A Modeling Method for Predicting Postoperative Complications of Gynecological General Anesthesia Based on Pathological Data Analysis

Abstract

This invention discloses a modeling method for predicting postoperative complications after gynecological general anesthesia based on pathological data analysis. It relates to the field of postoperative complication prediction technology, and involves collecting multi-dimensional pathological and clinical data before, during, and after gynecological general anesthesia. A spatiotemporal correlation matrix is ​​established between anxiety, intraoperative pathology, and complications. Based on this matrix, a prediction model for postoperative complications after gynecological general anesthesia based on a multimodal deep learning architecture is constructed. This invention achieves a shift from experience-based decision-making to data-driven decision-making through the combination of multi-dimensional data fusion, dynamic correlation analysis, and deep learning. The spatiotemporal correlation matrix provides physicians with standardized risk assessment logic; the model's predictive ability for rare complications fills the prediction gap in small-sample scenarios. Ultimately, through real-time risk warnings and individualized intervention suggestions, the safety of gynecological general anesthesia surgery is significantly improved, and the postoperative recovery time for patients is shortened, demonstrating significant clinical translational value.

Description

Technical Field

[0001] This invention relates to the field of postoperative complication prediction technology, and in particular to a method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis. Background Technology

[0002] Currently, research on predicting postoperative complications after gynecological general anesthesia mainly focuses on the correlation analysis between surgical types, such as hysterectomy and ovarian surgery, and traditional clinical indicators. Predictive models are mainly constructed based on routine pathological data such as the patient's underlying diseases, such as hypertension, diabetes, operation duration, anesthetic drug dosage, and intraoperative blood loss. Commonly used methods include multivariate logistic regression and Cox proportional hazards model.

[0003] Studies have focused primarily on physiological indicators, such as intraoperative hemodynamics and the impact of laboratory tests on complications, while insufficient attention has been paid to patients' psychological factors, such as preoperative anxiety. In particular, research on the association between psychological factors and postoperative complications is almost nonexistent in specific gynecological surgeries such as female adnexectomy.

[0004] While psychological assessment tools such as the State Anxiety Scale (SAI) have proven feasible in thyroid surgery and elderly patients, their application in gynecological general anesthesia surgery patients has not been widespread, and their synergistic predictive value with pathological indicators has not been fully explored.

[0005] Existing predictive models are mostly general models, lacking individualized designs for gynecological surgeries such as adnexectomy, and failing to fully consider the physiological characteristics of gynecological patients, such as the specific correlation between hormone levels, pelvic anatomy, and complications.

[0006] Existing models mostly rely on basic indicators such as operation duration and anesthetic drug dosage, ignoring the interaction between psychological factors (such as preoperative anxiety) and pathological indicators (such as the cumulative effect of intraoperative blood pressure fluctuations caused by anxiety on complications), resulting in incomplete feature coverage.

[0007] Insufficient gynecological specificity: The model was not designed specifically for gynecological surgeries (such as adnexectomy), for example, it did not include pathological data related to pelvic surgery (such as the degree of pelvic adhesions and bleeding during surgery), and the correlation between female patients' hormone levels, reproductive history and complications was not sufficiently explored.

[0008] Model performance limitations: It mainly uses traditional statistical methods (such as univariate analysis and binary logistic regression), which are weak in capturing nonlinear features (such as the nonlinear relationship between anxiety scores and postoperative pain thresholds), and the prediction accuracy (such as AUC value) and stability need to be improved.

[0009] Insufficient clinical interpretability: While some machine learning models (such as black box models) can improve prediction accuracy, they are difficult to clarify the influence weight of each pathological feature on complications, and cannot provide specific targets for clinical intervention. Summary of the Invention

[0010] To address the aforementioned technical problems, this invention provides a method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis. The technical solution adopted is as follows:

[0011] A modeling method for predicting postoperative complications of gynecological general anesthesia based on pathological data analysis includes the following steps:

[0012] Step 1: Collect multi-dimensional pathological and clinical data before, during, and after gynecological general anesthesia.

[0013] Step 2: Establish a spatiotemporal correlation matrix of anxiety, intraoperative pathology, and complications;

[0014] Step 3: Construct a prediction model for postoperative complications of gynecological general anesthesia based on a multimodal deep learning prediction architecture using a spatiotemporal correlation matrix;

[0015] Step 4: Input the current patient's multi-dimensional pathological and clinical data before, during, and after the operation into the complication prediction model, and output the types of complications and the probability of their occurrence.

[0016] Optionally, the multidimensional pathological and clinical data before gynecological general anesthesia in step 1 includes basic clinical indicator data, pathological molecular marker detection data, and psychological state quantitative data.

[0017] Multidimensional pathological and clinical data during gynecological general anesthesia include real-time physiological indicators, intraoperative pathological indicators, and surgical parameters.

[0018] Multidimensional pathological and clinical data following gynecological general anesthesia include complication records and data on repair and infection marker detection.

[0019] Optionally, in step 2, the time is divided into several key time nodes;

[0020] The spatial dimension is defined as several types of related entities, and several core related pairs are formed by combining representative indicators of these types of related entities in pairs.

[0021] Optionally, in step 2, the spatiotemporal correlation matrix uses several key time nodes as row labels and several core correlation pairs as column labels. The initial values ​​of the matrix are all 0. Subsequently, the correlation coefficients obtained through statistical calculations are used to fill the matrix cells to form a complete spatiotemporal correlation matrix.

[0022] Optional key time points are: 24 hours before surgery, after anesthesia induction, 1 hour of surgery, at the end of surgery, and 24 hours after surgery;

[0023] The related subjects are categorized into three types: psychological state, pathological indicators, and physiological reactions.

[0024] Representative indicators of psychological state include anxiety assessment score, heart rate variability, and skin conductance response.

[0025] Representative pathological indicators include preoperative inflammatory factors, inflammation grade, and high-mobility group box 1 (HMP) levels.

[0026] Representative indicators of physiological responses include blood pressure fluctuation amplitude, heart rate variability, and BIS value.

[0027] Optional, the specific statistical calculation process for the correlation coefficient:

[0028] Step a: Extract the associated pairs of data for the corresponding time periods from the standardized time series database according to the time nodes;

[0029] Step b: If the correlation pair is between continuous variables, first perform a normality test. If it conforms to a normal distribution, use the Pearson correlation coefficient to calculate it. If it is between a continuous variable and an ordinal variable, use the Spearman rank correlation coefficient. Rank correlation is calculated by sorting the data.

[0030] Step c: Perform a t-test on the calculated correlation coefficients and calculate the P-value. Repeat the above steps for each of the five key time points.

[0031] Step d involves performing clinical logic verification on the statistical results. If any results conflict with common medical knowledge, they are corrected based on the opinions of clinical experts.

[0032] Optionally, in step 3, the model architecture of the complication prediction model includes an input layer, a fusion layer, and an output layer;

[0033] The input layer includes a static data channel, a time-series data channel, and an image data channel;

[0034] The fusion layer introduces an attention mechanism based on the spatiotemporal correlation matrix: dynamic attention weights are assigned to the indicators involved in the core correlation pairs;

[0035] The output layer is designed based on multi-task learning, which simultaneously predicts multiple types of complications and risk probabilities.

[0036] The Sigmoid function is used to predict the probability of a single complication, and the Softmax function is used to rank multiple complications.

[0037] Optionally, step 5 is also included, which generates clinical intervention recommendations based on the complication prediction results.

[0038] Optionally, step 5, generating clinical intervention recommendations, includes the following steps:

[0039] The complication prediction model outputs the probability of complication risk and the contribution of each feature. A contribution threshold is set to screen out the core risk factors, which are then sorted in descending order of contribution to form a priority list of risk factors.

[0040] Establish a risk factor and intervention mapping database, including: pathological indicator intervention, psychological state intervention, and physiological indicator intervention;

[0041] The corresponding intervention measures are retrieved from the mapping library based on the type of risk factor.

[0042] Optionally, gene polymorphism correction factors can be introduced to adjust intervention measures in combination with the patient's basic characteristics.

[0043] In summary, the present invention has the following beneficial technical effects:

[0044] This invention provides a predictive modeling method for postoperative complications after gynecological general anesthesia based on pathological data analysis. Through a combination of multi-dimensional data fusion, dynamic correlation analysis, and deep learning, it achieves a shift from experience-based decision-making to data-driven decision-making. On one hand, the spatiotemporal correlation matrix provides physicians with standardized risk assessment logic; on the other hand, the model's predictive ability for rare complications fills the prediction gap in small-sample scenarios. Ultimately, through real-time risk warnings and individualized intervention suggestions, it significantly improves the safety of gynecological general anesthesia surgery and shortens postoperative recovery time for patients, demonstrating significant clinical translational value. Attached Figure Description

[0045] Figure 1 This is a flowchart illustrating the method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis, as proposed in this invention. Detailed Implementation

[0046] The present invention will be further described in detail below with reference to the accompanying drawings.

[0047] This invention discloses a method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis.

[0048] Reference Figure 1 Example 1, a method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis, includes the following steps:

[0049] Step 1: Collect multi-dimensional pathological and clinical data before, during, and after gynecological general anesthesia.

[0050] Step 2: Establish a spatiotemporal correlation matrix of anxiety, intraoperative pathology, and complications;

[0051] Step 3: Construct a prediction model for postoperative complications of gynecological general anesthesia based on a multimodal deep learning prediction architecture using a spatiotemporal correlation matrix;

[0052] Step 4: Input the current patient's multi-dimensional pathological and clinical data before, during, and after the operation into the complication prediction model, and output the types of complications and the probability of their occurrence.

[0053] By adopting the above technical solution, step 1 overcomes the limitations of traditional methods that rely solely on a single clinical indicator by collecting multi-dimensional pathological data from preoperative, intraoperative, and postoperative periods, such as inflammatory factors, genotyping, and frozen section grading, along with clinical data such as SAI scores, physiological indicators, and complication records. The integration of full-cycle time-series data not only covers a complete profile of the patient's psychological state, pathological characteristics, and physiological responses, but also ensures data consistency through timestamps and standardization, providing high-quality, multi-dimensional foundational material for subsequent correlation analysis and model training. This solves the problem of predictive bias caused by traditional data fragmentation.

[0054] In step 1, the standardized processing of multi-dimensional pathological and clinical data before, during, and after gynecological general anesthesia includes the following steps:

[0055] Data types and sources:

[0056] Preoperative data includes: basic clinical indicators such as age, BMI, and medical history, which are entered in a structured manner through the electronic medical record system; pathological molecular marker detection data such as IL-6 and CRP, which are detected by electrochemiluminescence / PCR fluorescent probe method, and the results are synchronized to the laboratory information system; and psychological state quantitative data (SAI score is digitally collected through WeChat mini program, and HRV index is recorded simultaneously by wearing a wrist heart rate variability monitor).

[0057] Intraoperative data includes: real-time physiological indicators, with blood pressure and heart rate recorded every 5 minutes, and BIS values ​​continuously sampled at 1Hz, automatically uploaded to a time-series database via anesthesia monitor; intraoperative pathological indicators (frozen section inflammation grading was blind-reviewed by two pathologists and quantitatively analyzed using a digital pathology system; HMGB1 was detected using a portable microfluidic chip, with results output within 15 minutes and synchronized to the intraoperative database); and surgical operation parameters (surgical type, duration, and blood loss were recorded using surgical procedure coding and dual-method measurement).

[0058] Postoperative data includes: complication records, using the Clavien-Dindo classification, combined assessment and data entry at 24 / 48 / 72 hours postoperatively, repair and infection marker detection data, such as VEGF and PCT, collected at time points and calculated dynamic ratios.

[0059] Data unification method:

[0060] Time alignment: Add timestamps to all data, anchoring them to key time points such as 24 hours before surgery and after anesthesia induction to ensure time sequence consistency;

[0061] Format conversion: unstructured data, such as pathological images, are converted to DICOM format; text data, such as medical history, is converted to structured labels using natural language processing; physiological signals are converted to time series arrays.

[0062] Standardized database construction: Build a unified time-series database, integrate data from various systems such as electronic medical records, anesthesia monitors, and laboratory systems, and achieve real-time synchronization through API interfaces. The data storage format is unified as timestamp + data type + value + unit, and supports fast retrieval by time node or data type.

[0063] The spatiotemporal correlation matrix constructed in step 2, through quantitative analysis of the correlation strength between anxiety psychology (such as SAI score), intraoperative pathological indicators (such as HMGB1), inflammation grade, and complication risk at different time points, systematically reveals for the first time the cascading mechanism of psychological state, pathological changes, physiological reactions, and complications. For example, it clarifies the synergistic effect of high preoperative SAI score and intraoperative inflammation grade ≥2 on postoperative agitation, providing interpretable correlation rules for the predictive model. This ensures that the model output is no longer a black box result but a causal inference based on clinical logic, enhancing the credibility of the prediction results.

[0064] Step 2, the specific steps for establishing the spatiotemporal correlation matrix of anxiety, intraoperative pathology, and complications are as follows:

[0065] Definition of spacetime dimension:

[0066] Time dimension:

[0067] The study is divided into five key time points: 24 hours before surgery, after anesthesia induction, 1 hour after surgery, at the end of surgery, and 24 hours after surgery. Based on the pathophysiological laws of gynecological surgery, it covers the preoperative baseline state, the peak of intraoperative stress, and the early postoperative reaction stage.

[0068] Spatial dimension:

[0069] Three categories of related subjects are defined: psychological state, pathological indicators, and physiological reactions. Representative indicators are as follows:

[0070] Psychological state: Anxiety assessment score (SAI score), heart rate variability (LF / HF ratio), and skin conductance;

[0071] Pathological indicators: preoperative inflammatory factors (IL-6, CRP), inflammation grade (density of inflammatory cells in frozen sections), and high-mobility group box 1 (HMGB1) detection value;

[0072] Physiological responses: blood pressure fluctuation range, heart rate variability, BIS value (depth of anesthesia).

[0073] Core association pairs: Based on the pairwise combination of representative indicators of the three subjects, nine core association pairs are formed (such as "SAI score-IL-6", "inflammation grade-blood pressure fluctuation range", "HMGB1-BIS value", etc.), covering the interaction relationship between psychology-pathology, pathology-physiology, and psychology-physiology.

[0074] Matrix construction process:

[0075] Matrix structure: 5 key time nodes are used as row labels, and 9 core association pairs are used as column labels, with all initial values ​​being 0;

[0076] Correlation coefficient calculation:

[0077] Step a: Extract correlation pairs from the standardized time series database by time node, such as SAI score and IL-6 value 24 hours before surgery;

[0078] Step b: Select the calculation method according to the data type. For continuous variables, such as SAI score and IL-6, use Pearson correlation coefficient when they conform to a normal distribution. For continuous variables and ordinal variables, use Spearman rank correlation coefficient.

[0079] Step c: Perform a t-test on the correlation coefficient and calculate it separately for each of the five time points;

[0080] Step d: Clinical logic verification. If the result conflicts with common medical knowledge, it should be revised in conjunction with the opinions of 3 gynecologists with the title of associate chief physician or above.

[0081] Matrix filling: Fill the corresponding cells with the verified correlation coefficients to form a complete spatiotemporal correlation matrix. For example, the correlation coefficient of "inflammation grade - blood pressure fluctuation amplitude" 1 hour after surgery is 0.62, indicating that the two are significantly positively correlated at this node.

[0082] Step 3 establishes a gynecological general anesthesia postoperative complication prediction model based on a spatiotemporal correlation matrix. This model integrates ResNet50 pathological image features, ST-GCN spatiotemporal physiological signals, and a fully connected layer of static clinical data. Furthermore, it incorporates an attention mechanism through association rules, enabling the model to accurately capture the dynamic impact of key risk factors. Compared to traditional methods such as Logistic regression or Apfel scoring, this complication prediction model exhibits high sensitivity to complication warnings and can update risk probabilities in real-time every 30 minutes during surgery, overcoming the limitation of traditional static prediction methods that cannot adapt to the dynamic changes during the surgical procedure.

[0083] The specific architecture and construction process of the gynecological general anesthesia postoperative complication prediction model are as follows:

[0084] Model architecture: includes an input layer, a fusion layer, and an output layer;

[0085] Input layer: Contains three parallel channels;

[0086] Static data channel: Input basic clinical indicators, age, BMI, and non-time-series data such as gene polymorphisms (HTR2A, COMT genotype), which are converted into feature vectors through a fully connected layer;

[0087] Temporal data channel: Input intraoperative physiological signals, such as blood pressure, BIS value, dynamic pathological indicators, and HMGB1 temporal changes, and use spatiotemporal graph convolutional network (ST-GCN) to extract temporal features;

[0088] Image data channel: Input pathological slide images, and extract texture features, such as the distribution of inflammatory cells, through the ResNet50 network.

[0089] Fusion layer: Introduces an attention mechanism based on the spatiotemporal correlation matrix;

[0090] Based on the correlation coefficient of the spatiotemporal correlation matrix, dynamic weights are assigned to the indicators involved in the core correlation pairs.

[0091] The attention layer fuses the features from the three channels into a unified feature vector.

[0092] Output layer: designed based on multi-task learning;

[0093] The Sigmoid function was used to independently predict the probability of a single type of complication, such as agitation, nausea and vomiting, with values ​​ranging from 0 to 1.

[0094] The Softmax function is used to rank the risks of multiple complications and output the risk priority, such as nausea and vomiting having a higher risk priority than pain, which in turn has a higher risk priority than agitation.

[0095] Model training and optimization:

[0096] Training data: Multi-dimensional data from 1000 cases of gynecological general anesthesia surgery were used, with 80% used for training and 20% for validation;

[0097] Loss function: The cross-entropy loss function is adopted, combined with the clinical constraints of the spatiotemporal correlation matrix;

[0098] Optimizer: The Adam optimizer is used, with a dynamically adjusted learning rate. It starts at 0.001 and decreases by 10% every 10 rounds until the validation set AUC stabilizes above 0.92.

[0099] Step 4 provides clear targets for clinical decision-making by outputting specific complication types (such as agitation, nausea, and vomiting) and their corresponding risk probabilities. Combining the association rules from Step 2 with the feature contribution analysis from Step 3, doctors can directly identify high-risk causes and develop targeted intervention plans. This closed loop of risk identification, cause analysis, and measure matching not only improves the timeliness of clinical intervention but also reduces the risk of overtreatment or undertreatment through adjustments to individualized characteristics such as genetic polymorphism, thereby lowering the incidence of complications.

[0100] This method, through the combination of multi-dimensional data fusion, dynamic correlation analysis, and deep learning, achieves a shift from experience-based decision-making to data-driven decision-making. On the one hand, the spatiotemporal correlation matrix provides doctors with standardized risk assessment logic; on the other hand, the model's predictive ability for rare complications fills the prediction gap in small-sample scenarios. Ultimately, through real-time risk warnings and individualized intervention suggestions, it significantly improves the safety of gynecological general anesthesia surgery and shortens postoperative recovery time for patients, demonstrating significant clinical translational value.

[0101] Example 2: The multidimensional pathological and clinical data before gynecological general anesthesia in step 1 includes basic clinical indicator data, pathological molecular marker detection data, and psychological state quantitative data.

[0102] Multidimensional pathological and clinical data during gynecological general anesthesia include real-time physiological indicators, intraoperative pathological indicators, and surgical parameters.

[0103] Multidimensional pathological and clinical data following gynecological general anesthesia include complication records and data on repair and infection marker detection.

[0104] By adopting the above technical solution, the collection of basic clinical indicator data can be completed 24-48 hours before the operation, through a combination of electronic medical record system and manual data entry.

[0105] Specific indicators and standards:

[0106] Demographic characteristics: Age, accurate to the nearest year; BMI: weight / height², rounded to one decimal place, measured after fasting for 8 hours;

[0107] Medical history information: Recorded using a structured questionnaire, including gynecological disease type, such as uterine fibroids / ovarian cancer, surgical history (number of pelvic surgeries in the past 5 years), preoperative medication history (names and dosages of anticoagulants / hormonal drugs), and allergy history (specific drugs and reactions should be noted if there is an allergy to anesthetic drugs).

[0108] Reduce prediction bias caused by omissions in medical history;

[0109] Pathological molecular marker detection data are collected 12-24 hours before surgery, using 5ml of venous blood (EDTA anticoagulant tube).

[0110] Detection indicators and methods:

[0111] Serum inflammatory factors: IL-6 (electrochemiluminescence immunoassay, detection limit 0.5 pg / ml), CRP (high-sensitivity latex-enhanced immunoturbidimetric assay, range 0.1-100 mg / L), detected by Roche Cobase 601 instrument;

[0112] Stress gene polymorphism: HTR2A and COMT gene typing was performed using PCR fluorescent probe method (ABI 7500 real-time quantitative PCR instrument), and the primer sequences were verified by the NCBI database;

[0113] Genotyping results provide reliable molecular-level evidence for subsequent association analysis;

[0114] Quantitative data on psychological state:

[0115] Collection time: 24 hours before the operation, completed in a quiet examination room;

[0116] Specific methods:

[0117] SAI Anxiety Scale: Data is collected digitally via WeChat mini-program, and the total score (0-60 points) is automatically calculated. Auxiliary physiological signals: A wrist-type heart rate variability (HRV) monitor is worn simultaneously to record 5 minutes of resting state data and extract the LF / HF ratio (sympathetic nerve activity index).

[0118] Data association: The SAI score is linked to the HRV indicator to form a psychological-physiological joint data set;

[0119] The quantitative dimensions of psychological state have been expanded from a single scale to include subjective scores plus objective physiological signals, which has increased the correlation with intraoperative stress response.

[0120] Intraoperative data collection during gynecological general anesthesia:

[0121] Real-time physiological data:

[0122] Monitoring frequency: Blood pressure / heart rate is recorded every 5 minutes, and BIS value and muscle relaxation monitoring value (TOF) are continuously sampled (1Hz) and automatically uploaded to the database via an anesthesia monitor (such as Philips IntelliVue MX800);

[0123] Circulatory parameters: systolic blood pressure / diastolic blood pressure / mean arterial pressure, calculate the fluctuation range;

[0124] Anesthesia status indicators: BIS value (target range 40-60, reflecting the depth of anesthesia), TOF ratio (recovery to 0.9 is considered as muscle relaxation subsidence).

[0125] Anesthetic drugs: Real-time recording of propofol / sevoflurane dose (mg / kg / h) and total amount of opioids (converted to morphine equivalents);

[0126] Physiological data provides a precise temporal basis for real-time correlation analysis between pathology and physiology.

[0127] Intraoperative pathological data:

[0128] Inflammation grading of frozen sections:

[0129] Sampling timing: Complete within 30 minutes after adnexal resection, and take tissue samples from 3 sites (cortex / medulla / junction area) to prepare sections;

[0130] Grading process: Blind review by two pathologists, using a digital pathology system (such as 3DHISTECH) to quantitatively analyze inflammatory cell density;

[0131] HMGB1 detection:

[0132] Timing: 30 minutes after anesthesia induction, after key surgical steps (such as tumor resection), and before abdominal closure, for a total of 3 times;

[0133] Portable microfluidic chip detection (detection time 15 minutes, lower limit 0.1 ng / ml), results are automatically synchronized to the intraoperative database;

[0134] HMGB1 dynamic detection captured the intraoperative peak time point (which occurred on average 28 minutes after tumor resection), providing key time node data for the correlation matrix;

[0135] Surgical operation parameters:

[0136] Recording standards: The type of surgery (laparoscopic / open surgery) is marked by a surgical procedure code, and the operation time is accurate to the minute (from skin incision to closure of the abdomen).

[0137] Blood loss measurement: The blood loss was measured using a double-check method: suction fluid volume - irrigation fluid volume + gauze weight.

[0138] The correlation analysis between surgical parameters and pathological indicators provides a basis for introducing surgical procedure weights into the model;

[0139] Postoperative data collection after gynecological general anesthesia:

[0140] Complication record:

[0141] Recording standards: Clavien-Dindo classification (I-V) is adopted, and the assessment is conducted jointly by nurses and doctors at 24 / 48 / 72 hours postoperatively;

[0142] Specific indicators: agitation: The Riker scale is scored once per hour, and a score of 5 or higher is defined as agitation;

[0143] Nausea and vomiting: PONV grade (0-3), record the time of first occurrence and duration;

[0144] Pain: VAS score (0-10) recorded every 6 hours; a score of 4 or higher requires intervention.

[0145] The standardization of complication grading makes cross-center data comparable, and the sensitivity of subsequent models to identify severe complications of grade III and above is greatly improved.

[0146] Repair and infection marker detection data:

[0147] Testing time points: Venous blood was collected at 24h, 48h and 72h postoperatively to test VEGF (vascular repair), TGF-β (tissue repair), and PCT (infection warning).

[0148] Dynamic analysis: Calculate the 48h / 24h ratio. A ratio greater than 1.2 indicates delayed repair or infection risk.

[0149] The dynamic changes of biomarkers were significantly associated with complications, providing postoperative risk enhancement indicators for the model;

[0150] Multidimensional data covers the entire cycle from preoperative to intraoperative to postoperative, expanding the feature dimensions of the prediction model in step 3 and increasing the AUC value.

[0151] Real-time data (such as intraoperative HMGB1) allows for early warning of complications during surgery (compared to traditional postoperative assessment), reducing the incidence of serious postoperative complications.

[0152] Example 3, in step 2, the time is divided into several key time nodes;

[0153] The spatial dimension is defined as several types of related entities, and several core related pairs are formed by combining representative indicators of these types of related entities in pairs.

[0154] In Example 4, in step 2, the spatiotemporal correlation matrix uses several key time nodes as row labels and several core correlation pairs as column labels. The initial values ​​of the matrix are all 0. Subsequently, the correlation coefficients obtained through statistical calculations are used to fill the matrix cells to form a complete spatiotemporal correlation matrix.

[0155] Example 5, several key time points are: 24 hours before surgery, after anesthesia induction, 1 hour of surgery, at the end of surgery, and 24 hours after surgery;

[0156] The related subjects are categorized into three types: psychological state, pathological indicators, and physiological reactions.

[0157] Representative indicators of psychological state include anxiety assessment score, heart rate variability, and skin conductance response.

[0158] Representative pathological indicators include preoperative inflammatory factors, inflammation grade, and high-mobility group box 1 (HMP) levels.

[0159] Representative indicators of physiological responses include blood pressure fluctuation amplitude, heart rate variability, and BIS value.

[0160] By adopting the above technical solution, the following key points were selected: 24 hours before surgery, after anesthesia induction, 1 hour of surgery, at the end of surgery, and 24 hours after surgery. The design was based on the pathophysiological laws of gynecological general anesthesia surgery:

[0161] 24 hours before surgery: Capture the patient's baseline status (such as chronic inflammation level, baseline anxiety) to provide a reference for changes during surgery;

[0162] Post-anesthesia induction: Reflects the initial effects of anesthetic drugs on the physiological state, at which point the psychological-pathological-physiological interaction begins to emerge;

[0163] One hour after surgery: The pathological reaction caused by surgical trauma (such as the release of inflammatory factors) reaches its peak, which is a critical window for warning of the risk of complications;

[0164] At the end of the surgery: the point where the cumulative trauma effect and residual anesthetic drugs overlap is directly related to early postoperative complications;

[0165] 24 hours post-surgery: a turning point between tissue repair and infection risk, allowing for assessment of the effectiveness of intraoperative interventions;

[0166] The three categories of related subjects are divided into three categories: based on the psychological-pathological-physiological triple interaction mechanism, psychological state (anxiety) affects pathological indicators (such as inflammatory factors) through the neuro-endocrine-immune network, while pathological indicators and physiological reactions (such as blood pressure fluctuations) are directly related to the occurrence of complications, forming a complete chain of psychological stress, pathological changes, physiological disorders and complications.

[0167] Selection of representative indicators:

[0168] Psychological state: The Anxiety Assessment Scale (SAI) reflects the subjective level of anxiety, while heart rate variability and skin conductance serve as objective physiological evidence, achieving dual quantification of subjective and objective factors.

[0169] Pathological indicators: Preoperative inflammatory factors (IL-6 / CRP) reflect the basic inflammatory state, intraoperative inflammation grading reflects local tissue damage, and HMGB1 detection value assesses ischemia-reperfusion injury, covering three levels of pathological dimensions: systemic, local, and cellular.

[0170] Physiological responses: Blood pressure fluctuations and heart rate reflect circulatory stability, while BIS values ​​monitor the depth of anesthesia. These three factors together constitute the core indicators of anesthesia safety.

[0171] The core association pair formation logic adopts the principle of intra-subject validation and inter-subject cross-validation. Indicators of the same subject (such as heart rate variability, which belongs to both psychological and physiological aspects) are used to verify consistency, and cross-subject indicator combinations (such as anxiety assessment value-inflammatory factors-BIS value) capture interactive relationships, ultimately forming 9 sets of core association pairs to achieve comprehensive coverage of multi-dimensional influences.

[0172] The matrix uses time nodes as the vertical axis and core correlation pairs as the horizontal axis. It quantifies the dynamic correlation strength through correlation coefficients. Essentially, it transforms the abstract interaction of psychology, pathology, and physiology into a visualized mathematical model.

[0173] Time dimension: Capture the timeliness of the correlation by the changes in coefficients at different nodes (e.g., the effect of anxiety on inflammation is strongest 24 hours before surgery and weakens 24 hours after surgery).

[0174] Spatial dimension: Key drivers are identified by the coefficient differences of cross-subject association pairs (e.g., the association strength between intraoperative inflammation grade and blood pressure fluctuation is usually higher than that between anxiety and BIS value).

[0175] Example 6: Detailed statistical calculation process for the correlation coefficient:

[0176] Step a: Extract the associated pairs of data for the corresponding time periods from the standardized time series database according to the time nodes;

[0177] Step b: If the correlation pair is between continuous variables, first perform a normality test. If it conforms to a normal distribution, use the Pearson correlation coefficient to calculate it. If it is between a continuous variable and an ordinal variable, use the Spearman rank correlation coefficient. Rank correlation is calculated by sorting the data.

[0178] Step c: Perform a t-test on the calculated correlation coefficients and calculate the P-value. Repeat the above steps for each of the five key time points.

[0179] Step d involves performing clinical logic verification on the statistical results. If any results conflict with common medical knowledge, they are corrected based on the opinions of clinical experts.

[0180] By employing the above technical solution, correlation pairs are extracted from a standardized time-series database according to time nodes, essentially achieving data alignment through time anchoring. Since the duration and pace of gynecological surgeries vary among patients, relying solely on absolute time (e.g., 3 hours post-surgery) makes it difficult to standardize the analytical dimensions. However, using key nodes such as 24 hours pre-surgery and 1 hour post-surgery as benchmarks ensures the clinical comparability of correlation pairs at the same time node. For example, all patients' SAI scores and blood pressure data after anesthesia induction are at the same diagnostic and treatment stage.

[0181] Continuous variable × continuous variable (such as preoperative inflammatory factor IL-6 and blood pressure fluctuation): Pearson correlation coefficient is suitable for data that are linearly related and conform to a normal distribution. It can accurately quantify the strength of the linear association between variables (range -1 to 1). Its mathematical principle is based on the ratio of covariance to standard deviation, reflecting the degree of co-variable change.

[0182] Continuous variable × ordinal variable (such as SAI score and inflammation grade): Spearman's rank correlation coefficient avoids the discontinuous and non-normal characteristics of ordinal variables by converting the data into ranks. It is more suitable for describing the monotonic association between ordered categorical data and continuous data, such as the consistency between the trend of increased inflammation grade and the increase of HMGB1 test value.

[0183] The t-test, by calculating the t-statistic and p-value corresponding to the correlation coefficient, determines whether the association strength is caused by random error. When p is less than 0.05, the association is considered statistically significant, excluding interference from accidental factors. Calculations were performed separately for the five key nodes to capture the dynamic changes in association strength; for example, the association between inflammation grade and heart rate was significant 1 hour after surgery, but not 24 hours before surgery.

[0184] Statistical results may be affected by sampling bias, leading to conclusions that conflict with common medical knowledge, such as a strong correlation between elevated inflammation grade and decreased blood pressure. Clinical logic validation, by incorporating expert experience, distinguishes between true weak associations and erroneous associations, ensuring that statistical results conform to pathophysiological laws, such as the fact that inflammatory responses are usually accompanied by elevated blood pressure, thus achieving a balance between statistical significance and clinical rationality.

[0185] Example 7, in step 3, the model architecture of the complication prediction model includes an input layer, a fusion layer, and an output layer;

[0186] The input layer includes a static data channel, a time-series data channel, and an image data channel;

[0187] The fusion layer introduces an attention mechanism based on the spatiotemporal correlation matrix: dynamic attention weights are assigned to the indicators involved in the core correlation pairs;

[0188] The output layer is designed based on multi-task learning, which simultaneously predicts multiple types of complications and risk probabilities.

[0189] The Sigmoid function is used to predict the probability of a single complication, and the Softmax function is used to rank multiple complications.

[0190] By adopting the above technical solution, static data reflects the patient's basic condition, time-series data captures dynamic changes during surgery, and image data provides pathological morphological information. The three types of data complement each other to cover the complete feature dimensions required for prediction.

[0191] The attention mechanism is embedded with clinical priors and uses the association matrix to guide weight allocation, so that the model automatically focuses on clinically significant and strongly correlated features, such as the high correlation between inflammation grade and HMGB1, thus avoiding the black box defects of data-driven models.

[0192] The Sigmoid function independently predicts the probability of various complications, while the Softmax function implements risk ranking, balancing the accuracy of single indicators with the priority requirements of clinical decision-making.

[0193] Multimodal fusion improves the AUC value compared to a single data model, especially in terms of early warning sensitivity for complications, which reaches over 85%.

[0194] Attention weight visualization can directly display key information such as the contribution of inflammation grading features, which improves doctors' acceptance.

[0195] The lightweight design of LSTM and ResNet50 enables single-sample prediction to take less than 0.5 seconds, meeting the requirements for real-time intraoperative early warning.

[0196] By constraining the model's learning direction through the association matrix, the model's performance degradation rate is less than 5% during cross-center validation. This architecture achieves a dual-drive approach of data-driven and clinically knowledge-guided approaches, leveraging the feature extraction advantages of deep learning while ensuring that the prediction logic aligns with medical cognition through association rules, thus providing reliable support for clinical decision-making.

[0197] Example 8 also includes step 5, generating clinical intervention recommendations based on complication prediction results.

[0198] Example 9, Step 5, generating clinical intervention recommendations includes the following steps:

[0199] The complication prediction model outputs the probability of complication risk and the contribution of each feature. A contribution threshold is set to screen out the core risk factors, which are then sorted in descending order of contribution to form a priority list of risk factors.

[0200] Establish a risk factor and intervention mapping database, including: pathological indicator intervention, psychological state intervention, and physiological indicator intervention;

[0201] The corresponding intervention measures are retrieved from the mapping library based on the type of risk factor.

[0202] Example 10: Introducing gene polymorphism correction factors to adjust intervention measures in combination with the patient's basic characteristics.

[0203] By employing the above technical approach, it is understood that the occurrence of complications is the result of multiple factors working together, and the contribution of different risk factors varies significantly. For example, inflammation grade may have a greater impact on postoperative infection than heart rate fluctuations. By setting a contribution threshold, such as greater than or equal to 10%, core risk factors can be screened and ranked according to their contribution, focusing on the factors with the most significant impact on complications and avoiding the dispersion of intervention resources. Essentially, this approach uses interpretable algorithms such as SHAP values ​​to transform the importance of black-box features output by the model into clinically understandable risk weights, ensuring that interventions directly target the core causes.

[0204] The core of establishing a risk factor-intervention mapping library is causal association matching based on evidence-based medicine. For example:

[0205] The mapping between pathological indicators, such as inflammation grade ≥2, and anti-inflammatory measures (glucocorticoids), stems from the pathological mechanisms of excessive release of inflammatory factors, tissue damage, and complications.

[0206] The mapping between psychological state, such as a SAI score greater than 40, and sedative medication is based on the physiological logic of anxiety, sympathetic nerve excitation, elevated stress hormones, and increased risk of complications.

[0207] The mapping between physiological indicators and vasoactive drugs is based on the clinical consensus of hemodynamic instability, insufficient organ perfusion, and delayed recovery.

[0208] The establishment of a mapping library ensures a direct correlation between intervention measures and risk causes, avoiding indiscriminate medication or procedures.

[0209] Genetic polymorphisms (such as COMTval158Met and HTR2Ars6311) affect drug metabolism efficiency and neuroendocrine responses, leading to significant differences in the effectiveness of the same intervention in different patients (e.g., patients with the Met / Met genotype are more sensitive to opioids). By adjusting for patient baseline characteristics (age, BMI, underlying diseases), the group-based intervention plan can be further refined into a personalized approach, aligning with the core principle of precision medicine: optimizing treatment based on individual biological characteristics.

[0210] By screening core risk factors, interventions for non-critical factors can be reduced by 30%-40%, avoiding the waste of broad-based medical resources.

[0211] Interventions based on evidence-based mapping libraries are directly matched with risk factors. For example, the use of dexamethasone in patients with grade 3 inflammation can reduce the incidence of postoperative agitation by more than 40%. After combining gene polymorphism adjustment, the incidence of adverse drug reactions (such as nausea and vomiting) is further reduced.

[0212] The structured intervention suggestion generation process significantly shortens doctors' decision-making time, especially in emergency intraoperative scenarios (such as sudden blood pressure fluctuations), it can quickly identify core risks and push corresponding measures, buying time for rescue.

[0213] Combining gene-correcting factors with adjustments to baseline characteristics allows intervention programs to be tailored to the physiological characteristics of different patients. For example, reducing the opioid dosage by 20% in elderly patients with COMTet / Met type opioids can maintain analgesic efficacy while reducing the risk of respiratory depression, improving clinical suitability by more than 50%.

[0214] The following describes the implementation principle of the present invention using specific embodiments:

[0215] The patient is a 45-year-old female who is scheduled to undergo laparoscopic ovarian cystectomy. She is classified as ASA II and has no history of hypertension or diabetes. She has not used any hormone medications before the operation.

[0216] Step 1: Multi-dimensional data collection:

[0217] Preoperative data:

[0218] Basic clinical indicators: age 45 years, BMI 23.5 kg / m², gynecological disease type benign cyst, no history of pelvic surgery, and no history of allergy to anesthetic drugs.

[0219] Pathological molecular markers: IL-6 was 12.3 pg / ml (electrochemiluminescence method) and CRP was 8.5 mg / L (high-sensitivity latex method) 18 hours before surgery; HTR2A genotype was C / C and COMT genotype was Met / Met.

[0220] Psychological status: 24h preoperative SAI score of 42 (collected via WeChat mini program), and simultaneous HRV test showed LF / HF=2.1 (high sympathetic nerve activity).

[0221] Intraoperative data:

[0222] Real-time physiological indicators: BIS value 45 after anesthesia induction, blood pressure fluctuation range 8 mmHg; heart rate 78 beats / min 1 hour after surgery, propofol dose 4 mg / kg / h.

[0223] Intraoperative pathological indicators: Frozen sections were completed 25 minutes after adnexal resection, and the inflammation grade was grade 2 (5-10 inflammatory cells / HPF) after blind review by two pathologists; the HMGB1 value was 6.2 ng / ml 1 hour after surgery (microfluidic chip method).

[0224] Surgical parameters: Laparoscopic surgery, duration 95 minutes, blood loss 50ml.

[0225] Postoperative data:

[0226] Complication record: 6 hours postoperatively, Riker scale score 4 (no agitation), PONV grade 1 (mild nausea), VAS score 3 (no intervention required).

[0227] Repair biomarkers: VEGF = 350 pg / ml at 24 h post-surgery, 48 h / 24 h ratio = 1.1 (no repair delay). Step 2: Spatiotemporal correlation matrix construction:

[0228] Key time points and core correlation pairs: 24 hours before surgery, after anesthesia induction, 1 hour of surgery, at the end of surgery, and 24 hours after surgery were selected. The core correlation pairs include 9 groups such as SAI score, IL-6 inflammation grade, and blood pressure fluctuation.

[0229] Correlation coefficient calculation:

[0230] The inflammation grade (grade 2) and blood pressure fluctuation range at 1 hour after surgery were calculated as a continuous variable × ordinal variable, and the Spearman coefficient was 0.62 (P < 0.01).

[0231] The 24-hour preoperative SAI score (42) - IL-6 (12.3 pg / ml) was a continuous variable × continuous variable. After the normality test was met, the Pearson coefficient was 0.58 (P < 0.05).

[0232] Clinical validation: All correlation coefficients are consistent with the medical logic that increased inflammation leads to increased blood pressure fluctuations, and there are no conflicting results.

[0233] Step 3: Complication prediction model output:

[0234] Multimodal fusion: Input preoperative genetic data (COMTMet / Met type), intraoperative pathological images (inflammation grade 2), and temporal physiological signals (blood pressure fluctuations). The attention mechanism of the fusion layer assigns a weight of 0.32 to the inflammation grade-HMGB1 association pair.

[0235] Predicted results: Postoperative agitation risk probability 18%, nausea and vomiting risk probability 42% (Sigmoid function); the order of multiple complications is nausea and vomiting > pain > agitation (Softmax function).

[0236] Steps 4-5: Generation of clinical intervention recommendations:

[0237] Core risk factors: Factors with a contribution of ≥10% were selected based on SHAP value, ranked as follows: HMGB1 (6.2ng / ml, 35%) > SAI score (42 points, 28%).

[0238] Intervention measures:

[0239] Call HMGB1>6ng / ml from the mapping library → monitor PCT (pathological index intervention) 6h postoperatively.

[0240] Based on the COMTMet / Met genotype (opioid sensitivity), the postoperative analgesia regimen was adjusted by reducing the morphine dose by 20%.

[0241] The results of comparing the technical effectiveness of this solution with traditional complication prediction solutions are shown in Table 1:

[0242] Table 1

[0243] index This technical solution Traditional method (Logistic regression) Traditional method (Apfel scoring) AUC (Prediction Accuracy) 0.92 0.75 0.68 Early warning sensitivity 85% 60% 55% Incidence of serious postoperative complications 4.2% 11.5% 13.8% Clinical intervention decision time 2.1 minutes 8.5 minutes 10.3 minutes Adverse drug reaction rate 3.5% (genetic adjustment) 9.8% 10.2%

[0244] This technical solution, through multi-dimensional data fusion, dynamic correlation analysis, and individualized intervention, significantly outperforms traditional methods in terms of prediction accuracy, clinical efficiency, and safety.

[0245] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis, characterized in that, Includes the following steps: Step 1: Collect multi-dimensional pathological and clinical data before, during, and after gynecological general anesthesia. Step 2: Establish a spatiotemporal correlation matrix of anxiety, intraoperative pathology, and complications; Step 3: Construct a prediction model for postoperative complications of gynecological general anesthesia based on a multimodal deep learning prediction architecture using a spatiotemporal correlation matrix; Step 4: The complication prediction model takes into account the patient's preoperative, intraoperative, and postoperative multidimensional pathological and clinical data, and outputs the types of complications and their probability of occurrence. In step 2, the time is divided into several key time nodes; The spatial dimension is defined as several types of related entities, and several core association pairs are formed by pairwise combinations of representative indicators of these types of related entities. The spatiotemporal association matrix uses several key time nodes as row labels and several core association pairs as column labels. The initial values ​​of the matrix are all 0, and the matrix cells are filled with correlation coefficients obtained through statistical calculations to form a complete spatiotemporal association matrix. In step 3, the model architecture of the complication prediction model includes an input layer, a fusion layer, and an output layer; The input layer includes a static data channel, a time-series data channel, and an image data channel; The fusion layer introduces an attention mechanism based on the spatiotemporal correlation matrix: dynamic attention weights are assigned to the indicators involved in the core correlation pairs; The output layer is designed based on multi-task learning, which simultaneously predicts multiple types of complications and risk probabilities. The Sigmoid function is used to predict the probability of a single complication, and the Softmax function is used to rank multiple complications.

2. The method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis according to claim 1, characterized in that, The multidimensional pathological and clinical data in step 1 before gynecological general anesthesia include basic clinical indicator data, pathological molecular marker detection data, and psychological state quantitative data. Multidimensional pathological and clinical data during gynecological general anesthesia include real-time physiological indicators, intraoperative pathological indicators, and surgical parameters. Multidimensional pathological and clinical data following gynecological general anesthesia include complication records and data on repair and infection marker detection.

3. The method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis according to claim 2, characterized in that, Several key time points are: 24 hours before surgery, after anesthesia induction, 1 hour of surgery, at the end of surgery, and 24 hours after surgery; The related subjects are categorized into three types: psychological state, pathological indicators, and physiological reactions. Representative indicators of psychological state include anxiety assessment score, heart rate variability, and skin conductance response. Representative pathological indicators include preoperative inflammatory factors, inflammation grade, and high-mobility group box 1 (HMP) levels. Representative indicators of physiological responses include blood pressure fluctuation amplitude, heart rate variability, and BIS value.

4. The method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis according to claim 3, characterized in that, The specific statistical calculation process for the correlation coefficient is as follows: Step a: Extract the associated pairs of data for the corresponding time periods from the standardized time series database according to the time nodes; Step b: If the correlation pair is between continuous variables, first perform a normality test. If it conforms to a normal distribution, use the Pearson correlation coefficient to calculate it. If it is between a continuous variable and an ordinal variable, use the Spearman rank correlation coefficient. Rank correlation is calculated by sorting the data. Step c: Perform a t-test on the calculated correlation coefficients and calculate the P-value. Calculate the P-value for each of the five key time nodes according to steps a to c. Step d involves performing clinical logic verification on the statistical results. If any results conflict with common medical knowledge, they are corrected based on the opinions of clinical experts.

5. The method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis according to claim 4, characterized in that, It also includes step 5, which generates clinical intervention recommendations based on the complication prediction results.

6. The method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis according to claim 5, characterized in that, Step 5, generating clinical intervention recommendations, includes the following steps: The complication prediction model outputs the probability of complication risk and the contribution of each feature. A contribution threshold is set to screen out the core risk factors, which are then sorted in descending order of contribution to form a priority list of risk factors. Establish a risk factor and intervention mapping database, including: pathological indicator intervention, psychological state intervention, and physiological indicator intervention; The corresponding intervention measures are retrieved from the mapping library based on the type of risk factor.

7. The method for predicting and modeling postoperative complications of gynecological general anesthesia based on pathological data analysis according to claim 6, characterized in that, Introduce gene polymorphism correction factors and adjust intervention measures in combination with the patient's basic characteristics.

Images