A safety air flow early warning and automatic adjustment system for a child respiratory device

Abstract

The present application relates to the technical field of respiratory equipment safety air flow early warning and automatic adjustment, and provides a child respiratory equipment safety air flow early warning and automatic adjustment system, the present application fuses respiratory signal data, blood oxygen signal data and carbon dioxide signal data, constructs an enhanced time sequence characteristic map, comprehensively judges the abnormal trend of multidimensional physiological parameters, identifies risks in advance when a single parameter has not exceeded the limit but multiple parameters show abnormal trends, significantly improves the accuracy and advance of early warning, realizes double confirmation through a blood oxygen-carbon dioxide consistency verification index, effectively reduces the false alarm or false adjustment risk caused by single sensor signal abnormality, introduces a blood oxygen-carbon dioxide double confirmation mechanism in risk level judgment, avoids false adjustment caused by sensor failure, and reduces the risk of inhalation and exhalation equipment use.

Description

Technical Field

[0001] This invention relates to the field of respiratory equipment safety airflow warning and automatic adjustment technology, and in particular to a child respiratory equipment safety airflow warning and automatic adjustment system. Background Technology

[0002] Pediatric respiratory support devices are widely used in the mechanical ventilation treatment of newborns and infants. However, due to the unique physiological structure of children, their respiratory management still faces many challenges. For example, setting safety data in the respiratory device based on the child's individual data, and implementing an automatic self-test mode before use. Current technologies neglect the combined analysis of respiratory signal data, blood oxygen signal data, and carbon dioxide signal data, relying solely on single data points for analysis. Furthermore, the lack of double verification of the data acquisition device fails to reduce the risk of false alarms or misadjustments caused by abnormal signals from the acquisition device, making it difficult to meet the high requirements for safety, timeliness, and accuracy in the use of respiratory devices. Summary of the Invention

[0003] This invention provides a safety airflow warning and automatic adjustment system for children's respiratory equipment, the main purpose of which is to solve the problem of safety airflow warning and automatic adjustment in respiratory equipment.

[0004] To achieve the above objectives, the present invention provides a safety airflow warning and automatic adjustment system for pediatric respiratory equipment, comprising: a device self-test module, which performs a self-test on the respiratory equipment and the monitoring link based on the hardware status data of the respiratory equipment and the connection status data of the additional monitoring equipment, and generates a device ready flag after the self-test passes, and calculates the initial ventilation parameters based on the child's basic physiological parameters.

[0005] The enhanced time-series feature map construction module calculates multidimensional respiratory-monitoring fusion feature vectors for each respiratory cycle based on respiratory signal data, blood oxygen signal data, and carbon dioxide signal data, and constructs an enhanced time-series feature map.

[0006] The index generation module, based on the enhanced time-series feature map, analyzes and derives a set of safety risk indices, a set of regulation and control parameters, and a blood oxygen-carbon dioxide consistency verification index.

[0007] The risk analysis module, based on a set of safety risk indices, sets tiered thresholds for each risk index and performs risk analysis.

[0008] The display terminal generates a complete historical dataset of each respiratory cycle and visualizes it.

[0009] In a preferred embodiment, the hardware status data of the respiratory device includes the zero-point drift value of the flow sensor, the opening response curve of the proportional solenoid valve, and the reference value of the pressure sensor; the connection status data of the additional monitoring device includes the connection status of the pulse oximeter probe, the signal quality index, the preheating status of the carbon dioxide analyzer, and the patency status of the sampling tubing; the basic physiological parameter data of the child includes age, weight, and preset breathing mode.

[0010] In a preferred embodiment, the step of performing a self-test on the respiratory device and monitoring link, generating a device readiness flag after the self-test passes, and calculating initial ventilation parameters based on the child's basic physiological parameters includes: S301, performing a device self-test on the respiratory device based on hardware status data and generating a respiratory device self-test result; performing a link self-test on the monitoring link based on the connection status data of the additional monitoring device and generating a monitoring link self-test result.

[0011] S302: When both the respiratory equipment self-test result and the monitoring link self-test result are in a pass state, generate a device ready flag; when either the respiratory equipment self-test result or the monitoring link self-test result is in a fail state, perform a fault handling operation.

[0012] S303 analyzes and obtains an initial ventilation parameter set based on the child's basic physiological parameter data; and sets alarm thresholds for low blood oxygen saturation and high carbon dioxide partial pressure.

[0013] In a preferred embodiment, the step of calculating the multidimensional respiratory-monitoring fusion feature vector for each respiratory cycle based on respiratory signal data, blood oxygen signal data, and carbon dioxide signal data, and constructing an enhanced time-series feature map, includes: S501, calculating the respiratory feature sub-vector for each respiratory cycle based on preprocessed respiratory signal data; and calculating the blood oxygen-carbon dioxide feature sub-vector for each respiratory cycle based on preprocessed blood oxygen signal data and carbon dioxide signal data.

[0014] S502, the respiratory feature vector and the blood oxygen-carbon dioxide feature vector are concatenated and fused to obtain a multidimensional respiratory-monitoring fusion feature vector; and the multidimensional respiratory-monitoring fusion feature vectors of several consecutive respiratory cycles are arranged in chronological order to construct an enhanced temporal feature map.

[0015] In a preferred embodiment, the blood oxygen-carbon dioxide feature subvector includes the mean blood oxygen saturation, the magnitude of the decrease in blood oxygen saturation, the partial pressure of carbon dioxide at the end of expiration, the upward slope of the carbon dioxide waveform, and the characteristic value of the plateau phase of the carbon dioxide waveform.

[0016] In a preferred embodiment, the calculation of the respiratory feature sub-vector for each respiratory cycle includes: S701, the formula for calculating the leakage is as follows:

[0017] in This represents the amount of air leakage, expressed in milliliters. Tidal volume, measured in milliliters, represents the total volume of air inhaled during one respiratory cycle. Tidal volume, measured in milliliters, represents the total volume of exhaled air during one respiratory cycle.

[0018] S702, the formula for calculating the rising slope of the waveform is as follows:

[0019] in The slope of the waveform is expressed in centimeters of water column per second. Peak pressure, expressed in centimeters of water column, represents the maximum airway pressure during one respiratory cycle; Inspiratory initiation pressure, expressed in centimeters of water column, represents the airway pressure at the start of a respiratory cycle. The inhalation rise time, measured in seconds, represents the time it takes to reach peak pressure from the start of inhalation.

[0020] S703, Inhaled tidal volume is the integral of the instantaneous inhaled flow rate over the duration during the respiratory cycle; Exhaled tidal volume is the absolute value of the integral of the instantaneous exhaled flow rate over the duration during the respiratory cycle; Respiratory rate is the value of 60 seconds divided by the duration of the respiratory cycle; Peak pressure is the maximum airway pressure during the respiratory cycle; Mean pressure is the arithmetic mean of the airway pressure during the respiratory cycle.

[0021] In a preferred embodiment, the step of analyzing and deriving the safety risk index set, the regulation control parameter set, and the blood oxygen-carbon dioxide consistency verification index based on the enhanced temporal feature map includes: inputting the enhanced temporal feature map into a pre-trained convolutional-recurrent hybrid neural network model to obtain a fused attention feature vector, and then performing a nonlinear mapping on the fused attention feature vector to obtain the safety risk index set, the regulation control parameter set, and the blood oxygen-carbon dioxide consistency verification index.

[0022] In a preferred embodiment, generating and visualizing the full-process historical dataset for each respiratory cycle includes: using each respiratory cycle as the recording unit, synchronously collecting the enhanced time-series feature map of the current respiratory cycle, the risk index values ​​in the safety risk index set, the adjustment parameter values ​​in the adjustment control parameter set, the blood oxygen-carbon dioxide consistency verification index value, and the types and timestamps of warning events triggered within the current respiratory cycle, forming a single-cycle respiratory event record, thereby generating a full-process historical dataset for breathing, and thus generating a multi-dimensional waveform trend chart.

[0023] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention provides a safety airflow warning and automatic adjustment system for pediatric respiratory equipment. By fusing respiratory signal data, blood oxygen signal data, and carbon dioxide signal data, an enhanced time-series feature map is constructed. Risk assessment is performed based on the abnormal trends of multiple physiological parameters. Risk is identified in advance when a single parameter has not exceeded its limit but multiple parameters show abnormal trends, significantly improving the accuracy of the warning. Simultaneously, a blood oxygen-carbon dioxide consistency verification index is used for dual confirmation, effectively reducing the risk of false alarms or misadjustments caused by abnormal signals from a single sensor. The introduction of a blood oxygen-carbon dioxide dual confirmation mechanism in risk level assessment avoids misadjustments caused by sensor malfunctions, reducing the risk of using respiratory equipment.

[0024] 2. This invention conducts a comprehensive check on the hardware status and link self-test before the use of children's respiratory equipment, and only equipment that passes the test can be used, thereby improving the safety of respiratory equipment use and reducing the risk of using respiratory equipment; furthermore, through the hierarchical early warning mechanism of the safety risk index set and the adaptive adjustment of the adjustment control parameter set, the children's respiratory equipment can automatically adjust, making data adjustment more accurate and reducing human error. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating a safety airflow warning and automatic adjustment system for a child's respiratory device according to an embodiment of the present invention. Figure 2 This is an example of a multidimensional waveform trend chart of the historical dataset of the entire respiratory process of this invention; Figure 3 This is a flowchart of the device self-test module of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0026] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0027] This invention provides a safety airflow warning and automatic adjustment system for pediatric respiratory equipment. The executing entity of this system includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the system provided in this invention: a server, a terminal, etc. In other words, the safety airflow warning and automatic adjustment system for pediatric respiratory equipment can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0028] Reference Figure 1 The diagram shown is a flowchart of a safety airflow warning and automatic adjustment system for children's breathing equipment provided in an embodiment of the present invention.

[0029] The equipment self-test module is connected to both the enhanced time-series feature map construction module and the risk analysis module. The self-test module sends a generated equipment readiness flag to the enhanced time-series feature map construction module, serving as a condition for the module to initiate data acquisition and map construction; that is, data acquisition and map construction can only begin after the self-test is passed. The enhanced time-series feature map construction module receives respiratory signal data, blood oxygen signal data, and carbon dioxide signal data. The index generation module performs neural network analysis on the received enhanced time-series feature map and transmits the output safety risk index set, regulation control parameter set, and blood oxygen-carbon dioxide consistency verification index to the risk analysis module. The risk analysis module performs hierarchical judgment based on the received index set and simultaneously synchronizes all data to the display terminal. The display terminal receives and records the entire process data generated by each module.

[0030] In this embodiment, the child respiratory equipment safety airflow warning and automatic adjustment system includes: a device self-test module, which performs a self-test on the respiratory equipment and the monitoring link based on the hardware status data of the respiratory equipment and the connection status data of the additional monitoring equipment, and generates a device ready flag after the self-test passes, and calculates the initial ventilation parameters based on the child's basic physiological parameters.

[0031] In this embodiment of the invention, the hardware status data of the respiratory device includes the zero-point drift value of the flow sensor, the opening response curve of the proportional solenoid valve, and the reference value of the pressure sensor; the connection status data of the additional monitoring device includes the connection status of the pulse oximeter probe, the signal quality index, the preheating status of the carbon dioxide analyzer, and the patency status of the sampling tubing; the basic physiological parameter data of the child includes age, weight, and preset breathing mode.

[0032] It should be noted that the hardware status data of the respiratory equipment is a multi-dimensional parameter set of the current working status of each core component; among them, the zero-point drift value of the flow sensor is the baseline offset when there is no airflow; the proportional solenoid valve opening response curve is the correspondence (ratio) between the valve opening command and the actual opening; and the pressure sensor reference value is the reference reading under normal pressure. The connection status data of the auxiliary monitoring equipment is a multi-dimensional parameter set of whether the pulse oximeter and carbon dioxide analyzer are in a normal data acquisition state; among them, the pulse oximeter probe connection status indicates whether the probe is in good contact with the child's skin; the signal quality index indicates the signal-to-noise ratio level of the blood oxygen signal, with a value ranging from 0 to 100, and a higher value indicates better signal quality; the carbon dioxide analyzer preheating status indicates whether the analyzer has completed preheating and reached the working temperature; and the sampling tubing patency status indicates whether there is blockage or condensation accumulation in the carbon dioxide sampling tubing. The child's basic physiological parameter data is the basic vital sign information of the child recorded before the use of the equipment, including preset respiratory modes including volume control mode and pressure control mode.

[0033] In this embodiment of the invention, the self-testing of the respiratory equipment and monitoring link, and the generation of a device readiness flag after the self-test passes, and the calculation of initial ventilation parameters based on the child's basic physiological parameters, include: S301 performs a device self-test on the respiratory device based on hardware status data and generates a respiratory device self-test result; it also performs a link self-test on the monitoring link based on the connection status data of the additional monitoring device and generates a monitoring link self-test result.

[0034] S302: When both the respiratory equipment self-test result and the monitoring link self-test result are in a pass state, generate a device ready flag; when either the respiratory equipment self-test result or the monitoring link self-test result is in a fail state, perform a fault handling operation.

[0035] S303 analyzes and obtains an initial ventilation parameter set based on the child's basic physiological parameter data; and sets alarm thresholds for low blood oxygen saturation and high carbon dioxide partial pressure.

[0036] It should be noted that the device self-test operation is a process of sending self-test commands to each hardware component of the respiratory device and receiving the status codes returned by each component to determine whether the component is working properly. The specific process is as follows: A zero-point calibration command is sent to the flow sensor, the returned zero-point drift value is read and compared with a preset drift tolerance threshold. If the zero-point drift value is within the drift tolerance threshold range, the flow sensor is considered to have passed the self-test; otherwise, it is considered to have failed. A set of test opening command sequences is sent to the proportional solenoid valve, the actual opening response value is read and compared with the theoretical response curve, and the deviation percentage of each test point is calculated. If the deviation percentage of all test points is less than the preset opening deviation tolerance threshold, the proportional solenoid valve is considered to have passed the self-test; otherwise, it is considered to have failed. A reference pressure test command is sent to the pressure sensor, the deviation between the returned reference value and the factory calibration value is read, and if the deviation is within the preset pressure reference tolerance threshold range, the pressure sensor is considered to have passed the self-test; otherwise, it is considered to have failed.

[0037] It should be noted that the link self-test operation sends status query commands to the pulse oximeter and carbon dioxide analyzer to determine whether the monitoring link is normal. The specific process is as follows: The current value of the pulse oximeter probe connection status register is read. If the connection status value is high, the pulse oximeter probe connection is considered normal; otherwise, it is considered abnormal. The signal quality index of the pulse oximeter is read. If the signal quality index is higher than the preset minimum signal quality threshold, the pulse oximeter signal quality is considered qualified; otherwise, it is considered unqualified. The preheating status register of the carbon dioxide analyzer is read. If the preheating status value is complete, the analyzer preheating is considered complete; otherwise, it is considered incomplete. A positive pressure pulse test signal is sent to the carbon dioxide sampling pipeline to detect the pressure holding capacity within the pipeline. If the pressure drop rate is lower than the preset pipeline leakage tolerance rate, the sampling pipeline patency is considered normal; otherwise, it is considered abnormal. Any of the above abnormalities results in a failure.

[0038] It should be noted that the complete logic judgment process is as follows: Simultaneously read the self-test results of the respiratory equipment and the self-test results of the monitoring link. If both are passed, a device ready flag is generated, and the device enters a standby state that can be started normally. If at least one fails, a device ready flag is not generated, and a fault classification judgment operation is triggered instead.

[0039] Further, a fault classification judgment operation is performed, and a fault severity level is generated. Based on the fault severity level, the corresponding fault handling operation is performed. The specific process of the fault classification judgment operation is as follows: First, extract the information of the specific components that failed the self-test results of the respiratory equipment and the information of the specific links that failed the self-test results of the monitoring links. For each failed component or link, identify the fault type and determine whether the fault type is a recoverable or unrecoverable fault. Among them, recoverable faults include: the flow sensor zero point offset exceeds the limit but does not exceed the hardware damage threshold, the pulse oximeter probe connection is loose, the carbon dioxide sampling pipeline has condensate blockage, and the proportional solenoid valve opening deviation is within the calibrable range. Unrecoverable faults include: the flow sensor signal is completely lost, the pressure sensor reference value deviates significantly from the factory calibration value and cannot be corrected, the proportional solenoid valve drive circuit is unresponsive, the pulse oximeter hardware fault code is reported, and the internal light source of the carbon dioxide analyzer fails. If all failed items are recoverable faults, the output fault severity level is Level 1 recoverable fault. If there is at least one unrecoverable fault, the output fault severity level is Level 2 unrecoverable fault.

[0040] It should be noted that the fault handling operations corresponding to the Level 1 recoverable fault are as follows: The specific faulty component prompts and troubleshooting guidance are output on the device display screen, including instructing staff to check whether the pulse oximeter probe is clamped tightly, check whether there is a bend or water accumulation in the carbon dioxide sampling tubing, and prompting the device to automatically re-execute the self-test after troubleshooting; at the same time, the timed retry mechanism is activated, and the retry waiting time is set to 10 seconds. After the waiting time expires, the self-test process is automatically re-executed. If there are still items that fail after 3 consecutive retries, the fault severity level is automatically upgraded to Level 2 unrecoverable fault, and the corresponding fault handling operations are executed.

[0041] Furthermore, the fault handling procedures for Level 2 unrecoverable faults are as follows: output a red highlighted fault alarm message on the device display screen, list the specific unrecoverable fault components, and prevent the breathing device from entering the working state; lock the air output valve of the device, generate a hardware fault log containing the fault component identifier, fault type code, device serial number, and occurrence timestamp, and continuously display the fault prompt on the screen until the staff manually confirms or replaces the fault component and then restarts the self-test process through manual command.

[0042] In this embodiment of the invention, the initial ventilation parameter set obtained by analyzing the child's basic physiological parameter data includes: S401, the calculation formula for the initial target tidal volume is:

[0043] in The initial target tidal volume is expressed in milliliters. The child's weight is expressed in kilograms. The tidal volume coefficient per unit body weight ranges from 4 to 6, with units of milliliters per kilogram. The default value is determined based on the preset breathing mode, with 6 in volume control mode and 4 in pressure control mode.

[0044] S402, the formula for calculating the initial upper limit of respiratory rate is:

[0045] in This is the initial upper limit of respiratory rate, expressed in breaths per minute. The baseline respiratory rate is based on age, measured in breaths per minute, and is obtained by looking up the age classification table. This is the frequency margin, measured in times per minute, with a default value of 10.

[0046] S403, the formula for calculating the initial pressure upper limit is:

[0047] in This represents the initial pressure limit, expressed in centimeters of water column. The upper limit of baseline pressure is based on age, and the unit is centimeters of water column, obtained by looking up the age classification table. This is the pressure margin, expressed in centimeters of water column, with a default value of 2.

[0048] S404, the formula for calculating the initial trigger sensitivity threshold is:

[0049] in This is the initial trigger sensitivity threshold, in milliliters; The trigger sensitivity benchmark is set per unit weight, in milliliters per kilogram, with a default value of 0.5; W represents the child's weight in kilograms.

[0050] It should be noted that the baseline respiratory rate, upper limit of baseline pressure, lower limit alarm threshold of blood oxygen saturation, and upper limit alarm threshold of carbon dioxide partial pressure are detailed in the table below: ; The enhanced time-series feature map construction module calculates multidimensional respiratory-monitoring fusion feature vectors for each respiratory cycle based on respiratory signal data, blood oxygen signal data, and carbon dioxide signal data, and constructs an enhanced time-series feature map.

[0051] In this embodiment of the invention, the step of calculating the multidimensional respiratory-monitoring fusion feature vector for each respiratory cycle based on respiratory signal data, blood oxygen signal data, and carbon dioxide signal data, and constructing an enhanced temporal feature map, includes: S501, calculating the respiratory feature sub-vector for each respiratory cycle based on preprocessed respiratory signal data; and calculating the blood oxygen-carbon dioxide feature sub-vector for each respiratory cycle based on preprocessed blood oxygen signal data and carbon dioxide signal data.

[0052] S502, the respiratory feature vector and the blood oxygen-carbon dioxide feature vector are concatenated and fused to obtain a multidimensional respiratory-monitoring fusion feature vector; and the multidimensional respiratory-monitoring fusion feature vectors of several consecutive respiratory cycles are arranged in chronological order to construct an enhanced temporal feature map.

[0053] It should be noted that respiratory signal data of children are collected in real time using a high-precision hot-wire flow sensor and pressure sensor at a sampling rate of no less than 2000 Hz. The respiratory signal data includes instantaneous inhalation flow rate, instantaneous exhalation flow rate, and airway pressure. Blood oxygen saturation signal data and pulse wave signal data are collected in real time using a pulse oximeter at a sampling rate of no less than 100 Hz. Carbon dioxide waveform signal data, end-tidal carbon dioxide partial pressure data, and respiratory rate data are collected in real time using a carbon dioxide analyzer at a sampling rate of no less than 200 Hz. Signal preprocessing operations are performed on the collected respiratory signal data, blood oxygen signal data, and carbon dioxide signal data to obtain preprocessed respiratory signal data, preprocessed blood oxygen signal data, and preprocessed carbon dioxide signal data.

[0054] Further, the signal preprocessing operation includes four sub-operations executed sequentially: power frequency noise filtering, motion artifact suppression, baseline correction, and abnormal spike removal. Power frequency noise filtering uses a notch filter with a notch frequency set to 50 Hz (or 60 Hz, depending on the local power grid frequency) to filter out periodic interference signals introduced by power line coupling. Motion artifact suppression uses a sliding window standard deviation method, calculating the signal standard deviation within a time window consisting of 50 sampling points before and after each sampling point. When this standard deviation exceeds a preset motion artifact detection threshold, the sampling point is determined to be subject to motion interference, and its signal value is replaced by linear interpolation of adjacent valid sampling points. Baseline correction uses a high-pass filter with a cutoff frequency set to 0.1 Hz to filter out sensor zero drift and low-frequency baseline drift components. Abnormal spike removal uses an amplitude threshold method, comparing the signal amplitude of each sampling point with a preset physiologically reasonable range. When the signal amplitude exceeds the physiologically reasonable range, it is determined to be an abnormal spike and removed, and the average of the preceding and following valid sampling points is used to fill the gap.

[0055] It should be noted that the respiratory feature subvectors include inhaled tidal volume, exhaled tidal volume, respiratory rate, peak pressure, mean pressure, leakage, and waveform rise slope. The respiratory feature subvectors are calculated based on preprocessed respiratory signal data through respiratory cycle segmentation and feature extraction. The specific process of respiratory cycle segmentation is as follows: using the zero-crossing point of the instantaneous inhaled flow rate in the preprocessed respiratory signal data as the cycle segmentation marker, the moment when the instantaneous inhaled flow rate changes from negative to positive and crosses zero is taken as the start point of a respiratory cycle, and the moment when the next instantaneous inhaled flow rate changes from negative to positive and crosses zero is taken as the end point of that respiratory cycle, thus determining the start and end times of each respiratory cycle.

[0056] In this embodiment of the invention, the blood oxygen-carbon dioxide feature subvector includes the mean blood oxygen saturation, the magnitude of the decrease in blood oxygen saturation, the partial pressure of carbon dioxide at the end of expiration, the upward slope of the carbon dioxide waveform, and the characteristic value of the plateau phase of the carbon dioxide waveform.

[0057] It should be noted that the blood oxygen-carbon dioxide feature subvector is calculated based on preprocessed blood oxygen and carbon dioxide signal data. The mean blood oxygen saturation is the arithmetic mean of all blood oxygen saturation sampling points within a respiratory cycle; the decrease in blood oxygen saturation is the difference between the baseline blood oxygen saturation value and the current mean blood oxygen saturation value, where the baseline blood oxygen saturation value is the maximum value of blood oxygen saturation sampling data within the last 60 seconds; the end-expiratory carbon dioxide partial pressure (EECP) is the peak value of the carbon dioxide waveform signal at the end of expiration within a respiratory cycle, taken as the current value of the EECP data output by the carbon dioxide analyzer for that respiratory cycle; the carbon dioxide waveform rise slope is the rate of change of the carbon dioxide waveform from the start of expiration to the peak value at the end of expiration within a respiratory cycle, calculated as the difference between the EECP value and the start-expiratory carbon dioxide partial pressure value, divided by the expiration rise time; the carbon dioxide waveform plateau characteristic value is the average value of the carbon dioxide waveform during the relatively stable phase after reaching its peak within a respiratory cycle, reflecting alveolar ventilation efficiency.

[0058] In this embodiment of the invention, the calculation of the respiratory feature sub-vectors for each respiratory cycle includes: S701, the formula for calculating the leakage is as follows:

[0059] in This represents the amount of air leakage, expressed in milliliters. Tidal volume, measured in milliliters, represents the total volume of air inhaled during one respiratory cycle. Tidal volume, measured in milliliters, represents the total volume of exhaled air during one respiratory cycle.

[0060] S702, the formula for calculating the rising slope of the waveform is as follows:

[0061] in The slope of the waveform is expressed in centimeters of water column per second. Peak pressure, expressed in centimeters of water column, represents the maximum airway pressure during one respiratory cycle; Inspiratory initiation pressure, expressed in centimeters of water column, represents the airway pressure at the start of a respiratory cycle. The inhalation rise time, measured in seconds, represents the time it takes to reach peak pressure from the start of inhalation.

[0062] S703, Inhaled tidal volume is the integral of the instantaneous inhaled flow rate over the duration during the respiratory cycle; Exhaled tidal volume is the absolute value of the integral of the instantaneous exhaled flow rate over the duration during the respiratory cycle; Respiratory rate is the value of 60 seconds divided by the duration of the respiratory cycle; Peak pressure is the maximum airway pressure during the respiratory cycle; Mean pressure is the arithmetic mean of the airway pressure during the respiratory cycle.

[0063] It should be noted that the inhaled tidal volume is the integral of the instantaneous inhaled flow rate over the duration during the respiratory cycle, and is calculated as follows: , This refers to the inhaled tidal volume, expressed in milliliters. This is a function of the instantaneous inhalation flow rate as a function of time, expressed in milliliters per second. This is the start time of the respiratory cycle. This is the time at the end of the respiratory cycle, expressed in seconds.

[0064] It should be noted that the splicing and fusion is a concatenation operation that connects the respiratory feature sub-vector and the blood oxygen-carbon dioxide feature sub-vector along the feature dimension. The respiratory feature sub-vector is a seven-dimensional vector containing seven feature values: inhaled tidal volume, exhaled tidal volume, respiratory rate, peak pressure, mean pressure, leakage, and waveform rise slope. The blood oxygen-carbon dioxide feature sub-vector is a five-dimensional vector containing five feature values: mean blood oxygen saturation, blood oxygen saturation decrease rate, end-expiratory carbon dioxide partial pressure, carbon dioxide waveform rise slope, and carbon dioxide waveform plateau phase feature value. The resulting multidimensional respiratory-monitoring fused feature vector after splicing and fusion is a twelve-dimensional vector.

[0065] It should be noted that the enhanced temporal feature map is a two-dimensional matrix data structure. Using the current time as a reference, the most recent N consecutive respiratory cycles are traced back (N is the preset time window length, with a default value of 30). The N twelve-dimensional multi-dimensional respiratory-monitoring fusion feature vectors corresponding to these N respiratory cycles are arranged chronologically from top to bottom, resulting in an N-row, 12-column two-dimensional matrix. The row number i represents the i-th respiratory cycle traced back (i=1 represents the most recent respiratory cycle, i=N represents the earliest respiratory cycle), and the column number j represents the j-th feature value among the twelve feature types.

[0066] The index generation module, based on the enhanced time-series feature map, analyzes and derives a set of safety risk indices, a set of regulation and control parameters, and a blood oxygen-carbon dioxide consistency verification index.

[0067] In this embodiment of the invention, the step of analyzing and obtaining a set of safety risk indices, a set of adjustment control parameters, and a blood oxygen-carbon dioxide consistency verification index based on an enhanced temporal feature map includes: inputting the enhanced temporal feature map into a pre-trained convolutional-recurrent hybrid neural network model to obtain a fused attention feature vector, and then performing a nonlinear mapping on the fused attention feature vector to obtain the set of safety risk indices, the set of adjustment control parameters, and the blood oxygen-carbon dioxide consistency verification index.

[0068] It should be noted that the convolutional-recurrent hybrid neural network model includes a three-branch parallel output structure after the attention weight allocation layer. The three branches share the same parameters for the convolutional feature extraction layer, the recurrent temporal coding layer, and the attention weight allocation layer, branching only at the final fully connected layer. The first branch extracts the hidden state component with the highest correlation to the respiratory signal features from the fused attention feature vector, outputting a first risk vector based solely on the respiratory signal. The second branch extracts the hidden state component with the highest correlation to the blood oxygen and carbon dioxide signal features from the fused attention feature vector, outputting a second risk vector based solely on the blood oxygen and carbon dioxide signals. The third branch integrates all feature information to output the safety risk index set and the regulation control parameter set. Furthermore, the blood oxygen-carbon dioxide consistency verification index is not directly output by the neural network, but is calculated externally based on the vector distance between the first and second risk vectors. The smaller the vector difference, the higher the consistency verification index, indicating a better match between the risk judgment results of the respiratory signal and the blood oxygen-carbon dioxide signal.

[0069] It should be noted that, through the convolutional feature extraction layer, two-dimensional convolution and max pooling operations are performed on the enhanced temporal feature map to obtain the feature map matrix; through the recurrent temporal coding layer, temporal modeling is performed on the feature map matrix along the time dimension based on the gated recurrent unit to obtain the temporal hidden state vector sequence; through the attention weight allocation layer, adaptive weighted calculation is performed on the contribution of each time step in the temporal hidden state vector sequence to obtain the fused attention feature vector.

[0070] It should be noted that the convolutional-recurrent hybrid neural network model is a pre-trained deep learning model. Its network architecture consists of four sub-network modules connected in sequence: a convolutional feature extraction layer, a recurrent temporal coding layer, an attention weight allocation layer, and a fully connected output layer. The input to the model is an enhanced temporal feature map, which is an N-row, 12-column two-dimensional matrix. N is the preset time window length, with a default value of 30, representing 30 consecutive respiratory cycles. The 12 columns represent twelve feature types, including inhaled tidal volume, exhaled tidal volume, respiratory rate, peak pressure, mean pressure, leakage, waveform rise slope, mean blood oxygen saturation, blood oxygen saturation drop, end-expiratory carbon dioxide partial pressure, carbon dioxide waveform rise slope, and carbon dioxide waveform plateau feature value.

[0071] Furthermore, the convolutional feature extraction layer performs a two-dimensional convolution operation on the input enhanced temporal feature map, using multiple convolutional kernels that slide along the time and feature dimensions to extract spatial correlation features of local regions in the map. After the convolution operation, max pooling is performed on the convolutional output, retaining the maximum value within each pooling window as the output, reducing the spatial size of the feature map while preserving the most salient feature responses. After processing by the convolutional feature extraction layer, a feature map matrix is ​​output.

[0072] It should be noted that the calculation formula for two-dimensional convolution is as follows:

[0073] in The value of the element at the i-th row and j-th column of the convolution output feature map; The input is the enhanced temporal feature map matrix; This is the size of the convolutional kernel in the time dimension, in the number of rows, with a default value of 3. This is the size of the convolution kernel along the feature dimension, in the number of columns; the default value is 3. This represents the weight value at the m-th row and n-th column of the convolution kernel weight matrix; For bias terms; The activation function is a linear rectified function.

[0074] Furthermore, the formula for calculating the max pooling operation is as follows:

[0075] in This represents the value of the element at the p-th row and q-th column in the pooled feature map matrix. The height of the pooled window, in rows, with a default value of 2. The width of the pooled window, in columns, with a default value of 2; This is the pooling step size, with a default value of 2.

[0076] It should be noted that the cyclic temporal coding layer uses gated recurrent units to perform temporal modeling of the feature map matrix along the time dimension. The feature map matrix is ​​split into sequential data according to time steps, with each time step corresponding to a respiratory cycle. The input for each time step is the row vector corresponding to that time step in the feature map matrix. The gated recurrent unit controls the degree of retention of historical information and the degree of fusion with the current input through two gating mechanisms: update gate and reset gate, thereby achieving modeling of the evolution of the respiratory state over time. After iterative calculation for all time steps, the temporal hidden state vector sequence is output.

[0077] Furthermore, the update calculation expression for the gated loop unit at each time step is as follows:

[0078]

[0079]

[0080]

[0081] in The update gate output value for the current time step t controls the proportion of the hidden state from the previous time step that is retained in the current time step; The reset gate output value for the current time step t controls the degree to which the hidden state of the previous time step is reset; Let be the candidate hidden state vector at the current time step t; Let be the hidden state output vector at the current time step t. The hidden state vectors of all time steps are arranged in chronological order to form a sequence of temporal hidden state vectors. This is the hidden state vector of the previous time step t-1; This refers to the row vector in the feature map matrix corresponding to the current time step t; Both are weight matrices; All are bias terms; The Sigmoid activation function maps the output value to the range of 0 to 1. The hyperbolic tangent activation function maps the output value to the range of -1 to 1. This indicates an element-wise multiplication operation.

[0082] Furthermore, the calculation formula for the attention weight allocation layer is as follows:

[0083]

[0084]

[0085] in Let t be the attention score at the t-th time step, representing the importance of that time step to the final decision; Let be the hidden state vector at time step t; The attention vector is a learnable parameter vector. This is the attention weight matrix; This refers to attention bias. Let be the normalized attention weight at time step t, with a value ranging from 0 to 1, and the sum of the attention weights at all time steps is 1. This is the total number of time steps, equal to the preset time window length N, with a default value of 30. It is an exponential function; The fused attention feature vector is the weighted sum of the hidden state vectors at all time steps according to the attention weights.

[0086] Furthermore, the computation expression for the fully connected output layer is as follows:

[0087]

[0088]

[0089] in The safety risk index set is a six-dimensional vector, with each dimension corresponding to the air leakage risk index, barotrauma risk index, apnea risk index, human-machine interaction risk index, oxygenation risk index, and inadequate ventilation risk index. The value of each risk index ranges from 0 to 1, with a larger value indicating a higher risk. The set of control parameters is a six-dimensional vector, with each dimension corresponding to the target gas flow compensation, trigger threshold adjustment, pressure upper limit dynamic constraint, base flow increment, oxygen concentration adjustment, and alarm delay adjustment. The blood oxygen-carbon dioxide consistency verification index is a scalar value that ranges from 0 to 1. The closer the value is to 1, the more consistent the respiratory signal judgment is with the blood oxygen or carbon dioxide signal judgment. These are the weight matrices for the three output branches; These are the bias terms for the three output branches; To fuse attention feature vectors.

[0090] It should be noted that the convolutional-recurrent hybrid neural network model was pre-trained using supervised learning. The training dataset consisted of historical clinical respiratory data, including synchronously collected respiratory signal data, blood oxygen signal data, carbon dioxide signal data, as well as safety event labels and physician adjustment records annotated by clinical experts. The loss function used during training was a weighted combination of mean squared error loss and cross-entropy loss. The mean squared error loss was used to constrain the output accuracy of the control parameter set, while the cross-entropy loss was used to constrain the classification accuracy of the safety risk index set. The optimizer used the Adam adaptive moment estimation optimizer, with an initial learning rate of 0.001, a batch size of 32, and 100 training epochs. An early stopping strategy was employed to prevent overfitting.

[0091] Furthermore, the training dataset consists of historical clinical respiratory therapy data, sourced from real clinical data collected over three years from multiple children's hospitals and neonatal intensive care units. The dataset includes respiratory therapy records of 300 children, ranging in age from 1 day to 3 years and in weight from 0.8 kg to 16 kg, covering the neonatal, infant, and toddler age groups. Each child's record duration ranges from 4 to 72 hours, accumulating over 5000 hours of valid data, totaling approximately 6 million complete respiratory cycles. The dataset annotation was independently completed by three experts with over 5 years of clinical experience in pediatric respiratory therapy. The annotation content includes the following two types of labels: The first type is safety event labels. The annotation method is as follows: After reviewing the synchronously displayed respiratory waveform, blood oxygen waveform, and carbon dioxide waveform, experts annotate each respiratory cycle to indicate whether a safety event exists, specifying the event type (including air leakage, barotrauma, apnea, patient-ventilator asynchrony, poor oxygenation, and inadequate ventilation), and also indicating the event's risk level (low risk, medium risk, high risk). For respiratory cycles with inconsistent annotations, a fourth, higher-level expert will arbitrate to determine the final label. The second type is physician adjustment record labels. The annotation method is as follows: Based on the child's real-time changes in condition, experts annotate the type and amount of adjustment operations that should be performed at the current moment, including target delivery flow compensation, trigger threshold adjustment, pressure upper limit dynamic constraint value, baseline flow increment value, oxygen concentration adjustment, and alarm delay adjustment. This annotation references the adjustment habits of physicians in similar scenarios in clinical practice, allowing the model to learn the adjustment strategies of experienced physicians. The training dataset is divided into training, validation, and test sets in an 8:1:1 ratio. All data is normalized before being input into the model to ensure that each feature value is uniformly mapped to the range of 0 to 1, thus avoiding the influence of features with different dimensions on model training.

[0092] The risk analysis module, based on a set of safety risk indices, sets tiered thresholds for each risk index and performs risk analysis.

[0093] In this embodiment of the invention, the step of setting graded thresholds for each risk index based on the safety risk index set and performing risk analysis includes: S901, setting corresponding low-risk thresholds, medium-risk thresholds, and high-risk thresholds based on the air leakage risk index, barotrauma risk index, apnea risk index, human-machine interaction risk index, oxygenation risk index, and insufficient ventilation risk index in the safety risk index set.

[0094] S902: When any risk index is greater than its corresponding medium-risk threshold but less than the high-risk threshold, if the blood oxygen-carbon dioxide consistency verification index is greater than the preset consistency threshold, a medium-level warning signal for the corresponding risk type is triggered, and the respiratory equipment is automatically adjusted based on the adjustment control parameter set; if the blood oxygen-carbon dioxide consistency verification index is less than or equal to the consistency threshold, the current risk event is downgraded, the abnormal event is recorded, and medical staff are prompted to check the sensor status, and the automatic adjustment operation is not performed.

[0095] S903: When any risk index is greater than or equal to its corresponding high-risk threshold, the highest level warning signal is triggered and mandatory safety protection operations are performed.

[0096] It should be noted that low-risk, medium-risk, and high-risk thresholds are set for each risk index; the thresholds at each level are determined based on the experience of clinical experts and statistical analysis of historical respiratory event data. The default threshold levels for each risk index are shown in the table below: ; It should be noted that the consistency threshold is a preset scalar value, ranging from 0 to 1, with a default value of 0.7. The blood oxygen-carbon dioxide consistency verification index represents the degree of agreement between the risk assessment results of respiratory signals and the risk assessment results of blood oxygen or carbon dioxide signals, and is output by the third output branch of the convolutional-recurrent hybrid neural network model. When the consistency verification index is higher than or equal to 0.7, it indicates that the signal assessment directions of the two dimensions are consistent, the risk assessment is reliable, and automatic adjustment is performed; when the consistency verification index is lower than 0.7, it indicates that there is a discrepancy between the signal assessments of the two dimensions, and the assessment may be due to sensor malfunction, signal interference, or special physiological conditions of children, in which case automatic adjustment is not performed.

[0097] Furthermore, the specific execution flow of the automatic adjustment operation is as follows: Based on the target gas delivery flow compensation amount in the adjustment control parameter set, an opening adjustment command is sent to the proportional solenoid valve to increase or decrease the valve opening to compensate for insufficient or excessive gas delivery caused by gas leakage; Based on the trigger threshold adjustment amount in the adjustment control parameter set, the current setting value of the respiratory trigger sensitivity threshold is updated to increase or decrease the trigger sensitivity to adapt to the current gas leakage state and avoid false triggering or trigger failure caused by gas leakage; Based on the base flow increment value in the adjustment control parameter set, the current setting value of the base flow rate is adjusted to increase or decrease the base flow to maintain a stable base flow in the pipeline and ensure the reliability of the trigger signal; Based on the oxygen concentration adjustment amount in the adjustment control parameter set, the oxygen mixing ratio is adjusted to increase or decrease the inhaled oxygen concentration to cope with the risk of decreased blood oxygen or oxygen toxicity; Based on the alarm delay adjustment amount in the adjustment control parameter set, the alarm response delay for the corresponding risk type is shortened so that medical staff can receive early warning information more promptly.

[0098] It should be noted that the highest level warning signal is the device's highest priority audible and visual alarm signal, which includes a continuous high-frequency buzzer and a flashing red alarm light. At the same time, the specific high-risk type and the corresponding risk index value are displayed in red on the device's display screen.

[0099] Furthermore, the mandatory safety protection operation includes the following four simultaneously executed sub-operations: The first sub-operation is to automatically switch the breathing mode to backup ventilation mode, which is a pressure control mode, to ensure ventilation at a preset safe pressure value in case the volume control mode may experience insufficient air delivery due to leakage; the second sub-operation is to forcibly reduce the upper limit of airway pressure to a preset safe value, which is 80% of the initial pressure upper limit in the initial ventilation parameter set, reducing the risk of barotrauma; the third sub-operation is to switch the low blood oxygen saturation alarm threshold and the high carbon dioxide partial pressure alarm threshold to stricter emergency thresholds, where the low blood oxygen saturation emergency threshold is increased by 2 percentage points from the original low limit alarm threshold, and the high carbon dioxide partial pressure emergency threshold is decreased by 5 mmHg from the original high limit alarm threshold, to enhance the sensitivity of monitoring the child's oxygenation and ventilation status; the fourth sub-operation is to generate an emergency event record containing high-risk type, risk index value, trigger timestamp, current equipment parameters, and the child's basic information, and send it to the monitoring terminal via a wired or wireless communication interface.

[0100] The display terminal generates a complete historical dataset of each respiratory cycle and visualizes it.

[0101] In this embodiment of the invention, generating and visualizing the complete historical dataset of each respiratory cycle includes: Using each respiratory cycle as the recording unit, the system synchronously collects the enhanced time-series feature map of the current respiratory cycle, the risk index values ​​in the safety risk index set, the regulation parameter values ​​in the regulation control parameter set, the blood oxygen-carbon dioxide consistency verification index value, and the types and timestamps of early warning events triggered within the current respiratory cycle. This constitutes a single-cycle respiratory event record, which in turn generates a historical dataset of the entire respiratory process, thereby generating a multidimensional waveform trend chart.

[0102] It should be noted that a single-cycle respiratory event record is a structured data recording unit containing the following fields: the start and end timestamps of the respiratory cycle, the twelve-dimensional feature vector corresponding to the respiratory cycle in the enhanced time-series feature map (including inhaled tidal volume, exhaled tidal volume, respiratory rate, peak pressure, mean pressure, leakage, waveform rise slope, mean blood oxygen saturation, blood oxygen saturation drop, end-expiratory carbon dioxide partial pressure, carbon dioxide waveform rise slope, and carbon dioxide waveform plateau characteristic value), six risk index values ​​in the safety risk index set (leakage risk index, barotrauma risk index, apnea risk index, human-ventilator asynchrony risk index, oxygenation risk index, and hypoventilation risk index), six regulation parameter values ​​in the regulation and control parameter set (target delivery flow compensation, trigger threshold adjustment, pressure upper limit dynamic constraint value, baseline flow increment value, oxygen concentration adjustment, and alarm delay adjustment), the blood oxygen-carbon dioxide consistency verification index value, the type of warning event triggered in the current respiratory cycle (no event, low-risk event, medium-risk event, and high-risk event), and the precise timestamp when the warning event was triggered. Append-to-memory operations to the device’s local memory are implemented using a circular queue data structure.

[0103] It should be noted that the multidimensional waveform trend chart is a visualization of specified data fields from the historical dataset of the entire respiratory process, presented graphically on the device display. The multidimensional waveform trend chart contains several independent but time-axis-aligned waveform sub-charts, including: a tidal volume trend sub-chart (horizontal axis: time; vertical axis: tidal volume value), showing the changes in both inhaled and exhaled tidal volume; an airway pressure trend sub-chart (horizontal axis: time; vertical axis: airway pressure value), showing the changes in peak and average pressure; a leakage quantum chart (horizontal axis: time; vertical axis: leakage value), showing the changes in leakage volume; a blood oxygen saturation trend sub-chart (horizontal axis: time; vertical axis: percentage of blood oxygen saturation), showing the changes in blood oxygen saturation; a carbon dioxide partial pressure trend sub-chart (horizontal axis: time; vertical axis: carbon dioxide partial pressure value), showing the changes in end-expiratory carbon dioxide partial pressure; and a blood oxygen-carbon dioxide consistency verification index value change chart.

[0104] Furthermore, the historical dataset of the entire respiratory process can be exported to an external storage device or uploaded to a cloud server via wired or wireless communication interfaces. The export format is a standardized medical data exchange format, which includes the child's anonymity identifier, device identifier, device usage start and end time, complete historical respiratory event records, and device parameter change logs.

[0105] In the several embodiments provided by this invention, it should be understood that the disclosed system can be implemented in other ways.

[0106] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0107] The embodiments of this invention can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, system, and technology that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0108] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A safety airflow warning and automatic adjustment system for pediatric breathing equipment, characterized in that, The system includes: The device self-test module performs a self-test on the respiratory device and the monitoring link based on the hardware status data of the respiratory device and the connection status data of the additional monitoring device. After the self-test passes, it generates a device ready flag and calculates the initial ventilation parameters based on the child's basic physiological parameters. The enhanced time-series feature map construction module calculates multidimensional respiratory-monitoring fusion feature vectors for each respiratory cycle based on respiratory signal data, blood oxygen signal data, and carbon dioxide signal data, and constructs an enhanced time-series feature map. The index generation module, based on the enhanced time-series feature map, analyzes and derives a set of safety risk indices, a set of regulation and control parameters, and a blood oxygen-carbon dioxide consistency verification index. The risk analysis module, based on a set of safety risk indices, sets tiered thresholds for each risk index and performs risk analysis. The display terminal generates a complete historical dataset of each respiratory cycle and visualizes it.

2. The child respiratory device safety airflow warning and automatic adjustment system as described in claim 1, characterized in that, The hardware status data of the respiratory equipment includes the zero-point drift value of the flow sensor, the opening response curve of the proportional solenoid valve, and the reference value of the pressure sensor; the connection status data of the additional monitoring equipment includes the connection status of the pulse oximeter probe, the signal quality index, the preheating status of the carbon dioxide analyzer, and the patency status of the sampling tubing; the basic physiological parameter data of the child includes age, weight, and preset breathing mode.

3. The child respiratory device safety airflow warning and automatic adjustment system as described in claim 1, characterized in that, The process involves self-checking the respiratory equipment and monitoring links, generating a device readiness flag upon successful self-check, and calculating initial ventilation parameters based on the child's baseline physiological parameters, including: S301 performs a device self-test on the respiratory device based on hardware status data and generates a respiratory device self-test result; performs a link self-test on the monitoring link based on the connection status data of the additional monitoring device and generates a monitoring link self-test result. S302: When both the respiratory equipment self-test result and the monitoring link self-test result are in a pass state, generate an equipment ready flag; when either the respiratory equipment self-test result or the monitoring link self-test result is in a fail state, perform a fault handling operation. S303 analyzes and obtains an initial ventilation parameter set based on the child's basic physiological parameter data; and sets alarm thresholds for low blood oxygen saturation and high carbon dioxide partial pressure.

4. The child breathing device safety airflow warning and automatic adjustment system as described in claim 3, characterized in that, The initial ventilation parameter set, derived from the analysis of the child's basic physiological parameter data, includes: S401, the formula for calculating the initial target tidal volume is: ; in The initial target tidal volume is expressed in milliliters. The child's weight is expressed in kilograms. The tidal volume coefficient per unit body weight ranges from 4 to 6, with the unit being milliliters per kilogram. The default value is determined based on the preset breathing mode, with 6 in volume control mode and 4 in pressure control mode. S402, the formula for calculating the initial upper limit of respiratory rate is: ; in This is the initial upper limit of respiratory rate, expressed in breaths per minute. The baseline respiratory rate is based on age, measured in breaths per minute, and is obtained by looking up the age classification table. This is the frequency margin, measured in times per minute, with a default value of 10. S403, the formula for calculating the initial pressure upper limit is: ; in This represents the initial pressure limit, expressed in centimeters of water column. The upper limit of baseline pressure is based on age, and the unit is centimeters of water column, obtained by looking up the age classification table. This is the pressure margin, expressed in centimeters of water column, with a default value of 2. S404, the formula for calculating the initial trigger sensitivity threshold is: ; in This is the initial trigger sensitivity threshold, in milliliters; The trigger sensitivity benchmark is set per unit weight, in milliliters per kilogram, with a default value of 0.5; W represents the child's weight in kilograms.

5. The child breathing device safety airflow warning and automatic adjustment system as described in claim 1, characterized in that, The method involves calculating a multidimensional respiratory-monitoring fusion feature vector for each respiratory cycle based on respiratory signal data, blood oxygen signal data, and carbon dioxide signal data, and constructing an enhanced time-series feature map, including: S501, based on the preprocessed respiratory signal data, calculate the respiratory feature sub-vectors for each respiratory cycle; and based on the preprocessed blood oxygen signal data and carbon dioxide signal data, calculate the blood oxygen-carbon dioxide feature sub-vectors for each respiratory cycle. S502, the respiratory feature vector and the blood oxygen-carbon dioxide feature vector are concatenated and fused to obtain a multidimensional respiratory-monitoring fusion feature vector; and the multidimensional respiratory-monitoring fusion feature vectors of several consecutive respiratory cycles are arranged in chronological order to construct an enhanced temporal feature map.

6. The child respiratory device safety airflow warning and automatic adjustment system as described in claim 5, characterized in that, The respiratory feature subvectors include inhaled tidal volume, exhaled tidal volume, respiratory rate, peak pressure, mean pressure, leakage, and waveform rise slope; the blood oxygen-carbon dioxide feature subvectors include mean blood oxygen saturation, blood oxygen saturation decrease, end-expiratory carbon dioxide partial pressure, carbon dioxide waveform rise slope, and carbon dioxide waveform plateau characteristic value.

7. The child breathing device safety airflow warning and automatic adjustment system as described in claim 5, characterized in that, The calculation of respiratory feature sub-vectors for each respiratory cycle includes: The formula for calculating the leakage rate of S701 is as follows: ; in This refers to the amount of air leakage, expressed in milliliters. Tidal volume, measured in milliliters, represents the total volume of air inhaled during one respiratory cycle. Tidal volume, measured in milliliters, represents the total volume of exhaled air during one respiratory cycle. S702, the formula for calculating the rising slope of the waveform is as follows: ; in The slope of the waveform is expressed in centimeters of water column per second. Peak pressure, expressed in centimeters of water column, represents the maximum airway pressure during one respiratory cycle; Inspiratory initiation pressure, expressed in centimeters of water column, represents the airway pressure at the start of a respiratory cycle. The inhalation rise time, in seconds, represents the time it takes to reach peak pressure from the start of inhalation. S703, Inhaled tidal volume is the integral of the instantaneous inhaled flow rate over the duration during the respiratory cycle; Exhaled tidal volume is the absolute value of the integral of the instantaneous exhaled flow rate over the duration during the respiratory cycle; Respiratory rate is the value of 60 seconds divided by the duration of the respiratory cycle; Peak pressure is the maximum airway pressure during the respiratory cycle; Mean pressure is the arithmetic mean of the airway pressure during the respiratory cycle.

8. The child respiratory device safety airflow warning and automatic adjustment system as described in claim 1, characterized in that, The enhanced time-series feature map analysis yields a set of safety risk indices, a set of regulatory control parameters, and a blood oxygen-carbon dioxide consistency verification index, including: The enhanced temporal feature map is input into a pre-trained convolutional-recurrent hybrid neural network model to obtain a fused attention feature vector. Then, a nonlinear mapping is performed on the fused attention feature vector to obtain a set of safety risk indices, a set of adjustment control parameters, and a blood oxygen-carbon dioxide consistency verification index.

9. The child breathing device safety airflow warning and automatic adjustment system as described in claim 1, characterized in that, The process involves setting tiered thresholds for each risk index based on a set of security risk indices and conducting risk analysis, including: S901 sets corresponding low-risk, medium-risk, and high-risk thresholds based on the air leakage risk index, barotrauma risk index, apnea risk index, human-machine interaction risk index, oxygenation risk index, and insufficient ventilation risk index of the safety risk index set. S902: When any risk index is greater than its corresponding medium-risk threshold but less than the high-risk threshold, if the blood oxygen-carbon dioxide consistency verification index is greater than the preset consistency threshold, a medium-level warning signal for the corresponding risk type is triggered, and the respiratory equipment is automatically adjusted based on the adjustment control parameter set; if the blood oxygen-carbon dioxide consistency verification index is less than or equal to the consistency threshold, the current risk event is downgraded, the abnormal event is recorded, and medical staff are prompted to check the sensor status, and the automatic adjustment operation is not performed. S903: When any risk index is greater than or equal to its corresponding high-risk threshold, the highest level warning signal is triggered and mandatory safety protection operations are performed.

10. The child breathing device safety airflow warning and automatic adjustment system as described in claim 1, characterized in that, The generation and visualization of the complete historical dataset for each respiratory cycle includes: Using each respiratory cycle as the recording unit, the system synchronously collects the enhanced time-series feature map of the current respiratory cycle, the risk index values ​​in the safety risk index set, the regulation parameter values ​​in the regulation control parameter set, the blood oxygen-carbon dioxide consistency verification index value, and the types and timestamps of early warning events triggered within the current respiratory cycle. This constitutes a single-cycle respiratory event record, which in turn generates a historical dataset of the entire respiratory process, thereby generating a multidimensional waveform trend chart.

Images