Real-time monitoring system and method for adverse events during perioperative period of hip fracture surgery in the elderly

Abstract

The application discloses a kind of elderly hip fracture surgery perioperative adverse event real-time monitoring system and method, it is related to perioperative monitoring technical field.The method includes: real-time acquisition perioperative monitoring data, carries out data pre-processing, and extracts benchmark feature;Identify the decline segment of cerebral oxygen saturation below benchmark feature, jointly assess the decline amplitude and persistence, obtain the weak trend characterization of brain oxygen;Analysis of the relative change characteristics of serum inflammatory markers, form the weak trend characterization of inflammation;Assess the deviation degree of vital signs relative to benchmark, obtain the vital signs fluctuation characterization, and further judge the delirium precursor risk, output risk prompt;Visual display various weak trend characterizations and delirium precursor risk, link label risk trigger information and contribution source.The present application solves the technical problems that the prior art is difficult to extract and identify early weak abnormal trend in perioperative multi-source physiological signals.

Description

Technical Field

[0001] This invention relates to the field of perioperative monitoring technology, and in particular to a real-time monitoring system and method for adverse events during the perioperative period of hip fracture surgery in elderly patients. Background Technology

[0002] Hip fractures in the elderly have become a common and serious injury type against the backdrop of an aging population, and their perioperative management involves the need for continuous monitoring and risk identification across multiple systems and stages. With the increase in surgical volume and the rising average age of patients, clinicians are placing increasing emphasis on continuous observation of perioperative physiological changes, dynamic risk management, and comprehensive information support. In the modern medical environment, the application of physiological monitoring equipment and information systems in perioperative management is becoming increasingly widespread, providing the foundation for real-time, objective, and data-driven presentation of patient status, and also promoting the development of perioperative risk management from experience-based to data-driven approaches.

[0003] For example, invention patent CN119970412A discloses an intermittent hyperoxia system that can monitor environmental parameters such as oxygen concentration, temperature, humidity, and pressure in a hyperoxia chamber in real time. It also features alarm and automatic correction functions, ensuring safety during hyperoxia treatment and preventing harm to the subject due to parameters exceeding set ranges. This intermittent hyperoxia system can be used for intermittent hyperoxia intervention in the acute phase of ischemic brain injury and other types of brain injury, effectively improving neurological function and reducing cerebral infarction volume, demonstrating significant application value and market potential.

[0004] For example, the invention patent with publication number CN119385792A discloses a microenvironment detection system for osteomyelitis wounds, which relates to the field of microenvironment detection technology for osteomyelitis wounds. It includes a leg support, with a detection support at the top. The detection support includes a positioning support and a movable support. A sensor mechanism is located at the end of the movable support near the latch, and a display is located on one side of the leg support. The sensor mechanism and the display are electrically connected via a data transmission line. This invention allows for adjustment of the height and position of the detection support, facilitating the selection of a suitable wound position based on the patient's height and weight. By integrating a sensor array, electrode system, signal converter, and data processing module into a portable device, it can monitor in real time relevant indicators such as pH, humidity, and pathogenic microorganisms around the bone infection wound. It effectively integrates and analyzes multiple environmental parameters, providing more comprehensive microenvironment information and enabling long-term accurate monitoring of bone infection wounds.

[0005] However, elderly patients with hip fractures often exhibit a series of slowly evolving, mild, and scattered early abnormal signals across different physiological systems during the perioperative period. These include mildly decreased cerebral oxygenation, persistently elevated serum inflammatory markers, and worsening fluctuations in vital signs. These warning signs typically do not reach the traditional clinical diagnostic thresholds, and their changes are gradual and insidious, making continuous monitoring and comprehensive assessment difficult during manual rounds, easily leading to the overlooking of potential risk stages. In particular, perioperative adverse events such as delirium are characterized by multiple factors driving them, and their precursor states often rapidly evolve from slight deviations to overt risks within a short period. Existing monitoring models struggle to identify subtle trends across indicators and time scales, and are unable to promptly indicate the direction of risk evolution, affecting the foresight and continuity of perioperative management.

[0006] Therefore, in order to address the above issues, there is an urgent need for a real-time monitoring system and method for adverse events during the perioperative period of hip fracture surgery in the elderly. Summary of the Invention

[0007] To address the technical problem of existing technologies that struggle to promptly identify early abnormal trends in perioperative multi-source physiological signals that are subtle, slowly evolving, and dispersed across indicators, thus hindering continuous, accurate, and prospective identification of precursors to potential adverse events, this invention provides a real-time monitoring system and method for perioperative adverse events in elderly patients undergoing hip fracture surgery. The technical solution is as follows: On the one hand, a method for real-time monitoring of adverse events during the perioperative period of hip fracture surgery in elderly patients is provided. This method includes: S1, real-time acquisition of perioperative monitoring data, data preprocessing of the perioperative monitoring data, and extraction of baseline features for each perioperative monitoring data point based on a sliding time window; S2, based on the perioperative monitoring data, identification of continuous segments where cerebral oxygen saturation falls below the baseline feature, joint assessment of the magnitude and persistence of the decline to obtain a cerebral oxygen weakness trend characterization, and generation of a cerebral oxygen weakness trend feature vector; S3, based on the perioperative monitoring data and baseline features, analysis of the relative changes in serum inflammatory markers, and extraction of features reflecting the slow accumulation of inflammation. The drift intensity of the product is used to form a weak inflammatory trend representation and generate an inflammatory feature vector; S4, combined with the temporal fluctuation characteristics of vital sign data in perioperative monitoring data, the degree of deviation from the baseline is assessed to obtain a vital sign fluctuation representation. The weak cerebral oxygenation trend representation, weak inflammatory trend representation, and vital sign fluctuation representation are integrated to determine the risk of delirium precursor and output risk warning information based on the delirium precursor risk; S5, the temporal evolution of the weak cerebral oxygenation trend representation, weak inflammatory trend representation, vital sign fluctuation representation, and delirium precursor risk is visualized and displayed, and the risk trigger information and main contribution sources are linked and labeled to realize the feedback recording of risk changes.

[0008] Optionally, the process of real-time acquisition of perioperative monitoring data, data preprocessing of the perioperative monitoring data, and extraction of baseline features of each perioperative monitoring data based on a sliding time window is as follows: Real-time acquisition of the patient's perioperative monitoring data, including: cerebral oxygen saturation, heart rate, respiratory rate, systolic blood pressure, interleukin-6 concentration, and C-reactive protein concentration; median filtering and moving average of continuous signals of cerebral oxygen saturation, heart rate, respiratory rate, and systolic blood pressure to remove spike noise and spurious signals from probe loosening; time stamp unification and sampling alignment of perioperative monitoring data from different sampling frequencies; and linear interpolation to construct continuous time series of discrete serum inflammatory indicators such as interleukin-6 concentration and C-reactive protein concentration; and processing of the perioperative monitoring data... Normalization was performed. Based on a sliding time window, the mean and standard deviation of cerebral oxygen saturation, interleukin-6 concentration, and C-reactive protein concentration were calculated, as well as the standard deviations of heart rate, respiratory rate, and systolic blood pressure. The median of the mean and standard deviation of perioperative monitoring data within each of the N historical sliding time windows was selected to obtain the baseline mean and standard deviation of cerebral oxygen, interleukin-6, C-reactive protein, heart rate, respiratory rate, and systolic blood pressure, thus constructing a baseline parameter set. A perioperative event monitoring database was established, and the raw and preprocessed perioperative monitoring data, along with the baseline parameter set, were written into the perioperative event monitoring database.

[0009] Optionally, based on perioperative monitoring data, the process of identifying consecutive segments of decreased cerebral oxygen saturation below the baseline characteristic and jointly assessing the magnitude and persistence of the decrease to obtain a characterization of weak cerebral oxygenation trends is as follows: Read the baseline mean and standard deviation of cerebral oxygen saturation; scan the cerebral oxygen saturation sequence within N historical sliding time windows point by point in chronological order; when cerebral oxygen saturation changes from not lower than the baseline mean to being lower than the baseline mean, mark it as the starting point of a decreasing segment; when cerebral oxygen saturation recovers to not lower than the baseline mean again, mark it as the ending point of the decreasing segment; record the number of consecutive sampling points within each decreasing segment as the number of continuous sampling points for the decreasing segment; take the median of the number of continuous sampling points for all decreasing segments to obtain... The baseline median decline is calculated. Preprocessed cerebral oxygen saturation (COS) sequences are read chronologically from the perioperative event monitoring database in real time. For each time point, the COS sequence is traced back to determine the number of consecutive sampling points where COS is less than the baseline mean, thus obtaining the current duration of continuous decline. Within the duration of continuous decline, the difference between the baseline mean and the corresponding COS is calculated and non-negated. The non-negated differences are accumulated to obtain the cumulative decline depth value. The baseline standard deviation is added to a minimum constant value, and multiplied by the baseline median decline to obtain the normalized scale value of COS decline. The cumulative decline depth value is divided by the normalized scale value of COS decline, and then added to a constant and the natural logarithm is taken to obtain the assessment value of micro-decline in COS.

[0010] Optionally, the specific process for generating the brain oxygen weakness trend feature vector is as follows: calculate the brain oxygen micro-decline assessment value in real time, construct a brain oxygen weakness trend sequence based on the brain oxygen micro-decline assessment values ​​at consecutive time points, and perform linear interpolation on abnormal jumps and short-term missing brain oxygen micro-decline assessment values; generate a brain oxygen feature vector based on the brain oxygen micro-decline assessment value at the current time, the change range of the brain oxygen micro-decline assessment value, and the current continuous decline duration; write the brain oxygen feature vector into the perioperative event monitoring database and perform timestamp alignment.

[0011] Optionally, based on perioperative monitoring data and baseline characteristics, the relative changes in serum inflammatory markers are analyzed, and the drift intensity reflecting the slow accumulation of inflammation is extracted to form a weak trend characterization of inflammation. The specific process for generating an inflammatory feature vector is as follows: Interleukin-6 (IL-6) and C-reactive protein (CRP) concentrations are read in real time, and a baseline parameter set is obtained; the normalized increase in IL-6 is obtained by subtracting the baseline mean of IL-6 from the current IL-6 concentration and dividing by the baseline standard deviation of IL-6, and the normalized increase in IL-6 is non-negatively processed; the normalized increase in CRP is obtained by subtracting the baseline mean of CRP from the current CRP concentration and dividing by the baseline standard deviation of CRP. The inflammatory drift intensity was calculated by normalizing the increase in C-reactive protein (CRP) and then non-negating it. The sum of squares of the normalized increases in CRP and CRP was calculated and the square root was taken to obtain the inflammatory drift intensity value. The inflammatory drift intensity value was calculated in real time, and an inflammatory drift intensity trend sequence was constructed. Abnormal jumps and short-term missing inflammatory drift intensity values ​​were linearly imputed. The duration of inflammatory elevation was determined based on consecutive moments when the inflammatory drift intensity value continuously increased compared to the previous moment. An inflammatory feature vector was constructed based on the current inflammatory drift intensity value, the magnitude of the change in inflammatory drift intensity value, and the duration of inflammatory elevation. This inflammatory feature vector was written into the perioperative event monitoring database and timestamped.

[0012] Optionally, by combining the temporal fluctuation characteristics of vital sign data in perioperative monitoring data, the degree of deviation relative to the baseline is assessed, and the specific process of representing vital sign fluctuations is as follows: Calculate the standard deviation of heart rate, respiratory rate, and systolic blood pressure within the current sliding time window, and obtain the baseline standard deviations of heart rate, respiratory rate, and systolic blood pressure; subtract the baseline standard deviation of heart rate from the current standard deviation of heart rate, and divide by the sum of the baseline standard deviation of heart rate and the minimum constant value to obtain the normalized fluctuation of heart rate; subtract the respiratory rate from the current standard deviation of respiratory rate. The normalized fluctuation of respiratory rate is obtained by dividing the baseline standard deviation by the sum of the baseline standard deviation of respiratory rate and the minimum constant value. The normalized fluctuation of systolic blood pressure is obtained by subtracting the baseline standard deviation of systolic blood pressure from the current standard deviation of systolic blood pressure and dividing by the sum of the baseline standard deviation of systolic blood pressure and the minimum constant value. The normalized fluctuations of heart rate, respiratory rate, and systolic blood pressure are non-negatively processed respectively. The sum of squares of the non-negative normalized fluctuations of heart rate, respiratory rate, and systolic blood pressure are calculated and the square root is taken to obtain the vital sign fluctuation value.

[0013] Optionally, the specific process for determining the prodromal risk of delirium by integrating the weak trend of brain oxygenation, the weak trend of inflammation, and the fluctuation of vital signs is as follows: read the brain oxygenation micro-decline assessment value, the inflammation drift intensity value, and the vital sign fluctuation value at the corresponding time, and calculate the average value after adding the brain oxygenation micro-decline assessment value, the inflammation drift intensity value, and the vital sign fluctuation value to obtain the comprehensive weak sign intensity value; take the negative number of the comprehensive weak sign intensity value as the exponent for natural exponent calculation, and subtract the natural exponent calculation result from the constant to obtain the delirium prodromal risk value.

[0014] Optionally, the specific process of outputting risk warning information based on the delirium precursor risk is as follows: The delirium precursor risk value is calculated in real time and written into the perioperative event monitoring database. Short-term missing values ​​are linearly imputed, and a robust smoothing correction is performed on sudden jumps using in-window trend regression to obtain the delirium risk change trajectory. The delirium precursor risk value is compared with the risk threshold in real time. If the duration of consecutive delirium precursor risk values ​​exceeding the risk threshold exceeds the allowable time threshold, a delirium precursor risk warning is triggered, and warning information is generated, including the delirium precursor risk value, the magnitude of change in the delirium precursor risk value, the duration of the risk, and the assessment value of a slight decrease in cerebral oxygenation, the intensity value of inflammatory drift, and the fluctuation value of vital signs. Simultaneously, based on the relative magnitude of the assessment value of a slight decrease in cerebral oxygenation, the intensity value of inflammatory drift, and the fluctuation value of vital signs within the overall weak symptom intensity value, the main physiological source leading to the increased risk is inferred, and a corresponding risk warning is generated.

[0015] Optionally, the temporal evolution of weak brain oxygenation trend, weak inflammation trend, vital sign fluctuation, and delirium precursor risk is visualized and linked with the annotation of risk trigger information and main contributing sources. The specific process for recording the feedback of risk changes is as follows: Based on the continuously calculated brain oxygenation micro-decline assessment value, inflammation drift intensity value, vital sign fluctuation value, and delirium precursor risk value, various weak sign indicators and their corresponding timestamps are visualized and rendered, displaying the brain oxygenation weak trend curve, inflammation drift trend curve, vital sign fluctuation trend curve, and delirium precursor risk change curve, and marking the risk threshold, risk trigger time, and risk duration, while linking with perioperative monitoring data; based on the risk warning information, the increase in the proportion of brain oxygenation-related fluctuations, the increase in the proportion of inflammation-related parameter fluctuations, and the increase in the proportion of vital sign fluctuation indicators are displayed in a hierarchical manner; at the same time, a delirium precursor risk warning report is generated and written into the perioperative event monitoring database.

[0016] On the other hand, a real-time monitoring system for perioperative adverse events in elderly patients undergoing hip fracture surgery is provided. This system includes: a monitoring data acquisition and processing module for real-time acquisition of perioperative monitoring data, data preprocessing, and extraction of baseline features for each perioperative monitoring data point based on a sliding time window; a brain oxygenation weakness trend extraction module for identifying continuous segments of brain oxygen saturation below the baseline feature based on perioperative monitoring data, jointly assessing the magnitude and persistence of the decline to obtain a brain oxygenation weakness trend characterization, and generating a brain oxygenation weakness trend feature vector; and an inflammation weakness trend extraction module for analyzing the relative changes in serum inflammatory markers based on perioperative monitoring data and baseline features, and extracting features reflecting... The drift intensity of slowly accumulating inflammation forms a weak trend representation of inflammation and generates an inflammatory feature vector. The multimodal weak sign fusion module is used to combine the temporal fluctuation characteristics of vital sign data in perioperative monitoring data to assess the degree of deviation from the baseline, obtain a vital sign fluctuation representation, fuse the weak trend representation of brain oxygenation, the weak trend representation of inflammation, and the vital sign fluctuation representation to determine the risk of delirium prodrome, and output risk warning information based on the risk of delirium prodrome. The visualization monitoring feedback module is used to visualize the temporal evolution of the weak trend representation of brain oxygenation, the weak trend representation of inflammation, the vital sign fluctuation representation, and the risk of delirium prodrome, and link and label the risk trigger information and the main contributing sources to realize the feedback recording of risk changes.

[0017] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: (1) By constructing a weak trend representation of brain oxygenation, a weak trend representation of inflammation, and a vital sign fluctuation representation based on baseline features, this invention can extract continuous trend information from physiological signals with slight amplitude, slow change, and dispersed across indicators, thereby achieving prospective identification of early perioperative abnormalities and significantly improving the ability to capture weak signs.

[0018] (2) This invention constructs a fusion strategy for three types of weak symptom representations, which comprehensively reflects the patient's overall physiological deviation by quantifying the intensity of weak symptom, and further maps it to delirium precursor risk value, so that physiological changes across channels and scales can be assessed within a unified risk framework, thereby improving the stability and accuracy of risk assessment.

[0019] (3) By continuously tracking the time evolution of trend representation and combining risk thresholds and persistent conditions to trigger early warning, this invention can issue a prompt when abnormal signs are still in the precursor stage, realize real-time and dynamic monitoring of perioperative adverse events such as delirium, and reduce the lag caused by reliance on manual observation.

[0020] (4) This invention visualizes the trend curves and risk trajectories of brain oxygen, inflammation and vital signs, and generates corresponding prompts based on the relative contribution of each weak sign to the risk, making the source of risk interpretable and forming a complete risk evolution record, providing a basis for subsequent clinical management. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a method for real-time monitoring of adverse events during the perioperative period of hip fracture surgery in elderly patients, provided in an embodiment of the present invention. Figure 2 This is a structural diagram of the real-time monitoring system for perioperative adverse events in elderly patients undergoing hip fracture surgery provided in an embodiment of the present invention; Figure 3 This is a trend chart of delirium precursor risk values ​​provided in an embodiment of the present invention; Figure 4 This is a radar chart of three indicators of weak symptoms provided in an embodiment of the present invention. Detailed Implementation

[0023] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0024] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0025] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0026] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0027] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0028] This invention provides a method for real-time monitoring of adverse events during the perioperative period of hip fracture surgery in elderly patients, such as... Figure 1 The flowchart shown is for a real-time monitoring method of perioperative adverse events in elderly patients undergoing hip fracture surgery. The process of this method may include the following steps: S1, real-time acquisition of perioperative monitoring data, data preprocessing of perioperative monitoring data, and extraction of baseline features of each perioperative monitoring data based on a sliding time window; S2, based on perioperative monitoring data, identifies continuous segments of brain oxygen saturation that are lower than the baseline characteristics, jointly assesses the magnitude and duration of the decline, obtains a brain oxygen weakness trend characterization, and generates a brain oxygen weakness trend feature vector. S3, based on perioperative monitoring data and baseline characteristics, analyzes the relative changes of serum inflammatory markers, extracts the drift intensity reflecting the slow accumulation of inflammation, forms a weak trend characterization of inflammation, and generates an inflammatory feature vector; S4. By combining the temporal fluctuation characteristics of vital signs data in perioperative monitoring data, the degree of deviation from the baseline is assessed to obtain the vital signs fluctuation characterization. The characterization of weak brain oxygenation trend, weak inflammation trend, and vital signs fluctuation characterization are integrated to determine the risk of delirium prodrome and output risk warning information based on the risk of delirium prodrome. S5 visualizes the temporal evolution of weak brain oxygenation trend, weak inflammation trend, vital sign fluctuations, and delirium precursor risk, and links risk trigger information and main contributors to achieve feedback recording of risk changes.

[0029] Optionally, the specific process of real-time acquisition of perioperative monitoring data, data preprocessing of perioperative monitoring data, and extraction of baseline features of each perioperative monitoring data based on a sliding time window is as follows: Real-time acquisition of the patient's perioperative monitoring data, including: cerebral oxygen saturation, heart rate, respiratory rate, systolic blood pressure, interleukin-6 concentration, and C-reactive protein concentration; wherein, cerebral oxygen saturation is obtained by continuous output of a near-infrared spectroscopy cerebral oxygen monitor; heart rate is calculated and output in real time by the electrocardiogram module of a multi-parameter monitor based on the RR interval; respiratory rate is automatically calculated and output by a flow sensor based on chest impedance respiratory signals of the monitor; systolic blood pressure is measured stroke-by-stroke by an arterial pressure sensor under invasive monitoring conditions, and by cuff oscillation method under non-invasive monitoring conditions; interleukin-6 concentration is obtained by quantitative detection of venous blood samples by an immunoassay device; C-reactive protein concentration is obtained by quantitative determination of serum or plasma samples by a clinical biochemical analyzer, forming a discrete inflammatory marker sequence. Median filtering and moving averages were applied to continuous signals of cerebral oxygen saturation, heart rate, respiratory rate, and systolic blood pressure to remove spike noise and spurious signals from probe loosening. Median filtering preferably used a sliding window of 3 to 7 sampling points to suppress instantaneous abnormal peaks, while moving averages were used to smooth high-frequency fluctuations over short periods, ensuring the stability and repeatability of subsequent calculations. Perioperative monitoring data from different sampling frequencies were time-stamped and aligned. Continuous time series of discrete serum inflammatory markers, such as interleukin-6 concentration and C-reactive protein concentration, were constructed using linear interpolation. The time-stamping step normalized and aligned the recording times of each monitoring channel, ensuring that all monitoring data had a time-by-time correspondence on the same time axis, guaranteeing the accuracy of subsequent multimodal feature fusion. Linear interpolation maps discrete detection values ​​to a continuous sequence at the second level, avoiding the inability to identify trends due to data sparsity. Z-score normalization was performed on perioperative monitoring data to eliminate numerical differences caused by different monitoring units, making subsequent statistical calculations of baseline characteristics more robust. Based on a sliding time window, the mean and standard deviation of cerebral oxygen saturation, interleukin-6 concentration, and C-reactive protein concentration were calculated, as well as the standard deviations of heart rate, respiratory rate, and systolic blood pressure. The length of the sliding time window can be adaptively set according to the clinical monitoring rhythm, preferably between 60 and 180 seconds, to ensure sensitivity to short-term trend changes and avoid excessive noise interference; the time step is preferably 1 second. The mean within the window reflects the overall level, and the standard deviation reflects the degree of fluctuation; both together constitute the basis for stability measurement.The median of the mean and standard deviation of perioperative monitoring data within N historical sliding time windows was selected to obtain the baseline mean and standard deviation of cerebral oxygenation, interleukin-6 (IL-6) baseline mean and standard deviation, C-reactive protein (CRP) baseline mean and standard deviation, heart rate baseline standard deviation, respiratory rate baseline standard deviation, and systolic blood pressure baseline standard deviation, thus constructing a baseline parameter set. N is preferably a recent sample of 30 to 50 windows, covering the reference state during the preoperative stable period or early anesthesia, ensuring the representativeness of the baseline parameters and reducing the impact of individual abnormal windows. A perioperative event monitoring database was established. The raw and preprocessed perioperative monitoring data, along with the baseline parameter set, were written into the database in a time series structure, supporting real-time read and write operations to ensure the continuity of the monitoring process and meet the real-time response requirements of clinical applications.

[0030] In this implementation plan, perioperative monitoring data is collected in real time, filtered and denoised, time-aligned, linearly interpolated and normalized. The mean and standard deviation are extracted based on a sliding window, and a baseline parameter set for each monitoring indicator is constructed using the historical window median. This allows for the acquisition of stable and reliable physiological baseline characteristics even under conditions of high noise, inconsistent sampling, and discrete inflammatory indicators. At the same time, the raw data, preprocessing results and baseline parameters are written into the event monitoring database to achieve continuous data management and real-time access. This provides a consistent and reliable foundation for subsequent weak sign identification and risk assessment, significantly improving the robustness and feasibility of perioperative monitoring.

[0031] Optionally, based on perioperative monitoring data, identifying consecutive segments of decreased cerebral oxygen saturation below the baseline characteristic, and jointly assessing the magnitude and persistence of the decrease to obtain a characterization of the weak cerebral oxygen trend, involves: reading the baseline mean and standard deviation of cerebral oxygen saturation, where the baseline mean and standard deviation are derived from sliding window statistics, reflecting the individualized reference level of the patient during the stable perioperative phase; scanning the cerebral oxygen saturation sequence within N historical sliding time windows point by point in chronological order, and marking the start point of a decreasing segment when cerebral oxygen saturation changes from not lower than the baseline mean; when cerebral oxygen saturation... When the cerebral oxygen saturation recovers to a level not lower than the baseline mean, it is marked as the termination point of the descent segment. The identification of descent segments is based on single-point threshold comparison, ensuring the implementation method is simple and adaptable to real-time computing scenarios. The number of consecutive sampling points within each descent segment is recorded as the number of continuous sampling points of the descent segment. The median of the number of continuous sampling points of all descent segments is taken to obtain the baseline descent median. The median can avoid the bias caused by extreme descent segments, making the subsequent descent trend assessment statistically robust. The preprocessed cerebral oxygen saturation sequence is read in real time from the perioperative event monitoring database in chronological order. The database is organized using a time-series structure to ensure... Data reading at any given time can be completed efficiently. For each time point, the brain oxygen saturation sequence is traced back to determine the number of sampling points where the brain oxygen saturation is continuously less than the baseline mean, thus obtaining the current duration of continuous decline. This duration can be updated in real time to dynamically reflect the changing trend of the patient's oxygen supply. Within the interval of continuous decline, the difference between the baseline mean brain oxygen saturation and the corresponding brain oxygen saturation is calculated and non-negatively processed. The non-negative differences are then summed to obtain the cumulative decline depth value. The non-negative processing involves comparing the calculated result with zero and taking the larger value to ensure that only the true decline is counted, avoiding the offsetting of values ​​during the recovery phase. The degree of decline is assessed by adding the baseline standard deviation of brain oxygen to a minimum constant value and multiplying this by the median decline to obtain a normalized scale value for brain oxygen decline. The product of the median decline and the baseline standard deviation of brain oxygen is chosen as the normalization scale because the median decline reflects the typical duration of brain oxygen decline during a relatively stable phase, while the baseline standard deviation reflects the normal fluctuation range of the indicator. Multiplying these two values ​​creates an individualized physiological decline reference, used to match the depth of decline with the patient's baseline level. This enhances sensitivity to subtle but persistent downward trends in assessment and avoids bias caused by short-term fluctuations. The cumulative decline value is divided by the normalized scale value of brain oxygen decline, added to a constant, and the natural logarithm is taken to obtain the assessment value for a slight decline in brain oxygen. This logarithmic form enhances the amplification of subtle downward trends, sensitively reflecting even mild but persistent declines, which is suitable for early perioperative risk identification.

[0032] The specific formula for assessing a slight decrease in brain oxygenation is as follows: ; In the formula, The brain oxygen micro-decline assessment value is used to uniformly quantify the magnitude and persistence of the decrease in brain oxygen saturation relative to the baseline, resulting in a stable and sensitive brain oxygen micro-decline assessment value, which is used to characterize early, mild and persistent trends of abnormal brain oxygen. The median of the baseline decline is used as a reference persistence indicator to normalize the cumulative decline depth, aligning the actual decline persistence with historical typical decline patterns and enhancing the robustness of the assessment results. It represents the baseline standard deviation of brain oxygen, which is used to characterize the baseline fluctuation range of brain oxygen saturation, provides a scale for normalizing the depth of decline, and makes data from different patients or different stages comparable. It indicates the duration of the current continuous decline, serving as a time scale for measuring the persistence of the decline, enabling the assessment results to simultaneously reflect how much the decline has occurred and how long it has lasted, thus achieving a joint representation of magnitude and persistence; It represents the baseline mean of brain oxygenation, which is used to provide a reference level for brain oxygen saturation and serve as a comparative baseline for judging whether it is below the baseline, so that the decline has a unified reference framework. It represents brain oxygen saturation, provides raw monitoring data at continuous time points, is used to calculate the difference between each time point and the baseline level, and is the basic input for constructing the descent depth sequence; This represents a very small constant value, used to avoid numerical instability caused by an excessively small denominator. The preferred value is 0.01.

[0033] This implementation scheme, by constructing an individualized cerebral oxygenation reference baseline, identifying continuous decline segments, and normalizing and quantifying them by combining the depth and persistence of the decline, can sensitively capture subtle but continuously developing trends of unstable cerebral oxygenation under real-time monitoring conditions. Simultaneously, steps such as non-negative accumulation, scale normalization, and logarithmic amplification enhance the ability to identify early abnormalities, ensuring that subtle declines are fully reflected in trend assessment. The method has a clear overall calculation process, traceable parameter sources, and good real-time performance, stability, and clinical applicability, effectively supporting continuous monitoring of perioperative weak cerebral oxygenation trends and identification of precursor risks.

[0034] Optionally, the specific process for generating the brain oxygen weakness trend feature vector is as follows: Brain oxygen micro-decline assessment values ​​are calculated in real time, and a brain oxygen weakness trend sequence is constructed based on the brain oxygen micro-decline assessment values ​​at consecutive time points. Abnormal jumps and short-term missing brain oxygen micro-decline assessment values ​​are linearly interpolated. Abnormal jumps are identified by comparing whether the sudden increase in adjacent sampling points exceeds a change threshold, and the mean of adjacent points is preferably used for smoothing. Short-term missing data is linearly interpolated based on the effective sampling points before and after the missing point to ensure the continuity and computability of the trend sequence. A brain oxygen feature vector is generated based on the current brain oxygen micro-decline assessment value, the magnitude of change in the brain oxygen micro-decline assessment value, and the current duration of continuous decline. The magnitude of change in the brain oxygen micro-decline assessment value is obtained by the difference between the current value and the previous value, and the duration of continuous decline is calculated by converting the number of sampling points that continuously meet the brain oxygen baseline mean, so that the brain oxygen feature vector simultaneously includes key parameters such as instantaneous state, short-term rate of change, and trend persistence. The brain oxygen feature vector is written into the perioperative event monitoring database and timestamped. The timestamp alignment is preferably based on a unified system clock. The brain oxygen feature vector is written into the corresponding time-series index position in the database according to the sampling frequency to ensure that the timestamps are consistent with those of other monitoring indicators, thereby supporting the synchronous operation of subsequent multimodal weak symptom fusion calculation and risk assessment processes.

[0035] In this implementation plan, by continuously calculating the assessment value of slight decrease in brain oxygenation, compensating for abnormal jumps and short-term missing values, and constructing a brain oxygenation feature vector by combining the current assessment value, the magnitude of change, and the duration of continuous decrease, a stable characterization of the weak trend of brain oxygenation is achieved. This not only maintains the integrity and continuity of time-series data and avoids misjudgment caused by signal noise or missing measurements, but also reflects the instantaneous state, rate of change, and trend persistence at the same time. This makes the expression of signs of weak brain oxygenation more accurate, robust, and usable for subsequent risk fusion judgment, thereby significantly improving the early identification ability of weak abnormal changes in the perioperative period.

[0036] Optionally, based on perioperative monitoring data and baseline characteristics, the specific process of analyzing the relative changes in serum inflammatory markers, extracting the drift intensity reflecting the slow accumulation of inflammation, forming a weak trend characterization of inflammation, and generating an inflammatory feature vector is as follows: Real-time reading of interleukin-6 (IL-6) and C-reactive protein (CRP) concentrations, and obtaining a baseline parameter set; subtracting the baseline mean of IL-6 from the current IL-6 concentration and dividing by the baseline standard deviation of IL-6 to obtain the normalized increase of IL-6, and then non-negating the normalized increase of IL-6; subtracting the baseline mean of CRP from the current CRP concentration and dividing by the baseline standard deviation of CRP to obtain the normalized increase of CRP, and then non-negating the normalized increase of CRP; the non-negation process ensures that the indicators only reflect the upward trend, avoiding interference from downward noise in trend judgment. The sum of squares of the normalized increase in non-negative interleukin-6 and C-reactive protein was calculated, and the square root was taken to obtain the inflammatory drift intensity value. This can simultaneously reflect the comprehensive deviation of the two types of inflammatory markers, avoiding misjudgments caused by fluctuations in a single indicator. The inflammatory drift intensity value was calculated in real time, and an inflammatory drift intensity trend sequence was constructed. Linear imputation was performed on inflammatory drift intensity values ​​with abnormal jumps and short-term missing values. Abnormal jumps can be identified by changing thresholds, thereby ensuring the stability and reliability of the compensation logic. The imputed sequence maintains temporal continuity, meeting the requirements for subsequent trend duration determination. The duration of inflammatory elevation was determined based on the continuous increase of the inflammatory drift intensity value compared to the previous moment. The duration of inflammatory elevation is used to quantify the slow accumulation process and can identify non-explosive but persistently increasing inflammatory trends. An inflammatory feature vector is constructed based on the current inflammatory drift intensity value, the magnitude of change in the inflammatory drift intensity value, and the duration of inflammatory elevation. The inflammatory feature vector is written into the perioperative event monitoring database and timestamped to ensure that the inflammatory features are synchronously integrated with the weak trends of brain oxygenation and vital signs on the same time axis, thereby improving the reliability of subsequent multimodal weak sign judgment.

[0037] The specific formula for the inflammation drift intensity value is as follows: ; In the formula, The value representing the intensity of inflammatory drift is a normalized integration of the upward deviation of interleukin-6 and C-reactive protein relative to their respective baseline levels, forming a quantifiable slow cumulative intensity of inflammation, used to characterize early, mild and persistent weak trends of inflammation. It represents the concentration of interleukin-6, indicating the immediate level of the body's inflammatory response at any given moment, and is a sensitive indicator of rapid changes in inflammation; It represents the baseline mean value of interleukin-6, which represents the patient's own IL-6 homeostatic reference level during the perioperative period and is a mean marker of the normal state; It represents the baseline standard deviation of interleukin-6, indicating the normal fluctuation range of IL-6, and serves as a reference scale; It represents the concentration of C-reactive protein, a slow-response indicator that represents the level of accumulated inflammation in the body; This represents the baseline mean C-reactive protein level, defines the normal level of C-reactive protein concentration for an individual patient, and is used to determine whether the current C-reactive protein concentration deviates from the steady state. This represents the baseline standard deviation of C-reactive protein, indicating the normal range of fluctuations in a patient's C-reactive protein concentration.

[0038] In this implementation scheme, the relative changes of interleukin-6 and C-reactive protein are normalized and non-negatively processed. The inflammatory drift intensity is constructed by combining the degree of joint deviation between the two. Furthermore, the duration of continuous rise and the magnitude of change are introduced to form an inflammatory feature vector, which can stably identify slowly accumulating weak inflammatory trends from discrete inflammatory indicators. At the same time, through abnormal jump correction, missing value imputation and timestamp alignment, the calculation process of inflammatory features is ensured to be continuous, robust and can be synchronously integrated with other weak physiological signs, thereby significantly improving the sensitivity and traceability of early risk identification and enhancing the overall monitoring system's ability to perceive the precursor state of potential adverse events.

[0039] Optionally, by combining the temporal fluctuation characteristics of vital sign data in perioperative monitoring data, the degree of deviation relative to the baseline is assessed, and the specific process for representing vital sign fluctuations is as follows: Calculate the standard deviation of heart rate, respiratory rate, and systolic blood pressure within the current sliding time window, and obtain the baseline standard deviations of heart rate, respiratory rate, and systolic blood pressure; subtract the baseline standard deviation of heart rate from the current standard deviation of heart rate, and divide by the sum of the baseline standard deviation of heart rate and the minimum constant value to obtain the normalized fluctuation of heart rate; subtract the baseline standard deviation of respiratory rate from the current standard deviation of respiratory rate. The difference is calculated and divided by the sum of the baseline standard deviation of respiratory rate and the minimum constant value to obtain the normalized fluctuation of respiratory rate. The standard deviation of current systolic blood pressure is subtracted from the baseline standard deviation of systolic blood pressure and divided by the sum of the baseline standard deviation of systolic blood pressure and the minimum constant value to obtain the normalized fluctuation of systolic blood pressure. The normalized fluctuations of heart rate, respiratory rate, and systolic blood pressure are non-negatively processed. The non-negativity processing adopts the method of taking the maximum value after comparison with zero to ensure that only the deviation relative to the baseline is retained and the fluctuations below the baseline are excluded, so as to enhance the sensitivity to abnormal fluctuation trends. The sum of squares of the non-negative normalized fluctuations of heart rate, respiratory rate, and systolic blood pressure are calculated and the square root is taken to obtain the vital sign fluctuation value. The square root operation of the sum of squares can integrate the relative deviation of the three fluctuations of heart rate, respiration, and blood pressure to form a time-domain fluctuation characterization value with a unified scale, which is used as the risk assessment input for the subsequent weak sign fusion module.

[0040] In this implementation scheme, the fluctuation characteristics of heart rate, respiratory rate, and systolic blood pressure within a window are uniformly normalized, non-negatively normalized, and comprehensively quantified to form a single-scale vital sign fluctuation characterization, which enables sensitive capture of subtle abnormal fluctuations in multiple vital signs. This method can maintain computational stability under different monitoring dimensions, avoid numerical instability caused by an excessively small denominator, and improve the overall identification ability of abnormal physiological fluctuations through multi-indicator fusion, thereby providing more accurate, continuous, and interpretable input features for subsequent delirium precursor risk assessment.

[0041] Optionally, the specific process for determining the prodromal risk of delirium by integrating the weak trend representation of cerebral oxygenation, the weak trend representation of inflammation, and the fluctuation representation of vital signs is as follows: Read the assessment value of the slight decrease in cerebral oxygenation, the intensity value of inflammation drift, and the fluctuation value of vital signs at the corresponding time point. Add these values ​​together and calculate the average to obtain a comprehensive weak symptom intensity value. This average value is used to map weak symptom representations from different sources to a unified scale, making cross-modal minor abnormalities comparable and avoiding bias caused by a single indicator mutation. The negative of the comprehensive weak symptom intensity value is taken as the exponent for natural exponent calculation. Natural exponent calculation is used to transform the linear change in weak symptom intensity into a non-linear growth relationship of risk, making the change in weak symptom intensity gradual at low levels and more sensitive at high levels, in line with the clinical characteristic of progressively accelerating adverse event risk. The result of the natural exponent calculation is subtracted from the constant to obtain the prodromal risk value of delirium, strictly limiting the prodromal risk value to between 0 and 1, forming a standardized risk probability representation, which is convenient for risk threshold comparison and early warning triggering.

[0042] The specific formula for the risk value of delirium precursors is as follows: ; In the formula, It represents the risk value of delirium precursors, which comprehensively reflects the overall risk level of three weak signs: decreased brain oxygenation, accumulated inflammation, and fluctuations in vital signs. It represents the assessment value of slight decrease in cerebral oxygen, which characterizes the combined intensity of the amplitude and persistence of the decrease in cerebral oxygen saturation relative to the baseline value, and is used to reflect potential oxygen insufficiency or a downward trend in perfusion. It represents the intensity of inflammatory drift, characterizing the slow, cumulative upward trend of interleukin-6 and C-reactive protein relative to the baseline, and is used to reflect the weak trend changes of subclinical inflammation or infection. It represents the standard deviation of heart rate, characterizing the temporal fluctuation of the current heart rate, and is used to detect abnormal heart rate variability and enhanced stress response; It represents the standard deviation of the heart rate baseline, characterizing the normal heart rate fluctuation range extracted from the historical window, and serving as a reference benchmark for judging whether the current fluctuation is abnormal; It represents the standard deviation of respiratory rate, characterizing the degree of fluctuation in the current respiratory rate, and is used to identify respiratory instability or compensatory changes; It represents the baseline standard deviation of respiratory rate, characterizes the range of normal respiratory variation in the historical window, and is used to compare whether the current fluctuation has increased abnormally. It represents the standard deviation of systolic blood pressure, characterizing the degree of fluctuation in current systolic blood pressure, and is used to reflect blood pressure stability or potential circulatory instability trends; It represents the baseline standard deviation of systolic blood pressure, characterizing the normal blood pressure fluctuation range within the historical window, and serving as a reference for deviation evaluation; This represents a very small constant value, used to avoid numerical instability caused by an excessively small denominator. The preferred value is 0.01.

[0043] In this embodiment, Table 1 is a data table of risk values ​​for the precursor of delirium. The baseline standard deviation of heart rate was set to 5, the baseline standard deviation of respiratory rate to 2, and the baseline standard deviation of systolic blood pressure to 8. The table details the assessment values ​​of slight decrease in cerebral oxygenation, the intensity of inflammatory drift, the standard deviation of heart rate, the standard deviation of respiratory rate, the standard deviation of systolic blood pressure, the fluctuation values ​​of vital signs, the intensity of comprehensive weak signs, and the risk values ​​for the precursor of delirium at five time points. Specifically, at time t1, the assessment value of slight decrease in cerebral oxygenation was 0.10, the intensity of inflammatory drift was 0.10, the standard deviation of heart rate was 5.5, and the standard deviation of respiratory rate was 2.2. The standard deviation of systolic blood pressure was 8.5, the variability of vital signs was 0.1541, the intensity of the overall weak signs was 0.1180, and the risk of delirium was 0.1113; the assessment of slight decrease in brain oxygenation at time t2 was 0.50, the intensity of inflammatory drift was 0.40, the standard deviation of heart rate was 6.0, the standard deviation of respiratory rate was 2.5, the standard deviation of systolic blood pressure was 9.0, the variability of vital signs was 0.3425, the intensity of the overall weak signs was 0.4142, and the risk of delirium was 0.3391. At time t3, the assessment value for slight decrease in cerebral oxygen saturation was 1.20, the intensity of inflammatory drift was 1.00, the standard deviation of heart rate was 8.0, the standard deviation of respiratory rate was 3.5, the standard deviation of systolic blood pressure was 12.0, the variability of vital signs was 1.0793, the intensity of overall weak signs was 1.0931, and the risk of delirium was 0.6648. At time t4, the assessment value for slight decrease in cerebral oxygen saturation was 1.50, the intensity of inflammatory drift was 1.40, the standard deviation of heart rate was 9.0, the standard deviation of respiratory rate was 4.0, and the systolic blood pressure was 12.0. The standard deviation of systolic blood pressure was 14.0, the variability of vital signs was 1.4794, the intensity of the overall weak signs was 1.4598, and the risk of delirium was 0.7677. At time t5, the assessment of slight decrease in brain oxygen was 0.80, the intensity of inflammatory drift was 1.00, the standard deviation of heart rate was 7.0, the standard deviation of respiratory rate was 3.0, the standard deviation of systolic blood pressure was 11.0, the variability of vital signs was 0.7397, the intensity of the overall weak signs was 0.8466, and the risk of delirium was 0.5711.

[0044] Table 1. Risk Values ​​of Precursors to Delirium

[0045] like Figure 3 The figure shows the trend of the risk value of delirium precursors. The figure illustrates the time-series change trend of the risk value of delirium precursors. The horizontal axis represents the time number, and the vertical axis represents the risk value of delirium precursors, with risk thresholds marked to visually distinguish possible risk stages. (This is in conjunction with Table 1 and...) Figure 3 It can be seen that the patient's risk level gradually increased over time. Both t1 and t2 were below the risk threshold, indicating a relatively safe overall condition. From t3 onwards, the risk value first exceeded the threshold, suggesting a significant increase in the three types of weak signs: decreased cerebral oxygenation, accumulated inflammation, and fluctuations in vital signs, indicating the entry into a potential risk stage. The risk reached its highest point at t4, indicating that the multi-source weak signs were most concentrated and had the strongest cumulative effect during this period, making it a high-risk interval requiring close monitoring. Although t5 was slightly lower than the peak, it remained above the risk threshold, indicating that the weak signs had not completely subsided and continued monitoring was necessary. Overall, the risk change trend was consistent with the intensity value of the comprehensive weak signs, enabling the identification of a continuous evolution from a safe period, an upward trend, to a high-risk period, providing a reliable basis for the prospective identification of perioperative risks.

[0046] like Figure 4 The image shows a radar chart of three indicators of weak signs. The radar chart uses three indicators of weak signs—brain oxygen saturation assessment, inflammatory drift intensity, and vital sign fluctuations—to illustrate the overall change pattern of weak signs at five time points from t1 to t5. The larger the outline of the line connecting each time point in the chart, the stronger the weak sign and the higher the potential risk. (This is in conjunction with Table 1 and...) Figure 4 It can be seen that the overall weak symptoms were mild in stages t1 and t2, with all three indicators at low levels and small radar map areas, indicating that the patient's condition was stable in the early stages and no significant risk accumulation had occurred. From t3 onwards, the weak symptoms significantly intensified, with all three indicators rising simultaneously and the radar map area expanding significantly, suggesting a synergistic increase in cerebral oxygenation, inflammation accumulation, and vital sign fluctuations, marking a key turning point of rising risk. T4 reached the peak of the three weak symptoms, with the largest radar map outline, indicating that all three types of symptoms were at their strongest, reflecting the most concentrated and significant weak symptoms at this time, representing the highest risk, corresponding to the peak position in the delirium prodromal risk map. T5 showed slight relief but remained higher than the early levels, with all three indicators decreasing compared to t4, but still significantly higher than t1 and t2, indicating that the weak symptoms had not fully recovered and continued monitoring was necessary.

[0047] In this implementation plan, cross-modal fusion of weak cerebral oxygenation trends, weak inflammatory trends, and vital sign fluctuations is performed, and the average value is used to uniformly quantify multi-source mild abnormalities. A nonlinear risk growth model is then constructed through exponential mapping, making the risk assessment more sensitive at high-level stages of mild symptoms. Simultaneously, the risk value is limited to between 0 and 1, forming a standardized, threshold-comparable risk representation. This method not only enhances the comparability of weak symptoms across indicators and avoids bias caused by single abnormalities, but also makes the changing pattern of delirium precursor risk more closely match the characteristics of progressively accumulating clinical risk, thereby improving early identification capabilities and warning accuracy.

[0048] Optionally, the specific process of outputting risk warning information based on the delirium precursor risk is as follows: The delirium precursor risk value is calculated in real time and written into the perioperative event monitoring database. Short-term missing values ​​are linearly interpolated. Linear interpolation is used to reconstruct a continuous trajectory when there are gaps of several seconds caused by momentary packet loss or slight detachment of the pulse oxygen probe, avoiding non-physiological breaks in the risk curve. Intra-window trend regression is used to robustly smooth sudden changes to suppress interference from non-pathological changes such as instrument jitter and motion artifacts on risk assessment, obtaining the delirium risk change trajectory. The delirium precursor risk value is compared with a risk threshold in real time. The risk threshold can be adaptively determined by historical sample statistics. If the time for which consecutive delirium precursor risk values ​​are higher than the risk threshold exceeds the allowable time threshold, the allowable time threshold is used to avoid false alarms triggered by single-point fluctuations. Preferably, the allowable time threshold is used. A sustained period exceeding the threshold of 30 to 60 seconds triggers a delirium precursor risk warning and generates warning information, including the delirium precursor risk value, the magnitude of change in the delirium precursor risk value, the duration of the risk, and assessment values ​​for slight decreases in brain oxygenation, inflammatory drift intensity, and vital sign fluctuations. Simultaneously, based on the relative magnitudes of the assessment values ​​for slight decreases in brain oxygenation, inflammatory drift intensity, and vital sign fluctuations within the overall weak symptom intensity value, the primary physiological source leading to the increased risk is inferred. By calculating the proportion of the three types of weak symptom, it can be determined whether the increased risk is primarily due to insufficient oxygen supply, accumulated inflammation, or abnormal vital sign fluctuations. The modality with the largest proportion among the three is marked as the primary contributing source. If the ratio of the second largest to the largest proportion is greater than 0.8, both are marked as common primary contributing sources. A corresponding risk warning is generated, enabling early identification of delirium precursor states.

[0049] In this implementation plan, by continuously updating, interpolating, and smoothing the risk value of delirium precursors, the stability and continuity of the risk curve can be maintained under conditions of noise, packet loss, or transient artifacts. By setting an adaptive risk threshold and a tolerance time mechanism for continuous over-threshold, false alarms caused by short-term fluctuations can be effectively avoided, improving the reliability of the warning. At the same time, by integrating the relative contributions of three weak signs—brain oxygenation, inflammation, and vital signs—the dominant physiological source of the risk increase can be accurately located, making the risk indication no longer a single value, but a guiding information with traceability. Thus, early, stable, and interpretable intelligent warnings of delirium precursor states are achieved, significantly improving the foresight and clinical applicability of perioperative risk identification.

[0050] Optionally, the temporal evolution of weak brain oxygenation trend, weak inflammation trend, vital sign fluctuations, and delirium precursor risk is visualized, and risk trigger information and main contributing sources are linked and labeled. The specific process for recording feedback on risk changes is as follows: Based on continuously calculated assessment values ​​of slight decrease in brain oxygenation, inflammation drift intensity, vital sign fluctuations, and delirium precursor risk, the above-mentioned weak signs and risk indicators are uniformly aligned with timestamps as the horizontal axis, so that all curves are rendered and displayed on the same time reference, thereby ensuring the comparability and continuity of indicator changes; the weak sign indicators and their corresponding timestamps are visualized and rendered, displaying the weak brain oxygenation trend curve, inflammation drift trend curve, vital sign fluctuation trend curve, and delirium precursor risk change curve, making the subtle changes in different physiological dimensions intuitively identifiable. The system marks risk thresholds, risk trigger times, and risk durations, and displays them in conjunction with perioperative monitoring data through a shared time index, synchronously showing the original monitoring data at the corresponding time points. This makes the correlation between risk increases and changes in physiological indicators clear and observable. Based on risk warning information, it displays warnings of increased proportions of brain oxygen-related fluctuations, increased proportions of inflammation-related parameter fluctuations, and increased proportions of vital sign fluctuations in a hierarchical manner, making the sources of contribution visually reviewable. Simultaneously, it generates a delirium precursor risk warning report, which includes the trend of delirium precursor risk value changes, the start and end time periods of risk triggering, the proportion ranking of contributing indicators, the corresponding weak symptom trends, and the threshold version. This delirium precursor risk warning report is written into the perioperative event monitoring database to support subsequent quality tracking, postoperative management, and clinical review.

[0051] This implementation plan presents a unified visualization of the temporal evolution of weak cerebral oxygenation trends, weak inflammatory trends, fluctuations in vital signs, and the risk of delirium precursors, and links this visualization with the original monitoring data. This allows for the intuitive identification of subtle changes in multi-source physiological signals, risk triggering processes, and major contributing sources. Simultaneously, through hierarchical labeling and early warning report output, the risk formation mechanism is more easily traced, and the basis for clinical intervention is clearer. This improves the efficiency and interpretability of identifying early abnormal trends during and after surgery, and enhances the continuity and operability of perioperative risk management.

[0052] On the other hand, embodiments of the present invention also provide a real-time monitoring system for perioperative adverse events in elderly patients undergoing hip fracture surgery, applied to the aforementioned method for real-time monitoring of perioperative adverse events in elderly patients undergoing hip fracture surgery. This system includes: a monitoring data acquisition and processing module, used to acquire perioperative monitoring data in real time, perform data preprocessing on the perioperative monitoring data, and extract baseline features of each perioperative monitoring data based on a sliding time window; a brain oxygenation weakness trend extraction module, used to identify continuous segments of brain oxygen saturation below the baseline feature based on the perioperative monitoring data, jointly evaluate the magnitude and persistence of the decline, obtain a brain oxygenation weakness trend characterization, and generate a brain oxygenation weakness trend feature vector; and an inflammation weakness trend extraction module, used to analyze serum... The system analyzes the relative changes of inflammatory markers and extracts the drift intensity reflecting the slow accumulation of inflammation to form a weak trend representation of inflammation, generating an inflammatory feature vector. A multimodal weak sign fusion module combines the temporal fluctuation characteristics of vital sign data from perioperative monitoring data to assess the degree of deviation from the baseline, obtaining a vital sign fluctuation representation. This module integrates the weak trend representation of cerebral oxygenation, the weak trend representation of inflammation, and the vital sign fluctuation representation to determine the risk of delirium precursors and output risk warning information based on the risk of delirium precursors. A visualization monitoring feedback module visualizes the temporal evolution of the weak trend representation of cerebral oxygenation, the weak trend representation of inflammation, the vital sign fluctuation representation, and the risk of delirium precursors, and links and labels risk trigger information and major contributing sources to achieve feedback recording of risk changes.

[0053] This implementation plan achieves real-time acquisition, weak trend extraction, cross-modal fusion, and risk assessment of perioperative multi-source monitoring data through modular division of labor, and presents the risk evolution trajectory and contribution sources in a visual manner. With a clear structure and complete signal processing link, it can realize the early identification of minor physiological abnormalities and transparent tracking of risk triggering processes, significantly enhancing the early perception of adverse events and improving the real-time performance, accuracy, and interpretability of perioperative monitoring, thus demonstrating good clinical applicability.

[0054] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0055] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0056] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0057] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0058] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0059] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0060] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0061] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0062] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0063] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0064] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for real-time monitoring of adverse events during the perioperative period of hip fracture surgery in elderly patients, characterized in that, The method includes: S1, real-time acquisition of perioperative monitoring data, data preprocessing of perioperative monitoring data, and extraction of baseline features of each perioperative monitoring data based on a sliding time window; S2, based on perioperative monitoring data, identifies continuous segments of brain oxygen saturation that are lower than the baseline characteristics, jointly assesses the magnitude and duration of the decline, obtains a brain oxygen weakness trend characterization, and generates a brain oxygen weakness trend feature vector. S3, based on perioperative monitoring data and baseline characteristics, analyzes the relative changes of serum inflammatory markers, extracts the drift intensity reflecting the slow accumulation of inflammation, forms a weak trend characterization of inflammation, and generates an inflammatory feature vector; S4. By combining the temporal fluctuation characteristics of vital signs data in perioperative monitoring data, the degree of deviation from the baseline is assessed to obtain the vital signs fluctuation characterization. The characterization of weak brain oxygenation trend, weak inflammation trend, and vital signs fluctuation characterization are integrated to determine the risk of delirium prodrome and output risk warning information based on the risk of delirium prodrome. S5 visualizes the temporal evolution of weak brain oxygenation trend, weak inflammation trend, vital sign fluctuations, and delirium precursor risk, and links risk trigger information and main contributors to achieve feedback recording of risk changes.

2. The method for real-time monitoring of perioperative adverse events in elderly patients undergoing hip fracture surgery according to claim 1, characterized in that, The specific process of real-time acquisition of perioperative monitoring data, data preprocessing of perioperative monitoring data, and extraction of baseline features of each perioperative monitoring data based on a sliding time window is as follows: Real-time collection of perioperative monitoring data of patients, including: cerebral oxygen saturation, heart rate, respiratory rate, systolic blood pressure, interleukin-6 concentration and C-reactive protein concentration; Median filtering and moving average were applied to continuous signals of cerebral oxygen saturation, heart rate, respiratory rate, and systolic blood pressure to remove spike noise and spurious signals from probe loosening. Perioperative monitoring data from different sampling frequencies were time-stamped and aligned. Continuous time series of discrete serum inflammatory markers, such as interleukin-6 concentration and C-reactive protein concentration, were constructed using linear interpolation. Perioperative monitoring data were then normalized. Based on a sliding time window, the mean and standard deviation of cerebral oxygen saturation, interleukin-6 concentration, and C-reactive protein concentration were calculated, as well as the standard deviations of heart rate, respiratory rate, and systolic blood pressure. The median of the mean and standard deviation of perioperative monitoring data within N historical sliding time windows was selected to obtain the baseline mean and standard deviation of cerebral oxygen saturation, interleukin-6 concentration, C-reactive protein concentration, heart rate, respiratory rate, and systolic blood pressure, thus constructing a baseline parameter set. A perioperative event monitoring database was established, and the raw and preprocessed perioperative monitoring data, along with the baseline parameter set, were written into the database.

3. The method for real-time monitoring of perioperative adverse events in elderly patients undergoing hip fracture surgery according to claim 1, characterized in that, The specific process of identifying continuous segments of decreased cerebral oxygen saturation below the baseline characteristic based on perioperative monitoring data, and jointly assessing the magnitude and persistence of the decrease to obtain a characterization of weak cerebral oxygenation trend is as follows: Read the baseline mean and baseline standard deviation of brain oxygen, and scan the brain oxygen saturation sequence within N historical sliding time windows point by point in chronological order. When the brain oxygen saturation changes from not lower than the baseline mean to lower than the baseline mean, it is marked as the starting point of a decreasing segment. When brain oxygen saturation recovers to a level not lower than the average brain oxygen baseline, it is marked as the termination point of the descending segment, and the number of consecutive sampling points within each descending segment is recorded as the number of continuous sampling points of the descending segment. The median of the baseline descent is obtained by taking the median number of continuous sampling points for all descent segments; The preprocessed brain oxygen saturation sequence is read in real time from the perioperative event monitoring database in chronological order. For each moment, the brain oxygen saturation sequence is traced back to determine the number of sampling points where the brain oxygen saturation is continuously less than the average brain oxygen baseline, and the current duration of continuous decline is obtained. Within the interval of continuous descent, the difference between the mean baseline brain oxygen value and the corresponding brain oxygen saturation is calculated and non-negatively processed. The non-negative differences are then summed to obtain the cumulative descent depth value. Add the baseline standard deviation of brain oxygen to the minimum constant value, and multiply it by the median of the baseline decrease to obtain the normalized scale value of brain oxygen decrease. Divide the cumulative value of the decrease depth by the normalized scale value of brain oxygen decrease, add it to the constant, and take the natural logarithm to obtain the assessment value of the micro decrease in brain oxygen.

4. The method for real-time monitoring of perioperative adverse events in elderly patients undergoing hip fracture surgery according to claim 1, characterized in that, The specific process for generating the brain oxygen weakness trend feature vector is as follows: The brain oxygen micro-decline assessment value is calculated in real time, and a weak trend sequence of brain oxygen is constructed based on the brain oxygen micro-decline assessment values ​​at continuous time. Linear interpolation is performed on the brain oxygen micro-decline assessment values ​​with abnormal jumps and short-term missing values. Based on the brain oxygen micro-decline assessment value at the current time, the change range of the brain oxygen micro-decline assessment value, and the current continuous decline duration, a brain oxygen feature vector is generated. Brain oxygenation feature vectors were written into the perioperative event monitoring database and timestamped.

5. The method for real-time monitoring of perioperative adverse events in elderly patients undergoing hip fracture surgery according to claim 1, characterized in that, The specific process of analyzing the relative changes in serum inflammatory markers based on perioperative monitoring data and baseline characteristics, extracting the drift intensity reflecting the slow accumulation of inflammation to form a weak trend characterization of inflammation, and generating an inflammatory feature vector is as follows: The system reads the concentrations of interleukin-6 (IL-6) and C-reactive protein (CRP) in real time and obtains a baseline parameter set. The normalized increase in IL-6 is obtained by subtracting the baseline mean of IL-6 from the current IL-6 concentration and then dividing by the baseline standard deviation of IL-6. The normalized increase in CRP is then non-negatively processed. Similarly, the normalized increase in CRP is obtained by subtracting the baseline mean of CRP from the current IL-6 concentration and then dividing by the baseline standard deviation of CRP. The normalized increase in CRP is then non-negatively processed. The sum of squares of the normalized increase in non-negative interleukin-6 and the normalized increase in C-reactive protein was calculated, and the square root was taken to obtain the inflammatory drift intensity value. The system calculates the inflammatory drift intensity value in real time, constructs an inflammatory drift intensity trend sequence, and performs linear interpolation on inflammatory drift intensity values ​​that are abnormally abrupt or short-term missing. It also determines the duration of inflammation elevation based on the continuous time when the inflammatory drift intensity value increases compared to the previous time. Based on the current inflammatory drift intensity value, the magnitude of the change in the inflammatory drift intensity value, and the duration of inflammation elevation, an inflammatory feature vector is constructed. The inflammatory feature vector is written into the perioperative event monitoring database and timestamp aligned.

6. The method for real-time monitoring of perioperative adverse events in elderly patients undergoing hip fracture surgery according to claim 1, characterized in that, The specific process of assessing the degree of deviation from the baseline by combining the temporal fluctuation characteristics of vital sign data in perioperative monitoring data to obtain the vital sign fluctuation characterization is as follows: Calculate the standard deviations of heart rate, respiratory rate, and systolic blood pressure within the current sliding time window, and obtain the baseline standard deviations of heart rate, respiratory rate, and systolic blood pressure. Subtract the baseline standard deviation of heart rate from the current standard deviation of heart rate and divide by the sum of the baseline standard deviation of heart rate and the minimum constant value to obtain the normalized fluctuation of heart rate. Subtract the baseline standard deviation of respiratory rate from the current standard deviation of respiratory rate and divide by the sum of the baseline standard deviation of respiratory rate and the minimum constant value to obtain the normalized fluctuation of respiratory rate. Subtract the baseline standard deviation of systolic blood pressure from the current standard deviation of systolic blood pressure and divide by the sum of the baseline standard deviation of systolic blood pressure and the minimum constant value to obtain the normalized fluctuation of systolic blood pressure. Nonnegate the normalized fluctuations of heart rate, respiratory rate, and systolic blood pressure. Calculate the sum of squares of the nonnegated normalized fluctuations of heart rate, respiratory rate, and systolic blood pressure and take the square root to obtain the vital sign fluctuation value.

7. The method for real-time monitoring of perioperative adverse events in elderly patients undergoing hip fracture surgery according to claim 1, characterized in that, The specific process for assessing the risk of delirium precursors by integrating weak brain oxygenation trend characteristics, weak inflammation trend characteristics, and vital sign fluctuation characteristics is as follows: Read the brain oxygen micro-decline assessment value, inflammation drift intensity value, and vital sign fluctuation value at the corresponding time. Add the brain oxygen micro-decline assessment value, inflammation drift intensity value, and vital sign fluctuation value and calculate the average value to obtain the comprehensive weak symptom intensity value. Take the negative number of the comprehensive weak symptom intensity value as the exponent for natural exponent calculation, and subtract the natural exponent calculation result from the constant to obtain the delirium precursor risk value.

8. The method for real-time monitoring of perioperative adverse events in elderly patients undergoing hip fracture surgery according to claim 1, characterized in that, The specific process of outputting risk warning information based on the risk of delirium precursors is as follows: The delirium precursor risk value is calculated in real time and written into the perioperative event monitoring database. Short-term missing values ​​are linearly imputed, and a robust smoothing correction is performed on sudden jumps using in-window trend regression to obtain the delirium risk change trajectory. The delirium precursor risk value is compared with the risk threshold in real time. If the duration of consecutive delirium precursor risk values ​​above the risk threshold exceeds the allowable time threshold, a delirium precursor risk warning is triggered and warning information is generated, including the delirium precursor risk value, the magnitude of change of the delirium precursor risk value, the duration of the risk, the assessment value of micro-decline in brain oxygen, the intensity value of inflammatory drift, and the value of vital sign fluctuations. Simultaneously, based on the relative magnitudes of brain oxygen micro-decline assessment values, inflammation drift intensity values, and vital sign fluctuation values ​​within the overall weak symptom intensity value, the main physiological sources leading to increased risk are inferred, and corresponding risk warnings are generated.

9. The method for real-time monitoring of perioperative adverse events in elderly patients undergoing hip fracture surgery according to claim 1, characterized in that, The process of visualizing the temporal evolution of weak brain oxygenation trends, weak inflammation trends, vital sign fluctuations, and delirium precursor risks, and linking them with risk trigger information and major contributing sources to achieve feedback recording of risk changes, is as follows: Based on the continuously calculated brain oxygen micro-decline assessment value, inflammation drift intensity value, vital sign fluctuation value, and delirium precursor risk value, various weak sign indicators and their corresponding timestamps are visualized and rendered, displaying brain oxygen weak trend curves, inflammation drift trend curves, vital sign fluctuation trend curves, and delirium precursor risk change curves, and marking risk thresholds, risk triggering times, and risk durations, while also linking with perioperative monitoring data for display; Based on risk warning information, the system displays warnings of increased proportions of fluctuations in brain oxygenation, inflammation-related parameters, and vital signs in a hierarchical manner; at the same time, it generates a delirium precursor risk warning report and writes it into the perioperative event monitoring database.

10. A real-time monitoring system for perioperative adverse events in elderly patients undergoing hip fracture surgery, used to implement the method as described in any one of claims 1-9, characterized in that, include: The monitoring data acquisition and processing module is used to collect perioperative monitoring data in real time, perform data preprocessing on the perioperative monitoring data, and extract the baseline features of each perioperative monitoring data based on a sliding time window. The brain oxygen deficiency trend extraction module is used to identify continuous segments of brain oxygen saturation that are lower than the baseline feature based on perioperative monitoring data, jointly evaluate the magnitude and persistence of the decline, obtain a brain oxygen deficiency trend characterization, and generate a brain oxygen deficiency trend feature vector. The weak trend extraction module for inflammation is used to analyze the relative changes of serum inflammatory markers based on perioperative monitoring data and baseline characteristics, extract the drift intensity that reflects the slow accumulation of inflammation, form a weak trend characterization of inflammation, and generate an inflammatory feature vector. The multimodal weak sign fusion module is used to combine the temporal fluctuation characteristics of vital sign data in perioperative monitoring data, assess the degree of deviation from the baseline, obtain vital sign fluctuation characteristics, fuse brain oxygen weak trend characteristics, inflammation weak trend characteristics and vital sign fluctuation characteristics, determine the risk of delirium prodrome, and output risk warning information based on the risk of delirium prodrome. The visualization monitoring and feedback module is used to visualize the temporal evolution of weak brain oxygenation trend, weak inflammation trend, vital sign fluctuations, and delirium precursor risk, and to link and label risk trigger information and main contributors to achieve feedback recording of risk changes.

Images