CN114451879B - Intelligent heart rate variability analysis system - Google Patents
Intelligent heart rate variability analysis system Download PDFInfo
- Publication number
- CN114451879B CN114451879B CN202210253155.0A CN202210253155A CN114451879B CN 114451879 B CN114451879 B CN 114451879B CN 202210253155 A CN202210253155 A CN 202210253155A CN 114451879 B CN114451879 B CN 114451879B
- Authority
- CN
- China
- Prior art keywords
- module
- heart rate
- unit
- rate variability
- training
- 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.)
- Active
Links
- 238000004458 analytical method Methods 0.000 title claims abstract description 100
- 238000012937 correction Methods 0.000 claims abstract description 66
- 238000007781 pre-processing Methods 0.000 claims abstract description 20
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 12
- 238000012545 processing Methods 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 71
- 238000013528 artificial neural network Methods 0.000 claims description 33
- 238000000034 method Methods 0.000 claims description 31
- 230000002452 interceptive effect Effects 0.000 claims description 18
- 238000005457 optimization Methods 0.000 claims description 15
- 238000003062 neural network model Methods 0.000 claims description 12
- 238000012216 screening Methods 0.000 claims description 11
- 238000001914 filtration Methods 0.000 claims description 10
- 238000013450 outlier detection Methods 0.000 claims description 10
- 238000011156 evaluation Methods 0.000 claims description 9
- 238000010183 spectrum analysis Methods 0.000 claims description 7
- 230000000747 cardiac effect Effects 0.000 claims description 6
- 238000000718 qrs complex Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 238000000638 solvent extraction Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 10
- 238000006243 chemical reaction Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 230000000007 visual effect Effects 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 3
- 206010015856 Extrasystoles Diseases 0.000 description 2
- 208000010496 Heart Arrest Diseases 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 239000000523 sample Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 208000024172 Cardiovascular disease Diseases 0.000 description 1
- 208000032131 Diabetic Neuropathies Diseases 0.000 description 1
- 206010019280 Heart failures Diseases 0.000 description 1
- 206010000891 acute myocardial infarction Diseases 0.000 description 1
- 210000003403 autonomic nervous system Anatomy 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001447 compensatory effect Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000013079 data visualisation Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 208000010125 myocardial infarction Diseases 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 238000012502 risk assessment Methods 0.000 description 1
- 230000036578 sleeping time Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000002889 sympathetic effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/024—Detecting, measuring or recording pulse rate or heart rate
- A61B5/02405—Determining heart rate variability
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/024—Detecting, measuring or recording pulse rate or heart rate
- A61B5/0245—Detecting, measuring or recording pulse rate or heart rate by using sensing means generating electric signals, i.e. ECG signals
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
- A61B5/346—Analysis of electrocardiograms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
- A61B5/346—Analysis of electrocardiograms
- A61B5/349—Detecting specific parameters of the electrocardiograph cycle
- A61B5/352—Detecting R peaks, e.g. for synchronising diagnostic apparatus; Estimating R-R interval
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
- A61B5/346—Analysis of electrocardiograms
- A61B5/349—Detecting specific parameters of the electrocardiograph cycle
- A61B5/366—Detecting abnormal QRS complex, e.g. widening
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
Abstract
The application discloses intelligent heart rate variability analysis system, this system includes: the system comprises a signal acquisition module, a preprocessing module, a correction module and a report output module; the signal acquisition module is used for acquiring an electrocardiogram digital signal; the preprocessing module is used for correcting the electrocardiogram digital signals to obtain electrocardiogram data; the correction module is used for processing the electrocardiogram data to obtain an NN interval sequence; and the report output module is used for obtaining a heart rate variability analysis report according to the NN interval sequence. The invention realizes the full utilization of the existing medical resources, can automatically improve the analysis algorithm of heart rate variability, improves the accuracy of the automatic analysis result of heart rate variability, and can meet the use requirements of multiple scenes and individuation of users.
Description
Technical Field
The invention relates to the technical field of signal processing, in particular to an intelligent heart rate variability analysis system.
Background
Heart rate variability refers to subtle temporal changes between successive cardiac cycles and its regularity, which are important indicators for evaluating sympathetic-complex sympathetic tone of the autonomic nervous system and its balance. Since the first acceptance of the clinical relevance of heart rate variability, subsequent studies have demonstrated that heart rate variability has important clinical value in the risk assessment of cardiovascular disease, such as heart failure, stroke, post-myocardial infarction, post-cardiac transplant, etc. The medical community has reached a general consensus: the heart rate variability can be used as a risk prediction index of acute myocardial infarction and an early warning index of diabetic neuropathy.
Currently, the industry standard analysis method of heart rate variability is: firstly, acquiring electrocardiosignals of a subject, and completing digital-analog conversion and artifact identification; secondly, identifying a QRS complex reference point in the electrocardiosignal to obtain an RR interval sequence and editing to obtain a sinus NN interval; and thirdly, carrying out time domain analysis on the NN intervals to obtain a correlation analysis result of heart rate variability. In analysis of heart rate variability, since time-domain and frequency-domain analysis is susceptible to inaccurate recognition of NN interval sequences, automated analysis results of heart rate variability are not accurate, requiring a clinician to spend a lot of time in RR interval editing phase to reject unacceptable RR intervals by visual inspection and manual correction. The existing automatic heart rate variability analysis system does not fully utilize the historical information such as accurate NN intervals, heart beat types and the like obtained by editing and correcting by the existing clinician, has the problems of low intelligent degree and low analysis efficiency, and cannot provide personalized intelligent service according to the use requirements of users.
Therefore, the main problem of the existing heart rate variability analysis system is that the heart rate variability automatic analysis system is low in intelligent degree, medical history data are not fully utilized, analysis results are inaccurate, and the use requirements of multiple scenes cannot be met. It is needed to design a heart rate variability analysis system, so as to fully utilize the existing medical resources, automatically improve the analysis algorithm of heart rate variability, improve the accuracy of the automatic analysis result, and meet the use requirements of multiple scenes and individuation of users.
Disclosure of Invention
In view of the foregoing, it is necessary to provide an intelligent heart rate variability analysis system for solving the problems of low intelligentization degree, inaccurate analysis result and inability to meet the personalized use demands of users in the heart rate variability analysis system in the prior art.
In order to solve the above problems, the present invention provides an intelligent heart rate variability analysis system, comprising:
the system comprises a signal acquisition module, a preprocessing module, a correction module and a report output module;
the signal acquisition module is used for acquiring an electrocardiogram digital signal;
the preprocessing module is used for preprocessing the electrocardiogram digital signals to obtain electrocardiogram data;
the correction module is used for analyzing the electrocardiogram data to obtain an NN interval sequence;
and the report output module is used for obtaining a heart rate variability analysis report according to the NN interval sequence.
Further, the preprocessing module includes: an outlier detection unit, a filter unit and an identification unit;
an outlier detection unit for determining abnormal data in the electrocardiogram digital signal;
a filter unit for filtering out artifacts in the electrocardiogram digital signal;
a QRS recognition unit for recognizing a QRS complex fiducial point in the electrocardiogram digital signal.
Further, the correction module comprises an automatic correction unit, a manual correction unit, a storage unit, a neural network training unit and a prediction unit;
an automatic correction unit for eliminating non-sinus cardiac RR intervals of the electrocardiographic data to obtain a first NN interval sequence;
the manual correction unit is used for providing an interactive interface for a user, so that the user can manually correct and mark the R wave position and type of the first NN interval sequence and the RR interval sequence;
the storage unit is used for storing the electrocardiogram data after the manual correction and the marking of the R wave position and type and the RR interval sequence;
the neural network training unit is used for establishing an initial correction model, and training the initial correction model by using the electrocardiogram data in the storage unit to obtain a complete correction model;
and the prediction unit is used for carrying out QRS recognition and RR interval automatic correction on the electrocardiogram data by utilizing the trained complete correction model.
Further, the automatic correction unit comprises a screening module and a calculation module;
the screening module is used for screening RR intervals meeting the first judgment condition in the RR interval sequence of the electrocardiogram data and judging whether the RR intervals meet the second judgment condition or not; if the second judging condition is met, recording the position of the RR interval in the RR interval sequence, and removing the RR interval from the RR interval sequence to obtain a first RR interval sequence; calculating the heart rate according to the first RR interval sequence, and eliminating RR intervals with the heart rate meeting a third judging condition to obtain a second RR interval sequence;
and the calculation module is used for processing the second RR interval sequence by using a preset algorithm to obtain an NN interval sequence.
Further, the neural network training unit comprises a selection module, an interactive training module and a self-training module;
the selection module is used for providing an interface for a user to enable the user to select an interactive training mode and a self-training mode;
the interactive training module is used for enabling a user to autonomously select an initial correction model and a neural network training method to train the neural network;
and the self-training module is used for training the neural network according to a preset neural network and training method.
Further, the interactive training module comprises a setting module for enabling a user to set the division ratio of the data set, select the neural network model and the super-parameter tuning method.
Further, the self-training module comprises a network searching module, an architecture optimizing module and a performance evaluating module,
the network searching module is used for storing a basic neural network model structure;
the architecture optimization module is used for performing super-parameter setting on the basic neural network model by using a preset architecture optimization method to obtain a candidate correction model;
and the performance evaluation module is used for performing performance evaluation on the candidate correction models with complete training, and selecting the heart rate analysis model with the best performance as the heart rate analysis model with complete training.
Further, the network search module comprises a basic operation unit, and the basic operation unit forms the basic neural network model structure.
Further, the architecture optimization method comprises a gradient descent-based micro-network searchable method.
Further, the report output module comprises a time domain analysis unit, a frequency domain analysis unit and a nonlinear analysis unit;
the time domain analysis unit is used for carrying out long-range and short-range heart rate variability time domain analysis according to the NN interval sequence to obtain a heart rate variability time domain analysis report;
the frequency domain analysis unit is used for carrying out long-range and short-range spectrum analysis according to the NN interval sequence to obtain a heart rate variability frequency domain analysis report;
and the nonlinear analysis unit is used for calculating nonlinear parameters of the NN interval sequence to obtain a heart rate variability nonlinear parameter analysis report.
Compared with the prior art, the invention has the beneficial effects that: performing digital-to-analog conversion on an electrocardiogram original signal through a signal acquisition module to obtain an electrocardiogram digital signal; preprocessing such as filtering is carried out on the electrocardiogram digital signals through a preprocessing module, and a data base is provided for subsequent heart rate variability analysis; the correction module fully utilizes the existing medical resources, so that the accuracy of automatic identification and correction of an analysis system is greatly improved, personalized function selection can be provided for a user in the use process, and the multi-scene and personalized use requirements of the user are met; and the heart rate variability is analyzed from multiple angles through the report output module, so that the analysis result is more comprehensive and accurate. The invention realizes the full utilization of the existing medical resources, can automatically improve the analysis algorithm of heart rate variability, improves the accuracy of the automatic analysis result of heart rate variability, and can meet the multi-scene and personalized use requirements of users.
Drawings
Fig. 1 is a schematic structural diagram of an embodiment of an intelligent heart rate variability analysis system according to the present invention;
FIG. 2 is a schematic diagram of a preprocessing module according to an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating a calibration module according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a workflow of an embodiment of a neural network unit according to the present invention;
FIG. 5 is a schematic diagram of a self-training module according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of another embodiment of a self-training module according to the present invention;
FIG. 7 is a schematic diagram illustrating a network search module according to an embodiment of the present invention;
FIG. 8 is a schematic diagram illustrating a report output module according to an embodiment of the present invention;
fig. 9 is a schematic diagram of a result obtained by performing spectrum analysis on the open source database nsr001 according to an embodiment of the frequency domain analysis unit provided by the present invention;
fig. 10 shows the corresponding poincare result of an embodiment of the nonlinear analysis unit according to the present invention after analyzing the data in the open source database nsr 001.
Detailed Description
Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form a part hereof, and together with the description serve to explain the principles of the invention, and are not intended to limit the scope of the invention.
The invention provides an intelligent heart rate variability analysis system, which is described in detail below.
As shown in fig. 1, an embodiment of the present invention provides an intelligent heart rate variability analysis system 100, which includes a signal acquisition module 101, a preprocessing module 102, a correction module 103, and a report output module 104;
the signal acquisition module 101 is used for acquiring an electrocardiogram digital signal;
the preprocessing module 102 is configured to preprocess the electrocardiogram digital signal to obtain electrocardiogram data;
the correction module 103 is configured to analyze the electrocardiographic data to obtain an NN interval sequence;
the report output module 104 is configured to obtain a heart rate variability analysis report according to the NN interval sequence.
Compared with the prior art, the intelligent heart rate variability analysis system provided by the embodiment carries out digital-to-analog conversion on an electrocardiogram original signal through the signal acquisition module to obtain an electrocardiogram digital signal; preprocessing such as filtering is carried out on the electrocardiogram digital signals through a preprocessing module, and a data base is provided for subsequent heart rate variability analysis; the correction module fully utilizes the existing medical resources, so that the accuracy of automatic identification and correction of an analysis system is greatly improved, personalized function selection can be provided for a user in the use process, and the multi-scene and personalized use requirements of the user are met; and the heart rate variability is analyzed from multiple angles through the report output module, so that the analysis result is more comprehensive and accurate.
As a specific embodiment, the signal acquisition module 101 includes a dynamic electrocardiograph, and is mainly configured to acquire an original electrocardiogram signal, and obtain an electrocardiogram digital signal after digital-to-analog conversion.
As a preferred embodiment, as shown in fig. 2, the preprocessing module 102 includes: an outlier detection unit 201, a filter unit 202, and an identification unit 203;
the outlier detection unit 201 is configured to determine abnormal data in the electrocardiogram digital signal;
the filter unit 202 is configured to filter out artifacts in the electrocardiogram digital signal;
the QRS recognition unit 203 is configured to recognize a QRS complex reference point in the electrocardiogram digital signal.
As a specific embodiment, the outlier detection unit 201 is mainly configured to find points inconsistent with other characteristics of the electrocardiographic sample points. Since the electrocardiographic data is one-dimensional data, the outlier detection unit 201 employs an outlier detection based on a statistical method in consideration of the arithmetic performance. Specifically, the process of outlier detection is: firstly, respectively calculating the mean value and the variance of all electrocardiographic samples, and then taking the value exceeding the range of the interval [ u-3 sigma, u+3 sigma ] as an outlier; where u represents the mean value of the electrocardiogram samples and σ represents the standard deviation of the electrocardiogram samples.
As a specific embodiment, the filter unit 202 is mainly used for filtering out various artifacts in the electrocardiogram. The specific treatment process comprises the following steps: and filtering baseline drift by using an FIR high-pass filter, filtering power frequency interference by using a Notch filter, and filtering high-frequency myoelectricity interference by using an IIR low-pass filter.
As a preferred embodiment, as shown in fig. 3, the correction module 103 includes an automatic correction unit 301, a manual correction unit 302, a storage unit 303, a neural network training unit 304, and a prediction unit 305;
the automatic correction unit 301 is configured to eliminate non-sinus cardiac RR intervals of the electrocardiographic data, to obtain a first NN interval sequence;
the manual correction unit 302 is configured to provide an interactive interface for a user, so that the user manually corrects and marks the R-wave position and type of the first NN interval sequence and the RR interval sequence;
the storage unit 303 is configured to store the electrocardiographic data after the manual correction and labeling of R-wave positions and types and RR interval sequences;
the neural network training unit 304 is configured to establish an initial correction model, and train the initial correction model by using the electrocardiogram data in the storage unit to obtain a complete correction model;
the prediction unit 305 is configured to perform QRS recognition and RR interval automatic correction on the electrocardiographic data by using the trained and complete correction model.
As a preferred embodiment, as shown in fig. 3, the automatic correction unit 301 includes a filtering module 3011 and a calculating module 3012;
the screening module 3011 is configured to screen an RR interval in the RR interval sequence of the electrocardiographic data, where the RR interval meets a first judgment condition, and judge whether the RR interval meets a second judgment condition; if the second judging condition is met, recording the position of the RR interval in the RR interval sequence, and removing the RR interval from the RR interval sequence to obtain a first RR interval sequence; calculating the heart rate according to the first RR interval sequence, and eliminating RR intervals with the heart rate meeting a third judging condition to obtain a second RR interval sequence;
the calculating module 3012 is configured to process the second RR interval sequence by using a preset algorithm to obtain a first NN interval sequence.
In general, sources of non-sinus cardiac signals mixed into the sinus RR interval sequence (also called NN interval sequence) are mainly: (a) The RR intervals produced by ectopic beats and the long RR intervals that follow it, such as premature heart beats and the compensatory intervals that follow it; (b) a long RR interval due to QRS complex missed detection; (c) False QRS due to QRS complex multi-detection may cause an otherwise normal sinus cardiac RR interval to be split into two short RR intervals, thereby producing a false RR interval sequence. In the scheme, three times of judgment are performed through the screening module 3011, and data processing is performed through the calculating module 3012, so that the detection accuracy of NN interval sequences is greatly improved.
As a specific embodiment, the first judgment condition is: RR interval >2s; the second judgment condition is as follows: cardiac arrest conditions; the third judgment condition is: RR intervals with heart rate greater than 200bpm or less than 30 bpm. Under the above preset conditions, the data correction process of the screening module 3011 is: firstly, screening data with RR interval more than 2s in electrocardiogram data, and judging whether the condition of cardiac arrest is met; if yes, recording the position of the data meeting the conditions, and eliminating the RR interval from the RR sequence; otherwise, no processing is performed. Secondly, calculating heart rate according to the RR intervals, and eliminating RR intervals with the heart rate more than 200bpm or less than 30 bpm.
The preset algorithm of the calculation module 3012 is linear interpolation or cubic spline interpolation; the RR intervals subjected to the reject operation by the screening module 3011 are interpolated to obtain sinus heart beat RR intervals for heart rate variability analysis, which are denoted as NN interval sequences.
As a preferred embodiment, as shown in fig. 3, the neural network training unit 304 includes a selection module 3041, an interactive training module 3042, and a self-training module 3043;
the selection module 3041 is configured to provide an interface for a user, so that the user selects an interactive training mode and a self-training mode;
the interactive training module 3042 is configured to enable a user to autonomously select an initial correction model and a neural network training method to perform neural network training;
the self-training module 3043 is configured to perform neural network training according to a preset neural network and a training method.
As a specific embodiment, the selection module 3041 may allow the user to autonomously select whether to turn on the neural network training mode. If the user chooses to start the neural network self-training, the system will first count the number of available label data in the designated storage unit, and to ensure the effectiveness of the neural network training, the lowest threshold of available label data is set to be 1 ten thousand. If the data volume is too small, prompting the user that the function cannot be started; if the data volume meets the requirement, prompting the user to select to start the network interaction training module or the self-training module (the self-training module is started by default to perform neural network training). When a user has scientific research requirements, the user may need to independently design a training method experiment, such as setting different models, deriving a related chart and analyzing results, and at this time, the interactive training module may be selectively started.
As a preferred embodiment, the interactive training module 3042 includes a setting module for enabling a user to set a data set division ratio, select a neural network model, and a super-parameter tuning method.
The neural network training unit 304 is described in detail below in conjunction with fig. 4.
As a specific embodiment, in the case where the interactive training module 3042 is turned on, the training process is an interactive training mode, specifically:
step S1: generating a visual probe analysis report for existing data, comprising:
(a) Descriptive statistics of the data, including conventional mean, variance, median, quantile, etc.;
(b) Data visualization charts including histograms of fields such as QRS wave position, QRS type, RR interval (visual data distribution characteristics), scatter plots (correlation between visual features), difference plots (describing differences in distribution and statistical information among multiple datasets), box plots (visual population differences), etc.;
step S2: setting a data set dividing ratio; there are two partitioning options: training set-test set division, training set-verification set-test set;
step S3: selecting a model in a preset model library; the model library provides common models such as a support vector machine, logistic regression, random forest, CNN, FCN, resNet, unet and the like for a user to select;
step S4: selecting an automatic super-parameter tuning algorithm; common algorithms such as Grid search, naive bayes evolution and the like are provided for users to select.
As a preferred embodiment, as shown in fig. 5, the self-training module 3043 includes a web search module 501, an architecture optimization module 502, and a performance evaluation module 503;
the network searching module 501 is configured to store a basic neural network model structure;
the architecture optimization module 502 performs super-parameter setting on the basic neural network model by using a preset architecture optimization method to obtain a candidate correction model;
the performance evaluation module 503 is configured to perform performance evaluation on the candidate calibration model with complete training, and select the calibration model with optimal performance as the candidate calibration model with complete training.
As a specific embodiment, as shown in fig. 6, the network searching module 501 further defines an algorithm for searching the type of the neural network by the user. Meanwhile, in order to improve the searching efficiency, the description mode of the neural network structure is also customized. In this embodiment, we define the neural network structure in the network search module 501 as a CNN and its variant network, FCN and its variant network, resNet and its variant network, and a network of the Unet series.
The architecture optimization method of the architecture optimization module 502 adopts a grid and random search mode by default, and the method enables a program to automatically run all parameters once through presetting a reasonable super parameter value range to obtain different performance index results. And finally, selecting a group of parameters with optimal performance indexes as the values of final super parameters. The architecture optimization module 502 can be utilized to model each of the network search modules 501: and performing super-parameter setting on the model 1, the model 2 and the … model n to obtain n candidate correction models.
The performance evaluation module 503 uses the F1 score as a performance evaluation index after each candidate correction model is trained, and selects one model with the best performance from the models 1, 2 and … n as a final training complete correction model.
As a preferred embodiment, the network search module 501 includes a basic operation unit, and the basic operation unit forms the basic neural network model structure.
As a specific embodiment, the network searching module 501 may be a hierarchical structure, that is, the unit structure generated in the previous step is used as a basic component of the unit structure in the next step, and the final network structure is obtained through iteration. As shown in fig. 7, fig. 7 is a schematic view of a three-layered structure. Combining the basic operation units of the first layer as basic operation units of the second layer; and combining the main operation units of the second layer to form a basic operation unit of the third layer.
As a preferred embodiment, the architecture optimization method comprises a gradient descent-based micro-network searchable method.
As a specific example, DARTS (Differentiable Architecture Search) is chosen as the architecture optimization method, which is a microminiatable grid search method based on a gradient descent method that uses softmax to relax the discrete space to search for neural structures in the continuous and microminiatable search space.
As a specific embodiment, the prediction unit 305 is started when the user starts the neural network self-training module and the training of the neural network is completed. After the prediction unit 305 is started, when the user subsequently re-uses the heart rate variability analysis function, the prediction unit 305 will be used for QRS identification and automatic correction of RR intervals.
As a preferred embodiment, as shown in fig. 8, the report output module 104 includes a time domain analysis unit 1041, a frequency domain analysis unit 1042, and a nonlinear analysis unit 1043;
the time domain analysis unit 1041 is configured to perform long-range and short-range time domain analysis of heart rate variability according to the NN interval sequence, so as to obtain a time domain analysis report of heart rate variability;
the frequency domain analysis unit 1042 is configured to perform long-range and short-range spectrum analysis according to the NN interval sequence to obtain a heart rate variability frequency domain analysis report;
the nonlinear analysis unit 1043 is configured to calculate nonlinear parameters of the NN interval sequence, and obtain a heart rate variability nonlinear parameter analysis report.
As a specific embodiment, the time domain analysis of the time domain analysis unit 1041 is used for short-range (e.g. 5 mins) heart rate variability detection and analysis, more mainly for long-range (e.g. 24 h) heart rate variability detection and analysis. Comprising the following steps:
(1) NN interval histogram: and counting NN interval distribution map in a certain time. Specifically, the number of heart beats in different NN intervals is counted at a predetermined NN interval (for example, 1/128s, i.e., 7.8124 ms). The abscissa is the length of the NN interval and the ordinate is the number of beats.
(2) NN interval difference histogram: distribution of differences between adjacent NN intervals. Typically, the NN interval of the previous heart beat is subtracted from the NN interval of the next heart beat. The abscissa is NN interval difference, the unit is ms, and the ordinate is the heart beat number.
(3) Day-night average NN interval difference (ms): the NN interval recorded in 24 hours is divided into two sections according to the daytime and the night (sleeping time), the average value of the NN intervals at all night is calculated, and the average value of the NN intervals at all the daytime is subtracted to obtain the day and night average NN interval difference. Since the heart rate is slow during night sleep and the NN interval is longer than during the day, the average NN interval difference between day and night is generally positive.
(4) NN interval Standard Deviation (SDNN): the standard deviation of all NN intervals within a preset time period is detected in ms.
(5) Standard Deviation (SDANN) of average of NN intervals: the unit is ms, NN interval data measured in 24h are divided into a plurality of sections (24 h and 288 sections in total) in a time sequence of taking every 5min as one section, the average value of NN intervals in each 5min time section is calculated, the average value of 288 NN intervals can be obtained, and the standard deviation of the 288 data is calculated.
(6) Root mean square (RMSSD) of adjacent NN interval differences: in ms. Here, adjacent NN intervals refer to two heartbeats being adjacent and both heartbeats meeting the criteria for analyzing raw data as heart rate variability. For example, where both heartbeats are sinus rhythms and there are no ectopic beats, missed heartbeats, etc., then RMSSD may be expressed as:
wherein DeltaNN i The difference between two adjacent NN intervals is shown, and N pairs are shared by the adjacent NN intervals in 24h.
(7) Average of NN interval standard deviation (sdn index): the NN intervals obtained by long-range detection are divided into one section every 5 minutes in time sequence in units of ms, the standard deviation of the NN intervals in each time section is calculated first, and then the average value of the standard deviations is calculated.
(8) Standard Deviation (SDSD) of adjacent NN interval differences: in ms, the difference between all adjacent NN intervals is calculated first, and then the standard deviation of the difference data is calculated.
(9) Heart beat number (NN 50) with a difference between adjacent NN intervals exceeding 50 ms: in heart beat number, which is typically used for long-range 24h heart rate variability analysis.
(10) The number of beats with a difference between adjacent NN intervals exceeding 50ms is a percentage of the total number of beats in the NN interval (pNN 50): substantially the same meaning as NN 50. In contrast, when NN50 is used, the heart rate variability measurement time must be strictly specified to be 24 hours.
(11) Heart rate variability triangular index (heart rate variability index): the total heart rate for an NN interval is divided by the height of the histogram for the NN interval (i.e., the heart rate at the highest point of the NN interval histogram). For normalization, typically the NN interval histogram must be plotted at 7.8125ms (1/128 s) intervals.
(12) Width of NN interval histogram (TRNN): in ms, it uses a triangle with the highest point of the NN interval histogram as the vertex to describe approximately the shape of the NN interval histogram.
(13) Heart rate variability differential index: the width of the histogram is calculated in ms at two set heights (e.g. 1000 heart beats and 10000 heart beats) of the adjacent NN interval difference histogram, the difference between these two widths being the differential index.
(14) Heart rate variability log index: in the adjacent NN interval difference histogram, the abscissa is the absolute value of the difference of NN intervals of two adjacent heartbeats.
(15) Lorenz scatter plot: the NN interval NNn of the previous heart beat and the NN interval NNn +1 of the next heart beat are taken as the abscissa and the adjacent NN intervals are taken as the ordinate. Lorenz scatter plots were obtained by plotting all of these points for NN interval data obtained from long-range detection.
As a specific example, the frequency domain analysis of the frequency domain analysis unit 1042 can be used for short range detection for 5min and long range detection for 24h, so that spectrum analysis can be performed. But short-range detection for 5min is recommended for spectral analysis. Short-range detection for 5min includes:
(1) Total Power (TP): unit ms 2 Is the integral value of the power spectrum density curve in the range of 0.0-0.4 Hz;
(2) Ultra low frequency power (ULF): unit ms2 is the integral value of the power spectrum density curve in the range of 0.003-0.04 Hz;
(3) Low frequency power (LF): in ms 2 LF is the integral value of the power spectral density curve in the range of 0.04-0.15 Hz;
(4) High frequency power (HF): in ms 2 HF is the integral value of the power spectrum density curve in the range of 0.15-0.4 Hz;
(5) Low frequency high frequency power ratio (LF/HF);
(6) Normalized LF power (lfnu);
(7) Normalized HF power (hfnu).
As shown in fig. 9, fig. 9 is a schematic diagram of a result obtained by performing spectral analysis on the open source database nsr001 by the frequency domain analysis unit 1042.
As a specific embodiment, the nonlinear analysis unit 1043 mainly calculates some relevant poincare nonlinear parameters. The specific analysis method comprises the following steps:
(1) SD1, standard deviation of Poincare image projected on a straight line perpendicular to an identity line;
(2) SD2: the poincare map projection y=standard deviation on the x line;
(3)SD2/SD1。
as shown in fig. 10, fig. 10 shows the corresponding poincare result after analyzing the nsr001 data of the open source database by the nonlinear analysis unit 1043. Wherein SD1 is 22.1526, SD2 is 236.9698, and SD2/SD1 value is 10.6971.
The invention discloses an intelligent heart rate variability analysis system, which is used for carrying out digital-to-analog conversion on an electrocardiogram original signal through a signal acquisition module to obtain an electrocardiogram digital signal; preprocessing such as filtering is carried out on the electrocardiogram digital signals through a preprocessing module, and a data base is provided for subsequent heart rate variability analysis; the correction module is used for fully utilizing the information corrected by doctors in a neural network learning mode, continuously improving the accuracy of an algorithm, fully utilizing the existing medical resources, greatly improving the accuracy of automatic identification and correction of an analysis system, providing personalized function selection for users in the use process, and meeting the use requirements of multiple scenes and individuation of the users; and the heart rate variability is analyzed from multiple angles through the report output module, so that the analysis result is more comprehensive and accurate.
The invention realizes the full utilization of the existing medical resources, can automatically improve the analysis algorithm of heart rate variability, improves the accuracy of the automatic analysis result of heart rate variability, and achieves the full automatic diagnosis of heart rate variability; and can meet the requirements of multiple scenes and individuation of users.
The present invention is not limited to the above-mentioned embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present invention are intended to be included in the scope of the present invention.
Claims (8)
1. An intelligent heart rate variability analysis system, comprising: the system comprises a signal acquisition module, a preprocessing module, a correction module and a report output module;
the signal acquisition module is used for acquiring an electrocardiogram digital signal;
the preprocessing module is used for preprocessing the electrocardiogram digital signals to obtain electrocardiogram data;
the correction module is used for analyzing the electrocardiogram data to obtain an NN interval sequence;
the report output module is used for obtaining a heart rate variability analysis report according to the NN interval sequence;
the correction module comprises an automatic correction unit, a manual correction unit, a storage unit, a neural network training unit and a prediction unit;
an automatic correction unit for eliminating non-sinus cardiac RR intervals of the electrocardiographic data to obtain a first NN interval sequence;
the manual correction unit is used for providing an interactive interface for a user, so that the user can manually correct and mark the R wave position and type of the first NN interval sequence and the RR interval sequence;
the storage unit is used for storing the electrocardiogram data after the manual correction and the marking of the R wave position and type and the RR interval sequence;
the neural network training unit is used for establishing an initial correction model, and training the initial correction model by using the electrocardiogram data in the storage unit to obtain a complete correction model;
the prediction unit is used for carrying out QRS recognition and RR interval automatic correction on the electrocardiogram data by utilizing the trained complete correction model;
the neural network training unit comprises a selection module, an interactive training module and a self-training module;
the selection module is used for providing an interface for a user to enable the user to select an interactive training mode and a self-training mode;
the interactive training module is used for enabling a user to autonomously select an initial correction model and a neural network training method to train the neural network;
and the self-training module is used for training the neural network according to a preset neural network and training method.
2. The intelligent heart rate variability analysis system of claim 1, wherein the preprocessing module comprises: an outlier detection unit, a filter unit and an identification unit;
an outlier detection unit for determining abnormal data in the electrocardiogram digital signal;
a filter unit for filtering out artifacts in the electrocardiogram digital signal;
a QRS recognition unit for recognizing a QRS complex fiducial point in the electrocardiogram digital signal.
3. The intelligent heart rate variability analysis system of claim 1, wherein the automatic correction unit comprises a screening module and a calculation module;
the screening module is used for screening RR intervals meeting the first judgment condition in the RR interval sequence of the electrocardiogram data and judging whether the RR intervals meet the second judgment condition or not; if the second judging condition is met, recording the position of the RR interval in the RR interval sequence, and removing the RR interval from the RR interval sequence to obtain a first RR interval sequence; calculating the heart rate according to the first RR interval sequence, and eliminating RR intervals with the heart rate meeting a third judging condition to obtain a second RR interval sequence;
and the calculation module is used for processing the second RR interval sequence by using a preset algorithm to obtain a first NN interval sequence.
4. The system of claim 1, wherein the interactive training module comprises a setting module for enabling a user to set data set partitioning ratios, select neural network models, and super-parametric tuning methods.
5. The intelligent heart rate variability analysis system of claim 1, wherein the self-training module comprises a web search module, an architecture optimization module, and a performance assessment module,
the network searching module is used for storing a basic neural network model structure;
the architecture optimization module is used for performing super-parameter setting on the basic neural network model by using a preset architecture optimization method to obtain a candidate correction model;
and the performance evaluation module is used for performing performance evaluation on the candidate correction models with complete training, and selecting the correction model with the best performance as the correction model with complete training.
6. The intelligent heart rate variability analysis system of claim 5, wherein the network search module comprises a basic arithmetic unit, the basic arithmetic unit comprising the basic neural network model structure.
7. The intelligent heart rate variability analysis system of claim 5, wherein the architecture optimization method comprises a gradient descent-based micro-web searchable method.
8. The intelligent heart rate variability analysis system of claim 1, wherein the report output module comprises a time domain analysis unit, a frequency domain analysis unit, and a non-linear analysis unit;
the time domain analysis unit is used for carrying out long-range and short-range heart rate variability time domain analysis according to the NN interval sequence to obtain a heart rate variability time domain analysis report;
the frequency domain analysis unit is used for carrying out long-range and short-range spectrum analysis according to the NN interval sequence to obtain a heart rate variability frequency domain analysis report;
and the nonlinear analysis unit is used for calculating nonlinear parameters of the NN interval sequence to obtain a heart rate variability nonlinear parameter analysis report.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210253155.0A CN114451879B (en) | 2022-03-15 | 2022-03-15 | Intelligent heart rate variability analysis system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210253155.0A CN114451879B (en) | 2022-03-15 | 2022-03-15 | Intelligent heart rate variability analysis system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114451879A CN114451879A (en) | 2022-05-10 |
| CN114451879B true CN114451879B (en) | 2024-04-09 |
Family
ID=81417972
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210253155.0A Active CN114451879B (en) | 2022-03-15 | 2022-03-15 | Intelligent heart rate variability analysis system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114451879B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1355680A (en) * | 1999-06-14 | 2002-06-26 | 宝洁公司 | Skin imaging and analysis systems and methods |
| CN101203747A (en) * | 2004-01-21 | 2008-06-18 | 医软信息科技有限公司 | Method and system for intelligent qualitative and quantitative analysis of digital radiography softcopy reading |
| CN104127194A (en) * | 2014-07-14 | 2014-11-05 | 华南理工大学 | Depression evaluating system and method based on heart rate variability analytical method |
| CN107343016A (en) * | 2016-12-05 | 2017-11-10 | 北京迈多多科技有限公司 | Biological electricity analyzes and processes platform |
| CN206792400U (en) * | 2017-01-16 | 2017-12-26 | 吉林东华原医疗设备有限责任公司 | HRV detection means |
| CN107951485A (en) * | 2017-11-27 | 2018-04-24 | 乐普（北京）医疗器械股份有限公司 | Ambulatory ECG analysis method and apparatus based on artificial intelligence self study |
| CN110393520A (en) * | 2018-11-16 | 2019-11-01 | 迪姆软件(北京)有限公司 | A kind of electrocardiogram (ECG) data analysis method and device, electrocardiograph analyzer, storage medium |
| CN110448289A (en) * | 2019-07-18 | 2019-11-15 | 深圳邦健生物医疗设备股份有限公司 | A kind of heart rate variability (HRV) analysis method, device, storage medium and equipment |
| CN215272821U (en) * | 2021-04-20 | 2021-12-24 | 李妙 | Calibrating device of electrocardio diagnosis and analysis system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9655532B2 (en) * | 2015-06-19 | 2017-05-23 | Michael Blake | Wearable physiological monitoring and notification system based on real-time heart rate variability analysis |
| US20220015711A1 (en) * | 2020-07-20 | 2022-01-20 | Board Of Regents, The University Of Texas System | System and method for automated analysis and detection of cardiac arrhythmias from electrocardiograms |
-
2022
- 2022-03-15 CN CN202210253155.0A patent/CN114451879B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1355680A (en) * | 1999-06-14 | 2002-06-26 | 宝洁公司 | Skin imaging and analysis systems and methods |
| CN101203747A (en) * | 2004-01-21 | 2008-06-18 | 医软信息科技有限公司 | Method and system for intelligent qualitative and quantitative analysis of digital radiography softcopy reading |
| CN104127194A (en) * | 2014-07-14 | 2014-11-05 | 华南理工大学 | Depression evaluating system and method based on heart rate variability analytical method |
| CN107343016A (en) * | 2016-12-05 | 2017-11-10 | 北京迈多多科技有限公司 | Biological electricity analyzes and processes platform |
| CN206792400U (en) * | 2017-01-16 | 2017-12-26 | 吉林东华原医疗设备有限责任公司 | HRV detection means |
| CN107951485A (en) * | 2017-11-27 | 2018-04-24 | 乐普（北京）医疗器械股份有限公司 | Ambulatory ECG analysis method and apparatus based on artificial intelligence self study |
| CN110393520A (en) * | 2018-11-16 | 2019-11-01 | 迪姆软件(北京)有限公司 | A kind of electrocardiogram (ECG) data analysis method and device, electrocardiograph analyzer, storage medium |
| CN110448289A (en) * | 2019-07-18 | 2019-11-15 | 深圳邦健生物医疗设备股份有限公司 | A kind of heart rate variability (HRV) analysis method, device, storage medium and equipment |
| CN215272821U (en) * | 2021-04-20 | 2021-12-24 | 李妙 | Calibrating device of electrocardio diagnosis and analysis system |
Non-Patent Citations (1)
| Title |
|---|
| 用于睡眠呼吸暂停检测的心电特征稳定性分析;董孝彤;曲新亮;魏守水;;生物医学工程研究;20200325(第01期);10-14 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114451879A (en) | 2022-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mousavi et al. | ECGNET: Learning where to attend for detection of atrial fibrillation with deep visual attention | |
| US20190167143A1 (en) | Electrocardiogram processing system for delineation and classification | |
| Narin et al. | Early prediction of paroxysmal atrial fibrillation based on short-term heart rate variability | |
| Kantelhardt et al. | Characterization of sleep stages by correlations in the magnitude and sign of heartbeat increments | |
| CN107951485A (en) | Ambulatory ECG analysis method and apparatus based on artificial intelligence self study | |
| WO2019161610A1 (en) | Electrocardiogram information processing method and electrocardiogram workstation system | |
| CN109411042B (en) | Electrocardio information processing method and electrocardio workstation | |
| CN111067503A (en) | Sleep staging method based on heart rate variability | |
| CN107451417A (en) | Dynamic ECG analysis intelligent diagnosis system and method | |
| US11678831B2 (en) | Electrocardiogram processing system for detecting and/or predicting cardiac events | |
| CN112932498B (en) | T waveform state classification system with generalization capability based on deep learning | |
| CN101427917A (en) | ECG abnormal detection method based on inherent trend subsequence mode decomposition | |
| WO2020012488A1 (en) | Heart rate variability analysis in mammalians | |
| Wang et al. | QRS detection in ECG signal based on residual network | |
| Bouaziz et al. | Diagnostic of ECG arrhythmia using wavelet analysis and K-nearest neighbor algorithm | |
| Venton et al. | Signal to image to classification: Transfer learning for ecg | |
| CN114451879B (en) | Intelligent heart rate variability analysis system | |
| Savostin et al. | Using the k-nearest neighbors algorithm for automated detection of myocardial infarction by electrocardiogram data entries | |
| WO2021164347A1 (en) | Method and apparatus for predicting blood pressure | |
| Krivenko et al. | Using mel-frequency cepstrum and amplitude-time heart variability as XGBoost handcrafted features for heart disease detection | |
| CN112656393A (en) | Heart rate variability detection method and system | |
| KR20140097039A (en) | Method and apparatus for classifying cardiac arrhythmia using an auto associative neural network | |
| CN107767934A (en) | A kind of HRV characteristic range methods of estimation for being used to describe pressure | |
| Jovic et al. | Biomedical time series preprocessing and expert-system based feature extraction in MULTISAB platform | |
| Hori et al. | Arrhythmia detection based on patient-specific normal ECGs using deep learning |
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 |