CN113854985A - Method and device for obtaining machine learning model samples for blood pressure prediction - Google Patents
Method and device for obtaining machine learning model samples for blood pressure prediction Download PDFInfo
- Publication number
- CN113854985A CN113854985A CN202110997501.1A CN202110997501A CN113854985A CN 113854985 A CN113854985 A CN 113854985A CN 202110997501 A CN202110997501 A CN 202110997501A CN 113854985 A CN113854985 A CN 113854985A
- Authority
- CN
- China
- Prior art keywords
- characteristic value
- blood pressure
- value
- determining
- electrocardiosignal
- 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.)
- Pending
Links
- 230000036772 blood pressure Effects 0.000 title claims abstract description 132
- 238000010801 machine learning Methods 0.000 title claims abstract description 84
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000001228 spectrum Methods 0.000 claims description 26
- 230000000747 cardiac effect Effects 0.000 claims description 21
- 238000011156 evaluation Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 15
- 238000005259 measurement Methods 0.000 abstract description 2
- 230000035487 diastolic blood pressure Effects 0.000 description 21
- 238000012417 linear regression Methods 0.000 description 21
- 230000035488 systolic blood pressure Effects 0.000 description 21
- 230000015654 memory Effects 0.000 description 16
- 238000003066 decision tree Methods 0.000 description 9
- 230000002159 abnormal effect Effects 0.000 description 7
- 238000004590 computer program Methods 0.000 description 5
- 230000005291 magnetic effect Effects 0.000 description 5
- 238000007637 random forest analysis Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 241000282414 Homo sapiens Species 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000001419 dependent effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000001615 p wave Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000003205 diastolic effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- QSHDDOUJBYECFT-UHFFFAOYSA-N mercury Chemical compound [Hg] QSHDDOUJBYECFT-UHFFFAOYSA-N 0.000 description 1
- 229910052753 mercury Inorganic materials 0.000 description 1
- 210000005037 parasympathetic nerve Anatomy 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000002889 sympathetic effect Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000003936 working memory Effects 0.000 description 1
Images
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
- A61B5/021—Measuring pressure in heart or blood vessels
-
- 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
-
- 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
- A61B5/353—Detecting P-waves
-
- 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
- A61B5/355—Detecting T-waves
-
- 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
- A61B5/358—Detecting ST segments
-
- 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
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Cardiology (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Biophysics (AREA)
- Public Health (AREA)
- Veterinary Medicine (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Molecular Biology (AREA)
- Medical Informatics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Artificial Intelligence (AREA)
- Physiology (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Mathematical Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Vascular Medicine (AREA)
- Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
Abstract
According to the technical scheme for obtaining the machine learning model sample for blood pressure prediction, electrocardiosignals and corresponding blood pressure values are obtained; determining a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal; and determining the blood pressure value as a label of a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal, and determining the first characteristic value, the second characteristic value, the third characteristic value and the label of the electrocardiosignal as the machine learning model sample. Further, a sample set of the machine learning model is constructed based on a plurality of samples, and the machine learning model learns based on the sample set to determine a blood pressure prediction model. According to the scheme, the blood pressure can be predicted through the multi-dimensional characteristic value of the electrocardiosignals, independent measurement is not needed, the user experience is improved, the blood pressure prediction model determined based on the diversity sample set can be used for predicting the blood pressure of the electrocardiosignals of different people, and the method has universal applicability.
Description
Technical Field
The application relates to the technical field of electrocardiosignal processing, in particular to a technology for acquiring machine learning model samples for blood pressure prediction.
Background
The electrocardiosignal is a periodic signal, has the elements of frequency, amplitude (or power), phase and the like, is easy to detect, has more intuitive regularity, and is a bioelectric signal which is researched by human beings at the earliest and applied to medical clinic. A typical cardiac electrical signal waveform is shown in fig. 1, where the x-axis is time (T) in milliseconds (mS) and the y-axis is voltage (V) in millivolts (mV), and can be thought of as being the exact superposition of a series of sine waves of different frequency, amplitude and phase.
Previously, clinicians mainly analyzed electrocardiosignals manually, which undoubtedly is time-consuming and labor-consuming and has low reliability. With the rapid development of computer technology, especially with the rapid development of artificial intelligence technology, more and more methods for analyzing electrocardiosignals by adopting machine learning technology appear, the characteristics of the electrocardiosignals are extracted, automatic analysis can be performed through a machine learning model, human physiological indexes corresponding to the electrocardiosignals are obtained, and for example, blood pressure corresponding to the electrocardiosignals can be predicted through the learned machine learning model.
In the existing method for predicting the blood pressure corresponding to the electrocardiosignals through the learned machine learning model, numerical value processing and characteristic value extraction are only carried out on the basis of individual parameters of the electrocardiosignals, and the established machine learning model needs to establish a model function aiming at the blood pressure of each person independently and has no universality.
Disclosure of Invention
The application aims to provide a technical scheme for obtaining machine learning model samples for blood pressure prediction, which is used for constructing a machine learning model which is suitable for different people and used for predicting blood pressure based on multi-dimensional characteristic values of electrocardiosignals.
According to an aspect of the present application, there is provided a method of machine learning model sample acquisition for blood pressure prediction, wherein the method comprises:
acquiring electrocardiosignals and corresponding blood pressure values thereof;
determining a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal;
and determining the blood pressure value as a label of a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal, and determining the first characteristic value, the second characteristic value, the third characteristic value and the label of the electrocardiosignal as the machine learning model sample.
Optionally, wherein determining the first characteristic value of the cardiac signal comprises:
determining each heartbeat waveform in the electrocardiosignal;
acquiring a characteristic value of each heartbeat waveform based on each heartbeat waveform,
and traversing all heartbeat waveforms of the electrocardiosignals, acquiring statistics of characteristic values of all heartbeat waveforms, and determining the statistics as a first characteristic value of the electrocardiosignals.
Optionally, wherein the characteristic value of each heartbeat waveform includes at least one of:
the P-wave duration of each heartbeat waveform;
PR intervals for each heartbeat waveform;
QRS time of each heartbeat waveform;
the ST segment duration of each heartbeat waveform;
the T-wave duration of each heartbeat waveform;
the duration of the TP segment of each heartbeat waveform;
QT interval of each heartbeat waveform;
the U-wave duration of each heartbeat waveform.
Optionally, the statistical values comprise at least one of:
root mean square;
the absolute deviation of the median;
the median coefficient of variation.
Optionally, wherein determining the second characteristic value of the cardiac signal comprises:
determining each heartbeat waveform in the electrocardiosignal;
acquiring characteristic values of two adjacent heartbeat waveforms based on the two adjacent heartbeat waveforms,
and traversing all heartbeat waveforms of the electrocardiosignals, acquiring statistics of the characteristic values of all two adjacent heartbeat waveforms, and determining the statistics as a second characteristic value of the electrocardiosignals.
Optionally, wherein the feature values of the two adjacent heartbeat waveforms include at least one of:
the PP interval of every two adjacent heartbeat waveforms;
the QQ interval of every two adjacent heartbeat waveforms;
RR intervals of two adjacent heartbeat waveforms;
SS intervals of every two adjacent heartbeat waveforms;
TT intervals of every two adjacent heartbeat waveforms;
UU intervals of two adjacent heartbeat waveforms.
Optionally, wherein the statistical values comprise at least one of:
root mean square;
the absolute deviation of the median;
a median coefficient of variation;
a number of consecutive interval differences greater than 50 milliseconds.
Optionally, wherein determining the third characteristic value of the cardiac signal comprises:
acquiring a power spectrum of a time domain waveform corresponding to the second characteristic value;
acquiring power values of a plurality of preset frequency domains based on the power spectrum;
and determining the ratio of the power values of the plurality of preset frequency domains as a third characteristic value of the electrocardiosignal.
Optionally, wherein the determining the third characteristic value of the cardiac signal further includes:
calculating an entropy value of the power spectrum based on the power spectrum;
and determining the entropy value of the power spectrum as a third characteristic value of the electrocardiosignal.
Optionally, the method for obtaining machine learning model samples for blood pressure prediction further comprises:
and determining a plurality of machine learning model samples based on a plurality of acquired historical electrocardiosignals and corresponding blood pressure values thereof so as to construct a sample set of the machine learning model.
Optionally, the method further comprises:
and training the machine learning model based on the sample set, and determining the trained machine learning model as a blood pressure prediction model when the evaluation index of the machine learning model accords with a preset threshold value.
Optionally, the method further comprises:
determining a first characteristic value, a second characteristic value and a third characteristic value of the obtained electrocardiosignals;
and inputting the first characteristic value, the second characteristic value and the third characteristic value of the electrocardiosignals into the blood pressure prediction model to predict the blood pressure value corresponding to the electrocardiosignals.
According to yet another aspect of the present application, there is also provided an apparatus for machine learning model sample acquisition for blood pressure prediction, wherein the apparatus comprises:
the first module is used for acquiring electrocardiosignals and corresponding blood pressure values thereof;
the second module is used for determining a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal;
and the third module is used for determining the blood pressure value as a label of the first characteristic value, the second characteristic value and the third characteristic value of the electrocardiosignal, and determining the first characteristic value, the second characteristic value, the third characteristic value and the label of the electrocardiosignal as the machine learning model sample.
Optionally, the apparatus for machine learning model sample acquisition for blood pressure prediction further comprises:
and the fourth module is used for determining a plurality of machine learning model samples based on a plurality of acquired historical electrocardiosignals and corresponding blood pressure values thereof so as to construct a sample set of the machine learning model.
Optionally, the apparatus further comprises:
and the fifth module is used for training the machine learning model based on the sample set, and determining the trained machine learning model as a blood pressure prediction model when the evaluation index of the machine learning model meets a preset threshold value.
Optionally, the apparatus further comprises:
the sixth module is used for determining the first characteristic value, the second characteristic value and the third characteristic value of the acquired electrocardiosignals;
and the seventh module is used for inputting the first characteristic value, the second characteristic value and the third characteristic value of the electrocardiosignals into the blood pressure prediction model to predict the blood pressure value corresponding to the electrocardiosignals.
Compared with the prior art, the electrocardiosignal and the corresponding blood pressure value are obtained through the technical scheme of obtaining the machine learning model sample for blood pressure prediction; determining a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal; and determining the blood pressure value as a label of a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal, and determining the first characteristic value, the second characteristic value, the third characteristic value and the label of the electrocardiosignal as the machine learning model sample. Further, a plurality of machine learning model samples are determined based on a plurality of acquired historical electrocardiosignals and corresponding blood pressure values thereof, so as to construct a sample set of the machine learning model. And training the machine learning model based on the sample set, and determining a blood pressure prediction model for predicting blood pressure through multi-dimensional characteristic values of electrocardiosignals. According to the technical scheme, the blood pressure is predicted based on the multi-dimensional characteristic value of the electrocardiosignals, the prediction accuracy is high, independent measurement is not needed, the user experience is improved, the blood pressure prediction model determined based on the diversity sample set training can be used for predicting the blood pressure of the electrocardiosignals of different people, and the method and the device have universal applicability.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments made with reference to the following drawings:
FIG. 1 illustrates a typical cardiac electrical signal waveform schematic;
FIG. 2 illustrates a flow diagram of a method for machine learning model sample acquisition for blood pressure prediction, according to one aspect of the present application;
FIG. 3 illustrates an apparatus diagram of machine learning model sample acquisition for blood pressure prediction according to another aspect of the present application;
the same or similar reference numbers in the drawings identify the same or similar elements.
Detailed Description
The present invention is described in further detail below with reference to the attached drawing figures.
In a typical configuration of the present application, each module of the device, apparatus, and/or trusted party includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.
Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include non-transitory computer readable media (transient media), such as modulated data signals and carrier waves.
In order to further explain the technical means and effects adopted by the present application, the following description clearly and completely describes the technical solution of the present application with reference to the accompanying drawings and alternative embodiments.
FIG. 2 illustrates a flow diagram of a method for machine learning model sample acquisition for blood pressure prediction in one aspect of the present application, wherein the method of an embodiment comprises:
s21 obtaining the electrocardiosignals and the corresponding blood pressure values;
s22, determining a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal;
s23 determines the blood pressure value as a label of the first, second, and third eigenvalues of the electrocardiographic signal, and determines the first, second, third, and label of the electrocardiographic signal as the machine learning model sample.
In the present application, the method is performed by the apparatus 100, wherein the apparatus 100 is a computer device and/or cloud installed with related computer programs, the computer device includes but is not limited to a personal computer, a notebook computer, an industrial computer, a network host, a single network server, a plurality of network server sets; the Cloud is made up of a large number of computers or web servers based on Cloud Computing (Cloud Computing), which is a type of distributed Computing, a virtual supercomputer consisting of a collection of loosely coupled computers.
The computer device and/or cloud are merely examples, and other existing or future devices and/or resource sharing platforms, as applicable to the present application, are also intended to be included within the scope of the present application and are hereby incorporated by reference.
In this embodiment, in step S21, the device 100 acquires an electrocardiographic signal acquired from an acquisition object and a corresponding blood pressure value thereof, where the blood pressure value includes two values, namely, diastolic pressure and systolic pressure, and the acquired electrocardiographic signal may be an electrocardiographic signal waveform (voltage-time waveform, i.e., electrocardiogram) stored in a digital format after being acquired by a dedicated electrocardiographic signal acquisition device, or an electrocardiographic signal waveform which is acquired by a dedicated electrocardiographic signal acquisition device and transmitted to the device 100, and the device 100 acquires and converts the acquired electrocardiographic signal waveform into the digital format. The obtained electrocardiosignals can be single-lead electrocardiosignals or multi-lead electrocardiosignals, wherein in order to ensure that enough characteristic information and corresponding characteristic values can be obtained, the sampling frequency of each lead electrocardiosignal is not less than 100Hz, and the electrocardiosignals last for at least 25 seconds. Wherein the obtained blood pressure value may be an average value of blood pressure values measured for the same subject a plurality of times during the acquisition of the cardiac signal, including an average value of diastolic blood pressure and an average value of systolic blood pressure.
In this embodiment, in step S22, the apparatus 100 determines a first characteristic value, a second characteristic value and a third characteristic value of the cardiac signal.
The electrocardiographic signal waveform obtained in step S21 is processed, the characteristic information in the entire electrocardiographic signal waveform is labeled, and the first characteristic value, the second characteristic value, and the third characteristic value of the electrocardiographic signal are determined. After the characteristic information is labeled, the whole section of electrocardiosignal waveform and the characteristic information need to be checked, and invalid electrocardiosignal waveforms, invalid heartbeat waveforms in the electrocardiosignal waveforms and characteristic information with wrong labeling are eliminated, so that the correctness of various subsequently determined characteristic values is ensured. Preferably, after the invalid electrocardiosignal waveforms, the invalid heartbeat waveforms in the electrocardiosignal waveforms and the characteristic information marked with errors are eliminated, the retained normal electrocardiosignal waveforms at least comprise 5 groups of heartbeat waveforms adjacent to each other in pairs so as to ensure that the characteristic values with enough quantity can be determined. The more the two adjacent heartbeat waveforms included in the retained normal electrocardiosignal waveform are, the more the number of the characteristic values is correspondingly obtained.
Optionally, wherein determining the first characteristic value of the cardiac signal comprises:
determining each heartbeat waveform in the electrocardiosignal;
acquiring a characteristic value of each heartbeat waveform based on each heartbeat waveform,
and traversing all heartbeat waveforms of the electrocardiosignals, acquiring statistics of characteristic values of all heartbeat waveforms, and determining the statistics as a first characteristic value of the electrocardiosignals.
The electrocardiographic signal waveform obtained by the apparatus 100 generally consists of a plurality of heartbeat waveforms, and the apparatus 100 processes the electrocardiographic signal waveform, labels characteristic information in the whole section of electrocardiographic signal waveform, including determining each heartbeat waveform in the electrocardiographic signal waveform, and labels the characteristic information of each heartbeat waveform. The characteristic information of each heartbeat waveform includes: p wave, PR interval, QRS, ST segment, T wave, U wave, TP segment, QT interval, etc. Acquiring the characteristic values corresponding to various types of characteristic information of each heartbeat waveform, on this basis, the device 100 traverses all heartbeat waveforms of the whole section of electrocardiosignal waveform, acquires the respective statistical values of the characteristic values corresponding to various types of characteristic information of all heartbeat waveforms, and determines each statistical value as the first characteristic value of the section of electrocardiosignal.
Optionally, wherein the characteristic value of each heartbeat waveform includes at least one of:
the P-wave duration of each heartbeat waveform;
PR intervals for each heartbeat waveform;
QRS time of each heartbeat waveform;
the ST segment duration of each heartbeat waveform;
the T-wave duration of each heartbeat waveform;
the duration of the TP segment of each heartbeat waveform;
QT interval of each heartbeat waveform;
the U-wave duration of each heartbeat waveform.
Wherein, the characteristic information of each heartbeat waveform comprises: p wave, PR interval, QRS, ST section, T wave, U wave, TP section and QT interval etc. the characteristic value that each characteristic information of every heartbeat waveform corresponds can include: the time length of P wave corresponding to P wave, PR interval corresponding to PR interval, QRS time corresponding to QRS, ST segment time length corresponding to ST segment, T wave time length corresponding to T wave, TP segment time length corresponding to TP segment, QT interval corresponding to QT interval and U wave time length corresponding to U wave. The statistical values corresponding to the characteristic values of all heartbeat waveforms of the whole cardiac signal segment may include: the statistical value of characteristic value P wave duration corresponding to P waves of all heart beat waveforms of the whole section of electrocardiosignal, the statistical value of characteristic value PR interval corresponding to PR interval of all heart beat waveforms of the whole section of electrocardiosignal, the statistical value of characteristic value QRS time corresponding to QRS of all heart beat waveforms of the whole section of electrocardiosignal, the statistical value of characteristic value ST segment duration corresponding to ST segment of all heart beat waveforms of the whole section of electrocardiosignal, the statistical value of characteristic value T wave duration corresponding to T wave of all heart beat waveforms of the whole section of electrocardiosignal, the statistical value of characteristic value TP segment duration corresponding to TP segment of all heart beat waveforms of the whole section of electrocardiosignal, the statistical value of characteristic value QT interval corresponding to QT interval of all heart beat waveforms of the whole section of electrocardiosignal and the statistical value of characteristic value U wave duration corresponding to U wave of all heart beat waveforms of the whole section of electrocardiosignal.
Optionally, wherein the statistical value of the feature values of all heartbeat waveforms includes at least one of:
root mean square;
the absolute deviation of the median;
the median coefficient of variation.
The statistical value may be a root mean square of each characteristic value corresponding to each type of characteristic information of all heartbeat waveforms of the entire section of the electrocardiographic signal, and/or an absolute deviation of a value thereof, and/or a median variation coefficient, or may be other statistical values of each characteristic value corresponding to each type of characteristic information of all heartbeat waveforms of the entire section of the electrocardiographic signal, and herein, no specific limitation is made, and any statistical value form of each characteristic value corresponding to each type of characteristic information of all heartbeat waveforms of the entire section of the electrocardiographic signal should be within a protection range of the present application.
Optionally, wherein determining the second characteristic value of the cardiac signal comprises:
determining each heartbeat waveform in the electrocardiosignal;
acquiring characteristic values of two adjacent heartbeat waveforms based on the two adjacent heartbeat waveforms,
and traversing all heartbeat waveforms of the electrocardiosignals, acquiring statistics of the characteristic values of all two adjacent heartbeat waveforms, and determining the statistics as a second characteristic value of the electrocardiosignals.
On the basis of processing the electrocardiosignal waveform and marking characteristic information of each heartbeat waveform of the whole section of electrocardiosignal, such as a P wave, a PR interval, a QRS, an ST section, a T wave, a U wave, a TP section and a QT interval, the device 100 acquires characteristic values corresponding to the same type of characteristic information of two adjacent heartbeat waveforms according to the two adjacent heartbeat waveforms (the two continuous heartbeat waveforms which are adjacent to each other and are not allowed to be spliced to obtain the two adjacent heartbeat waveforms), on the basis, the device 100 traverses all heartbeat waveforms of the whole section of electrocardiosignal waveform, acquires respective statistical values of the characteristic values corresponding to the same type of characteristic information of the two adjacent heartbeat waveforms, and determines the statistical values as second characteristic values of the section of electrocardiosignal.
Optionally, wherein the feature values of the two adjacent heartbeat waveforms include at least one of:
the PP interval of every two adjacent heartbeat waveforms;
the QQ interval of every two adjacent heartbeat waveforms;
RR intervals of two adjacent heartbeat waveforms;
SS intervals of every two adjacent heartbeat waveforms;
TT intervals of every two adjacent heartbeat waveforms;
UU intervals of two adjacent heartbeat waveforms.
Wherein, the characteristic information in each heartbeat waveform of the electrocardiosignal can comprise: the characteristic values corresponding to the characteristic information of two adjacent heartbeat waveforms are respectively a PP interval, a QQ interval, an RR interval, an SS interval, a TT interval and a UU interval, and the statistical value of each characteristic value corresponding to the characteristic information of all two adjacent heartbeat waveforms of the whole section of electrocardiosignal can include: the statistical value of the PP intervals corresponding to the P waves of all two adjacent heartbeat waveforms, the statistical value of the QQ intervals corresponding to the Q points of all two adjacent heartbeat waveforms, the statistical value of the RR intervals corresponding to the R points of all two adjacent heartbeat waveforms, the statistical value of the SS intervals corresponding to the S points of all two adjacent heartbeat waveforms, the statistical value of the TT intervals corresponding to the T waves of all two adjacent heartbeat waveforms and the statistical value of the UU intervals corresponding to the U waves of all two adjacent heartbeat waveforms.
If abnormal characteristic information exists in the characteristic information labeled by the electrocardiosignal waveform corresponding to the whole section of electrocardiosignal, the abnormal characteristic information and the whole heartbeat waveform where the abnormal characteristic information exists need to be deleted, under the condition, the electrocardiosignal waveform of the whole section of electrocardiosignal is divided into a plurality of sections, characteristic values corresponding to various kinds of characteristic information of two adjacent heartbeat waveforms of the plurality of sections of electrocardiosignal waveforms are respectively determined, and then the characteristic values are summarized to determine a statistical value.
Optionally, wherein the statistical value of the feature values of all two adjacent heartbeat waveforms includes at least one of:
root mean square;
the absolute deviation of the median;
a median coefficient of variation;
a number of consecutive interval differences greater than 50 milliseconds.
The statistical value may be a root mean square of characteristic values corresponding to various kinds of characteristic information of all two adjacent heartbeat waveforms of the whole section of the electrocardiographic signal, and/or an absolute deviation of a value thereof, and/or a median variation coefficient, and/or a continuous interval difference number greater than 50 milliseconds therein, or may be other statistical values of characteristic values corresponding to various kinds of characteristic information of all two adjacent heartbeat waveforms of the whole section of the electrocardiographic signal.
Optionally, wherein determining the third characteristic value of the cardiac signal comprises:
acquiring a power spectrum of a time domain waveform corresponding to the second characteristic value;
acquiring power values of a plurality of preset frequency domains based on the power spectrum;
and determining the ratio of the power values of the plurality of preset frequency domains as a third characteristic value of the electrocardiosignal.
In the field of electrocardiosignal research, time domain waveforms corresponding to second characteristic values of two adjacent heartbeat waveforms, such as a PP interval, a QQ interval, an RR interval, an SS interval, a TT interval and/or a UU interval and the like, can reflect the activity of human body sympathetic and parasympathetic nerves of an acquisition object, so that the time domain waveforms corresponding to the second characteristic values can better reflect the relevance with other physiological indexes, such as blood pressure (diastolic pressure and systolic pressure) and the like compared with the time domain waveforms of electrocardiosignals. The power spectrum of the time-domain waveform corresponding to the second eigenvalue reflects the frequency distribution of the time-domain waveform speed corresponding to the second eigenvalue, and more features can be obtained from the power spectrum and used in the machine learning model described in the application.
Before the power spectrum of the time domain waveform corresponding to the second characteristic value is obtained, if abnormal characteristic information exists in characteristic information labeled on the electrocardiosignal waveform corresponding to the whole section of electrocardiosignal, the abnormal characteristic information and the whole heartbeat waveform where the abnormal characteristic information exists need to be deleted.
Combining the characteristics of the time domain waveform corresponding to the second eigenvalue, after the time domain waveform corresponding to the second eigenvalue is converted into the power spectrum, most energy (power) of the power spectrum is concentrated in a frequency domain range of 0.04-0.15 hertz (Hz), so as to facilitate the learning of a subsequent machine learning model, the frequency domain range can be divided into a plurality of preset frequency domains, and a power value in each preset frequency domain is obtained, wherein the sum of the power values in each preset frequency domain is normalized to be 1. And determining the ratio of the power values in the preset frequency domains as a third characteristic value of the electrocardiosignal. For example, the preset frequency domain may be divided as follows: VLF (Very Low Frequency), less than 0.04 Hz; LF (Low Frequency), 0.04-0.15 Hz (including two endpoints); HF (High Frequency), which is greater than 0.15Hz, where the normalized power values in the three preset Frequency domains are P1, P2, and P3, P1+ P2+ P3 is 1, P1/P2 is a, P2/P3 is b, and P1/P3 is c, and a, b, and c can be determined as the third characteristic value of the electrocardiographic signal.
In order to obtain a blood pressure prediction model with higher prediction accuracy, optionally, the determining the third characteristic value of the cardiac electrical signal further includes:
calculating an entropy value of the power spectrum based on the power spectrum;
and determining the entropy value of the power spectrum as a third characteristic value of the electrocardiosignal.
After obtaining the power spectrums of all normal heartbeat signal waveforms in the whole segment of electrocardiosignal waveform, further calculating an entropy value of the power tiling, and determining the entropy value of the power spectrum as a third characteristic value of the electrocardiosignal, that is, the third characteristic value may include: the ratio of the power value of the power spectrum of the electrocardiosignal in a plurality of preset frequency domains and the entropy value of the power spectrum.
In the embodiment, in the step S23, the apparatus 100 determines the blood pressure value obtained in the step S21 contemporaneously with the cardiac electrical signal as the label of the first feature value, the second feature value and the third feature value of the cardiac electrical signal determined in the step S22, and determines the first feature value, the second feature value, the third feature value and the label of the cardiac electrical signal as the machine learning model sample, wherein the blood pressure value includes two values, namely, diastolic pressure and systolic pressure.
Optionally, the method of the embodiment further includes:
s24 (not shown) obtains a plurality of historical electrocardiographic signals and corresponding blood pressure values, and determines a plurality of samples of the machine learning model by using the method of the above embodiment, so as to construct a sample set of the machine learning model for training the machine learning model.
In order to make the samples of the machine learning model have wider diversity, so that the trained machine learning model is suitable for different people, the blood pressure values of the samples in the selected sample set should cover different blood pressure value ranges, such as: samples corresponding to diastolic <60mmHg (millimeters of mercury), systolic <90 mmHg; samples with diastolic pressure of 60-79 mmHg and systolic pressure of 90-119 mmHg; samples with diastolic pressure of 80-89 mmHg and systolic pressure of 120-139 mmHg and samples with diastolic pressure of 89mmHg and systolic pressure of 139mmHg are used, the number of samples in the sample set in different blood pressure value ranges can be basically the same, for example, the ratio of the samples in the four blood pressure value ranges in the sample set is 25%. In addition, the electrocardiographic signals and corresponding blood pressures can be acquired based on the collected objects of different sexes and ages, for example, samples corresponding to diastolic pressure less than 60mmHg and systolic pressure less than 90mmHg can be acquired by collecting electrocardiographic signals of different collected objects and blood pressure values of the same period according to the sexes (male and female) and ages (for example, 18 to 30 years, 31 to 60 years and more than 61 years). The number of samples in the sample set for different genders and age groups should also be substantially the same.
If the acquired original electrocardiosignal and/or the corresponding blood pressure value are abnormal, for example, the duration of the acquired electrocardiosignal is less than 25 seconds, and a sufficient characteristic point cannot be determined according to the acquired electrocardiosignal to obtain the first characteristic value and the second characteristic value, the original electrocardiosignal and the corresponding blood pressure value are abandoned.
The more the first characteristic values and the second characteristic values of the samples in the selected sample set include, the higher the prediction accuracy of the blood pressure prediction model obtained after the machine learning model is trained.
Optionally, the method of the embodiment further includes:
s25 (not shown), training the machine learning model based on the sample set, and determining the trained machine learning model as a blood pressure prediction model when the evaluation index of the machine learning model meets a preset threshold.
In an alternative embodiment, the machine learning model may be a multiple linear regression model. The multiple linear regression model can be expressed by the following formula:
y＝b0+b1x1+b2x2+…+bkxk+e
wherein, b0Is a constant term, b1、b2…bkIs a partial regression coefficient (b)1Is x1、x2…xkAt fixed time, independent variable x1The effect on the dependent variable y per unit increase; same principle b2Is x1、x2…xkAt fixed time, independent variable x2The meaning of the other regression coefficients is known in the same way for each increase in the effect of one unit on the dependent variable y). e is an error term.
In this alternative embodiment, the multiple linear regression model may be expressed by the following formula:
s＝a0+a1x1+a2x2+…+aixi+b1y1+b2y2+…+bjyj+c1z1+c2z2+…+ckzk+e
wherein, a0Is a constant term of1、a2…aiFor a plurality of first characteristic values x of the electrocardiosignal1、x2…xiPartial regression coefficients of; b1、b2…bjA plurality of second characteristic values y of the electrocardiosignals1、y2…yjPartial regression coefficients of; c. C1、c2…ckFor a plurality of third characteristic values z of the electrocardiosignal1、z2…zkPartial regression coefficients of; s is a regression value, namely a blood pressure value corresponding to the electrocardiosignal obtained by regression based on the multiple linear regression model.
By using the samples in the sample set, the multiple linear regression model may be trained by using methods such as SVM (Support Vector Machines), SGD (Stochastic Gradient Descent), and the like, the constant term and each partial regression coefficient of the multiple linear regression model are determined based on the preset evaluation index threshold of the multiple linear regression model, and the trained multiple linear regression model conforming to the evaluation index threshold is determined as the blood pressure prediction model. Based on the samples in the sample set, training the multiple linear regression model by using a first characteristic value, a second characteristic value, a third characteristic value and a diastolic pressure label of a blood pressure value of electrocardiosignals in the samples, and determining the trained multiple linear regression model as a blood pressure prediction model for predicting the diastolic pressure of the blood pressure value when an evaluation index of the trained multiple linear regression model meets a preset threshold value; and training the multiple linear regression model by adopting a first characteristic value, a second characteristic value, a third characteristic value and a systolic pressure label of a blood pressure value of the electrocardiosignals in the sample based on the sample in the sample set, and determining the trained multiple linear regression model as a blood pressure prediction model for predicting systolic pressure of the blood pressure value when the evaluation index of the trained multiple linear regression model accords with a preset threshold value.
The evaluation index of the multiple linear regression model is used for evaluating the fitting degree of the multiple linear regression model, and the evaluation index can be Mean Squared Error (MSE) of a regression value or Root Mean Squared Error (RMSE) of the regression value, and when the evaluation index meets a preset threshold value, the fitting degree of the multiple linear regression model trained based on the sample set meets the requirement, and the trained multiple linear regression model can be determined as the blood pressure prediction model. If the evaluation index of the multiple linear regression model after multiple iterative training based on the sample set cannot meet the preset threshold, the trained multiple linear regression model corresponding to the optimal evaluation index may be determined as the blood pressure prediction model, for example, if MSE is used as the evaluation index, the trained multiple linear regression model with the minimum MSE is determined as the blood pressure prediction model.
In another optional embodiment, the machine learning model may also be a random forest regression model, the multiple samples of the put-back samples are sampled based on a sample set to obtain a plurality of sub-sample sets, then continuous decision and opportunity node training are performed on the random forest regression model respectively based on a first feature value, a second feature value and a third feature value of each sample in the plurality of sub-sample sets to generate a decision tree corresponding to each decision tree, each decision tree outputs one result, each decision tree is voted based on a label in the plurality of sub-sample sets, and the random forest regression model corresponding to the decision tree whose evaluation index meets a preset threshold is determined as the blood pressure prediction model according to the voting result. Voting is carried out on each decision tree based on the diastolic pressure labels of the blood pressure values in the plurality of sub-sample sets, and according to the voting result, a random forest regression model corresponding to the decision tree of which the evaluation index meets a preset threshold value is determined as a blood pressure prediction model for predicting the diastolic pressure of the blood pressure values; voting is carried out on each decision tree based on the systolic pressure labels of the blood pressure values in the plurality of sub-sample sets, and according to the voting result, a random forest regression model corresponding to the decision tree of which the evaluation index meets a preset threshold value is determined as a blood pressure prediction model for predicting the systolic pressure of the blood pressure values.
Optionally, the method of the above embodiment further includes:
s26 (not shown), determining a first characteristic value, a second characteristic value, and a third characteristic value of the acquired electrocardiographic signal;
s27 (not shown) inputs the first, second, and third feature values of the electrocardiographic signal into the blood pressure prediction model, and predicts a blood pressure value corresponding to the electrocardiographic signal.
The device 100 obtains an electrocardiographic signal which has a duration of not less than 25 seconds and contains a plurality of complete heartbeat waveforms of an acquisition object, determines a characteristic point thereof by referring to the method of the foregoing embodiment, obtains a first characteristic value, a second characteristic value and a third characteristic value thereof, and inputs the obtained first characteristic value, second characteristic value and third characteristic value of the electrocardiographic signal into the blood pressure prediction model for predicting the diastolic pressure of a blood pressure value to predict a blood pressure diastolic pressure value corresponding to the electrocardiographic signal; inputting the obtained first characteristic value, second characteristic value and third characteristic value of the electrocardiosignal into the blood pressure prediction model for predicting the systolic pressure of the blood pressure value so as to predict the systolic pressure value of the blood pressure corresponding to the electrocardiosignal.
Fig. 3 shows a schematic diagram of an apparatus for machine learning model sample acquisition for blood pressure prediction according to another aspect of the present application, wherein the apparatus of an embodiment comprises:
a first module 31, configured to obtain an electrocardiographic signal and a corresponding blood pressure value thereof;
a second module 32, configured to determine a first characteristic value, a second characteristic value, and a third characteristic value of the electrocardiographic signal;
a third module 33, configured to determine the blood pressure value as a label of the first feature value, the second feature value, and the third feature value of the electrocardiographic signal, and determine the first feature value, the second feature value, the third feature value, and the label of the electrocardiographic signal as the machine learning model sample.
In this embodiment, the apparatus is used to implement the method of the previous embodiment, and the apparatus is integrated in the same device as the software and hardware configuration environment of the previous device 100.
The first module 31 of the device acquires an electrocardiographic signal acquired from an acquisition object and a corresponding blood pressure value thereof, wherein the blood pressure value includes two numerical values of diastolic pressure and systolic pressure, and the acquired electrocardiographic signal may be an electrocardiographic signal waveform (voltage-time waveform, namely, electrocardiogram) stored in a digital format after being acquired by a special electrocardiographic signal acquisition device, or an electrocardiographic signal waveform which is acquired by the special electrocardiographic signal acquisition device and then transmitted to the device and is converted into the electrocardiographic signal waveform in the digital format after being acquired by the device. The obtained electrocardiosignals can be single-lead electrocardiosignals or multi-lead electrocardiosignals, wherein in order to ensure that enough characteristic information and corresponding characteristic values can be obtained, the sampling frequency of each lead electrocardiosignal is not less than 100Hz and lasts for at least 25 seconds. Wherein the obtained blood pressure value may be an average value of blood pressure values measured for the same subject a plurality of times during the acquisition of the cardiac signal, including an average value of diastolic blood pressure and an average value of systolic blood pressure.
The second module 32 of the apparatus processes the electrocardiographic signal waveform obtained in the first module 31, labels the characteristic information in the entire section of electrocardiographic signal waveform, and determines a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiographic signal. After the characteristic information is labeled, the whole section of electrocardiosignal waveform and the characteristic information need to be checked, and invalid electrocardiosignal waveforms, invalid heartbeat waveforms in the electrocardiosignal waveforms and characteristic information with wrong labeling are eliminated, so that the correctness of various subsequently determined characteristic values is ensured.
The third module 33 of the apparatus determines the blood pressure value obtained in the first module 31 in synchronization with the electrocardiographic signal as the first characteristic value, the second characteristic value and the label of the electrocardiographic signal determined in the second module 32, and determines the first characteristic value, the second characteristic value, the third characteristic value and the label of the electrocardiographic signal as the machine learning model sample, wherein the blood pressure value includes two values of diastolic pressure and systolic pressure.
Further, the apparatus further includes a fourth module 34 (not shown) for obtaining a plurality of historical electrocardiographic signals and corresponding blood pressure values thereof, and determining a plurality of samples of the machine learning model by using the method of the above embodiment to construct a sample set of the machine learning model for learning of the machine learning model.
Further, the apparatus further includes a fifth module 35 (not shown) for training the machine learning model based on the sample set constructed in the fourth module 34, and determining the trained machine learning model as the blood pressure prediction model when the evaluation index of the machine learning model meets a preset threshold. Based on the samples in the sample set, training the machine learning model by adopting a first characteristic value, a second characteristic value, a third characteristic value and a diastolic pressure label of a blood pressure value of electrocardiosignals in the samples, and determining the trained machine learning model as a blood pressure prediction model for predicting the diastolic pressure of the blood pressure value when an evaluation index of the trained machine learning model meets a preset threshold value; and training the machine learning model by adopting a first characteristic value, a second characteristic value, a third characteristic value and a systolic pressure label of a blood pressure value of the electrocardiosignals in the sample based on the sample in the sample set, and determining the trained machine learning model as a blood pressure prediction model for predicting systolic pressure of the blood pressure value when the evaluation index of the trained machine learning model accords with a preset threshold value.
Further, the apparatus further includes a sixth module 36 (not shown) configured to obtain an electrocardiographic signal of the collected object, determine a first characteristic value, a second characteristic value, and a third characteristic value of the electrocardiographic signal, input the first characteristic value, the second characteristic value, and the third characteristic value of the electrocardiographic signal into the blood pressure prediction model, and predict a blood pressure value corresponding to the electrocardiographic signal. Inputting the first characteristic value, the second characteristic value and the third characteristic value of the electrocardiosignal into a blood pressure prediction model for predicting diastolic pressure of a blood pressure value to obtain a predicted value of the diastolic pressure of the blood pressure; and inputting the first characteristic value, the second characteristic value and the third characteristic value of the electrocardiosignal into a blood pressure prediction model for predicting the systolic pressure of the blood pressure value to obtain a predicted value of the systolic pressure of the blood pressure.
According to yet another aspect of the present application, there is also provided a computer readable medium having stored thereon computer readable instructions executable by a processor to implement the foregoing method.
It should be noted that the present application may be implemented in software and/or a combination of software and hardware, for example, implemented using Application Specific Integrated Circuits (ASICs), general purpose computers or any other similar hardware devices. In one embodiment, the software programs of the present application may be executed by a processor to implement the steps or functions described above. Likewise, the software programs (including associated data structures) of the present application may be stored in a computer readable recording medium, such as RAM memory, magnetic or optical drive or diskette and the like. Additionally, some of the steps or functions of the present application may be implemented in hardware, for example, as circuitry that cooperates with the processor to perform various steps or functions.
In addition, some of the present application may be implemented as a computer program product, such as computer program instructions, which when executed by a computer, may invoke or provide methods and/or techniques in accordance with the present application through the operation of the computer. Program instructions which invoke the methods of the present application may be stored on a fixed or removable recording medium and/or transmitted via a data stream on a broadcast or other signal-bearing medium and/or stored within a working memory of a computer device operating in accordance with the program instructions. An embodiment according to the present application comprises a device comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein the computer program instructions, when executed by the processor, trigger the device to perform a method and/or a solution according to the aforementioned embodiments of the present application.
According to yet another aspect of the present application, there is also provided an apparatus for machine learning model sample acquisition for blood pressure prediction, wherein the apparatus comprises:
one or more processors; and
a memory storing computer readable instructions that, when executed, cause the processor to perform operations of the method as previously described.
For example, the computer readable instructions, when executed, cause the one or more processors to: acquiring electrocardiosignals and corresponding blood pressure values thereof; determining a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignals, wherein the first characteristic value of the electrocardiosignals comprises statistics of characteristic values of all heartbeat waveforms of the electrocardiosignals, the second characteristic value of the electrocardiosignals comprises statistics of characteristic values of all two adjacent heartbeat waveforms in each section of continuous normal electrocardiosignal waveforms of the electrocardiosignals, and the third characteristic value of the electrocardiosignals comprises a ratio of powers of a plurality of preset frequency bands in a power spectrum corresponding to the electrocardiosignals and an entropy value of the power spectrum; and determining the blood pressure value as a label of a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal, and determining the first characteristic value, the second characteristic value, the third characteristic value and the label of the electrocardiosignal as the machine learning model sample. Further, based on the obtained plurality of historical electrocardiosignals and corresponding blood pressure values thereof, the method is adopted to determine a plurality of machine learning model samples so as to construct a sample set of the machine learning model. Further, training the machine learning model based on the sample set, and determining the trained machine learning model as a blood pressure prediction model when the evaluation index of the machine learning model meets a preset threshold value. Further, determining a first characteristic value, a second characteristic value and a third characteristic value of the acquired electrocardiosignals; and inputting the first characteristic value, the second characteristic value and the third characteristic value of the electrocardiosignals into the blood pressure prediction model to predict the blood pressure value corresponding to the electrocardiosignals.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned. Furthermore, it is obvious that the word "comprising" does not exclude other elements or steps, and the singular does not exclude the plural. A plurality of units or means recited in the apparatus claims may also be implemented by one unit or means in software and/or hardware. The terms first, second, etc. are used to denote names, but not any particular order.
Claims (13)
1. A method of machine learning model sample acquisition for blood pressure prediction, the method comprising:
acquiring electrocardiosignals and corresponding blood pressure values thereof;
determining a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal;
and determining the blood pressure value as a label of a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal, and determining the first characteristic value, the second characteristic value, the third characteristic value and the label of the electrocardiosignal as the machine learning model sample.
2. The method of claim 1, wherein determining the first characteristic value of the cardiac signal comprises:
determining each heartbeat waveform in the electrocardiosignal;
acquiring a characteristic value of each heartbeat waveform based on each heartbeat waveform,
and traversing all heartbeat waveforms of the electrocardiosignals, acquiring statistics of characteristic values of all heartbeat waveforms, and determining the statistics as a first characteristic value of the electrocardiosignals.
3. The method of claim 2, wherein the characteristic value of each heartbeat waveform includes at least one of:
the P-wave duration of each heartbeat waveform;
PR intervals for each heartbeat waveform;
QRS time of each heartbeat waveform;
the ST segment duration of each heartbeat waveform;
the T-wave duration of each heartbeat waveform;
the duration of the TP segment of each heartbeat waveform;
QT interval of each heartbeat waveform;
the U-wave duration of each heartbeat waveform.
4. The method of claim 2, wherein the statistical values comprise at least one of:
root mean square;
the absolute deviation of the median;
the median coefficient of variation.
5. The method of claim 1, wherein determining the second characteristic value of the cardiac signal comprises:
determining each heartbeat waveform in the electrocardiosignal;
acquiring characteristic values of two adjacent heartbeat waveforms based on the two adjacent heartbeat waveforms,
and traversing all heartbeat waveforms of the electrocardiosignals, acquiring statistics of the characteristic values of all two adjacent heartbeat waveforms, and determining the statistics as a second characteristic value of the electrocardiosignals.
6. The method according to claim 5, wherein the feature values of the two adjacent heartbeat waveforms include at least one of:
the PP interval of every two adjacent heartbeat waveforms;
the QQ interval of every two adjacent heartbeat waveforms;
RR intervals of two adjacent heartbeat waveforms;
SS intervals of every two adjacent heartbeat waveforms;
TT intervals of every two adjacent heartbeat waveforms;
UU intervals of two adjacent heartbeat waveforms.
7. The method of claim 5, wherein the statistical values comprise at least one of:
root mean square;
the absolute deviation of the median;
a median coefficient of variation;
a number of consecutive interval differences greater than 50 milliseconds.
8. The method of claim 5, wherein determining the third characteristic value of the cardiac signal comprises:
acquiring a power spectrum of a time domain waveform corresponding to the second characteristic value;
acquiring power values of a plurality of preset frequency domains based on the power spectrum;
and determining the ratio of the power values of the plurality of preset frequency domains as a third characteristic value of the electrocardiosignal.
9. The method of claim 8, wherein determining a third characteristic value of the cardiac signal further comprises:
calculating an entropy value of the power spectrum based on the power spectrum;
and determining the entropy value of the power spectrum as a third characteristic value of the electrocardiosignal.
10. The method according to any one of claims 1 to 9, further comprising:
and determining a plurality of machine learning model samples based on a plurality of acquired historical electrocardiosignals and corresponding blood pressure values thereof so as to construct a sample set of the machine learning model.
11. The method of claim 10, further comprising:
and training the machine learning model based on the sample set, and determining the trained machine learning model as a blood pressure prediction model when the evaluation index of the machine learning model accords with a preset threshold value.
12. The method of claim 11, further comprising:
determining a first characteristic value, a second characteristic value and a third characteristic value of the obtained electrocardiosignals;
and inputting the first characteristic value, the second characteristic value and the third characteristic value of the electrocardiosignals into the blood pressure prediction model to predict the blood pressure value corresponding to the electrocardiosignals.
13. An apparatus for machine learning model sample acquisition for blood pressure prediction, the apparatus comprising:
the first module is used for acquiring electrocardiosignals and corresponding blood pressure values thereof;
the second module is used for determining a first characteristic value, a second characteristic value and a third characteristic value of the electrocardiosignal;
and the third module is used for determining the blood pressure value as a label of the first characteristic value, the second characteristic value and the third characteristic value of the electrocardiosignal, and determining the first characteristic value, the second characteristic value, the third characteristic value and the label of the electrocardiosignal as the machine learning model sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110997501.1A CN113854985A (en) | 2021-08-27 | 2021-08-27 | Method and device for obtaining machine learning model samples for blood pressure prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110997501.1A CN113854985A (en) | 2021-08-27 | 2021-08-27 | Method and device for obtaining machine learning model samples for blood pressure prediction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113854985A true CN113854985A (en) | 2021-12-31 |
Family
ID=78988646
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110997501.1A Pending CN113854985A (en) | 2021-08-27 | 2021-08-27 | Method and device for obtaining machine learning model samples for blood pressure prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113854985A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2024033658A1 (en) * | 2022-08-12 | 2024-02-15 | Digital & Future Technologies Limited | Continuous blood pressure monitor |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106413534A (en) * | 2015-08-08 | 2017-02-15 | 深圳先进技术研究院 | Blood-pressure continuous-measurement device, measurement model establishment method, and system |
| CN108420424A (en) * | 2017-02-14 | 2018-08-21 | 路提科技股份有限公司 | Blood pressure measuring method and device |
| CN108652604A (en) * | 2018-02-06 | 2018-10-16 | 北京大学深圳研究生院 | It is a kind of based on electrocardiosignal without air bag blood pressure detecting method and system |
-
2021
- 2021-08-27 CN CN202110997501.1A patent/CN113854985A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106413534A (en) * | 2015-08-08 | 2017-02-15 | 深圳先进技术研究院 | Blood-pressure continuous-measurement device, measurement model establishment method, and system |
| CN108420424A (en) * | 2017-02-14 | 2018-08-21 | 路提科技股份有限公司 | Blood pressure measuring method and device |
| CN108652604A (en) * | 2018-02-06 | 2018-10-16 | 北京大学深圳研究生院 | It is a kind of based on electrocardiosignal without air bag blood pressure detecting method and system |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2024033658A1 (en) * | 2022-08-12 | 2024-02-15 | Digital & Future Technologies Limited | Continuous blood pressure monitor |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11147500B2 (en) | Electrocardiogram processing system for delineation and classification | |
| Vest et al. | An open source benchmarked toolbox for cardiovascular waveform and interval analysis | |
| EP3692900B1 (en) | Artificial intelligence self-learning-based static electrocardiography analysis method and apparatus | |
| US20230143594A1 (en) | Systems and methods for reduced lead electrocardiogram diagnosis using deep neural networks and rule-based systems | |
| US8543194B2 (en) | System and method of detecting abnormal movement of a physical object | |
| Lake et al. | Accurate estimation of entropy in very short physiological time series: the problem of atrial fibrillation detection in implanted ventricular devices | |
| CN110495872B (en) | Electrocardiogram analysis method, device, equipment and medium based on picture and heartbeat information | |
| US20180235487A1 (en) | Method and system for cuffless blood pressure estimation using photoplethysmogram features and pulse transit time | |
| Silva et al. | Robust detection of heart beats in multimodal data | |
| Raj et al. | A personalized point-of-care platform for real-time ECG monitoring | |
| Kayikcioglu et al. | Time-frequency approach to ECG classification of myocardial infarction | |
| Stapelberg et al. | The sensitivity of 38 heart rate variability measures to the addition of artifact in human and artificial 24‐hr cardiac recordings | |
| Poddar et al. | Automated diagnosis of coronary artery diseased patients by heart rate variability analysis using linear and non-linear methods | |
| Antink et al. | Detection of heart beats in multimodal data: a robust beat-to-beat interval estimation approach | |
| US20150265174A1 (en) | Non-Invasive Evaluation of Cardiac Repolarisation Instability for Risk Stratification of Sudden Cardiac Death | |
| Zhang et al. | Interpretable detection and location of myocardial infarction based on ventricular fusion rule features | |
| US11672464B2 (en) | Electrocardiogram processing system for delineation and classification | |
| CN113854985A (en) | Method and device for obtaining machine learning model samples for blood pressure prediction | |
| CN113855007A (en) | Method and device for obtaining machine learning model samples for blood glucose prediction | |
| CN116504398A (en) | Methods and systems for arrhythmia prediction using a transducer-based neural network | |
| Yao et al. | Arrhythmia classification from single lead ecg by multi-scale convolutional neural networks | |
| Arenas et al. | Morphological and temporal ecg features for myocardial infarction detection using support vector machines | |
| Rodriguez et al. | Embedded system based on an ARM microcontroller to analyze heart rate variability in real time using wavelets | |
| Krasteva et al. | Delineation of 12-Lead ECG Representative Beats Using Convolutional Encoder–Decoders with Residual and Recurrent Connections | |
| Setiawan et al. | Development of cloud computing algorithm for arrhythmia detection |
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 |