Anesthesia scheme evaluation system based on big data analysis

Abstract

The invention relates to the technical field of anesthesia management, in particular to an anesthesia scheme evaluation system based on big data analysis, and the system comprises a response section recognition module, a lag offset linkage module, a risk association mapping module, a risk association mapping module and a scheme structure evaluation module. According to the method, by constructing a multi-section anesthesia depth dynamic structure based on time step division and slope analysis and extracting time synchronization changes of medicine injection information and physiological indexes, boundary recognition and rule induction of multi-section response characteristics in an operation are achieved, and the dynamic anesthesia depth is obtained by combining response delay characteristics and starting point difference calculation. A dose response relation label corresponding to index change is obtained, a preoperative visual induction task and a postoperative trajectory offset record are fused, intervention association between a path error area and a medication path is mined, structured mapping and label evaluation between an anesthesia strategy and postoperative behavior performance of a patient are achieved, and the patient postoperative behavior performance is evaluated. And the structure evaluation pertinence and response precision of the anesthesia scheme are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of anesthesia management technology, and in particular to an anesthesia protocol evaluation system based on big data analysis. Background Technology

[0002] The field of anesthesia management technology involves the comprehensive management of anesthetic interventions during surgery or medical procedures. Core aspects include the selection of anesthesia methods, determination of anesthetic drug dosages and types, intraoperative monitoring of anesthesia depth, postoperative recovery assessment and guidance, and complication risk control. Within this field, medical personnel must comprehensively analyze patient history, physiological parameters, and surgical type to develop a scientifically sound anesthesia plan, and conduct real-time adjustments and assessments during its execution. In recent years, with the development of information technology, big data analysis, clinical decision support systems, and intelligent monitoring methods have been gradually applied to anesthesia management processes to assist in anesthesia decision-making and optimize evaluation mechanisms, thereby improving the safety and standardization of the anesthesia process. Traditional anesthesia protocol evaluation systems refer to technical systems that perform preoperative prediction, intraoperative adjustment, and postoperative retrospective analysis of previously implemented anesthesia methods and processes. These systems primarily assist anesthesiologists in judging the rationality and safety of anesthesia strategies. The key technical issue addressed is how to organize and interpret complex and heterogeneous clinical anesthesia information to support protocol evaluation and optimization. Traditional anesthesia protocol evaluation systems are based on structured medical records extracted from clinical databases and intraoperative monitoring records. They use static rules to score indicators and assess risks, including ASA classification based on patient basic information, setting drug dosage ranges and drug combination lists based on past experience, judging the trend of anesthesia depth changes by combining intraoperative vital sign curves such as blood pressure and heart rate, and using established standards to conduct comparative analysis on postoperative recovery time, thus forming empirical evaluation results of the anesthesia protocol.

[0003] Traditional anesthesia protocol evaluation systems are based on static rule settings and rely excessively on structured data from medical records and intraoperative monitoring. They lack in-depth analysis of dynamic trends during anesthesia, resulting in insufficient integration when processing multi-source heterogeneous clinical information. Short-term fluctuations in intraoperative vital signs are difficult to correlate with the drug action process, and nonlinear changes in key response segments cannot be identified. There is a lack of effective means to track the behavioral trajectory between functional performance during postoperative recovery and preoperative state. This results in significant limitations in the accuracy and adaptability of protocol evaluation results, affecting the judgment of the rationality of anesthesia strategies in complex surgical scenarios and limiting the support that protocol evaluation can provide for clinical practice. Summary of the Invention

[0004] To address the shortcomings of existing technologies, such as over-reliance on structured data from medical records and intraoperative monitoring, lack of in-depth analysis of dynamic trends during anesthesia, insufficient integration when processing multi-source heterogeneous clinical information, difficulty in effectively correlating short-term fluctuations in intraoperative vital signs with drug action, inability to identify nonlinear changes in key response segments, and lack of effective means to track the behavioral trajectory between postoperative functional performance and preoperative state, the technical problems of limited accuracy and adaptability in protocol evaluation results affect the judgment of the rationality of anesthesia strategies in complex surgical scenarios and limit the support of protocol evaluation for clinical practice, this invention provides an anesthesia protocol evaluation system based on big data analysis. The technical solution is as follows: On the one hand, a big data analytics-based anesthesia protocol evaluation system is provided, which includes: The response segment identification module acquires the intraoperative anesthesia depth sequence, divides it into equal-length segments according to time steps, calculates the slope change, filters out continuous segments with abnormal slope, performs time mapping and response synchronicity judgment, and generates a multi-segment anesthesia response dynamic feature set. Based on the multi-segment anesthesia response dynamic feature set, the hysteresis offset linkage module extracts the corresponding drug injection time, sets the response detection window, calculates the difference between the index response start point and the injection time, calculates the index change difference, and obtains the dose response hysteresis matching label set. The visual induction trajectory module obtains preoperative visual task eye movement trajectory and direction judgment data based on the dose response lag matching tag set, identifies visual response mutation points during the induction period, extracts eye movement offset direction and duration, records postoperative graphic trajectory error information and matches it with intraoperative tasks to obtain the correlation structure between induction and postoperative visual response trajectories. Based on the induction and postoperative visual response trajectory association structure, the risk association mapping module filters repeated path offset segments, extracts drug injection information and path change nodes during the induction period, calculates the postoperative trajectory error density, marks hot zone paths that meet the feature conditions, and generates a set of path association risk intervention segments.

[0005] As a further aspect of the present invention, the multi-segment anesthesia response dynamic feature set includes response boundary structure, response synchronization mode, dynamic change amplitude, and characteristic time points; the dose response lag matching label set includes response lag time difference, response intensity label, dose correspondence, and initiation threshold marker; the induction and postoperative visual response trajectory association structure includes induction deviation mode, postoperative trajectory error type, visual response matching path, and eye movement direction distribution; and the path-associated risk intervention segment set includes path deviation segment, error high-density region, intervention drug node, and risk path mapping relationship.

[0006] As a further aspect of the present invention, the response segment identification module includes: The slope calculation submodule acquires the intraoperative anesthesia depth sequence, divides it into sequence segments of equal length according to time steps, acquires the anesthesia depth values ​​of adjacent sampling points within the segment, calculates the slope change value between adjacent points within the segment by performing a ratio operation on the difference of anesthesia depth values ​​in consecutive time steps and the time difference, and uses the segment as a unit to calculate the mean and range of slope within the segment, generating a segment slope change sequence. The abnormal segment screening submodule judges the difference of the slope change value of the segment based on the slope change sequence, uses the difference between the segment slope change value and the overall slope mean as the threshold judgment, sets the slope abnormality threshold as the screening benchmark value, screens the segment sequence that continuously meets the slope abnormality threshold condition, calls the drug injection time information and physiological index sequence corresponding to the abnormal segment, establishes the correspondence on the time axis, and generates a response synchronization mapping control set; The multi-segment feature recognition submodule calls the response synchronization mapping comparison set, which combines the time tags of abnormal segments with the corresponding drug injection information and physiological index values. It aligns the data according to the time mapping order, measures the sequence difference of the physiological index curve trend changes between adjacent segments, and judges the change boundary by combining the drug injection type change node marker. It identifies the boundary position and response relationship of the trend change breakpoint within the segment group and generates a multi-segment anesthesia response dynamic feature set.

[0007] As a further aspect of the present invention, the hysteresis offset linkage module includes: The response time difference calculation submodule extracts the injection time corresponding to each feature based on the drug injection time recorded in the multi-segment anesthesia response dynamic feature set, sets a response detection window of fixed length, collects the corresponding physiological index sequence from the injection time to the back, obtains the position of the response start point on the time axis by detecting the position of the inflection point of the continuous change trend in the index curve, and then calculates the time difference with the injection time point to generate a set of index response time offsets. The response tag allocation submodule calls the time difference results in the indicator response time offset set, combines them with the original indicator curve data in the response detection window, calculates the average value of the indicator in the two time periods before and after injection, and takes the difference as the indicator response intensity. Combining the combination of time offset and response intensity, the response mode corresponding to the injection time is mapped to the matching state, and a dose response lag matching tag set is generated.

[0008] As a further aspect of the present invention, the risk association mapping module includes: The visual mutation identification submodule obtains visual task eye movement trajectory data and direction judgment data within the corresponding time period before surgery based on the time position of the matching tags in the dose response lag matching tag set. It performs time-series analysis on the eye movement trajectory, identifies the inflection point position of the trajectory curve where there is a change in direction or speed, and confirms whether the inflection point is a change associated with the task by combining the direction judgment data, thus obtaining a set of visually induced mutation points. The eye movement feature extraction submodule calls the eye movement trajectory data corresponding to the mutation point in the set of visually induced mutation points, obtains the eye movement direction change value and trajectory duration length in a continuous time period after the mutation, calculates the angle amplitude of the direction change, extracts the duration, constructs a two-dimensional feature vector, and labels the two-dimensional feature vector under the corresponding induction task label to generate a continuous feature set of eye movement offset direction. The trajectory association submodule obtains multiple error indices from the postoperative graphic trajectory test results based on the continuous feature set of eye movement offset direction, calculates the geometric offset value and cumulative error between the postoperative trajectory and the preoperative induction trajectory, matches the target trajectory segment corresponding to the intraoperative induction task, and performs mapping association based on task type and time sequence to generate an induction and postoperative visual response trajectory association structure.

[0009] As a further aspect of the present invention, the process of identifying the inflection point position of the trajectory curve where there is a sudden change in direction or a change in speed is specifically as follows: by setting a sliding window with a fixed time window length of 100 milliseconds in the eye movement trajectory, the change in eye movement position within each sliding window is vector calculated. If the angle of change in vector direction between two adjacent windows is greater than 25 degrees and the rate of change in vector length exceeds a set threshold of 10%, then the associated time point is recorded as the inflection point position. The process of confirming whether the inflection point is a task-induced change by combining the direction judgment data is specifically as follows: determine whether the inflection point position appears within ±500 milliseconds of the time position corresponding to the matching label, and whether it is consistent with the target reaction direction in the direction judgment data. Only the inflection point positions that meet both conditions are retained to form a set of visually induced mutation points.

[0010] As a further aspect of the present invention, the risk association mapping module includes: The path offset filtering submodule, based on the trajectory matching path in the induction and postoperative visual response trajectory association structure, retrieves segments that spatially overlap between the preoperative induction trajectory and the postoperative trajectory, compares the start and end positions and deformation trends of the corresponding path segments, determines whether there are repeated trajectory segments with the same direction but offset amplitude exceeding a set threshold, extracts sequence information that meets the path offset conditions, and generates a set of repeated path offset segments. The error hot zone annotation submodule calls the spatial region of the path segment in the set of repeated path offset segments, extracts the corresponding drug injection time and path change nodes during the induction period, and calculates the cumulative error per unit area of ​​each segment as the trajectory error density by combining the error distribution of the postoperative trajectory in the same region. It then filters out regions where the error density exceeds the preset trajectory error density threshold and generates a path error density hot zone annotation set. The risk intervention segment generation submodule obtains the path segment number based on the coordinates of the hot zone region in the path error density hot zone annotation set, and matches the corresponding drug injection time, trajectory change node and task type information. It then annotates and integrates the regional segments that meet the requirements of induced deviation path and error density threshold to generate a path-related risk intervention segment set.

[0011] As a further aspect of the present invention, the process of determining whether there are repeated trajectory segments with the same direction but an offset exceeding a set threshold specifically involves: based on the tangential angle change sequence of each corresponding segment in the path matching, if the angle difference does not exceed 15 degrees, it is considered to be in the same direction; at the same time, the Euclidean distance offset between the start and end points of the trajectory segment is calculated, and if the Euclidean distance offset is greater than the set threshold of 3 mm, it is considered to be an offset exceeding the threshold. The process of extracting sequence information that meets the path offset conditions specifically involves filtering trajectory segments that simultaneously meet the conditions of consistent direction and excessive offset, recording the start and end timestamps, spatial coordinate range, and trajectory number in the order of trajectory time, and forming a set of repeated path offset segments.

[0012] As a further aspect of the present invention, the system also includes a solution structure evaluation module: The protocol structure evaluation module extracts the frequency of each drug combination and the input structure based on the path-related risk intervention segment set, calculates the matching distribution with the postoperative trajectory error, statistically analyzes the frequency and deviation level, and generates an anesthesia protocol structure evaluation tag set by combining the correspondence between intervention structure behavior and response results. The anesthesia protocol structure assessment label set includes drug combination labels, frequency distribution levels, error offset levels, and intervention effect markers.

[0013] As a further aspect of the present invention, the structure evaluation module includes: The drug structure statistics submodule extracts the drug combination configuration and input order within the drug injection time period corresponding to the path-related risk intervention segment set, calculates the frequency of occurrence of multiple drug combinations in the intervention segment, calculates the spatial distribution similarity value between the combination input structure and the postoperative trajectory error, records the frequency of error matching degree under each structure in the segment set, and generates a drug structure matching distribution statistics table. The protocol label generation submodule calls the drug structure matching distribution statistics table, combines the intervention structure behavior associated with multiple drug combinations and the offset level of the corresponding postoperative response results, quantifies and classifies the correspondence between differentiated drug combinations and response results, assigns classification labels, and generates an anesthesia protocol structure evaluation label set.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By constructing a multi-segment dynamic structure of anesthesia depth based on time step division and slope analysis, the temporal synchronicity changes of drug injection information and physiological indicators are extracted, enabling boundary identification and pattern summarization of intraoperative multi-segment response characteristics. Combining response delay characteristics and initiation point differences, dose-response relationship labels corresponding to indicator changes are obtained. By integrating preoperative visual induction tasks and postoperative trajectory deviation records, a dynamic correlation structure of eye-tracking response behavior is established, and the intervention correlation between path error regions and drug administration pathways is explored. Through drug combination input and error matching distribution analysis, a structured mapping and label evaluation between anesthesia strategies and postoperative patient behavior is achieved, effectively improving the pertinence and response accuracy of anesthesia protocol structure evaluation. Attached Figure Description

[0015] 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.

[0016] Figure 1 This is a schematic diagram of the system provided by the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention; Figure 3 This is a flowchart of the response segment identification module in this invention; Figure 4 This is a flowchart of the hysteresis offset linkage module in this invention; Figure 5 This is a flowchart of the risk association mapping module in this invention; Figure 6 This is a flowchart of the risk association mapping module in this invention; Figure 7 This is a flowchart of the scheme structure evaluation module in this invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] 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.

[0019] 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.

[0020] 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.

[0021] 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.

[0022] This invention provides an anesthesia protocol evaluation system based on big data analysis, such as... Figure 1-2 The diagram shown illustrates an anesthesia protocol evaluation system based on big data analytics. This system includes: The response segment identification module acquires the intraoperative anesthesia depth sequence, divides it into equal-length segments according to time steps, calculates the slope change, filters out continuous segments with abnormal slope, extracts the corresponding drug injection information and physiological index sequence, performs time mapping and response synchronicity judgment, calculates the boundary structure of multi-segment response changes, and generates a multi-segment anesthesia response dynamic feature set. The hysteresis linkage module is based on the dynamic feature set of multi-segment anesthesia response, extracts the corresponding drug injection time, sets the response detection window, calculates the difference between the index response start point and the injection time, calculates the index change difference, and obtains the dose response hysteresis matching label set. The visual induction trajectory module obtains preoperative visual task eye movement trajectory and direction judgment data based on dose response lag matching label set, identifies visual response mutation points during induction, extracts eye movement offset direction and duration, records postoperative graphic trajectory error information and matches intraoperative tasks, and obtains the correlation structure between induction and postoperative visual response trajectories. The risk association mapping module is based on the association structure of induction and postoperative visual response trajectories. It filters repeated path offset segments, extracts drug injection information and path change nodes during the induction period, calculates postoperative trajectory error density, marks hot zone paths that meet the feature conditions, and generates a set of path association risk intervention segments. The protocol structure assessment module extracts the frequency of each drug combination and the input structure based on the path-related risk intervention segment set, calculates the matching distribution with the postoperative trajectory error, statistically analyzes the frequency and deviation level, and generates an anesthesia protocol structure assessment tag set by combining the correspondence between intervention structure behavior and response results. The multi-segment anesthesia response dynamic feature set includes response boundary structure, response synchronization mode, dynamic change amplitude, and characteristic time points. The dose response lag matching label set includes response lag time difference, response intensity label, dose correspondence, and initiation threshold marker. The induction and postoperative visual response trajectory association structure includes induction deviation mode, postoperative trajectory error type, visual response matching path, and eye movement direction distribution. The path-related risk intervention segment set includes path deviation segment, error high-density region, intervention drug node, and risk path mapping relationship. The anesthesia protocol structure assessment label set includes drug combination label, frequency distribution level, error deviation level, and intervention effect marker.

[0023] Specifically, such as Figure 2 , 3 As shown, the response segment identification module includes: The slope calculation submodule acquires the intraoperative anesthesia depth sequence, divides it into sequence segments of equal length according to time steps, acquires the anesthesia depth values ​​of adjacent sampling points within the segment, calculates the slope change value between adjacent points within the segment by performing a ratio operation on the difference of anesthesia depth values ​​in consecutive time steps and the time difference, and uses the segment as a unit to calculate the mean and range of slope within the segment, generating a segment slope change sequence. The data from the entire surgical procedure are segmented according to fixed time steps, for example, every 60 seconds is a time window. A 3-hour intraoperative anesthesia process is divided into 180 segments, each containing dozens to hundreds of sampling points. The numerical difference between adjacent anesthesia depth data points within each segment is extracted, and then the ratio is processed according to the time interval between adjacent data points to form a set of change values ​​reflecting the rate of increase or decrease of the data. Each change value represents the rate of change of anesthesia depth between the previous time point. The slope values ​​are traversed within each segment, and the average, maximum, and minimum slope values ​​are calculated. The difference in slope, or the range of slope variation, serves as a statistical characteristic of the segment. The above steps are repeated until the entire anesthesia sequence is completed, forming a sequence composed of the average slope and range of variation of multiple segments. This serves as an important basis for identifying abnormal fluctuations. For example, in a segment of intraoperative anesthesia monitoring, if the anesthesia depth values ​​are continuously sampled as 60, 58, 57, 59, and 62 within a certain 1-minute segment, the slope between the consecutive points shows a trend of first decreasing and then increasing. The average slope is close to 0, but the range of variation is large, indicating that there is significant fluctuation in the anesthesia depth within the segment, generating a sequence of changes in the slope of the segment.

[0024] The abnormal segment screening submodule judges the difference of the slope change value of the segment based on the segment slope change sequence, uses the difference between the segment slope change value and the overall slope mean as the threshold judgment, sets the slope abnormality threshold as the screening benchmark value, screens the segment sequence that continuously meets the slope abnormality threshold condition, calls the drug injection time information and physiological index sequence corresponding to the abnormal segment, establishes the correspondence on the time axis, and generates a response synchronization mapping control set; The mean slope of each segment is compared with the overall mean slope during the entire anesthesia process, and the difference is used as a preliminary criterion for determining whether it is abnormal. If the difference reaches a set abnormal slope threshold, it is considered an abnormal segment with suspicious fluctuations. The judgment process involves sequentially filtering within the segments and checking for abnormal sequences formed by multiple consecutive abnormal segments to enhance the stability of abnormal trend identification. After filtering out abnormal segment sequences, the actual time tag of the segment during the operation is retrieved, and the anesthetic drug injection record information consistent with the time period is retrieved, including drug name, injection rate, and injection method. If abnormal segments occur consecutively between the 45th and 48th minute of the surgery, the system will simultaneously search whether propofol or remifentanil was continuously injected during the time period, and obtain information such as injection frequency, interval, and rate. At the same time, the system will also synchronously call the patient's physiological indicators corresponding to the time period, such as heart rate, blood pressure, and blood oxygen saturation, extract the original value sequence of the indicators and align them one by one with the abnormal segment time, completing the time synchronization process of multi-source data. The system integrates the time tag, drug injection data, and physiological indicator sequence into a control set to generate a response synchronization mapping control set.

[0025] The multi-segment feature recognition submodule calls the response synchronization mapping comparison set, which combines the time labels of abnormal segments with the corresponding drug injection information and physiological index values, performs alignment processing according to the time mapping order, measures the sequence difference of the trend changes of physiological index curves between adjacent segments, and judges the change boundary by combining the drug injection type change node markers, identifies the boundary position and response relationship of the trend change breakpoint within the segment group, and generates a multi-segment anesthesia response dynamic feature set. Abnormal segment information is sorted and processed chronologically to ensure that data from different sources are aligned along a unified timeline. Within each pair of adjacent segments, the physiological indicator values ​​are compared to determine if there are significant fluctuations or turning points, such as whether heart rate shows a continuous increase, sudden increase, or sudden decrease. If there are significant differences in the changes of physiological indicators between two adjacent segments, the system marks them as trend change nodes and checks whether these two segments are accompanied by changes in the type of anesthetic drug injection; for example, the previous segment might be a continuous injection of propofol, while the subsequent segment might be an intermittent injection of remifentanil. If the drug injection method changes and physiological indicators also show a sudden change in trend, the node is marked as an important change boundary. Within a complete multi-segment group, the system will identify multiple boundary nodes and summarize the trend change direction between the breakpoints and the relationship with drug changes. For example, during a certain operation, the system detects that in the segment from the 75th to the 78th minute of the operation, the patient's heart rate continuously rises from 80 to 95, while the anesthetic drug is changed from propofol alone to remifentanil in combination. The node will be identified as a multi-segment feature boundary and a multi-segment anesthesia response dynamic feature set will be generated.

[0026] Specifically, such as Figure 2 , 4 As shown, the hysteresis offset linkage module includes: The response time difference calculation submodule is based on the drug injection time recorded in the dynamic feature set of multiple anesthesia response. It extracts the injection time corresponding to each feature, sets a fixed-length response detection window, collects the corresponding physiological index sequence from the start of the injection time, obtains the position of the response start point on the time axis by detecting the position of the inflection point of the continuous change trend in the index curve, and then calculates the time difference with the injection time point to generate a set of index response time offsets. Starting with the drug injection time recorded in the dynamic feature set of multi-segment anesthesia response, the system extracts the drug injection time corresponding to each feature segment segment. A fixed-length detection window is set, such as 30 seconds or 60 seconds after the start of injection, to observe the dynamic changes of corresponding physiological indicators after drug injection. Within each detection window, the system continuously collects the patient's target physiological indicator sequence, such as heart rate, blood pressure, or blood oxygen saturation data at the second level, and performs trend analysis on the curves formed by the changes in the indicators. The trend analysis method calculates the directionality of the indicator changes point by point, determines whether the indicator is continuously rising, falling, or remaining flat, and finds the time point when the trend shows a significant turning point as the indicator. The response initiation point, for example, is the point at which the heart rate suddenly and rapidly increases from a steady state 20 seconds after the start of drug injection. This point is considered the time when the indicator responds to the drug. The system compares the response initiation time with the drug injection start time to obtain the specific time interval, which is used as the response time difference of the physiological indicator to the current drug injection. Such differences will be calculated and summarized one by one in the records. Taking a surgical scenario as an example, within 30 seconds after the start of remifentanil injection, the blood pressure drop was detected to begin at the 18th second and the heart rate drop to begin at the 22nd second. The differences are recorded as 18 seconds and 22 seconds, respectively, and are included in the key reference indicators for response label evaluation, generating a set of indicator response time offsets.

[0027] The response tag allocation submodule calls the time difference results in the indicator response time offset set, combines them with the original indicator curve data in the response detection window, calculates the average value of the indicator in the two time periods before and after injection, and takes the difference as the indicator response intensity. Combining the combination of time offset and response intensity, the response mode corresponding to the injection time is mapped to the matching state, and a dose response lag matching tag set is generated. By combining the original indicator sequence data from the two time periods before and after injection, and comparing and analyzing the actual impact of injection on indicator values, the system divides the response detection window into two parts. For example, the first 15 seconds are designated as the pre-injection interval, and the next 15 seconds as the post-injection interval. The average value of the physiological indicator data in the two time periods is calculated. For example, if the average heart rate is 85 before injection and 92 after injection, the response intensity is 7. The response intensity reflects the magnitude of change in physiological indicators caused by drug injection. The system combines the intensity with the previously calculated response time offset to analyze the matching relationship. For example, if the response time difference is less than 10 seconds and the response intensity exceeds 5, it is judged as a "strong match". If the time difference is between 10 and 20 seconds and the intensity is low, it can be marked as "medium match"; if it exceeds 20 seconds but the intensity is significant, it is classified as "lagging strong match" or similar states. By setting the standards for these match states, the system establishes a fixed mapping relationship between each drug injection behavior and the corresponding indicator response pattern. For example, during a patient's surgery, a turning point in blood pressure occurs 12 seconds after propofol injection, and the change before and after is 11 units. This is marked as "medium lag high response", reflecting that the injection behavior produces a significant but slightly delayed physiological effect. The match state is recorded in the label set accordingly, forming a data structure that can be used for subsequent analysis, classification and modeling, generating a dose response lag match label set.

[0028] Specifically, such as Figure 2 , 5 As shown, the risk association mapping module includes: The visual mutation identification submodule obtains visual task eye movement trajectory data and direction judgment data within the corresponding time period before surgery based on the time position of the matching tags in the dose response lag matching tag set. It performs time series analysis on the eye movement trajectory, identifies the inflection point position of the trajectory curve where there is a change in direction or speed, and confirms whether the inflection point is a change associated with the task by combining the direction judgment data, thus obtaining a set of visually induced mutation points. The system retrieves visual task data from a preoperative time window, primarily including eye movement trajectory information and visual direction judgment results. During the preoperative visual induction task, subjects complete visual recognition by focusing on a moving target point on the screen. Simultaneously, the system records eye movement trajectory data, such as the complete sequence of changes in gaze coordinates in the horizontal and vertical directions over time. The submodule performs frame-by-frame temporal analysis on these trajectory sequences to identify inflection points where velocity or direction changes occur. For example, if the gaze direction changes from left to right within three consecutive sampling points, or if the eye movement velocity suddenly increases from a stable state to a significant shift within a very short time, the system identifies these as inflection points. The system then compares these inflection points with the direction judgment data, i.e., the target judgment results recorded in the preoperative task, to determine whether the target direction was correctly judged. This helps determine whether the inflection point was indeed caused by task induction rather than random disturbances or equipment errors. If the trajectory inflection point and the direction judgment event are temporally consistent, and the trajectory characteristics match the induced target direction in the task, a set of visual induction inflection points is obtained.

[0029] The eye movement feature extraction submodule calls the eye movement trajectory data corresponding to the mutation point in the visual induced mutation point set, obtains the eye movement direction change value and trajectory duration length within a continuous time period after the mutation, calculates the angle amplitude of the direction change, extracts the duration, constructs a two-dimensional feature vector, and labels the two-dimensional feature vector under the corresponding induced task label to generate a continuous feature set of eye movement offset direction. The system performs feature extraction on the eye movement trajectory after each mutation point, focusing on the change in eye movement direction and duration within a certain time range after the mutation. The system tracks the time series of the gaze coordinates backward from the mutation point, extracting the changes in the horizontal and vertical directions over several consecutive seconds, and calculating the magnitude of the directional angle formed by the changes, such as the rotation angle formed by the gaze shifting from upward to right. The rotation angle is a key measure of the change in eye movement direction, used to represent the magnitude of the change in the direction of the eye movement trajectory after the mutation. At the same time, the system counts the time taken from the mutation point to the point where the direction remains relatively stable, which is regarded as the duration of the eye movement trajectory in the direction. These two data are combined to form a two-dimensional feature vector, corresponding to the eye movement response features of the mutation point under the visual induction task. For example, in a certain mutation event, the direction change angle is 90 degrees and the duration is 1.8 seconds. The two-dimensional feature vector is labeled under the induction task. The system integrates the feature vectors generated by multiple mutation points to generate a feature set of eye movement offset direction persistence.

[0030] The trajectory association submodule obtains multiple error indicators from the postoperative graphic trajectory test results based on the continuous feature set of eye movement offset direction, calculates the geometric offset value and error accumulation between the postoperative trajectory and the preoperative induction period trajectory, matches the target trajectory segment corresponding to the intraoperative induction task, and performs mapping association based on task type and time sequence to generate the induction and postoperative visual response trajectory association structure. The system retrieves postoperative trajectory testing data, including records of subjects tracking or drawing target graphics on a screen. It statistically analyzes multiple error indicators in the postoperative trajectory data, such as the geometric distance of the trajectory deviating from the original target path, cumulative deviation value, and directional shift angle. After summarizing the errors, a set of performance characteristics for the postoperative trajectory is formed. The system matches the eye movement trajectory characteristics generated during the preoperative induction period with the postoperative trajectory data, based on similar trajectory trends and eye movement direction characteristics. Simultaneously, the system maps and matches task types with corresponding time series information. For example, if a right-side induction task is performed during a certain time period during surgery, the system compares the error indicators of the relevant motion trajectories in the right-side corner of the postoperative graphics. In this way, a structured mapping relationship is established between the intraoperative visual induction task and the postoperative eye movement response behavior, generating a correlation structure between the induction and postoperative visual response trajectories. For instance, the system can output a sudden change in the trajectory of a certain intraoperative induction right-side motion task, and if a continuous error shift occurs in the same area of ​​the trajectory during the postoperative task, a correlation structure between the induction and postoperative visual response trajectories is generated.

[0031] Specifically, such as Figure 2 , 6 As shown, the risk association mapping module includes: The path offset filtering submodule is based on the trajectory matching path in the correlation structure of the induction and postoperative visual response trajectory. It searches for segments that spatially overlap between the preoperative induction trajectory and the postoperative trajectory, compares the start and end positions and deformation trends of the corresponding path segments, determines whether there are repeated trajectory segments with the same direction but offset amplitude exceeding the set threshold, extracts the sequence information that meets the path offset conditions, and generates a set of repeated path offset segments. The system's trajectory comparison program automatically retrieves trajectory segments that overlap in spatial location between the eye movement trajectory generated in the preoperative induction task and the postoperative trajectory test results. Each trajectory segment is analyzed according to the trajectory coordinate sequence, and the spatial relationship between the starting and ending points is determined. The system analyzes whether the path direction is consistent, including whether the movement trends in the horizontal, vertical, and diagonal directions are approximately overlapping. Under the premise of consistent spatial path direction, the system calculates the positional offset of the two trajectories in the same direction and compares the offset distance with a preset offset judgment threshold. If the offset exceeds the threshold, the trajectory segment is marked as a repeated offset path segment. The system traverses comparable trajectory combinations, extracts multiple sets of trajectory segments that meet the offset judgment conditions, and records information such as start and end times, spatial area range, and matching task type. For example, in a certain test, the induction task guided the eye movement from coordinates (10, 20) to (40, 60), while the postoperative trajectory, although in the same direction, had a starting point of (15, 25) and an ending point offset to (45, 65). The spatial offset exceeded the preset offset threshold by 2 units in both the horizontal and vertical directions, generating a set of repeated path offset segments.

[0032] The error hot zone annotation submodule calls the spatial region of the path segment in the set of repeated path offset segments, extracts the corresponding drug injection time and path change nodes during the induction period, combines the error distribution of the postoperative trajectory in the same region, calculates the cumulative error per unit area of ​​each segment as the trajectory error density, filters the region whose error density exceeds the preset trajectory error density threshold, and generates a path error density hot zone annotation set. The system extracts key trajectory nodes from the drug injection time and start / end coordinates during the induction period for each path segment, such as trajectory direction turning points and velocity change points. It also identifies the positions of the segmented trajectory in the postoperative repeat path, analyzes the error distribution of the postoperative trajectory within the repeat area, and statistically analyzes the spatial coordinates and error magnitudes of multiple error points within the area, such as positional deviation values. By calculating the total error value of the error points within the area and dividing it by the spatial area occupied by multiple trajectory segments, the system calculates the cumulative error per unit area of ​​the path segment, i.e., the error density index. The system presets an error density threshold, for example, 5 error units per square pixel. If the error density of a trajectory segment in a certain area exceeds the threshold, it is identified as an error hotspot area. Such areas are characterized by significant concentrated errors. The system spatially labels these areas as hotspots and records metadata such as the associated induction task, start / end time period, and path segment number. The system integrates and outputs these data to form a path error density hotspot annotation set. For example, in an upward trajectory segment of an induction task, if the postoperative trajectory shows concentrated offset with a cumulative error of 75 units, the trajectory occupies 15 square pixels, and the error density is 5.0, which equals the set threshold, thus generating a path error density hotspot annotation set.

[0033] The risk intervention segment generation submodule obtains the path segment number based on the coordinates of the hot zone region in the path error density hot zone annotation set, and matches the corresponding drug injection time, trajectory change node and task type information. It then annotates and integrates the regional segments that meet the requirements of induced deviation path and error density threshold to generate a path-related risk intervention segment set. Based on the spatial coordinates of each hot zone in the path error density hot zone annotation set, the system retrospectively identifies the path segment number to which it belongs and retrieves information such as drug injection time nodes, key points of trajectory changes, and task types associated with the path segment. During the process, the system filters trajectory segments that meet the dual conditions of induced deviation path and error density threshold, retaining only the regions that have both significant spatial repetition deviation and exceed the risk threshold in error density. The trajectory segments are standardized, numbered, and labeled, and integrated into a unified data structure to form a complete set of path-related risk intervention segments. For example, in a visual tracking test, the system identifies a trajectory segment located between coordinates (20, 40) and (60, 80) that simultaneously meets the conditions of postoperative repetition deviation distance exceeding the standard and unit error density greater than 5. It also corresponds to the induction task of injecting remifentanil at the 35th minute of the operation, with the task type being left-lower direction eye movement induction. The system will mark it as an intervention segment that needs attention and generate a set of path-related risk intervention segments.

[0034] Specifically, such as Figure 2 , 7 As shown, the scheme structure evaluation module includes: The drug structure statistics submodule is based on the path-related risk intervention segment set. It extracts the drug combination configuration and input order within the drug injection time period corresponding to the segment content, counts the frequency of occurrence of multiple drug combinations in the intervention segment, calculates the spatial distribution similarity value between the combination input structure and the postoperative trajectory error, records the frequency of error matching degree under each structure in the segment set, and generates a drug structure matching distribution statistics table. The system extracts the drug injection time intervals corresponding to each segment and identifies the drug combination structures within those time periods, including drug types, the start and end times of each drug injection, input order, and interval time. The system then structurally categorizes these drug combination structures, generating statistical entries based on different combination types, such as single propofol injection, propofol combined with remifentanil, propofol-remifentanil-sequential injection, etc. After extracting the drug combination information associated with the trajectory segments, the system retrieves the frequency of each combination structure within the intervention segments and performs aggregated statistics, recording the total number of occurrences of each structure and the path segment number it covers. Based on this, it further analyzes and classifies the drug combination structures. The system performs a matching analysis between the spatial distribution of postoperative trajectory errors of the same drug structure combination and the corresponding path segment. It compares whether the error concentration areas of each path segment under the same combination structure have spatial consistency. The spatial distribution similarity value is calculated by the degree of spatial overlap of the matching areas. If multiple trajectory segments repeatedly show error hotspots in the same area, the similarity is high. The system records the matching degree between each drug input structure and trajectory error and summarizes the frequency of the structure in all intervention segments. For example, the alternating injection structure of propofol and remifentanil appears 18 times in the total segment, with 12 showing high spatial overlap. A statistical table of drug structure matching distribution is generated.

[0035] The protocol label generation submodule calls the drug structure matching distribution statistics table, combines the intervention structure behavior associated with multiple drug combinations and the offset level of the corresponding postoperative response results, quantifies and classifies the correspondence between differentiated drug combinations and response results, assigns classification labels, and generates an anesthesia protocol structure evaluation label set. The system categorizes the postoperative visual trajectory response results corresponding to each drug combination structure, paying particular attention to the correspondence between offset levels and error hotspot distributions. During the process, each drug combination structure is assigned a unique identification label, and all associated intervention segments are retrieved. The response error of each segment is graded by intensity, for example, segmented into low offset (less than 5 units), medium offset (5 to 10 units), and high offset (more than 10 units). Simultaneously, the frequency of structure occurrence at each offset level is recorded. The system forms a differential distribution matrix by pairing the drug structures with the response distributions, identifying correlated drug configuration patterns, and then classifying them according to the group... The stability of the combination is quantitatively assessed. For example, if a structure exhibits a high offset response in 90% of the associated fragments, it can be labeled as "high-risk" even if the combination is complex. On the other hand, if some structures do not trigger error hot zones in all associated fragments, they can be classified as "low-risk". The system converts the grading results in the above assessment process into structured classification labels, which are labeled according to the drug structure type. The content includes the combination drug structure, trajectory offset level, error spatial distribution characteristics, frequency of occurrence and suggested labels, such as "Structure A - Medium offset - Low regional overlap - Stable" and other types of labels. This provides a decision-making basis for the feasibility grading of different schemes and generates a set of structural assessment labels for anesthesia schemes.

[0036] The above are merely specific embodiments 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. An anesthesia protocol evaluation system based on big data analysis, characterized in that, The system includes: The response segment identification module acquires the intraoperative anesthesia depth sequence, divides it into equal-length segments according to time steps, calculates the slope change, filters out continuous segments with abnormal slope, performs time mapping and response synchronicity judgment, and generates a multi-segment anesthesia response dynamic feature set. Based on the multi-segment anesthesia response dynamic feature set, the hysteresis offset linkage module extracts the corresponding drug injection time, sets the response detection window, calculates the difference between the index response start point and the injection time, calculates the index change difference, and obtains the dose response hysteresis matching label set. The visual induction trajectory module obtains preoperative visual task eye movement trajectory and direction judgment data based on the dose response lag matching tag set, identifies visual response mutation points during the induction period, extracts eye movement offset direction and duration, records postoperative graphic trajectory error information and matches it with intraoperative tasks to obtain the correlation structure between induction and postoperative visual response trajectories. Based on the induction and postoperative visual response trajectory association structure, the risk association mapping module filters repeated path offset segments, extracts drug injection information and path change nodes during the induction period, calculates the postoperative trajectory error density, marks hot zone paths that meet the feature conditions, and generates a set of path association risk intervention segments.

2. The anesthesia protocol evaluation system based on big data analysis according to claim 1, characterized in that: The multi-segment anesthesia response dynamic feature set includes response boundary structure, response synchronization mode, dynamic change amplitude, and characteristic time points. The dose response lag matching label set includes response lag time difference, response intensity label, dose correspondence, and initiation threshold marker. The induction and postoperative visual response trajectory association structure includes induction deviation mode, postoperative trajectory error type, visual response matching path, and eye movement direction distribution. The path-associated risk intervention segment set includes path deviation segment, error high-density region, intervention drug node, and risk path mapping relationship.

3. The anesthesia protocol evaluation system based on big data analysis according to claim 1, characterized in that: The response segment identification module includes: The slope calculation submodule acquires the intraoperative anesthesia depth sequence, divides it into sequence segments of equal length according to time steps, acquires the anesthesia depth values ​​of adjacent sampling points within the segment, calculates the slope change value between adjacent points within the segment by performing a ratio operation on the difference of anesthesia depth values ​​in consecutive time steps and the time difference, and uses the segment as a unit to calculate the mean and range of slope within the segment, generating a segment slope change sequence. The abnormal segment screening submodule judges the difference of the slope change value of the segment based on the slope change sequence, uses the difference between the segment slope change value and the overall slope mean as the threshold judgment, sets the slope abnormality threshold as the screening benchmark value, screens the segment sequence that continuously meets the slope abnormality threshold condition, calls the drug injection time information and physiological index sequence corresponding to the abnormal segment, establishes the correspondence on the time axis, and generates a response synchronization mapping control set; The multi-segment feature recognition submodule calls the response synchronization mapping comparison set, which combines the time tags of abnormal segments with the corresponding drug injection information and physiological index values. It aligns the data according to the time mapping order, measures the sequence difference of the physiological index curve trend changes between adjacent segments, and judges the change boundary by combining the drug injection type change node marker. It identifies the boundary position and response relationship of the trend change breakpoint within the segment group and generates a multi-segment anesthesia response dynamic feature set.

4. The anesthesia protocol evaluation system based on big data analysis according to claim 3, characterized in that: The hysteresis offset linkage module includes: The response time difference calculation submodule extracts the injection time corresponding to each feature based on the drug injection time recorded in the multi-segment anesthesia response dynamic feature set, sets a response detection window of fixed length, collects the corresponding physiological index sequence from the injection time to the back, obtains the position of the response start point on the time axis by detecting the position of the inflection point of the continuous change trend in the index curve, and then calculates the time difference with the injection time point to generate a set of index response time offsets. The response tag allocation submodule calls the time difference results in the indicator response time offset set, combines them with the original indicator curve data in the response detection window, calculates the average value of the indicator in the two time periods before and after injection, and takes the difference as the indicator response intensity. Combining the combination of time offset and response intensity, the response mode corresponding to the injection time is mapped to the matching state, and a dose response lag matching tag set is generated.

5. The anesthesia protocol evaluation system based on big data analysis according to claim 4, characterized in that: The risk association mapping module includes: The visual mutation identification submodule obtains visual task eye movement trajectory data and direction judgment data within the corresponding time period before surgery based on the time position of the matching tags in the dose response lag matching tag set. It performs time-series analysis on the eye movement trajectory, identifies the inflection point position of the trajectory curve where there is a change in direction or speed, and confirms whether the inflection point is a change associated with the task by combining the direction judgment data, thus obtaining a set of visually induced mutation points. The eye movement feature extraction submodule calls the eye movement trajectory data corresponding to the mutation point in the set of visually induced mutation points, obtains the eye movement direction change value and trajectory duration length in a continuous time period after the mutation, calculates the angle amplitude of the direction change, extracts the duration, constructs a two-dimensional feature vector, and labels the two-dimensional feature vector under the corresponding induction task label to generate a continuous feature set of eye movement offset direction. The trajectory association submodule obtains multiple error indices from the postoperative graphic trajectory test results based on the continuous feature set of eye movement offset direction, calculates the geometric offset value and cumulative error between the postoperative trajectory and the preoperative induction trajectory, matches the target trajectory segment corresponding to the intraoperative induction task, and performs mapping association based on task type and time sequence to generate an induction and postoperative visual response trajectory association structure.

6. The anesthesia protocol evaluation system based on big data analysis according to claim 5, characterized in that: The process of identifying the inflection point position of the trajectory curve where there is a sudden change in direction or a change in speed is specifically as follows: by setting a sliding window with a fixed time window length of 100 milliseconds in the eye movement trajectory, the change in eye movement position within each sliding window is vector calculated. If the angle of change in vector direction between two adjacent windows is greater than 25 degrees and the rate of change in vector length exceeds a set threshold of 10%, then the associated time point is recorded as the inflection point position. The process of confirming whether the inflection point is a task-induced change by combining the direction judgment data is specifically as follows: determine whether the inflection point position appears within ±500 milliseconds of the time position corresponding to the matching label, and whether it is consistent with the target reaction direction in the direction judgment data. Only the inflection point positions that meet both conditions are retained to form a set of visually induced mutation points.

7. The anesthesia protocol evaluation system based on big data analysis according to claim 5, characterized in that: The risk association mapping module includes: The path offset filtering submodule, based on the trajectory matching path in the induction and postoperative visual response trajectory association structure, retrieves segments that spatially overlap between the preoperative induction trajectory and the postoperative trajectory, compares the start and end positions and deformation trends of the corresponding path segments, determines whether there are repeated trajectory segments with the same direction but offset amplitude exceeding a set threshold, extracts sequence information that meets the path offset conditions, and generates a set of repeated path offset segments. The error hot zone annotation submodule calls the spatial region of the path segment in the set of repeated path offset segments, extracts the corresponding drug injection time and path change nodes during the induction period, and calculates the cumulative error per unit area of ​​each segment as the trajectory error density by combining the error distribution of the postoperative trajectory in the same region. It then filters out regions where the error density exceeds the preset trajectory error density threshold and generates a path error density hot zone annotation set. The risk intervention segment generation submodule obtains the path segment number based on the coordinates of the hot zone region in the path error density hot zone annotation set, and matches the corresponding drug injection time, trajectory change node and task type information. It then annotates and integrates the regional segments that meet the requirements of induced deviation path and error density threshold to generate a path-related risk intervention segment set.

8. The anesthesia protocol evaluation system based on big data analysis according to claim 7, characterized in that: The process of determining whether there are repeated trajectory segments with the same direction but an offset exceeding a set threshold is as follows: based on the tangential angle change sequence of each corresponding segment in the path matching, if the angle difference does not exceed 15 degrees, it is considered to be in the same direction; at the same time, the Euclidean distance offset between the start and end points of the trajectory segment is calculated. If the Euclidean distance offset is greater than the set threshold of 3 mm, it is considered to be an offset exceeding the threshold. The process of extracting sequence information that meets the path offset conditions specifically involves filtering trajectory segments that simultaneously meet the conditions of consistent direction and excessive offset, recording the start and end timestamps, spatial coordinate range, and trajectory number in the order of trajectory time, and forming a set of repeated path offset segments.

9. The anesthesia protocol evaluation system based on big data analysis according to claim 1, characterized in that: The system also includes a scheme structure evaluation module: The protocol structure evaluation module extracts the frequency of each drug combination and the input structure based on the path-related risk intervention segment set, calculates the matching distribution with the postoperative trajectory error, statistically analyzes the frequency and deviation level, and generates an anesthesia protocol structure evaluation tag set by combining the correspondence between intervention structure behavior and response results. The anesthesia protocol structure assessment label set includes drug combination labels, frequency distribution levels, error offset levels, and intervention effect markers.

10. The anesthesia protocol evaluation system based on big data analysis according to claim 9, characterized in that: The scheme structure evaluation module includes: The drug structure statistics submodule extracts the drug combination configuration and input order within the drug injection time period corresponding to the path-related risk intervention segment set, calculates the frequency of occurrence of multiple drug combinations in the intervention segment, calculates the spatial distribution similarity value between the combination input structure and the postoperative trajectory error, records the frequency of error matching degree under each structure in the segment set, and generates a drug structure matching distribution statistics table. The protocol label generation submodule calls the drug structure matching distribution statistics table, combines the intervention structure behavior associated with multiple drug combinations and the offset level of the corresponding postoperative response results, quantifies and classifies the correspondence between differentiated drug combinations and response results, assigns classification labels, and generates an anesthesia protocol structure evaluation label set.

Images