CN111738305A - Mechanical ventilation man-machine asynchronous rapid identification method based on DBA-DTW-KNN - Google Patents
Mechanical ventilation man-machine asynchronous rapid identification method based on DBA-DTW-KNN Download PDFInfo
- Publication number
- CN111738305A CN111738305A CN202010474275.4A CN202010474275A CN111738305A CN 111738305 A CN111738305 A CN 111738305A CN 202010474275 A CN202010474275 A CN 202010474275A CN 111738305 A CN111738305 A CN 111738305A
- Authority
- CN
- China
- Prior art keywords
- sequence
- template
- dtw
- dba
- knn
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/30—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
- G16H40/60—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the operation of medical equipment or devices
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Anesthesiology (AREA)
- Animal Behavior & Ethology (AREA)
- Hematology (AREA)
- Veterinary Medicine (AREA)
- Business, Economics & Management (AREA)
- General Business, Economics & Management (AREA)
- Heart & Thoracic Surgery (AREA)
- Pulmonology (AREA)
- Emergency Medicine (AREA)
- Biophysics (AREA)
- Physical Education & Sports Medicine (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention provides a mechanical ventilation man-machine asynchronous rapid identification method based on DBA-DTW-KNN, which comprises the following steps: reading respiratory waveform data in real time to form a test sequence, and carrying out standardization processing on the respiratory waveform data; the DTW distance of the test sequence from all sequences in the training set is then calculated. And calculating the similarity distance by adopting DTW, and classifying the test sequence samples according to the set K value by combining the KNN classification idea. Wherein the training set uses DBA and DTW based data set compression. The invention can be used for judging the existence of invalid inspiration effort phenomenon in man-machine asynchronism, further evaluating the reasonability of the parameter setting of the breathing machine and laying a foundation for medical staff to adjust the parameter setting of the breathing machine.
Description
Technical Field
The invention relates to a quick identification method for man-machine dyssynchrony in mechanical ventilation, which is characterized in that data set compression is realized based on DBA and DTW, ineffective inspiration effort of a patient in mechanical ventilation is identified based on a classification thought of KNN, and further reasonability of breathing machine parameter setting can be evaluated, and auxiliary teaching is provided for related fields of adjusting breathing machine parameter setting.
Background
In the Intensive Care Unit (ICU), Mechanical Ventilation (MV) is an important life support for patients with acute respiratory failure. However, when the patient's respiratory demand does not match the ventilator setting parameters, man-machine dyssynchrony can result, which can lead to a series of adverse clinical outcomes. Common types of dyssynchrony between humans include ineffective inspiratory effort, double triggering, too short a cycle, too long a cycle, etc. The present invention is directed primarily to identification of the type of invalid inhalation Effort (IEE). Ineffective inspiratory effort refers to the patient not triggering the ventilator to deliver air after an inspiratory effort, which is primarily manifested in the respiratory waveform as a protrusion in the expiratory phase flow rate with a depression in the pressure waveform.
The most common early human-machine asynchronous identification method was based on visual observation and evaluation of ventilator waveforms at the bedside, which required a significant expenditure of medical personnel resources. There are also reports in the literature that IEE identification can be achieved by a rule-based approach, i.e. finding a maximum value in the flow-rate expiratory phase, then finding a minimum value between the maximum value and the expiratory end point, calculating the difference between the maximum value and the minimum value, and setting a threshold value, and when the difference is greater than the threshold value, considering the breath as an ineffective inspiratory effort; random Forest (RF) and adaptive boost (AdaBoost) are two common methods in the field of machine learning, and are widely used in various fields. Therefore, in recent years, researchers have proposed using random forests and adaptive reinforcement learning methods to achieve IEE identification. However, the above methods have great limitations in IEE identification: the rule-based IEE identification method is sensitive to the selection of the threshold and requires accurate detection of the expiratory starting point; the IEE identification method based on RF and AdaBoost also involves detection of expiratory points when extracting features, and the calculation of each feature is also troublesome. Therefore, it is necessary to design a more convenient method for automatically identifying the man-machine dyssynchrony of mechanical ventilation.
Disclosure of Invention
In order to solve the problems that accurate detection of an expiratory starting point is difficult, threshold setting is uncertain, KNN classification consumes time and the like, the invention provides a mechanical ventilation man-machine asynchronous recognition method based on DBA-DTW-KNN.
The technical scheme adopted by the invention for solving the technical problems is as follows:
a mechanical ventilation man-machine asynchronous rapid identification method based on DBA-DTW-KNN comprises the following steps: real-time reading respiratory waveform data to form test sequence Ql=(q1,q2,...,qp) And standardizing the respiratory waveform data; then calculating the test sequence QlAnd training set C ═ C (C)1,C2,...,CE) DTW distance of all sequences in (a). Calculating similarity distance by adopting DTW (dynamic time warping), combining KNN (K-nearest neighbor) classification thought, and according to set K value, testing sequence sample QlAnd D, classifying, wherein K is a positive integer. Wherein, training set C ═ C (C)1,C2,...,CE) The construction of (A) comprises the following steps:
a acquiring pre-annotated respiratory waveform data as an original training data set S ═ S (S)1,S2,...,SM) Where M represents the number of sequences in the original training data set.
And b, preprocessing all sequence samples in the original training data set. Z-score normalization is performed on all labeled respiratory sequence samples, and the sequence S is normalizedm=(s1,s2,...,sn) M ∈ (1, 2.. multidot.m) normalization transformation intoThe formula is shown as follows:
where μ is the sequence mean, σ is the sequence standard deviation, m represents the sequence number, and i represents the number of sample points.
c, compressing the raw training data set data after preprocessing. The compression steps are as follows:
c1 first construct a template libraryInitial template set T1Then randomly selecting a sequence from the preprocessed original training data set as an initial template sequence, and placing the initial template sequence into a template set T1;
c2 selecting a non-initial template sequence from the preprocessed original training data setDTW distance calculation is carried out on the template sequence and each template sequence in each template set in the template library, and DTW distance calculation is obtainedAverage DTW distance from each template set;
c3 calculating the minimum average DTW distance mean _ D obtained in the step c2mWhen mean _ D is compared with a set threshold valuemWhen the sequence is less than or equal toAddition of DmIn the corresponding template set, judging whether the total number of template sequences in the template set reaches a set threshold lambda, if so, calculating an average template sequence of the template set by using a DBA algorithm, emptying all sequences in the template set, and adding the average template sequence to the template set; if mean _ Di> then create a new template set TlL represents the serial number of the template set and willIs added to Tl. Where is a real number and λ is a positive integer.
c4 repeats steps c2-c3 until the original training data set is traversed.
c5 match board libraryEach template set in (1) is compressed again by using the DBA algorithm, and finallyObtaining a compressed training set C ═ C1,C2,...,CE) And E denotes the size of the training set, i.e., the number of template sets.
Further, in the step a, the respiratory waveform data is pre-labeled by 5 doctors with lower seniority firstly, and then the labeled result of the first round is checked by 2 doctors with higher seniority to ensure the accuracy of the labeled result, and then all ineffective inspiratory effort waveforms are extracted, and the same number of non-ineffective inspiratory effort waveforms are randomly selected to form an experimental data set.
Further, the DTW distance is calculated by the following formula:
where X, Y represent two different sequence samples, X, Y represent sample points from the X, Y sequences, respectively, subscripts i, j represent sample point numbers, d (X)i,yj)=(xi-yj)2. D (i, j) represents the minimum accumulation distance; finally, the DTW distance of the two sequences, i.e., DTW (X, Y), is calculated by the following formula.
Further, the specific steps of compressing by using the DBA algorithm are as follows:
s1 extracting target template set Tk(Tk,1,Tk,2,...,Tk,H) Randomly selecting an initial sequence Tk,jAs an average sequence, where k is a template set number, k ∈ (1, 2.. E), j ∈ (1, 2.... H) represents a sequence number of the sequence in the template set;
s2 orderly selecting target template set Tk(Tk,1,Tk,2,...,Tk,H) Inner sequence Tk,i(non-initial sequence T)k,j) I ≠ j, i ∈ (1, 2.. multidot.h), based on the idea of dynamic programming, will Tk,iAnd Tk,jAligning;
s3 repeating step S2 until target template set Tk(Tk,1,Tk,2,...,Tk,H) Completing traversal;
s4 calculates the average sequence mean _ T of all sequences after alignmentk;
S5 ifMean _ T is usedkReplacing the ordered columns in the target dataset; otherwise, let Tk,j=mean_TkAnd repeatedly executing the steps S2-S5, wherein the execution times are increased by 1, and ξ is any real number.
S6 when the execution times exceeds the set threshold, the mean _ T obtained last timekReplacing all sequences in the target template set. Is any integer.
The invention has the following beneficial effects: similarity calculation is carried out on the respiration waveform by using a DBA-DTW-KNN-based method, the problems of difficulty in accurate detection of the expiratory starting point, uncertainty in threshold setting and time consumption of KNN classification are successfully solved, and a foundation is provided for realizing automatic and rapid identification of man-machine asynchronism.
Drawings
FIG. 1 is a schematic flow chart of the present invention.
Fig. 2(a) is a schematic diagram after preprocessing the original waveform, and (B) is a schematic diagram after the sequence is DTW aligned.
FIG. 3 is a flow chart of data set compression.
Fig. 4 is a schematic diagram of KNN classification, wherein triangles and squares represent different classes known and circles represent classes to be classified.
FIG. 5 is a training set generation time diagram under different generation template threshold conditions of the method of the present invention.
FIG. 6 is a test timing diagram of the optimal K value under different generated template threshold conditions of the method of the present invention.
Fig. 7 is a test time chart of the standard DTW _ KNN method under different K values.
FIG. 8 is a graph of the accuracy of the optimal K value under different conditions of template threshold generation by the method of the present invention.
Fig. 9 is a graph of accuracy of the standard DTW _ KNN method under different K values.
FIG. 10 is a F1 score plot of optimal K values under different generated template threshold conditions of the method of the present invention.
Fig. 11 is a F1 score plot for different K values of the standard DTW _ KNN method.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
The invention provides a mechanical ventilation man-machine asynchronous rapid identification method based on DBA-DTW-KNN, which is characterized by comprising the following steps: real-time reading respiratory waveform data to form test sequence Ql=(q1,q2,...,qp) And standardizing the respiratory waveform data; then calculating the test sequence QlAnd training set C ═ C (C)1,C2,...,CE) DTW distance of all sequences inside. Calculating similarity distance by adopting DTW (dynamic time warping), combining KNN (K-nearest neighbor) classification thought, and according to set K value, testing sequence sample QlAnd (6) classifying. Wherein, training set C ═ C (C)1,C2,...,CE) Obtained by compressing pre-annotated respiratory waveform data. Wherein the value of K is a positive integer.
Preferably, the acquired respiratory waveform data used to construct the training set are sample-equalized by performing a first round of pre-labeling by 5 physicians with lower seniority and then reviewing the labeling results of the first round by 2 physicians with higher seniority to ensure the accuracy of the labeling results, and then extracting all invalid inspiratory effort waveforms and randomly selecting the same number of non-invalid inspiratory effort waveforms. In this embodiment, a total of 12305 breaths were taken clinically for 17 patients, with 4689 of the futile inspiratory effort breath waveforms and 7616 of the non-futile inspiratory effort breath waveforms.
As shown in the flowchart of fig. 1, a training set construction process of a mechanical ventilation man-machine asynchronous fast recognition method based on DBA-DTW-KNN includes the following steps:
a extracting all ineffective inspiratory effort waveforms randomly divided into 10 equal parts and randomly picking the same number of non-ineffective inspiratory effortsThe force waveforms are grouped into groups such that the data sets are sample equalized. Wherein 9 are the original training set S ═ (S)1,S2,...,SM) And 1 part is test set Q ═ Q (Q)1,Q2,...,QR);
b, preprocessing all sequence samples. Z-score normalization is performed on all labeled respiratory sequence samples respectively, and the sequence S is usedm=(s1,s2,...,sn) M ∈ (1, 2.. multidot.M) to MFor example, the normalization formula is shown as:
where μ is the sequence mean, σ is the sequence standard deviation, m represents the sequence number, and i represents the number of sample points.
The Z-score normalization of the sequences was performed to avoid problems due to differences in baseline levels.
c, compressing the original training set data after preprocessing. As shown in fig. 3, the compression steps are as follows:
c1 first construct a template libraryInitial template set T1Then randomly selecting a sequence from the original training set after pretreatment as an initial template sequence, and placing the initial template sequence into a template set T1;
c2 selecting a sequence from the preprocessed raw training data set(non-initial template sequence), carrying out DTW distance calculation on the template sequence and each template sequence in each template set in the template library, and obtainingAverage DTW distance from each template set;
c3 distance mean _ D the smallest average DTWmWhen mean _ D is compared with a set threshold valuemWhen the sequence is less than or equal toAddition of DmIn the corresponding template set, judging whether the total number of template sequences in the template set exceeds a set threshold lambda, if so, calculating an average template sequence of the template set by using a DBA algorithm, and then adding the average template sequence to the template set; if mean _ Dm> then create a new template set TlL represents the serial number of the template set and willIs added to TlWherein the threshold value λ can be selected according to the actual situation, usually ∈ [3,35 ]],λ=[5,15]。
c4 repeats steps c2-c3 until the training data set is traversed.
c5 match board libraryEach template set in the training set is compressed again by using the DBA algorithm, and finally, a compressed training set C (C) is obtained1,C2,...,CE) And E denotes the size of the training set.
d calculating the test sequence QlAnd training set C ═ C (C)1,C2,...,CE) DTW distance of all sequences inside.
e, a KNN classification method based on DTW is adopted, the classification schematic diagram is shown in figure 4, and a test sequence sample Q is obtained according to a set K valuelAnd (6) classifying.
Specifically, the DTW distance is calculated by the following formula: taking DTW distance calculation in step d as an example, selecting a test sequence sample Q in the test setl=(q1,q2,...,qp) R, calculating Q ═ 1,2lAll samples of training setThe DTW distance D ═ D1,d2,...,dE). Wherein for two unequal time sequences Ql=(q1,q2,...,qp) 1,2, R and Ck(ck,1,ck,2,...,ck,r) E, DTW can find Q by dynamic programminglAnd CkThe best regular path therebetween, and then the minimum accumulated distance D (i, j) is obtained by the following formula.
Wherein the subscripts i, j denote the number of sampling points, d (q)i,ck,j)=(qi-ck,j)2. Finally, the DTW distance of the two sequences, namely DTW (Q), can be calculated by the following formula (7)l,Ck)。
In the method, the DBA algorithm is realized by the following steps:
s1 extracting target template set Tk(Tk,1,Tk,2,...,Tk,H) Randomly selecting an initial sequence Tk,jAs an average sequence, where k is a template set number, k ∈ (1, 2.. E), j ∈ (1, 2...., H) indicates a sequence number of the sequence in the template set;
s2 orderly selecting target template set Tk(Tk,1,Tk,2,...,Tk,H) Inner sequence Tk,i(non-initial sequence T)k,j) I ≠ j, i ∈ (1, 2.. multidot.h), based on the idea of dynamic programming, will Tk,iAnd Tk,jAligning;
s3 repeating step S2 until target template set Tk(Tk,1,Tk,2,...,Tk,H) Completing traversal;
s4 calculates the average sequence mean _ T of all sequences after alignmentk;
S5 ifMean _ T is usedkReplacing the ordered columns in the target dataset; otherwise, let Tk,j=mean_TkRepeating the steps 2-5, adding 1 to the execution times, wherein ξ is any real number, and usually ξ can be selected to be 0.001
S6 when the execution times exceeds the set threshold, the mean _ T obtained last timekReplacing all sequences in the target template set. Here, the integer is arbitrary, and in the present embodiment, the integer is set to 50.
In addition, different thresholds can be set, a plurality of training sets can be constructed, and the pre-marked test set can be adopted for preferential selection. In the present embodiment, K is set in steps of 2 from 1 to 15, the threshold is set in steps of 1 from 6 to 17, and the threshold λ is set to λ 15.
Through test set detection, the accuracy and the F1 score of the method are equivalent to those of the standard DTW _ KNN classification method, but the generation time and the test running time of the training set are obviously reduced, and the result is 1/8-1/2 of the test time of the standard DTW _ KNN classification method, and the experimental result is shown in a pair in fig. 5-11. The method can accurately and quickly judge the man-machine asynchrony of mechanical ventilation, and can be used for auxiliary teaching.
The IEE of mechanical ventilation man-machine asynchronism is rapidly and automatically identified through the DBA-DTW-KNN, and the problems that the accurate detection of the expiratory starting point is difficult, the threshold setting is uncertain and the KNN classification consumes time are solved.
Claims (4)
1. A mechanical ventilation man-machine asynchronous rapid identification method based on DBA-DTW-KNN is characterized by comprising the following steps: real-time reading respiratory waveform data to form test sequence Ql=(q1,q2,...,qp) And standardizing the respiratory waveform data; then calculating the test sequence QlAnd training set C ═ C (C)1,C2,...,CE) DTW distance of all sequences in (a). Calculating similarity distance by adopting DTW (dynamic time warping), combining KNN (K-nearest neighbor) classification thought, and according to set K value, testing sequence sample QlIs classified, K is a positive integer. Wherein, training set C ═ C (C)1,C2,...,CE) The construction of (A) comprises the following steps:
a acquiring pre-annotated respiratory waveform data as an original training data set S ═ S (S)1,S2,...,SM) Where M represents the size of the original training data set.
And b, preprocessing all sequence samples in the original training data set. Z-score normalization is performed on all labeled respiratory sequence samples, and the sequence S is normalizedm=(s1,s2,...,sn) M ∈ (1, 2.. multidot.m) normalization transformation intoThe formula is shown as follows:
where μ is the sequence mean, σ is the sequence standard deviation, m represents the sequence number, and i represents the number of sample points.
c, compressing the raw training data set data after preprocessing. The compression steps are as follows:
c1 first construct a template libraryInitial template set T1Then randomly selecting a sequence from the preprocessed original training data set as an initial template sequence, and placing the initial template sequence into a template set T1;
c2 selecting a non-initial template sequence from the preprocessed original training data setDTW distance calculation is carried out on the template sequence and each template sequence in each template set in the template library, and DTW distance calculation is obtainedAverage DTW distance from each template set;
c3 calculating the minimum average DTW distance mean _ D obtained in the step c2mWhen mean _ D is compared with a set threshold valuemWhen the sequence is less than or equal toAddition of DmIn the corresponding template set, judging whether the total number of template sequences in the template set reaches a set threshold lambda, if so, calculating an average template sequence of the template set by using a DBA algorithm, emptying all sequences in the template set, and adding the average template sequence to the template set; if mean _ Dm> then create a new template set TlL represents the serial number of the template set and willIs added to Tl. Where is a real number and λ is a positive integer.
c4 repeats steps c2-c3 until the original training data set is traversed.
2. The DBA-DTW-KNN-based mechanical ventilation man-machine dyssynchrony rapid identification method according to claim 1, wherein in the step a, the respiratory waveform data is pre-labeled by 5 physicians with lower seniority for the first round, and then the labeled result of the first round is reviewed by 2 physicians with higher seniority for ensuring the accuracy of the labeled result, and then all invalid inspiratory effort waveforms are extracted, and the same number of non-invalid inspiratory effort waveforms are randomly selected to form the original training data set.
3. The DBA-DTW-KNN based mechanical ventilation dyssynchrony rapid identification method according to claim 1, wherein the DTW distance is calculated by the following formula:
where X, Y represent two different sequence samples, X, Y represent sample points from the X, Y sequences, respectively, subscripts i, j represent sample point numbers, d (X)i,yj)=(xi-yj)2. D (i, j) represents the minimum accumulation distance; finally, the DTW distance of the two sequences, i.e., DTW (X, Y), is calculated by the following formula.
4. The mechanical ventilation man-machine asynchronous rapid identification method based on DBA-DTW-KNN as claimed in claim 1, wherein the concrete steps of adopting DBA algorithm compression are as follows:
s1 extracting target template set Tk(Tk,1,Tk,2,...,Tk,H) Randomly selecting an initial sequence Tk,jAs an average sequence, where k is a template set number, k ∈ (1, 2.. E), j ∈ (1, 2.... H) represents a sequence number of the sequence in the template set;
s2 orderly selecting target template set Tk(Tk,1,Tk,2,...,Tk,H) Inner sequence Tk,iI ≠ j, i ∈ (1, 2.. multidot.h), based on the idea of dynamic programming, will Tk,iAnd Tk,jAligning;
s3 repeating step S2 until target template set Tk(Tk,1,Tk,2,...,Tk,H) Completing traversal;
s4 calculates the average sequence mean _ T of all sequences after alignmentk;
S5 ifMean _ T is usedkReplacing all sequences in the target dataset; otherwise, let Tk,j=mean_TkAnd repeatedly executing the steps S2-S5, wherein the execution times are increased by 1, and ξ is any real number.
S6 when the execution times exceeds the set threshold, the mean _ T obtained last timekReplacing all sequences in the target template set. Is any integer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010474275.4A CN111738305B (en) | 2020-05-29 | 2020-05-29 | Mechanical ventilation man-machine asynchronous rapid identification method based on DBA-DTW-KNN |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010474275.4A CN111738305B (en) | 2020-05-29 | 2020-05-29 | Mechanical ventilation man-machine asynchronous rapid identification method based on DBA-DTW-KNN |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111738305A true CN111738305A (en) | 2020-10-02 |
CN111738305B CN111738305B (en) | 2022-06-24 |
Family
ID=72646597
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010474275.4A Active CN111738305B (en) | 2020-05-29 | 2020-05-29 | Mechanical ventilation man-machine asynchronous rapid identification method based on DBA-DTW-KNN |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111738305B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112906768A (en) * | 2021-02-04 | 2021-06-04 | 南京邮电大学 | Cross-scene pre-selection method for wireless behavior perception model |
CN113539398A (en) * | 2021-06-25 | 2021-10-22 | 中国科学院深圳先进技术研究院 | Breathing machine man-machine asynchronous classification method, system, terminal and storage medium |
CN113539501A (en) * | 2021-06-25 | 2021-10-22 | 中国科学院深圳先进技术研究院 | Breathing machine man-machine asynchronous classification method, system, terminal and storage medium |
CN113642682A (en) * | 2021-10-14 | 2021-11-12 | 北京理工大学 | Trajectory primitive extraction and analysis method and system under multi-vehicle interaction environment |
WO2023102820A1 (en) * | 2021-12-09 | 2023-06-15 | 深圳迈瑞生物医疗电子股份有限公司 | Medical device and ventilation state recognition method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109893732A (en) * | 2019-02-28 | 2019-06-18 | 杭州智瑞思科技有限公司 | A kind of mechanical ventilation patient-ventilator asynchrony detection method based on Recognition with Recurrent Neural Network |
CN110163130A (en) * | 2019-05-08 | 2019-08-23 | 清华大学 | A kind of random forest grader and classification method of the feature pre-align for gesture identification |
US20200149998A1 (en) * | 2018-11-12 | 2020-05-14 | Kabushiki Kaisha Toshiba | Method of detecting anomalies in waveforms, and system thereof |
-
2020
- 2020-05-29 CN CN202010474275.4A patent/CN111738305B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200149998A1 (en) * | 2018-11-12 | 2020-05-14 | Kabushiki Kaisha Toshiba | Method of detecting anomalies in waveforms, and system thereof |
CN109893732A (en) * | 2019-02-28 | 2019-06-18 | 杭州智瑞思科技有限公司 | A kind of mechanical ventilation patient-ventilator asynchrony detection method based on Recognition with Recurrent Neural Network |
CN110163130A (en) * | 2019-05-08 | 2019-08-23 | 清华大学 | A kind of random forest grader and classification method of the feature pre-align for gesture identification |
Non-Patent Citations (3)
Title |
---|
XIANGLEI ZHU 等: "《Research on Driving Behaviors Based on Machine Learning Algorithms》", 《2020 IEEE 4TH INFORMATION TECHNOLOGY, NETWORKING, ELECTRONIC AND AUTOMATION CONTROL CONFERENCE 》 * |
傅致远 等: "《基于DTW和DBA算法的CNC刀具破损在线检测》", 《组合机床与自动化加工技术》 * |
王鑫 等: "《辅助机械通气时吸气起始点判断方法的比较》", 《临床荟萃》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112906768A (en) * | 2021-02-04 | 2021-06-04 | 南京邮电大学 | Cross-scene pre-selection method for wireless behavior perception model |
CN113539398A (en) * | 2021-06-25 | 2021-10-22 | 中国科学院深圳先进技术研究院 | Breathing machine man-machine asynchronous classification method, system, terminal and storage medium |
CN113539501A (en) * | 2021-06-25 | 2021-10-22 | 中国科学院深圳先进技术研究院 | Breathing machine man-machine asynchronous classification method, system, terminal and storage medium |
CN113642682A (en) * | 2021-10-14 | 2021-11-12 | 北京理工大学 | Trajectory primitive extraction and analysis method and system under multi-vehicle interaction environment |
WO2023102820A1 (en) * | 2021-12-09 | 2023-06-15 | 深圳迈瑞生物医疗电子股份有限公司 | Medical device and ventilation state recognition method |
Also Published As
Publication number | Publication date |
---|---|
CN111738305B (en) | 2022-06-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111738305B (en) | Mechanical ventilation man-machine asynchronous rapid identification method based on DBA-DTW-KNN | |
EP3706845B1 (en) | Clinical decision support system for patient-ventilator asynchrony detection and management | |
RU2515401C2 (en) | System and method of identifying respiratory insufficiency of subject's respiration | |
CN110533658A (en) | Intelligent pulmonary emphysema diagnostic message processing system and method, information data processing terminal | |
WO2022242123A1 (en) | Mechanical ventilation man-machine asynchronous detection method and apparatus, and computer-readable storage medium | |
CN113539501A (en) | Breathing machine man-machine asynchronous classification method, system, terminal and storage medium | |
WO2022267382A1 (en) | Human-machine asynchronous classification method and system for ventilator, terminal and storage medium | |
Herry et al. | Segmentation and classification of capnograms: application in respiratory variability analysis | |
Chen et al. | Cough detection using selected informative features from audio signals | |
CN111227792B (en) | Apnea detection method and system, electronic device and storage medium | |
Tahir et al. | A CNN-ELM-based method for ballistocardiogram classification in a clinical environment | |
CN111227793A (en) | Apnea recognition method and system, electronic equipment and storage medium | |
CN115814222B (en) | Man-machine asynchronous waveform identification method under hybrid mechanical ventilation mode and related equipment | |
CN114842978B (en) | Intelligent blood gas analysis detection system and method based on medical big data | |
CN115273176A (en) | Pain multi-algorithm objective assessment method based on vital signs and expressions | |
Garde et al. | Patients on weaning trials classified with support vector machines | |
Sottile et al. | The development, optimization, and validation of four different machine learning algorithms to identify ventilator dyssynchrony | |
Arizmendi et al. | Data mining of patients on weaning trials from mechanical ventilation using cluster analysis and neural networks | |
CN114288500B (en) | Man-machine ventilation asynchronous detection model and device based on fuzzy entropy feature extraction | |
Asaithambi et al. | Clustering based analysis of spirometric data using principal component analysis and self organizing map | |
Tataw et al. | Quantitative analysis of live-cell growth at the shoot apex of Arabidopsis thaliana: algorithms for feature measurement and temporal alignment | |
CN111276243B (en) | Multi-variable classification system and method based on biomarker | |
KR102678040B1 (en) | Apparatus and method of classifying cement dust-exposed subject based on machine learning for computed tomography images | |
Swapna et al. | An improved machine learning approach to classify sleep stages and apnea events | |
Dongyao et al. | A method of detecting human body falling action in a complex background |
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 |