CN-122025065-A - Automatic flow control detection method and system for medical breathing machine

CN122025065ACN 122025065 ACN122025065 ACN 122025065ACN-122025065-A

Abstract

The invention discloses a medical breathing machine automatic flow control detection method and system, wherein the method synchronously collects breathing machine operation parameters through a multi-sensor network, fuses the breathing machine operation parameters into comprehensive state data after time alignment and normalization pretreatment, performs time sequence feature analysis based on an improved long-short-period memory network, enhances the capture of breathing cycle features by introducing cycle gates and utilizing the breathing cycle features, fuses static parameters of a patient to generate risk feature vectors, dynamically adjusts control parameters such as air suction pressure, breathing frequency, oxygen concentration and the like by adopting a particle swarm optimization algorithm with minimum risk as a target, comprehensively calculates time sequence risk, control deviation and real-time abnormal indexes by utilizing a weighted fusion strategy, executes self-adaptive control instructions according to risk grades, and realizes continuous optimization through a closed loop feedback mechanism. The corresponding system integrates data acquisition, time sequence analysis, decision control and man-machine interaction modules on the embedded platform, and supports real-time data processing.

Inventors

ZHONG YANING
PENG JINHONG
WU ZHANGLI

Assignees

与实检测技术服务(苏州)有限公司

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (8)

1. The automatic flow control detection method and system for the medical breathing machine are characterized by comprising the following steps: S1, collecting operation parameters of a breathing machine through a multi-sensor network, preprocessing the parameters and fusing the parameters to generate comprehensive state data; S2, based on the comprehensive state data, performing time sequence feature analysis through an improved long-short-term memory network, and fusing static parameters of a patient to generate a risk feature vector; S3, adopting an optimization algorithm, and dynamically adjusting control parameters of the breathing machine with the aim of minimizing risk, wherein the control parameters at least comprise the air suction pressure, the breathing frequency and the oxygen concentration; And S4, calculating a comprehensive risk value through a weighted fusion strategy based on the risk feature vector and the optimized control parameter, and executing a corresponding automatic control instruction according to the classified risk level.
2. The method for detecting the automatic flow control of the medical ventilator according to claim 1, wherein the step of generating the integrated status data in S1 includes: initializing a sensor network and parameter configuration, initializing a multi-mode sensor network consisting of an airflow sensor, a pressure sensor, an oxygen concentration sensor and an event sensor, and setting sampling frequency and a communication protocol; through synchronous acquisition of multi-mode original data, parallel triggering of sensor data acquisition, adding of time stamps to all data, and outputting of original data stream with time stamps Wherein For the air flow rate, As a value of the pressure, the pressure value, In order to obtain the blood oxygen saturation level, In the event data of the event data, For time window, linear interpolation is used based on air flow sensor data Aligning other sensor data to a unified time point to obtain a synchronous data set Wherein A 10ms equally spaced time sequence; For the synchronous data set Time series data in (a) Applying a min-max normalization process, the formula is Mapping values to Interval, at the same time, event data is encoded into binary vectors to obtain normalized data set ; The normalized time sequence data and the event data are fused through characteristic splicing to form a comprehensive state vector Outputting the comprehensive state data set 。
3. The automated ventilator flow control detection method of claim 1, wherein the modified long and short term memory network of S2 is configured by introducing a cycle gate To enhance capture of respiratory cycle characteristics, the cell state update formula is: Wherein, the In order to adjust the term(s) periodically, For respiratory cycle feature vectors extracted from the airflow data, 、、 A forget gate, an input gate and an output gate of a standard LSTM, In order to be a candidate cell state, Representing element-by-element multiplication.
4. The automated ventilator control detection method of claim 3, wherein the risk feature vector is generated in S2 The process of (1) comprises: Extracting the hidden state of the final time step of the long-period memory network As a time series feature vector Encoding static parameters of patient age and medical history into numerical vectors Splicing the time sequence feature vector and the static feature vector into a fusion vector ; Applying a linear transformation and a ReLU activation function to the fusion vector to generate a risk feature vector 。
5. The method for detecting the automatic flow control of the medical ventilator according to claim 1, wherein the optimization algorithm in S3 is a particle swarm optimization algorithm, and the process comprises: initializing a population of particles within a safe boundary of ventilator control parameters, the position of each particle representing a set of control parameter combinations; Evaluating particle quality based on a fitness function defined as Wherein To calculate the resulting risk score by a pre-trained risk prediction model based on the risk feature vector, In order to control the amplitude of the variation of the parameter, For trade-off coefficients, for balancing risk minimization with control stationarity; By iteratively updating particle velocity And position Searching for an optimal solution; Terminating iteration when the adaptability converges, the maximum iteration number is reached or the risk is lower than a preset threshold value, and outputting a global optimal position as an optimal control parameter vector 。
6. The method for detecting the automatic flow control of the medical ventilator according to claim 1, wherein the weighted fusion strategy in S4 comprises: Calculating time sequence risk index Control deviation index And real-time anomaly index And normalize each index to A section; Historical confidence based on metrics Dynamically assigning weights The weight satisfies the normalization condition ; Using a weighted average formula Calculating a comprehensive risk value ; Will be according to the preset threshold value Dividing into low, medium and high risk grades, and generating corresponding control instructions, wherein the low risk maintains the current parameters, and the risk is adjusted according to a fine tuning formula Smooth transition to optimal parameters, high risk immediate application And triggers an alarm.
7. The method for detecting automated ventilator process control according to claim 6, wherein S4 further comprises a closed loop execution and feedback process of automated control instructions: After the instruction is issued, the absolute deviation between the actual parameter and the target value is monitored ; According to the execution result, an exponential moving average formula is used Updating confidence levels of risk indicators, wherein To perform accuracy; if the risk level is continuously medium or high and the deviation is high And if the flow is larger than the set threshold, re-triggering the flows of the S3 and the S4 to form the self-adaptive closed-loop control.
8. A medical ventilator automated process control detection system for implementing the method of any of claims 1-7, comprising: The data acquisition module is used for acquiring the operation parameters of the breathing machine through the multi-mode sensor array; the time sequence analysis module is used for running the improved long-term and short-term memory network to generate a risk characteristic vector; The decision control module is used for executing the optimization algorithm and generating a control instruction; The man-machine interaction module is used for displaying the risk level and the alarm information; the system is integrated on the embedded platform and supports real-time data stream processing and control.

Description

Automatic flow control detection method and system for medical breathing machine Technical Field The invention relates to the technical field of intelligent control of medical equipment, in particular to a method and a system for detecting automatic flow control of a medical breathing machine. Background The breathing machine is used as important life support equipment, is widely applied to the fields of operation anesthesia, critical care and the like, has the core functions of maintaining oxygenation and ventilation of a patient through mechanical ventilation, and has the advantages that a traditional breathing machine parameter control is seriously dependent on manual adjustment by a clinician according to blood gas analysis results and physiological indexes, and the mode has several obvious defects that firstly, the manual adjustment has hysteresis and cannot respond to the dynamic change of the breathing state of the patient in real time, especially under the condition that the illness state such as acute respiratory distress syndrome and the like rapidly progresses, the adjustment delay possibly causes insufficient ventilation or increased risk of lung injury, secondly, the physiological characteristics and pathological states of different patients have obvious differences, the fixed parameter setting is difficult to realize personalized accurate treatment, and in addition, in the long-time ventilation process, medical staff need to continuously monitor a plurality of parameters, the workload is large, and the artificial error is easy to cause due to fatigue. In order to improve the automation level of the control of the breathing machine, some auxiliary control schemes exist in the prior art, for example, part of systems adopt control logic based on fixed rules or a classical PID controller is introduced to adjust single parameters, however, the methods still have obvious limitations that the rule type system has poor flexibility and cannot adapt to complex clinical scenes, the PID controller can realize continuous adjustment, but the PID controller has a plurality of independent variables, lacks comprehensive optimization on the coupling relation of multiple parameters, and has difficulty in processing inherent nonlinearity and time-varying characteristics of the breathing system, and in addition, the existing methods mostly depend on limited data sources, cannot fully utilize multi-mode sensor information to perform comprehensive risk assessment, and have defects in predictive control and prospective intervention. Therefore, there is a need for an intelligent ventilator system that can integrate multisource information, evaluate respiratory risk in real time, and automatically optimize control parameters to overcome the hysteresis, versatility, and limitations of the prior art, and promote the safety and effectiveness of ventilation therapy. Disclosure of Invention In order to overcome the defects in the prior art, the invention provides the following technical scheme for achieving the purposes: A medical breathing machine automatic flow control detection method and system comprises the following steps: S1, collecting operation parameters of a breathing machine through a multi-sensor network, preprocessing the parameters and fusing the parameters to generate comprehensive state data; S2, based on the comprehensive state data, performing time sequence feature analysis through an improved long-short-term memory network, and fusing static parameters of a patient to generate a risk feature vector; S3, adopting an optimization algorithm, and dynamically adjusting control parameters of the breathing machine with the aim of minimizing risk, wherein the control parameters at least comprise the air suction pressure, the breathing frequency and the oxygen concentration; And S4, calculating a comprehensive risk value through a weighted fusion strategy based on the risk feature vector and the optimized control parameter, and executing a corresponding automatic control instruction according to the classified risk level. Further, the process of generating the integrated status data in S1 includes: initializing a sensor network and parameter configuration, initializing a multi-mode sensor network consisting of an airflow sensor, a pressure sensor, an oxygen concentration sensor and an event sensor, and setting sampling frequency and a communication protocol; through synchronous acquisition of multi-mode original data, parallel triggering of sensor data acquisition, adding of time stamps to all data, and outputting of original data stream with time stamps WhereinFor the air flow rate,As a value of the pressure, the pressure value,In order to obtain the blood oxygen saturation level,In the event data of the event data,For time window, linear interpolation is used based on air flow sensor dataAligning other sensor data to a unified time point to obtain a synchronous data setWhereinA 10ms equally spaced time sequence