WO2023060478A1 - 一种人机不同步识别方法、系统、终端以及存储介质 - Google Patents

一种人机不同步识别方法、系统、终端以及存储介质 Download PDF

Info

Publication number
WO2023060478A1
WO2023060478A1 PCT/CN2021/123558 CN2021123558W WO2023060478A1 WO 2023060478 A1 WO2023060478 A1 WO 2023060478A1 CN 2021123558 W CN2021123558 W CN 2021123558W WO 2023060478 A1 WO2023060478 A1 WO 2023060478A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
respiratory
human
breathing
machine
Prior art date
Application number
PCT/CN2021/123558
Other languages
English (en)
French (fr)
Inventor
仲为
李慧慧
熊富海
颜延
王磊
马良
Original Assignee
中国科学院深圳先进技术研究院
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Priority to PCT/CN2021/123558 priority Critical patent/WO2023060478A1/zh
Publication of WO2023060478A1 publication Critical patent/WO2023060478A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61M—DEVICES FOR INTRODUCING MEDIA INTO, OR ONTO, THE BODY; DEVICES FOR TRANSDUCING BODY MEDIA OR FOR TAKING MEDIA FROM THE BODY; DEVICES FOR PRODUCING OR ENDING SLEEP OR STUPOR
    • A61M16/00—Devices for influencing the respiratory system of patients by gas treatment, e.g. mouth-to-mouth respiration; Tracheal tubes

Definitions

  • the present application belongs to the technical field of physiological data analysis, and in particular relates to a method, system, terminal and storage medium for identifying human-computer asynchrony.
  • the ventilator As an effective means of artificially replacing the spontaneous ventilation function, has been widely used in respiratory failure caused by various reasons, anesthesia respiratory management during major surgery, respiratory support treatment and emergency resuscitation. It occupies a very important position in the field of modern medicine. Ventilator is a vital medical equipment that can prevent and treat respiratory failure, reduce complications, save and prolong the life of patients.
  • human-machine asynchrony caused by abnormal ventilation often occurs, such as invalid inspiratory effort, double-trigger inhalation, trigger delay, short cycle, long cycle, reverse trigger etc. Human-machine out-of-synchronization will cause many negative impacts on patients, usually relying on doctors to make judgments and adjust ventilator parameters in a timely manner, which is relatively inefficient.
  • machine learning methods With the development of machine learning, applying machine learning methods to automatically detect and classify human-machine asynchronous phenomena can greatly improve the detection efficiency.
  • machine learning methods need to input a large amount of raw data into a multi-layer neural network for machine learning.
  • the algorithm process is generally very complicated, and the amount of calculation is very large, so it is impossible to perform instant and effective waveform classification.
  • the present application provides a method, system, terminal and storage medium for identifying human-computer out-of-synchronization, aiming to solve one of the above-mentioned technical problems in the prior art at least to a certain extent.
  • a human-machine asynchronous identification method comprising:
  • the division of the respiration data into at least two pieces of data with the same number of data points includes:
  • the input of the feature value into the trained respiratory waveform classification model includes:
  • the machine learning classification algorithm includes a support vector machine algorithm, a nearest neighbor node algorithm or a simple shell Yeesian algorithm.
  • the technical solution adopted in the embodiment of the present application also includes: the acquisition of the data set used for training the model is specifically:
  • the continuous respiration waveform signals include simulated respiration signals or real respiration signals;
  • the continuous respiratory signal is divided into respiratory data with multiple respiratory cycles, and the respiratory data of each respiratory cycle are classified and marked according to the waveform characteristics , store the tags of all breathing data in a tag list in turn;
  • a one-to-one correspondence between the breathing type and the eigenvalues in the label list and the eigenvalue list is used to generate a data set for model training.
  • the breathing data is airway pressure data
  • the breathing type includes normal breathing or abnormal breathing including double-triggered inspiratory or ineffective inspiratory effort.
  • the identification of man-machine asynchronous phenomena according to the breathing type classification results includes:
  • the breathing type classification result is normal breathing, it is determined that there is no human-machine asynchronous phenomenon
  • the output type of respiration is abnormal respiration, it is determined that there is a human-computer out-of-synchronization phenomenon, and a human-machine out-of-synchronization prompt message is issued.
  • a human-machine asynchronous recognition system including:
  • Data acquisition module used to acquire respiratory data in the current respiratory cycle
  • Eigenvalue calculation module used to divide the breathing data into at least two pieces of data with the same number of data points, and calculate the variance of each piece of data respectively, and use the variance calculation result of each piece of data as the breathing data eigenvalues;
  • Waveform classification module used for inputting the feature value into the trained respiratory waveform classification model, classifying the respiratory type of the respiratory data through the respiratory waveform classification model, and man-machine asynchrony according to the respiratory type classification result phenomenon is identified.
  • a terminal includes a processor and a memory coupled to the processor, wherein,
  • the memory stores program instructions for realizing the human-machine asynchronous identification method
  • the processor is configured to execute the program instructions stored in the memory to control human-machine out-of-synchronization recognition.
  • a storage medium storing program instructions executable by a processor, and the program instructions are used to execute the human-machine asynchronous identification method.
  • the beneficial effect produced by the embodiments of the present application lies in that the human-machine asynchronous recognition method, system, terminal and storage medium of the embodiments of the present application segment the breathing data of each breathing cycle and calculate the score Segment variance, the segment variance is used as the feature value of the corresponding respiratory cycle, and the feature value is classified into the breathing type through the machine learning classification algorithm, and the man-machine asynchronous phenomenon is identified according to the classification result.
  • the embodiment of the present application is easy to operate, and can accurately identify the man-machine asynchronous phenomenon in real time, thereby assisting the medical staff to monitor and accelerate the judgment of the man-machine asynchronous phenomenon, which greatly improves the feasibility in practical application.
  • Fig. 1 is the flow chart of the human-computer asynchronous recognition method of the embodiment of the present application
  • Fig. 2 is a normal respiratory airway pressure waveform diagram of a complete respiratory cycle
  • Fig. 3 is a double-trigger respiratory airway pressure waveform diagram of a complete respiratory cycle
  • Fig. 4 is a waveform diagram of airway pressure of invalid inspiratory effort for a complete breathing cycle
  • FIG. 5 is a schematic structural diagram of a human-machine asynchronous recognition system according to an embodiment of the present application.
  • FIG. 6 is a schematic structural diagram of a terminal according to an embodiment of the present application.
  • FIG. 7 is a schematic structural diagram of a storage medium according to an embodiment of the present application.
  • FIG. 1 is a flow chart of a method for identifying human-computer asynchrony according to an embodiment of the present application.
  • the human-machine asynchronous identification method in the embodiment of the present application includes the following steps:
  • the collected continuous respiration waveform signal includes a simulated respiration signal or a real respiration signal of the patient.
  • the collection of simulated breathing signals of patients with acute respiratory distress syndrome (ARDS) is taken as an example. It can be understood that the present application is also applicable to human-machine asynchronous recognition in other ventilator application scenarios.
  • the acquisition method of the continuous breathing waveform signal is as follows: the breathing mode of the simulated lung is set to , the ventilation mode of the ventilator is set to CPAP/PSV mode, the sampling frequency is 50HZ, and the breathing rate is 21 times per minute.
  • Test simulated lung can output 15 channels of respiratory data such as flow, airway pressure, tidal volume, alveolar pressure, pleural cavity pressure, and cardiac pressure. Since the waveform characteristics of airway pressure data are relatively obvious, it is easy to distinguish different types of waveforms. And the values are all positive, so in the following embodiments of the present invention, the airway pressure data is used as an example to train the classification model. Specifically as shown in Figures 2 to 4, Figure 2 is a normal respiratory airway pressure waveform for a complete respiratory cycle, Figure 3 is a dual-trigger respiratory airway pressure waveform for a complete respiratory cycle, and Figure 4 is a complete respiratory cycle Ineffective inspiratory effort airway pressure waveform.
  • the ordinate in the figure is the value of the airway pressure (unit: cmH 2 O), and the abscissa is the serial number of the period data point. It can be seen from the figure that the waveform characteristics of the airway pressure data are relatively obvious, and it is easy to distinguish the waveforms of different breathing types. It can be understood that the present invention is also applicable to human-computer out-of-synchronization recognition of tidal volume and other respiratory data.
  • S20 Divide the continuous respiration signal into respiration data with multiple respiration cycles, classify and label the respiration data of each respiration cycle according to the waveform characteristics, and store the tags of all respiration data in a tag list in turn;
  • the division principle of the respiration data is: divide the respiration data between every two troughs as a complete respiration cycle.
  • the number of data points of each section of respiratory data can be different, but the data point difference needs to be within the set threshold range (the threshold range is set to 1 to 10 in the embodiment of the present application), as a preferred
  • the number of data points of each segment of breathing data is set to be between 80 and 90. Since the number of data points of each segment of breathing data can be different, the feasibility in practical application can be greatly improved.
  • Breathing types include normal breathing and abnormal breathing such as trigger delay, short cycle, long cycle, reverse triggering, double-trigger inhalation, and invalid inspiratory effort. Take the three types of invalid inspiratory effort as an example, and the number of breath data of the three types is 1000 respectively. Store the three types of breathing data in three different folders, and mark the corresponding breathing types on the folders, for example: mark the breathing data of normal breathing as 0, and mark the breathing data of double-triggered breathing is 1, and marks the breath data of invalid inspiratory effort as 2.
  • the division method of respiratory data is: if the number of data points of a segment of respiratory data is odd, the data of the first segment is rounded down, and the data of the latter segment Round up; then calculate the variance of the data in the front section and the data in the back section respectively, and record the variance calculation results as: before Var, after Var, and record the eigenvalues as [before Var, after Var]. It can be understood that calculating the standard deviation or mean value of the data in the previous period and the data in the latter period can also be used as an alternative to the variance.
  • the machine learning classification algorithm includes, but is not limited to, a support vector machine algorithm, a nearest neighbor node algorithm, a naive Bayesian algorithm, and the like.
  • S60 Use the trained respiratory waveform classification model to perform waveform classification on the patient's respiratory data to obtain the patient's respiratory type, and identify human-machine asynchronous phenomena according to the respiratory type;
  • the automatic sampling point rules are set in advance according to different ventilator models.
  • the original respiratory signal of the patient is obtained, and the airway pressure data in the original respiratory signal is extracted immediately at the end of each breathing process, and the segmental variance of the airway pressure data is calculated as the breath's Eigenvalue, input the eigenvalue into the trained respiratory waveform classification model, and output the respiratory type classification result of this breath through the respiratory waveform classification model. If the output breathing type is normal breathing, it is determined that there is no man-machine asynchronous phenomenon.
  • the output breathing type is abnormal breathing such as double-trigger breathing or invalid inspiratory effort
  • a prompt message will be issued to remind the doctor to adjust the ventilator parameters in time, or configure the ventilator in advance , so that the ventilator can automatically adjust the parameters of the ventilator when it receives the prompt message of man-machine out-of-sync, so as to realize the real-time detection and classification of man-machine out-of-sync phenomena.
  • the recognition effect of the present application is verified through experiments.
  • the data of 1000 normal breathing cycles, 1000 double-trigger breathing cycles and 1000 invalid inspiratory effort cycles were selected.
  • the eigenvalue extraction operation proposed in this application use the support vector machine algorithm, the nearest neighbor node algorithm, the logistic regression algorithm, the decision tree algorithm, the naive Bayesian algorithm, and the random forest algorithm to test the classification effect.
  • the experimental results show that the support vector
  • the classification accuracy rate of computer algorithm, nearest neighbor node algorithm and logistic regression algorithm is as high as 100.00%, and all 990 classification tasks are classified correctly.
  • the decision tree algorithm is 99.90% accurate.
  • the Naive Bayes algorithm is 99.49% accurate.
  • the random forest algorithm has an accuracy rate of 98.48%.
  • the human-machine out-of-synchronization recognition method in the embodiment of the present application segments the respiratory data of each respiratory cycle, and calculates the segment variance, uses the segment variance as the feature value of the corresponding respiratory cycle, and classifies it through machine learning
  • the algorithm classifies the breathing type on the feature value, and identifies the man-machine asynchronous phenomenon according to the classification result.
  • the embodiment of the present application is easy to operate, and can accurately identify the man-machine asynchronous phenomenon in real time, thereby assisting the medical staff to monitor and accelerate the judgment of the man-machine asynchronous phenomenon, which greatly improves the feasibility in practical application.
  • the embodiment of the present application has universal applicability to the identification and analysis of various types of man-machine asynchrony, and the present invention can also be extended to the detection and analysis of one-dimensional signals such as electrocardiographic signals.
  • FIG. 5 is a schematic structural diagram of a human-machine asynchronous recognition system according to an embodiment of the present application.
  • the man-machine asynchronous recognition system 40 of the embodiment of the present application includes:
  • Data acquisition module 41 used to acquire respiratory data in the current respiratory cycle
  • Eigenvalue calculation module 42 used to divide the breathing data into at least two pieces of data with the same number of data points, and calculate the variance of each piece of data respectively, and use the variance calculation result of each piece of data as the eigenvalue of the breathing data;
  • Waveform classification module 43 used to input the characteristic value into the trained respiratory waveform classification model, classify the respiratory type of the respiratory data through the respiratory waveform classification model, and identify the man-machine asynchronous phenomenon according to the respiratory type classification result; wherein, if If the output breathing type is normal breathing, it is determined that there is no man-machine out-of-sync phenomenon.
  • the output breathing type is abnormal breathing such as double-trigger breathing or invalid inspiratory effort
  • a prompt message will be issued to remind the doctor to adjust the ventilator parameters in time, or configure the ventilator in advance , so that the ventilator can automatically adjust the parameters of the ventilator when it receives the prompt message of man-machine out-of-sync, so as to realize the real-time detection and classification of man-machine out-of-sync phenomena.
  • the man-machine out-of-synchronization identification system in the embodiment of the present application segments the breathing data of each breathing cycle, and calculates the segment variance, uses the segment variance as the feature value of the corresponding breathing cycle, and classifies the features through the machine learning classification algorithm. Classify the breathing type according to the value, and identify the man-machine asynchronous phenomenon according to the classification result.
  • the embodiment of the present application is easy to operate, and can accurately identify the man-machine asynchronous phenomenon in real time, thereby assisting the medical staff to monitor and accelerate the judgment of the man-machine asynchronous phenomenon, which greatly improves the feasibility in practical application.
  • FIG. 6 is a schematic diagram of a terminal structure according to an embodiment of the present application.
  • the terminal 50 includes a processor 51 and a memory 52 coupled to the processor 51 .
  • the memory 52 stores program instructions for realizing the above-mentioned human-machine out-of-synchronization identification method.
  • the processor 51 is used to execute the program instructions stored in the memory 52 to control human-machine asynchronous recognition.
  • the processor 51 may also be referred to as a CPU (Central Processing Unit, central processing unit).
  • the processor 51 may be an integrated circuit chip with signal processing capability.
  • the processor 51 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • DSP digital signal processor
  • ASIC application-specific integrated circuit
  • FPGA off-the-shelf programmable gate array
  • a general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.
  • FIG. 7 is a schematic structural diagram of a storage medium according to an embodiment of the present application.
  • the storage medium of the embodiment of the present application stores a program file 61 capable of realizing all the above-mentioned methods, wherein the program file 61 can be stored in the above-mentioned storage medium in the form of a software product, and includes several instructions to make a computer device (which can It is a personal computer, a server, or a network device, etc.) or a processor (processor) that executes all or part of the steps of the methods in various embodiments of the present invention.
  • a computer device which can It is a personal computer, a server, or a network device, etc.
  • processor processor
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disc, etc., which can store program codes. media, or terminal devices such as computers, servers, mobile phones, and tablets.

Landscapes

  • Health & Medical Sciences (AREA)
  • Emergency Medicine (AREA)
  • Pulmonology (AREA)
  • Engineering & Computer Science (AREA)
  • Anesthesiology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Hematology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

本申请涉及一种人机不同步识别方法、系统、终端以及存储介质。包括：获取当前呼吸周期内的呼吸数据；将所述呼吸数据划分为具有相同数据点个数的至少两段数据，并分别计算每段数据的方差，将所述每段数据的方差计算结果作为所述呼吸数据的特征值；将所述特征值输入训练好的呼吸波形分类模型，通过所述呼吸波形分类模型对所述呼吸数据的呼吸类型进行分类，根据所述呼吸类型分类结果对人机不同步现象进行识别。本申请实施例通过对每个呼吸周期的呼吸数据进行分段，并计算分段方差，将分段方差作为对应呼吸周期的特征值，并通过机器学习分类算法进行呼吸类型的分类，根据分类结果对人机不同步现象进行识别，可以实时准确的识别出人机不同步现象。

Description

一种人机不同步识别方法、系统、终端以及存储介质 技术领域
本申请属于生理数据分析技术领域，特别涉及一种人机不同步识别方法、系统、终端以及存储介质。
背景技术
在现代临床医学中，呼吸机作为一项能人工替代自主通气功能的有效手段，已普遍应用于各种原因所致的呼吸衰竭、大手术期间的麻醉呼吸管理、呼吸支持治疗和急救复苏中，在现代医学领域内占有十分重要的位置。呼吸机是一种能够起到预防和治疗呼吸衰竭，减少并发症，挽救及延长病人生命的至关重要的医疗设备。然而，在用呼吸机给患者进行机械通气的过程中，往往会出现通气异常导致的人机不同步现象，例如无效吸气努力、双触发吸气、触发延迟、短循环、长循环、反向触发等。人机不同步现象会对患者造成很多负面的影响，通常需要依靠医生进行判断，并及时调整呼吸机参数，效率较为低下。
随着机器学习的发展，应用机器学习方法对人机不同步现象进行自动检测和分类可以大大提高检测效率。机器学习方法在数据处理以及特征提取方面需要输入大量的原始数据进入多层的神经网络进行机器学习，算法流程普遍非常复杂，运算量非常大，无法进行即时有效的波形分类。
发明内容
本申请提供了一种人机不同步识别方法、系统、终端以及存储介质，旨在至少在一定程度上解决现有技术中的上述技术问题之一。
为了解决上述问题，本申请提供了如下技术方案：
一种人机不同步识别方法，包括：
获取当前呼吸周期内的呼吸数据；
将所述呼吸数据划分为具有相同数据点个数的至少两段数据，并分别计算 每段数据的方差，将所述每段数据的方差计算结果作为所述呼吸数据的特征值；
将所述特征值输入训练好的呼吸波形分类模型，通过所述呼吸波形分类模型对所述呼吸数据的呼吸类型进行分类，根据所述呼吸类型分类结果对人机不同步现象进行识别。
本申请实施例采取的技术方案还包括：所述将所述呼吸数据划分为具有相同数据点个数的至少两段数据包括：
如果所述呼吸数据的数据点个数为奇数，则按照向下或向上取整的方式对所述呼吸数据进行划分。
本申请实施例采取的技术方案还包括：所述将所述特征值输入训练好的呼吸波形分类模型包括：
获取用于训练模型的数据集，采用机器学习分类算法对所述数据集进行训练，得到训练好的呼吸波形分类模型；所述机器学习分类算法包括支持向量机算法、最邻近节点算法或朴素贝叶斯算法。
本申请实施例采取的技术方案还包括：所述获取用于训练模型的数据集具体为：
采集连续呼吸波形信号；所述连续呼吸波形信号包括模拟呼吸信号或真实呼吸信号；
按照每两个波谷间的呼吸数据为一个完整呼吸周期的分割原则，将所述连续呼吸信号划分为具有多个呼吸周期的呼吸数据，并根据波形特征对各个呼吸周期的呼吸数据进行分类及标注，依次将所有呼吸数据的标签存入一个标签列表中；
分别将每个呼吸周期的呼吸数据划分为具有相同数据点个数的至少两段数据，分别计算每段数据的方差，将方差计算结果作为对应呼吸周期的特征值，并依次将所有呼吸周期的特征值存入一个特征值列表中；
将所述标签列表与特征值列表中的呼吸类型与特征值一一对应，生成用于模型训练的数据集。
本申请实施例采取的技术方案还包括：
所述呼吸数据为气道压数据；
所述呼吸类型包括正常呼吸或异常呼吸，所述异常呼吸包括双触发吸气或无效吸气努力。
本申请实施例采取的技术方案还包括：所述根据所述呼吸类型分类结果对人机不同步现象进行识别包括：
如果呼吸类型分类结果为正常呼吸，则判定不存在人机不同步现象；
如果输出的呼吸类型为异常呼吸，则判定存在人机不同步现象，并发出人机不同步提示信息。
本申请实施例采取的另一技术方案为：一种人机不同步识别系统，包括：
数据获取模块：用于获取当前呼吸周期内的呼吸数据；
特征值计算模块：用于将所述呼吸数据划分为具有相同数据点个数的至少两段数据，并分别计算每段数据的方差，将所述每段数据的方差计算结果作为所述呼吸数据的特征值；
波形分类模块：用于将所述特征值输入训练好的呼吸波形分类模型，通过所述呼吸波形分类模型对所述呼吸数据的呼吸类型进行分类，根据所述呼吸类型分类结果对人机不同步现象进行识别。
本申请实施例采取的又一技术方案为：一种终端，所述终端包括处理器、与所述处理器耦接的存储器，其中，
所述存储器存储有用于实现所述人机不同步识别方法的程序指令；
所述处理器用于执行所述存储器存储的所述程序指令以控制人机不同步识别。
本申请实施例采取的又一技术方案为：一种存储介质，存储有处理器可运行的程序指令，所述程序指令用于执行所述人机不同步识别方法。
相对于现有技术，本申请实施例产生的有益效果在于：本申请实施例的人机不同步识别方法、系统、终端以及存储介质通过对每个呼吸周期的呼吸数据进行分段，并计算分段方差，将分段方差作为对应呼吸周期的特征值，并通过机器学习分类算法对特征值进行呼吸类型的分类，根据分类结果对人机不同步现象进行识别。本申请实施例操作简单，可以实时准确的识别出人机不同步现象，从而辅助医护人员对人机不同步现象进行监测与加速判断，大大提高了实际应用中的可行性。
附图说明
图1是本申请实施例的人机不同步识别方法的流程图；
图2为一个完整呼吸周期的正常呼吸气道压波形图；
图3为一个完整呼吸周期的双触发呼吸气道压波形图；
图4为一个完整呼吸周期的无效吸气努力气道压波形图；
图5为本申请实施例的人机不同步识别系统结构示意图；
图6为本申请实施例的终端结构示意图；
图7为本申请实施例的存储介质的结构示意图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本申请，并不用于限定本申请。
请参阅图1，是本申请实施例的人机不同步识别方法的流程图。本申请实施例的人机不同步识别方法包括以下步骤：
S10：采集连续呼吸波形信号；
本步骤中，采集的连续呼吸波形信号包括模拟呼吸信号或者患者的真实呼吸信号。本申请实施例中以采集急性呼吸窘迫综合症(Acute Respiratory Distress Syndrome,简称为ARDS)患者的模拟呼吸信号为例，可以理解，本申请同样适用于其他呼吸机应用场景的人机不同步识别。连续呼吸波形信号的采集方式具体为：将模拟肺的呼吸模式设定为，呼吸机通气模式设定为CPAP/PSV模式，采样频率为50HZ，呼吸频率为每分钟21次。Test模拟肺可输出流量、气道压、潮气量、肺泡压、胸膜腔内压、心脏压力等15个通道的呼吸数据，由于气道压数据的波形特征较为明显，易于区分不同类型的波形，且数值都为正，因此本发明以下实施例中以选用气道压数据为例进行分类模型的训练。具 体如图2至图4所示，图2为一个完整呼吸周期的正常呼吸气道压波形图，图3为一个完整呼吸周期的双触发呼吸气道压波形图，图4为一个完整呼吸周期的无效吸气努力气道压波形图。图中纵坐标为气道压的值(单位：cmH 2O)，横坐标为此周期数据点的序号。从图中可以看出，气道压数据的波形特征较为明显，易于区分不同呼吸类型的波形。可以理解，本发明同样适用于潮气量等其他呼吸数据的人机不同步识别。
S20：将连续呼吸信号分割为具有多个呼吸周期的呼吸数据，并根据波形特征对各个呼吸周期的呼吸数据进行分类及标注，依次将所有呼吸数据的标签存入一个标签列表中；
本步骤中，呼吸数据的分割原则为：将每两个波谷间的呼吸数据作为一个完整的呼吸周期进行分割。在进行呼吸周期的划分时，每段呼吸数据的数据点个数可以不同，但数据点差值需在设定阈值范围(本申请实施例设定该阈值范围为1～10)内，作为优选，本申请实施例设定每段呼吸数据的数据点个数分别在80到90之间。由于每段呼吸数据的数据点的个数可以不同，可以大大提高实际应用中的可行性。在完成数据划分后，将每个呼吸周期的呼吸数据单独存为表格文件。呼吸类型包括正常呼吸以及触发延迟、短循环、长循环、反向触发、双触发吸气、无效吸气努力等异常呼吸，为便于说明，本申请实施例仅以正常呼吸、双触发吸气以及无效吸气努力三种类型为例，三种类型的呼吸数据的数量分别为1000个。将三种类型的呼吸数据分别存放在三个不同的文件夹中，并分别在文件夹上标注对应的呼吸类型，例如：将正常呼吸的呼吸数据标注为0，将双触发呼吸的呼吸数据标注为1，将无效吸气努力的呼吸数据标注为2。
S30：分别将每个呼吸周期的呼吸数据划分为具有相同数据点个数的至少 两段数据，并分别计算每段数据的方差，将方差计算结果作为对应呼吸周期的特征值，并将所有呼吸周期的特征值依次存入一个特征值列表中；
本步骤中，以将每个呼吸周期的呼吸数据划分为两段为例，呼吸数据划分方式为：如果某段呼吸数据的数据点个数为奇数，则前段数据向下取整，后段数据向上取整；然后分别计算前段数据和后段数据的方差，并将方差计算结果记为：Var前、Var后，将特征值记为[Var前、Var后]。可以理解，还可以通过计算前段数据和后段数据的标准差或均值作为方差的替代方案。
S40：将标签列表与特征值列表中的呼吸类型与特征值一一对应，生成用于模型训练的数据集，并将数据集分为训练集与测试集；
本步骤中，随机将数据集的67％划分为训练集，33％划分为测试集，具体划分比例可根据实际应用进行调整。
S50：采用机器学习分类算法对训练集与测试集进行训练与测试，得到训练好的呼吸波形分类模型；
本步骤中，机器学习分类算法包括但不限于支持向量机算法、最邻近节点算法、朴素贝叶斯算法等。
S60：通过训练好的呼吸波形分类模型对患者的呼吸数据进行波形分类，得到患者的呼吸类型，根据呼吸类型对人机不同步现象进行识别；
本步骤中，在进行呼吸类型分类前，根据不同的呼吸机型号提前设置自动采样点的规则。在进行呼吸类型分类时，获取患者的原始呼吸信号，并在每一次呼吸过程结束时即时提取出原始呼吸信号中的气道压数据，计算出气道压数据的分段方差，作为此次呼吸的特征值，将特征值输入训练好的呼吸波形分类模型，通过呼吸波形分类模型输出本次呼吸的呼吸类型分类结果。如果输出的呼吸类型为正常呼吸，则判定不存在人机不同步现象。如果输出的呼吸类型为 双触发呼吸或无效吸气努力等异常呼吸，则判定当前呼吸机存在人机不同步现象，并发出提示信息，提醒医生及时调整呼吸机参数，或预先对呼吸机进行配置，使呼吸机收到人机不同步提示信息时自行调整呼吸机参数，从而实现人机不同步现象的即时检测与分类。
为了证明本申请实施例的可行性和有效性，通过实验对本申请的识别效果进行验证。在实验中，共选取1000个正常呼吸周期的数据、1000个双触发呼吸周期的数据以及1000个无效吸气努力周期的数据。经过本申请提出的特征值提取操作后，分别使用支持向量机算法、最邻近节点算法、逻辑回归算法、决策树算法、朴素贝叶斯算法、随机森林算法检验分类效果，实验结果证明，支持向量机算法、最邻近节点算法、逻辑回归算法的分类准确率高达100.00％，990项分类任务全部分类正确。决策树算法的准确率为99.90％。朴素贝叶斯算法的准确率为99.49％。随机森林算法的准确率为98.48％。在得到分类准确率结果后，可择优选取并保存呼吸波形分类模型。
基于上述，本申请实施例的人机不同步识别方法通过对每个呼吸周期的呼吸数据进行分段，并计算分段方差，将分段方差作为对应呼吸周期的特征值，并通过机器学习分类算法对特征值进行呼吸类型的分类，根据分类结果对人机不同步现象进行识别。本申请实施例操作简单，可以实时准确的识别出人机不同步现象，从而辅助医护人员对人机不同步现象进行监测与加速判断，大大提高了实际应用中的可行性。本申请实施例对于各种类型的人机不同步识别和分析具有普遍适用性，本发明还可扩展至心电信号等一维信号的检测与分析。
请参阅图5，为本申请实施例的人机不同步识别系统结构示意图。本申请实施例的人机不同步识别系统40包括：
数据获取模块41：用于获取当前呼吸周期内的呼吸数据；
特征值计算模块42：用于将呼吸数据划分为具有相同数据点个数的至少两段数据，并分别计算每段数据的方差，将每段数据的方差计算结果作为呼吸数据的特征值；
波形分类模块43：用于将特征值输入训练好的呼吸波形分类模型，通过呼吸波形分类模型对呼吸数据的呼吸类型进行分类，根据呼吸类型分类结果对人机不同步现象进行识别；其中，如果输出的呼吸类型为正常呼吸，则判定不存在人机不同步现象。如果输出的呼吸类型为双触发呼吸或无效吸气努力等异常呼吸，则判定当前呼吸机存在人机不同步现象，并发出提示信息，提醒医生及时调整呼吸机参数，或预先对呼吸机进行配置，使呼吸机收到人机不同步提示信息时自行调整呼吸机参数，从而实现人机不同步现象的即时检测与分类。
本申请实施例的人机不同步识别系统通过对每个呼吸周期的呼吸数据进行分段，并计算分段方差，将分段方差作为对应呼吸周期的特征值，并通过机器学习分类算法对特征值进行呼吸类型的分类，根据分类结果对人机不同步现象进行识别。本申请实施例操作简单，可以实时准确的识别出人机不同步现象，从而辅助医护人员对人机不同步现象进行监测与加速判断，大大提高了实际应用中的可行性。
请参阅图6，为本申请实施例的终端结构示意图。该终端50包括处理器51、与处理器51耦接的存储器52。
存储器52存储有用于实现上述人机不同步识别方法的程序指令。
处理器51用于执行存储器52存储的程序指令以控制人机不同步识别。
其中，处理器51还可以称为CPU(Centra l Process i ng Un it，中央处理单元)。处理器51可能是一种集成电路芯片，具有信号的处理能力。处理器51还可以是通用处理器、数字信号处理器(DSP)、专用集成电路(AS I C)、现成 可编程门阵列(FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
请参阅图7，为本申请实施例的存储介质的结构示意图。本申请实施例的存储介质存储有能够实现上述所有方法的程序文件61，其中，该程序文件61可以以软件产品的形式存储在上述存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)或处理器(processor)执行本发明各个实施方式方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(ROM，Read-On l y Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质，或者是计算机、服务器、手机、平板等终端设备。
对所公开的实施例的上述说明，使本领域专业技术人员能够实现或使用本发明。对这些实施例的多种修改对本领域的专业技术人员来说将是显而易见的，本发明中所定义的一般原理可以在不脱离本发明的精神或范围的情况下，在其它实施例中实现。因此，本发明将不会被限制于本发明所示的这些实施例，而是要符合与本发明所公开的原理和新颖特点相一致的最宽的范围。

Claims (9)

  1. 一种人机不同步识别方法，其特征在于，包括：
    获取当前呼吸周期内的呼吸数据；
    将所述呼吸数据划分为具有相同数据点个数的至少两段数据，并分别计算每段数据的方差，将所述每段数据的方差计算结果作为所述呼吸数据的特征值；
    将所述特征值输入训练好的呼吸波形分类模型，通过所述呼吸波形分类模型对所述呼吸数据的呼吸类型进行分类，根据所述呼吸类型分类结果对人机不同步现象进行识别。
  2. 根据权利要求1所述的人机不同步识别方法，其特征在于，所述将所述呼吸数据划分为具有相同数据点个数的至少两段数据包括：
    如果所述呼吸数据的数据点个数为奇数，则按照向下或向上取整的方式对所述呼吸数据进行划分。
  3. 根据权利要求1或2所述的人机不同步识别方法，其特征在于，所述将所述特征值输入训练好的呼吸波形分类模型包括：
    获取用于训练模型的数据集，采用机器学习分类算法对所述数据集进行训练，得到训练好的呼吸波形分类模型；所述机器学习分类算法包括支持向量机算法、最邻近节点算法或朴素贝叶斯算法。
  4. 根据权利要求3所述的人机不同步识别方法，其特征在于，所述获取用于训练模型的数据集具体为：
    采集连续呼吸波形信号；所述连续呼吸波形信号包括模拟呼吸信号或真实呼吸信号；
    按照每两个波谷间的呼吸数据为一个完整呼吸周期的分割原则，将所述连续 呼吸信号划分为具有多个呼吸周期的呼吸数据，并根据波形特征对各个呼吸周期的呼吸数据进行分类及标注，依次将所有呼吸数据的标签存入一个标签列表中；
    分别将每个呼吸周期的呼吸数据划分为具有相同数据点个数的至少两段数据，分别计算每段数据的方差，将方差计算结果作为对应呼吸周期的特征值，并依次将所有呼吸周期的特征值存入一个特征值列表中；
    将所述标签列表与特征值列表中的呼吸类型与特征值一一对应，生成用于模型训练的数据集。
  5. 根据权利要求4所述的人机不同步识别方法，其特征在于，
    所述呼吸数据为气道压数据；
    所述呼吸类型包括正常呼吸或异常呼吸，所述异常呼吸包括双触发吸气或无效吸气努力。
  6. 根据权利要求5所述的人机不同步识别方法，其特征在于，所述根据所述呼吸类型分类结果对人机不同步现象进行识别包括：
    如果呼吸类型分类结果为正常呼吸，则判定不存在人机不同步现象；
    如果输出的呼吸类型为异常呼吸，则判定存在人机不同步现象，并发出人机不同步提示信息。
  7. 一种人机不同步识别系统，其特征在于，包括：
    数据获取模块：用于获取当前呼吸周期内的呼吸数据；
    特征值计算模块：用于将所述呼吸数据划分为具有相同数据点个数的至少两段数据，并分别计算每段数据的方差，将所述每段数据的方差计算结果作为所述呼吸数据的特征值；
    波形分类模块：用于将所述特征值输入训练好的呼吸波形分类模型，通过所述呼吸波形分类模型对所述呼吸数据的呼吸类型进行分类，根据所述呼吸类型分类结果对人机不同步现象进行识别。
  8. 一种终端，其特征在于，所述终端包括处理器、与所述处理器耦接的存储器，其中，
    所述存储器存储有用于实现权利要求1-6任一项所述的人机不同步识别方法的程序指令；
    所述处理器用于执行所述存储器存储的所述程序指令以控制人机不同步识别。
  9. 一种存储介质，其特征在于，存储有处理器可运行的程序指令，所述程序指令用于执行权利要求1至6任一项所述人机不同步识别方法。
PCT/CN2021/123558 2021-10-13 2021-10-13 一种人机不同步识别方法、系统、终端以及存储介质 WO2023060478A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/123558 WO2023060478A1 (zh) 2021-10-13 2021-10-13 一种人机不同步识别方法、系统、终端以及存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/123558 WO2023060478A1 (zh) 2021-10-13 2021-10-13 一种人机不同步识别方法、系统、终端以及存储介质

Publications (1)

Publication Number Publication Date
WO2023060478A1 true WO2023060478A1 (zh) 2023-04-20

Family

ID=85987913

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/123558 WO2023060478A1 (zh) 2021-10-13 2021-10-13 一种人机不同步识别方法、系统、终端以及存储介质

Country Status (1)

Country Link
WO (1) WO2023060478A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109893732A (zh) * 2019-02-28 2019-06-18 杭州智瑞思科技有限公司 一种基于循环神经网络的机械通气人机不同步检测方法
US20200261674A1 (en) * 2017-11-09 2020-08-20 Autonomous Healthcare, Inc. Clinical Decision Support System for Patient-Ventilator Asynchrony Detection and Management
CN112819093A (zh) * 2021-02-24 2021-05-18 浙江工业大学 基于小数据集与卷积神经网络的人机不同步识别方法
CN113313045A (zh) * 2021-06-11 2021-08-27 浙江工业大学 基于多任务学习和类激活图反馈的人机不同步识别方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200261674A1 (en) * 2017-11-09 2020-08-20 Autonomous Healthcare, Inc. Clinical Decision Support System for Patient-Ventilator Asynchrony Detection and Management
CN109893732A (zh) * 2019-02-28 2019-06-18 杭州智瑞思科技有限公司 一种基于循环神经网络的机械通气人机不同步检测方法
CN112819093A (zh) * 2021-02-24 2021-05-18 浙江工业大学 基于小数据集与卷积神经网络的人机不同步识别方法
CN113313045A (zh) * 2021-06-11 2021-08-27 浙江工业大学 基于多任务学习和类激活图反馈的人机不同步识别方法

Similar Documents

Publication Publication Date Title
WO2022267381A1 (zh) 一种呼吸机人机异步分类方法、系统、终端以及存储介质
CN105342563B (zh) 睡眠状况的检测
CN109893732B (zh) 一种基于循环神经网络的机械通气人机不同步检测方法
JP5639152B2 (ja) 非同期性の検出
CN107106028B (zh) 用于心肺睡眠阶段分类的系统和方法
CN112675393B (zh) 一种呼吸机撤机管理系统及方法
WO2022267382A1 (zh) 一种呼吸机人机异步分类方法、系统、终端以及存储介质
Gutta et al. Cardiorespiratory model-based data-driven approach for sleep apnea detection
WO2022242123A1 (zh) 机械通气人机异步检测方法、装置及计算机可读存储介质
Stanculescu et al. A Hierarchical Switching Linear Dynamical System Applied to the Detection of Sepsis in Neonatal Condition Monitoring.
De Chazal et al. Sleep apnoea classification using heart rate variability, ECG derived respiration and cardiopulmonary coupling parameters
WO2021208656A1 (zh) 睡眠风险预测方法、装置和终端设备
CN113941061B (zh) 一种人机不同步识别方法、系统、终端以及存储介质
WO2023060478A1 (zh) 一种人机不同步识别方法、系统、终端以及存储介质
CN110277158A (zh) 一种适合于各类呼吸机进行自动化适配并整理存储数据的方法
EP4031003A1 (en) System and method for determining sleep stages based on non-cardiac body signals
CN115414026A (zh) 一种基于流速波形的呼吸自动切分方法及其系统
Subramanian et al. Arousal detection in obstructive sleep apnea using physiology-driven features
Sisodia et al. Sleep order detection model using support vector machines and features extracted from brain ECG signals
Ozdemir et al. An early warning algorithm to predict obstructive sleep apnea (OSA) episodes
US20220323700A1 (en) Tracking respiratory mechanics of a patient
Charbonnier On-line extraction of successive temporal sequences from ICU high-frequency data for decision support information
EP3973851A1 (en) Obtaining respiratory related sounds from an audio recording
CN114288500B (zh) 一种基于模糊熵特征提取的人机通气异步检测模型及装置
CN115068760A (zh) 一种呼吸机-患者通气异步现象智能识别分类方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21960219

Country of ref document: EP

Kind code of ref document: A1