of chronic underlying diseases, the system extracted scores for diabetes status, cerebrovascular disease history, and hypertension history. Patients without the corresponding medical history were assigned a score of 0. Patients with a medical history but stable long-term medication control and no significant complications were uniformly assigned a score of 0.50. For patients with poor long-term blood glucose or blood pressure control, combined target organ microvascular damage, previous stroke, or significant white matter lesions, the system assigned a score of 1.00 to the corresponding dimension to quantify their impaired cerebral microcirculation compensatory buffering capacity.

[0080] Considering the impact of current surgical physical trauma on the central nervous system, a systematic assessment of the surgical type, surgical duration, or perioperative event load score is performed. Low-trauma or short-duration surgeries are assigned a score of 0; moderate-trauma surgeries or surgeries with an expected duration of 2 to 4 hours are assigned a score of 0.50; major surgeries, surgeries lasting 4 hours or longer, surgeries accompanied by massive bleeding, or emergency surgeries are assigned a score of 1.00.

[0081] In step S502, the brain vulnerability assessment unit performs a weighted linear combination of the scores for each of the standardized sub-items mentioned above, and calculates and outputs the final basic brain vulnerability score. The logic is shown in the following formula: ; in, The baseline brain vulnerability score is used to characterize the patient's basic susceptibility to perioperative delayed neurocognitive recovery. The higher the score, the worse the brain's basic compensatory reserve. Indicates age-stratified scoring; This represents the converted value of the preoperative cognitive function score; Indicates a diabetes status score; Indicates the history of cerebrovascular disease score; Indicates the history of hypertension; Indicates the type of surgery, duration of surgery, or perioperative event load score; ~ These are the preset weight coefficients for each dimension.

[0082] Based on the pathological contribution of various basic clinical characteristics to postoperative neurocognitive outcomes, ~ The initial configuration values ​​can be set to 0.20, 0.25, 0.15, 0.20, 0.10 and 0.10 respectively, and the arithmetic sum of the weight coefficients is configured to 1.0.

[0083] The specific values ​​of the above parameters can be obtained by refitting and calibrating the multicenter clinical training cohort based on real neurocognitive outcome data through logistic regression analysis.

[0084] To ensure the completeness of the assessment logic in a real medical information environment and to address scenarios where some patients lack specific basic information due to emergency admission, incomplete medical records, or incomplete cognitive scales, this unit is configured with a weight normalization mechanism. If a certain piece of preoperative basic information cannot be obtained, the system does not directly default to the corresponding high-risk or no-risk level to prevent introducing assessment bias. The system removes the missing item from the calculation formula and proportionally amplifies the weights of the remaining valid basic features, so that the arithmetic sum of the weights involved in the actual calculation is normalized to 1.0. The final weighted summation generates a continuous value between 0 and 1, which is established as the basic brain vulnerability score. This score is transmitted to the coupling integration and risk output module as an inherent base-modulating variable for the final risk prediction.

[0085] See attached document Figure 5 The coupling integration and risk output module is responsible for aggregating the multidimensional independent features generated by the front-end status assessment module and performing cross-dimensional interactive risk calculations based on these features.

[0086] The general technical principle underlying this assessment process is that, under normal physiological homeostasis, a tight metabolic coupling mechanism exists within the central nervous system. A transient increase in neuronal activity potential prompts adjacent astrocytes to synchronously upregulate substrate uptake and lactate output rates, thereby maintaining a supply-demand balance. However, under severe perioperative stress, if neuronal energy demand surges while the mediating function of astrocytes' physiological energy supply declines, it will lead to a temporal and spatial mismatch between energy supply and demand. Simply performing a linear weighted sum of the two cannot objectively reveal the nonlinear damage aggravation effect caused by local metabolic substrate depletion under this dual injury.

[0087] This embodiment constructs a mathematical representation of supply-demand mismatch. Its calculation process specifically includes: In step S601, the coupling integration and risk output module synchronously acquires the neural activity demand score and astrocyte energy supply impairment score within the current analysis time window output by the state assessment module. Since these two scores originate from different physiological signal and biochemical marker characteristic systems, their original values ​​inherently differ in absolute scale and fluctuation range. To establish a unified mathematical measurement benchmark, the system performs synchronous dimensionless processing on them.

[0088] The system extracts the historical minimum and maximum values ​​of the subject's neural activity demand score and astrocyte energy supply impairment score within the current monitoring span. The system uses a linear transformation operation of maximum-minimum scaling to calculate the difference between the current feature score and its corresponding historical minimum value, and then divides this difference by the range between the historical maximum and minimum values. To avoid computational bias and ensure the system's operational stability when the subject's physiological characteristics remain constant (i.e., the historical maximum and minimum values ​​are occasionally equal), the system routinely adds a small positive real number to the denominator of this range. (Usually 10) -6 ~10 -8 After smoothing transformation, the scores for neural activity demand and astrocyte energy supply disorder were mapped to a standardized range of 0 to 1.

[0089] In step S602, after completing the unified normalization of the data dimensions, the coupling integration and risk output module performs a non-negative truncated overflow combination operation on these two normalized features to calculate the final interaction risk amplification score. This step aims to mathematically separate and quantify the degree of supply-demand mismatch deterioration beyond conventional physiological compensation. Its core calculation logic is shown in the following formula: ; in, The interaction risk amplification score is used independently to characterize the degree of additional mismatch damage that occurs when the energy demand of neural activity surges while the energy support capacity of astrocytes lags behind. It is a non-negative truncation function, designed to prevent invalid calculation results in non-mismatch states from interfering with the system's total score; This represents the normalized score for neural activity demand. This represents the normalized score for star glue power supply barrier. The constant 1 is the threshold for the sum of normalized scores. Its physiological meaning is: when both the normalized neural activity demand score and the astrocyte energy supply impairment score are at their respective historical monitoring median levels, their sum is 1.0, representing the critical equilibrium point of the basal compensatory state. This is achieved by arithmetically adding the normalized neural activity demand score and the astrocyte energy supply impairment score, and inputting the overflow difference exceeding the constant 1 into a non-negative truncation function. When the sum of the two normalized scores is less than or equal to the preset constant threshold 1, it indicates that the current increase in neuronal energy demand is still within the support range of the residual energy supply capacity of astrocytes. At this time, the system's basal compensatory buffer mechanism has not been exhausted, and the two have not yet formed a damaging interactive amplification effect. The system assigns a value of 0 to the interaction risk amplification score within the specified time window. Conversely, when the sum of the two scores is greater than 1, it indicates a simultaneous overlap between high metabolic demand and inefficient astrocyte-mediated support. In this case, the excess value exceeding 1 reflects the numerical difference beyond compensation. This numerical difference is fully preserved and established as a continuous interaction risk amplification score, providing amplified risk input parameter for the subsequent construction of a global imbalance early warning index.

[0090] See attached document Figure 5 The coupling integration and risk output module is responsible for aggregating all independent assessment features generated from the front end to generate a unified quantitative risk assessment index and clinical auxiliary decision-making information.

[0091] The general technical principle underlying this assessment process is that perioperative delayed neurocognitive recovery is not caused by a single-dimensional physiological abnormality, but rather by the intertwined and combined effects of multiple factors in time and space, including the patient's basic vulnerability, transient surges in neural activity, lack of metabolic substrates, and decline in the astrocyte energy network. To avoid clinical decision-making confusion caused by multi-source heterogeneous parameters, the system globally couples multi-dimensional physical variables within a mathematical space, mapping abstract dimensionless scores to probabilities of occurrence that conform to clinical pathogenesis. The specific process of constructing this comprehensive imbalance index and outputting the risk includes: In step S701, the coupling integration and risk output module extracts the neural activity demand score, glucose-lactic acid metabolic resource deficiency score, astrocyte energy supply impairment score, basic brain vulnerability score, and interaction risk amplification score within the current time window, and performs global weight fusion. As a preferred method, its core calculation logic is shown in the following formula: ; in, The index represents the imbalance between energy supply and demand of neural activity mediated by astrocytes during the perioperative period. This index comprehensively reflects the energy demand end of neurons, the energy supply end of peripheral blood, the microscopic astrocyte mediating end, the patient's inherent susceptibility end, and the interactive mismatch end caused by the time mismatch between supply and demand, forming a complete metabolic coupling assessment closed loop. This indicates the score for neural activity demand; This indicates insufficient glucose-lactic acid metabolism resources; This indicates the score for star glue power supply obstacle; This indicates the score for amplified interaction risks; This indicates a basic brain vulnerability score; ~ These are the preset weighting coefficients for the corresponding feature scores.

[0092] Based on the different pathological contributions of the above five dimensions to the pathogenesis of delayed neurocognitive recovery. ~ The initial configuration values ​​can be set to 0.15, 0.25, 0.25, 0.20 and 0.15 respectively. The above weight coefficients are obtained by fitting the clinical training cohort, and the arithmetic sum of each weight coefficient is configured to 1.0.

[0093] The purpose of the above parameter settings is to highlight the main driving role of insufficient metabolic resources, impaired astrocyte-colloid energy supply, and interaction risk mismatch in brain injury, while preserving the moderating effect of neural demand base and basic vulnerability. To ensure the robustness of the algorithm in real clinical applications, and to address abnormal situations where the continuous absence of certain specific physiological signals or laboratory results renders a single score completely uncalculation, the system activates a weight normalization mechanism. The system removes score items that become invalid due to missing data from the calculation formula and proportionally amplifies the weights of the remaining valid score items, so that the arithmetic sum of the weights involved in the actual calculation is renormalized to 1.0, thereby avoiding output evaluation bias caused by missing parameters.

[0094] In step S702, after constructing the comprehensive imbalance index, the coupling integration and risk output module converts the linear dimensionless score into a predicted probability of disease incidence. Considering that the incidence of clinical physiological complications typically exhibits an S-shaped distribution with the accumulation of risk factors—that is, low-level imbalances are compensated for by the body, while the risk increases non-linearly after exceeding a critical point—the system uses a logistic regression mapping function for probability conversion. Its calculation logic is shown in the following formula: ; in, It is a natural constant; This represents the probability of delayed neurocognitive recovery during the perioperative period, with its numerical range mapped to 0–1; The imbalance index calculated in the above steps; This represents the intercept parameter of the regression model; This represents the slope weight parameter corresponding to the imbalance index.

[0095] In one exemplary implementation, when a clinical training cohort has not yet been established at the center, the following can be adopted: = 2.197 and =4.394 was used as the initial probability mapping parameter. To achieve continuous iteration of the system's prediction accuracy, the system uses historical multicenter perioperative clinical cohorts... As single-dimensional or multi-dimensional feature inputs, the actual delayed neurocognitive recovery outcome, clinically confirmed in the early postoperative period (especially within 30 days postoperatively), is used as the binary classification supervision label. Iterative solutions are obtained using a conventional logistic regression algorithm, and the intercept parameter is dynamically calibrated and updated. With model parameters This allows the predicted probability to better fit the distribution of real-world data.

[0096] In step S703, the coupling integration and risk output module compares the calculated predicted probability with the system's preset risk grading judgment threshold (obtained by fitting the clinical training queue) and outputs the metabolic supply and demand imbalance warning status and structured clinical auxiliary intervention prompts to the display terminal.

[0097] The system has two preset evaluation thresholds: a threshold for excluding low-risk cases. Set the high-risk warning threshold to 0.20. Set to 0.40. When When the value is less than 0.20, the system outputs a low-risk status indicator; when 0.20 ≤ When <0.40, the system outputs a medium-risk status indicator and triggers a medium-level warning; when When the value is ≥0.40, the system outputs a high-risk status indicator and triggers an advanced alarm.

[0098] While issuing high-risk warnings, the module executes an automatic anomaly feature backtracking mechanism. The system compares the relative contribution of each sub-item score leading to the increased imbalance index to the total score, identifying the core cause. If the glucose-lactic acid metabolic resource deficiency score dominates, the system terminal simultaneously generates a brain metabolic resource insufficiency warning or a blood glucose fluctuation management warning; if the astrocyte-collagen energy supply impairment score dominates, the system terminal simultaneously generates a astrocyte-collagen energy insufficiency warning or a postoperative cognitive function follow-up warning; if the neural activity demand score or the interaction risk amplification score dominates, an increased neural activity demand warning or an anesthesia depth adjustment warning is generated. This structured backtracking feedback mechanism realizes the technical transformation from single numerical alarms to specific and actionable medical guidance, but this auxiliary prompt information does not directly replace the doctor's diagnostic or treatment decisions.

[0099] Specific application examples: Background setting: The subject was a 68-year-old male patient scheduled for an open radical gastrectomy for gastric cancer, expected to last longer than 4 hours, under general anesthesia. The patient had a history of hypertension, which was well-controlled with long-term medication, and no history of diabetes or cerebrovascular disease. His preoperative Montreal Cognitive Assessment score was 24, placing him in the mild cognitive impairment range. The current assessment endpoint was set at 25 minutes after the anesthesia recovery period following surgery.

[0100] Implementation steps and numerical calculations: In step S10, the data acquisition and processing module acquires multi-source data before and after 25 minutes of the patient's anesthesia recovery period. After system time reference alignment and extreme value scaling normalization, a standardized perioperative metabolic coupling assessment dataset is generated for use by downstream units.

[0101] In step S20, the neural activity demand assessment unit calls the normalized EEG feature data of the current time window and obtains: EEG energy score of 0.80, EEG complexity recovery score of 0.70, burst inhibition recovery score of 0.90, and awakening conversion score of 0.80.

[0102] Based on the neural activity demand score calculation logic, the system multiplies the above four normalized feature scores by their preset weights and then performs a summation operation. In this embodiment, the preset weights are set to 0.30, 0.25, 0.20, and 0.25, respectively. 0.30×0.80+0.25×0.70+0.20×0.90+0.25×0.80=0.795. This value reflects that the subject is in a high-energy-consuming stage of neural activity recovery.

[0103] In step S30, the brain metabolic resource supply assessment unit jointly derived the current glucose substrate availability insufficiency score as 0.60 and lactate-like metabolic pressure score as 0.50.

[0104] The system performs a weighted average calculation on the above two scores based on the set resource allocation parameters. In this embodiment, the weight of both scores is configured to 0.50. 0.50 × 0.60 + 0.50 × 0.50 = 0.550. This value quantifies the current scarcity of metabolic resources in brain tissue.

[0105] In step S40, the astral colloid energy assessment unit retrieved biomarker data. It was found that the relative abundance of monocarboxylic acid transporters and aldolases decreased, while the abundance of glial fibrillary acidic protein increased slightly.

[0106] The system calculates the deviation of each marker from the baseline based on the direction determination rules, and sets protective items that are opposite to the trend of damage risk change to 0 to eliminate interference. Subsequently, each effective damage sub-score is proportionally restandardized and weighted and summed, outputting an astrocyte energy supply impairment score of 0.650. This value quantifies the degree of impairment in the mediating function of astrocytes in transporting lactate substrates to neurons.

[0107] Step S50: The brain vulnerability assessment unit extracts basic information scores: age stratification score is 0.50; cognitive function conversion value is 0.50; hypertension score is 0.50; surgical burden score is 1.00; diabetes and cerebrovascular disease scores are both 0.

[0108] The system performs a weighted linear combination calculation on the above scores based on preset weight coefficients. In this embodiment, the weight coefficients are 0.20, 0.25, 0.15, 0.20, 0.10, and 0.10, respectively. 0.20×0.50+0.25×0.50+0.15×0+0.20×0+0.10×0.50+0.10×1.00=0.375. This value serves as the baseline brain vulnerability score.

[0109] Step S60: The coupling integration and risk output module synchronously obtains the normalized neural activity demand score. The value is 0.795, which is consistent with the score for energy supply barrier in star glue. The value is 0.650. The system call formula for non-negative truncation overflow combination calculation is as follows: ; Substitute the values ​​and perform the calculation: This eigenvalue separates and quantifies the additional damaging effects that occur when a surge in neural energy demand overlaps with a decline in astrocyte energy supply.

[0110] Step S70: The system obtains the scores of the five dimensions, performs global weight fusion, and assigns corresponding weights. ~ Configure them to 0.15, 0.25, 0.25, 0.20, and 0.15 respectively, and call the comprehensive imbalance index calculation formula: ; Substitute the values ​​and perform the calculation: =0.15×0.795+0.25×0.550+0.25×0.650+0.20×0.445+0.15×0.375=0.119+0.138+0.163+0.089+0.056=0.565.

[0111] Subsequently, the system calls the logistic regression function to calculate the predicted probability of disease incidence, where the initial parameters... It is -2.197. The value is 4.394. ; Substitute the values ​​and perform the calculation: Mapping variable value = 2.197 + 4.394 × 0.565 = 0.286; Incidence rate .

[0112] Comparative conclusions and dynamic evolution trend analysis: To visually present the pathological significance of the above calculation results over time, please refer to the appendix. Figure 6 The attached figure illustrates the dynamic evolution of key characteristic scores in subjects during the early recovery period (0–60 minutes). Figure 6 The horizontal axis represents the evaluation time in minutes; the vertical axis represents the quantitative value of the indicator, ranging from 0 to 1.

[0113] Based on system comparison and determination, the subject's current incidence probability is 0.571, which is greater than the high-risk warning threshold of 0.40. Using the incidence probability of 0.40 corresponding to the set high-risk warning threshold, the system's overall imbalance index warning baseline is estimated to be approximately 0.41. Combined with... Figure 6 The intersection trajectory of the curves can be seen: around the 25th minute, even Figure 6 The neural activity demand shown by the solid line has not yet reached its peak for the entire cycle, but due to the superposition of... Figure 6 The persistent, damaged star glue power supply disorder, indicated by the dashed line, led to... Figure 6 The overall imbalance index, represented by the dotted line, continues to rise and has broken through the high-risk warning baseline represented by the dotted line.

[0114] If relying on conventional multi-parameter monitoring, the patient's EEG was on a normal awakening trajectory, and peripheral blood glucose was normal, so the system would not trigger an alert. However, this assessment system effectively identified impaired microscopic energy supply and supply-demand mismatch in astrocytes, thus outputting a high-risk indicator. Since the score for astrocyte energy supply impairment accounts for a relatively high proportion of the total score, the system terminal simultaneously generates an astrocyte energy deficiency warning, suggesting auxiliary intervention information to ensure a smooth transition and avoid excessive consumption of metabolic resources caused by rapid awakening.

[0115] Experimental verification and effect comparison: To verify the effectiveness of this assessment system, the research team retrieved the perioperative characteristics database of 200 elderly patients recently treated at our center. Among them, 95 patients experienced delayed neurocognitive recovery during postoperative follow-up, while the remaining 105 did not.

[0116] The verification scheme inputs the same batch of data into: a conventional basic early warning system based on EEG parameters and peripheral blood oxygen and blood pressure thresholds; an internal control model with RMS interaction terms removed (only linear feature weighting is applied to demand, supply, star glue and vulnerability scores); and the system of this invention, which includes complete RMS combined calculations, for testing and comparison.

[0117] Verification conclusion: The area under the receiver operating characteristic curve (AUC) of the conventional basic early warning system in identifying this complication was only 0.68, indicating low sensitivity to transient imbalances in the local brain microenvironment during the recovery period.

[0118] The area under the curve of the internal control model with RMS interaction removed in the test set was 0.79, which is higher than that of the conventional early warning system. However, when dealing with the high-concurrency pathological state of high neuronal energy consumption and impaired astrocyte energy supply, it is prone to missed alarms because it cannot reflect the nonlinear superposition of damage.

[0119] The complete system of this invention further improves the area under the curve (AUC) to 0.88 on the test set by introducing the RMS interactive risk amplification calculation mechanism. Ablation experiments demonstrate the core contribution of the RMS feature extraction formula itself in quantifying nonlinear superposition damage. For elderly patients with a baseline brain vulnerability score greater than 0.30, the system of this invention can identify the metabolic compensation threshold an average of 12 minutes in advance, effectively improving the accuracy of clinical early warning.

[0120] The differences in AUC among the three models were statistically significant according to the DeLong test (complete system vs. model without RMS control, P<0.01; complete system vs. conventional basic early warning system, P<0.001). The 95% confidence interval for the complete system was 0.83–0.93. The model discrimination was verified by 10-fold cross-validation, and the AUC fluctuation range was less than 0.03.

Claims

1. A perioperative brain metabolic coupling risk assessment system, characterized in that, It includes a data acquisition and processing module, which is used to acquire multi-source data of subjects during the perioperative period and perform time-based alignment and normalization to generate a standardized perioperative metabolic coupling assessment dataset. The status assessment module is used to receive the perioperative metabolic coupling assessment dataset, output the neural activity demand score through the neural activity demand assessment unit, output the glucose-lactic acid metabolic resource deficiency score through the brain metabolic resource supply assessment unit, assign positive or negative polarity determination coefficients according to the physiological characteristics of the biomarkers, calculate the energy supply impairment sub-scores corresponding to each specific biomarker by retaining the changing trend consistent with the physiological impairment risk through a non-negative truncation function, perform weighted summation on the energy supply impairment sub-scores through the astrocyte-colloid energy supply assessment unit to output the astrocyte-colloid energy supply impairment score, and output the basic brain vulnerability score through the brain vulnerability assessment unit. The coupling integration and risk output module is used to perform dimensionless processing on the neural activity demand score and the star colloid energy supply impairment score to obtain a normalized score, to perform arithmetic summation on each of the normalized scores to obtain the sum of normalized scores, and to input the difference between the sum of normalized scores and a constant 1 into a non-negative truncation function to generate an interactive risk amplification score. When the sum of the normalized scores is not greater than a constant 1, the interactive risk amplification score is assigned a value of 0; The neural activity demand score, glucose-lactic acid metabolic resource deficiency score, astrocyte energy supply impairment score, interaction risk amplification score, and basic brain vulnerability score are globally weighted and integrated to construct a comprehensive imbalance index. The probability of occurrence is predicted based on the comprehensive imbalance index. The predicted probability of occurrence is compared with the preset risk classification judgment threshold obtained by fitting a clinical training cohort, and the metabolic supply and demand imbalance warning status and auxiliary prompt information are output.

2. The Perioperative brain metabolism coupling risk assessment system according to claim 1, characterized in that, The perioperative multi-source data includes continuous physiological monitoring data, intermittent biochemical data, and specific biomarker data. The process of the data acquisition and processing module performing time base alignment includes: The monitoring time span is divided into continuous analysis time windows. The average and maximum values ​​of the continuous physiological monitoring data within the same analysis time window are extracted. The detection values ​​of the intermittent biochemical data and the specific biomarker data are mapped onto the time axis. Within the corresponding analysis time window, the detection values ​​are maintained as physiological state reference values ​​for subsequent adjacent analysis time windows.

3. The system of claim 1, wherein, The neural activity demand assessment unit extracts the degree of brainwave energy recovery, the degree of brainwave complexity recovery, the degree of burst inhibition recovery, and the degree of awakening from the continuous brainwave time series features in the perioperative metabolic coupling assessment dataset; The neural activity demand assessment unit performs dimensionless scaling on the degree of brainwave energy recovery, the degree of brainwave complexity recovery, the degree of burst inhibition recovery, and the degree of awakening transition, and outputs the neural activity demand score through a weighted summation operation; When performing the division operation with the dimensionless scaling process, a small positive real number is superimposed on the denominator.

4. The Perioperative brain metabolism coupling risk assessment system of claim 1, wherein, The brain metabolic resource supply assessment unit extracts the change in total hemoglobin concentration, the rate of change in oxyhemoglobin concentration and mean arterial pressure to construct brain perfusion proxy indicators, and calculates the estimated value of local brain glucose uptake rate by combining arterial blood glucose concentration. The brain metabolic resource supply assessment unit uses arterial blood oxygen content and tissue oxygen extraction rate to estimate the brain oxygen metabolism rate. It then divides the estimated brain oxygen metabolism rate with the estimated local brain glucose uptake rate to obtain the oxygen-glucose metabolism ratio. The unit calculates the deviation gap between the oxygen-glucose metabolism ratio and a preset reference ratio. It constructs an instantaneous anaerobic metabolic load by combining the estimated local brain glucose uptake rate with the instantaneous anaerobic metabolic load. Based on the estimated local brain glucose uptake rate and the instantaneous anaerobic metabolic load, it performs weighted integration and outputs the glucose-lactic acid metabolic resource deficiency score.

5. The Perioperative brain metabolism coupling risk assessment system of claim 1, wherein, Before calculating the energy supply barrier sub-scores corresponding to each specific biomarker, the method further includes: Extract the absolute difference between the current monitored value and the reference baseline value of the specific biomarker.

6. The Perioperative brain metabolism coupling risk assessment system according to claim 5, wherein, The brain vulnerability assessment unit extracts multidimensional clinical baseline characteristics, including age, preoperative cognitive function, diabetic status, history of cerebrovascular disease, history of hypertension, and surgical trauma load. The brain vulnerability assessment unit converts the multidimensional clinical basic features into standardized sub-item scores ranging from 0 to 1 based on a preset segmented assignment rule obtained by fitting a clinical training cohort. The standardized sub-item scores are then weighted and linearly combined to output the basic brain vulnerability score.

7. The Perioperative brain metabolism coupling risk assessment system according to claim 6, characterized in that, When the specific biomarker or the multidimensional clinical baseline feature is missing, the status assessment module removes the missing item from the weighted calculation, performs proportional amplification of the weights of the remaining valid items, and normalizes the arithmetic sum of the weights of the remaining valid items involved in the calculation to a value of 1.

8. The Perioperative brain metabolism coupling risk assessment system of claim 1, wherein, The process by which the coupling integration and risk output module constructs the comprehensive imbalance index and calculates the predicted probability of occurrence includes: The global weight integration is completed by multiplying the neural activity demand score, the glucose-lactic acid metabolic resource deficiency score, the astrocyte energy supply impairment score, the interaction risk amplification score, and the basic brain vulnerability score by a preset weight coefficient obtained by fitting a clinical training cohort, and performing a weighted summation to generate the comprehensive imbalance index. Substituting the comprehensive imbalance index into the logistic regression mapping function, the predicted probability of occurrence is output with a value in the range of 0 to 1.

9. A method of perioperative brain metabolic coupling risk assessment, characterized in that, The perioperative brain-metabolic coupling risk assessment system according to any one of claims 1-8 includes the following steps: We acquired multi-source perioperative data from subjects and performed time-baseline alignment and normalization to generate a standardized perioperative metabolic coupling assessment dataset. The system receives the perioperative metabolic coupling assessment dataset, outputs a neural activity demand score through the neural activity demand assessment unit, assigns a positive or negative polarity determination coefficient based on the physiological characteristics of the biomarkers, calculates the energy supply impairment sub-scores corresponding to each specific biomarker by retaining the changing trend consistent with the physiological impairment risk through a non-negative truncation function, performs a weighted summation of the energy supply impairment sub-scores through the astrocyte-colloid energy supply assessment unit to output the astrocyte-colloid energy supply impairment score, and outputs a basic brain vulnerability score through the brain vulnerability assessment unit. The neural activity demand score and the astrocyte energy supply impairment score are dimensionlessly processed to obtain normalized scores. The normalized scores are arithmetically added together to obtain the sum of normalized scores. The difference between the sum of normalized scores and a constant 1 is input into a non-negative truncation function to generate an interaction risk amplification score. When the sum of normalized scores is not greater than the constant 1, the interaction risk amplification score is assigned a value of 0. The neural activity demand score, glucose-lactic acid metabolic resource deficiency score, astrocyte energy supply impairment score, interaction risk amplification score, and basic brain vulnerability score are globally weighted and integrated to construct a comprehensive imbalance index. The probability of occurrence is predicted based on the comprehensive imbalance index. The predicted probability of occurrence is compared with a preset risk grading threshold obtained by fitting a clinical training cohort. The metabolic supply and demand imbalance warning status and auxiliary prompt information are output. The auxiliary prompt information does not directly replace clinical diagnostic decisions.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by the processor, it implements the perioperative brain-metabolic coupling risk assessment method as described in claim 9.