A method, system, storage medium, and program product for early warning of hysteroscopic surgery.

Abstract

A method, system, storage medium, and program product for early warning of hysteroscopic surgery, relating to the field of electronic digital data processing, are disclosed. The method includes: acquiring historical surgical data for hysteroscopic surgery; the historical surgical data includes historical sensor data, patient data, surgical parameters, and surgical anomaly annotations; based on the surgical anomaly annotations, the historical surgical data is segmented temporally to construct a temporal risk dataset; based on the temporal risk dataset, the correlation between data sequences is established to construct an abnormal event knowledge graph; risk path features are extracted from the abnormal event knowledge graph to obtain risk warning conditions, and an abnormal event detection model is constructed; real-time monitoring data of hysteroscopic surgery is input into the abnormal event detection model according to time windows to obtain abnormal detection results and result confidence levels; the surgical risk level is determined, and a surgical warning is issued based on the surgical risk level. Implementing this application enables more accurate abnormal warnings in hysteroscopic surgery.

Description

Technical Field

[0001] This application relates to the field of electronic digital data processing, and in particular to a method, system, storage medium, and program product for early warning of hysteroscopic surgery. Background Technology

[0002] In recent years, with the rapid development of minimally invasive surgical techniques, hysteroscopic surgery has become the preferred treatment for gynecological diseases such as endometrial polyps, uterine fibroids, and intrauterine adhesions due to its advantages of minimal trauma and rapid recovery. During hysteroscopic surgery, in order to obtain a clear surgical field and establish surgical space, it is necessary to inject distending fluid into the uterine cavity, and to monitor and control the pressure of the distending fluid in real time. This is of great significance for ensuring the safety and effectiveness of the surgery.

[0003] In related technologies, the monitoring and early warning methods used by medical institutions mainly involve recording the infusion and recovery volume of distension fluid through a fluid monitoring system, calculating the difference to estimate the amount of fluid absorbed by the patient; medical staff observe changes in the patient's vital signs and combine this with experience to judge whether there is a risk of complications; and an early warning system based on a fixed threshold is established to issue an alarm when the fluid absorption volume reaches the preset threshold.

[0004] However, fixed-threshold warning methods are difficult to adapt to individual differences among patients, which may lead to untimely warnings or false alarms. Summary of the Invention

[0005] This application provides a method, system, storage medium, and program product for early warning of hysteroscopic surgery, which can achieve more accurate early warning of abnormalities during hysteroscopic surgery.

[0006] Firstly, this application provides an early warning method for hysteroscopic surgery, applied to a surgical early warning system. The method includes: acquiring historical surgical data for hysteroscopic surgery; the historical surgical data includes sensor data, patient data, surgical parameters, and surgical anomaly annotations from various historical surgical records; the sensor data includes uterine cavity image data, distension fluid irrigation difference data, uterine cavity size data, and uterine cavity pressure values; the surgical parameters include anesthesia type, anesthetic drug data, and distension fluid data; based on surgical anomaly annotations, the historical surgical data is segmented temporally to construct a temporal risk dataset; based on the temporal risk dataset, the correlation between data sequences is established to construct an abnormal event knowledge graph; risk path features are extracted from the abnormal event knowledge graph to obtain risk warning conditions, and a model is trained based on these conditions to obtain an abnormal event detection model; real-time monitoring data of hysteroscopic surgery is input into the abnormal event detection model according to a time window to obtain abnormal detection results and result confidence levels; the surgical risk level is determined based on the abnormal detection results and result confidence levels, and a surgical early warning is issued according to the surgical risk level.

[0007] In the above embodiments, the surgical early warning system acquires historical surgical data and performs time-series segmentation based on anomaly annotations to establish an abnormal event knowledge graph, extracts risk path features for model training, and ultimately achieves real-time early warning of surgical risks. It performs correlation analysis on multi-dimensional surgical data and establishes the correlation between data sequences, making risk warnings more accurate and avoiding the limitations of traditional fixed threshold early warning methods.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, the step of constructing a time-series risk dataset by temporally segmenting historical surgical data based on surgical anomaly annotation specifically includes: using the anomaly time of the surgical anomaly annotation as a reference point, tracing back a preset time period to obtain historical surgical data within the preset time period as risk evolution data; segmenting the risk evolution data according to time intervals, marking a preset number of data segments before the anomaly occurrence time as high-risk intervals, and marking other data segments as progressive risk intervals of different levels according to the time distance to the anomaly occurrence time; extracting data features of the risk evolution data within each risk interval; the data features include mean, standard deviation, trend, mutation point, and correlation coefficient; and constructing a time-series risk dataset containing risk level annotations based on the data features.

[0009] In the above embodiments, the surgical early warning system traces historical data for a preset duration based on abnormal moments, segments and marks different risk intervals according to time intervals, extracts data features to construct a time-series risk dataset, and realizes dynamic assessment of surgical risks through fine-grained analysis of the risk evolution process, enabling the early warning system to identify potential risks earlier.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, the steps of extracting risk path features from the abnormal event knowledge graph to obtain risk warning conditions, and training a model based on the risk warning conditions to obtain an abnormal event detection model, specifically include: extracting entity combinations with a correlation degree higher than a preset correlation threshold based on the entity association strength in the abnormal event knowledge graph; determining the change patterns of each entity indicator in the entity combination in different risk intervals; the change patterns include change rate, acceleration, and collaborative change features; establishing a risk path feature library based on the change patterns, and binding risk levels to the risk path features in the risk path feature library; and training the risk path features in conjunction with the risk levels to obtain the abnormal event detection model.

[0011] In the above embodiments, the surgical early warning system extracts highly associated entity combinations based on the entity association strength in the knowledge graph, analyzes their changing patterns in different risk ranges, establishes and trains an abnormal event detection model, and makes full use of the association relationships between multidimensional data to improve the accuracy and timeliness of early warning.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, the step of establishing a risk path feature library based on change patterns and binding risk levels to risk path features in the risk path feature library specifically includes: performing time-series clustering analysis on change patterns to classify change patterns with the same change trend into the same risk path feature; calculating the association probability between each risk path feature and different risk levels based on the risk level labels in the time-series risk dataset; binding risk levels with association probabilities higher than a preset probability threshold to the corresponding risk path features; setting risk weight coefficients for each risk path feature based on the change rate, acceleration, and collaborative change characteristics of the risk path features; and adjusting the risk levels of the risk path features according to the risk weight coefficients to obtain the risk path feature library.

[0013] In the above embodiments, the surgical early warning system performs time-series cluster analysis on the change pattern, calculates the correlation probability between risk path characteristics and risk level, and sets weight coefficients for adjustment, thereby achieving accurate quantification and classification of risk path characteristics and improving the accuracy of early warning.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, prior to the step of acquiring historical surgical data of hysteroscopic surgery, the method further includes: grouping and labeling historical surgical records based on surgical complexity, patient risk level, and surgical outcome to obtain multiple convergent surgical groups; performing time-series feature analysis on surgical records within each convergent surgical group to generate typical risk evolution patterns and key early warning indicators; and establishing a differentiated risk assessment library containing multiple convergent surgical groups based on typical risk evolution patterns and key early warning indicators.

[0015] In the above embodiments, the surgical early warning system groups historical surgical records based on factors such as surgical complexity, generates typical risk evolution patterns and key early warning indicators, establishes a differentiated risk assessment library, and realizes personalized risk assessment for different types of surgeries.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, after establishing a differentiated risk assessment library containing multiple convergent surgical groups based on typical risk evolution patterns and key early warning indicators, the method further includes: obtaining basic information and surgical plan information of the patient to be operated on; the basic information includes age, past medical history, organ function indicators, and surgical contraindications; the surgical plan information includes surgical type, surgical difficulty level, and estimated surgical duration; matching corresponding convergent surgical groups from the differentiated risk assessment library based on the basic information and surgical plan information; extracting corresponding typical risk evolution patterns and key early warning indicators from the matched convergent surgical groups; extracting risk factors from the basic information, and adjusting the risk thresholds in the typical risk evolution patterns based on the risk factors; inputting the adjusted risk thresholds and key early warning indicators into an abnormal event detection model to obtain personalized detection results.

[0017] In the above embodiments, the surgical early warning system acquires the basic information and surgical plan information of the patient to be operated on, matches the corresponding convergent surgical group, adjusts the risk threshold to obtain personalized detection results, and achieves precise adaptation to individual patient differences.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, after determining the surgical risk level based on the anomaly detection results and the confidence level of the results, and issuing a surgical warning based on the surgical risk level, the method further includes: recording all warning events during the surgical process to obtain an intraoperative warning record; the warning events include the warning time, the combination of indicators that triggered the warning, the risk level, and the doctor's handling measures; after the surgery, based on the intraoperative monitoring data and the intraoperative warning record, analyzing the accuracy rate, false alarm rate, and false negative rate of the warning to obtain a warning assessment result; and adjusting the parameters of the abnormal event detection model according to the warning assessment result.

[0019] In the above embodiments, the surgical early warning system records intraoperative early warning events, analyzes indicators such as the accuracy of early warnings, and adjusts model parameters, thereby realizing the self-optimization and continuous improvement of the early warning system.

[0020] In a second aspect, embodiments of this application provide a surgical early warning system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the surgical early warning system to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a surgical early warning system, cause the surgical early warning system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a surgical early warning system, cause the surgical early warning system to perform the method described in the first aspect and any possible implementation thereof.

[0023] Understandably, the surgical early warning system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. By employing a method that constructs a time-series risk dataset and anomaly event knowledge graph based on historical surgical data, and analyzing and issuing early warnings for real-time monitoring data, it is possible to capture surgical risk characteristics from multiple dimensions and establish correlations between data sequences. This effectively solves the limitations of existing technologies that rely solely on a single indicator or fixed threshold for early warning, thereby achieving more accurate surgical risk early warnings.

[0025] 2. By employing time-series cluster analysis of change patterns to calculate the correlation probability between risk path characteristics and risk levels, and adjusting the weight coefficients, it is possible to accurately quantify and classify risk path characteristics. This effectively solves the problems of coarse risk level classification and lack of precise quantitative standards in existing technologies, thereby achieving a more scientific and accurate risk assessment.

[0026] 3. By adopting a method that matches patients' basic information and surgical plan information to convergent surgical groups and makes personalized adjustments to risk thresholds, it can provide accurate early warnings based on the individual characteristics of different patients. This effectively solves the problem that the existing technology has a uniform early warning standard and cannot adapt to individual differences, thus realizing personalized risk warnings. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating an early warning method for hysteroscopic surgery in an embodiment of this application; Figure 2 This is another flowchart illustrating the early warning method for hysteroscopic surgery in the embodiments of this application; Figure 3 This is a schematic diagram of the physical device structure of a surgical early warning system in the embodiments of this application. Detailed Implementation

[0028] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0029] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0030] To facilitate understanding, the application scenarios of the embodiments of this application are described below.

[0031] At the gynecology center of a top-tier hospital, approximately 200 hysteroscopic surgeries are performed monthly. During one procedure, a 58-year-old patient underwent a myomectomy, during which distending fluid was used to establish the surgical field. About 40 minutes into the procedure, the patient suddenly experienced a drop in blood pressure and an increased heart rate. Examination revealed that this was due to excessive absorption of the distending fluid. Although timely intervention prevented serious consequences, such situations are not uncommon in hysteroscopic surgeries. Traditional monitoring methods rely primarily on the experience and judgment of medical staff, making it difficult to detect potential risks promptly. Especially under conditions of high surgical volume and limited medical resources, intelligent early warning systems are essential.

[0032] In related technologies, surgical risk warnings can be achieved by employing a fixed threshold monitoring and early warning system. This system assesses risk by setting standardized parameters such as a threshold for uterine distension fluid absorption and a range of vital sign fluctuations, and issues an early warning signal when the threshold is reached. The following describes a scenario where this early warning method for hysteroscopic surgery is used.

[0033] The hospital's original early warning system primarily used a fixed threshold (e.g., 1000ml) for the difference between the infusion and recovery volumes of distension fluid. In one surgery, the same warning threshold was used for a 50kg patient and a 75kg patient, resulting in insufficient timely warnings for the lighter patient and potentially excessive warnings for the heavier patient. Furthermore, the system focused on only a single indicator, failing to consider individual differences such as age and underlying medical conditions, and neglecting factors like surgical procedures and anesthesia methods, leading to insufficient accuracy and timeliness in the warnings.

[0034] The early warning method for hysteroscopic surgery described in this application constructs a multi-dimensional abnormal event knowledge graph, combines time-series risk data analysis and personalized threshold adjustment, and achieves more accurate risk warnings. This not only adapts to individual differences among patients but also identifies potential risks in advance, providing timely decision support for doctors. The following describes scenarios where the early warning method for hysteroscopic surgery described in this application is used.

[0035] After adopting the early warning system described in this application, the system first analyzed data from 3,000 hysteroscopic surgeries performed at the hospital over the past five years, establishing a multi-dimensional knowledge graph including surgical type, patient characteristics, and surgical parameters. In a new surgery, the system collects various patient indicators in real time and automatically adjusts the early warning threshold based on the patient's individual characteristics (such as age 65 and history of hypertension). When the system detects an accelerated absorption rate of the distension fluid accompanied by slight fluctuations in blood pressure, it issues an early warning, enabling doctors to adjust their surgical strategies promptly and effectively prevent complications.

[0036] As can be seen, the hysteroscopic surgery early warning method in this application embodiment can not only achieve surgical risk early warning, but also effectively solve the problems of insufficient accuracy and low personalization of traditional fixed threshold early warning methods, thereby achieving more intelligent and accurate surgical early warning.

[0037] To facilitate understanding, the method provided in this implementation will be described in detail below, using the above scenario as an example. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating an early warning method for hysteroscopic surgery in an embodiment of this application.

[0038] S101. Obtain historical surgical data for hysteroscopic surgery.

[0039] Historical surgical data refers to the complete collection of past hysteroscopic surgical records stored in the surgical early warning system, including sensor data, patient data, surgical parameters, and surgical anomaly annotations. Sensor data represents real-time data collected by various monitoring devices during the procedure, including surgical field images acquired via hysteroscopy, the volume difference between the injection and recovery of distending fluid, uterine cavity size parameters measured by ultrasound and other equipment, and uterine cavity pressure monitoring values. Patient data refers to basic patient information and clinical indicators related to the surgery. Surgical parameters represent key control indicators during the procedure, such as the anesthesia method used, the dosage and timing of anesthetic drugs, and the type and concentration of distending fluid. Surgical anomaly annotations represent the physician's records and classifications of various abnormal situations that occurred during the procedure.

[0040] Before building its early warning model, the surgical early warning system first needs to acquire sufficient historical data as a training foundation. Specifically, the system connects to the hospital's surgical database to extract complete records of past hysteroscopic surgeries, including sensor data collected throughout the procedure, patient clinical information, surgical parameters, and any abnormal events recorded by the physician. The system preprocesses the acquired data, including data cleaning, format standardization, and outlier handling, to ensure data quality and usability. Simultaneously, the system standardizes the data to make it comparable across different sources and time periods.

[0041] In some embodiments, historical surgical data acquisition and preprocessing can be achieved in multiple ways: Optionally, the system can directly acquire data through a hospital information system (HIS) interface, parse and store the data according to a preset data template, perform quality assessment and screening, and finally integrate qualified datasets into the database of the early warning system; Optionally, the system can import locally stored surgical record files in batches, extract key information using natural language processing technology, perform structured processing on the extracted information, and then perform data verification and completion. It is understood that other methods can also be used to acquire and preprocess historical surgical data, such as real-time synchronization through a remote data interface or distributed data acquisition, etc., which are not limited here.

[0042] S102. Based on surgical anomaly annotation, historical surgical data is segmented into time series to construct a time series risk dataset.

[0043] In this context, temporal segmentation refers to dividing continuous surgical data into multiple interconnected data segments according to the time dimension. The temporal risk dataset represents a structured collection of data with risk level labels after segmentation, where each data segment carries a corresponding timestamp and risk marker.

[0044] After acquiring historical surgical data, the surgical early warning system needs to perform time-series analysis and risk labeling. Specifically, the system uses the time of the abnormal event as a baseline and traces back a preset time length to extract all monitoring data within that period. The extracted data is then segmented according to fixed time intervals, and the data characteristics of each time period are extracted and analyzed. Based on the time of the abnormal event, the system marks data segments closer to the abnormal event as high-risk intervals, while other data segments are marked as different levels of risk intervals based on their time distance from the abnormal event.

[0045] In some embodiments, the construction of time-series segmentation and risk datasets can be achieved in several ways: Optionally, the system uses a sliding time window approach to segment the data, calculates the statistical characteristics within each window, determines the risk level based on the changing trends of the characteristics, and finally clusters data segments with the same risk characteristics; Optionally, the system uses a time-series pattern mining algorithm to identify key change points in the data, segments the data using these change points as boundaries, analyzes the correlation between each data segment and the abnormal event, and constructs a training set with risk labels. It is understood that other methods can also be used to achieve the construction of time-series segmentation and risk datasets, such as using deep learning algorithms to automatically extract time-series features, etc., which are not limited here.

[0046] S103. Based on the time-series risk dataset, establish the correlation between data sequences and construct an abnormal event knowledge graph.

[0047] Here, a data sequence represents a sequence of numerical values ​​for various monitoring indicators arranged chronologically. Relationships refer to the mutual influence and dependencies between different data sequences. An anomaly event knowledge graph is a structured knowledge network describing the mechanisms of surgical anomalies, comprising three elements: entities, attributes, and relationships.

[0048] After obtaining the time-series risk dataset, the surgical early warning system needs to mine the deep correlations between the data. Specifically, the system analyzes the changing patterns of different monitoring indicators over time, calculating the correlation coefficients and mutual information between the indicators. For combinations of indicators with significant correlations, the system further analyzes their causal relationships and degree of influence. Based on these analytical results, the system constructs a knowledge graph containing entity nodes and relational edges, where nodes represent various monitoring indicators, edges represent the correlations between indicators, and the weights of the edges reflect the strength of the correlations.

[0049] S104. Extract risk path features from the abnormal event knowledge graph to obtain risk warning conditions, and train the model based on the risk warning conditions to obtain an abnormal event detection model.

[0050] Among them, risk path features represent characteristic patterns that demonstrate the development law of abnormal events in the knowledge graph. Risk warning conditions refer to the specific thresholds and combinations of conditions that trigger a warning.

[0051] After constructing a knowledge graph, the surgical early warning system needs to extract feature patterns that can be used for early warning. Specifically, the system analyzes the topological structure and connectivity of each node in the knowledge graph to identify feature combinations that have a significant predictive effect on abnormal events. For each feature combination, the system analyzes its performance characteristics at different risk levels and establishes corresponding early warning rules. Then, the system uses these rules as training objectives and combines them with historical data to train an abnormal event detection model, enabling it to assess surgical risks in real time.

[0052] It should be noted that the abnormal event detection model in this application is trained using historical surgical data as the training set. Input data includes sensor data during surgery (uterine cavity images, distension fluid perfusion difference, uterine cavity size, uterine cavity pressure), patient data (age, past medical history, organ function indicators), and surgical parameters (anesthesia type, medication data, distension fluid data). Training criteria are based on surgical abnormal events annotated by physicians, with supervised learning using risk level annotations obtained through temporal segmentation. The system first extracts risk path features from the knowledge graph, establishes a feature-risk level mapping relationship, and then uses path features with risk level annotations as training samples. Cross-validation is used to evaluate model performance, ensuring the model has good generalization ability. This abnormal event detection model is a deep learning-based temporal classification model, comprising three main modules: feature extraction, risk assessment, and early warning output. The feature extraction module uses a multi-layer neural network to process multi-dimensional input data and extract temporal features; the risk assessment module analyzes the importance weights of features based on an attention mechanism and performs risk inference by combining the relationships in the knowledge graph; the early warning output module uses a softmax classifier to determine the risk level of the current state and outputs a confidence score. In practical applications, this abnormal event detection model can receive real-time monitoring data as input, including all monitoring indicators within the current time window. The abnormal event detection model first performs feature extraction and matching, comparing real-time data with known risk path features, then dynamically adjusts the risk assessment threshold based on individual patient characteristics, and finally outputs the risk level assessment result and confidence score. When the risk level exceeds the preset threshold, the system triggers different levels of early warning mechanisms based on the confidence level, providing doctors with specific risk alerts and intervention suggestions.

[0053] S105. Input the real-time monitoring data of hysteroscopic surgery into the abnormal event detection model according to the time window to obtain the abnormal detection results and the confidence level of the results.

[0054] Real-time monitoring data represents the data stream collected in real time by various sensors during the surgical procedure. A time window is a fixed-length segment used for data analysis. Anomaly detection results represent the model's risk assessment output for the current state. Result confidence represents the model's degree of certainty about the predicted outcome, expressed as a percentage between 0 and 1.

[0055] In actual operation, the surgical early warning system requires continuous analysis of real-time data. Specifically, the system segments the real-time monitoring data according to a preset time window length (e.g., 30 seconds or 1 minute), extracts features and preprocesses the data within each time window to meet the model's input requirements. The processed data is then input into a trained anomaly detection model, which outputs an anomaly risk assessment result for the current state, along with a confidence value for the prediction, indicating the reliability of the prediction.

[0056] In some embodiments, real-time data processing and anomaly detection can be achieved in several ways: Optionally, the system uses a sliding window approach to process the data stream in real time, employs a fast feature extraction algorithm to calculate feature values, and performs parallel computation through a model to achieve low-latency real-time prediction; alternatively, the system establishes a multi-level caching mechanism to store real-time data, uses incremental learning methods to update features, and employs an ensemble prediction strategy to integrate analysis results from multiple time scales. It is understood that other methods can also be used to achieve real-time data processing and anomaly detection, such as using streaming computing frameworks, and are not limited here.

[0057] S106. Determine the surgical risk level based on the abnormality detection results and the confidence level of the results, and issue a surgical warning based on the surgical risk level.

[0058] The surgical risk level represents the graded assessment of the degree of danger in the current surgical state. Anomaly detection results refer to the model's assessment output of various risk factors. Surgical alerts are used to represent risk warning information issued by the system to medical staff, including the alert level, reason, and recommended measures.

[0059] After receiving abnormal detection results, the surgical early warning system needs to make appropriate early warning decisions. Specifically, the system determines the risk level of the current state based on the type and confidence level of the abnormal detection results, referring to preset risk level classification standards. For different risk levels, the system will trigger corresponding early warning mechanisms, including visual cues, audible alarms, or push notifications. The early warning information will include risk source analysis, possible consequences, and suggested intervention measures to help doctors make timely judgments and take appropriate actions.

[0060] In some embodiments, risk level assessment and early warning can be implemented in multiple ways: Optionally, the system uses fuzzy logic to comprehensively evaluate multiple risk factors, establish dynamic early warning thresholds, adaptively adjust early warning strategies based on doctor feedback, and send early warning information through multiple channels; Optionally, the system uses rule-based decision trees to determine risk levels, combines expert knowledge bases to provide treatment suggestions, and uses intelligent push to ensure that key early warning information is delivered to relevant personnel in a timely manner. It is understood that other methods can also be used to achieve risk level assessment and early warning, such as using probabilistic graphical models for risk reasoning, etc., which are not limited here.

[0061] In the above embodiment, a complete risk assessment system was established through in-depth mining and multi-dimensional analysis of historical surgical data. In practical applications, the system can dynamically adjust the early warning strategy based on individual patient characteristics and specific surgical circumstances, achieving precise early warning. The following section supplements the scenario described in this embodiment.

[0062] With continued use, the early warning model is constantly optimized. The system can dynamically adjust its early warning strategy based on the operational characteristics of different surgeons. For example, for doctors who have just completed their residency training, the system provides more detailed risk warnings; for experienced chief physicians, it mainly provides reminders of key risk points. Simultaneously, the system continuously optimizes the accuracy of the early warning model by analyzing postoperative follow-up data. After discovering that a certain type of patient (such as those with cardiac dysfunction) is more sensitive to distending fluid, the system automatically adjusts the risk assessment criteria for that patient group, achieving more precise and personalized early warnings.

[0063] In light of the above scenarios, the method provided in this implementation will now be described in more detail. Please refer to [link / reference]. Figure 2 This is another flowchart illustrating the early warning method for hysteroscopic surgery in this application embodiment.

[0064] S201. Based on the complexity of the surgery, the patient's risk level, and the surgical outcome, historical surgical records are grouped and labeled to obtain multiple convergent surgical groups.

[0065] Surgical complexity represents the level of difficulty in surgical procedures, including factors such as surgical type, operational difficulty, and expected risks. Patient risk level refers to the surgical risk classification based on the patient's baseline condition. Surgical outcome represents the final evaluation of the surgical result, including surgical completion, complication rate, and recovery status. Convergent surgery group represents a set of surgical cases with similar characteristics and risk patterns.

[0066] Before building an early warning model, the surgical early warning system needs to classify and organize historical data. Specifically, the system first assesses the complexity of the surgery based on factors such as surgery type, duration, and difficulty. It then determines the patient's risk level by combining this with information such as patient age, underlying diseases, and organ function, and considers surgical outcomes such as completion status and complication occurrence. Next, a multi-dimensional clustering algorithm is used to group surgical cases with similar characteristics, forming several representative convergent surgical groups, providing a foundation for subsequent risk analysis.

[0067] In some embodiments, surgical records can be grouped and labeled in multiple ways: Optionally, the system uses a hierarchical clustering algorithm to analyze surgical features, calculate the similarity matrix between cases, set a clustering threshold for automatic grouping, and finally have experts confirm the rationality of the grouping results; Optionally, the system uses a decision tree algorithm to perform preliminary grouping according to preset classification rules, calculate the feature consistency of cases within the group, dynamically adjust the grouping threshold, and finally form a stable grouping result. It is understood that other methods can also be used to group and label surgical records, such as using neural networks for feature learning, etc., which are not limited here.

[0068] S202. Perform time-series feature analysis on surgical records within each convergent surgical group to generate typical risk evolution patterns and key early warning indicators.

[0069] Among these, time-series feature analysis represents the study of the changes in various indicators over time during the surgical procedure. Typical risk evolution patterns refer to the risk development processes and characteristic sequences commonly seen in specific surgical types. Key early warning indicators are used to represent core monitoring parameters that are of significant value for risk prediction. Surgical records represent historical cases containing complete surgical procedure data.

[0070] After identifying similar surgical groups, the surgical early warning system needs to conduct in-depth analysis of the risk characteristics within each group. Specifically, the system performs time-series analysis on cases in each similar surgical group, using data mining methods to identify common risk evolution characteristics. The system extracts typical change patterns of various monitoring indicators during the risk evolution process, including the temporal regularity of indicator changes, critical values, and inflection point characteristics. Simultaneously, through importance analysis, it identifies the most valuable combination of key indicators for risk prediction, serving as characteristic early warning indicators for this type of surgery.

[0071] In some embodiments, time series feature analysis and risk pattern extraction can be implemented in multiple ways: Optionally, the system uses time series decomposition technology to extract trend features, applies dynamic time warping algorithm to align time series patterns, calculates feature importance scores, and finally generates a risk evolution feature library; Optionally, the system uses recurrent neural networks to analyze time series data, identify key time points and state transition features, construct a risk transition probability matrix, and form a risk evolution prediction model. It is understood that other methods can also be used to implement time series feature analysis and risk pattern extraction, such as frequency domain analysis, etc., which are not limited here.

[0072] S203. Based on typical risk evolution patterns and key early warning indicators, establish a differentiated risk assessment database that includes multiple convergent surgical groups.

[0073] The differentiated risk assessment database refers to a set of risk assessment standards established for different types of surgeries. Typical risk evolution patterns represent the development paths of high-risk events in various surgeries. Key early warning indicators are used to represent core parameters that require close monitoring.

[0074] After obtaining the characteristic analysis results of each surgical group, the surgical early warning system needs to establish a systematic risk assessment system. Specifically, the system integrates typical risk evolution patterns and key early warning indicators of various converging surgical groups to establish a hierarchical risk assessment standard library. For each type of surgery, the system sets differentiated early warning thresholds and risk judgment rules, and establishes a correlation assessment mechanism between indicators. These assessment standards will be dynamically adjusted and optimized based on actual application results.

[0075] In some embodiments, the establishment of a differentiated risk assessment library can be achieved in several ways: Optionally, the system constructs initial assessment rules based on expert knowledge, optimizes threshold parameters through machine learning methods, establishes a multi-level assessment standard system, and realizes a dynamic update mechanism; Optionally, the system uses a multi-objective optimization algorithm to balance the sensitivity and specificity of early warnings, constructs an adaptive assessment model, and continuously improves the assessment rules through a feedback mechanism. It is understood that other methods can also be used to establish a differentiated risk assessment library, such as using deep learning algorithms to automatically learn assessment rules, etc., which are not limited here.

[0076] In some embodiments, the surgical early warning system also acquires basic information and surgical plan information of the patient to be operated on; the basic information includes age, past medical history, organ function indicators, and surgical contraindications; the surgical plan information includes surgical type, surgical difficulty level, and estimated surgical duration; based on the basic information and surgical plan information, it matches corresponding convergent surgical groups from a differentiated risk assessment database; it extracts corresponding typical risk evolution patterns and key early warning indicators from the matched convergent surgical groups; it extracts risk factors from the basic information and adjusts the risk thresholds in the typical risk evolution patterns based on the risk factors; and it inputs the adjusted risk thresholds and key early warning indicators into an abnormal event detection model to obtain personalized detection results.

[0077] Here, basic information represents a collection of patient personal characteristics and health status data. Surgical planning information refers to detailed information on the implementation plan developed for a specific surgery. Convergent surgery groups represent a collection of historical surgical cases with similar characteristics and risk patterns. Typical risk evolution patterns are used to represent common risk development paths in a certain type of surgery. Key early warning indicators are core monitoring parameters that are of significant value for risk prediction. Personalized test results represent risk assessment results adjusted according to patient characteristics.

[0078] Before initiating surgical monitoring, the surgical early warning system requires personalized configuration of early warning parameters. Specifically, the system first acquires the patient's detailed basic information and surgical plan, then searches for highly matched concordant surgical groups in a differentiated risk assessment database. Based on the matching results, the system extracts typical risk characteristics and key monitoring indicators for this type of surgery. Based on individual patient differences (such as age, underlying diseases, and other risk factors), the system adjusts the risk thresholds accordingly. Finally, the adjusted parameters are input into the detection model to generate a personalized risk assessment plan for the patient.

[0079] In some embodiments, personalized early warning parameters can be configured in several ways: Optionally, the system uses a multi-dimensional similarity calculation method to match cases, establish patient feature vectors, calculate the matching degree with historical cases, select the most similar convergent surgical group, and extract the corresponding risk patterns; Optionally, the system uses an expert rule system to assess patient risk factors, establish a risk factor weight matrix, dynamically adjust the early warning threshold, and generate a personalized early warning strategy. It is understood that other methods can also be used to configure personalized early warning parameters, such as using deep learning algorithms to automatically learn parameter adjustment strategies, etc., which are not limited here.

[0080] S204. Obtain historical surgical data for hysteroscopic surgeries.

[0081] Referring to step S101, the surgical warning system will acquire historical surgical data.

[0082] S205. Using the abnormal moment marked by the surgical abnormality as the benchmark, trace back a preset time period to obtain historical surgical data within the preset time period as risk evolution data.

[0083] Here, "abnormal moment" refers to the specific time point during the surgical procedure when a risk event occurred. "Preset duration" refers to the length of a fixed time interval for backward tracing. "Risk evolution data" represents the complete monitoring data sequence prior to the occurrence of the abnormal event. "Baseline point" represents the reference time used for time localization.

[0084] When analyzing risk development patterns, surgical early warning systems need to extract key data prior to the occurrence of abnormal events. Specifically, the system uses the occurrence time of each labeled abnormal event as a baseline and traces back a pre-set time period (such as 30 or 60 minutes) to extract all monitoring data within this period, including various sensor data, surgical parameters, and patient indicators. This data comprehensively records the entire evolution process of risk from potential to manifestation, constituting the core training data for the risk early warning model.

[0085] In some embodiments, risk evolution data can be extracted in multiple ways: Optionally, the system sets a dynamic tracing duration algorithm, adaptively adjusts the tracing range according to the anomaly type, filters data integrity, and establishes a data quality assessment mechanism; Optionally, the system adopts a multi-scale time window method to simultaneously acquire data samples from different time spans, perform data completion and outlier processing, and ensure data continuity. It is understood that other methods can also be used to extract risk evolution data, such as intelligent sampling, which are not limited here.

[0086] S206. Divide the risk evolution data into segments according to time intervals, mark a preset number of data segments before the time of anomaly occurrence as high-risk intervals, and mark other data segments as progressive risk intervals of different levels according to the time distance to the time of anomaly occurrence.

[0087] The high-risk zone refers to the critical period before an abnormal event is about to occur. The progressive risk zone is used to represent the transitional period during which risk gradually accumulates.

[0088] After acquiring risk evolution data, the surgical early warning system needs to perform fine-grained segmentation along the time dimension. Specifically, the system first divides the entire time series into continuous data segments at fixed intervals (e.g., 1 minute). For a predetermined number of data segments close to the abnormal event (e.g., the last 5 segments), the system marks them as high-risk intervals. For other data segments, the system marks them as progressively risky intervals of different levels according to their time distance from the time of the abnormality, based on preset risk level classification rules, forming a complete risk evolution sequence.

[0089] In some embodiments, data segmentation and risk labeling can be implemented in multiple ways: Optionally, the system uses an adaptive segmentation algorithm to dynamically adjust the segment length based on data change characteristics, calculate risk level weights, and construct a risk level transformation matrix; alternatively, the system employs a fuzzy time division method to establish a smooth transition mechanism for risk levels, comprehensively assesses the degree of risk by combining multi-dimensional indicators, and generates a continuous risk change curve. It is understood that other methods can also be used to implement data segmentation and risk labeling, such as time-series pattern recognition, which are not limited here.

[0090] S207. Extract the data features of risk evolution data within each risk interval.

[0091] Data characteristics refer to the set of feature values ​​obtained from statistical analysis of the original data, including mean, standard deviation, trend, mutation point, and correlation coefficient. The mean refers to the average level of the data over a specific time period. The standard deviation represents the degree of data fluctuation. The trend indicates the direction of data development. A mutation point is the moment when the data undergoes a significant change. The correlation coefficient represents the degree of correlation between different indicators.

[0092] After defining risk intervals, the surgical early warning system needs to extract feature information for each interval. Specifically, the system performs multi-dimensional feature extraction on the data within each risk interval, including calculating the statistical characteristics (mean, standard deviation) of each monitoring indicator, analyzing the changing characteristics of the data (trend, acceleration), identifying key points of change (abrupt changes, inflection points), and assessing the interrelationships between indicators (correlation, causality). These data features together constitute a feature vector describing the risk evolution process.

[0093] In some embodiments, data feature extraction can be achieved in multiple ways: Optionally, the system uses a sliding window to calculate statistical features, applies regression analysis to identify trend features, employs a mutation detection algorithm to locate key time points, and calculates a multidimensional correlation matrix; alternatively, the system uses wavelet transform to extract time-frequency features, uses entropy value method to quantify data fluctuations, determines indicator correlations through causal analysis, and constructs a feature vector space. It is understood that other methods can also be used to achieve data feature extraction, such as deep feature learning, which are not limited here.

[0094] S208. Based on data characteristics, construct a time-series risk dataset that includes risk level labels.

[0095] In this context, the time-series risk dataset represents a structured dataset containing time information and risk level labels. Feature vectors are multi-dimensional numerical combinations describing data characteristics. Risk level labels are used to represent the risk level classification for a corresponding time period. The data organization structure represents the data storage and indexing methods.

[0096] After acquiring data features from each time interval, the surgical early warning system needs to construct a standardized training dataset. Specifically, the system associates the feature vectors of each time period with the corresponding risk level labels, forming feature-label pairs with temporal attributes. The system standardizes the feature data to ensure the comparability of features across different dimensions. Simultaneously, an efficient data index structure is established to support rapid retrieval and update operations. This dataset will serve as the foundational data source for model training.

[0097] In some embodiments, the construction of a time-series risk dataset can be achieved in several ways: Optionally, the system uses a time-series database to store feature vectors, establishes a multi-dimensional index structure, realizes the associated storage of feature-label pairs, and constructs a data quality verification mechanism; Optionally, the system uses a distributed storage framework to organize data, realizes incremental updates of features, establishes data version control, and ensures data consistency and traceability. It is understood that other methods can also be used to construct the time-series risk dataset, such as graph databases, etc., which are not limited here.

[0098] S209. Based on the time-series risk dataset, establish the correlation between data sequences and construct an abnormal event knowledge graph.

[0099] Referring to step S103, the surgical early warning system will construct an abnormal event knowledge graph.

[0100] S210. Based on the entity association strength in the abnormal event knowledge graph, extract entity combinations with an association degree higher than a preset association threshold.

[0101] In this context, entity association strength represents the weight value of the connection between nodes in the knowledge graph. The preset association threshold refers to the minimum weight standard for selecting strongly related entities. Entity combinations are used to represent a set of monitoring indicators with significant correlation.

[0102] After constructing the knowledge graph, the surgical early warning system needs to extract important entity association patterns. Specifically, the system first calculates the association strength between all entity nodes in the knowledge graph, including a comprehensive evaluation of direct and indirect associations. Then, the system compares the association strength with a preset threshold and filters out entity pairs with an association strength higher than the threshold. For these highly associated entity pairs, the system further analyzes their combination characteristics to form a set of entity combinations with predictive value.

[0103] In some embodiments, entity association analysis can be implemented in multiple ways: Optionally, the system uses a graph traversal algorithm to calculate the association paths between nodes, evaluate the strength of multi-hop associations, apply the PageRank algorithm to calculate node importance, and generate key entity combinations; alternatively, the system employs association rule mining methods to calculate mutual information between entities, construct a hierarchical association network, and identify strongly related entity groups. It is understood that other methods can also be used to implement entity association analysis, such as knowledge graph embedding, which are not limited here.

[0104] S211. Determine the changing patterns of the indicators of each entity in the entity portfolio across different risk ranges.

[0105] The change pattern represents the evolution characteristics of entity indicators over time. The risk range refers to the time period corresponding to different risk levels.

[0106] After acquiring entity combinations, the surgical early warning system needs to analyze their dynamic change characteristics. Specifically, the system analyzes the performance of each entity combination within different risk ranges, calculating the rate of change and acceleration characteristics of each indicator. Simultaneously, the system analyzes the collaborative change relationships within entity combinations, including the synchronicity, causality, and complementarity of indicator changes. These change characteristics will be used to construct the basis for risk warning judgments. The rate of change refers to how quickly the indicator value changes over time; acceleration represents the trend of the rate of change; and collaborative change characteristics represent the linkage relationship between multiple indicators.

[0107] In some embodiments, change pattern analysis can be implemented in multiple ways: Optionally, the system uses polynomial fitting to calculate the change trend, applies numerical differentiation methods to obtain the change rate, constructs a co-change matrix, and identifies key change patterns; alternatively, the system employs time series decomposition technology to extract periodic change features, analyzes nonlinear change laws, and establishes a multidimensional change pattern library. It is understood that other methods can also be used to implement change pattern analysis, such as dynamic system modeling, which are not limited here.

[0108] S212. Establish a risk path feature library based on the change pattern, and bind risk levels to the risk path features in the risk path feature library.

[0109] The risk path feature library represents a knowledge base storing typical patterns of risk development. Risk path features refer to the feature sequences that describe the risk evolution process.

[0110] After analyzing change patterns, the surgical early warning system needs to establish a systematic feature library. Specifically, the system categorizes and organizes the identified change patterns to form standardized risk path feature descriptions. For each feature, the system analyzes its correspondence with different risk levels and establishes feature-risk level mapping rules. Simultaneously, the system assigns weight coefficients to different features to reflect their importance in risk early warning.

[0111] In some embodiments, the risk path feature database can be established in several ways: Optionally, the system uses clustering analysis to classify features, calculates the correlation between features and risk levels, establishes a hierarchical feature index, and enables rapid feature retrieval; alternatively, the system uses an expert rule system to define feature weights, constructs a feature scoring mechanism, and designs a feature update strategy to maintain the timeliness of the feature database. It is understood that other methods can also be used to establish the risk path feature database, such as transfer learning, which are not limited here.

[0112] In some embodiments, the surgical early warning system performs time-series clustering analysis on change patterns, classifying change patterns with the same trend into the same risk path feature; based on the risk level labels in the time-series risk dataset, it calculates the association probability between each risk path feature and different risk levels; it binds risk levels with association probabilities higher than a preset probability threshold to the corresponding risk path features; based on the change rate, acceleration, and co-change characteristics of the risk path features, it sets a risk weight coefficient for each risk path feature; and it adjusts the risk level of the risk path features according to the risk weight coefficient to obtain a risk path feature library.

[0113] Temporal clustering analysis is an analytical method for grouping and classifying time-series data. Trend refers to the direction and characteristics of data changes over time. Risk path features represent feature sequences describing the evolution of risk. Association probability represents the likelihood of a correspondence between a feature and a risk level. Risk weight coefficients are numerical values ​​reflecting the importance of features. The risk path feature library represents a knowledge base storing typical patterns of risk development.

[0114] After acquiring change pattern data, the surgical early warning system needs to establish a standardized risk feature classification system. Specifically, the system first performs time-series cluster analysis on all change patterns to identify pattern groups with similar trends. Then, based on risk level labels in historical data, it statistically analyzes the correspondence between each feature pattern and different risk levels, calculating the association probability. For feature-risk level pairs with association probabilities higher than a threshold, the system performs association binding. Considering the differences in the importance of different features, the system sets weight coefficients based on multiple dimensions of the change features and fine-tunes the risk levels accordingly, ultimately forming a complete risk path feature library.

[0115] In some embodiments, risk feature classification and assessment can be achieved in multiple ways: Optionally, the system uses a dynamic time warping algorithm to align time-series patterns, calculates a pattern similarity matrix, applies hierarchical clustering methods to construct feature family trees, and establishes feature classification standards; Optionally, the system uses machine learning methods to train a feature weight model, analyzes the contribution of features to risk prediction, constructs a multi-level feature scoring mechanism, and dynamically updates feature weights. It is understood that other methods can also be used to classify and assess risk features, such as using deep learning algorithms to automatically extract feature classification rules, etc., which are not limited here.

[0116] S213. Combine the risk level with the risk path features to train the abnormal event detection model.

[0117] The risk level indicates the severity of the risk.

[0118] After establishing a feature database, the surgical early warning system needs to train an early warning model. Specifically, the system uses path features labeled with risk levels as training samples and employs machine learning algorithms to construct an abnormal event detection model. During training, the system adjusts the model parameters through algorithm optimization to ensure it can accurately identify feature patterns corresponding to different risk levels. Simultaneously, the system evaluates model performance through cross-validation to ensure the model has good generalization ability.

[0119] In some embodiments, model training can be implemented in multiple ways: Optionally, the system uses a deep learning framework to build the model, designs a multi-layer feature extraction network, applies gradient descent to optimize parameters, and achieves end-to-end risk detection; alternatively, the system employs an ensemble learning method to combine multiple base models, achieving voting or weighted fusion to improve the robustness of the model. It is understood that other methods can also be used for model training, such as reinforcement learning, which are not limited here.

[0120] S214. Input the real-time monitoring data of hysteroscopic surgery into the abnormal event detection model according to the time window to obtain the abnormal detection results and the confidence level of the results.

[0121] Referring to step S105, the surgical early warning system will generate abnormal detection results and result confidence levels.

[0122] S215. Determine the surgical risk level based on the abnormality detection results and the confidence level of the results, and issue a surgical warning based on the surgical risk level.

[0123] Referring to step S106, the surgical warning system will issue a surgical warning based on the surgical risk level.

[0124] In some embodiments, the surgical early warning system records all early warning events during the surgical procedure to obtain intraoperative early warning records. The early warning events include the early warning time, the combination of indicators that triggered the early warning, the risk level, and the doctor's handling measures. After the surgery, based on the intraoperative monitoring data and the intraoperative early warning records, the accuracy rate, false alarm rate, and false negative rate of the early warning are analyzed to obtain the early warning evaluation results. Based on the early warning evaluation results, the parameters of the abnormal event detection model are adjusted.

[0125] Among them, an early warning event refers to a specific record of a risk warning issued by the system. An intraoperative early warning record refers to a complete collection of all early warning information during the surgical procedure. Early warning evaluation results are used to represent a comprehensive assessment of the early warning system's performance. Accuracy rate indicates the correctness of the early warning judgment. False alarm rate refers to the proportion of incorrect early warnings. Missed alarm rate indicates the proportion of early warnings not issued in a timely manner.

[0126] After completing surgical monitoring, the surgical early warning system needs to evaluate and optimize its effectiveness. Specifically, the system records all early warning events throughout the surgery, including trigger time, relevant indicators, risk level, and the surgeon's handling. After the surgery, the system compares and analyzes the early warning records with the actual monitoring data, calculating performance indicators such as accuracy, false alarm rate, and false negative rate. Based on these evaluation results, the system adjusts the parameters of the abnormal event detection model accordingly to improve the accuracy and reliability of the early warnings.

[0127] In some embodiments, early warning effectiveness evaluation and model optimization can be achieved in multiple ways: Optionally, the system establishes an early warning effectiveness evaluation matrix, calculates various performance indicators, analyzes the causes of early warning failure, generates optimization suggestions, and updates model parameters; Optionally, the system adopts an incremental learning method to integrate new early warning cases, adjust model weights, optimize early warning thresholds, and improve the model's adaptability. It is understood that other methods can also be used to achieve early warning effectiveness evaluation and model optimization, such as using reinforcement learning methods to dynamically optimize early warning strategies, etc., which are not limited here.

[0128] In this embodiment, by employing a technical solution that constructs a time-series risk dataset and anomaly event knowledge graph based on historical surgical data, and combines this with machine learning methods for model training, multi-dimensional analysis and dynamic assessment of surgical risks can be achieved. This effectively solves the problems of untimely warnings, insufficient accuracy, and low personalization in traditional early warning methods, thereby realizing more intelligent and accurate surgical risk warnings. Real-time data analysis and personalized threshold adjustments improve the accuracy and timeliness of warnings; continuous model optimization and iterative learning enable dynamic optimization of the early warning system, enhancing the safety of hysteroscopic surgery.

[0129] The surgical early warning system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference]. Figure 3 This is a schematic diagram of the physical device structure of a surgical early warning system in an embodiment of this application.

[0130] It should be noted that, Figure 3 The structure of the surgical early warning system shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0131] like Figure 3 As shown, the surgical early warning system includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 302 or programs loaded from storage section 308 into Random Access Memory (RAM) 303, such as performing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0132] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0133] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the various functions defined in the present invention.

[0134] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0135] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0136] Specifically, the surgical early warning system of this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the early warning method for hysteroscopic surgery provided in the above embodiment.

[0137] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the surgical warning system described in the above embodiments; or it may exist independently and not incorporated into the surgical warning system. The storage medium carries one or more computer programs, which, when executed by a processor of the surgical warning system, cause the surgical warning system to implement the hysteroscopic surgery warning method provided in the above embodiments.

[0138] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0139] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0140] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for early warning during hysteroscopic surgery, characterized in that, The method, applied to a surgical early warning system, includes: Acquire historical surgical data for hysteroscopic surgeries; the historical surgical data includes sensor data, patient data, surgical parameters, and surgical abnormality annotations from each historical surgical record; the sensor data includes uterine cavity image data, distension fluid irrigation difference data, uterine cavity size data, and uterine cavity pressure values; the surgical parameters include anesthesia type, anesthetic drug data, and distension fluid data; Based on the surgical anomaly annotations, the historical surgical data is segmented in time series to construct a time series risk dataset; Based on the aforementioned time-series risk dataset, establish the correlation between data sequences and construct an abnormal event knowledge graph; Risk path features are extracted from the abnormal event knowledge graph to obtain risk warning conditions, and the model is trained based on the risk warning conditions to obtain an abnormal event detection model. Real-time monitoring data of hysteroscopic surgery is input into the abnormal event detection model according to a time window to obtain abnormal detection results and result confidence levels; The surgical risk level is determined based on the anomaly detection results and the confidence level of the results, and a surgical warning is issued based on the surgical risk level.

2. The method according to claim 1, characterized in that, The step of constructing a time-series risk dataset by performing time-series segmentation of the historical surgical data based on the surgical anomaly annotations specifically includes: Using the abnormal moment marked by the surgical abnormality as the benchmark, trace back a preset time period to obtain historical surgical data within the preset time period as risk evolution data; The risk evolution data is segmented according to time intervals. A predetermined number of data segments before the time of an anomaly are marked as high-risk intervals, and other data segments are marked as progressive risk intervals of different levels according to the time distance to the time of an anomaly. Extract data features of risk evolution data within each risk interval; the data features include mean, standard deviation, trend, abrupt change points, and correlation coefficient; Based on the aforementioned data characteristics, a time-series risk dataset containing risk level labels is constructed.

3. The method according to claim 1, characterized in that, The steps of extracting risk path features from the abnormal event knowledge graph to obtain risk warning conditions, and training a model based on the risk warning conditions to obtain an abnormal event detection model, specifically include: Based on the entity association strength in the abnormal event knowledge graph, entity combinations with an association degree higher than a preset association threshold are extracted. Determine the change patterns of each entity indicator in the entity portfolio across different risk ranges; the change patterns include change rate, acceleration, and collaborative change characteristics. A risk path feature library is established based on the change pattern, and risk levels are bound to the risk path features in the risk path feature library. The risk path features are trained based on the risk level to obtain an abnormal event detection model.

4. The method according to claim 3, characterized in that, The step of establishing a risk path feature library based on the change pattern and binding risk levels to the risk path features in the risk path feature library specifically includes: Temporal clustering analysis was performed on the aforementioned change patterns to classify change patterns with the same trend into the same risk path characteristics; Based on the risk level labels in the time-series risk dataset, calculate the association probability between each risk path feature and different risk levels; The risk level with an association probability higher than a preset probability threshold is bound to the corresponding risk path characteristics; Based on the rate of change, acceleration, and collaborative change characteristics of the risk path features, a risk weight coefficient is set for each risk path feature; The risk level of the risk path feature is adjusted according to the risk weight coefficient to obtain the risk path feature library.

5. The method according to claim 1, characterized in that, Prior to the step of acquiring historical surgical data for hysteroscopic procedures, the method further includes: Based on the complexity of the surgery, the patient's risk level, and the surgical outcome, historical surgical records were grouped and labeled to obtain multiple convergent surgical groups. Time-series feature analysis was performed on the surgical records within each convergent surgical group to generate typical risk evolution patterns and key early warning indicators. Based on the aforementioned typical risk evolution patterns and key early warning indicators, a differentiated risk assessment database containing multiple convergent surgical groups is established.

6. The method according to claim 5, characterized in that, Following the step of establishing a differentiated risk assessment library containing multiple convergent surgical groups based on the typical risk evolution pattern and the key early warning indicators, the method further includes: Obtain basic information and surgical plan information of the patient to be operated on; the basic information includes age, medical history, organ function indicators and surgical contraindications; the surgical plan information includes surgical type, surgical difficulty level and estimated surgical duration. Based on the basic information and the surgical plan information, a corresponding convergent surgical group is matched from the differential risk assessment database; Extract typical risk evolution patterns and key early warning indicators from the matched convergent surgical groups; Extract risk factors from the basic information, and adjust the risk thresholds in the typical risk evolution pattern based on the risk factors; The adjusted risk threshold and the key early warning indicators are input into the abnormal event detection model to obtain personalized detection results.

7. The method according to claim 1, characterized in that, After the steps of determining the surgical risk level based on the anomaly detection results and the confidence level of the results, and issuing a surgical warning based on the surgical risk level, the method further includes: All warning events during the surgical procedure are recorded to obtain an intraoperative warning record; the warning events include the warning time, the combination of indicators that triggered the warning, the risk level, and the doctor's handling measures; After the operation, based on the intraoperative monitoring data and the intraoperative early warning record, the accuracy rate, false alarm rate and false negative rate of the early warning are analyzed to obtain the early warning evaluation results; Based on the early warning assessment results, the parameters of the abnormal event detection model are adjusted.

8. A surgical early warning system, characterized in that, The surgical early warning system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the surgical early warning system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the surgical warning system, the surgical warning system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the surgical early warning system, it causes the surgical early warning system to perform the method as described in any one of claims 1-7.

Images