CN112245728A - Respirator false positive alarm signal identification method and system based on integrated tree - Google Patents

Respirator false positive alarm signal identification method and system based on integrated tree Download PDF

Info

Publication number
CN112245728A
CN112245728A CN202010492039.5A CN202010492039A CN112245728A CN 112245728 A CN112245728 A CN 112245728A CN 202010492039 A CN202010492039 A CN 202010492039A CN 112245728 A CN112245728 A CN 112245728A
Authority
CN
China
Prior art keywords
data
alarm signal
data set
characteristic
label information
Prior art date
Legal status (The legal status 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 status listed.)
Granted
Application number
CN202010492039.5A
Other languages
Chinese (zh)
Other versions
CN112245728B (en
Inventor
刘佳明
李想
范皓玥
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Chemical Technology
Original Assignee
Beijing University of Chemical Technology
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 Beijing University of Chemical Technology filed Critical Beijing University of Chemical Technology
Priority to CN202010492039.5A priority Critical patent/CN112245728B/en
Publication of CN112245728A publication Critical patent/CN112245728A/en
Application granted granted Critical
Publication of CN112245728B publication Critical patent/CN112245728B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • 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
    • A61M16/0051—Devices for influencing the respiratory system of patients by gas treatment, e.g. mouth-to-mouth respiration; Tracheal tubes with alarm devices
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Emergency Medicine (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Pulmonology (AREA)
  • Anesthesiology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Hematology (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

The invention discloses a method and a system for identifying false positive alarm signals of a breathing machine based on an integrated tree, which comprises the following steps: s1, data collection: collecting monitoring data of a patient from a hospital ventilator and a monitor; s2, preprocessing data: processing missing values, abnormal values and standardization in the data set, and generating an identification rule of a false positive alarm signal; s3, feature extraction: sorting the features by using a random forest, and selecting the features with good identification capability; s4, false positive alarm signal identification: and establishing a false positive alarm signal identification method of the breathing machine and the monitor. Experimental results show that the method has excellent identification performance of false positive alarm signals, and the identification effect of the method is stable.

Description

Respirator false positive alarm signal identification method and system based on integrated tree
Technical Field
The invention relates to a method and a system for identifying false positive alarm signals of a hospital respirator-monitor, in particular to a method and a system for identifying false positive alarm signals of a respirator based on an integrated tree.
Background
A ventilator is widely used in modern clinical medicine as a kind of medical equipment for emergency treatment and life support, for example, for treating patients suffering from respiratory failure due to various causes, anesthesia and breathing management during major surgery, respiratory support therapy, and emergency resuscitation. Since the ventilator is mainly used for patients with high risk of illness, it is usually used with a monitor. When the breathing of the patient is abnormal or the equipment fails, the breathing machine-monitor sends out an alarm signal, and medical personnel check the state of the corresponding patient according to the alarm signal and check the running condition of the equipment. The effective alarm signal can help medical personnel to correctly identify and timely process alarm of the breathing machine, and normal work of the breathing machine and safety of patients are guaranteed.
However, during use of the ventilator-monitor, there are often situations where most of the alarm signals are false positive alarm signals. According to statistics, medical staff resources in many developing countries are in short supply, and particularly, during epidemic situations or emergencies, the medical staff is subjected to greater working pressure due to false positive alarm of the breathing machine. Frequent false positive alarms can cause medical personnel to alarm fatigue on the alarm signal, affecting the speed of response of the medical personnel to the alarm signal. When a plurality of breathing machine-monitors alarm simultaneously, the real dangerous patient is not checked in time and the best treatment opportunity is missed probably because of the occurrence of false positive alarm condition.
False positive alarm signals are identified based on real human body data monitored by a breathing machine and a monitor, the working pressure of medical personnel can be reduced, and the alertness of the medical personnel to the alarm signals can be improved. In addition, the accurate identification of false positive alarm signals can also strive for more timely medical aid for truly risky patients, relieve the alarm of a breathing machine, and enable the patients to be treated safely and effectively. At present, research on methods for analyzing and processing reasons of ventilator alarms is available, but in general, research on recognition of ventilator-monitor false positive alarm signals by means of a machine learning method has not been developed, and most research contents are discussed around problems caused by ventilator false positive alarms, so that an effective ventilator-monitor false positive alarm signal recognition method and system are not available at present. Therefore, it is very important to develop the identification method of false positive alarm signal for hospital respirator-monitor.
Therefore, there is an urgent need for a new method for identifying false positive alarm signals of a ventilator-monitor, which satisfies the following technical requirements: 1) the interpretation capability of the identification result can be effectively improved, and the sign indexes playing a key role in false positive alarm signals can be found; 2) the method and the system have good classification and identification effects and performance, and are a method and a system for effectively identifying false positive alarm signals of a breathing machine-monitor.
Disclosure of Invention
The invention solves the problems: the method and the system for identifying the false positive alarm signal of the breathing machine based on the integrated tree are provided to solve the problems that the subjective judgment or the identification effect is poor in the current identification of the false positive alarm signal of the breathing machine-monitor, and the problems that medical staff is high in pressure and misses the optimal treatment opportunity and the like caused by false positive alarm are solved.
The technical scheme adopted by the invention is as follows:
the invention provides a ventilator false positive alarm signal identification method based on an integrated tree, which comprises the following steps:
step 1) data collection: collecting sample data of a plurality of real breathing machines and a plurality of monitors of a patient from a hospital, and combining the sample data of each monitor and the sample data of each breathing machine to be used as an original data set, wherein the original data set comprises a plurality of characteristic data and corresponding alarm signals;
step 2) data preprocessing: carrying out missing value processing, abnormal value processing and data standardization processing on the characteristic data of the original data set in the step 1), and carrying out identification processing on alarm signals of the original data set so as to obtain a preprocessed data set, wherein the identification processing is to respectively identify different types of label information for the alarm signals according to a set rule, and the different types of label information are true positive alarm signals or false positive alarm signals;
step 3) feature selection: performing feature screening on the feature data of the preprocessed data set in the step 2) by using a random forest, reserving screened features, and forming a training data set by the screened feature data in the preprocessed data set and corresponding early warning signal label information;
step 4), false positive alarm signal identification: training the parameters of the gradient boosting decision tree classifier by using the training data set in the step 3) to obtain the trained alarm signal label information category identifier, and identifying the category of the corresponding alarm signal label information to be a true positive alarm signal or a false positive alarm signal according to newly input screened feature data and the corresponding early warning signal by the identifier.
Further, in the step 1:
the sample frequency of the real breathing machine-monitor monitoring data collected from the hospital is in seconds, and the data is collected three times per second;
the plurality of features comprises 16 individual feature features, wherein the 16 individual feature features are minute expiratory volume, average pressure, oxygen input port pressure, inspiratory oxygen concentration, respiratory non-positive pressure, spontaneous respiratory frequency, inspiratory tidal volume, expiratory tidal volume, peak pressure, invasive blood pressure mean, invasive blood pressure high value, invasive blood pressure low value, central venous pressure, blood oxygen concentration and heart rate respectively;
the specific implementation of combining each monitor data sample and each ventilator data sample is to combine each monitor data sample and each ventilator data sample into one sample by using a matching method and taking a ventilator as a main timestamp.
Further, in the step 2:
the missing value processing is specifically realized by adopting a feature mean value method to screen missing values of each feature data in the original data set and fill the missing values into a mean value of each feature data, wherein the value x 'of the missing value of the jth sample of the ith feature data in the original data set after filling the missing value through missing value processing'missing(i，j)，
Figure BDA0002521433560000031
Wherein x isi1，xi2，...，xinRespectively representing the 1 st, 2 nd, … th samples under the ith characteristic in the original data set, wherein n represents the samplesThe number;
the abnormal value processing is specifically realized by adopting a triple standard deviation method, firstly screening an abnormal value of which the difference between each characteristic data in the original data set and the mean value of the characteristic data is more than triple of the standard deviation of the characteristic data, and adjusting the abnormal value to be the sum of the mean value of the characteristic data and triple of the standard deviation of the characteristic data; then screening abnormal values in each feature data in the original data set, wherein the abnormal values in each feature data in the original data set are smaller than the difference of the mean value of the feature data by three times of the inverse number of the standard deviation of the feature data, and adjusting the abnormal values to be the difference of the mean value of the feature data and three times of the standard deviation of the feature data, wherein the abnormal values of the ith feature data of the jth sample in the original data set are adjusted to be values x 'after abnormal value processing'outlier(i，j)，
Figure BDA0002521433560000032
Wherein x isijRepresents the value of the j sample under the ith characteristic data in the original data set, muiRepresenting the mean, σ, of the ith characteristic data in the original data setiRepresenting a standard deviation of ith characteristic data in the original data set;
the normalization processing is realized by replacing the numerical value of each characteristic data in the original data set by the z-score of each characteristic data by using a z-score method, wherein the numerical value of the ith characteristic data of the jth sample in the original data set is replaced by the numerical value x 'after the abnormal value processing'norm(i，j)，
Figure BDA0002521433560000033
Wherein x isijA value, μ, representing the j sample under the i characteristic data in the original data setiRepresenting the mean, σ, of the ith characteristic data in the original data setiRepresenting a standard deviation of ith characteristic data in the original data set;
the specific implementation of the identification processing is that the established identification rule includes: when the respirator and the monitor alarm at the same time, the label information of the alarm signal is a true positive alarm signal; when continuous uninterrupted alarm occurs on the respirator or the monitor and the alarm duration times exceed 3 times, label information of the alarm signal is a true positive alarm signal; when the sign characteristic data exceeds the threshold range of the sign characteristic data set by the respirator, the label information of the alarm signal is a true positive alarm signal; the threshold range of the sign characteristic data comprises: the minute expiration amount is 0.5-180L/min, and the allowable error range is +/-3%; average pressure is-2 to 12kPa, and the allowable error range is +/-0.1 kPa; the concentration of the inhaled oxygen is 21-100%, and the allowable error range is +/-3%; the non-positive pressure of respiration is-12 to 12kPa, and the allowable error range is +/-0.05 kPa; the spontaneous respiratory frequency and the respiratory frequency are 1-150 times/minute, and the allowable error range is +/-3%; the inspiration/expiration tidal volume is-10L, and the allowable error range is +/-3%.
Further, in the step 3, the feature screening is performed by using a random forest, and the specific implementation of retaining the screened features is as follows:
calculating the difference information Gain (L, F) between the information Entropy (Encopy (L) of the alarm signal label information and the information Entropy (Encopy (L, F)) of the alarm signal label under the characteristic F for each characteristic F in the preprocessed data set,
Gain(L，F)＝Entropy(L)-Entropy(L，F)，
if Gain (L, F) > theta, keeping the characteristic F as the screened characteristic, and if Gain (L, F) < theta, deleting the characteristic F, and enabling theta to be a set threshold;
Figure BDA0002521433560000041
wherein L represents alarm tag information of the preprocessed data set, piThe probability that label information of the ith category of the alarm signal appears in the preprocessed data set is represented;
Figure BDA0002521433560000042
wherein L represents alarm signal label information of the preprocessed data set, v represents the number of values of the preprocessed data set under the characteristic F, and LjAnd representing the number of jth values of the preprocessed data set under the characteristic F.
The screened characteristics comprise peak pressure, heart rate, respiratory rate, spontaneous respiratory rate, expiratory tidal volume, inspiratory tidal volume, minute respiratory volume, average pressure and respiratory unpressurized pressure, and preferably comprise peak pressure, heart rate, respiratory rate and spontaneous respiratory rate.
Further, in the step 4, the setting range of the number of decision trees of the gradient boosting decision tree classifier is [50,150], the step size is 10, the setting range of the tree height is [3,10], the step size is 1, the setting range of the number of leaf nodes is [5,15], and the step size is 1.
The specific implementation of the step 4 is as follows:
41) taking the screened features obtained in the step 3) as an input feature vector space, and if the early warning signal label information output by the gradient lifting decision tree classifier of the (m-1) th round is Fm-1(x) Then the loss function L (y, F)m-1(x))＝y-Fm-1(x) Wherein x is a sample, and y is real early warning signal label information of the sample;
42) by L (y, F)m-1(x) Pair F)m-1(x) Derivation of the deviation
Figure BDA0002521433560000051
Obtaining the optimization direction and the learning rate gamma of the gradient lifting decision tree classifier of the mth roundm-1Controlling the contribution degree of the early warning signal label information output by the gradient lifting decision tree classifier in the m-1 th round, wherein the early warning signal label information output by the gradient lifting decision tree classifier in the m-1 th round is
Figure BDA0002521433560000052
43) Iteratively repeating steps 41) -42) Until the m-th round and the m-1 th round, the early warning signal label information F output by the gradient lifting decision tree classifier is identifiedm(x) And Fm-1(x) When the difference is smaller than a set threshold value, iteration is repeatedly stopped to obtain the trained alarm signal label information category identifier;
44) and the identifier identifies the type of the corresponding early warning signal label information as a true positive alarm signal or a false positive alarm signal according to the newly input screened characteristic data and the corresponding early warning signal.
The invention also provides a ventilator false positive alarm signal identification system based on the integrated tree, which comprises:
the system comprises a data acquisition module, a data preprocessing module and an alarm signal label information category identifier;
the data acquisition module acquires a monitoring data set of the breathing machine-monitor input by a user and sends the monitoring data set to the data preprocessing module, wherein the monitoring data set comprises a plurality of characteristic data and alarm signals, and the plurality of characteristics comprise peak pressure, heart rate, respiratory rate, spontaneous respiratory rate, expiratory tidal volume, inspiratory tidal volume, minute respiratory volume, average pressure and respiratory non-positive pressure, preferably peak pressure, heart rate, respiratory rate and spontaneous respiratory rate;
the data preprocessing module receives the monitoring data set sent by the data acquisition module, performs missing value processing, abnormal value processing and data standardization processing on the characteristic data in the detection data set, and sends the preprocessed monitoring data set to the alarm signal label information category identifier;
the alarm signal label information category recognizer is a trained gradient lifting decision tree classifier, receives the preprocessed monitoring data set sent by the data preprocessing module, and recognizes and outputs whether the category of the early warning signal label information is a true positive alarm signal or a false positive alarm signal.
Compared with the prior art, the invention has the advantages that:
(1) the invention discloses a method for recognizing false positive alarm signals of a hospital respirator and a monitor based on a gradient lifting decision tree method, which comprises the steps of firstly collecting real-time monitoring data of a patient from the hospital respirator and the monitor, then carrying out preprocessing operation on the data and identifying the false positive alarm signals, ensuring the integrity and the effectiveness of the data, then extracting effective characteristics of the data by using a random forest method, and finally recognizing and verifying the false positive alarm signals of the hospital respirator and the monitor by using the gradient lifting decision tree method;
(2) the false positive alarm signal identification method provided by the invention has good interpretation capability and provides a basis for finding key classification indexes;
(3) the method has excellent classification and identification performances, and has the best identification result in the aspects of accuracy, AUC, F1-SCORE and the like compared with other machine learning methods.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate an exemplary embodiment of the invention and, together with the description, serve to explain the invention and to make the aforementioned advantages of the invention more apparent. Wherein the content of the first and second substances,
FIG. 1 is a flow chart of a false positive alarm signal identification method of the present invention;
FIG. 2 is a data acquisition device, a ventilator and a monitor;
FIG. 3 is a comparison of the GBDT, SVM, NB, LR four methods F1-Score index, wherein (a) the training set test set ratio 80: 2; (b) training set to test set ratio 70: 30; (c) the ratio of a test set of a training set is 60:40, GBDT represents a gradient lifting decision tree, SVM represents a support vector machine, NB represents a naive Bayes classifier, and LR represents Logistic regression.
Detailed Description
In order to make the objects, technical solutions, implementation steps and advantages of the present invention more apparent, the following description is further detailed with reference to the accompanying drawings and implementation examples. It should be noted that the specific implementation examples of the present disclosure are only used for explaining the present invention, and are not used for limiting the present invention, and the technical solutions formed by combining the respective parts in the implementation examples are within the protection scope of the present invention.
The invention mainly aims at the problem that the working pressure of medical staff is increased rapidly because a breathing machine and a monitor matched with a serious patient frequently generate false positive alarm signals in a hospital environment, and provides an integrated tree-based breathing machine false positive alarm signal identification method for identifying the false positive alarm signals to relieve the working pressure of the medical staff, which comprises the following steps: s1, data collection: collecting monitoring data of a patient from a hospital respirator and a monitor as a raw data set; s2, preprocessing data: carrying out missing value, abnormal value and standardization processing on an original data set, carrying out identification processing on an alarm signal of the original data set, and respectively identifying different types of label information, namely a true positive alarm signal or a false positive alarm signal, for the alarm signal according to a set rule; s3, feature extraction: performing feature screening by using a random forest, and reserving screened features to further construct a training data set; s4, false positive alarm signal identification: training the parameters of the gradient boosting decision tree classifier by using a training set, establishing an alarm signal label information category recognizer, and recognizing and outputting the category of the corresponding alarm signal label information as a true positive alarm signal or a false positive alarm signal according to newly input screened feature data and the corresponding alarm signal. The invention firstly collects the real physical sign data of the patient from the breathing machine and the monitor of the hospital, carries out feature selection based on random forest after preprocessing the data, then realizes the identification work of false positive alarm signals of the breathing machine and the monitor by adopting a gradient lifting decision tree method, and carries out experimental verification. Experimental results show that the method has excellent identification performance of false positive alarm signals, and the identification effect of the method is stable.
The process of the method mainly comprises the following steps:
step 1) data collection: collecting sample data of a plurality of real breathing machines and a plurality of monitors of a patient from a hospital, and combining the sample data of each monitor and the sample data of each breathing machine to be used as an original data set, wherein the original data set comprises a plurality of characteristic data and corresponding alarm signals;
step 2) data preprocessing: carrying out missing value processing, abnormal value processing and data standardization processing on the characteristic data of the original data set in the step 1), and carrying out identification processing on alarm signals of the original data set so as to obtain a preprocessed data set, wherein the identification processing is to respectively identify different types of label information for the alarm signals according to a set rule, and the different types of label information are true positive alarm signals or false positive alarm signals;
step 3) feature selection: performing feature screening on the feature data of the preprocessed data set in the step 2) by using a random forest, reserving screened features, and forming a training data set by the screened feature data in the preprocessed data set and corresponding early warning signal label information;
further, in step 1):
according to the actual running states of the breathing machine and the monitor, the real state of the patient can be more accurately displayed by considering the data with higher resolution, the physical sign monitoring data of the patient are collected from the breathing machine and the monitor in a hospital, the frequency of the collected samples is in units of seconds, and the samples are collected for three times per second.
The collected information includes 16 individual characteristics, which are respectively minute expiratory volume, average pressure, oxygen input port pressure, inspiratory oxygen concentration, respiratory non-positive pressure, spontaneous respiratory frequency, inspiratory tidal volume, expiratory tidal volume, peak pressure, invasive blood pressure average value, invasive blood pressure high value, invasive blood pressure low value, central venous pressure, blood oxygen concentration and heart rate, and a label information for identifying an alarm signal.
The corresponding set of sign features is denoted as X ═ { X ═ X1，x2，...，x16Label information of 1 alarm signal is represented as Y ═ 0,1, where 0 indicates that no alarm occurs and 1 indicates that an alarm occurs.
Because the data is from two devices, namely a breathing machine and a detector, the problem that the sample time stamps of the two data sources are not uniform exists. And (3) taking a breathing machine as a main time stamp, and combining each monitor data sample and each breathing machine data sample into one sample by adopting a matching method. The concrete implementation is as follows: firstly, based on the acquisition time, respectively sequencing samples of a breathing machine and a monitor, correspondingly combining the monitor samples with the same time to the breathing machine samples, and deleting the samples corresponding to the breathing machine at the time when the breathing machine has the time sample and the monitor does not have the time sample, or vice versa.
Further, in step 2):
in the missing value processing step, due to unavoidable factors such as machine faults, the situation that completely random missing data occurs in the data set needs to be processed, a feature averaging method is adopted to fill the missing values, and the calculation formula is as follows:
Figure BDA0002521433560000071
wherein x isi1，xi2，...，xinRespectively representing the 1 st, 2 nd, the.
In the abnormal value processing step, due to reasons such as recording errors and machine abnormality, samples with obvious differences appear in data, in order to avoid negative influence of abnormal values in the data on identification of false positive alarm signals, the abnormal values are processed by using a variance of three times of standard deviation, namely, the data exceeding three times of standard deviation is adjusted to three times of standard deviation, so that the problem of the abnormal values is solved, and a specific calculation formula is as follows:
Figure BDA0002521433560000081
wherein x isijIndicating that the j sample under the ith characteristic data in the original data set is an abnormal value muiPresentation instrumentMean, σ, of ith feature data in the original data setiRepresenting the standard deviation of the ith feature data in the raw data set.
A standardization processing step, namely, in order to eliminate dimension problems among different physical sign characteristics and avoid the phenomenon that a classification result excessively deviates to a certain dimension larger characteristic, a z-score method is adopted to carry out standardization processing on the data characteristics, and a specific processing formula is as follows:
Figure BDA0002521433560000082
wherein x isijRepresents the j sample under the i characteristic data in the original data set, muiRepresenting the mean, σ, of the ith characteristic data in the original data setiRepresenting the standard deviation of the ith feature data in the raw data set.
And a step of marking false positive alarm signals, wherein the marking of the false positive alarm signals needs professional background knowledge of experts in relevant medical fields, after full comprehensive discussion, a rule of true positive alarm signals is provided, and the true positive alarm signals are marked from all the alarm signals, and the rest part is automatically divided into the false positive alarm signals. The identification rule includes three aspects: (1) when the breathing machine and the monitor alarm at the same time, the alarm signal is a true positive alarm signal; (2) when continuous uninterrupted alarm occurs in the breathing machine or the monitor, the alarm is a true positive alarm signal; (3) and when the physical sign data of the patient exceeds a set threshold value of the breathing machine, the alarm signal is a true positive alarm signal. The physical characteristics comprise minute breathing rate, flow range: (0.5-180) L/min, and the allowable error range is as follows: ± 3%, average pressure, pressure range: (-2 to 12) kPa, the allowable error range: 0.1kPa, inhaled oxygen concentration, range: 21% -100%, allowable error range: ± 3%, respiratory non-positive pressure, range: 12kPa, allowable error range: ± 0.05kPa, spontaneous and respiratory rate, frequency range: (1-150) times per minute, and the allowable error range is as follows: ± 3%, inspiratory/expiratory tidal volume, tidal volume: 10L, allowable error range: 3 percent.
Based on the rules, the identification work of the true positive and false positive alarm signals is completed, and a scientific data basis is provided for a subsequent classification method of supervised learning.
Further, in step 3):
the collected original data contains 16 individual characteristic data of the patient, wherein the characteristic with weaker performance for identifying false positive alarm signals does not exist, so that the part of the content adopts a characteristic selection method of Random Forest (Random Forest) to select the characteristics in the original data set, the number of the characteristics is reduced on the basis of not reducing the identification precision, and the calculation efficiency and the identification accuracy are improved.
The main idea of feature selection of the random forest is based on the idea of information gain in a decision tree. The random forest generates a plurality of decision trees with differences by performing double disturbance on the characteristics and samples of data, the idea of selecting the characteristics of the random forest is the same as or different from that of the decision trees, but the random forest has the advantage of integrating a plurality of decision trees, and can be expressed as follows in a model mode:
calculating the difference information Gain (L, F) between the information Entropy (Encopy (L) of the alarm signal label information and the information Entropy (Encopy (L, F)) of the alarm signal label under the characteristic F for each characteristic F in the preprocessed data set,
Gain(L，F)＝Entropy(L)-Entropy(L，F)，
if Gain (L, F) > theta, keeping the characteristic F as the screened characteristic, and if Gain (L, F) < theta, deleting the characteristic F, and enabling theta to be a set threshold;
information entropy of the alarm signal label information
Figure BDA0002521433560000091
Wherein L represents alarm tag information of the preprocessed data set, piProbability, p, of occurrence of label information representing an ith category of alarm signal in said preprocessed data setiThe method is obtained by calculating the number proportion of the false positive alarm signal samples in the preprocessed data set samples;
information of the alarm signal tag under the characteristic FEntropy of the entropy
Figure BDA0002521433560000092
Wherein L represents alarm signal tag information of the preprocessed data set, LjThe number of the characteristic F of the preprocessed data set which takes a certain numerical value is represented, v represents the number of different values of the preprocessed data set under the characteristic F, and j represents the index of the jth value of the preprocessed data set under the characteristic F.
The random forest method is realized by the following steps:
step 1), inputting training data of an initial feature set, calculating and outputting an information entropy (L) of an alarm signal, and providing a basis for calculating information gain in step 3);
step 2) inputting training data of an initial feature set, calculating and outputting alarm signal conditional Entropy (L, F) of a feature F under each tree, and providing basis for calculating information gain in the step 3);
and 3) taking the results of the step 1) and the step 2) as input, calculating the information Gain (L, F) of the characteristic F under each tree, and taking the average value of the information Gain according to the number of the trees. The larger the information gain, the more important the feature is to the classification result. And setting a threshold value theta for selecting the characteristic, and keeping the characteristic when Gain (L, F) exceeds theta, otherwise deleting the characteristic.
Further, in step 4):
gradient boosting decision tree parameter initialization process: the main parameters influencing the classification performance of the gradient lifting decision tree comprise the number of trees, the height of the trees and the number of leaf nodes, the setting range of the number of decision trees of the gradient lifting decision tree classifier is [50,150], the step length is 10, the setting range of the tree height is [3,10], the step length is 1, the setting range of the leaf node number is [5,15] and the step length is 1.
Taking the feature subset obtained in the step 3 as an input feature vector space of the step, and assuming that the classifier classification result of the (m-1) th round is F in order to eliminate the residual error because the calculation purpose of gradient lifting is to reduce the residual error of the last calculation resultm-1(x) Then the loss function is defined as: l (y, F)m-1(x))＝y-Fm-1(x) Where x is the sample and y is the true alarm signal value of the sample. Gradient boosting decision tree by pair penalty function L (y, F)m-1(x) Predicted value F of)m-1(xi) Derivation of the deviation
Figure BDA0002521433560000101
Obtaining the optimized direction of the next round of decision tree and using the learning rate gammamControlling the contribution degree of decision tree of each round to the classification result, the classifier result of the mth round can be expressed as
Figure BDA0002521433560000102
Iteratively repeating the steps until the early warning signal label information category F identified by the gradient lifting decision tree classifier of the mth roundm(x) Early warning signal label information category F identified by the gradient boosting decision tree classifier of round m-1m-1(x) When the difference is smaller than a set threshold value, iteration is stopped repeatedly, the gradient lifting decision tree classifier finishes training to obtain the trained alarm signal label information category identifier, wherein Fm(xi) Represents the predicted label result of the ith sample of the mth iteration, yiTrue tag information, L (y), representing a sample ii，Fm(xi) Is a loss function, i.e. the error between the true tag value and the predicted tag value.
And the identifier identifies the type of the corresponding early warning signal label information as a true positive alarm signal or a false positive alarm signal according to the newly input screened characteristic data and the corresponding early warning signal.
The invention relates to a hospital respirator-monitor false positive alarm signal identification system, which comprises:
a data acquisition module: the method comprises the steps of obtaining a monitoring data set of the breathing machine-monitor input by a user, and sending the monitoring data set to a data preprocessing module, wherein the monitoring data set comprises a plurality of characteristic data and alarm signals, the plurality of characteristics comprise peak pressure, heart rate, respiratory frequency, spontaneous respiratory frequency, expiratory tidal volume, inspiratory tidal volume, minute respiratory volume, average pressure and respiratory unpressurized pressure, and the plurality of characteristics are preferably peak pressure, heart rate, respiratory frequency and spontaneous respiratory frequency;
a data preprocessing module: receiving a monitoring data set sent by the data acquisition module, performing missing value processing, abnormal value processing and data standardization processing on characteristic data in the detection data set, and sending the preprocessed monitoring data set to an alarm signal label information category identifier;
and the alarm signal label information category identifier is a trained gradient lifting decision tree classifier, receives the preprocessed monitoring data set sent by the data preprocessing module, and identifies and outputs the category of the alarm signal label information as a true positive alarm signal or a false positive alarm signal.
In order to verify the performance of the method in the identification of false positive alarm signals of a ventilator-monitor, empirical experiments were performed, collecting real patient monitoring data of a plurality of ventilators and monitors from an intensive care unit of a hospital, the ventilator-monitor being as shown in fig. 2. The data sample size was 15006 patient signs and alarm signal recordings.
For the purpose of performance demonstration, three common Machine learning classification methods are selected as comparison methods, namely, Logistic Regression (LR), Support Vector Machine (SVM) and Naive Bayes classifier (Naive Bayes, NB) are compared with the method provided by the invention, and the adopted judgment indexes of false positive alarm signal identification performance comprise Accuracy (Accuracy), first Type error rate (Type I error), second Type error rate (Type II error), AUC and F1-score. The experimental flow is shown in figure 1:
in order to provide an intuitive understanding of the data sets collected by hospital ventilators and monitors, data samples are used as shown in table 1:
table 1: monitoring information dataset samples
x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12 x13 x14 x15 x16 y
8.1 7.2 56.6 50 5.4 29.9 29.9 258 291 15.5 79 111 65 30 99 111 1
7.9 7.5 56.6 50 5.4 30 30 257 260 15.6 80 112 66 19 100 109 1
7.6 7.6 56.6 50 5.3 30.1 30.1 272 292 15 80 111 66 19 100 108 1
7.7 7.5 56.6 50 5 30.1 30.1 232 230 15.4 78 108 65 20 100 107 1
6.7 7.8 56.6 50 5.3 30.3 30.3 254 288 15.2 78 109 65 21 99 105 1
8 7.4 56.6 50 5.6 29.5 29.5 255 254 15.6 79 110 65 19 100 109 0
7.7 7.5 56.6 50 5.3 28 28 237 257 15.3 78 109 64 19 98 110 0
7.2 7.7 56.6 50 5.1 26.5 26.5 300 297 15.3 78 109 64 19 100 109 0
6.9 7.5 56.6 50 5.7 25.5 25.5 275 256 15.4 78 109 64 19 100 109 0
7.3 7.3 56.6 50 5.4 27.9 27.9 240 255 15.6 78 110 64 19 100 109 0
In Table 1, x1…x16Corresponding to 16 physical sign information (minute expiration volume, average pressure, oxygen inlet pressure, inspired oxygen concentration, breath non-positive pressure, spontaneous respiratory frequency, inspiratory tidal volume, expiratory tidal volume, peak pressure, invasive blood pressure average value, invasive blood pressure high value, invasive blood pressure low value, central venous pressure, blood oxygen concentration and heart rate) representing the patient, y represents the type of the alarm signal (y is 1 represents a false positive alarm signal, and y is 0 represents a true positive alarm signalPositive alarm signal)
In order to avoid randomness possibly caused by one experiment, 30 experiments are carried out in a random sampling mode, wherein the division ratio of the training sample and the test sample is 20%, 30% and 40%, and finally the average result and variance of the 30 experiments are taken to judge the performance of the method. The results of the method proposed by the invention and the results of the comparative method are listed in tables 2-4, respectively:
table 2: comparison of the Performance of the GBDT method and of the comparison method (LR, SVM and NB) (training set: test set: 80:20)
Figure BDA0002521433560000111
Figure BDA0002521433560000121
As can be seen from the classification results in Table 2, the gradient boosting decision tree GBDT obtains the best recognition effect of false positive alarm signals, compared with the other four methods, the method obtains very good classification results on the accuracy, the second type error rate, the AUC and the F1-Score, and the AUC reaches 97.6%. Although the error rate of the gradient boosting decision tree is higher than that of Logistic regression and a support vector machine in the first type of error rate, the invention aims at the field of false positive alarm signal identification, namely, the second type of error rate really reflects the identification success rate of a method for false positive alarm signals. Therefore, even though Logistic regression has a good effect on the first type of error rate, the second type of error rate is very high, and the result shows 56.4%, which indicates that the method has a poor effect on identifying the false positive alarm type. In addition, AUC and F1-Score are comprehensive indexes for measuring the classification effect of two types of signals, and the gradient boost decision tree is far superior to other methods in the performance of the two indexes. The F1-Score pair for the four models is shown in FIG. 3 (a). Therefore, the method provided by the invention has a very good function of identifying the false positive alarm model.
Table 3: comparison of the Performance of the GBDT method and of the comparison method (LR, SVM and NB) (training set: test set: 70:30)
Metrics GBDT LR SVM NB
AUC 0.972(0.02) 0.691(0.03) 0.837(0.03) 0.837(0.03)
Accuracy 0.997(0.00) 0.989(0.00) 0.994(0.00) 0.941(0.01)
TypeIError 0.002(0.00) 0.000(0.00) 0.000(0.00) 0.060(0.01)
TypeIIError 0.054(0.04) 0.619(0.06) 0.325(0.06) 0.012(0.01)
F1-Score 0.914(0.04) 0.547(0.06) 0.794(0.04) 0.364(0.03)
Table 3 reflects the recognition effect of each method at different training set and test set ratios, and the overall results are consistent with those of table 2. The proposed gradient lifting decision tree method has robustness in the identification of false positive alarm signals, and the signal identification effect is stable. The F1-Score pair for the four models is shown in FIG. 3 (b).
Table 4: comparison of the Performance of the GBDT method and of the comparison method (LR, SVM and NB) (training set: test set ═ 60:40)
Figure BDA0002521433560000122
Figure BDA0002521433560000131
The classification effect reflected in table 4 and tables 2 and 3 is the same, and the signal identification performance of all methods is reduced as a whole due to the reduction of the sample size of the training set, but the whole judgment is not influenced, namely, the gradient lifting decision tree has the optimal identification function of false positive alarm signals. In addition, the variance result of 30 experimental results in parentheses is observed, so that the performance of the gradient lifting decision tree is very stable and no large fluctuation is generated. For further clear demonstration of the experimental results, reference may be further made to the comparison results of the F1-SCORE indexes of the four methods GBDT, LR, SVM and NB provided in (c) of fig. 3.
The invention discloses a method for identifying false positive alarm signals of a breathing machine based on an integrated tree. Experimental results show that the method has excellent identification performance of false positive alarm signals, and the identification effect of the method is stable.
The above examples are provided only for the purpose of describing the present invention, and are not intended to limit the scope of the present invention. The scope of the invention is defined by the appended claims. Various equivalent substitutions and modifications can be made without departing from the spirit and principles of the invention, and are intended to be within the scope of the invention.

Claims (8)

1. A respiratory machine false positive alarm signal identification method based on an integrated tree is characterized by comprising the following steps:
step 1) data collection: collecting sample data of a plurality of real breathing machines and a plurality of monitors of a patient from a hospital, and combining the sample data of each monitor and the sample data of each breathing machine to be used as an original data set, wherein the original data set comprises a plurality of characteristic data and corresponding alarm signals;
step 2) data preprocessing: carrying out missing value processing, abnormal value processing and data standardization processing on the characteristic data of the original data set in the step 1), and carrying out identification processing on alarm signals of the original data set so as to obtain a preprocessed data set, wherein the identification processing is to respectively identify different types of label information for the alarm signals according to a set rule, and the different types of label information are true positive alarm signals or false positive alarm signals;
step 3) feature selection: performing feature screening on the feature data of the preprocessed data set in the step 2) by using a random forest, reserving screened features, and forming a training data set by the screened feature data in the preprocessed data set and corresponding early warning signal label information;
step 4), false positive alarm signal identification: training the parameters of the gradient boosting decision tree classifier by using the training data set in the step 3) to obtain the trained alarm signal label information category identifier, and identifying the category of the corresponding alarm signal label information to be a true positive alarm signal or a false positive alarm signal according to newly input screened feature data and the corresponding early warning signal by the identifier.
2. The integrated tree based ventilator false positive alarm signal identification method as claimed in claim 1, wherein in the step 1:
the sample frequency of the real breathing machine-monitor monitoring data collected from the hospital is in seconds, and the data is collected three times per second;
the plurality of features comprises 16 individual feature features, wherein the 16 individual feature features are minute expiratory volume, average pressure, oxygen input port pressure, inspiratory oxygen concentration, respiratory non-positive pressure, spontaneous respiratory frequency, inspiratory tidal volume, expiratory tidal volume, peak pressure, invasive blood pressure mean, invasive blood pressure high value, invasive blood pressure low value, central venous pressure, blood oxygen concentration and heart rate respectively;
the specific implementation of combining each monitor data sample and each ventilator data sample is to combine each monitor data sample and each ventilator data sample into one sample by using a matching method and taking a ventilator as a main timestamp.
3. The integrated tree based ventilator false positive alarm signal identification method as claimed in claim 1, wherein in the step 2:
the missing value processing is specifically realized by adopting a feature averaging method to screen the missing value of each feature data in the original data set and fill the missing value into the average value of each feature data, wherein the missing value is processed by the original data setThe value x 'of the missing value of the jth sample of the ith characteristic data after being padded by missing value processing'missing(i，j)，
Figure FDA0002521433550000021
Wherein x isi1，xi2，...，xinRespectively representing the 1 st, 2 nd, … th samples under the ith characteristic in the original data set, wherein n represents the number of the samples;
the abnormal value processing is specifically realized by adopting a triple standard deviation method, firstly screening an abnormal value of which the difference between each characteristic data in the original data set and the mean value of the characteristic data is more than triple of the standard deviation of the characteristic data, and adjusting the abnormal value to be the sum of the mean value of the characteristic data and triple of the standard deviation of the characteristic data; then screening abnormal values in each feature data in the original data set, wherein the abnormal values in each feature data in the original data set are smaller than the difference of the mean value of the feature data by three times of the inverse number of the standard deviation of the feature data, and adjusting the abnormal values to be the difference of the mean value of the feature data and three times of the standard deviation of the feature data, wherein the abnormal values of the ith feature data of the jth sample in the original data set are adjusted to be values x 'after abnormal value processing'outlier(i，j)，
Figure FDA0002521433550000022
Wherein x isijRepresents the value of the j sample under the ith characteristic data in the original data set, muiRepresenting the mean, σ, of the ith characteristic data in the original data setiRepresenting a standard deviation of ith characteristic data in the original data set;
the normalization processing is realized by replacing the numerical value of each characteristic data in the original data set by the z-score of each characteristic data by using a z-score method, wherein the numerical value of the ith characteristic data of the jth sample in the original data set is replaced by the numerical value x 'after the abnormal value processing'norm(i，j)，
Figure FDA0002521433550000023
Wherein x isijA value, μ, representing the j sample under the i characteristic data in the original data setiRepresenting the mean, σ, of the ith characteristic data in the original data setiRepresenting a standard deviation of ith characteristic data in the original data set;
the specific implementation of the identification processing is that the established identification rule includes: when the respirator and the monitor alarm at the same time, the label information of the alarm signal is a true positive alarm signal; when continuous uninterrupted alarm occurs on the respirator or the monitor and the alarm duration times exceed 3 times, label information of the alarm signal is a true positive alarm signal; when the sign characteristic data exceeds the threshold range of the sign characteristic data set by the respirator, the label information of the alarm signal is a true positive alarm signal; the threshold range of the sign characteristic data comprises: the minute expiration amount is 0.5-180L/min, and the allowable error range is +/-3%; average pressure is-2 to 12kPa, and the allowable error range is +/-0.1 kPa; the concentration of the inhaled oxygen is 21-100%, and the allowable error range is +/-3%; the non-positive pressure of respiration is-12 to 12kPa, and the allowable error range is +/-0.05 kPa; the spontaneous respiratory frequency and the respiratory frequency are 1-150 times/minute, and the allowable error range is +/-3%; the inspiration/expiration tidal volume is-10L, and the allowable error range is +/-3%.
4. The integrated tree based ventilator false positive alarm signal identification method as claimed in claim 1, wherein in the step 3, the feature screening is performed by using random forest, and the specific implementation of retaining the screened features is as follows:
calculating the difference information Gain (L, F) between the information Entropy (Encopy (L) of the alarm signal label information and the information Entropy (Encopy (L, F)) of the alarm signal label under the characteristic F for each characteristic F in the preprocessed data set,
Gain(L，F)＝Entropy(L)-Entropy(L，F)，
if Gain (L, F) > theta, keeping the characteristic F as the screened characteristic, and if Gain (L, F) < theta, deleting the characteristic F, and enabling theta to be a set threshold;
Figure FDA0002521433550000031
wherein L represents alarm tag information of the preprocessed data set, piThe probability that label information of the ith category of the alarm signal appears in the preprocessed data set is represented;
Figure FDA0002521433550000032
wherein L represents alarm signal label information of the preprocessed data set, v represents the number of values of the preprocessed data set under the characteristic F, and LjAnd representing the number of jth values of the preprocessed data set under the characteristic F.
5. The integrated tree based method of identifying false positive alarm signals for ventilators as claimed in claim 1, wherein in step 3, the filtered features comprise peak pressure, heart rate, respiratory rate, spontaneous respiratory rate, expiratory tidal volume, inspiratory tidal volume, minute respiratory volume, mean pressure, and respiratory non-positive pressure.
6. The integrated tree based ventilator false positive alarm signal recognition method of claim 1, wherein in the step 4, the number of decision trees of the gradient boosting decision tree classifier is set to be in a range of [50,150], the step size is 10, the tree height is set to be in a range of [3,10], the step size is 1, the number of leaf nodes is set to be in a range of [5,15], and the step size is 1.
7. The integrated tree based ventilator false positive alarm signal identification method as claimed in claim 1, wherein the step 4 is implemented as follows:
41) taking the screened features obtained in the step 3) as an input feature vector space, and if the early warning signal label information output by the gradient lifting decision tree classifier of the (m-1) th round is Fm-1(x) Then the loss function L (y, F)m-1(x))＝y-Fm-1(x) Wherein x is a sample, and y is real early warning signal label information corresponding to the sample;
42) by L (y, F)m-1(x) Pair F)m-1(x) Derivation of the deviation
Figure FDA0002521433550000033
Obtaining the optimization direction and the learning rate gamma of the gradient lifting decision tree classifier of the mth roundm-1Controlling the contribution degree of the early warning signal label information output by the gradient lifting decision tree classifier in the m-1 th round, wherein the early warning signal label information output by the gradient lifting decision tree classifier in the m-1 th round is
Figure FDA0002521433550000034
43) Iteratively repeating the steps 41) to 42) until the gradient lifting decision tree classifiers of the mth round and the mth-1 round identify the output early warning signal label information Fm(x) And Fm-1(x) When the difference is smaller than a set threshold value, iteration is repeatedly stopped to obtain the trained alarm signal label information category identifier;
44) and the identifier identifies the type of the corresponding early warning signal label information as a true positive alarm signal or a false positive alarm signal according to the newly input screened characteristic data and the corresponding early warning signal.
8. An integrated tree based ventilator false positive alarm signal identification system, the system comprising:
the system comprises a data acquisition module, a data preprocessing module and an alarm signal label information category identifier;
the data acquisition module acquires a monitoring data set of the breathing machine-monitor input by a user and sends the monitoring data set to the data preprocessing module, wherein the monitoring data set comprises a plurality of characteristic data and alarm signals, and the plurality of characteristics comprise peak pressure, heart rate, respiratory rate, spontaneous respiratory rate, expiratory tidal volume, inspiratory tidal volume, minute respiratory volume, average pressure and non-positive respiratory pressure;
the data preprocessing module receives the monitoring data set sent by the data acquisition module, performs missing value processing, abnormal value processing and data standardization processing on the characteristic data in the detection data set, and sends the preprocessed monitoring data set to the alarm signal label information category identifier;
the alarm signal label information category recognizer is a trained gradient lifting decision tree classifier, receives the preprocessed monitoring data set sent by the data preprocessing module, and recognizes and outputs whether the category of the early warning signal label information is a true positive alarm signal or a false positive alarm signal.
CN202010492039.5A 2020-06-03 2020-06-03 Respirator false positive alarm signal identification method and system based on integrated tree Active CN112245728B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010492039.5A CN112245728B (en) 2020-06-03 2020-06-03 Respirator false positive alarm signal identification method and system based on integrated tree

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010492039.5A CN112245728B (en) 2020-06-03 2020-06-03 Respirator false positive alarm signal identification method and system based on integrated tree

Publications (2)

Publication Number Publication Date
CN112245728A true CN112245728A (en) 2021-01-22
CN112245728B CN112245728B (en) 2022-11-29

Family

ID=74224209

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010492039.5A Active CN112245728B (en) 2020-06-03 2020-06-03 Respirator false positive alarm signal identification method and system based on integrated tree

Country Status (1)

Country Link
CN (1) CN112245728B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112800983A (en) * 2021-02-01 2021-05-14 玉林师范学院 Non-line-of-sight signal identification method based on random forest
CN113349746A (en) * 2021-07-21 2021-09-07 中南大学湘雅医院 Vital sign monitoring alarm system
CN115399738A (en) * 2022-08-17 2022-11-29 中南大学湘雅医院 Quick ICU false alarm identification method
CN117612725A (en) * 2024-01-23 2024-02-27 南通大学附属医院 Respirator alarm management method and system for intensive care unit

Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120050074A1 (en) * 2010-02-26 2012-03-01 Bechtel Jon H Automatic vehicle equipment monitoring, warning, and control system
CN104414636A (en) * 2013-08-23 2015-03-18 北京大学 Magnetic resonance image based cerebral micro-bleeding computer auxiliary detection system
CN106730209A (en) * 2017-01-18 2017-05-31 湖南明康中锦医疗科技发展有限公司 The method and lung ventilator of ventilator alarm
CN106961249A (en) * 2017-03-17 2017-07-18 广西大学 A kind of diagnosing failure of photovoltaic array and method for early warning
CN109087482A (en) * 2018-09-18 2018-12-25 西安交通大学 A kind of falling detection device and method
CN109117956A (en) * 2018-07-05 2019-01-01 浙江大学 A kind of determination method of optimal feature subset
CN109489800A (en) * 2018-12-14 2019-03-19 广东世港信息科技有限公司 A kind of disturbance event recognition methods in distribution optic cable vibration safety pre-warning system
CN110024043A (en) * 2016-11-29 2019-07-16 皇家飞利浦有限公司 False alarm detection
CN110298085A (en) * 2019-06-11 2019-10-01 东南大学 Analog-circuit fault diagnosis method based on XGBoost and random forests algorithm
CN110519128A (en) * 2019-09-20 2019-11-29 西安交通大学 A kind of operating system recognition methods based on random forest
CN110544373A (en) * 2019-08-21 2019-12-06 北京交通大学 truck early warning information extraction and risk identification method based on Beidou Internet of vehicles

Patent Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120050074A1 (en) * 2010-02-26 2012-03-01 Bechtel Jon H Automatic vehicle equipment monitoring, warning, and control system
CN104414636A (en) * 2013-08-23 2015-03-18 北京大学 Magnetic resonance image based cerebral micro-bleeding computer auxiliary detection system
CN110024043A (en) * 2016-11-29 2019-07-16 皇家飞利浦有限公司 False alarm detection
CN106730209A (en) * 2017-01-18 2017-05-31 湖南明康中锦医疗科技发展有限公司 The method and lung ventilator of ventilator alarm
CN106961249A (en) * 2017-03-17 2017-07-18 广西大学 A kind of diagnosing failure of photovoltaic array and method for early warning
CN109117956A (en) * 2018-07-05 2019-01-01 浙江大学 A kind of determination method of optimal feature subset
CN109087482A (en) * 2018-09-18 2018-12-25 西安交通大学 A kind of falling detection device and method
CN109489800A (en) * 2018-12-14 2019-03-19 广东世港信息科技有限公司 A kind of disturbance event recognition methods in distribution optic cable vibration safety pre-warning system
CN110298085A (en) * 2019-06-11 2019-10-01 东南大学 Analog-circuit fault diagnosis method based on XGBoost and random forests algorithm
CN110544373A (en) * 2019-08-21 2019-12-06 北京交通大学 truck early warning information extraction and risk identification method based on Beidou Internet of vehicles
CN110519128A (en) * 2019-09-20 2019-11-29 西安交通大学 A kind of operating system recognition methods based on random forest

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
吴爽等: "HOG结合随机森林的新型手势识别框架", 《湘潭大学自然科学学报》 *
曹玉珍等: "基于深度学习的癫痫脑电通道选择与发作检测", 《天津大学学报(自然科学与工程技术版)》 *
李天庆 等: "呼吸机假阳性报警的现状与应对策略探讨", 《医疗卫生装备》 *
杨玉志: "医疗设备报警提示系统的探讨", 《中国医疗设备》 *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112800983A (en) * 2021-02-01 2021-05-14 玉林师范学院 Non-line-of-sight signal identification method based on random forest
CN112800983B (en) * 2021-02-01 2024-03-08 玉林师范学院 Random forest-based non-line-of-sight signal identification method
CN113349746A (en) * 2021-07-21 2021-09-07 中南大学湘雅医院 Vital sign monitoring alarm system
CN115399738A (en) * 2022-08-17 2022-11-29 中南大学湘雅医院 Quick ICU false alarm identification method
CN115399738B (en) * 2022-08-17 2023-05-16 中南大学湘雅医院 Rapid ICU false alarm identification method
CN117612725A (en) * 2024-01-23 2024-02-27 南通大学附属医院 Respirator alarm management method and system for intensive care unit
CN117612725B (en) * 2024-01-23 2024-03-29 南通大学附属医院 Respirator alarm management method and system for intensive care unit

Also Published As

Publication number Publication date
CN112245728B (en) 2022-11-29

Similar Documents

Publication Publication Date Title
CN112245728B (en) Respirator false positive alarm signal identification method and system based on integrated tree
Rose et al. Cough augmentation techniques for extubation or weaning critically ill patients from mechanical ventilation
RU2515401C2 (en) System and method of identifying respiratory insufficiency of subject&#39;s respiration
US11577037B2 (en) Automatic detection of airway device, endotracheal intubation, and tube misplacement in children during the anesthesia procedure
JP2006502481A (en) Neural network in sedation and analgesia system
Jaffe Using the features of the time and volumetric capnogram for classification and prediction
US10340039B2 (en) Managing patient devices based on sensor data
CN107887032A (en) A kind of data processing method and device
Kamio et al. Mechanical ventilation-related safety incidents in general care wards and ICU settings
KR20210066271A (en) Order system using medical deep learning in the field of anesthesia
Durmuşoğlu et al. Remembering Medical Ventilators and Masks in the Days of COVID-19: Patenting in the Last Decade in Respiratory Technologies
CN116530943B (en) Anesthesia depth detection device based on blood gas data
Mahmud et al. Res-se-convnet: A deep neural network for hypoxemia severity prediction for hospital in-patients using photoplethysmograph signal
Sá et al. Automated breath detection on long-duration signals using feedforward backpropagation artificial neural networks
Douglas et al. Performance of first aid trained staff using a modified START triage tool at achieving appropriate triage compared to a physiology-based triage strategy at Australian mass gatherings
CN115050454B (en) Method, device, equipment and storage medium for predicting mechanical ventilation offline
Betancourt et al. Segmented wavelet decomposition for capnogram feature extraction in asthma classification
CN115089140A (en) Sleep analysis system and method capable of monitoring heart rate and respiration
EP4070327A1 (en) Assistance in the detection of pulmonary diseases
Oliveira et al. Feature selection for detecting patients with weaning failures in Intensive Medicine
CN111383764A (en) Correlation detection system for mechanical ventilation driving pressure and related events of breathing machine
CN117612725B (en) Respirator alarm management method and system for intensive care unit
Armarego et al. High‐flow nasal cannula therapy for infants with bronchiolitis
CN116469568A (en) Breathing training model creation mode and equipment based on breathing training acquisition data
CN113257399B (en) Automatic emergency first-aid article acquisition method and system based on semantic analysis

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant