Digital intelligent anesthesia safety management method and related equipment

Abstract

The invention discloses a digital intelligent anesthesia safety management method and related equipment. The method comprises the steps of collecting an operation scene, medicine infusion data, anesthesia machine ventilation parameters, vital sign parameters and patient basic characteristics of an operation patient; based on the operation scene, the medicine infusion data, the ventilation parameters, the vital sign parameters and the basic features of the patient, the risk level of the operation patient is evaluated; and performing graded alarm according to the risk grade of the surgical patient obtained by evaluation. The problems of low early warning accuracy, weak decision support capability, poor interactivity and insufficient emergency response efficiency caused by false alarm or delayed alarm of an existing anesthesia monitoring system in a complex operation scene can be solved, and meanwhile, high dependence on judgment of an anesthetist is reduced; therefore, the restriction on the improvement of clinical anesthesia safety and quality due to the huge gap of anesthetists is relieved.

Description

Technical Field

[0001] This application relates to the field of smart healthcare, and more specifically, to a digital and intelligent anesthesia safety management method and related equipment. Background Technology

[0002] With the advancement of medical technology, surgery has become an important means of treating diseases. During surgery, anesthesia management is directly related to patient safety and surgical outcomes. Traditional anesthesia management mainly relies on medical staff to observe and monitor vital signs such as heart rate, blood pressure, and respiratory rate. However, this method has obvious limitations: on the one hand, it depends on the clinical experience of medical staff, and on the other hand, it has limited ability to identify complex cases or potential risks. Summary of the Invention

[0003] The summary section introduces a series of simplified concepts, which will be further explained in detail in the detailed description section. The summary section of this invention is not intended to limit the key features and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.

[0004] To address the shortcomings of existing anesthesia monitoring systems in terms of early warning accuracy, decision support capabilities, and emergency response efficiency in complex surgical scenarios, which hinder the improvement of clinical anesthesia safety and quality, this invention proposes, firstly, a digital and intelligent anesthesia safety management method, comprising: Collect surgical scene, drug infusion data, ventilation parameters, vital signs parameters and basic patient characteristics of surgical patients; Based on the surgical scenario, drug infusion data, ventilation parameters, vital sign parameters, and patient baseline characteristics, the risk level of the surgical patient is assessed. A tiered alarm is triggered based on the risk level of the surgical patient as determined by the assessment.

[0005] Optionally, the assessment of the surgical patient's risk level based on the surgical scenario, drug infusion data, ventilation parameters, vital sign parameters, and patient baseline characteristics includes: Based on the surgical scenario, drug infusion data, ventilation parameters, vital sign parameters, and patient baseline characteristics, a proactive risk assessment model is used to predict potential anesthesia risks in order to assess the risk level of the surgical patient.

[0006] Optionally, the step of predicting potential anesthesia risks using an active risk assessment model based on the surgical scenario, drug infusion data, ventilation parameters, vital sign parameters, and patient baseline characteristics to assess the risk level of the surgical patient includes: Based on the surgical scenario, drug infusion data, ventilation parameters, vital sign parameters, and patient baseline characteristics, a proactive risk assessment model is used to predict multi-dimensional risk indicators to establish a dynamic risk matrix. The dynamic risk matrix includes real-time risk values, risk evolution trends, and a list of abnormal related physiological parameters.

[0007] Optional, also includes: Intelligent decision-making suggestions are generated based on the aforementioned multi-dimensional risk indicators.

[0008] Optional, also includes: Record the actual actions taken by doctors upon receiving the intelligent decision-making suggestions; When there is a deviation between the actual processing operation and the intelligent decision-making suggestion, the deviation event is recorded so as to carry out reinforcement learning of the risk assessment model and the decision-making model based on the deviation event.

[0009] Optional, also includes: The surgical scenario, drug infusion data, and ventilation parameters identify the current surgical anesthesia stage, and the anesthesia safety alarm threshold for vital signs parameters is dynamically calculated based on the current surgical anesthesia stage.

[0010] Secondly, the present invention also proposes a digital intelligent anesthesia safety management device, comprising: The data acquisition unit is used to collect data on the surgical scene, drug infusion, ventilation parameters, vital signs, and basic characteristics of the surgical patient. The analysis unit is used to assess the risk level of the surgical patient based on the surgical scenario, drug infusion data, ventilation parameters, vital sign parameters, and patient baseline characteristics. The decision-making unit is used to issue graded alarms based on the risk level of the surgical patient obtained from the assessment.

[0011] Thirdly, an electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program stored in the memory to implement the steps of the digital anesthesia safety management method as described in any of the first aspects above.

[0012] Fourthly, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the digital and intelligent anesthesia safety management method of any of the above claims in the first aspect.

[0013] In summary, the digital anesthesia safety management method proposed in this application collects the surgical patient's drug infusion data, ventilation parameters, and vital sign parameters; identifies the current surgical anesthesia stage based on the drug infusion data and vital sign parameters; and dynamically calculates the anesthesia safety alarm threshold for vital sign parameters according to the current surgical anesthesia stage. Considering that existing systems use a fixed threshold alarm mechanism, failing to take into account individual differences among patients such as whether they have underlying diseases like hypertension, age, and gender, and the varying tolerance requirements for physiological parameter fluctuations at different stages of the anesthesia process, such as the induction period, surgical stimulation period, or recovery period, resulting in low alarm accuracy, this solution operates the entire system in a dynamic feedback closed loop. Whenever vital signs change significantly or the stage state changes, the threshold calculation model and risk assessment engine are recalculated accordingly, ensuring that the alarm response mechanism always matches the patient's current state. Furthermore, ventilation parameters, especially end-expiratory CO2 partial pressure (EtCO2), airway pressure, and airway resistance, often change in the early stages of a crisis, even earlier than traditional vital sign changes such as SpO2, HR, and SBP, exhibiting extremely high predictive sensitivity. By introducing a real-time stage recognition and dynamic threshold adaptive adjustment mechanism, this method significantly reduces the false alarm rate of traditional threshold-based alarms and improves the ability to provide early warning of important clinical events. It is especially suitable for the precise management of potential anesthetic complications in high-risk patients or during long-term surgeries.

[0014] The digital and intelligent anesthesia safety management method of the present invention, and other advantages, objectives and features of the present invention will be partly apparent from the following description, and partly understood by those skilled in the art through study and practice of the present invention. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit this specification. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This application provides a schematic diagram of a digital and intelligent anesthesia safety management method. Figure 1a A schematic flowchart illustrating another digital and intelligent anesthesia safety management method provided in this application embodiment; Figure 2 A schematic diagram of a digital anesthesia safety management device provided in this application embodiment; Figure 3 This is a schematic diagram of the structure of an electronic device for digital anesthesia safety management provided in an embodiment of this application. Detailed Implementation

[0016] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. The technical solutions of the embodiments of this application will now be clearly and completely described in conjunction with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0017] Currently, there is a lack of effective coordination between monitor alarm data and surgical procedures, anesthesia machine, and patient clinical status, leading to incorrect or delayed risk assessment. For example, if the monitor indicates arrhythmia, it may actually be caused by electrosurgical interference, but the system cannot verify this in conjunction with the status of the electrosurgical unit in use. Similarly, after a breathing circuit detaches, an alarm is only triggered after SpO2 drops to a threshold, but real-time airway pressure changes from the anesthesia machine are not included in the warning logic. Existing research largely focuses on monitor data analysis, but it lacks sufficient integration of key factors affecting surgical procedures, such as airway safety factors like catheter displacement and pneumothorax, drug interactions like excessive anesthesia depth, and hemodynamic changes caused by pneumoperitoneum, making it difficult to cover complex intraoperative scenarios. When risks occur, only simple alarms are provided, lacking personalized management recommendations based on individual patient characteristics.

[0018] These technological shortcomings collectively result in insufficient accuracy, timeliness, and decision support in early warning systems for complex surgical scenarios, hindering the improvement of clinical anesthesia safety and quality. Against this backdrop, constructing a data-integrated, intelligent anesthesia safety management and early warning system to achieve earlier risk detection, more accurate event identification, and more intelligent decision support is particularly important. To address these issues, such as... Figure 1 The present invention provides a digital and intelligent anesthesia safety management method, including steps S110 to S130.

[0019] S110 collects surgical scene data, drug infusion data, ventilation parameters, vital sign parameters and basic patient characteristics of surgical patients; S120, based on the surgical scenario, drug infusion data, ventilation parameters, vital sign parameters and patient baseline characteristics, assess the risk level of the surgical patient; S130, tiered alarms are generated based on the risk level of the surgical patient as determined by the assessment.

[0020] like Figure 1a and Figure 2 As shown, one embodiment of the digital anesthesia safety management device in this application may include: The data acquisition unit 21 is used to collect surgical scene, drug infusion data, ventilation parameters, vital sign parameters and basic characteristics of surgical patients. Analysis unit 22 is used to assess the risk level of the surgical patient based on the surgical scenario, drug infusion data, ventilation parameters, vital sign parameters and patient baseline characteristics; Decision unit 23 is used to issue graded alarms based on the risk level of the surgical patient obtained from the assessment.

[0021] For example, the acquisition unit is used to collect and integrate the surgical patient's surgical condition, medication infusion data, ventilation parameters, monitor data, and basic patient characteristics; the analysis unit is used to dynamically analyze existing and potential risks and issue graded alarms based on the surgical condition, medication infusion data, ventilation parameters, monitor data, and basic patient characteristics; the decision-making unit is used to provide corresponding treatment suggestions based on the evaluation results of the analysis unit. For example, the acquisition unit uses intelligent data acquisition and fusion technology to achieve real-time synchronous acquisition of multi-dimensional physiological parameters and medical equipment status during surgery. It covers all elements of information from patient vital signs to medical equipment operation, ensuring the comparability of equipment data at different sampling frequencies, and includes a built-in encrypted transmission mechanism for patient data.

[0022] For example, the analysis unit achieves risk analysis by integrating multi-dimensional data from monitoring equipment, anesthesia equipment, and the surgical site in real time, overcoming the limitations of traditional single-parameter alarms. When any medical device detects an abnormal physiological parameter, the unit automatically activates the collaborative verification mechanism of the associated devices, combining the patient's individual characteristics with the status of the surgical process to achieve essential identification of clinical critical events. For example, when abnormal parameters occur in devices such as monitors and anesthesia machines, the analysis unit first identifies the type and characteristics of the abnormal parameters and automatically matches the relevant device categories to be invoked. For instance, when pulse oximetry saturation decreases, the system prioritizes associating anesthesia machine ventilation parameters and surgical video data. The authenticity of the abnormality is verified by comparing physiological parameters collected from different devices with device status data. Typical scenarios include verifying whether a decrease in pulse oximetry saturation stems from a breathing circuit problem by checking changes in anesthesia machine airway pressure, or confirming whether an abnormal electrocardiogram is related to electrosurgical interference by checking real-time surgical video. After completing data verification between devices, the unit combines patient history, surgical procedures, and other information to output a structured conclusion including the nature of the event, severity level alarm, and confidence level. For example, it distinguishes whether hypotension is caused by excessive anesthesia, acute bleeding, or an allergic reaction. The decision-making unit transforms the structured conclusions output by the analysis unit, such as event type, severity, and confidence level, into actionable treatment plans. The decision-making unit operates according to a process of receiving, matching, and generating recommendations: it acquires the structured conclusions from the analysis unit in real time, including validated clinical event types, severity levels, and system confidence scores. Based on a built-in clinical knowledge base, it automatically matches treatment plans related to the event and personalizes them according to individual patient characteristics. It generates a complete plan including immediate measures, follow-up observation points, smart device linkage, and remote expert consultation. In some examples, the system consists of an acquisition unit, an analysis unit, and a decision-making unit, interconnected via a dedicated communication protocol. The acquisition unit, acting as the system's data perception layer, achieves standardized integration of data from different medical devices in the operating room through multimodal data synchronous acquisition and intelligent preprocessing technology. This unit first establishes an acquisition system covering five major data categories: the surgical scene is captured in real time by a high-definition operating room camera system, including real-time assessment of bleeding volume and tissue perfusion status; instrument operations such as electrosurgical usage and suction rate; physical characteristics such as patient facial color and endotracheal tube position; and the behavior of all personnel in the operating room. Video streams are encrypted via algorithms to ensure patient privacy and security. Medication infusion data is accurately recorded, including drug type, concentration, and real-time infusion rate, with a reading calibration every 5 seconds to ensure cumulative dose accuracy. Anesthesia machine ventilation parameters are sampled at high frequency, comprehensively acquiring key indicators such as airway pressure (Paw), end-tidal carbon dioxide (EtCO2), airway resistance (Raw), tidal volume (VT), and respiratory rate (RR). Monitor data synchronously receives vital signs such as heart rate (HR), blood pressure (SBP / DBP), and bispectral index (BIS). Patient baseline information is automatically obtained through the electronic medical record system interface, including ASA classification and medical history.

[0023] For example, to address the challenge of data synchronization between devices with different sampling frequencies, the acquisition unit employs a solution combining dynamic time windows and interpolation compensation algorithms. By setting adaptive buffers for devices with different frequencies, such as monitors, anesthesia machines, and video streams, and using cubic spline interpolation to compensate for low-frequency data and optical flow analysis to correct video frame timing, multi-source data is ultimately unified to the same time axis, achieving phase consistency matching such as pulse oximetry (SpO2) and end-tidal carbon dioxide (EtCO2) waveforms.

[0024] For example, the frequencies of each device are 1Hz for the monitor, 10Hz for the anesthesia machine, and 25fps for the video stream. The 1-second interval data from the monitor is matched with 10 sampling data points from the anesthesia machine on a unified time axis to ensure phase consistency between end-tidal carbon dioxide (EtCO2) and blood oxygen saturation (SpO2). The system precisely correlates the visual signal activated by the electrosurgical unit with the interference peak of the electrocardiogram (ECG) using image timestamps, thereby verifying the effectiveness of data synchronization between devices.

[0025] For example, the analysis unit performs real-time risk analysis through a closed-loop system that integrates multi-dimensional data from monitoring equipment, anesthesia equipment, and the surgical site, overcoming the limitations of traditional single-parameter alarms. The workflow follows the principles of triggering, verification, and discrimination: When a first-class device, such as a monitor, detects an abnormal parameter like the disappearance of ECG heart rate, the unit automatically retrieves a pre-defined list of associated devices, such as anesthesia machines and surgical imaging systems. Real-time data from a second-class device is retrieved for cross-verification, such as whether the EtCO2 waveform of the anesthesia machine is continuous and whether the electrosurgical unit is activated in the video. If the associated data supports the authenticity of the abnormality, such as the simultaneous disappearance of EtCO2, it is confirmed as a clinical event, and a structured conclusion is output, including event type, severity, and confidence level. If there are contradictions, such as the electrosurgical unit interfering with the ECG but the SpO2 pulse wave being normal, it is marked as device interference, and false alarms are suppressed.

[0026] It should be noted that the system establishes an airway safety early warning system through multi-parameter dynamic monitoring of the anesthesia machine, breaking through the traditional lagging alarm mode that relies on pulse oxygen saturation. It performs real-time analysis of the three-dimensional correlation between expiratory and inspiratory tidal volume balance (VTe / VTi), airway pressure-flow curve morphology, and EtCO2 waveform characteristics. When it detects a characteristic change of a sudden drop in expiratory tidal volume accompanied by airway pressure abnormalities and the disappearance of the end-expiratory carbon dioxide waveform, the system can immediately identify respiratory circuit abnormalities even if SpO2 has not yet decreased.

[0027] For example, when the system detects a sudden drop in expiratory tidal volume (VTe) from 450 ml to 120 ml and a significant imbalance with the inspiratory tidal volume (VTi), it simultaneously verifies the loss of fluctuation in the airway pressure curve, such as a drop in peak pressure from 18 cmH2O to 3 cmH2O, and the disappearance of the end-expiratory CO2 waveform, such as a drop from 35 mmHg to 5 mmHg. These synchronous abnormalities in these three anesthesia machine-specific indicators allow the system to determine respiratory circuit disengagement even before SpO2 decreases, providing a significantly earlier warning than traditional methods and resolving the lag problem of traditional single-parameter monitoring.

[0028] For example, the system dynamically classifies the severity of risks into three levels: Class I low-risk events involve minor deviations in parameters, such as blood pressure fluctuations within ±20% of baseline or transient abnormalities, such as a transient drop in SpO2 to 88%, requiring only log recording and a flashing alert; Class II moderate-risk events require a single parameter with a moderate abnormality, such as SpO2 between 85% and 89% lasting for 1 minute, and the presence of overlapping risk factors, such as ST-segment elevation of 1 mm in patients with coronary artery disease, triggering a pop-up prompt requiring manual confirmation; Class III high-risk events involve extreme abnormalities, such as SpO2 <80% lasting for 30 seconds or multiple system abnormalities, such as SpO2 <90% accompanied by blood pressure <60 mmHg, immediately triggering an audible and visual alarm. The classification logic further integrates individual patient characteristics for threshold correction.

[0029] For example, the SpO2 alarm threshold for ASA stage III patients is automatically raised by one level, meaning that when SpO2 in COPD patients drops to 85%, a level I response is initiated, while the same indicator in healthy patients only triggers a level II response. The final output confidence score quantifies the certainty of the system's judgment using a weighted model (0.4 × data consistency + 0.3 × device synergy + 0.3 × historical concordance rate). When the confidence score is ≥90%, direct device pretreatment, such as pausing the infusion pump, is initiated; at 70%-90%, explicit treatment recommendations are generated; and at <70%, observational recommendations are generated, forming a tiered decision support system.

[0030] For example, when the system detects isolated blood pressure fluctuations, such as with a confidence level of 65%, it only displays a gray prompt box on the working interface and does not trigger an audible or visual alarm.

[0031] In some examples, the decision unit serves as the system's clinical intervention engine, transforming the structured conclusions of the analysis unit into actionable solutions. The unit incorporates a dynamic knowledge base, integrating the latest clinical guidelines and institutional treatment protocols, and automatically adjusts recommendations based on individual patient characteristics such as ASA classification and liver and kidney function. The output recommendations include comprehensive solutions encompassing immediate interventions, follow-up observation points, smart device integration, and remote expert consultation. In smart device integration, all device control commands undergo double confirmation: the doctor clicks to confirm the operation suggestion, and the system displays a secondary confirmation dialog box showing operation details and the patient's real-time status. Each operation generates a complete audit log, including the operator, timestamp, and execution deviations.

[0032] For example, when the endotracheal tube is dislodged, the system detects a synchronous decrease in SpO2 and EtCO2 with a confidence level of 96%, classifying it as a Level III event. The decision unit displays: High risk and ventilation failure. Immediately perform manual ventilation and check the tube position. Equipment recommendations requiring confirmation are generated: pause propofol infusion and switch the anesthesia machine to standby mode. The system will only send the pause command to the infusion pump after the physician clicks to pause the propofol infusion.

[0033] Understandably, the system automatically presets the status of related equipment, such as pausing the infusion pump and adjusting anesthesia machine parameters, based on the treatment suggestions output by the decision-making unit. However, these actions require secondary confirmation from medical staff before execution to ensure safety and maintain human control. After medical staff respond to the system's suggestions via mobile devices or operating room terminals, the execution results, such as drug dosage adjustments and airway manipulations, are transmitted back to the data acquisition unit in real time, forming a new monitoring data stream. The system continuously tracks changes in physiological parameters after execution, such as the rate of blood pressure recovery and SpO2 recovery time. If the expected effect is not achieved, it triggers an upgrade suggestion or recommends alternative solutions, such as switching from laryngeal mask airway adjustment to endotracheal intubation. The treatment effects are compared with historical cases to optimize the confidence calculation model and suggestion priority for similar scenarios in the future.

[0034] For example, when the system determines that a patient's SpO2 drop due to laryngeal mask airway displacement is a Grade III event with a confidence level of 92%, it automatically executes the following closed-loop procedure: pausing the infusion pump, switching to manual ventilation mode, and executing after doctor confirmation. After the doctor adjusts the laryngeal mask airway position, the procedure is marked as completed on the terminal; if the system monitors and the SpO2 recovers to above 95% within 2 minutes, the alarm is deactivated; if there is no improvement, endotracheal intubation is recommended; the time and effect of this procedure are recorded to optimize the confidence calculation weight for future laryngeal mask airway displacement events.

[0035] For example, when the operating room camera detects projectile bleeding (bleeding rate > 200 ml / min) in the surgical field, it immediately compares in real time the blood pressure drop curve on the monitor (e.g., SBP drops by 30% within 5 minutes), the precipitous drop in EtCO2 of the anesthesia machine, and the surgical progress information (currently in the liver vessel separation stage). The system completes cross-validation within 3 seconds and determines it as a Grade III major hemorrhage event (95% confidence). The decision-making unit simultaneously triggers three emergency measures: pushes a treatment guide containing a diagram of the compression hemostasis points to the operating room, automatically sends an emergency blood matching request (including the patient's blood type and estimated blood volume) to the blood bank through the hospital information system, and simultaneously calls the anesthesia resuscitation team. The system continuously tracks the slope of blood pressure recovery after transfusion. If the expected recovery curve (SBP > 90 mmHg) is not reached within 15 minutes, a higher-level transfusion plan is generated, and the infusion pump is linked to adjust the crystalloid / colloid infusion ratio. The timeline data of the entire event (from bleeding detection to blood pressure stabilization) is encrypted and recorded to optimize the warning threshold and treatment strategy for similar scenarios in the future.

[0036] For example, when the acquisition unit detects a sudden increase in airway resistance (from 15 to 28 cmH2O) accompanied by a characteristic sawtooth EtCO2 waveform on the anesthesia machine, while the blood pressure (82 / 45 mmHg) and heart rate (115 beats / min) on the monitor have not yet reached the alarm threshold, the analysis unit, combined with images captured by the operating room camera showing the infusion of antibiotics and the patient's systemic rash, determines that the patient has experienced anaphylactic shock. This triggers a Level III high-risk event warning (89% confidence level), and the decision-making unit immediately initiates a multi-device collaborative treatment process. The system first sends a command to the infusion pump to pause the current medication, and simultaneously displays a warning message on the anesthesia machine interface: "High-risk alarm: Anaphylactic shock suspected, switching to manual ventilation mode is recommended." After the medical staff clicks to confirm, the system further displays a secondary verification dialog box, listing the patient's real-time vital signs (blood pressure 82 / 45 mmHg, heart rate 115 beats / min) and the proposed procedure (intravenous injection of 0.1 mg of epinephrine). After the doctor's second confirmation, the dedicated epinephrine drawer in the smart medicine cabinet automatically pops open and illuminates with a red light, and the infusion pump simultaneously switches to the preset epinephrine infusion channel (concentration 0.1 mg / ml). All operational steps (including the initial click time, final execution time, actual infusion rate, and deviation from the recommended value) are encrypted and recorded in the blockchain audit log. If the system detects a critical operational delay (such as epinephrine not being dispensed for more than 3 minutes), it will automatically trigger a broadcast call to the highest-ranking physician.

[0037] The digital anesthesia safety management method provided in this application further includes: collecting drug infusion data, ventilation parameters, and vital sign parameters of surgical patients; identifying the current surgical anesthesia stage based on the drug infusion data and vital sign parameters; and dynamically calculating the anesthesia safety alarm threshold of vital sign parameters according to the current surgical anesthesia stage. Considering that existing systems use fixed threshold alarm mechanisms, such as triggering an alarm when systolic blood pressure <90 mmHg, they fail to consider individual differences among patients, such as whether they have underlying diseases like hypertension, age, gender, etc., and the varying tolerance requirements for physiological parameter fluctuations at different stages of the anesthesia process, such as the induction period, surgical stimulation period, or recovery period, resulting in low alarm accuracy. Compared to existing methods, this solution continuously operates in a dynamic feedback closed loop. Whenever vital signs change significantly or the stage state switches, the threshold calculation model and risk assessment engine are recalculated to ensure that the alarm response mechanism always matches the patient's current state. By introducing a real-time stage identification and dynamic threshold adaptive adjustment mechanism, this method significantly reduces the false alarm rate of traditional threshold-based alarms and improves the ability to provide early warning of important clinical events, making it particularly suitable for the precise management of potential anesthetic complications in high-risk patients or during long surgeries.

[0038] For example, the system integrates with medical devices such as anesthesia pumps, monitors, and anesthesia information management systems (AIMS) to continuously collect multi-source data from surgical patients. This includes, but is not limited to, anesthetic drug infusion information such as drug type, infusion rate, and start / stop time; key vital signs parameters such as heart rate, systolic and diastolic blood pressure, SpO2, end-tidal carbon dioxide concentration, respiratory rate, and depth of anesthesia index (BIS); and basic patient characteristics such as age, gender, BMI, ASA score, and past medical history. All data is cached in a dynamic monitoring database according to a unified timestamp format, and a high-dimensional feature vector sequence arranged by a time sliding window is constructed to provide input for subsequent stage discrimination and model calculation.

[0039] For example, based on the aforementioned data sequence, the system automatically identifies the patient's current anesthesia stage using time-series feature extraction-based classification algorithms such as Bi-LSTM networks or Transformer models incorporating attention mechanisms. During system training, clinically labeled standard anesthesia stage data is incorporated to extract characteristic patterns: the induction phase is typically characterized by a sudden increase in drug infusion, a decrease in blood pressure, and a rapid decline in BIS values; the maintenance phase is characterized by stable vital signs with minor fluctuations and stable drug dosage; and the recovery phase is often characterized by drug reduction / discontinuation, a rebound in BIS, and increased fluctuations in blood pressure and heart rate. By using the trained model to annotate real-time data, the system achieves accurate identification of the current anesthesia state, providing core criteria for subsequent personalized threshold adjustments.

[0040] For example, based on the identified anesthesia stage information, individual patient parameters, and dynamic physiological indicator trends, the system calculates the safety thresholds for various vital signs in real time using a built-in personalized physiological parameter threshold adjustment model. This model is constructed based on stage identification labels as the primary control condition, combined with individual characteristics such as advanced age, hypertension, and beta-blocker use, to differentiate the upper and lower limits of indicators such as SBP, HR, and SpO2. For instance, for a hypertensive patient with a baseline systolic blood pressure of 160 mmHg, the system will automatically use a 20% lower threshold (e.g., 128 mmHg) as the hypotension alarm threshold, rather than the general standard for healthy individuals (e.g., 90 mmHg), thus achieving more clinically significant and accurate early warnings. The threshold adjustment model used in this stage can be formally represented as a multi-factor weighted linear model or a decision tree-based rule engine, and the calculation results are synchronously updated to the monitoring engine in a structured form within the system.

[0041] For example, the system compares the monitored vital signs parameters with dynamically calculated personalized thresholds in real time to determine if any abnormal deviations occur. In the event of a deviation, the system further introduces a multi-parameter fusion assessment mechanism, such as logistic regression scoring or a fuzzy inference engine based on the combined trends of BIS, HR, and SpO2, to calculate a risk level score by integrating features such as the current deviation value, duration, and direction of fluctuation, and triggers a tiered alarm response accordingly. Alarms are categorized into alert, warning, and critical levels, and are linked to suggested treatment measures, such as prompts like "Insufficient minute ventilation, suggest adjusting respiratory parameters" or "HR rising with SpO2 falling, suggest investigating hypovolemia and ventilation / oxygenation disorders caused by hemorrhage," thus providing decision support. Alarm information is simultaneously displayed on the anesthesiologist's terminal and can be presented through multiple channels such as voice and graphics to enhance perception efficiency.

[0042] For example, by synchronizing and serializing the drug infusion data collected during the patient's surgery, such as the type, dosage, infusion rate, and infusion start and end time of drugs like propofol, remifentanil, and atracurium besylate, with multiple real-time vital sign parameters such as heart rate (HR), systolic blood pressure (SBP), respiratory rate (RR), oxygen saturation (SpO2), end-tidal carbon dioxide partial pressure (EtCO2), and bispectral index (BIS), a continuous characteristic time series is constructed. Based on this, a time series analysis model is deployed to perform real-time inference on this time series. This model can employ a deep neural network structure based on Long Short-Term Memory (LSTM) or a Transformer model incorporating attention mechanisms to learn the implicit dynamic patterns between drug input and vital sign changes at different stages of anesthesia. Specifically, the induction phase typically shows a rapid increase in drug input rate accompanied by a decrease in BIS and a rapid decline in blood pressure; the maintenance phase shows a state where parameter fluctuations tend to stabilize; and the recovery phase corresponds to a gradual decrease in drug infusion, a gradual recovery of BIS, and a certain degree of rebound in HR and SBP. During the model training phase, labels are learned based on existing surgical case data. During inference, the model performs real-time stage classification based on the feature sequence within the latest time window and outputs the current anesthesia stage label. This analytical method overcomes the limitations of previous stage divisions relying solely on manually set time segments or subjective judgment by medical staff, achieving objective, dynamic, and high-precision stage identification. This provides core criteria for subsequent dynamic threshold adjustments and alarm strategy formulation. It effectively improves the intelligence and individualization of anesthesia management, and is particularly suitable for situations with complex surgical rhythms, drastic changes in anesthetic medications, or large fluctuations in the patient's baseline condition, significantly enhancing the system's ability to identify risk points and its intervention response efficiency.

[0043] In some examples, the system first continuously collects multi-source data streams from surgical patients, specifically including anesthetic drug infusion parameters such as propofol / remifentanil dosage, ventilation parameters such as airway pressure, end-tidal carbon dioxide partial pressure, airway resistance, tidal volume, respiratory rate, and traditional vital sign parameters such as heart rate, systolic blood pressure, bispectral index of EEG, and arterial oxygen saturation. Subsequently, the system uses a time-series recognition model integrating multimodal physiological features to jointly analyze the data, identifying the current surgical anesthesia stage of the patient, including the patient's current surgical progress, depth of anesthesia, and physiological state. After identifying the specific surgical anesthesia stage of the patient, the system further calls the dynamic threshold mapping model under the corresponding stage to automatically calculate personalized alarm thresholds for vital sign parameters. Unlike traditional methods that rely solely on HR or SBP fluctuations, this approach prioritizes rate-of-change analysis of ventilation parameters. By introducing respiratory dynamics leading factors such as an EtCO2 decrease of >5 mmHg / 30s and a sustained increase in Paw ≥3 cmH2O accompanied by increased Raw, it serves as a precursor to potential metabolic crisis or airway obstruction, achieving an alarm response tens of seconds earlier than traditional methods. Technically, the system utilizes deep temporal sensing networks, such as Temporal Fusion Transformer (TCN), built upon the integration of ventilation, medication, and circulation parameters. These networks can output real-time stage predictions and risk probability distributions, and dynamically invoke threshold adjustment functions for specific stages. For example, when the system detects a rapid decrease in end-tidal carbon dioxide (EtCO2) (e.g., from 40 mmHg to 30 mmHg within 1 minute), it will increase the "trend sensitivity coefficient" of the heart rate (HR) alarm model by 50% in real time. This means that any slight decrease in heart rate will be amplified and assessed, revealing earlier the potential compensatory decrease that may occur under conditions of insufficient ventilation, thus achieving risk feedforward early warning across physiological systems. Furthermore, the system can be set to a ventilation-circulation inconsistency warning mode, triggering an early warning of deep anesthesia risk or airway complications when ventilation parameters have fluctuated drastically while traditional vital signs remain unchanged, for physician reference. Through these methods, the system can rapidly identify and dynamically reconstruct anesthetic crisis precursors before changes in vital signs, significantly improving the response window and risk management efficiency for intraoperative emergencies.

[0044] For example, ventilation parameters include: tidal volume, ventilation mode, airway pressure, respiratory circuit flow, pressure waveform, and CO2 waveform. Vital signs parameters include: non-invasive / invasive blood pressure, heart rate, heart rhythm, pulse oxygen saturation, bispectral index, end-tidal carbon dioxide, body temperature, and central venous pressure, as well as integrated blood gas analysis data such as pH, PaO2, and lactate.

[0045] For example, during general anesthesia, if the system detects that the expiratory tidal volume is lower than the inspiratory tidal volume and the CO2 waveform amplitude is reduced, it will automatically trigger a breathing circuit leak alarm and suggest checking the endotracheal tube cuff or circuit connection.

[0046] For example, the system continuously collects multiple vital sign parameters, including heart rate (HR), systolic blood pressure (SBP), bispectral index (BIS), and end-tidal carbon dioxide partial pressure (EtCO2). A sliding time window, such as the past 60 seconds, is set to extract multidimensional data points within that time period, forming a four-dimensional parameter point cloud. Next, the complex algorithm in topological data analysis is used to construct a topological structure from this point cloud at different distance scales. This converts the multidimensional changes in vital signs during surgery into an analytical topological structure, used to detect the global shape characteristics of the data. Topological holes or cyclic structures are extracted by calculating its first-order homology group H1, forming a topological perturbation index (TP) to measure the degree of state-space perturbation. This TP value reflects the current system structural stability; a higher value indicates poorer coordination between parameters and that the system is closer to the edge of non-steady-state conditions. After obtaining the TP value, the system dynamically adjusts the safety alarm thresholds for key vital signs based on the set perturbation response rules. For example, when the TP value continuously rises above 5, the system automatically raises the lower limit of SBP from 85 mmHg to 90 mmHg, while simultaneously increasing the tolerance of the lower limit of BIS, thereby enhancing sensitivity to potential risks of hypoperfusion or loss of consciousness. When the TP value falls back below 2 and remains below it for a period of time, the system gradually restores the original threshold, maintaining a sufficiently lenient alarm criterion to reduce false alarms. Furthermore, the system can set the TP threshold change trend as a high-risk alarm trigger condition. When topological perturbation continuously increases and is accompanied by a significant increase in the fluctuation period, the system sends an early warning of system structural instability to the anesthesiologist, prompting them to assess the impact of current operations such as traction or electrocautery on the depth of anesthesia or the circulatory system, and adjust the operation rhythm or supplement sedative drugs if necessary. In this way, the system not only perceives the overall risk situation of the evolution of the anesthetic state from a structural level, but also avoids the misjudgment problems of traditional single-parameter fluctuation-based methods, achieving accurate modeling and dynamic management of stage boundaries and physiological tolerance zones without relying on stage labels. Compared with existing rule-based or stage-based classification algorithms, this method exhibits higher adaptability to complex surgeries, stronger ability to identify parameter couplings, and enhanced system stability analysis capabilities, significantly improving the accuracy and response sensitivity of anesthesia safety alarm systems under highly dynamic intraoperative conditions. By introducing topological data analysis methods to structurally model the multi-parameter coupled fluctuations of patients' vital signs under anesthesia, the method utilizes the topological perturbation characteristics of physiological signals during surgical procedures to optimize stage identification and dynamic threshold coordinated control.

[0047] In some examples, it also includes: Real-time acquisition of surgical operation characteristic data, including the usage parameters of surgical instruments; The operation influencing factor is calculated based on real-time collected vital sign parameters, surgical operation characteristic data, and operation-physiological response correlation model. An early warning will be issued if the operational impact factor exceeds a preset safety threshold.

[0048] For example, real-time acquisition of surgical operation characteristic data includes usage parameters of at least one surgical instrument, such as electrosurgical energy output level, electrocoagulation duration, laparoscopic grasping forceps holding time and frequency, bone drill rotation speed, and tissue traction tension parameters. This data can be automatically collected through data interfaces with intelligent surgical instruments, energy platforms, or integrated surgical robot systems, with the acquisition frequency synchronized with vital sign monitoring, such as once per second. After acquiring the surgical instrument operation data, the system combines current real-time vital sign parameters such as systolic blood pressure, heart rate, blood oxygen saturation, end-tidal carbon dioxide partial pressure (EtCO2), and bispectral index (BIS) and inputs them into a pre-constructed operation-physiological response correlation model to calculate the operation impact factor. This model can be a regression model or support vector machine based on supervised learning, or a fuzzy logic rule system constructed by combining medical expert knowledge. Its core objective is to quantify the immediate impact of surgical operations on the patient's physiological state within a specific time window during surgery. The OII (Operation Impact Index) is the output result; a higher value indicates a stronger physiological response and greater risk to the patient triggered by the current operation. The system compares the OII value with a preset safety threshold. If the operational influencing factors exceed the set safety limits (e.g., OII > 0.8, or a sharp increase of more than 50% compared to the previous window value), the system automatically triggers an intraoperative warning. The warning may include descriptions of operational behaviors such as continuous high-frequency electrosurgical output for more than 10 seconds, and corresponding physiological responses such as a decrease in SpO2 combined with an increase in HR. It also provides treatment suggestions, such as pausing energy operations and assessing circulatory status. This mechanism not only enhances the ability to detect potential anesthetic risks at high-risk intraoperative points such as bleeding, nerve injury, or excessive traction, but also transforms the anesthesia monitoring system from passive physiological alarms to proactive risk trend warnings, effectively improving the efficiency of multidisciplinary collaborative decision-making and the level of patient safety during surgery.

[0049] For example, at each time slice t, the system extracts several parameter vectors representing the characteristics of the surgical procedure. For example: x1: Current electrosurgical power (W); x2: Electrocoagulation duration (s); x3: Tension value of the traction device (N); x4: Operation frequency (times / min); ...x n Other quantitative characteristics include operation duration. The operation impact factor is defined as:

[0050] in, The influence weight of the corresponding operating parameters represents the ability of this feature to disturb the patient's physiological state.

[0051] For example, to obtain medically credible and highly generalizable... Large-scale surgical data is required for regression optimization training. Ridge regression or Lasso can be used to avoid overfitting. Modeling can be performed separately for each type of surgery to adapt to the operation-response coupling characteristics of different surgical procedures. Dynamic time window weighting can be introduced to improve the sensitivity to short-term drastic disturbances.

[0052] For example, during a breast nodule resection under general anesthesia and laryngeal mask airway insertion, the system monitors surgical operation characteristics data in real time: intermittent activation of the electrosurgical unit is detected, energy output is stable, the suction device performs low-speed intermittent aspiration, and the efficiency of fluid removal from the surgical field is normal; the retractor position is stable, the surgical field is adequately exposed, and there is no abnormal displacement; at the same time, the system monitors ventilation parameters in real time: SpO2 decreases, airway pressure increases; the system analyzes and calculates operation influencing factors: the correlation between electrosurgical unit activation duration and tissue thermal damage risk is normal; the matching rate between suction device negative pressure and surgical field dryness is normal; the laryngeal mask airway displacement risk index is abnormal, triggering an alarm.

[0053] In some examples, it also includes: Based on real-time collected surgical procedure characteristic data, vital sign parameters, anesthesia machine ventilation parameters, drug infusion data, and patient baseline characteristics, a proactive risk assessment model is used to predict potential anesthesia risks.

[0054] For example, an active risk assessment model can be built and run to comprehensively predict potential anesthesia-related risk events based on real-time collected surgical operation characteristic data, vital sign parameters, anesthesia machine parameters, drug infusion data, and patient baseline characteristics, thereby achieving early warning, dynamic intervention, and intelligent decision support. Specifically, the system first collects three types of key data in real time: first, surgical operation characteristic data, including but not limited to the output power, duration, and frequency of energy surgical equipment such as electrosurgical units and ultrasonic scalpels, the amplitude and intensity of traction / clamping instruments, intra-abdominal pressure, and irrigation fluid flow rate; second, vital sign data, including heart rate, blood pressure, SpO2, EtCO2, and BIS; and third, anesthetic drug-related data, including drug type, unit dose, total dose, infusion method, duration / intermittent time, and cumulative medication time. The above data is synchronized temporally and feature extracted in the system to form a fused dynamic input feature sequence, which is then input into the pre-trained active risk assessment model.

[0055] For example, the model can be built based on deep neural networks such as bidirectional LSTM, GRU, Transformer structures, or ensemble learning models such as XGBoost and LightGBM combined with temporal window embedding, aiming to capture the nonlinear coupling relationship between multidimensional parameters and short-to-medium time lag response characteristics. During the model training phase, real surgical case data is used, with risk events such as intraoperative hypotension, sudden hypercapnia, awakening of consciousness, and increased airway pressure serving as supervised labels to achieve early identification of high-risk prodromal states during surgery. The model can output a risk score, which represents the probability or confidence level of the patient entering a high-risk event range in their current state. When the risk score exceeds a set threshold, such as >0.75, the system will trigger a potential anesthesia risk warning and simultaneously inform the anesthesiologist of the risk type, associated operations or parameters, and intervention suggestions, such as: a predicted hypotension risk score of 0.81 within 5 minutes; current electrosurgical output power is too high, blood pressure shows a significant downward trend, and the cumulative dose of remifentanil is high; suggestions include: reducing sedation, monitoring blood volume, and preparing vasopressors.

[0056] Understandably, unlike traditional alarm mechanisms that only respond passively after an event has occurred, this proactive risk assessment mechanism, through feedforward learning of surgical behavior, physiological changes, and pharmacological load, possesses the ability to predict and prevent risk situations. It is particularly suitable for individualized risk management in highly dynamic intraoperative scenarios. Clinically, this method can effectively reduce the incidence of serious adverse events, shorten event response time, improve the collaborative efficiency of cross-disciplinary teams, and significantly enhance patient safety and the level of intelligence in anesthesia management.

[0057] In some examples, the prediction of potential anesthetic risks based on real-time acquired surgical procedure characteristic data, vital sign parameters, anesthesia machine ventilation parameters, drug infusion data, and patient baseline characteristics through an active risk assessment model includes: Based on real-time collected surgical operation characteristic data, vital sign parameters, anesthesia machine ventilation parameters, drug infusion data, and patient basic characteristics, a proactive risk assessment model is used to predict multi-dimensional risk indicators to establish a dynamic risk matrix. The dynamic risk matrix includes immediate risk values, risk evolution trends, and a list of abnormal related physiological parameters.

[0058] For example, based on real-time collected surgical operation characteristic data, vital sign parameters, anesthesia machine ventilation parameters, drug infusion data and patient basic characteristics, a multi-dimensional potential anesthesia-related risk index is predicted through an active risk assessment model, thereby constructing a multi-dimensional dynamic risk matrix for continuous monitoring and visual assessment of the patient's risk status and physiological stress response during surgery. The process specifically includes: First, real-time collection of three types of core data from surgical equipment and monitoring systems, including but not limited to the following: (1) Surgical operation characteristic data, such as the output power, duration of action, frequency of intermittent action, changes in tissue traction force, pneumoperitoneum pressure, etc. of electrosurgical instruments; (2) Vital sign parameters, such as heart rate (HR), systolic blood pressure / diastolic blood pressure, SpO2, EtCO2, RR, BIS index, etc.; (3) Anesthesia machine ventilation parameters, such as airway pressure, airway resistance, and end-expiratory CO2 partial pressure; (4) Drug infusion information, including the real-time infusion rate, cumulative dosage and type of anesthetics, analgesics and muscle relaxants. (5) Basic patient characteristics, which may include basic information such as height, weight, gender, age, medical history, and ASA classification. These multi-source heterogeneous data are input into the active risk assessment model after unified time synchronization. This model integrates temporal deep learning networks such as bidirectional LSTM with multiple input channels and attention mechanisms, which can identify nonlinear correlations and potential time delay effects between physiological responses and intraoperative operations while maintaining temporal continuity.

[0059] For example, the model output is not limited to a single risk score, but generates multiple quantitative risk indicators in parallel from multiple dimensions, including: (1) RRI (Real-Time Risk Index): reflecting the confidence level of the patient facing a serious risk of anesthesia complications at the current time point; (2) Risk Evolution Slope: fitting the risk score through a sliding window to represent the speed and direction of the risk value increase / decrease, used to identify a high-risk gradual state; (3) Physio-Abnormal List: listing the key vital signs parameters that currently drive the risk increase and their abnormal manifestations, such as BIS consistently below 40, SBP decrease, and SpO2 decrease of more than 5% from baseline, used by doctors to quickly determine the pathological cause. The above multidimensional outputs are summarized to form a structured dynamic risk matrix, which is displayed in real time in a graphical interface. The vertical axis is the risk dimension, the horizontal axis is the time process, and the risk value heat is distinguished by color. The matrix supports interactive clicking to view the historical fluctuation curves and influencing factor analysis of individual indicators, realizing the transition from risk score to pathological state cognition.

[0060] Understandably, the construction and dynamic updating mechanism of this risk matrix not only enhances the system's ability to identify and intervene in advance of sudden anesthesia risks, but also provides anesthesiologists with a clear and structured decision support view. Especially in complex surgeries such as neurosurgery, thoracic surgery, organ transplantation, or patients with multiple system comorbidities, it can effectively address the problem of delayed early warning caused by disturbances in multi-source parameters, thereby significantly improving the initiative, accuracy, and response speed of intraoperative management.

[0061] In some examples, it also includes: Risk levels are assessed based on the aforementioned multi-dimensional risk indicators to enable tiered early warning and expert resource matching; and / or, Intelligent decision-making suggestions are generated based on the aforementioned multi-dimensional risk indicators.

[0062] For example, based on the construction of a dynamic risk matrix, the system comprehensively quantifies the patient's overall anesthesia risk status at the current moment based on preset multi-dimensional risk indicator weighting rules or fusion decision models such as a fuzzy logic system with hierarchical weighting mechanisms or a rule tree-based multi-label classifier, and classifies it into clear risk levels, such as levels I to IV or four categories: warning / suggestion / serious / critical, serving as the triggering basis for graded early warning. The system dynamically allocates response resources by comparing the degree of matching between the risk level and the current surgical risk level, such as ASA classification or surgical procedure classification. For example, when the system identifies a patient in a high-risk situation, such as risk level IV, it will automatically call the expert collaborative scheduling interface to send remote assistance requests to the anesthesiologist in charge of the surgery or area, the risk management team, and cardiologists / ICU doctors, or prompt on-duty experts to intervene, forming a closed-loop processing flow; while in low to medium risk situations, the system only prompts the front-line anesthesiologist to pay closer attention, avoiding waste of human resources. This expert matching mechanism can match doctors with corresponding specialist backgrounds based on risk type, such as circulatory risk / respiratory risk / consciousness risk, to improve intervention effectiveness.

[0063] For example, the system can also automatically generate intelligent decision-making suggestions based on a combination of multidimensional risk indicators and their driving factors. This suggestion module is built upon a case knowledge graph and an intervention strategy library constructed based on an inference engine. It combines specific anomalies such as a persistently low BIS of 40 accompanied by decreased SBP and EtCO2, patient individual characteristics such as past cardiovascular history, and current medication status such as high-dose remifentanil maintenance. Through condition matching, parameter fitting, or logical backtracking mechanisms, it outputs a set of clinically actionable response suggestions. These suggestions may include: assessing the possibility of sedative overdose and attempting to slow the propofol infusion rate; recommending fluid resuscitation to maintain circulating volume and preparing for low-dose dopamine infusion; or considering large fluctuations in SpO2 during resuscitation and unstable ventilation strategies, suggesting adjustments to tidal volume or respiratory rate. These suggestions can be pushed to the anesthesiologist's intelligent assistant interface, supporting voice broadcasting, text-image interaction, and trend playback, helping medical staff respond quickly under high information load.

[0064] Understandably, this risk-level-driven response mechanism and intelligent intervention suggestion output system not only significantly improves the ability of anesthesia management to shift from monitoring abnormal parameters to guiding clinical behavior, but also provides technical support for building a cross-disciplinary collaborative, risk value chain-based closed-loop management system, significantly enhancing the safety assurance capabilities and clinical response efficiency for high-risk patients during surgery.

[0065] In some examples, it also includes: Record the actual actions taken by doctors upon receiving the intelligent decision-making suggestions; When there is a deviation between the actual processing operation and the intelligent decision-making suggestion, the deviation event is recorded so as to carry out reinforcement learning of the risk assessment model and the decision-making model based on the deviation event.

[0066] For example, when the system generates intelligent decision-making suggestions based on surgical operation characteristics, vital sign parameters, anesthesia machine ventilation parameters, drug infusion data, and patient baseline characteristics, such as suggesting slowing the propofol infusion rate or increasing fluid load, the suggestion will be pushed to the anesthesiologist's interface, and the system will simultaneously activate the operation behavior monitoring module. This monitoring module, in conjunction with the infusion pump, ventilator control interface, and vital sign change detection module, records in real time the physician's adjustments to key parameters after the suggestion is issued. This includes, but is not limited to, drug adjustment actions such as the object, dosage, and timing of adjustments; changes in ventilation parameters; use of vasoactive drugs; and fluid replenishment.

[0067] For example, the system compares the doctor's actual actions with the suggested actions at the corresponding moment at both the semantic and operational levels. If there is a clear deviation, such as the system suggesting a reduction in medication but the doctor choosing to increase it, or suggesting adjusting breathing but taking no action, the system identifies this behavior as a deviation event. The deviation event record not only includes the specific deviation item and its value, but also associates it with contextual data, such as the patient's vital signs curve at the time, the type of suggestion, the doctor's role information, and the stage of the operation, forming a structured deviation dataset.

[0068] For example, this bias dataset is then incorporated into a sample pool for model self-learning, and the proactive risk assessment model and decision reasoning model in the system are optimized through reinforcement learning mechanisms. Specifically, policy inversion methods based on policy gradients, proximal policy optimization (PPO), or inverse reinforcement learning (IRL) can be used. By treating bias events as environmental feedback, the system dynamically adjusts the model's output strategy by evaluating changes in risk, clinical outcomes, or expert review opinions after the bias, thereby optimizing the accuracy and clinical acceptability of the recommendations. For instance, if a biased behavior, such as a doctor not following the system's recommendations, results in better vital signs, the system will lower the output weight of such recommendations in the strategy; conversely, it will enhance the model's sensitivity to such situations, increasing the frequency and priority of recommendation outputs.

[0069] Understandably, by introducing this method, the system possesses the ability to continuously iterate and update its knowledge, gradually adapting to the actual operating styles, clinical preferences, and patient differences of different hospitals and doctor groups. This enhances the adaptability and trustworthiness of the intelligent system to complex clinical environments, propelling clinical intelligent assistance from static algorithm output to a new stage of dynamic behavior co-construction and intelligent evolution.

[0070] According to some embodiments, the system may include a user operation module and a security management and early warning system module.

[0071] The user interface module includes a clinical decision-making interaction module, an operation feedback calibration module, and a mobile collaboration module. The clinical decision interaction module is used to visually present risk assessment results and decision recommendations. The operation feedback calibration module records the deviation between the doctor's actual treatment and the system's generated suggestions, and uses closed-loop reinforcement learning to optimize the decision-making scheme. The mobile collaboration module supports mobile alert push notifications and access to remote expert guidance. The safety management and early warning module includes an anesthesia phase perception module, a multimodal correlation analysis module, an intelligent risk assessment module, a clinical decision center module, and an emergency response execution module; The anesthesia stage sensing module collects ventilation parameters, drug infusion data, and vital sign parameters from the anesthesia machine. It calculates risk probabilities using a time-series analysis model and dynamically adjusts the global alert level and alarm sensitivity of the monitoring system. The anesthesia stage sensing module includes a data acquisition module, a stage identification module, and a threshold adjustment module. The data acquisition module synchronously acquires the type and rate of anesthetic drug infusion and vital sign parameters in real time. The stage identification module identifies the current anesthesia stage and its classification. The threshold adjustment module dynamically calculates alarm thresholds based on the identified anesthesia stage, optimizes the threshold range by combining it with the patient's personalized baseline data, and supports a manual calibration mode for anesthesiologists to adjust.

[0072] The multimodal correlation analysis module is used to analyze the impact of surgical procedures and anesthetic interventions on physiological parameters in real time; it acquires surgical procedure feature data through visual perception technology; it constructs a causal analysis model to assess the correlation between procedures and physiological parameters; and it triggers a correlation warning when the procedure impact factor exceeds a preset threshold. The procedure impact factor is calculated using a weighted summation method, with weight coefficients obtained through training on a large amount of surgical data. The multimodal correlation analysis module includes a surgical procedure perception module, a physiological parameter analysis module, and a causal correlation module. The surgical procedure perception module acquires real-time surgical field video streams through visual acquisition devices, extracts key surgical procedures using image recognition technology, and establishes a time series of procedure events, including the number of times electrocoagulation equipment is used and the activation duration of the suction device. The physiological parameter analysis module receives real-time vital sign data from monitoring equipment and extracts dynamic change characteristics of physiological parameters, including blood pressure variability and heart rate variability. The causal correlation management module constructs a procedure-physiological response correlation model, calculates the procedure impact factor, and generates a warning signal when the procedure impact factor value exceeds a preset safety threshold.

[0073] For example, during laparoscopic cholecystectomy under general anesthesia, the multimodal correlation analysis module monitors the following parameters in real time: Surgical operation perception module: identifies continuous activation of the electrocoagulation device, accelerated suction by the aspirator, and active bleeding in the surgical field; Physiological parameter analysis module: detects significant increase in blood pressure, significant rise in blood pressure, compensatory increase in heart rate, and significant increase in pulse pressure variability; Causal correlation module: calculates operational influencing factors and triggers corresponding alarm levels when thresholds are reached.

[0074] The intelligent risk assessment module processes feature data from multiple sources and generates risk predictions. Based on a historical case database, it generates treatment plan suggestions. The multi-dimensional feature data includes at least five categories: vital signs, anesthesia machine ventilation parameters, drug infusion parameters, surgical instrument status, and patient baseline characteristics. The intelligent risk assessment module comprises a multi-source data fusion module, a feature engineering module, and a risk output module. The feature engineering module includes a data extraction module and a standardization processing module. The multi-source data fusion module receives and integrates core data in real time, including vital signs data, anesthesia machine ventilation parameters, drug infusion parameters, surgical instrument status, and patient baseline characteristics, establishing a data matrix with a unified timeline. The data extraction module extracts features from multiple dimensions, and the standardization processing module standardizes heterogeneous medical data into a unified risk assessment feature set. The risk output module outputs high-risk event probability predictions.

[0075] For example, during surgery, the multi-source data fusion module collects the following parameters in real time: vital signs: sudden drop in blood pressure, increased heart rate; ventilation parameters: increased airway pressure; decreased or absent breath sounds as heard by the anesthesiologist; medication data: antibiotics recently administered; instrument status: no special procedures performed. The data extraction module extracts feature vectors: abnormal rate of blood pressure drop, and the temporal correlation between drug and blood pressure changes conforms to allergy characteristics. The risk output module outputs a high-risk anaphylactic shock assessment.

[0076] For example, during the surgery, the multi-source data fusion module collects the following parameters in real time: Vital signs: decreased blood pressure, increased heart rate, or arrhythmia; Electrocardiogram: ST segment elevation or depression, T wave inversion, or new-onset atrioventricular block; Ventilation parameters: normal airway pressure; Anesthesiologist observation: no abnormal breath sounds, but jugular venous distension may be observed; Medication data: no recent use of vasoactive drugs, no allergic drug administration; Instrument status: the surgical procedure did not directly stimulate the heart. The data extraction module extracts feature vectors: abnormal blood pressure-ECG correlation; decreased heart rate variability; no increase in airway pressure, ruling out pulmonary embolism or allergic reaction. The risk output module outputs a high-risk judgment: high probability of intraoperative acute myocardial infarction, recommending immediate screening of myocardial enzymes, suspension of surgical stimulation, and initiation of cardiology consultation.

[0077] The Clinical Decision Center module analyzes multidimensional data to construct an active risk assessment model, predict potential anesthesia risks, and generate intelligent decision-making suggestions. The Clinical Decision Center module includes a risk integration module and a decision reasoning module. The risk integration module receives multidimensional risk indicators from the intelligent risk assessment module and establishes a dynamic risk matrix, including: real-time risk values, risk evolution trends, and a list of abnormal related physiological parameters. The decision reasoning module outputs treatment plans.

[0078] For example, during surgery, the risk integration module receives a high-probability prediction of a malignant hyperthermia event generated by the intelligent risk assessment module. The decision reasoning module matches the feature patterns of malignant hyperthermia in the case library. If the similarity is high, a treatment plan is generated.

[0079] The emergency response execution module automatically initiates a multi-level response mechanism based on the severity of the emergency, enabling intelligent allocation of expert resources. This module includes a severity assessment module and a resource scheduling module. The severity assessment module receives risk level data from the clinical decision-making center and implements multi-level response classification. The resource scheduling module generates expert matching strategies and intelligently coordinates with emergency medical equipment.

[0080] For example, when the clinical decision center outputs a major bleeding crisis, the urgency assessment module determines it to be a high-level response, the resource scheduling module intelligently calls the blood bank, notifies the chief surgeon and anesthesiologist, and coordinates with the surgical shadowless lamp to increase its brightness and with the surgical air conditioning system to increase the room temperature.

[0081] The safety management and early warning system module also includes a data security and privacy protection module, which is used to encrypt and store anesthesia data to ensure the security of all anesthesia data and the protection of patient privacy.

[0082] In some cases, considering that a patient's airway may be nearing decompensation even when apparent vital signs are stable, traditional monitoring methods struggle to detect this early. The system can periodically apply very small, short-duration programmed perturbations to tidal volume or inspiratory pressure without altering effective ventilation, while simultaneously acquiring airway pressure, airway resistance, end-expiratory carbon dioxide partial pressure, and flow rate curves. It calculates the incremental response coefficient and recovery time constant of respiratory mechanics before and after the perturbation, thereby assessing the vulnerability index of airway compliance reserve and carbon dioxide expulsion efficiency. When the index deteriorates and heart rate or blood pressure remains unchanged afterward, it is marked as a potential destabilization zone in the stage identification, dynamically tightening the alarm thresholds for blood pressure and blood oxygenation, and providing early warnings for ventilation strategy adjustments. This can advance the crisis recognition window by tens of seconds to several minutes.

[0083] For example, after the patient enters stable mechanical ventilation, several respiratory cycles are continuously collected as an individualized baseline. Key quantities such as peak and plateau pressure of airway pressure, peak and average airway flow, actual tidal volume, respiratory rate, set positive end-expiratory pressure, carbon dioxide expiratory plateau value, and pressure-volume loop area are fully recorded for each cycle. At the same time, the reliability of these quantities is verified by ventilation mode and loop information, and the range, sampling frequency, and time synchronization status of the monitoring equipment are self-checked. Interference periods such as changes in body position, suctioning, and instrument connection are eliminated before the baseline is solidified. Subsequently, based on the patient's age, weight, underlying lung disease, depth of anesthesia, and current surgical position, a safe boundary for the perturbation allowable zone is set. This includes that the perturbation amplitude should not exceed a small percentage of the current tidal volume or a small percentage of the inspiratory target pressure, should not cross the end of the expiratory phase, and should not be triggered when blood oxygen, carbon dioxide, and peak pressure are close to the danger threshold. Therefore, by calibrating the patient's current mechanical and metabolic status through an individualized baseline and controlling perturbations within physiologically acceptable ranges using multiple safety gates, any subsequent observed incremental changes have a clear reference without affecting effective ventilation or causing hemodynamic adverse events. The perturbations are implemented using small, short-duration, programmed step movements, allowing for a slight increase in the target tidal volume or a slight increase in the target airway pressure of the current inspiration, with either option selectable and the upper limit of the amplitude constrained by the aforementioned safety boundaries. To avoid confusion from nursing procedures or surgical traction, the system pseudo-randomizes the sequence number of the triggered breath and the trigger phase within the inspirational phase within a preset observation window, ensuring sufficient recovery respiratory cycles between two perturbations to minimize cumulative effects. Perturbation commands are issued through a validated anesthesia machine control interface, and all pressure, flow, carbon dioxide, and volume data are simultaneously collected at a sampling rate higher than routine monitoring and precisely aligned with timestamps. Therefore, by using mild external disturbances combined with the identification of system response and return to steady state, potential subclinical abnormalities can be amplified and quantified in a short time. Early signals such as decreased compliance reserve, increased resistance, or slower carbon dioxide elution can be detected without changing the ventilation mode and settings. The system performs drift and noise reduction processing on the acquired airway pressure and flow signals, and finely segments each respiratory cycle according to the zero-crossing point, extreme point, and plateau stability of the inspiratory-expiratory transition. It accurately marks the target cycle in which the perturbation occurs and several adjacent cycles before and after it. To ensure the reliability of cross-cycle comparisons, these cycles are resampled to a consistent time length. At the same time, the system extracts indicators such as the inspiratory peak, plateau average, actual tidal volume, pressure-volume loop area, carbon dioxide plateau level, and subsequent recovery trajectory for each cycle. The target cycle and adjacent cycles are packaged into local time-series segments before, during, and after the perturbation. Thus, with strict consistency and time alignment, the system separates the minute differences in the signal from the measurement noise, providing a high signal-to-noise ratio and repeatable input basis for subsequent incremental comparisons and recovery rate assessments, and avoiding misjudgments caused by phase mismatch or sampling jitter.Within the mid-inspiratory phase and adjacent stable plateau of the target cycle, the system estimates equivalent compliance and equivalent resistance for the two states before and after the perturbation, compares their relative changes, and observes whether atypical changes occur in the shape of the pressure-volume loop, such as increased area, intensified separation of the ascending and descending branches, or inflection point drift. After the perturbation ends, the system tracks the rate at which airway pressure and flow return to baseline over several subsequent breaths and fits the trajectory of carbon dioxide expiratory plateau value recovery over time to obtain quantitative characterizations of mechanical recovery rate and carbon dioxide elution recovery rate. Thus, by treating the respiratory system as a weakly nonlinear dynamic object under the current settings and patient conditions, and by comparing the two states before and after the perturbation to capture abnormal responses that are more sensitive to the same input, have slower recovery, and higher energy consumption, the system can significantly improve the detection rate of early, hidden problems such as decreased compliance, increased resistance, and slower carbon dioxide expulsion, even when heart rate, blood pressure, and blood oxygen are still within the reference range, providing reliable destabilization indications. The system integrates multiple observables, such as incremental compliance, incremental resistance, mechanical recovery rate, carbon dioxide elution recovery rate, pressure-volume loop area change, and overshoot occurrence, into a single score after dimensional normalization and range clipping, serving as a respiratory vulnerability index. The index's weights are obtained through retrospective annotation and prospective validation of a large amount of surgical data. Annotation criteria may include objective events such as the need for suctioning, bronchodilator use, circuit or ventilation parameter adjustments, and the occurrence of oxygen saturation decline or the need for vasopressors within a certain time window. After deployment, the index employs a combined smoothing and rate-of-change monitoring strategy over time to balance sensitivity and stability. This integration converges multiple weak signals into a single-valued indicator with consistent physical meaning, establishing a stable mapping relationship with clinical outcomes. This enables universal early risk quantification across surgical procedures and patients, while retaining interpretability to support rapid understanding and adoption by physicians. When the respiratory vulnerability index exceeds the high-risk threshold, or shows a significant accelerated deterioration within the medium-risk range, the system automatically refines the current surgical anesthesia stage into a ventilation destabilization zone. It then proactively tightens the vital sign alarm thresholds according to predetermined mapping rules, such as slightly adjusting the lower limits of systolic blood pressure, blood oxygen saturation, and electroencephalogram (EEG) index towards a more cautious approach. Simultaneously, it increases sensitivity to declining blood pressure and blood oxygen saturation trends and shortens the reassessment cycle. If the index deteriorates but heart rate and blood pressure remain unchanged, the system still performs threshold reconstruction and clearly indicates on the interface that this reconstruction was triggered by ventilation vulnerability, guiding priority to address ventilation issues rather than blindly deepening anesthesia or prematurely increasing pressure. By using ventilation leaders to integrate and synchronously switch between stages and thresholds, the alarm strategy is transformed from passive response to proactive prevention, advancing the crisis identification window, reducing false alarm rates, increasing the recall rate of necessary alarms, and minimizing unnecessary drug and circulatory interventions.

[0084] In some examples, considering that small leaks in the circuit or fatigue of the one-way valve may initially manifest only as high-frequency sawtoothing of the expiratory flow curve and weak resonance of airway pressure, traditional thresholds are difficult to trigger. Based on the acoustic and mechanical coupling characteristics of the circuit, high-frequency detail energy and the stability of the main resonant frequency band in the mid-to-late expiratory phase can be extracted in real time. When both rise synchronously while blood oxygenation remains normal, the patient is placed in the device coupling abnormality sub-stage, dynamically lowering the alarm threshold for the decrease in end-expiratory carbon dioxide partial pressure, increasing sensitivity to blood pressure drops, and prompting a check of the breathing valve and connector seals.

[0085] For example, after the ventilation settings stabilize, the system first continuously collects a baseline data segment covering multiple complete respiratory cycles, recording the expiratory flow rate curve, airway pressure curve, end-expiratory carbon dioxide waveform, ventilator alarm log, assembly timestamps of disposable valves and connectors, and soda lime replacement records. Simultaneously, it performs a low-pressure circuit self-check and leak test, writing the results into the current case context. This uses the combined steady-state of the device and the patient as a reference surface, ensuring that subsequently identified high-frequency detail changes do not originate from confounding factors beyond initial differences or assembly defects. This provides a reliable reference for subsequent microstructure comparisons and makes the implicit variable of circuit health explicit, reducing the noise floor of false alarms. The system increases the sampling rate of airway pressure and flow rate on the ventilator side to a level sufficient to reflect high-frequency details and strictly aligns it with the carbon dioxide waveform and machine event log. Simultaneously, it monitors the status of the working ports of devices that may introduce pseudo-high-frequency components, such as electrosurgical units, suction devices, and orthopedic drills, and generates interference marker trajectories. Subsequently, it corrects all curves for temperature drift and baseline drift, limiting the analysis window to the mid-to-late expiratory phase, automatically avoiding the end-inspiratory transition and the high-speed phase at the beginning of expiration. Therefore, by separating the actual loop, high-frequency components of airway coupling, and external surgical instrument noise in terms of time and frequency, the signal-to-noise ratio of subsequent serration and resonance stability extraction can be significantly improved, and false positives triggered by external interference can be reduced. The system uses a joint criterion of zero flow crossover, volume integral continuity, and airway pressure inflection point to finely divide the expiratory phase into acceleration, plateau, and terminal segments, and provides data quality labels for each segment, including the presence of events such as coughing, airway suction, sudden changes in body position, and abnormal opening and closing of the expiratory valve; segments that do not meet the quality standards are directly eliminated or downweighted. Considering that serration and resonance often occur in the middle and later stages when the gas flow rate decreases significantly and the valve participation increases, it is necessary to cleanly cut this critical interval from the overall expiratory phase to provide a stable and reproducible target interval for subsequent microstructure analysis and avoid the inclusion of irrelevant transient disturbances. In the already located mid-to-late expiratory phase, the system performs a microstructural scan of the flow curve, focusing on detecting small, equally or nearly equally spaced small peaks, shoulder-shaped undulations, and repeated troughs. It then calculates a sawtooth significance score based on the duration, number of repetitions, and energy percentage of these peaks. Simultaneously, the system cross-validates this score with the tidal volume stability and end-expiratory carbon dioxide plateau integrity within the same cycle to eliminate spurious textures caused by the patient's spontaneous breathing effort or monitoring circuit jitter. Considering that small leaks in the circuit, fatigue of the one-way valve reed, or valve seat contamination can induce regular perturbations during valve micro-opening and return path switching, thus forming rhythmic, fine serrations on the flow curve, the system quantifies details that are difficult to reliably identify with the naked eye into a robust score, serving as the first chain of evidence for subsequent coupling anomaly assessment.The system searches for the main resonant frequency band related to the inherent frequency characteristics of valve opening and closing, loop volumetric elasticity, and tubing length on the synchronized airway pressure curve. It observes whether this band consistently appears and maintains high stability in the mid-to-late expiratory phase, while simultaneously verifying a fixed proportional relationship between this main frequency and the rhythm of flow serrations. If the main resonance is accompanied by an increase in sideband components and maintains continuity in position and amplitude across multiple cycles, the system identifies this phenomenon as a sign of device-side resonance. Therefore, when valve materials age, sealing surfaces become contaminated, or joints leak slightly, the mechatronic characteristics of the loop will exhibit a weak but continuous resonant peak in the mid-to-late expiratory phase, where load changes are most sensitive, and this peak is synchronized with the flow serrations. This provides a second chain of evidence corroborating the serrations, allowing the judgment to no longer rely on a single curve and significantly suppressing the influence of occasional noise. When both the significance of the sawtooth pattern in the flow rate curve and the stability of the airway pressure master resonance reach a preset threshold within the same time window, and the interference marker trajectory does not indicate external device interference, the system will comprehensively provide a highly confident conclusion of abnormal loop valve coupling, and further label the current surgical anesthesia stage as a sub-stage of abnormal device coupling. Simultaneously, the system will also examine whether there are signs of decreased rebreathing / ventilation efficiency, such as shortening or elevation of the end-expiratory carbon dioxide plateau, to further subdivide it into leakage tendency type or valve hysteresis type. Therefore, by using a multi-dimensional joint judgment combining dual-channel consistency judgment with interference elimination and end-expiratory waveform evidence, specificity is improved. Before changes in blood oxygen and blood pressure or before triggering routine alarms, the system can initially pinpoint the problem originating from a high-confidence state of the loop or valve, shortening the location time. Once the system enters the "mechanical coupling anomaly sub-stage," it immediately makes proactive adjustments to its alarm strategy: increasing sensitivity to end-tidal carbon dioxide decline, shortening the detection cycle for blood pressure and blood oxygen trends and appropriately tightening the lower threshold, while reducing the response intensity to single low-amplitude heart rate fluctuations to focus on early ventilation-related risks; highlighting loops and valves with prominent colors on the interface for priority investigation, and temporarily suppressing non-critical alerts to reduce physician distraction. Based on the highly reliable judgment that the problem originates at the loop and valve end, the system first adjusts the attention weight of the monitored objects, allocating alarm resources to the channels most likely to evolve into decreased ventilation efficiency, which can improve the recall rate of necessary alarms, reduce the frequency of irrelevant alarms, and advance the intervention timing.

[0086] like Figure 3 As shown, this application embodiment also provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 320 and executable on the processor. When the processor 320 executes the computer program 311, it implements the steps of any of the above-mentioned methods for digital and intelligent anesthesia safety management.

[0087] Since the electronic device described in this embodiment is the device used to implement the digital anesthesia safety management device in the embodiments of this application, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the method described in the embodiments of this application. Therefore, how the electronic device implements the method in the embodiments of this application will not be described in detail here. Any device used by those skilled in the art to implement the method in the embodiments of this application is within the scope of protection of this application.

[0088] In practical implementation, when the computer program 311 is executed by the processor, it can achieve the following: Figure 1 Any of the corresponding implementation methods in the embodiments.

[0089] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

Claims

1. A digital anesthesia safety management method, characterized in that, The method comprises: collecting the operation scene of a surgical patient, drug infusion data, ventilation parameters, vital sign parameters, and patient basic characteristics; based on the operation scene, drug infusion data, ventilation parameters, vital sign parameters, and patient basic characteristics, evaluating the risk level of the surgical patient; performing hierarchical alarm according to the risk level of the surgical patient obtained by evaluation.

2. The method of claim 1, wherein, The method of evaluating the risk level of the surgical patient based on the operation scene, drug infusion data, ventilation parameters, vital sign parameters, and patient basic characteristics comprises: based on the operation scene, drug infusion data, ventilation parameters, vital sign parameters, and patient basic characteristics, predicting potential anesthesia risks by an active risk assessment model to evaluate the risk level of the surgical patient.

3. The method of claim 2, wherein, The method of evaluating the risk level of the surgical patient based on the operation scene, drug infusion data, ventilation parameters, vital sign parameters, and patient basic characteristics comprises: based on the operation scene, drug infusion data, ventilation parameters, vital sign parameters, and patient basic characteristics, predicting multi-dimensional risk indicators by an active risk assessment model to establish a dynamic risk matrix, wherein the dynamic risk matrix comprises an instant risk value, a risk evolution trend, and a list of associated physiological parameter abnormalities.

4. The method of claim 1, wherein, The method further comprises: generating intelligent decision suggestions according to the multi-dimensional risk indicators.

5. The method of claim 4, wherein, The method further comprises: performing expert resource matching according to the risk level.

6. The method of claim 5, wherein, The method further comprises: recording the actual handling operation of a doctor in the case of receiving the intelligent decision suggestions; in the case of deviation between the actual handling operation and the intelligent decision suggestions, recording a deviation event to perform reinforcement learning of a risk assessment model and a decision model according to the deviation event.

7. The method of claim 6, wherein, The method further comprises: The operation scene, drug infusion data, and ventilation parameters identify the current operation anesthesia stage, and the anesthesia safety alarm threshold of the vital sign parameters is dynamically calculated according to the current operation anesthesia stage.

8. A digital anesthesia safety management device, characterized in that, The method comprises: a collection unit configured to collect the operation scene of a surgical patient, drug infusion data, ventilation parameters, vital sign parameters, and patient basic characteristics; an analysis unit configured to evaluate the risk level of the surgical patient based on the operation scene, drug infusion data, ventilation parameters, vital sign parameters, and patient basic characteristics; a decision unit configured to perform hierarchical alarm according to the risk level of the surgical patient obtained by evaluation.

9. An electronic device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is configured to implement the steps of the digital anesthesia safety management method according to any one of claims 1-7 when executing the computer program stored in the memory.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executable by the processor to implement the digital anesthesia safety management method according to any one of claims 1-7.

Images