WO2023097780A1 - 机械通气过程中人机异步现象的分类方法和分类装置 - Google Patents

机械通气过程中人机异步现象的分类方法和分类装置 Download PDF

Info

Publication number
WO2023097780A1
WO2023097780A1 PCT/CN2021/138034 CN2021138034W WO2023097780A1 WO 2023097780 A1 WO2023097780 A1 WO 2023097780A1 CN 2021138034 W CN2021138034 W CN 2021138034W WO 2023097780 A1 WO2023097780 A1 WO 2023097780A1
Authority
WO
WIPO (PCT)
Prior art keywords
waveform data
mechanical ventilation
classification
man
machine
Prior art date
Application number
PCT/CN2021/138034
Other languages
English (en)
French (fr)
Inventor
马良
仲为
熊富海
廖天正
颜延
李慧慧
王磊
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
Publication of WO2023097780A1 publication Critical patent/WO2023097780A1/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
    • A61M16/0003—Accessories therefor, e.g. sensors, vibrators, negative pressure
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/08—Detecting, measuring or recording devices for evaluating the respiratory organs
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • 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
    • A61M16/021—Devices for influencing the respiratory system of patients by gas treatment, e.g. mouth-to-mouth respiration; Tracheal tubes operated by electrical means
    • A61M16/022—Control means therefor
    • 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
    • A61M16/021—Devices for influencing the respiratory system of patients by gas treatment, e.g. mouth-to-mouth respiration; Tracheal tubes operated by electrical means
    • A61M16/022—Control means therefor
    • A61M16/024—Control means therefor including calculation means, e.g. using a processor
    • 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
    • A61M16/021—Devices for influencing the respiratory system of patients by gas treatment, e.g. mouth-to-mouth respiration; Tracheal tubes operated by electrical means
    • A61M16/022—Control means therefor
    • A61M16/024—Control means therefor including calculation means, e.g. using a processor
    • A61M16/026—Control means therefor including calculation means, e.g. using a processor specially adapted for predicting, e.g. for determining an information representative of a flow limitation during a ventilation cycle by using a root square technique or a regression analysis
    • 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
    • A61M2230/00—Measuring parameters of the user
    • A61M2230/40—Respiratory characteristics

Definitions

  • the invention belongs to the technical field of electrophysiological detection and monitoring, and specifically relates to a classification method, a classification device, a computer-readable storage medium, and a computer device for human-machine asynchronous phenomena in the process of mechanical ventilation.
  • MV mechanical ventilation
  • PCV pressure control ventilation
  • VCV volume control ventilation
  • PSV pressure support ventilation
  • Figure 1 shows the typical waveform diagrams of the above four common asynchrony.
  • human-machine asynchrony often leads to increasing chances of ventilator-induced lung injury, prolonging hospital stay and even increasing the mortality rate of patients in the ICU. Therefore, the recognition and response of human-machine asynchrony continues to attract research interests from clinicians to related professionals.
  • the most commonly used clinical method is the observation method, that is, the bedside doctor observes the ventilation waveform on the screen of the ventilator, including the airway pressure time waveform, flow velocity time waveform, and tidal volume time waveform, to determine which asynchrony has occurred, and then according to Take corresponding actions according to your own professional training.
  • Rule-based algorithms that is, using clinical expertise and metadata in the mechanical ventilation process to formulate relevant detection rules, such as calculating the ratio of the tidal volume during the expiratory process to the tidal volume during the expiratory process, and the inspiratory process. ratio of inhalation time to exhalation time.
  • the rule-based method often needs to set the threshold according to the professional knowledge of clinical personnel, and the selection of the threshold can only be used as the decision-making basis for the current sample, and cannot be flexibly adapted to more scenarios.
  • the academic circles focus on the research of man-machine asynchrony on algorithms based on machine learning and deep learning, that is, a large amount of respiratory waveform data is divided according to the respiratory cycle, and then features are extracted from them, and the features are used as the input of the algorithm model.
  • Cao Rui et al. used fuzzy entropy to analyze EEG signals in the patent (CN201610289472.2) "A Method for Classifying Epilepsy EEG Signals Based on Fuzzy Entropy", and then extracted the corresponding features of EEG signals through feature selection. The fuzzy entropy under the electrode is used as the input feature, and the feature is finally used for classification. The invention solves the problem of binary classification.
  • Lu Yunfei and others first used wavelet scale transformation to transform the original respiratory waveform in the journal article "Application of wavelet multi-scale features to detect mechanical ventilation human-machine asynchrony". On this basis, various entropy features were used to extract nonlinear features. Using After the previous selection algorithm selects the best feature combination, it is used as the input of the support vector machine classification algorithm for classification. This invention still only classifies the man-machine asynchronous phenomenon of invalid inhalation effort, which belongs to the binary classification task.
  • the technical problem solved by the present invention is: how to realize multiple man-machine asynchronous types that can be classified under multiple ventilation modes at the same time.
  • a classification method for man-machine asynchrony during mechanical ventilation comprising:
  • the Poincaré graph features are input to a pre-trained classification model, and the classification model outputs the human-machine asynchronous type corresponding to the real-time respiratory waveform data.
  • the classification method also includes:
  • the training samples are used to train the classification model to be trained to obtain the trained classification model.
  • the Poincaré map feature extraction operation is performed on the waveform data of the three channels of pressure time waveform data, flow time waveform data and volume time waveform data in sequence to obtain the Poincaré map feature corresponding to the waveform data of each channel, so as to The training samples that make up the current epoch.
  • performing the Poincaré graph feature extraction operation on the waveform data of each channel includes the following steps:
  • the calculation formula of the short-term standard deviation SD1 is as follows:
  • X(i) represents the signal of the i-th sampling point in the current segment
  • N represents the length of the data in the current segment
  • the formula for calculating the long-term standard deviation is as follows:
  • X(i) represents the signal of the i-th sampling point of the current segment, Indicates the average value of the current segment signal, and N indicates the length of the current segment data.
  • the multiple different man-machine asynchronous types at least include double trigger type, invalid effort type, early switching type, and delayed switching type.
  • the present application also discloses a classification device for man-machine asynchrony during mechanical ventilation, the classification device includes:
  • a waveform acquisition unit configured to acquire real-time respiratory waveform data of the subject to be measured during mechanical ventilation
  • a feature extraction unit is used to extract the Poincaré graph feature of the real-time respiratory waveform data
  • the type prediction unit is used to predict and obtain the human-machine asynchronous type corresponding to the real-time respiratory waveform data according to the input Poincaré graph features.
  • the present application also discloses a computer-readable storage medium.
  • the computer-readable storage medium stores a classification program for human-computer asynchrony during mechanical ventilation, and the classification program for human-computer asynchrony during mechanical ventilation is processed by a processor. During execution, the above-mentioned classification method of man-machine asynchronous phenomena in the process of mechanical ventilation is realized.
  • the present application also discloses a computer device, which includes a computer-readable storage medium, a processor, and a program for classifying human-machine asynchrony during mechanical ventilation stored in the computer-readable storage medium.
  • a program for classifying human-machine asynchronous phenomena during ventilation is executed by the processor, the above-mentioned method for classifying human-computer asynchronous phenomena during mechanical ventilation is realized.
  • the invention discloses a classification method and a classification device for man-machine asynchronous phenomena in the process of mechanical ventilation. Compared with the existing method, it has the following technical effects:
  • This method uses the Poincaré diagram to extract features from the original waveform, which does not depend on other factors other than the waveform, but is only related to the shape of the waveform, and the extracted features can reflect information such as the shape change of the waveform;
  • This method proposes to use the built machine learning model directly for various man-machine asynchronous classification tasks under various mechanical ventilation modes, avoiding the process of repeated modeling.
  • Figure 1 shows four common asynchronous typical waveforms.
  • Fig. 2 is a flow chart of the classification method of man-machine asynchrony during mechanical ventilation according to Embodiment 1 of the present invention
  • FIG. 3 is a diagram of the training process of the classification model of the man-machine asynchronous phenomenon in the process of mechanical ventilation according to Embodiment 1 of the present invention
  • Fig. 4 is the schematic diagram of the extraction process of the Poincaré characteristic map of the respiratory waveform data of Embodiment 1 of the present invention.
  • FIG. 5 is a schematic structural diagram of a device for classifying human-machine asynchronous phenomena during mechanical ventilation according to Embodiment 2 of the present invention.
  • FIG. 6 is a schematic diagram of computer equipment according to Embodiment 4 of the present invention.
  • this application provides a classification method for man-machine asynchronous phenomena in the process of mechanical ventilation, which extracts the Poincaré graph features of the original respiratory waveform data corresponding to various types of man-machine asynchrony, and constitutes a training sample.
  • the learning model is trained to obtain a classification model that can classify a variety of man-machine asynchronous phenomena at the same time, and then use the classification model to predict the real-time respiratory waveform data during mechanical ventilation, and obtain the corresponding man-machine asynchronous type.
  • the first embodiment discloses a classification method for man-machine asynchrony during mechanical ventilation, including the following steps:
  • Step S11 Obtain real-time respiratory waveform data of the subject to be measured during mechanical ventilation
  • Step S12 extracting the Poincaré map features of the real-time respiratory waveform data
  • Step S13 Input the Poincaré graph features into the pre-trained classification model, and the classification model outputs the human-machine asynchronous type corresponding to the real-time respiratory waveform data.
  • the first embodiment discloses a classification method for man-machine asynchrony during mechanical ventilation, which also includes the following steps:
  • Step S21 Obtain historical respiratory waveform data during mechanical ventilation
  • Step S22 Perform waveform segmentation and data labeling on the historical respiratory waveform data to obtain multiple cycles of respiratory waveform data, wherein the multiple cycles of respiratory waveform data correspond to at least four different man-machine asynchronous types;
  • Step S23 sequentially extracting the Poincaré graph features of the respiratory waveform data of each cycle to form a training sample
  • Step S24 using the training samples to train the classification model to be trained to obtain a trained classification model.
  • the first embodiment discloses a classification method for man-machine asynchronous phenomena during mechanical ventilation, including model training (steps S21 to S24), and model prediction (steps S11 to S13).
  • model training steps S21 to S24
  • model prediction steps S11 to S13
  • step S21 first use TestLung (a simulated lung device) to simulate patients with respiratory diseases. By adjusting different parameters, patients with different respiratory rates, different respiratory intensities and different respiratory diseases can be simulated. Then use the ventilator and TestLung for mechanical ventilation, and adjust the ventilator ventilation mode configuration to obtain different man-machine asynchronous data under different ventilation modes.
  • the respiratory waveform data of TestLung can be exported by software.
  • the historical respiratory waveform data used in the first embodiment is the data of three channels: pressure-time waveform, flow-velocity-time waveform and volume-time waveform.
  • the waveform segmentation method for the historical respiratory waveform data is: segment the data obtained in step S21 according to the respiratory cycle.
  • the human-machine asynchronous types contained in the historical respiratory waveform data collected in the first embodiment include four common types: double trigger type (DT), invalid effort type (IE), early switching type (PC) and delayed switching type (DC).
  • DT double trigger type
  • IE invalid effort type
  • PC early switching type
  • DC delayed switching type
  • Asynchronous type for these four asynchronous types, after observation, formulate segmentation rules, and then write corresponding codes for segmentation. Some of the segmented data cannot distinguish its asynchronous type, and this type of data is usually discarded.
  • the segmented data is obtained It will be a respiratory cycle one by one, and the data of each respiratory cycle will be saved in the csv file in the order of appearance, and the file name will be named from 1 in the order of appearance.
  • step S22 the method of data labeling the historical respiratory waveform data is as follows: since the nature of the problem determines the need to use supervised learning to train the model, the segmented data also needs to be labeled, that is, for each cycle Waveform labeling.
  • the process of adding tags is completed by three relevant professionals who have been trained in human-machine asynchronous recognition, and the annotation of each tag will go through the process of one person marking and another person reviewing. If the annotation disagrees, the cycle data will be discarded.
  • the above-mentioned four types of asynchronous plus Other types are digitalized as 0-4, and the Other type is composed of any one that does not belong to DT, IE, PC, and DC.
  • a csv file is generated by the code.
  • the file consists of two columns, one column is the file name of each breathing cycle, and the other column corresponds to the label of the cycle.
  • step S23 the Poincaré graph features of the respiratory waveform data of each cycle are sequentially extracted, and the method of forming a training sample is as follows: the waveform data of the three channels of pressure-time waveform data, flow-time waveform data and volume-time waveform data Carry out the Poincaré graph feature extraction operation to obtain the Poincaré graph features corresponding to the waveform data of each channel to form the training samples of the current cycle.
  • step S23 the Poincaré graph feature extraction operation includes the following steps:
  • X(i) represents the signal of the i-th sampling point of the current segment, Indicates the average value of the current segment signal, and N indicates the length of the current segment.
  • step S24 the specific process of using the training samples to train the classification model to be trained is: using the existing machine learning framework to build various machine learning models, including models such as random forest, support vector machine and logistic regression. Divide 80% of the training samples obtained in step S23 into a training set and the remaining 20% into a test set, use the training set to train the model, use the test set to evaluate the model, and obtain the classification model with the best performance.
  • models such as random forest, support vector machine and logistic regression.
  • step S11 to step S13 is used to predict the man-machine asynchronous type corresponding to the real-time respiratory waveform data of the subject under mechanical ventilation.
  • the second embodiment also discloses a device for classifying man-machine asynchronous phenomena during mechanical ventilation.
  • the device for classifying includes a waveform acquisition unit 100 , a feature extraction unit 200 and a type prediction unit 300 .
  • the waveform acquisition unit 100 is used to acquire the real-time respiratory waveform data of the subject under mechanical ventilation;
  • the feature extraction unit 200 is used to extract the Poincare map features of the real-time respiratory waveform data;
  • the type prediction unit 300 is used to input According to the Poincaré graph feature prediction, the human-machine asynchronous type corresponding to the real-time respiratory waveform data is obtained.
  • the category prediction unit 300 is a pre-trained classification model, and the training process of the classification model can be obtained by referring to the method of step S21 to step S24 in the first embodiment, and will not be repeated here.
  • the third embodiment also discloses a computer-readable storage medium.
  • the computer-readable storage medium stores a classification program for man-machine asynchrony during mechanical ventilation, and the classification program for man-machine asynchrony during mechanical ventilation is When the processor executes, the method for classifying human-machine asynchronous phenomena in the process of mechanical ventilation according to the first embodiment is realized.
  • Embodiment 4 also discloses a computer device.
  • the terminal includes a processor 12 , an internal bus 13 , a network interface 14 , and a computer-readable storage medium 11 .
  • the processor 12 reads the corresponding computer program from the computer-readable storage medium and executes it, forming a request processing device on a logical level.
  • one or more embodiments of this specification do not exclude other implementations, such as logic devices or a combination of software and hardware, etc., that is to say, the execution subject of the following processing flow is not limited to each A logic unit, which can also be a hardware or logic device.
  • the computer-readable storage medium 11 stores a classification program for the human-machine asynchronous phenomenon in the mechanical ventilation process, and when the classification program for the human-machine asynchronous phenomenon in the mechanical ventilation process is executed by the processor, the mechanical ventilation process of the first embodiment is realized.
  • Computer-readable storage media includes both volatile and non-permanent, removable and non-removable media by any method or technology for storage of information.
  • Information may be computer readable instructions, data structures, modules of a program, or other data.
  • Examples of computer readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage , magnetic cassettes, disk storage, quantum memory, graphene-based storage media or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by computing devices.
  • PRAM phase change memory
  • SRAM static random access memory
  • DRAM dynamic random access memory
  • RAM random access memory
  • ROM read-only memory
  • EEPROM electrically erasable

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Veterinary Medicine (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Physics & Mathematics (AREA)
  • Pulmonology (AREA)
  • Artificial Intelligence (AREA)
  • Hematology (AREA)
  • Anesthesiology (AREA)
  • Emergency Medicine (AREA)
  • Biophysics (AREA)
  • Physiology (AREA)
  • Pathology (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • Evolutionary Computation (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

一种机械通气过程中人机异步现象的分类方法和分类装置。该分类方法包括：获取待测对象在机械通气过程中的实时呼吸波形数据（S11）；提取所述实时呼吸波形数据的庞加莱图特征（S12）；将所述庞加莱图特征输入至预先训练好的分类模型，分类模型输出所述实时呼吸波形数据对应的人机异步类型（S13）。本方法从原始波形中提取庞加莱图特征，不依赖于波形之外的其它因素，以更好地反映波形的形态变化等信息，从训练得到可用于多种机械通气模式下的多种人机异步分类任务的分类模型，最终实现对实时呼吸波形数据进行准确的分类。

Description

机械通气过程中人机异步现象的分类方法和分类装置 技术领域
本发明属于电生理检测监护技术领域，具体地讲，涉及一种机械通气过程中人机异步现象的分类方法、分类装置、计算机可读存储介质、计算机设备。
背景技术
呼吸机作为各大医院重症监护病房(intensive care unit,ICU)中的重要设备，其在生命支持系统中扮演着极其重要的角色。呼吸机与患有呼吸系统疾病患者的交互过程称为机械通气(mechanical ventilation,MV)，该过程主要有两种模式：1)控制模式，这种模式下往往患者会被注射镇静剂或麻醉剂，此时由呼吸机控制患者的呼吸节律，如压力控制通气(pressure control ventilation,PCV)和容量控制通气(volume control ventilation,VCV)；2)辅助模式，如压力支持通气(pressure support ventilation,PSV)等，这种模式下，患者存在一定程度的呼吸努力，呼吸机起到了辅助患者呼吸的作用，这样可以减小患者的呼吸做功。然而无论是哪种情况，二者之间的交互都不会那么顺利，均会出现一定程度上的不协调，这种不协调被普遍地称之为人机异步(patient-ventilator asynchrony,PVA)。
由于不同的患者拥有不同的病情，甚至是同一个患者在机械通气过程随着时间的推移其病情也会发生变化，因此在机械通气发展的几十年中，直到现在一共发现了多种类型的人机异步，如常见的双触发(double triggering,DT)、无效努力(ineffective effort,IE)、提前切换(premature cycling,PC)、延迟切换(delayed cycling,DC)及反向触发(reverse triggering,RT)等，图1展示了以上四种常见异步典型的波形图。随着医学界多年的观察研究，发现人机异步往往会带来不断增加地呼吸机诱发肺损伤发生几率、延长住院时间甚至导致患者ICU中死亡率增加。因此，人机异步的识别与应对不断吸引着从临床医生到相关专业人员的研究兴趣。
临床上最为常用的方法为观察法，即床旁医生观察呼吸机屏幕上的通气波形，包括气道压力时间波形、流速时间波形以及潮气量时间波形，来判断到底发生了哪种异步，然后根据自己的专业训练作出相应的应对动作。而也有研究 使用基于规则的算法，即结合使用临床专业知识以及机械通气过程中的元数据共同制定相关的检测规则，如计算呼气过程的潮气量与呼气过程中的潮气量之比，吸气时间与呼气时间之比等。
由于观察法往往是耗时的，且对护理人员的专业知识能力要求较高，需要受到一定训练的人才能快速且准确地识别当前所面对的呼吸周期是否包含异步波形。而基于规则的方法往往需要根据临床人员的专业知识设定阈值，且该阈值的选择仅仅只能作为当前样本的决策依据，无法灵活地适应更多的场景。
目前学术界将研究人机异步的焦点放在基于机器学习与深度学习的算法，即将大量的呼吸波形数据按照呼吸周期分割，然后从中提取特征，将特征作为算法模型的输入，利用现有的计算资源优势，通过有监督的学习，最终让模型学习到识别各种人机异步的能力。
以下是目前该领域相关研究的调查：
1、毛科栋等人在专利(CN201910149798.9)《一种基于循环神经网络的机械通气人机异步检测方法》中提到了使用循环神经网络的算法来检测人机异步，两个通道的GRU(门控循环单元)分别提取压力波形特征和流速时间波形特征，然后将两种特征融合后使用BGRU(双向门控循环神经单元)提取更高维的特征，最后使用softmax全连接层得到对人机异步类型的分类结果。在这篇专利中，他们的数据集标注由专业医生标注，共检测四大类人机异步：流速、触发、周期和其它。
2、李冰等人在专利(CN202011431650.3)《基于一维可解释卷积神经网络的人机异步识别方法》中提出将由医生标注的数据集经过预处理后输入到一维卷积神经网络中进行学习训练，得到一个基于神经网络的预测模型。在预测过程中，通过梯度加权类激活映射的方式，可以获得该模型分类决策的可视化解释。
3、潘清等人在专利(CN202110208054.7)《基于小数据集与卷积神经网络的人机异步识别方法》中通过把采集到的原始呼吸信号转换成二维图像，先使用公开的图像数据集ImageNet训练二维图像多分类的模型，之后以迁移学习的方式，将呼吸波形构成的二维图像输入到模型中并对最后一层的全连接层以上层的权重作微调，得到可用于呼吸波形分类的卷积神经网络。
4、葛慧青等人在专利(CN202010474275.4)《一种基于DBA-DTW-KNN 的机械通气人机异步快速识别方法》中实时读取呼吸波形数据构成测试序列，经过标准化之后计算测试序列与训练集里面的所有序列的DTW距离，而后用DTW计算相似性距离，再结合KNN的聚类思想，对测试序列进行分类。此发明用于判断人机异步现象中无效吸气努力。
5、曹锐等人在专利(CN201610289472.2)《一种基于模糊熵的癫痫脑电信号分类方法》中使用模糊熵对脑电信号分析，接着通过特征选择提取出反应脑电信号特征的对应电极下模糊熵作为输入特征，最后将特征用于分类，该发明解决的属于二分类问题。
6、陆云飞等人在期刊文章《应用小波多尺度特征检测机械通气人机异步》中首先采用小波尺度变换对原始呼吸波形作一次变换，在此基础上使用多种熵特征提取非线性特征，使用前项选择算法选择出最佳的特征组合后，将其作为支持向量机分类算法的输入进行分类，该发明仍仅分类无效吸气努力这一种人机异步现象，属于二分类任务。
研究发现，机器学习和深度学习二者在分类效果上性能相似，差别是机器学习算法的性能受到特征选择的影响，而且目前特征的选择多是统计特征及临床上的一些数据组合构成；而深度学习算法的模型建立相对前者更为复杂。重要的是，关于人机异步现象多种通气模式的多分类任务现有的研究基本没有。
发明内容
(一)本发明所要解决的技术问题
本发明解决的技术问题是：如何实现可同时分类多种通气模式下的多种人机异步类型。
(二)本发明所采用的技术方案
一种机械通气过程中人机异步现象的分类方法，所述分类方法包括：
获取待测对象在机械通气过程中的实时呼吸波形数据；
提取所述实时呼吸波形数据的庞加莱图特征；
将所述庞加莱图特征输入至预先训练好的分类模型，分类模型输出所述实时呼吸波形数据对应的人机异步类型。
优选地，所述分类方法还包括：
获取机械通气过程中的历史呼吸波形数据；
对所述历史呼吸波形数据进行波形分割和数据标注，获得多个周期的呼吸波形数据，其中多个周期的呼吸波形数据对应多种不同人机异步类型；
依次提取每个周期的呼吸波形数据的庞加莱图特征，构成训练样本；
利用训练样本对待训练的分类模型进行训练，获得训练好的分类模型。
优选地，依次对压力时间波形数据、流量时间波形数据和容量时间波形数据三个通道的波形数据进行庞加莱图特征提取操作，获得每个通道的波形数据对应的庞加莱图特征，以构成当前周期的训练样本。
优选地，对每个通道的波形数据进行庞加莱图特征提取操作包括如下步骤：
将当前通道的波形数据划分为n段数据，n≥2；
依次计算每一段数据的短时标准差、长时标准差以及短时标准差与长时标准差的比值，获得n*3维的特征数据，作为当前通道的波形数据对应的庞加莱图特征。
优选地，所述短时标准差SD1的计算公式如下：
Figure PCTCN2021138034-appb-000001
Figure PCTCN2021138034-appb-000002
其中，X(i)表示当前段第i个采样点的信号，N表示当前段数据的长度。
优选地，所述长时标准差的计算公式如下：
Figure PCTCN2021138034-appb-000003
Figure PCTCN2021138034-appb-000004
其中，X(i)表示当前段第i个采样点的信号，
Figure PCTCN2021138034-appb-000005
表示当前段信号的均值，N表示当前段数据的长度。
优选地，多种不同人机异步类型至少包括双触发类型、无效努力类型、提前切换类型、延迟切换类型。
本申请还公开了一种机械通气过程中人机异步现象的分类装置，所述分类装置包括：
波形获取单元，用于获取待测对象在机械通气过程中的实时呼吸波形数据；
特征提取单元，用于提取所述实时呼吸波形数据的庞加莱图特征；
类型预测单元，用于根据输入的所述庞加莱图特征预测得到所述实时呼吸波形数据对应的人机异步类型。
本申请还公开了一种计算机可读存储介质，所述计算机可读存储介质存储有机械通气过程中人机异步现象的分类程序，所述机械通气过程中人机异步现象的分类程序被处理器执行时实现上述的机械通气过程中人机异步现象的分类方法。
本申请还公开了一种计算机设备，所述计算机设备包括计算机可读存储介质、处理器和存储在所述计算机可读存储介质中的机械通气过程中人机异步现象的分类程序，所述机械通气过程中人机异步现象的分类程序被处理器执行时实现上述的机械通气过程中人机异步现象的分类方法。
(三)有益效果
本发明公开了一种机械通气过程中人机异步现象的分类方法和分类装置，相对于现有方法，具有如下技术效果：
1、本方法使用庞加莱图从原始波形中提取特征，不依赖于波形之外的其它因素，只与波形的形态有关，提取的特征更可以反映波形的形态变化等信息；
2、使用本方法搭建的机器学习模型相较于深度学习模型更为简单，更易于部署；
3、本方法提出将搭建好的机器学习模型直接用于多种机械通气模式下的多种人机异步分类任务，避免了反复建模的过程。
附图说明
图1为四种常见异步典型的波形图。
图2为本发明的实施例一的机械通气过程中人机异步现象的分类方法的流程图；
图3为本发明的实施例一的机械通气过程中人机异步现象的分类模型的训练过程图；
图4为本发明的实施例一的呼吸波形数据的庞加莱特征图的提取过程示意图；
图5为本发明的实施例二的机械通气过程中人机异步现象的分类装置的结构示意图；
图6为本发明的实施例四的计算机设备示意图。
具体实施方式
为了使本发明的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本发明进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。
在详细描述本申请的各个实施例之前，首先简单描述本申请的发明构思：现有的判断人机异步现象的方法包括观察法和基于机器学习的方法，前者一般仅适用于当前个体，不具有适用性，且高度依赖于医生的经验，后者依赖于所提取的特征，目前所采用的特征多是统计特征及临床数据组合，训练得到的模型准确率不高，且多数适用于二分类任务，未实现多分类任务。为此，本申请提供了一种机械通气过程中人机异步现象的分类方法，提取多种类型人机异步对应的原始呼吸波形数据的庞加莱图特征，构成训练样本，对现有的机器学习模型进行训练，得到可同时对多种人机异步现象进行分类的分类模型，接着利用该分类模型对机械通气过程中的实时呼吸波形数据进行预测，获得对应的人机异步类型。
如图2所示，本实施例一公开了机械通气过程中人机异步现象的分类方法包括如下步骤：
步骤S11：获取待测对象在机械通气过程中的实时呼吸波形数据；
步骤S12：提取所述实时呼吸波形数据的庞加莱图特征；
步骤S13：将所述庞加莱图特征输入至预先训练好的分类模型，分类模型输出所述实时呼吸波形数据对应的人机异步类型。
进一步地，如图3所示，本实施例一公开了机械通气过程中人机异步现象的分类方法还包括如下步骤：
步骤S21：获取机械通气过程中的历史呼吸波形数据；
步骤S22：对所述历史呼吸波形数据进行波形分割和数据标注，获得多个周期的呼吸波形数据，其中多个周期的呼吸波形数据至少对应四种不同人机异步类型；
步骤S23：依次提取每个周期的呼吸波形数据的庞加莱图特征，构成训练样本；
步骤S24：利用训练样本对待训练的分类模型进行训练，获得训练好的分类模型。
具体来说，本实施例一公开了机械通气过程中人机异步现象的分类方法包括模型训练即步骤S21至步骤S24，以及模型预测即步骤S11至步骤S13两部分，下面着重描述模型训练部分。
在步骤S21中，首先使用TestLung(一种模拟肺设备)模拟患有呼吸疾病的病人，通过调节不同的参数，可以模拟不同呼吸频率、不同呼吸强度且不同呼吸疾病的患者。然后使用呼吸机与TestLung进行机械通气，调节呼吸机通气模式配置便可得到不同通气模式下不同的人机异步数据。TestLung的呼吸波形数据可以通过软件导出。本实施例一使用的历史呼吸波形数据为压力时间波形、流速时间波形以及容量时间波形三种通道的数据。
在步骤S22中对历史呼吸波形数据进行波形分割的方法为：将步骤S21中得到的数据按照呼吸周期分割。本实施例一所采集的历史呼吸波形数据中包含的人机异步类型有双触发类型(DT)、无效努力类型(IE)、提前切换类型(PC)和延迟切换类型(DC)四种常见的异步类型，对于这四种异步类型分别通过观察后制定分割规则，然后编写相应的代码进行分割，分割后的数据有部分是无法辨别其异步类型的，这类数据通常会舍弃，最终分割得到的将是一个一个的呼吸周期，每个呼吸周期的数据按照出现的顺序分别保存到csv文件中，文件名则按出现顺序由1开始命名。
在步骤S22中对历史呼吸波形数据进行数据标注的方法为：由于问题的性质决定了需要使用有监督学习的方式来训练模型，因此对于分割后的数据还需要进行标注，即为每个周期的波形添加标签。添加标签的过程由三位经过人机异步识别训练的相关专业人员完成，并且每个标签的标注均会经过一个人标注另外一个人审查的过程，如果标注意见不一致则会舍弃该周期数据。其中，对上述提到的四种异步加上Other类型共五种类型的标签数字化为0-4，其中Other类型是由不属于DT、IE、PC和DC中的任何一种构成。
标注过程，首先将每个周期的数据使用代码读出，其次将三个通道的数据按照压力、流量和容量的顺序画在一幅图中，然后由标注人员观察该图形的形态，并判断该周期属于哪一类异步。当标注完成后，由代码生成一个csv文件，文件由两列构成，一列为每个呼吸周期的文件名称，另一列对应为该周期的标签。
在步骤S23中，依次提取每个周期的呼吸波形数据的庞加莱图特征，构成训练样本的方法为：依次对压力时间波形数据、流量时间波形数据和容量时间波形数据三个通道的波形数据进行庞加莱图特征提取操作，获得每个通道的波形数据对应的庞加莱图特征，以构成当前周期的训练样本。
具体来讲，不直接使用原始波形作为模型的输入有两个原因，一个是原始数据的维度过大，直接输入模型会增加训练难度；二是由于通常每个数据周期的长度也不一致，对于模型来说一般是不允许的。因此对原始数据提取关键特征来降低数据维度，而且最终可以得到一致的维度作为模型的输入。我们利用庞加莱图算法对标注好的每个呼吸周期数据提取特征，其中分别从每个通道中提取3个特征然后扁平化作为该呼吸周期三个通常波形的整体特征，最后这些特征将作为机器学习模型的训练样本。
在步骤S23中，如4所示，庞加莱图特征提取操作包括如下步骤：
将当前通道的波形数据划分为n段数据，n≥2；
依次计算每一段数据的短时标准差、长时标准差以及短时标准差与长时标准差的比值，获得n*3维的特征数据，作为当前通道的波形数据对应的庞加莱图特征。最终获得当前周期的三个通道的波形数据对应的庞加莱图特征，即n*3*3维的特征数据，作为当前周期的训练样本。
其中，短时标准差SD1的计算公式如下：
Figure PCTCN2021138034-appb-000006
Figure PCTCN2021138034-appb-000007
长时标准差的计算公式如下：
Figure PCTCN2021138034-appb-000008
Figure PCTCN2021138034-appb-000009
其中，X(i)表示当前段第i个采样点的信号，
Figure PCTCN2021138034-appb-000010
表示当前段信号的均值，N表示当前段的长度。
短时标准差与长时标准差的比值Ratio＝SD1/SD2。
在步骤S24中，利用训练样本对待训练的分类模型进行训练的具体过程为：利用现有的机器学习框架搭建多种机器学习模型，包括随机森林、支持向量机和逻辑回归等模型。将步骤S23得到的训练样本的80％划分成训练集和其余20％划分为测试集，使用训练集对模型进行训练，使用测试集对模型进行评估，获得性能最优的分类模型。
经过实验验证，经过逻辑回归模型(Logistic Regression)、支持向量机模型(SVM)、随机森林模型(Random Forest)、Voting模型和XGBoost模型的训练和测试，结果表明对于一般的模型均能达到80％以上的准确率，对于绝大多数模型而言，该方案所提取的特征可使人机异步的分类总体准确率95％以上，结果如表1所示。
表1各模型测试集结果
序号 模型 Accuracy Cohen kappa
1 Logistic regression 0.8294 0.7868
2 Random forest 0.9538 0.9422
3 SVM 0.9076 0.8845
4 Voting 0.9562 0.9452
5 XGBoost 0.9550 0.9437
进一步地，在利用步骤S21至步骤S24的方法训练得到分类模型之后，采用步骤S11至步骤S13的方法预测得到待测对象在机械通气过程中的实时呼吸波形数据对应的人机异步类型。
如图5所示，本实施例二还公开了一种机械通气过程中人机异步现象的分类装置，该分类装置包括波形获取单元100、特征提取单元200和类型预测单元300。该波形获取单元100用于获取待测对象在机械通气过程中的实时呼吸波形数据；特征提取单元200用于提取所述实时呼吸波形数据的庞加莱图特征；类型预测单元300用于根据输入的所述庞加莱图特征预测得到所述实时呼吸波形数据对应的人机异步类型。
其中，类型预测单元300为预先训练好的分类模型，分类模型的训练过程参考实施例一中的步骤S21至步骤S24的方法训练得到，在此不进行赘述。
本实施例三还公开了一种计算机可读存储介质，所述计算机可读存储介质存储有机械通气过程中人机异步现象的分类程序，所述机械通气过程中人机异 步现象的分类程序被处理器执行时实现实施例一的机械通气过程中人机异步现象的分类方法。
本实施例四还公开了一种计算机设备，在硬件层面，如图6所示，该终端包括处理器12、内部总线13、网络接口14、计算机可读存储介质11。处理器12从计算机可读存储介质中读取对应的计算机程序然后运行，在逻辑层面上形成请求处理装置。当然，除了软件实现方式之外，本说明书一个或多个实施例并不排除其他实现方式，比如逻辑器件抑或软硬件结合的方式等等，也就是说以下处理流程的执行主体并不限定于各个逻辑单元，也可以是硬件或逻辑器件。所述计算机可读存储介质11上存储有机械通气过程中人机异步现象的分类程序，所述机械通气过程中人机异步现象的分类程序被处理器执行时实现实施例一的机械通气过程中人机异步现象的分类方法。
计算机可读存储介质包括永久性和非永久性、可移动和非可移动媒体可以由任何方法或技术来实现信息存储。信息可以是计算机可读指令、数据结构、程序的模块或其他数据。计算机可读存储介质的例子包括，但不限于相变内存(PRAM)、静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、其他类型的随机存取存储器(RAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、快闪记忆体或其他内存技术、只读光盘只读存储器(CD-ROM)、数字多功能光盘(DVD)或其他光学存储、磁盒式磁带、磁盘存储、量子存储器、基于石墨烯的存储介质或其他磁性存储设备或任何其他非传输介质，可用于存储可以被计算设备访问的信息。
上面对本发明的具体实施方式进行了详细描述，虽然已表示和描述了一些实施例，但本领域技术人员应该理解，在不脱离由权利要求及其等同物限定其范围的本发明的原理和精神的情况下，可以对这些实施例进行修改和完善，这些修改和完善也应在本发明的保护范围内。

Claims (10)

  1. 一种机械通气过程中人机异步现象的分类方法，其特征在于，所述分类方法包括：
    获取待测对象在机械通气过程中的实时呼吸波形数据；
    提取所述实时呼吸波形数据的庞加莱图特征；
    将所述庞加莱图特征输入至预先训练好的分类模型，分类模型输出所述实时呼吸波形数据对应的人机异步类型。
  2. 根据权利要求1所述的机械通气过程中人机异步现象的分类方法，其特征在于，所述分类方法还包括：
    获取机械通气过程中的历史呼吸波形数据；
    对所述历史呼吸波形数据进行波形分割和数据标注，获得多个周期的呼吸波形数据，其中多个周期的呼吸波形数据对应多种不同人机异步类型；
    依次提取每个周期的呼吸波形数据的庞加莱图特征，构成训练样本；
    利用训练样本对待训练的分类模型进行训练，获得训练好的分类模型。
  3. 根据权利要求2所述的机械通气过程中人机异步现象的分类方法，其特征在于，提取每个周期的呼吸波形数据的庞加莱图特征的方法包括：
    获取每个周期的压力时间波形数据、流量时间波形数据和容量时间波形数据；
    依次对压力时间波形数据、流量时间波形数据和容量时间波形数据三个通道的波形数据进行庞加莱图特征提取操作，获得每个通道的波形数据对应的庞加莱图特征，以构成当前周期的训练样本。
  4. 根据权利要求3所述的机械通气过程中人机异步现象的分类方法，其特征在于，对每个通道的波形数据进行庞加莱图特征提取操作包括如下步骤：
    将当前通道的波形数据划分为n段数据，n≥2；
    依次计算每一段数据的短时标准差、长时标准差以及短时标准差与长时标准差的比值，获得n*3维的特征数据，作为当前通道的波形数据对应的庞加莱图特征。
  5. 根据权利要求3所述的机械通气过程中人机异步现象的分类方法，其特征在于，所述短时标准差SD1的计算公式如下：
    Figure PCTCN2021138034-appb-100001
    Figure PCTCN2021138034-appb-100002
    其中，X(i)表示当前段第i个采样点的信号，N表示当前段数据的长度。
  6. 根据权利要求3所述的机械通气过程中人机异步现象的分类方法，其特征在于，所述长时标准差的计算公式如下：
    Figure PCTCN2021138034-appb-100003
    Figure PCTCN2021138034-appb-100004
    其中，X(i)表示当前段第i个采样点的信号，
    Figure PCTCN2021138034-appb-100005
    表示当前段信号的均值，N表示当前段数据的长度。
  7. 根据权利要求4所述的机械通气过程中人机异步现象的分类方法，其特征在于，多种不同人机异步类型至少包括双触发类型、无效努力类型、提前切换类型、延迟切换类型。
  8. 一种机械通气过程中人机异步现象的分类装置，其特征在于，所述分类装置包括：
    波形获取单元，用于获取待测对象在机械通气过程中的实时呼吸波形数据；
    特征提取单元，用于提取所述实时呼吸波形数据的庞加莱图特征；
    类型预测单元，用于根据输入的所述庞加莱图特征预测得到所述实时呼吸波形数据对应的人机异步类型。
  9. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质存储有机械通气过程中人机异步现象的分类程序，所述机械通气过程中人机异步现象的分类程序被处理器执行时实现权利要求1至7任一项所述的机械通气过程中人机异步现象的分类方法。
  10. 一种计算机设备，其特征在于，所述计算机设备包括计算机可读存储介质、处理器和存储在所述计算机可读存储介质中的机械通气过程中人机异步现象的分类程序，所述机械通气过程中人机异步现象的分类程序被处理器执行时实现权利要求1至7任一项所述的机械通气过程中人机异步现象的分类方法。
PCT/CN2021/138034 2021-12-01 2021-12-14 机械通气过程中人机异步现象的分类方法和分类装置 WO2023097780A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111452509.6A CN114191665A (zh) 2021-12-01 2021-12-01 机械通气过程中人机异步现象的分类方法和分类装置
CN202111452509.6 2021-12-01

Publications (1)

Publication Number Publication Date
WO2023097780A1 true WO2023097780A1 (zh) 2023-06-08

Family

ID=80649916

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/138034 WO2023097780A1 (zh) 2021-12-01 2021-12-14 机械通气过程中人机异步现象的分类方法和分类装置

Country Status (2)

Country Link
CN (1) CN114191665A (zh)
WO (1) WO2023097780A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115040109B (zh) * 2022-06-20 2024-03-22 徐州工程学院 一种呼吸模式分类方法及系统
CN117045913B (zh) * 2023-07-14 2024-04-30 南通大学附属医院 一种基于呼吸变量监测的机械通气模式智能切换系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120037159A1 (en) * 2009-04-22 2012-02-16 Resmed Ltd Detection of asynchrony
US20200261674A1 (en) * 2017-11-09 2020-08-20 Autonomous Healthcare, Inc. Clinical Decision Support System for Patient-Ventilator Asynchrony Detection and Management
CN113539398A (zh) * 2021-06-25 2021-10-22 中国科学院深圳先进技术研究院 一种呼吸机人机异步分类方法、系统、终端以及存储介质
CN113521460A (zh) * 2021-05-20 2021-10-22 深圳先进技术研究院 机械通气人机异步检测方法、装置及计算机可读存储介质

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2009012599A1 (en) * 2007-07-26 2009-01-29 Uti Limited Partnership Transient intervention for modifying the breathing of a patient
CN103169476B (zh) * 2013-03-14 2015-03-25 中山大学 一种用于呼吸波形图像识别与预警的装置
CN111563451B (zh) * 2020-05-06 2023-09-12 浙江工业大学 基于多尺度小波特征的机械通气无效吸气努力识别方法
CN113539501A (zh) * 2021-06-25 2021-10-22 中国科学院深圳先进技术研究院 一种呼吸机人机异步分类方法、系统、终端以及存储介质
CN113689948A (zh) * 2021-08-18 2021-11-23 深圳先进技术研究院 呼吸机机械通气的人机异步检测方法、装置和相关设备
CN113642512B (zh) * 2021-08-30 2023-10-24 深圳先进技术研究院 呼吸机人机异步检测方法、装置、设备及存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120037159A1 (en) * 2009-04-22 2012-02-16 Resmed Ltd Detection of asynchrony
US20200261674A1 (en) * 2017-11-09 2020-08-20 Autonomous Healthcare, Inc. Clinical Decision Support System for Patient-Ventilator Asynchrony Detection and Management
CN113521460A (zh) * 2021-05-20 2021-10-22 深圳先进技术研究院 机械通气人机异步检测方法、装置及计算机可读存储介质
CN113539398A (zh) * 2021-06-25 2021-10-22 中国科学院深圳先进技术研究院 一种呼吸机人机异步分类方法、系统、终端以及存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
YUN-FEI LU, LU FEI, FANG LU-PING, GE HUI-QING, PAN QING: "Application of Wavelet Multi-scale Characteristics to Detect Patient-ventilator Asynchrony in Mechanical Ventilation", JOURNAL OF CHINESE COMPUTER SYSTEMS, vol. 41, no. 12, 12 December 2020 (2020-12-12), pages 2677 - 2682, XP093068729 *

Also Published As

Publication number Publication date
CN114191665A (zh) 2022-03-18

Similar Documents

Publication Publication Date Title
Gao et al. An effective LSTM recurrent network to detect arrhythmia on imbalanced ECG dataset
Yadav et al. Prediction of heart disease using feature selection and random forest ensemble method
Ambekar et al. Disease risk prediction by using convolutional neural network
Lin et al. An explainable deep fusion network for affect recognition using physiological signals
CN110459328B (zh) 临床监护设备
Li et al. Interpretability analysis of heartbeat classification based on heartbeat activity’s global sequence features and BiLSTM-attention neural network
CN112365978B (zh) 心动过速事件早期风险评估的模型的建立方法及其装置
Xu et al. A novel ensemble of random forest for assisting diagnosis of Parkinson's disease on small handwritten dynamics dataset
WO2023097780A1 (zh) 机械通气过程中人机异步现象的分类方法和分类装置
Long et al. A scoping review on monitoring mental health using smart wearable devices
Shen et al. A High‐Precision Fatigue Detecting Method for Air Traffic Controllers Based on Revised Fractal Dimension Feature
Aggarwal et al. A structured learning approach with neural conditional random fields for sleep staging
KR102421172B1 (ko) 앙상블 딥러닝과 형상 융합 기반 심장병 예측을 위한 스마트 헬스케어 모니터링 방법 및 시스템
CN112562808A (zh) 患者画像的生成方法、装置、电子设备及存储介质
Shan et al. Abnormal ECG detection based on an adversarial autoencoder
Mijwil et al. A scoping review of machine learning techniques and their utilisation in predicting heart diseases
CN115024725A (zh) 融合心理状态多参数检测的肿瘤治疗辅助决策系统
Lu et al. Video-based neonatal pain expression recognition with cross-stream attention
Nazlı et al. Classification of Coronary Artery Disease Using Different Machine Learning Algorithms
Hong et al. Gated temporal convolutional neural network and expert features for diagnosing and explaining physiological time series: a case study on heart rates
CN112487980B (zh) 基于微表情治疗方法、装置、系统与计算机可读存储介质
Ern et al. Classification of arrhythmia signals using hybrid convolutional neural network (cnn) model
Tang et al. Electrocardiogram Classification Using Long Short-Term Memory Networks
Gilani Machine learning classifiers for critical cardiac conditions
Sasikala et al. Transforming Healthcare with Deep Learning Cardiovascular Disease Prediction

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: 21966204

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE