Laparoscope risk prediction model based on big data and construction method thereof

Abstract

The invention relates to the technical field of laparoscope data processing, in particular to a laparoscope risk prediction model based on big data and a construction method thereof.The construction method comprises the following steps that S1, patient sign data before a laparoscope operation is collected from a big data platform, the patient sign data is preprocessed and converted into sign time sequence data, and the sign time sequence data is stored in a database; dividing the physical sign time sequence data into a plurality of time windows with fixed lengths, acquiring statistical characteristics of the physical sign time sequence data in each time window, judging abnormal time points in the time windows according to a preset statistical characteristic threshold value, and generating abnormal physical sign time sequence data. According to the method, through time serialization processing of the physical sign data of the patient, extraction of statistical features in the time window and recognition of abnormal points, the complex fluctuation condition of the state of the patient can be reflected comprehensively. By expanding the abnormal sign time sequence data to the multi-dimensional phase space, evolution trajectories of different types of abnormal signs can be described.

Description

Technical Field

[0001] The present invention relates to the technical field of laparoscopic data processing, and in particular to a laparoscopic risk prediction model based on big data and a construction method thereof. Background Art

[0002] The laparoscopic risk prediction model is an intelligent tool that predicts possible complications and other risk situations during laparoscopic surgery in a data-driven way. The model usually predicts personalized risk assessment results based on the patient's preoperative information and the complexity of the operation, helping doctors to fully identify high-risk patients before surgery and formulate corresponding intervention strategies.

[0003] In the risk prediction of laparoscopic surgery, the existing technology mainly relies on the static analysis of preoperative information, and it is difficult to dynamically capture the time series change characteristics of vital sign data. The lack of ability to identify short-term abnormal vital signs may lead to the neglect of certain key information related to intraoperative complications. For example, the patient's preoperative heart rate changes may fluctuate dramatically within a certain period of time, making it difficult to quickly capture its potential risks. In addition, the existing technology's use of complication data in historical cases only focuses on the overall risk, which is not conducive to in-depth mining of specific correlation data, limiting the model's ability to parse complex causal relationships. For example, the same type of complication may be associated with multiple abnormal vital signs, and it is difficult for existing technology to quantify its specific contribution through conditional probability models, resulting in insufficient accuracy of the prediction results. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a laparoscopic risk prediction model based on big data and a construction method thereof.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: a method for constructing a laparoscopic risk prediction model based on big data, comprising the following steps: S1: Collect patient vital sign data before laparoscopic surgery from the big data platform, perform preprocessing operations on the patient vital sign data and convert them into vital sign time series data, divide the vital sign time series data into multiple time windows of fixed length, obtain the statistical features of the vital sign time series data in each time window, judge the abnormal time points in the time window according to the preset statistical feature threshold, and generate abnormal vital sign time series data; S2: Expanding the abnormal vital sign time series data of different categories into independent multidimensional phase spaces respectively, by capturing the change of the motion trajectory of each category of abnormal vital sign time series data in the independent multidimensional phase space, calculating the Lyapunov exponent of the abnormal vital sign time series data, quantifying the change characteristics of the abnormal vital sign state of the patient in the independent multidimensional phase space, and generating the patient's vital sign state analysis results; S3: Based on the patient's vital sign status analysis results, the Lyapunov index of the abnormal vital sign time series data is divided into state stability, and the time window of the abnormal vital sign time series data with a positive Lyapunov index is marked as a potential risk time window, and a preoperative patient status record is generated; S4: Collect intraoperative complications that have occurred in historical laparoscopic surgery cases from the big data platform, extract data marked as potential risk time windows based on the preoperative patient status records, analyze the correlation between physical sign fluctuations and intraoperative complications within the potential risk time window, and obtain a risk probability distribution data set for intraoperative complications. Based on the risk probability distribution data set for intraoperative complications, a laparoscopic risk prediction model is constructed using the Bayesian inference method. The laparoscopic risk prediction model is used to output risk level labels for preoperative patients.

[0006] As a further solution of the present invention, the step of obtaining the statistical characteristics of the vital sign time series data in each time window is specifically as follows: S111: By accessing the big data platform, extracting the patient's vital sign data before laparoscopic surgery from the database of the big data platform, including intra-abdominal pressure, blood pressure, heart rate and respiratory rate, cleaning and standardizing the patient's vital sign data and sorting them according to timestamps to generate vital sign time series data; S112: Divide the vital sign time series data into multiple time windows according to fixed lengths, extract statistical features of the vital sign time series data in each time window, including mean, variance, skewness and kurtosis, and generate a time window feature data set.

[0007] As a further solution of the present invention, the step of acquiring the abnormal vital sign time series data is specifically as follows: S121: comparing the statistical features of the vital sign time series data in the time window feature data set with a preset statistical feature threshold, and when the statistical features of any vital sign time series data in the time window exceed the statistical feature threshold range, marking the time point in the time window as an abnormal time point, and generating an abnormal time point analysis result; S122: Based on the abnormal time point analysis result, the marked abnormal time point information, including the abnormal patient vital sign data name and the time range of the abnormal time point, is integrated to generate abnormal vital sign time series data.

[0008] As a further solution of the present invention, the acquisition step of capturing the change of the motion trajectory of each type of abnormal vital sign time series data in an independent multidimensional phase space is specifically as follows: S211: expanding the abnormal vital sign time series data into a multidimensional phase space by a delay embedding method, selecting a time delay in the multidimensional phase space and determining an embedding dimension of the multidimensional phase space, thereby generating multidimensional phase space data of the abnormal vital sign; S212: Analyze the changes in the motion trajectories of the abnormal signs in the multidimensional phase space data of the abnormal signs, and divide the changes in the motion trajectories of the abnormal signs into multiple dynamic modes, reveal the dynamic characteristics of the abnormal signs in the multidimensional phase space, and generate abnormal sign trajectory change analysis results.

[0009] As a further solution of the present invention, the step of obtaining the patient's physical sign status analysis result is specifically: S221: Based on the abnormal vital sign trajectory change analysis result, the formula is used: Calculation of Lyapunov exponents for abnormal physical signs time series data , integrate to get the Lyapunov index set; in, is the sampling duration of abnormal physical sign time series data in the multidimensional phase space, is the total number of abnormal signs in the abnormal sign time series data in the multidimensional phase space, It is the first The weight of abnormal signs, is the first order of abnormal physical sign time series data in multidimensional phase space. Abnormal signs during data sampling time The Euclidean distance of is the first order of abnormal physical sign time series data in multidimensional phase space. Abnormal signs in the initial data sampling time The Euclidean distance when is the natural logarithm, which is used to scale and linearize the ratio of distance changes in multidimensional phase space; S222: Based on the Lyapunov exponent set, analyze the sensitivity of changes in the patient's abnormal vital signs time series data in the multidimensional phase space to generate a patient vital signs status analysis result.

[0010] As a further solution of the present invention, the steps of obtaining the preoperative patient status record are specifically as follows: S311: Based on the patient's vital sign status analysis result, the Lyapunov index of the abnormal vital sign time series data is classified into states, the abnormal vital sign time series data with a positive Lyapunov index is marked as having unstable characteristics, and the corresponding time window in the abnormal vital sign time series data is marked to generate potential risk time window information; S312: The abnormal vital sign time series data with a negative Lyapunov index is marked as a stable state, the potential risk time window information is integrated, and a preoperative patient status record is generated.

[0011] As a further solution of the present invention, the steps for obtaining the risk probability distribution data set of intraoperative complications are specifically as follows: S321: Collect the intraoperative complications that have occurred in the historical cases of laparoscopic surgery from the big data platform, and based on the potential risk time window information in the preoperative patient status record, use the formula: Calculate the abnormal sign fluctuation conditions selected in the potential risk time window information Next, complications Probability of occurrence , generate prediction results of abnormal physical sign fluctuations and intraoperative complication probability; in, Indicates specific complications that have occurred during laparoscopic surgery from historical cases Abnormal physical sign fluctuation conditions in potential risk time window information in the event of occurrence The probability of occurrence, Indicates the intraoperative complications that have occurred obtained from the historical cases of laparoscopic surgery The overall probability of occurrence, Indicates abnormal physical sign fluctuation conditions in potential risk time window information The overall probability of S322: Compare the preset probability threshold of the correlation strength with the abnormal physical sign fluctuation and the prediction result of the intraoperative complication probability, analyze the correlation between the abnormal physical sign fluctuation and the historical intraoperative complication data, and generate a risk probability distribution data set of intraoperative complications.

[0012] As a further solution of the present invention, the acquisition steps of constructing a laparoscopic risk prediction model are specifically: S411: organizing the risk probability distribution data set of the intraoperative complications into conditional input data for Bayesian inference, analyzing the occurrence risk of each intraoperative complication under the current abnormal physical sign fluctuation condition, and establishing a laparoscopic risk prediction model; S412: Based on the laparoscopic risk prediction model, the patient's preoperative abnormal physical sign fluctuation data pre-input into the laparoscopic risk prediction model is combined with the intraoperative complication events that have occurred in the historical cases of laparoscopic surgery, and the posterior probability and risk level of the current patient's intraoperative complications are output to generate the output result of the laparoscopic risk prediction model.

[0013] A laparoscopic risk prediction model based on big data is obtained by a method for constructing a laparoscopic risk prediction model based on big data.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by processing the patient's vital signs data in a time series and extracting statistical features and identifying abnormal points within the time window, it is beneficial to comprehensively reflect the complex fluctuations of the patient's state. By expanding the abnormal sign time series data to a multidimensional phase space, it is not only possible to characterize the evolution trajectory of abnormal signs of different categories, but also to quantify their stability through the Lyapunov exponent, further clarify the potential time period of high-risk states, and lay the foundation for subsequent risk assessment. In addition, the correlation analysis of complication data and abnormal signs in historical surgical cases is used, and a conditional probability model is constructed, so that the prediction results can more carefully reflect the causal relationship between abnormal preoperative signs and intraoperative complications. The risk level is accurately calibrated by the Bayesian inference method, which improves the model's adaptability to complex clinical situations, enhances the reliability and practicality of the prediction, and provides data support for surgical decision-making by assisting in identifying the risk status of high-risk patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the main steps of the present invention; Figure 2 A flow chart for obtaining statistical characteristics of vital sign time series data in each time window in the present invention; Figure 3 A flow chart for obtaining abnormal physical sign time series data of the present invention; Figure 4 A flow chart of the present invention for capturing the changes in the motion trajectory of each type of abnormal vital sign time series data in an independent multi-dimensional phase space; Figure 5 A flow chart for obtaining the patient's vital sign status analysis results for the present invention; Figure 6 A flowchart for obtaining a preoperative patient status record for the present invention; Figure 7 A flow chart for obtaining a risk probability distribution data set of intraoperative complications for the present invention; Figure 8 A flow chart of constructing a laparoscopic risk prediction model for the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0018] See also Figure 1 The present invention provides a technical solution: a method for constructing a laparoscopic risk prediction model based on big data, comprising the following steps: S1: Collect patient vital sign data before laparoscopic surgery from the big data platform, including intra-abdominal pressure, blood pressure, heart rate, and respiratory rate, perform pre-processing operations on the patient vital sign data and convert it into vital sign time series data, divide the vital sign time series data into multiple time windows of fixed length, obtain the statistical characteristics of the vital sign time series data in each time window, including mean, variance, skewness, and kurtosis, judge the abnormal time points in the time window according to the preset statistical feature threshold, and generate abnormal vital sign time series data; S2: Through the delayed embedding method, the abnormal vital sign time series data of different categories are expanded to independent multidimensional phase spaces respectively. The independent multidimensional phase space contains multiple points, and each point represents the abnormal vital sign time series data of the same category. For example, the abnormal vital sign time series data of heart rate is processed by the delayed embedding method. Each point in the multidimensional phase space represents the state of the heart rate under multiple time delays, and the collection of multiple points describes the evolution trajectory of the heart rate in the multidimensional space. By capturing the changes in the motion trajectory of each type of abnormal vital sign time series data in the independent multidimensional phase space, the Lyapunov exponent of the abnormal vital sign time series data is calculated, and the change characteristics of the abnormal vital sign state of the patient in the independent multidimensional phase space are quantified, and the patient's vital sign state analysis results are generated; S3: Based on the results of the patient's vital sign status analysis, the Lyapunov index of the abnormal vital sign time series data is divided into state stability. For the abnormal vital sign time series data with a positive Lyapunov index, it is marked as having unstable characteristics, and the corresponding time window is marked as a potential risk time window. For the abnormal vital sign time series data with a negative Lyapunov index, it is marked as a stable state, and a preoperative patient status record is generated; S4: Collect intraoperative complications that have occurred in historical cases of laparoscopic surgery from the big data platform. Based on the preoperative patient status records, extract the data marked as the potential risk time window, analyze the correlation between physical sign fluctuations and intraoperative complications within the potential risk time window, establish the conditional probability relationship between physical sign fluctuations and possible intraoperative complications, and obtain the risk probability distribution data set of intraoperative complications. Based on the risk probability distribution data set of intraoperative complications, build a laparoscopic risk prediction model through the Bayesian inference method. The laparoscopic risk prediction model outputs the risk level label of the preoperative patient, including high risk, medium risk or low risk, to provide data support for the formulation of surgical plans and preoperative interventions; The abnormal vital sign time series data include the marked abnormal time window and the specific vital sign fluctuation values ​​it contains. The patient vital sign status analysis results include the evolution trend of abnormal vital signs and the corresponding status assessment values. The preoperative patient status record includes the status classification label and fluctuation characteristic description of the patient's vital signs. The laparoscopic risk prediction model includes comprehensive information on the relationship between risk status records and conditional probability.

[0019] See also Figure 2 , the specific steps for obtaining the statistical characteristics of the vital sign time series data in each time window are: S111: By accessing the big data platform, extracting the patient's vital sign data before laparoscopic surgery from the database of the big data platform, including intra-abdominal pressure, blood pressure, heart rate and respiratory rate, cleaning and standardizing the patient's vital sign data and sorting them according to timestamps to generate vital sign time series data; By accessing the big data platform that integrates the hospital's electronic medical records and surgical data, Hadoop tools are used to extract the vital sign data of patients undergoing laparoscopic surgery before surgery, including intra-abdominal pressure, blood pressure, heart rate and respiratory rate, from the database of the big data platform. The extracted data are cleaned, and abnormal data are eliminated by setting the normal physiological range of each sign. Linear interpolation is used to fill in the entries with fewer missing values, while records with a large proportion of missing values ​​are directly deleted. Subsequently, standardization tools such as Python's pandas library are used to standardize the data, and vital sign data of different dimensions and ranges are converted into consistent numerical forms. Finally, they are sorted and organized according to timestamps to complete the preprocessing of patient vital sign data and convert them into time series data.

[0020] S112: Divide the vital sign time series data into multiple time windows according to a fixed length, extract statistical features of the vital sign time series data in each time window, including mean, variance, skewness and kurtosis, and generate a time window feature data set; After completing the cleaning and standardization preprocessing of the patient's vital sign data before laparoscopic surgery, and converting the data of intra-abdominal pressure (in mmHg), blood pressure (in mmHg), heart rate (in bpm), and respiratory rate (in times / minute) into time series, the time series was divided into multiple fixed-length time windows based on the timestamp information of the time series using Python's pandas library, for example, each window contains all continuous vital sign data points for 30 seconds or 1 minute. In each time window, the mean function of the pandas library is used to calculate the mean, which reflects the average levels of intra-abdominal pressure, blood pressure, heart rate and respiratory rate in the time period; the variance is calculated by the var function to measure the fluctuation amplitude of each vital sign data. For example, the variance can be used to evaluate whether the intra-abdominal pressure or blood pressure fluctuates greatly in a certain period of time, which is of great significance for monitoring the pneumoperitoneum pressure or the stability of the patient's blood circulation during laparoscopic surgery; the skew function is used to calculate the skewness to describe the symmetry of the distribution of vital sign data. For example, the skewness value can reveal whether the heart rate tends to be abnormally high or abnormally low in a specific time period, thereby helping to predict whether the patient may be at risk of heart rate arrhythmia during surgery; the kurt function is used to calculate the kurtosis to evaluate the steepness of the distribution of vital sign data. For example, a high kurtosis value may indicate that there is a risk of a sharp change in respiratory rate in a short period of time, which may affect intraoperative ventilation management, and the calculation results are finally recorded as the characteristic data set for each time window.

[0021] See also Figure 3 ,The specific steps for obtaining abnormal physical sign time series data are: S121: Compare the statistical features of the vital sign time series data in the time window feature data set with a preset statistical feature threshold. When the statistical features of any vital sign time series data in the time window exceed the statistical feature threshold, mark the time point in the time window as an abnormal time point, and generate an abnormal time point analysis result. In each time window, the statistical features of the four types of vital signs data, namely, intra-abdominal pressure, blood pressure, heart rate and respiratory rate, are calculated respectively, and compared with the set statistical feature thresholds one by one to determine whether to mark the time point in the time window as an abnormal time point. The set statistical feature thresholds are: the mean statistical feature threshold range is -1 to 1, the variance statistical feature threshold is 0.5, the skewness statistical feature threshold is an absolute value less than 0.5, and the kurtosis statistical feature threshold is 2. If the time series is divided into multiple time windows, assuming that there are two time windows of 0-30 seconds and 30-60 seconds, the judgment process is as follows: Taking the time window of 0-30 seconds as an example, the intra-abdominal pressure is judged: assuming that the standardized intra-abdominal pressure statistical feature is calculated, its mean statistical feature is 1.2, which exceeds the mean statistical feature threshold range of -1 to 1; the variance statistical feature is 0.6, which is higher than the variance statistical feature threshold of 0.5; the skewness statistical feature is 0.7, and the absolute value is greater than the skewness statistical feature threshold of 0.5; the kurtosis statistical feature is 2.5, which exceeds the kurtosis statistical feature threshold of 2. Therefore, the time points in the 0-30 second time window are marked as abnormal time points due to abnormal intra-abdominal pressure. Judging blood pressure: Assuming that the standardized blood pressure statistical features are calculated, its mean statistical feature is -0.8, which is within the mean statistical feature threshold range of -1 to 1; the variance statistical feature is 0.4, which is less than the variance statistical feature threshold of 0.5; the skewness statistical feature is 0.3, and the absolute value is less than the skewness statistical feature threshold of 0.5; the kurtosis statistical feature is 1.8, which is less than the kurtosis statistical feature threshold of 2. Therefore, the time points in the 0-30 second time window are not marked due to abnormal blood pressure. Judging heart rate: Assuming that the standardized heart rate statistical features are calculated, its mean statistical feature is 0.2, which is within the mean statistical feature threshold range of -1 to 1; the variance statistical feature is 0.3, which is less than the variance statistical feature threshold of 0.5; the skewness statistical feature is -0.4, and the absolute value is less than the skewness statistical feature threshold of 0.5; the kurtosis statistical feature is 1.5, which is less than the kurtosis statistical feature threshold of 2. Therefore, the time points in the 0-30 second time window are not marked as abnormal heart rate. Judging the respiratory rate: Assuming that the standardized respiratory rate statistical feature is calculated, its mean statistical feature is 1.5, which exceeds the mean statistical feature threshold range of -1 to 1; the variance statistical feature is 0.9, which is higher than the variance statistical feature threshold of 0.5; the skewness statistical feature is 0.6, and the absolute value is greater than the skewness statistical feature threshold of 0.5; the kurtosis statistical feature is 2.8, which exceeds the kurtosis statistical feature threshold of 2. Therefore, the time points in the 0-30 second time window are marked as abnormal time points due to abnormal respiratory rate.

[0022] S122: Based on the abnormal time point analysis results, the marked abnormal time point information is integrated, including the abnormal patient vital sign data name and the time range of the abnormal time point, to generate abnormal vital sign time series data; After completing the statistical feature judgment of each vital sign data in each time window and marking the abnormal time points, the time point information marked as abnormal in all time windows is integrated. For example, for the 0-30 second time window, the time points in this window are marked as abnormal because the statistical features of the intra-abdominal pressure and respiratory rate exceed the threshold. The marked information includes the specific abnormal sign name (such as intra-abdominal pressure and respiratory rate) and the corresponding time point range (0-30 seconds). Assuming that the 30-60 second time window is not marked as abnormal, after integrating these marks, an abnormal sign time series data is generated, recording the sign category and specific time range of each abnormal time point, providing data support for subsequent diagnostic analysis.

[0023] See also Figure 4 , the specific steps for capturing the changes in the motion trajectory of each type of abnormal physical sign time series data in an independent multidimensional phase space are: S211: expanding the abnormal physical sign time series data into a multidimensional phase space by a delay embedding method, selecting a time delay in the multidimensional phase space and determining an embedding dimension of the multidimensional phase space, thereby generating multidimensional phase space data of the abnormal physical sign; After completing the labeling and integration of abnormal sign time series data, the delayed embedding method is used to process each type of abnormal sign time series data and expand it into an independent multidimensional phase space. First, the appropriate time delay is determined by analyzing the autocorrelation function of the time series. The autocorrelation function is calculated by Python's SciPy library to evaluate the correlation between different time points in the time series. The time delay is found by finding the point where the autocorrelation value first drops to zero or approaches zero. For example, for the abnormal heart rate sign time series data, the reference point of the time delay can be found by calculating its autocorrelation function. By selecting a reasonable time delay, it is ensured that the time series does not contain too much redundant information after delay embedding. If the time delay is too small, the data points are highly correlated, and the points in the multidimensional phase space will be too dense to fully reveal the dynamic characteristics. The embedding dimension is determined by combining the pseudo-neighbor point algorithm, and then the delayed embedding method is used to construct a multidimensional matrix. Each row represents the value of the heart rate at different time delays. Each row in the matrix is ​​mapped to a point in the multidimensional phase space. Finally, the independent multidimensional phase space is displayed by MATLAB.

[0024] S212: analyzing the changes in the motion trajectories of the abnormal signs in the multidimensional phase space data of the abnormal signs, and dividing the changes in the motion trajectories of the abnormal signs into multiple dynamic modes, revealing the dynamic characteristics of the abnormal signs in the multidimensional phase space, and generating abnormal signs trajectory change analysis results; Capture the changes in the motion trajectory of each type of abnormal sign time series data in an independent multidimensional phase space. Take the abnormal signs of heart rate, blood pressure, intra-abdominal pressure and respiratory rate as examples. Assume that the abnormal time series data of these signs are processed by the delayed embedding method and expanded into a set of points in the three-dimensional phase space. The three-dimensional coordinates of each point represent the state of the corresponding sign under three time delays. By connecting these points in time sequence, a trajectory can be formed. The distribution pattern of the points and their dynamic characteristics are as follows. For abnormal heart rate signs, periodic trajectory: The trajectory of the points is distributed in a regular closed loop, reflecting that the abnormal heart rate may be caused by persistent stress or anxiety in patients before surgery. For example, in preoperative monitoring, the fluctuation of heart rate presents a highly repetitive rhythm, which may be related to long-term psychological stress. Chaotic trajectory: The trajectory of the points is densely distributed in the local area but without repetition, indicating that the change of heart rate may be caused by complex nonlinear mechanisms, such as imbalance of the autonomic nervous system or potential cardiovascular disease. Random trajectories: The distribution of points is completely irregular, which may reflect that the heart rate before surgery is affected by multiple interference factors, such as sudden environmental stress or emotional fluctuations. In this case, the trajectory shows a chaotic distribution pattern. For abnormal blood pressure signs, periodic trajectories: The distribution of points forms a closed trajectory ring, indicating that the preoperative blood pressure may be affected by periodic factors, such as the patient taking antihypertensive drugs regularly, which causes the blood pressure changes to show periodic fluctuations. Chaotic trajectories: The trajectory of points is irregular but not completely random in an area, which may reflect complex endocrine regulation imbalances or nonlinear fluctuations in the course of chronic hypertension. Random trajectories: The distribution of points has no obvious regularity, which may be due to dehydration, pain or other unexpected physiological disturbances in the patient before surgery, causing abnormal blood pressure fluctuations. For abnormal intra-abdominal pressure signs, periodic trajectories: The trajectory is closed, which may correspond to the patient's experience of periodic eating or drinking before surgery, which affects the changes in intra-abdominal pressure, such as periodic fluctuations caused by preoperative fasting and gastrointestinal motility. Chaotic trajectories: The points are densely distributed but not repeated, which may reflect the complex fluctuations of intra-abdominal pressure before surgery caused by intestinal distension or ascites. Random trajectories: The points are scattered and irregular, indicating that the intra-abdominal pressure of the patient before surgery may be subject to unpredictable interference, such as acute abdominal pain or other sudden preoperative factors. For abnormal signs of respiratory rate, periodic trajectories: The points are distributed in a closed ring, which may reflect the periodic changes in the respiratory rate of the patient before surgery due to calm breathing or deep breathing training. Chaotic trajectories: The distribution of points is dense but not repetitive, which may indicate that the respiratory rate of the patient before surgery is affected by nonlinear regulation, such as respiratory disorders caused by preoperative anxiety. Random trajectories: The points are completely irregularly distributed, which may indicate that the preoperative respiratory rate is disturbed by sudden stress or irregular breathing patterns, such as preoperative fear or discomfort. By capturing the distribution patterns and trajectory changes of these points, the dynamic characteristics of each abnormal sign in the multidimensional space can be intuitively reflected, revealing the potential physiological or pathological mechanisms, and providing more comprehensive data support for preoperative risk assessment.The characteristics and distribution differences of trajectories can help identify specific abnormal patterns and provide guidance for the formulation of subsequent intervention measures.

[0025] See also Figure 5 , the specific steps for obtaining the patient's vital signs status analysis results are: S221: Based on the analysis results of abnormal vital signs trajectory changes, the formula is used: Calculation of Lyapunov exponents for abnormal physical signs time series data , integrate to get the Lyapunov index set; in, It is the sampling duration of abnormal vital signs time series data in multidimensional phase space, which is determined by the recording interval and accumulation duration of monitoring data. It is the total number of abnormal signs of abnormal sign time series data in multidimensional phase space. The number of signs is determined through the analysis process, such as intra-abdominal pressure, blood pressure, heart rate, and respiratory rate. is the first order of abnormal physical sign time series data in multidimensional phase space. The weight of abnormal signs is calculated as follows: , It is The standard deviation of abnormal signs, Yes all The sum of the standard deviations of the abnormal signs is used for normalization. By setting the weights, the contribution of each abnormal sign to the final result can be adjusted according to its standard deviation when calculating the Lyapunov index. For example, the larger the standard deviation of the sign, the stronger its volatility in the time series, and the weight value will be increased accordingly, giving it a higher influence in the index calculation. The weight setting can ensure that there is a reasonable comparison and proportional relationship between different signs, thereby improving the reliability of the analysis results. is the traversal variable of the summation symbol, In the multidimensional phase space, Abnormal signs during data sampling time The Euclidean distance represents the state change of the sign in the time series, and the calculation formula is: , Based on the data sampling time When Abnormal physical sign values ​​of dimensional Based on the initial data sampling duration At that time, Abnormal values ​​of the dimensional body are determined by initial measurements, such as resting intra-abdominal pressure, is the number of dimensions of abnormal vital signs data. If it is a single vital sign such as heart rate or respiratory rate, the dimension is 1. In the multidimensional phase space, Abnormal signs in the initial data sampling time The Euclidean distance when is the natural logarithm, which is used to scale and linearize the ratio of distance changes in multidimensional phase space, Is the limit operator, indicating that the data sampling time When it approaches infinity, the stable state of abnormal signs is approximated using the maximum data sampling time.

[0026] Assume that the status changes of the four types of abnormal signs monitored from the initial 0 seconds to the maximum data sampling time of 10 seconds are as follows: intra-abdominal pressure: baseline value 10mmHg, 15mmHg after 10 seconds, blood pressure: baseline value 120mmHg, 135mmHg after 10 seconds, heart rate: baseline value 80bpm, 100bpm after 10 seconds, respiratory rate: baseline value 16 times / min, 15 times / min after 10 seconds, and the standard deviations are: intra-abdominal pressure ,blood pressure , heart rate , respiratory rate .

[0027] For calculation of Lyapunov exponent of intra-abdominal pressure: For weight , the logarithm of the distance change ratio: , Lyapunov exponent: .

[0028] Calculation of the Lyapunov exponent for blood pressure: For weight , the logarithm of the distance change ratio: , Lyapunov exponent: .

[0029] For the calculation of Lyapunov exponent of heart rate: For weight , the logarithm of the distance change ratio: , Lyapunov exponent: .

[0030] For the calculation of the Lyapunov exponent of the respiratory rate: For weight , the logarithm of the distance change ratio: , Lyapunov exponent: .

[0031] The results showed that the intra-abdominal pressure index , blood pressure index , heart rate index , respiratory rate index .

[0032] S222: Based on the Lyapunov exponent set, analyze the sensitivity of the patient's abnormal vital signs time series data to change in the multidimensional phase space, and generate the patient's vital signs status analysis results; The delayed embedding method is used to expand the patient's vital sign data, such as heart rate, blood pressure, etc., into an independent multidimensional phase space, where each point represents the patient's vital sign status at a specific time delay. This method allows the observation of changes in patient vital signs over time, thereby identifying possible risk patterns. By calculating the distribution and changes of these data points in the multidimensional phase space, the Lyapunov exponent can be used to quantify the characteristics of these changes. Specifically, the Lyapunov exponent helps determine the sensitivity of changes in vital sign data, that is, how small initial differences expand over time, which is crucial for predicting possible complications during surgery.

[0033] See also Figure 6 ,The specific steps for obtaining the preoperative patient status record are: S311: Based on the patient's vital sign status analysis results, the Lyapunov index of the abnormal vital sign time series data is classified into states, the abnormal vital sign time series data with a positive Lyapunov index is marked as having unstable characteristics, and the corresponding time window in the abnormal vital sign time series data is marked to generate potential risk time window information; The Lyapunov index of abnormal vital signs time series data was divided into states. According to the calculation results, the Lyapunov index of intra-abdominal pressure was 0.012, which was a positive value, indicating that the intra-abdominal pressure was easily disturbed by external factors in the preoperative stage and fluctuated greatly, making it difficult to maintain a stable state. For example, the intra-abdominal pressure may increase or become unstable due to the patient's preoperative tension, position adjustment, or the influence of preoperative drugs. The Lyapunov index of heart rate was 0.0047, which was also a positive value, indicating that the patient's heart rate may fluctuate rapidly before surgery, and it is easy to cause irregular changes in heart rate due to anxiety or stimulation. The Lyapunov index of blood pressure was 0.003, which was a positive value, indicating that the patient's blood pressure may change significantly in the preoperative stage due to emotional fluctuations or interventions in preoperative preparations. Therefore, the above-mentioned vital signs time series were marked as unstable characteristics. These corresponding time windows were recorded as potential risk time windows, and it was recommended to focus on monitoring the changes in intra-abdominal pressure, heart rate, and blood pressure before surgery so that timely intervention measures could be taken.

[0034] S312: For abnormal vital sign time series data with negative Lyapunov index, mark it as stable state, integrate potential risk time window information, and generate preoperative patient status record; For abnormal vital sign time series data with negative Lyapunov index, after analyzing its change characteristics, it is marked as a stable state. For example, the Lyapunov exponent of the respiratory rate is -0.0014, which is a negative value. This indicates that the patient's respiratory rate fluctuates less in the preoperative stage and the overall change tends to be stable, and it is not easy to cause significant deviations due to external interference or environmental changes. Specifically, this may be related to the patient's good preoperative respiratory regulation or effective preoperative intervention. By marking the respiratory rate as a stable state, the resource investment in its preoperative dynamic monitoring can be reduced. Finally, these analysis results are integrated to generate a record of the patient's preoperative status, providing clear data support for preoperative preparation and surgical planning.

[0035] See also Figure 7 , the steps for obtaining the risk probability distribution data set of intraoperative complications are as follows: S321: Collect the intraoperative complications that have occurred in the historical cases of laparoscopic surgery from the big data platform, and use the formula based on the potential risk time window information in the preoperative patient status record: Calculate the abnormal sign fluctuation conditions selected in the potential risk time window information (such as heart rate, blood pressure fluctuations), complications (such as low blood pressure or abnormal heart rate) , generate prediction results of abnormal physical sign fluctuations and intraoperative complication probability; in, Indicates specific complications that have occurred during laparoscopic surgery from historical cases Abnormal physical sign fluctuation conditions in potential risk time window information in the event of occurrence The probability of occurrence of complications is calculated by recording historical intraoperative complications, and the frequency of corresponding physical sign fluctuations when complications occur is counted. For example, the proportion of abnormal heart rate fluctuations when hypotension occurred in previous surgeries is counted. Indicates the intraoperative complications that have occurred obtained from the historical cases of laparoscopic surgery The overall probability of occurrence can be directly obtained by counting the intraoperative complication records. For example, if the number of hypotension occurrences in 100 laparoscopic surgeries is 30, then =30%, Indicates abnormal physical sign fluctuation conditions in potential risk time window information The overall probability of the abnormal heart rate fluctuation is calculated by the frequency of occurrence of the physical sign fluctuation values ​​in the preoperative physical sign records. For example, the time proportion of abnormal heart rate fluctuation is analyzed. Assuming that the total time is 100 minutes and the abnormal heart rate fluctuation lasts for 15 minutes, then .

[0036] For the probability of association between intra-abdominal pressure fluctuation and complications, it is assumed that the parameter : In 100 surgeries, 20 cases had intraoperative hypotension. : In 20 cases of intraoperative hypotension, the intra-abdominal pressure fluctuations exceeded the normal range in 12 cases. : Among all surgeries, 30 cases had abnormal fluctuations in intra-abdominal pressure. Calculation process:

[0037] The results showed that the probability of intraoperative hypotension was 40% under conditions of intra-abdominal pressure fluctuations.

[0038] For the probability of association between heart rate fluctuation and complications, assume that the parameter : In 100 surgeries, 25 cases had arrhythmias during the operation. : Among the 25 cases of arrhythmia, 17 had abnormal heart rate fluctuations. : Among all surgeries, there were 40 cases of abnormal heart rate fluctuations. The calculation process was:

[0039] The results showed that the probability of arrhythmia under heart rate fluctuation conditions was 43.75%.

[0040] For the probability of blood pressure fluctuations being associated with complications, assume that the parameter : In 100 surgeries, 15 cases had intraoperative hypertension. : Among the 15 cases of intraoperative hypertension, 7.5 cases had blood pressure fluctuations outside the normal range (values ​​were taken proportionally based on statistical data). : Among all surgeries, there were 20 cases with abnormal blood pressure fluctuations. The calculation process was: Results: Under the condition of blood pressure fluctuation, the probability of intraoperative hypertension was 37.5%.

[0041] S322: comparing the preset probability threshold of the correlation strength with the abnormal physical sign fluctuation and the intraoperative complication probability prediction result, analyzing the correlation between the abnormal physical sign fluctuation and the historical intraoperative complication data, and generating a risk probability distribution data set of the intraoperative complication; Based on the calculated probability results, the correlation between physical sign fluctuations and intraoperative complications within the potential risk time window was analyzed. It is known that the probability of abnormal intra-abdominal pressure fluctuations is 40%, the probability of abnormal heart rate fluctuations is 43.75%, and the probability of abnormal blood pressure fluctuations is 37.5%. The probability threshold of association strength is set to 40%, and the probability threshold of association strength is exceeded as a high-intensity association. On this basis, both abnormal intra-abdominal pressure fluctuations and abnormal heart rate fluctuations exceed the probability threshold of association strength and are judged to be high-intensity associations. For abnormal intra-abdominal pressure fluctuations, the abnormal fluctuation amplitude and frequency are statistically analyzed with the time period of intraoperative hypotension, and the conditional probability formula is used to quantify its impact on intraoperative hypotension; for abnormal heart rate fluctuations, the conditional probability relationship between abnormal heart rate fluctuations and intraoperative arrhythmia is established by analyzing the fluctuation characteristics within the time window, and the strength of the association is quantified. For abnormal blood pressure fluctuations, the probability is 37.5%, which is lower than the probability threshold of association strength, so it is not listed as a high-intensity association for the time being, but its fluctuation characteristic data are still retained for the preliminary analysis of the conditional probability relationship between abnormal blood pressure fluctuations and intraoperative hypertension. Ultimately, an analysis report on the intensity of the correlation between physical sign fluctuations and intraoperative complications within the potential risk time window is generated, and key monitoring recommendations for high-intensity correlated signs are clarified to provide support for intraoperative management and preoperative intervention.

[0042] See also Figure 8 ,The specific steps for constructing the laparoscopic risk prediction model are: S411: Organize the risk probability distribution data set of intraoperative complications into conditional input data for Bayesian inference, analyze the risk of each intraoperative complication under the current abnormal sign fluctuation conditions, and establish a laparoscopic risk prediction model; According to the risk probability distribution data set of intraoperative complications, a laparoscopic risk prediction model was constructed by the Bayesian inference method. First, the collected risk probability distribution data were sorted according to the intraoperative complication category and the characteristics of physical sign fluctuations, and the abnormal physical sign fluctuations generated in the potential risk time window were combined with the intraoperative complication probability prediction results. These probability results were used as conditional input data for Bayesian inference, and Bayesian inference was performed using statistical analysis software (such as Python's scikit-learn library or R language's Bayes package). The specific operation is to use the patient's real-time physical sign fluctuation data (such as dynamic monitoring values ​​of heart rate, blood pressure, and intra-abdominal pressure) as observation data, combined with the prior probability of each intraoperative complication in historical cases, and analyze the risk of each intraoperative complication under current conditions by updating the posterior probability distribution. For example, the predicted probability results of abnormal heart rate and arrhythmia are directly used, combined with the prior probability of arrhythmia during surgery, to infer the posterior probability value of arrhythmia in the current patient. By analyzing the posterior probabilities of all intraoperative complications and matching them with the set risk level threshold, a laparoscopic risk prediction model was finally established.

[0043] S412: Based on the laparoscopic risk prediction model, the abnormal preoperative physical sign fluctuation data of the patient pre-input into the laparoscopic risk prediction model is combined with the intraoperative complication events that have occurred in the historical cases of laparoscopic surgery, and the posterior probability and risk level of the intraoperative complications of the current patient are output to generate the output result of the laparoscopic risk prediction model; Based on the preoperative patient risk level label output by the laparoscopic risk prediction model, the model output includes the patient's specific intraoperative complication probability (such as a 45% probability of arrhythmia, a 70% probability of hypotension, and a 25% probability of hypertension) and the overall risk level label (such as high risk). The model combines the input patient's preoperative physical sign fluctuation data (such as abnormal heart rate amplitude, intra-abdominal pressure fluctuation time, and blood pressure fluctuation peak) with historical data. The historical data includes the patient's intraoperative complications recorded in laparoscopic surgery cases and their related preoperative physical sign fluctuation characteristics, and outputs the posterior probability of each complication and the corresponding risk level. For example, for a patient, the model may output: the probability of hypotension is 70%, which belongs to the high-risk label; the probability of arrhythmia is 45%, which belongs to the medium-risk label; the probability of hypertension is 25%, which belongs to the low-risk label. Based on these data, all patients are classified according to risk levels. For example, below 30% is low risk, 30%-60% is medium risk, and above 60% is high risk. Further analysis is then conducted based on the physical sign fluctuation characteristics of high-risk patients. Through correlation analysis, key intervention measures that require preoperative intervention are determined, and a comprehensive label report covering patients at different risk levels is generated, providing clear support for doctors to formulate preoperative interventions and optimize surgical plans.

[0044] A laparoscopic risk prediction model based on big data is obtained by a method for constructing a laparoscopic risk prediction model based on big data.

[0045] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for constructing a laparoscopic risk prediction model based on big data, characterized in that: The following steps are involved: S1: Collect patient vital sign data before laparoscopic surgery from the big data platform, perform preprocessing operations on the patient vital sign data and convert them into vital sign time series data, divide the vital sign time series data into multiple time windows of fixed length, obtain the statistical features of the vital sign time series data in each time window, judge the abnormal time points in the time window according to the preset statistical feature threshold, and generate abnormal vital sign time series data; S2: Expanding the abnormal vital sign time series data of different categories into independent multidimensional phase spaces respectively, by capturing the change of the motion trajectory of each category of abnormal vital sign time series data in the independent multidimensional phase space, calculating the Lyapunov exponent of the abnormal vital sign time series data, quantifying the change characteristics of the abnormal vital sign state of the patient in the independent multidimensional phase space, and generating the patient's vital sign state analysis results; S3: Based on the patient's vital sign status analysis results, the Lyapunov index of the abnormal vital sign time series data is divided into state stability, and the time window of the abnormal vital sign time series data with a positive Lyapunov index is marked as a potential risk time window, and a preoperative patient status record is generated; S4: Collect intraoperative complications that have occurred in historical laparoscopic surgery cases from the big data platform, extract data marked as potential risk time windows based on the preoperative patient status records, analyze the correlation between physical sign fluctuations and intraoperative complications within the potential risk time window, and obtain a risk probability distribution data set for intraoperative complications. Based on the risk probability distribution data set for intraoperative complications, a laparoscopic risk prediction model is constructed using the Bayesian inference method. The laparoscopic risk prediction model is used to output risk level labels for preoperative patients.

2. The method for constructing a laparoscopic risk prediction model based on big data according to claim 1, characterized in that: The steps for obtaining the statistical features of the vital sign time series data in each time window are specifically as follows: S111: By accessing the big data platform, extracting the patient's vital sign data before laparoscopic surgery from the database of the big data platform, including intra-abdominal pressure, blood pressure, heart rate and respiratory rate, cleaning and standardizing the patient's vital sign data and sorting them according to timestamps to generate vital sign time series data; S112: Divide the vital sign time series data into multiple time windows according to fixed lengths, extract statistical features of the vital sign time series data in each time window, including mean, variance, skewness and kurtosis, and generate a time window feature data set.

3. The method for constructing a laparoscopic risk prediction model based on big data according to claim 2, characterized in that: The steps for acquiring the abnormal physical sign time series data are specifically as follows: S121: comparing the statistical features of the vital sign time series data in the time window feature data set with a preset statistical feature threshold, and when the statistical features of any vital sign time series data in the time window exceed the statistical feature threshold range, marking the time point in the time window as an abnormal time point, and generating an abnormal time point analysis result; S122: Based on the abnormal time point analysis result, the marked abnormal time point information, including the abnormal patient vital sign data name and the time range of the abnormal time point, is integrated to generate abnormal vital sign time series data.

4. The method for constructing a laparoscopic risk prediction model based on big data according to claim 1, characterized in that: The steps for capturing the changes in the motion trajectory of each type of abnormal vital sign time series data in an independent multidimensional phase space are specifically as follows: S211: expanding the abnormal vital sign time series data into a multidimensional phase space by a delay embedding method, selecting a time delay in the multidimensional phase space and determining an embedding dimension of the multidimensional phase space, thereby generating multidimensional phase space data of the abnormal vital sign; S212: Analyze the changes in the motion trajectories of the abnormal signs in the multidimensional phase space data of the abnormal signs, and divide the changes in the motion trajectories of the abnormal signs into multiple dynamic modes, reveal the dynamic characteristics of the abnormal signs in the multidimensional phase space, and generate abnormal sign trajectory change analysis results.

5. The method for constructing a laparoscopic risk prediction model based on big data according to claim 4, characterized in that: The steps for obtaining the patient's physical sign status analysis results are specifically as follows: S221: Based on the abnormal vital sign trajectory change analysis result, the formula is used: Calculation of Lyapunov exponents for abnormal physical signs time series data , integrate to get the Lyapunov index set; in, is the sampling duration of abnormal physical sign time series data in the multidimensional phase space, is the total number of abnormal signs in the abnormal sign time series data in the multidimensional phase space, It is the first The weight of abnormal signs, is the first order of abnormal physical sign time series data in multidimensional phase space. Abnormal signs during data sampling time The Euclidean distance of is the first order of abnormal physical sign time series data in multidimensional phase space. Abnormal signs in the initial data sampling time The Euclidean distance when is the natural logarithm, which is used to scale and linearize the ratio of distance changes in multidimensional phase space; S222: Based on the Lyapunov exponent set, analyze the sensitivity of changes in the patient's abnormal vital signs time series data in the multidimensional phase space to generate a patient vital signs status analysis result.

6. The method for constructing a laparoscopic risk prediction model based on big data according to claim 1, characterized in that: The steps for obtaining the preoperative patient status record are specifically as follows: S311: Based on the patient's vital sign status analysis result, the Lyapunov index of the abnormal vital sign time series data is classified into states, the abnormal vital sign time series data with a positive Lyapunov index is marked as having unstable characteristics, and the corresponding time window in the abnormal vital sign time series data is marked to generate potential risk time window information; S312: The abnormal vital sign time series data with a negative Lyapunov index is marked as a stable state, and the potential risk time window information is integrated to generate a preoperative patient status record.

7. The method for constructing a laparoscopic risk prediction model based on big data according to claim 1, characterized in that: The steps for obtaining the risk probability distribution data set of intraoperative complications are specifically as follows: S321: Collect the intraoperative complications that have occurred in the historical cases of laparoscopic surgery from the big data platform, and based on the potential risk time window information in the preoperative patient status record, use the formula: Calculate the abnormal sign fluctuation conditions selected in the potential risk time window information Next, complications Probability of occurrence , generate prediction results of abnormal physical sign fluctuations and intraoperative complication probability; in, Indicates specific complications that have occurred during laparoscopic surgery from historical cases Abnormal physical sign fluctuation conditions in potential risk time window information in the event of occurrence The probability of occurrence, Indicates the intraoperative complications that have occurred obtained from the historical cases of laparoscopic surgery The overall probability of occurrence, Indicates abnormal physical sign fluctuation conditions in potential risk time window information The overall probability of S322: Compare the preset probability threshold of the correlation strength with the abnormal physical sign fluctuation and the prediction result of the intraoperative complication probability, analyze the correlation between the abnormal physical sign fluctuation and the historical intraoperative complication data, and generate a risk probability distribution data set of intraoperative complications.

8. The method for constructing a laparoscopic risk prediction model based on big data according to claim 1, characterized in that: The steps for constructing the laparoscopic risk prediction model are specifically as follows: S411: organizing the risk probability distribution data set of the intraoperative complications into conditional input data for Bayesian inference, analyzing the occurrence risk of each intraoperative complication under the current abnormal physical sign fluctuation condition, and establishing a laparoscopic risk prediction model; S412: Based on the laparoscopic risk prediction model, the patient's preoperative abnormal physical sign fluctuation data pre-input into the laparoscopic risk prediction model is combined with the intraoperative complication events that have occurred in the historical cases of laparoscopic surgery, and the posterior probability and risk level of the current patient's intraoperative complications are output to generate the output result of the laparoscopic risk prediction model.

9. Laparoscopic risk prediction model based on big data, characterized by: Obtained by the method for constructing a laparoscopic risk prediction model based on big data as described in any one of claims 1-8.

Images