An airway cannula flow monitoring risk assessment method for oral surgery

Abstract

The application discloses an airway cannula flow monitoring risk assessment method for oral anesthesia surgery, collects real-time ventilation monitoring data of an airway cannula patient in oral anesthesia surgery; according to a pre-set screening condition, an instant is screened from the collected ventilation monitoring data as a monitoring reference instant, three types of mapping relationships of a linear correlation model, a waveform correlation model and a multi-parameter coupling model are constructed; based on a pre-constructed normal ventilation data set, a single parameter fluctuation threshold library and an associated parameter fluctuation threshold library are respectively constructed and classified, and thresholds in the single parameter fluctuation threshold library and the associated parameter fluctuation threshold library are updated in real-time monitoring; the real-time collected ventilation monitoring data are compared with the thresholds in the single parameter fluctuation threshold library and the associated parameter fluctuation threshold library of the corresponding category, a risk is automatically judged and marked according to a comparison result by using a Boolean algebra expression, and a risk assessment report is output.

Description

Technical Field

[0001] This invention relates to the field of airway intubation monitoring technology, and in particular to a risk assessment method for monitoring airway intubation flow in oral anesthesia surgery. Background Technology

[0002] With the widespread use of oral anesthesia in surgery, endotracheal intubation, as a core means of ensuring patient ventilation safety during surgery, directly impacts surgical risk control and patient safety. Under general anesthesia, patients' spontaneous breathing is suppressed, relying entirely on mechanical ventilation to maintain gas exchange. Therefore, real-time and accurate monitoring of ventilation parameters is crucial for preventing complications such as hypoxemia, barotrauma, and inadequate ventilation. In current clinical practice, parameters such as tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow rate waveform have been incorporated into routine monitoring systems to reflect airway patency and lung compliance.

[0003] Among these, risk assessment methods based on multi-parameter fusion are gradually becoming a research direction for improving the intelligence level of airway management. An ideal risk warning mechanism should be able to capture the inherent physiological coupling relationship between parameters, rather than judging whether a single indicator exceeds its limit in isolation. For example, changes in tidal volume are usually accompanied by corresponding adjustments in peak airway pressure, and the morphological characteristics of the flow waveform are also closely related to the duration of the inspiratory phase and minute ventilation. However, existing monitoring systems generally use fixed thresholds to trigger alarms, lacking the ability to model the dynamic correlation of parameters, leading to high false alarm rates or missed reports of key abnormal events.

[0004] Existing technologies for airway risk assessment suffer from three limitations: First, they rely solely on preset static thresholds to judge single parameter exceedances, ignoring the differences in reasonable fluctuation ranges between parameters under different ventilation states. Second, they fail to establish quantifiable mapping relationships between parameters, making it impossible to identify hidden risks such as "normal numerical values ​​but abnormal combinations." Third, risk output often incorporates clinical inferences, blurring the boundary between monitoring data and disease diagnosis, and easily leading to medical liability disputes. These shortcomings significantly weaken the reliability and practicality of monitoring systems, especially in scenarios with limited operating space and frequent airway interference, such as oral and maxillofacial surgery. Therefore, there is an urgent need for a risk assessment method for airway intubation flow monitoring that is based solely on the intrinsic correlation of multi-source ventilation data, achieves objective quantification through a dual-dimensional threshold mechanism, and strictly avoids diagnostic conclusions. Summary of the Invention

[0005] The purpose of this invention is to provide a risk assessment method for airway intubation flow monitoring in oral anesthesia surgery, addressing the problems existing in the prior art.

[0006] The technical solution adopted to achieve the purpose of this invention is: A method for risk assessment of airway intubation flow monitoring in oral anesthesia surgery includes the following steps: Step 1: Collect real-time ventilation monitoring data of patients with airway intubation during oral anesthesia surgery. The collected ventilation monitoring data includes five types of parameters: tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform. Step 2: Based on the pre-set screening conditions, select a time point as the monitoring baseline time point T from the ventilation monitoring data collected in Step 1. Extract the tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform at the monitoring baseline time point T. Decompose the flow waveform into three feature values: peak flow rate, waveform morphology, and inspiratory phase duration. Based on the statistical modeling principle of using a single parameter as the independent variable and other parameters as the dependent variable, construct three types of mapping relationships: linear correlation model, waveform correlation model, and multi-parameter coupling model. Step 3: Based on the pre-built normal ventilation dataset, construct a single parameter fluctuation threshold library and a related parameter fluctuation threshold library respectively, and classify the thresholds according to different patient groups and different surgical types to obtain the corresponding single parameter fluctuation threshold library and related parameter fluctuation threshold library. During real-time monitoring, update the thresholds in the single parameter fluctuation threshold library and related parameter fluctuation threshold library according to the real-time collected ventilation monitoring data. Step 4: Compare the real-time collected ventilation monitoring data with the thresholds in the corresponding category of single parameter fluctuation threshold library and related parameter fluctuation threshold library obtained in Step 3. Based on the comparison results, use Boolean algebra expressions to automatically determine the risk and mark it, and output a risk assessment report.

[0007] In the above technical solution, during the real-time acquisition of ventilation monitoring data, an interference identification subroutine is continuously run. This subroutine identifies transient interference events such as sensor calibration deviation, intubation adjustment, or suctioning operation based on accelerometer signals or pressure change slope markers for acquiring ventilation monitoring data, and automatically removes data segments containing transient interference event markers, retaining only valid data segments under continuous and stable ventilation conditions for subsequent analysis.

[0008] In the above technical solution, step 1 includes the following steps: S1.1: Simultaneously acquire ventilation monitoring data of patients with airway intubation during oral anesthesia surgery through a high-precision sensor array integrated into the anesthesia machine or independent monitor; S1.2: The signal acquired by the high-precision sensor is processed using a second-order Butterworth low-pass filter; S1.3: Perform analog-to-digital conversion on the processed high-precision sensor signal to obtain ventilation monitoring data.

[0009] In the above technical solution, the pre-set screening conditions are that the fluctuation amplitude of each parameter within 5 consecutive complete respiratory cycles is less than 5% of its average value over the past 10 minutes, the flow waveform remains consistent, and there are no interference event markers.

[0010] In the above technical solution, the flow waveform at the monitoring reference time T is further decomposed into three characteristic values: the peak flow rate is obtained by finding the maximum flow rate in the intake phase; the waveform morphology is obtained by calculating the skewness coefficient γ of the flow curve in the intake phase. Classify, when γ When the value is -0.5, it is determined to be a decreasing wave, and -0.5 ≤ γ. When ≤0.5, it is determined to be a square wave, γ A value of 0.5 is considered a sine wave; the duration of the inspiratory phase is obtained by detecting the time interval between the flow rate rising from the baseline of 10% to the peak and then falling back to 10%.

[0011] In the above technical solution, the different patient groups include children, adults and the elderly, with children defined as under 12 years old and the elderly as over 65 years old; the different surgical types include tooth extraction and maxillofacial surgery.

[0012] In the above technical solution, step 3 includes the following steps: S3.1: In the normal ventilation dataset, the initial threshold is calculated by using the 95% confidence interval method to calculate the distribution of each parameter in the normal population, and the initial threshold is reasonably corrected in combination with clinical ventilation practice, and a single parameter fluctuation threshold library is constructed. S3.2: Based on the three types of mapping relationships constructed in step 2—linear correlation model, waveform correlation model, and multi-parameter coupling model—calculate the actual parameter values ​​Y in the normal ventilation dataset. actual The predicted value Y of linear correlation model, waveform correlation model, and multi-parameter coupling model predicted The absolute deviation between |Y actual -Y predicted Collect all deviation values ​​and calculate the upper limit of the maximum deviation ΔY using the interquartile range method. max A library of correlation parameter fluctuation thresholds is constructed based on the maximum deviation upper limit; S3.3: Based on different patient groups and different surgical types, the constructed single parameter fluctuation threshold library and related parameter fluctuation threshold library are classified to obtain the corresponding type of single parameter fluctuation threshold library and related parameter fluctuation threshold library; S3.4: During real-time monitoring, based on the real-time collected ventilation monitoring data, the moving average and standard deviation of each parameter are repeatedly updated within a preset time. Based on the update results, the corresponding single parameter fluctuation threshold library and related parameter fluctuation threshold library are dynamically adjusted. Furthermore, the first-order difference sign consistency is used to determine whether the parameters in all collected ventilation monitoring data show a monotonically increasing or decreasing trend within 10 consecutive sampling points. If so, the threshold adaptive update is triggered to prevent hysteresis alarms caused by slow drift.

[0013] In the above technical solution, the calculation of the maximum deviation upper limit ΔY using the interquartile range method is described. max The expression is as follows:

[0014] In the formula, This represents the upper limit of the maximum deviation; Represents the interquartile range ( ); The upper quartile, representing all deviation values, can be considered as the 75th percentile of the quartile method; Q 1 represents the lower quartile of all deviation values, which can be regarded as the 25th percentile of the quartile method.

[0015] In the above technical solution, step 4 includes the following steps: S4.1: Compare the five parameters collected in real time—tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform—with the corresponding thresholds in the single-parameter fluctuation threshold library obtained in step 3. If a parameter collected in real time exceeds the threshold range, Boolean algebra expressions are used to automatically determine that the parameter is an abnormal risk and mark it.

[0016] S4.2: Calculate the absolute deviations between the five parameters collected in real time—tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform—and the predicted values ​​of the linear correlation model, waveform correlation model, and multi-parameter coupling model, respectively. Compare these deviations with the thresholds in the correlation parameter fluctuation threshold library obtained in step 3. If any absolute deviation exceeds the threshold range, use a Boolean algebra expression to automatically determine that the parameter corresponding to the absolute deviation is at risk of abnormality and mark it.

[0017] In the above technical solution, the risk assessment report is displayed intuitively through a graphical interface, using four colors—green, yellow, orange, and red—to indicate normal, mild, moderate, and severe risk states. Abnormal parameters are highlighted by flashing, and a text summary conforming to the HL7 standard is generated for electronic medical record archiving.

[0018] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention achieves a technological leap from isolated parameter monitoring to dynamic coupling analysis between parameters by constructing a two-dimensional threshold system based on the intrinsic correlation of multi-source ventilation parameters, effectively overcoming the problems of high false alarm rate and hidden anomalies in traditional fixed threshold alarm mechanisms.

[0019] 2. By establishing two types of quantitative standards—the fluctuation threshold of a single parameter and the fluctuation threshold of related parameters—this invention can accurately identify complex risk scenarios where “the values ​​are normal but the combinations are abnormal,” significantly improving the sensitivity and specificity of risk warnings.

[0020] 3. By introducing a local modeling strategy for monitoring the baseline time T and a dynamic threshold adjustment mechanism, this invention enhances the system's adaptability to individual differences and intraoperative physiological changes, thereby improving the personalization of the assessment results.

[0021] 4. This invention clarifies the functional boundaries of the monitoring tool by strictly limiting the risk output to only a description of data anomalies without attaching any clinical diagnostic conclusions, thereby reducing the risk of medical liability disputes. The overall solution relies on objective data-driven and mathematical modeling to achieve the standardization, quantification and de-subjectivity of airway intubation status risk assessment, providing reliable technical support for ventilation safety management in oral anesthesia surgery. Attached Figure Description

[0022] Figure 1 The diagram shown is a flowchart of the airway intubation flow monitoring risk assessment method described in this invention. Detailed Implementation

[0023] The present invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. Example 1

[0024] A risk assessment method for airway intubation flow monitoring in oral anesthesia surgery, see [link to relevant documentation]. Figure 1 This includes the following steps: Step 1: Collect real-time ventilation monitoring data for patients with airway intubation during oral anesthesia surgery. The collected ventilation monitoring data includes five types of parameters: tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow rate waveform.

[0025] In this embodiment, during the real-time acquisition of ventilation monitoring data, an interference identification subroutine is continuously run. This subroutine is based on accelerometer signals (if equipped with intubation) or pressure change slope (dP / dt). Transient interference events such as sensor calibration deviation, intubation adjustment, or suctioning operations that occur when ventilation monitoring data is collected (with a continuous flow of 50 cmH2O / s for more than 200 milliseconds) are marked. Data segments containing transient interference event markers are automatically removed, and only valid data segments under continuous and stable ventilation conditions are retained for subsequent analysis.

[0026] Step 1 includes the following steps: S1.1: Ventilation monitoring data of patients with airway intubation during oral anesthesia surgery is synchronously acquired through a high-precision sensor array integrated into the anesthesia machine or independent monitor. Specifically, the sensor for tidal volume uses a thermal mass flow meter principle with a sampling frequency set to 125Hz; the sensor for peak airway pressure is a piezoresistive silicon micromechanical sensor with a range covering 0 to 60 cmH2O and a sampling frequency of 200Hz; the respiratory rate is calculated in real-time using a flow integral zero-crossing detection algorithm; the flow waveform is sampled at a raw rate of 1000Hz after high-pass filtering by the thermal sensor, and then downsampled to 100Hz to match the system master clock; the positive end-expiratory pressure is the average end-expiratory airway pressure over a continuous sampling period.

[0027] In this embodiment, the high-precision sensor has a sampling frequency of no less than 100Hz, which ensures that the consistency error of the timestamps of various parameters in the collected ventilation monitoring data does not exceed 1 millisecond. Furthermore, it eliminates instantaneous operational interference such as sensor calibration deviations and intubation adjustments during the acquisition process, retaining only the valid data segments under stable ventilation conditions. S1.2: The signal acquired by the high-precision sensor is processed using a second-order Butterworth low-pass filter (the cutoff frequency is set to 45Hz to eliminate high-frequency electromagnetic interference).

[0028] S1.3: Perform analog-to-digital conversion on the processed high-precision sensor signal to obtain ventilation monitoring data. The ventilation monitoring data includes a 16-byte header (containing timestamp, patient ID, and device serial number) and a 32-byte payload (containing 5 main parameters and their derived features).

[0029] In this embodiment, tidal volume is standardized to mL / kg based on patient weight, and peak airway pressure and positive end-expiratory pressure are measured in cmH2O. O, respiratory rate is measured in breaths / min, and the core characteristics of the flow waveform include peak flow (in mL / s), waveform morphology (divided into square wave, sine wave, and decrementing wave), and inspiratory phase duration (in seconds).

[0030] Step 2: Based on pre-set screening criteria, a time point T is selected from the ventilation monitoring data collected in Step 1 as the monitoring baseline time point. Tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform are extracted from this monitoring baseline time point T. The flow waveform is decomposed into three feature values: peak flow rate, waveform morphology, and inspiratory phase duration. Based on the statistical modeling principle of using a single parameter as the independent variable and other parameters as the dependent variable, three types of mapping relationships are constructed: linear correlation model, waveform correlation model, and multi-parameter coupling model. The tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform parameters extracted from the monitoring baseline time point T are used to initialize the localized calibration of the three mapping models. This involves weighted fusion of the global model coefficients with the measured values ​​at time T to generate a personalized prediction model suitable for the current patient, ensuring that subsequent predictions conform to the current patient's physiological response characteristics.

[0031] The pre-set screening criteria are that the fluctuation amplitude (defined as the difference between the maximum and minimum values) of each parameter (tidal volume, peak airway pressure, positive end-expiratory pressure, and respiratory rate) within 5 consecutive complete respiratory cycles is less than 5% of its average value over the past 10 minutes, and the flow waveform remains consistent (i.e., it belongs to square wave, sine wave, or decreasing wave), while there are no interference event markers. The midpoint of the 5 complete respiratory cycles is taken as the monitoring baseline time T, and the instantaneous values ​​of all parameters at the monitoring baseline time T are recorded for initializing the local calibration of the mapping model to ensure that subsequent predictions conform to the current physiological response characteristics of the patient.

[0032] The flow waveform at the monitoring reference time T is further decomposed into three characteristic values: peak flow rate is obtained by finding the maximum flow rate in the intake phase; waveform morphology is obtained by calculating the skewness coefficient γ of the flow curve in the intake phase. Classify, when γ When the value is -0.5, it is determined to be a decreasing wave, and -0.5 ≤ γ. When ≤0.5, it is determined to be a square wave, γ A value of 0.5 is considered a sine wave; the duration of the inspiratory phase is obtained by detecting the time interval between the flow rate rising from the baseline of 10% to the peak and then falling back to 10%.

[0033] The linear correlation model calculates the linear correlation between pairs of parameters using the Pearson correlation coefficient and establishes a regression equation. For example, the relationship between tidal volume (VT) and peak airway pressure (PIP) is expressed as PIP = a × VT + b, where the regression coefficients a and b are obtained by fitting normal ventilation data from no less than 5,000 historical oral anesthesia surgeries, and the model's coefficient of determination (R²) must be no less than 0.75 to be valid. Similarly, the relationship between respiratory rate (RR) and minute ventilation (VE), and the relationship between positive end-expiratory pressure (PEEP) and lung compliance (C) are also established using the same method.

[0034] The waveform correlation model calculates tidal volume based on peak flow rate and inspiratory phase duration combined with waveform correction coefficients, and establishes a functional relationship between peak airway pressure, peak flow rate, and positive end-expiratory pressure; the expression for tidal volume is as follows:

[0035] In the formula, Represents tidal volume; The waveform correction coefficient is set to 1.0 for square wave, 0.8 for sine wave, and 0.9 for decreasing wave, depending on the waveform. This coefficient is determined through regression analysis of large sample data to compensate for the nonlinear error of flow-volume integral under different waveforms. Represents peak traffic; This represents the duration of the inspiratory phase.

[0036] The functional relationship between peak airway pressure, peak flow rate, and positive end-expiratory pressure is expressed as follows:

[0037] In the formula, This represents the functional relationship between peak airway pressure, peak flow rate, and positive end-expiratory pressure. , , These represent the fitting coefficients for historical data; Represents peak traffic; It represents positive end-expiratory pressure.

[0038] The multi-parameter coupling model employs partial least squares regression, using tidal volume, respiratory rate, and flow waveform (characterized by peak flow and inspiratory duration) as input variables X to predict the output values ​​Y of peak airway pressure and positive end-expiratory pressure. Its core lies in extracting the latent components of the input variables to maximize the explanation of the variance of the output variables, thereby covering the indirect physiological coupling paths between parameters. For example, changes in tidal volume indirectly alter airway resistance by affecting alveolar expansion, thus influencing peak airway pressure. The cross-validation accuracy of the multi-parameter coupling model is no less than 90%. Input variables are preprocessed using Z-score standardization before being fed into the multi-parameter coupling model, i.e., (x - μ) / σ, where μ and σ are the mean and standard deviation of the training set, respectively.

[0039] Step 3: Based on the pre-constructed normal ventilation dataset, construct single-parameter fluctuation threshold libraries and associated-parameter fluctuation threshold libraries respectively. Classify the thresholds according to different patient groups and different surgical types to obtain corresponding single-parameter fluctuation threshold libraries and associated-parameter fluctuation threshold libraries. During real-time monitoring, update the thresholds in the single-parameter fluctuation threshold libraries and associated-parameter fluctuation threshold libraries based on the real-time collected ventilation monitoring data. The normal ventilation dataset is a large-sample normal ventilation dataset (constructed by merging ventilation data from healthy, normal individuals of each age group in children, adults, and the elderly), with no fewer than 2000 cases. The different patient groups include children, adults, and the elderly; children are defined as under 12 years old, and the elderly as over 65 years old. The thresholds for each age group are independently generated based on large-sample statistical data of the corresponding population. The different surgical types include tooth extraction and maxillofacial surgery.

[0040] Step 3 includes the following steps: S3.1: In the normal ventilation dataset, the initial threshold is calculated by using the 95% confidence interval method to calculate the distribution of each parameter in the normal population (i.e., the threshold range is equal to the mean plus or minus 1.96 times the standard deviation, the lower threshold = μ - 1.96σ, the upper threshold = μ + 1.96σ, where μ and σ represent the mean and standard deviation of the parameter in the normal ventilation dataset, respectively). The initial threshold is then reasonably corrected in combination with clinical ventilation practice, and a single parameter fluctuation threshold library is constructed.

[0041] For example, if the adult tidal volume μ is 8 mL / kg and σ is 1.2 mL / kg, then the baseline threshold range is 5.65 to 10.35 mL / kg. A committee composed of at least three senior anesthesiologists conducts a clinical feasibility review of the threshold range of the tidal volume parameter to avoid the incorrect exclusion of extreme but reasonable physiological manifestations (such as patients with acute respiratory distress syndrome who are receiving protective lung ventilation strategies, whose tidal volume is often set at 6 to 8 mL / kg). Ultimately, a standardized threshold library of clinically acceptable tidal volume fluctuation thresholds is formed.

[0042] S3.2: Based on the three types of mapping relationships constructed in step 2—linear correlation model, waveform correlation model, and multi-parameter coupling model—calculate the actual parameter values ​​Y in the normal ventilation dataset. actual The real-time values ​​collected during monitoring are compared with the predicted values ​​Y from linear correlation models, waveform correlation models, and multi-parameter coupling models. predicted The absolute deviation between |Y actual -Y predicted Collect all deviation values ​​and calculate the upper limit of the maximum deviation ΔY using the interquartile range method. max A threshold library for correlation parameter fluctuations is constructed based on the maximum deviation upper limit. The interquartile range method is used to calculate the maximum deviation upper limit, which effectively resists outlier interference and ensures threshold robustness.

[0043] The maximum deviation upper limit ΔY is calculated using the interquartile range method. max The expression is as follows:

[0044] In the formula, This represents the upper limit of the maximum deviation; Represents the interquartile range ( ); The upper quartile, representing all deviation values, can be considered as the 75th percentile of the quartile method; Q 1 represents the lower quartile of all deviation values, which can be regarded as the 25th percentile of the quartile method.

[0045] S3.3: Based on different patient groups and different surgical types, the constructed single parameter fluctuation threshold library and related parameter fluctuation threshold library are classified to obtain the corresponding type of single parameter fluctuation threshold library and related parameter fluctuation threshold library.

[0046] S3.4: During real-time monitoring, based on the real-time collected ventilation monitoring data, the moving average and standard deviation of each parameter are repeatedly updated within a preset time. Based on the update results, the corresponding single parameter fluctuation threshold library and related parameter fluctuation threshold library are dynamically adjusted. Furthermore, the first-order difference sign consistency is used to determine whether any parameter in all collected ventilation monitoring data shows a monotonically increasing or decreasing trend within 10 consecutive sampling points (100 milliseconds). If so, an adaptive threshold update is triggered to prevent delayed alarms caused by slow drift (such as the endotracheal tube gradually kinking).

[0047] Specifically, in this embodiment, the data is updated every 30 seconds during real-time monitoring. The moving average and standard deviation of each parameter within 30 seconds are calculated and compared with the moving average and standard deviation corresponding to the current single parameter fluctuation threshold and the associated parameter fluctuation threshold. If the difference is greater than 5%, the thresholds in the current single parameter fluctuation threshold library and the associated parameter fluctuation threshold library of the corresponding type are dynamically fine-tuned. The adjustment range does not exceed 10% of the previous threshold to adapt to individualized ventilation status changes.

[0048] Step 4: Compare the real-time collected ventilation monitoring data with the thresholds in the corresponding category's single parameter fluctuation threshold library and related parameter fluctuation threshold library obtained in Step 3. Based on the comparison results, Boolean algebra expressions are used to automatically determine the risk and mark it, and a risk assessment report is output. The risk assessment report is a structured report containing the risk level, details of abnormal parameters, and their deviation degree (e.g., "tidal volume exceeds the upper limit by 12.3%", "VT-PIP correlation deviation reaches 35.7%").

[0049] Step 4 includes the following steps: S4.1: Compare the five parameters collected in real time—tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform—with the corresponding thresholds in the single-parameter fluctuation threshold library obtained in step 3. If a parameter collected in real time exceeds the threshold range, Boolean algebra expressions are used to automatically determine that the parameter is an abnormal risk and mark it.

[0050] S4.2: Calculate the absolute deviations between the five parameters collected in real time—tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform—and the predicted values ​​of the linear correlation model, waveform correlation model, and multi-parameter coupling model, respectively. Compare these deviations with the thresholds in the correlation parameter fluctuation threshold library obtained in step 3. If any absolute deviation exceeds the threshold range, use a Boolean algebra expression to automatically determine that the parameter corresponding to the absolute deviation is at risk of abnormality and mark it.

[0051] Among them, single parameter anomalies are connected by OR gates, that is, any parameter anomaly will trigger the single parameter anomaly risk flag. Composite risks are determined by AND-OR hybrid logic gate structure, specifically (number of single parameter anomalies ≥ 2) OR (number of single parameter anomalies ≥ 1 AND number of associated parameter anomalies ≥ 1).

[0052] Specifically, this embodiment uses an embedded ARM Cortex-M7 processor to complete the anomaly risk assessment within milliseconds, ensuring real-time performance meets clinical needs. The anomaly risk level is determined based on the extent to which the parameter exceeds the threshold: a single parameter exceeding the threshold by no more than 10% or a correlation deviation of no more than 20% without any composite anomalies is considered mild risk; a single parameter exceeding the threshold by 10% to 30% or a correlation deviation of 20% to 50%, or the presence of one set of composite anomalies, is considered moderate risk; a single parameter exceeding the threshold by 30% or more, or a correlation deviation of 50% or more, or the presence of two or more sets of composite anomalies, is considered severe risk. When a severe risk or a continuously escalating risk sequence is identified (such as escalating from mild to severe within 10 seconds), a multi-channel alarm mechanism is activated. This involves simultaneously triggering an audible and visual alarm (buzzer frequency 2000Hz, LED flashing frequency 2Hz), remotely notifying the anesthesiologist's mobile terminal via the hospital's internal 5G private network, and uploading an event snapshot to the hospital's data center. The alarm delay time does not exceed 2 seconds. The event record includes a complete parameter curve from 30 seconds before the risk occurs to 60 seconds after. The sampling rate is maintained at 100Hz, and the data is stored after being encrypted with AES-256 for easy post-event review and analysis.

[0053] The risk assessment report is displayed intuitively through a graphical interface, using green, yellow, orange, and red to indicate normal, mild, moderate, and severe risk states. Abnormal parameters are highlighted by flashing. At the same time, a text summary conforming to the HL7 standard is generated for electronic medical record archiving. The information obtained by the airway intubation flow monitoring risk assessment method in this embodiment is strictly limited to the scope of data anomaly description and does not involve etiological speculation, pathological naming, or treatment suggestions. An additional statement is attached: "This risk assessment report is generated only based on abnormal airway intubation monitoring data and does not constitute any clinical disease diagnosis conclusion. It is only for medical staff to refer to in order to adjust ventilation strategies." Example 2

[0054] Based on Example 1, in order to further illustrate the practical application effect of the airway intubation flow monitoring risk assessment method described in Example 1, an example is given of a 58-year-old male patient weighing 70kg who underwent mandibular osteotomy under general anesthesia for a mandibular cyst.

[0055] After anesthesia induction and endotracheal intubation, the system corresponding to the airway intubation flow monitoring risk assessment method described in Example 1 was connected. The system entered a stable monitoring state 15 minutes after the start of surgery. First, ventilation monitoring data was continuously collected at a sampling frequency of 125Hz, with a timestamp error... 0.5 milliseconds. Secondly, a stable window of 5 respiratory cycles was identified 20 minutes after the start of surgery (parameter fluctuations...). 4%), selected the monitoring baseline time T, extracted VT=8.2 mL / kg, PIP=18 cmH O, PEEP = 5cmH O, RR = 12 times / min, flow waveform is square wave, PF = 600 mL / s, Ti = 1.2 s. Based on this, the system initializes a localized model, for example, the coefficients of the VT-PIP linear model are a = 1.8, b = 3.2 (derived from the fusion of the global model and the T-time value). Then, a threshold library specific to adult maxillofacial surgery is loaded, with VT thresholds of 6~10 mL / kg and PIP thresholds of 15~22 cmH. The endotracheal tube was adjusted every 30 seconds. At the 45-minute mark of the surgery, a change in the patient's head and neck position caused a slight kink in the endotracheal tube. The final result showed that VT had decreased to 5.8 mL / kg (exceeding the lower limit by 3.3%, a single abnormal parameter), while the VT-PIP model predicted a PIP of 13.6 cmH. O, but the actual measured PIP was 24cmH. O, the deviation reached 10.4 cmH O, exceeding the correlation threshold ΔY max =8.5cmH O (abnormal related parameters) constitutes a compound anomaly risk; due to VT deviating by 12.7% ( If the airway malfunctions (within 10%) and multiple abnormalities are present, the system classifies it as a moderate risk, displays an orange warning on the interface, highlights the VT and PIP parameters, and pushes a notification to the anesthesiologist's handheld terminal. The physician immediately checks the airway and finds the tube kinked; after repositioning, the collected parameters return to normal. The entire process, from the occurrence of the abnormality to the alarm output, takes 1.8 seconds, effectively preventing further deterioration of insufficient ventilation. Example 3

[0056] Building upon Examples 1-2, the normal ventilation dataset described in Example 1 comprises an oral anesthesia ventilation database with no fewer than 100,000 records. This database is deployed on a cloud server cluster using a distributed NoSQL architecture (such as a MongoDB sharded cluster) to support high-concurrency writes and complex queries. Each record contains complete patient metadata (age, gender, weight, height, body mass index, underlying diseases such as COPD or asthma, ASA classification), surgical information (type, duration, position), full sequence of ventilation parameters (sampling at 100Hz, continuous throughout the surgical procedure), risk event markers (generated by the system in Example 1), and postoperative outcome data (such as whether a hypoxic event occurred, extubation time). The data undergoes rigorous cleaning before being stored: removing missing data. For 5% of the records, outliers are detected and corrected using the Isolation Forest algorithm, and the time series are aligned to a unified UTC timestamp. This database is used for continuous optimization of the mapping model and threshold library: a batch model training task is performed weekly, using the Spark MLlib framework for parallel processing on a cluster. For linear association models, the system recalculates the Pearson correlation coefficient and regression coefficient for all parameter pairs and updates the coefficient of determination R². If the new model's R² improves by more than 0.02 and the p-value is... If the value is 0.01, the old model is replaced. For waveform correlation models, the system analyzes the optimal values ​​of different waveform correction coefficients k in various patient subgroups (such as COPD patients) and dynamically adjusts the rules for determining the value of k. For multi-parameter coupling models, the system attempts to introduce new machine learning algorithms, such as gradient boosting trees (GBDT) or long short-term memory networks (LSTM), to capture nonlinearity and time dependence, and selects the best performer through cross-validation.

[0057] The system corresponding to the airway intubation flow monitoring risk assessment method described in Example 1 uses unsupervised learning algorithms (such as DBSCAN clustering) to mine potential new association patterns: the high-dimensional ventilation parameter space is divided into several clusters, and the covariance of parameters within each cluster is analyzed. If a specific parameter combination (such as low VT, high RR, and decreasing wave) is frequently found to be accompanied by risk events in a certain cluster, new association rules are generated and incorporated into the mapping model system in step 2 of Example 1. The optimized model and threshold library are packaged into a model update package and pushed to each terminal device through a secure OTA (Over-The-Air) mechanism. The terminal device automatically downloads and verifies the update package (through digital signature verification) during non-surgical periods and activates the new model upon the next power-on. This closed-loop learning mechanism significantly improves the long-term applicability of the system, enabling it to adapt to constantly changing clinical practices and patient characteristics. For example, after accumulating sufficient data on elderly obese patients in the database, the system automatically identifies the "high PEEP-low VT" safety combination unique to this group and adjusts its exclusive threshold library accordingly to avoid false alarms for such normal ventilation modes.

[0058] For ease of explanation, spatial relative terms such as “up,” “down,” “left,” and “right” are used in the embodiments to describe the relationship of one element or feature shown in the figures relative to another element or feature. It should be understood that, in addition to the orientations shown in the figures, spatial terms are intended to include different orientations of the device in use or operation. For example, if the device in the figures is inverted, an element described as being “down” of other elements or features would be positioned “up” of those other elements or features. Therefore, the exemplary term “down” can encompass both up and down orientations. The device may be positioned in other ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used herein will be interpreted accordingly.

[0059] Moreover, relational terms such as “first” and “second” are used merely to distinguish one component from another that has the same name, without necessarily requiring or implying any such actual relationship or order between the components.

[0060] The above description is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for risk assessment of airway intubation flow monitoring in oral anesthesia surgery, characterized in that, Includes the following steps: Step 1: Collect real-time ventilation monitoring data of patients with airway intubation during oral anesthesia surgery. The collected ventilation monitoring data includes five types of parameters: tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform. Step 2: Based on the pre-set screening conditions, select a time point as the monitoring baseline time point T from the ventilation monitoring data collected in Step 1. Extract the tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform at the monitoring baseline time point T. Decompose the flow waveform into three feature values: peak flow rate, waveform morphology, and inspiratory phase duration. Based on the statistical modeling principle of using a single parameter as the independent variable and other parameters as the dependent variable, construct three types of mapping relationships: linear correlation model, waveform correlation model, and multi-parameter coupling model. Step 3: Based on the pre-built normal ventilation dataset, construct a single parameter fluctuation threshold library and a related parameter fluctuation threshold library respectively, and classify the thresholds according to different patient groups and different surgical types to obtain the corresponding single parameter fluctuation threshold library and related parameter fluctuation threshold library. During real-time monitoring, update the thresholds in the single parameter fluctuation threshold library and related parameter fluctuation threshold library according to the real-time collected ventilation monitoring data. Step 4: Compare the real-time collected ventilation monitoring data with the thresholds in the corresponding category of single parameter fluctuation threshold library and related parameter fluctuation threshold library obtained in Step 3. Based on the comparison results, use Boolean algebra expressions to automatically determine the risk and mark it, and output a risk assessment report.

2. The risk assessment method for airway intubation flow monitoring according to claim 1, characterized in that, During the real-time acquisition of ventilation monitoring data, an interference identification subroutine is continuously run. This subroutine identifies transient interference events such as sensor calibration deviations, intubation adjustments, or suctioning operations based on accelerometer signals or pressure change slope markers. It automatically removes data segments marked with transient interference events and retains only valid data segments under continuous and stable ventilation conditions for subsequent analysis.

3. The risk assessment method for airway intubation flow monitoring according to claim 1, characterized in that, Step 1 includes the following steps: S1.1: Simultaneously acquire ventilation monitoring data of patients with airway intubation during oral anesthesia surgery through a high-precision sensor array integrated into the anesthesia machine or independent monitor; S1.2: The signal acquired by the high-precision sensor is processed using a second-order Butterworth low-pass filter; S1.3: Perform analog-to-digital conversion on the processed high-precision sensor signal to obtain ventilation monitoring data.

4. The risk assessment method for airway intubation flow monitoring according to claim 1, characterized in that, The pre-set screening criteria are that the fluctuation range of each parameter within 5 consecutive complete respiratory cycles is less than 5% of its average value over the past 10 minutes, the flow waveform remains consistent, and there are no interference event markers.

5. The risk assessment method for airway intubation flow monitoring according to claim 1, characterized in that, The flow waveform at the monitoring reference time T is further decomposed into three characteristic values: peak flow rate is obtained by finding the maximum flow rate in the intake phase; waveform morphology is obtained by calculating the skewness coefficient γ of the flow curve in the intake phase. Classify, when γ When the value is -0.5, it is determined to be a decreasing wave, and -0.5 ≤ γ. When ≤0.5, it is determined to be a square wave, γ A value of 0.5 is considered a sine wave; the duration of the inspiratory phase is obtained by detecting the time interval between the flow rate rising from the baseline of 10% to the peak and then falling back to 10%.

6. The risk assessment method for airway intubation flow monitoring according to claim 1, characterized in that, The different patient groups include children, adults, and the elderly, with children defined as those under 12 years old and the elderly as those over 65 years old; the different surgical types include tooth extraction and maxillofacial surgery.

7. The risk assessment method for airway intubation flow monitoring according to claim 1, characterized in that, Step 3 includes the following steps: S3.1: In the normal ventilation dataset, the initial threshold is calculated by using the 95% confidence interval method to calculate the distribution of each parameter in the normal population, and the initial threshold is reasonably corrected in combination with clinical ventilation practice, and a single parameter fluctuation threshold library is constructed. S3.2: Based on the three types of mapping relationships constructed in step 2—linear correlation model, waveform correlation model, and multi-parameter coupling model—calculate the actual parameter values ​​Y in the normal ventilation dataset. actual The predicted value Y of linear correlation model, waveform correlation model, and multi-parameter coupling model predicted The absolute deviation between |Y actual -Y predicted Collect all deviation values ​​and calculate the upper limit of the maximum deviation ΔY using the interquartile range method. max A library of correlation parameter fluctuation thresholds is constructed based on the maximum deviation upper limit; S3.3: Based on different patient groups and different surgical types, the constructed single parameter fluctuation threshold library and related parameter fluctuation threshold library are classified to obtain the corresponding type of single parameter fluctuation threshold library and related parameter fluctuation threshold library; S3.4: During real-time monitoring, based on the real-time collected ventilation monitoring data, the moving average and standard deviation of each parameter are repeatedly updated within a preset time. Based on the update results, the corresponding single parameter fluctuation threshold library and related parameter fluctuation threshold library are dynamically adjusted. Furthermore, the first-order difference sign consistency is used to determine whether the parameters in all collected ventilation monitoring data show a monotonically increasing or decreasing trend within 10 consecutive sampling points. If so, the threshold adaptive update is triggered to prevent hysteresis alarms caused by slow drift.

8. The method for risk assessment of airway intubation flow monitoring according to claim 7, characterized in that, The maximum deviation upper limit ΔY is calculated using the interquartile range method. max The expression is as follows: In the formula, This represents the upper limit of the maximum deviation; Represents the interquartile range ( ); The upper quartile, representing all deviation values, can be considered as the 75th percentile of the quartile method; Q 1 represents the lower quartile of all deviation values, which can be regarded as the 25th percentile of the quartile method.

9. The risk assessment method for airway intubation flow monitoring according to claim 1, characterized in that, Step 4 includes the following steps: S4.1: Compare the five parameters collected in real time—tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform—with the corresponding thresholds in the single-parameter fluctuation threshold library obtained in step 3. If a parameter collected in real time exceeds the threshold range, Boolean algebra expressions are used to automatically determine that the parameter is an abnormal risk and mark it. S4.2: Calculate the absolute deviations between the five parameters collected in real time—tidal volume, peak airway pressure, positive end-expiratory pressure, respiratory rate, and flow waveform—and the predicted values ​​of the linear correlation model, waveform correlation model, and multi-parameter coupling model, respectively. Compare these deviations with the thresholds in the correlation parameter fluctuation threshold library obtained in step 3. If any absolute deviation exceeds the threshold range, use a Boolean algebra expression to automatically determine that the parameter corresponding to the absolute deviation is at risk of abnormality and mark it.

10. The method for risk assessment of airway intubation flow monitoring according to claim 1, characterized in that, The risk assessment report is displayed intuitively through a graphical interface, using four colors—green, yellow, orange, and red—to indicate normal, mild, moderate, and severe risk statuses. Abnormal parameters are highlighted by flashing. At the same time, a text summary conforming to the HL7 standard is generated for electronic medical record archiving.

Images