CN118161171B - Hidden atrial fibrillation detection analysis system and method based on RR interval frequency domain parameters - Google Patents

Abstract

The present invention discloses a system and method for detecting and analyzing latent atrial fibrillation based on RR interval frequency domain parameters, including the following steps: collecting and saving the heart rate variability signal to be analyzed; performing frequency domain analysis and calculation on the RR interval of the heart rate variability digital signal to obtain the power values and spectrum envelope diagram of the frequency domain parameters LFnorm, HFnorm, and LF/HF; judging the risk level of latent atrial fibrillation and classifying it based on the power values and spectrum envelope diagram. The present invention collects the electrocardiogram signal to be analyzed, performs frequency domain analysis on the RR interval of the heart rate variability of the electrocardiogram signal, obtains the power values and spectrum envelope diagram of the frequency domain parameters respectively, and then analyzes and judges the risk level of latent atrial fibrillation based on the power values and spectrum envelope diagram of the frequency domain parameters respectively, thereby ensuring the accuracy of risk prediction and improving the sensitivity of monitoring.

Description

Latent atrial fibrillation detection and analysis system and method based on RR interval frequency domain parameters

Technical Field

The present invention relates to the technical field of electrocardiogram signal detection and analysis, and in particular to a system and method for detecting and analyzing latent atrial fibrillation based on RR interval frequency domain parameters.

Background Art

Atrial fibrillation is a rapid, irregular atrial rhythm with symptoms including palpitations, sometimes fatigue, decreased strength and presyncope. It may form atrial emboli and has a significant risk of causing embolic stroke. Diagnosis is based on electrocardiogram, as shown in Figure 1. Normal electrocardiograms and atrial fibrillation electrocardiograms can be distinguished by naked eyes. At this time, atrial fibrillation has reached a certain degree of severity. In the early stage of atrial fibrillation (mild), the wavelet of its tremor is only about 10 microvolts or even smaller. It is difficult to tell whether there is atrial fibrillation by electrocardiogram. This mild condition can be called latent atrial fibrillation. Once it is determined that mild atrial fibrillation has occurred, timely treatment can be given, including the use of drugs to control heart rate, the use of anticoagulants to prevent thromboembolism, and sometimes the use of drugs or cardioversion to convert atrial fibrillation to sinus rhythm.

There are five aspects of adverse consequences of atrial fibrillation:

(1) Discomfort: palpitations, shortness of breath, chest tightness, chest pain, dizziness, fatigue, insomnia, psychological pressure, tension, fear, etc.;

(2) Heart failure: decreased cardiac function, reduced diastolic function, bradycardia and heart failure;

(3) Stroke: Atrial fibrillation can easily cause blood clots to form, leading to embolism, which can easily block cerebral blood vessels and cause cerebral thrombosis and hemiplegia;

(4) Ischemia: If there is coronary heart disease and myocardial ischemia, atrial fibrillation will aggravate myocardial ischemia;

(5) Sudden death: Atrial fibrillation is a serious arrhythmia, and the sudden death rate of arrhythmia is generally 5-6 times higher than that of the general population;

According to data from the China National Stroke Registry, the recurrence rate, disability rate, and mortality rate of stroke in patients with atrial fibrillation are approximately 32.35%, 51.58%, and 34.23%, respectively. The risk of recurrence in stroke patients with atrial fibrillation is 3.7 times higher than that in patients without atrial fibrillation. The Alliance of Atrial Fibrillation Centers conducted an epidemiological survey on atrial fibrillation from July 2020 to September 2021. The results showed that the prevalence of atrial fibrillation in China was 1.6%. In opportunistic screening of nearly 600,000 people, the number of patients with atrial fibrillation in China is currently as high as 20 million. According to the survey, among patients diagnosed with atrial fibrillation, 36% did not know that they had atrial fibrillation.

Therefore, it is urgent to solve the above problems.

Summary of the invention

Purpose of the invention: The first purpose of the present invention is to provide a latent atrial fibrillation detection and analysis method based on RR interval frequency domain parameters that can accurately monitor the risk level of latent atrial fibrillation.

The second object of the present invention is to provide a latent atrial fibrillation detection and analysis system based on RR interval frequency domain parameters.

Technical solution: To achieve the above objectives, the present invention discloses a method for detecting and analyzing latent atrial fibrillation based on RR interval frequency domain parameters, comprising the following steps:

(1) collecting and saving the heart rate variability signal to be analyzed;

(2) Perform frequency domain analysis and calculation on the RR interval of the heart rate variability digital signal, calculate the power values of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF by the average cycle method, and obtain the spectrum envelope diagram of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF of the ECG data to be analyzed by the AR model method;

(3) If the power value of the frequency domain parameter LF/HF is within the first threshold range, it indicates that the risk level of latent atrial fibrillation is low risk; if the power value of the frequency domain parameter LF/HF is within the second threshold range, it indicates that the risk level of latent atrial fibrillation is medium risk; if the power value of the frequency domain parameter LF/HF is within the third threshold range, it indicates that the risk level of latent atrial fibrillation is high risk; or if the spectrum envelope diagram of the heart rate variability signal to be analyzed is within the first standard group diagram, it indicates that the risk level of latent atrial fibrillation is low risk; if the spectrum envelope diagram of the heart rate variability signal to be analyzed is within the second standard group diagram, it indicates that the risk level of latent atrial fibrillation is medium risk; if the spectrum envelope diagram of the heart rate variability signal to be analyzed is within the third standard group diagram, it indicates that the risk level of latent atrial fibrillation is high risk.

Preferably, the specific steps of calculating the power values of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF by the average cycle method in step (2) are:

The power spectral density PSD is calculated by performing fast Fourier transform on the RR interval data, where the power spectral density PSD can be expressed as a function of frequency, denoted as P(f), as follows:

In the formula, x(n) (n = 0, 1, ..., N-1) represents the RR interval sequence of the heart beat, N is the length of the RR interval sequence to be analyzed, usually N = 256 or N = 512; Δt is the average sampling interval of the RR interval sequence, that is, the average value of the RR interval sequence in this section X(f) is the discrete Fourier transform of x(n);

The power spectrum density PSD is divided into the following four frequency bands: ultra-low frequency band ULF: <0.003HZ, that is, PSD _ULF , very low frequency band VLF: 0.003～0.04HZ, that is, PSD _VLF , low frequency band LF: 0.04～0.15HZ, that is, PSD _LF , high frequency band HF0.15～0.40HZ, that is, PSD _HF , total frequency band: ≤0.4HZ, that is, PSD _total ;

The power spectral density of the total frequency band PSD _total = PSD _ULF + PSD _LF + PSD _HF + PSD _VLF ,

LFnorm＝100×PSD _LF /(PSD _total －PSD _VLF ),

HFnorm＝100×PSD _HF /(PSD _total －PSD _VLF ),

LF/HF＝PSD _LF /PSD _HF .

Furthermore, the specific steps of obtaining the spectrum envelope diagram of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF by the AR model method in step (2) are as follows:

The difference equation of the AR model, i.e. the autoregressive model, can be expressed as:

Where x(n) is the RR interval sequence of the heart beat, u(n) is a white noise sequence with a mean of zero and a variance of σ ² ; p is the model order, p = 15, a _k is the coefficient of the AR model, k = 1, 2, 3, ..., p;

According to the autocorrelation function of x(n), the canonical equation of the AR model, namely the Yule-Walker equation, is constructed. After solving it, the estimated values of each coefficient a _k and σ ² are obtained; then the final power spectrum P _AR (f) of x(n) can be expressed as:

The envelope diagram of the ECG data to be analyzed, which is a power density spectrum curve of the RR interval every 5 minutes, is obtained, that is, the spectrum envelope diagram of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF.

Furthermore, the method for selecting the determined value of the model order p in step (2) is as follows: performing frequency domain analysis and calculation on the RR interval of the heart rate variability digital signal, calculating the power values of the frequency domain parameters LFnorm, HFnorm, and LF/HF of the autonomic nervous function by the average period method, and drawing the corresponding power density spectrum curve; setting the initial value of the model order p, and obtaining the envelope diagram of the power density spectrum curve of the RR interval by the AR model method, that is, the spectrum diagram of the frequency domain parameters LFnorm, HFnorm, and LF/HF of the autonomic nervous function; if the power density spectrum curve matches the spectrum envelope diagram, then the model order p is the determined value; if the power density spectrum curve does not match the spectrum envelope diagram, then adjusting the value of the model order p, and re-obtaining the envelope diagram of the power density spectrum curve of the RR interval by the AR model method, until the power density spectrum curve matches the spectrum envelope diagram, and obtaining the determined value of the model order p.

Among them, the first threshold is 0.4-3, the second threshold is 0.2-0.4, and the third threshold is <0.2.

Furthermore, the method for selecting the first standard group graph in step (3) is as follows: a heart rate variability signal whose power value of the frequency domain parameter LF/HF is within the first threshold range is selected as the first standard heart rate variability digital signal, the standard heart rate variability digital signal is input into the AR model to obtain a cluster of standard AR model graphs, and then several graphs with a similarity greater than 0.7 to the standard AR model graph are selected to form the first standard group graph.

Preferably, the method for selecting the second standard group graph in step (3) is: selecting a heart rate variability signal whose power value of the frequency domain parameter LF/HF is within the second threshold range as the second standard heart rate variability digital signal, inputting the standard heart rate variability digital signal into the AR model to obtain a cluster of standard AR model graphs, and then selecting several graphs with a similarity degree greater than 0.7 to the standard AR model graph to form a second standard group graph.

Furthermore, the method for selecting the third standard group graph in step (3) is as follows: a heart rate variability signal whose power value of the frequency domain parameter LF/HF is within the third threshold range is selected as the third standard heart rate variability digital signal, the standard heart rate variability digital signal is input into the AR model to obtain a cluster of standard AR model graphs, and then several graphs with a similarity greater than 0.7 to the standard AR model graph are selected to form a third standard group graph.

The present invention provides a latent atrial fibrillation detection and analysis system based on RR interval frequency domain parameters, comprising a signal acquisition and storage module, which is used to acquire and store the heart rate variability signal to be analyzed;

The frequency domain calculation module performs frequency domain analysis and calculation on the RR interval of the heart rate variability digital signal, calculates the power values of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF by the average cycle method, and obtains the spectrum envelope diagram of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF of the ECG data to be analyzed by the AR model method;

Risk judgment module, if the power value of the frequency domain parameter LF/HF is within the first threshold range, it indicates that the risk level of latent atrial fibrillation is low risk, if the power value of the frequency domain parameter LF/HF is within the second threshold range, it indicates that the risk level of latent atrial fibrillation is medium risk, if the power value of the frequency domain parameter LF/HF is within the third threshold range, it indicates that the risk level of latent atrial fibrillation is high risk; or if the frequency spectrum envelope diagram of the heart rate variability signal to be analyzed is within the first standard group diagram, it indicates that the risk level of latent atrial fibrillation is low risk, if the frequency spectrum envelope diagram of the heart rate variability signal to be analyzed is within the second standard group diagram, it indicates that the risk level of latent atrial fibrillation is medium risk, if the frequency spectrum envelope diagram of the heart rate variability signal to be analyzed is within the third standard group diagram, it indicates that the risk level of latent atrial fibrillation is high risk.

Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: the present invention collects the ECG signal to be analyzed, performs frequency domain analysis on the ECG signal, obtains the power value and spectrum envelope diagram of the frequency domain parameters respectively, and then analyzes and judges the risk level of latent atrial fibrillation based on the power value and spectrum envelope diagram of the frequency domain parameters respectively, thereby ensuring the accuracy of risk prediction and improving the sensitivity of monitoring.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram showing a comparison between a normal electrocardiogram and an atrial fibrillation electrocardiogram in the present invention;

FIG2 is a schematic diagram of an ECG waveform in the present invention;

FIG3 is a schematic diagram of the FFT power spectrum of the HRV signal in a 5-minute period in the present invention;

FIG4 is a schematic diagram of an AR model curve and FFT power spectrum of a 5-minute HRV signal in the present invention;

FIG5 is a schematic diagram of an AR model curve of a 5-minute HRV signal in the present invention;

FIG6 is a power spectrum curve diagram of an AR model obtained from an electrocardiogram signal of a 5-minute test period in a simulation experiment of the present invention;

FIG. 7 is a schematic diagram of the flow of time-frequency analysis in the present invention.

DETAILED DESCRIPTION

The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

Example 1: The human heart rate (in beats per minute, or bpm) is the number of heartbeats in 1 minute. Electrocardiogram (ECG) is a physiological method for recording the electrical activity of the heart. The individual heartbeat waves in the ECG signal show the change of electrical potential, and its components are called P wave, QRS complex and T wave, as shown in Figure 2. The human heart rate is not absolutely regular, and there are tens of milliseconds or even more than one hundred milliseconds of time difference between two heartbeats. The change of the heartbeat interval of a healthy person is due to the change of the sympathetic and parasympathetic nerves with factors such as breathing, which is a normal phenomenon. Heart rate variability (HRV) refers to the time difference between successive heartbeats, which is usually defined as the slight fluctuation of the R-R interval of successive heartbeats. The R-R interval is analyzed and calculated from the ECG data, so the heart rate variability signal R-R interval can reflect the activity of the autonomic nerves. The patient group compared the resting state of the normal group with the psychophysiological indexes, heart rate (HR), low frequency (LF) in the heart rate variability power spectrum, and low frequency high frequency ratio (LF/HF) in the patient group were all higher than those in the control group. The electrocardiogram of normal people in a healthy state shows obvious periodicity. However, the results of careful observation and measurement of the electrocardiogram show that the electrocardiogram of a normal person also fluctuates slowly and slightly within a certain range, and two adjacent RR intervals (cardiac cycles) usually differ by tens of milliseconds or even more than one hundred milliseconds. Heart rate variability (HRV) refers to the slight change of instantaneous heart rate over time, which is usually defined as the slight fluctuation of the RR interval of a continuous normal heartbeat or the instantaneous heart rate. A sequence composed of a series of continuous RR intervals or instantaneous heart rates of the electrocardiogram can be used to record the changes of heart rate over time, so it is also called a heart rate variability signal. The heart rate variability signal of the present invention adopts a RR interval sequence. The autonomy of the heart is an inherent characteristic of the cardiac pacemaker tissue, but the heart rate and rhythm are largely regulated by the ANS (Autonomic Nerve System). The PPS (Parasympathetic System) affects the heart rate by releasing acetylcholine; the SPS (Sympathetic System) affects the heart rate by releasing adrenaline and noradrenaline.

Embodiment 1 discloses a method for detecting and analyzing latent atrial fibrillation based on RR interval frequency domain parameters, comprising the following steps:

(1) collecting and saving the heart rate variability signal to be analyzed; specifically, performing single-lead ECG signal detection on the subject for 15 to 45 minutes to obtain the HRV RR interval signal;

(2) Perform frequency domain analysis and calculation on the RR interval of the heart rate variability digital signal, calculate the power values of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF by the average cycle method, and obtain the spectrum envelope diagram of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF of the ECG data to be analyzed by the AR model method;

(3) If the power value of the frequency domain parameter LF/HF is within the first threshold range, which is 0.4 to 3, it indicates that the risk level of latent atrial fibrillation is low risk; if the power value of the frequency domain parameter LF/HF is within the second threshold range, which is 0.2 to 0.4, it indicates that the risk level of latent atrial fibrillation is medium risk; if the power value of the frequency domain parameter LF/HF is within the third threshold range, which is <0.2, it indicates that the risk level of latent atrial fibrillation is high risk; or if the spectrum envelope of the heart rate variability signal to be analyzed is within the first standard group diagram, it indicates that the risk level of latent atrial fibrillation is low risk; if the spectrum envelope of the heart rate variability signal to be analyzed is within the second standard group diagram, it indicates that the risk level of latent atrial fibrillation is medium risk; if the spectrum envelope of the heart rate variability signal to be analyzed is within the third standard group diagram, it indicates that the risk level of latent atrial fibrillation is high risk; as the amount of data measured in practice increases, the threshold value can be changed through statistical analysis;

Frequency domain analysis is to convert the heart rate variability signal in the time domain into the frequency domain to obtain the power spectrum. The present invention uses two methods to obtain the power spectrum: the average period method and the AR model method. The average period method uses the fast Fourier transform FFT to obtain the power value, and the AR model method obtains a smooth power spectrum line close to the FFT. The average period method is based on FFT. The average period method makes an assumption of period extension for the signal, and longer data is required to obtain a higher spectral resolution. Because the heart rate variability signal records the RR intervals of successive heartbeats, it is a non-uniform sampling. In the actual FFT processing, a relatively long period of successive RR intervals is often regarded as a uniform time series, and the sampling rate is defined as is the average RR interval of this data. If the collected RR interval of heart rate data is transformed by FFT with a step length of 5 minutes, the FFT power spectrum of the HRV signal frequency domain in this 5-minute period can be obtained, as shown in Figure 3.

The specific steps of calculating the power values of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF by the average cycle method in step (3) are as follows:

The power spectral density PSD is calculated by performing fast Fourier transform on the RR interval data, where the power spectral density PSD can be expressed as a function of frequency, denoted as P(f), as follows:

In the formula, x(n) (n = 0, 1, ..., N-1) represents the RR interval sequence of the heart beat, N is the length of the RR interval sequence to be analyzed, usually N = 256 or N = 512; Δt is the average sampling interval of the RR interval sequence, that is, the average value of the RR interval sequence in this section X(f) is the discrete Fourier transform of x(n);

The power spectrum density PSD is divided into the following four frequency bands: ultra-low frequency band ULF: <0.003HZ, that is, PSD _ULF , very low frequency band VLF: 0.003～0.04HZ, that is, PSD _VLF , low frequency band LF: 0.04～0.15HZ, that is, PSD _LF , high frequency band HF0.15～0.40HZ, that is, PSD _HF , total frequency band: ≤0.4HZ, that is, PSD _total ;

The power spectral density of the total frequency band PSD _total = PSD _ULF + PSD _LF + PSD _HF + PSD _VLF ,

LFnorm＝100×PSD _LF /(PSD _total －PSD _VLF ),

HFnorm＝100×PSD _HF /(PSD _total －PSD _VLF ),

LF/HF＝PSD _LF /PSD _HF ;

The high-frequency HF component in the heart rate variability signal mainly reflects the activity of the cardiac vagus nerve, and the low-frequency LF component reflects the activity of the cardiac sympathetic nerve or the joint activity of the sympathetic vagus nerve. Therefore, the ratio of the low- and high-frequency components of the signal LF/HF reflects the balance of sympathetic and vagal activities.

The AR model method establishes a parameter model for the heart rate signal and reflects the characteristics of the signal by transferring the parameters of the function. The AR model method belongs to the modern spectrum estimation method, which requires short data and high resolution. Compared with the spectrum of FFT analysis, the AR model curve is like a smooth envelope, as shown in the smooth curve in Figure 4 and the curve in Figure 5. Moreover, as the order of the AR model increases, the curve of the AR model fluctuates more and is closer to the spectrum line of FFT, as shown in the uneven curve in Figure 4. This shows that the results of the two analysis methods are very consistent, so for the frequency domain analysis of non-stationary heart rate variability signals, the AR model is often used to draw the spectrum line of the analysis, and the FFT method is used to calculate the power spectrum density value.

The specific steps of obtaining the spectrum envelope diagram of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF by the AR model method in step (3) are as follows:

The difference equation of the AR model, i.e. the autoregressive model, can be expressed as:

Where x(n) is the RR interval sequence of the heart beat, u(n) is a white noise sequence with a mean of zero and a variance of σ ² ; p is the model order, p = 15, a _k is the coefficient of the AR model, k = 1, 2, 3, ..., p;

According to the autocorrelation function of x(n), the canonical equation of the AR model, namely the Yule-Walker equation, is constructed. After solving it, the estimated values of each coefficient a _k and σ ² are obtained; then the final power spectrum P _AR (f) of x(n) can be expressed as:

Obtaining the envelope diagram of the ECG data to be analyzed, which is a power density spectrum curve of the RR interval every 5 minutes, is the spectrum envelope diagram of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF;

The method for selecting the determined value of the model order p in step (3) is as follows: performing frequency domain analysis and calculation on the RR interval of the heart rate variability digital signal, calculating the power values of the frequency domain parameters LFnorm, HFnorm, and LF/HF of the autonomic nervous function by the average period method, and drawing the corresponding power density spectrum curve; setting the initial value of the model order p, and obtaining the envelope diagram of the power density spectrum curve of the RR interval by the AR model method, that is, the spectrum diagram of the frequency domain parameters LFnorm, HFnorm, and LF/HF of the autonomic nervous function; if the power density spectrum curve matches the spectrum envelope diagram, then the model order p is the determined value; if the power density spectrum curve does not match the spectrum envelope diagram, then adjusting the value of the model order p, and re-obtaining the envelope diagram of the power density spectrum curve of the RR interval by the AR model method, until the power density spectrum curve matches the spectrum envelope diagram, and obtaining the determined value of the model order p.

The method for selecting the first standard group graph is as follows: a heart rate variability signal whose power value of the frequency domain parameter LF/HF is within the first threshold range is selected as the first standard heart rate variability digital signal, the standard heart rate variability digital signal is input into the AR model to obtain a cluster of standard AR model graphs, and then several graphs with a similarity greater than 0.7 to the standard AR model graphs are selected to form the first standard group graph.

The method for selecting the second standard group graph is as follows: a heart rate variability signal whose power value of the frequency domain parameter LF/HF is within the second threshold range is selected as the second standard heart rate variability digital signal, the standard heart rate variability digital signal is input into the AR model to obtain a cluster of standard AR model graphs, and then several graphs with a similarity greater than 0.7 to the standard AR model graph are selected to form the second standard group graph.

The method for selecting the third standard group graph is as follows: a heart rate variability signal whose power value of the frequency domain parameter LF/HF is within the third threshold range is selected as the third standard heart rate variability digital signal, the standard heart rate variability digital signal is input into the AR model to obtain a cluster of standard AR model graphs, and then several graphs with a similarity degree greater than 0.7 to the standard AR model graph are selected to form the third standard group graph.

The index thresholds recommended in the present invention are shown in Tables 1 and 2 below.

Table 1 Threshold values of frequency domain analysis parameters of autonomic nervous function in short time (5 min)

Table 2 Thresholds for risk level of latent atrial fibrillation

The threshold of the short-term (5 minutes) frequency domain parameter can determine the risk level of latent atrial fibrillation in the subject as long as it meets the above-mentioned frequency threshold LF/HF range.

Autoregressive analysis belongs to the modern spectrum estimation method, which uses parameter models to separate and organically combine the randomness of random processes and a certain degree of predictability. The excitation white noise reflects the randomness of the process, and the deterministic model reflects the predictability of the process. Therefore, it can better summarize the properties of random signals. The power spectrum obtained is a continuous function of frequency, avoiding the random fluctuations of periodic spectrum estimation, and better spectrum estimation can be obtained with relatively short data, which is very beneficial for signals with strong non-stationarity such as heart rate variability.

Compared with the spectrum analyzed by FFT, the AR model curve is like a smooth envelope, and as the order of the AR model increases, the AR model curve fluctuates more and is closer to the FFT spectrum. This shows that the results of the two analysis methods are very consistent. Because the heart rate variability signal is a relatively non-stationary data, it is more appropriate to use the AR model to analyze the spectrum.

There is a problem of determining the order when using the AR model to calculate the spectrum. For heart rate variability analysis, the 15th order is most appropriate. If the order is too high, the peak line will split and false spectrum peaks will appear; if the order is too low, the spectrum line cannot correctly reflect the changes in the spectrum peak. The order can be adjusted for different data. There are three algorithms for the AR model: L-D algorithm, Burg algorithm and Marple algorithm. Among them, the Marple algorithm is the fastest and has the smallest error, but the results of the three algorithms for heart rate variability analysis are basically the same.

The power spectrum of human heart rate variability generally ranges from 0 to 0.5 Hz, and it can have three peaks, as shown in Figure 5. Due to the influence of environmental conditions and different body positions, the peaks can also change or temporarily disappear:

(1) The first peak is around 0.03 Hz (0.02-0.09), which is a low-frequency peak and is related to multiple factors such as peripheral vasoconstriction, temperature regulation, and renin-angiotensin system activity;

(2) The second peak is around 0.10 Hz (0.09-0.15), which is a mid-frequency peak and is related to blood pressure regulation;

(3) The third peak is around 0.25 Hz (0.15-0.40), which is a high-frequency peak and is related to the heart rate changes caused by the respiratory cycle;

In the autonomic nervous system regulation mechanism, the energy changes of low and medium frequency peaks are affected by both the sympathetic nerves and the vagus nerves, while the high frequency peaks are only affected by the vagus nerves. In clinical practice, 0.15Hz is often used as the boundary to divide it into low-frequency component LF and high-frequency component HF. The LF component is affected by both the sympathetic nerves and the vagus nerves, with the sympathetic nerves being dominant. The HF component reflects the activity of the vagus nerve. As shown in Figure 6, a middle-aged person tested the electrocardiogram signal for a period of 5 minutes to obtain an AR model power spectrum curve, and its LF/HF value is 0.39, which is used to assess the risk level of latent atrial fibrillation, which is medium risk.

Embodiment 2: Embodiment 2 discloses a latent atrial fibrillation detection and analysis system based on RR interval frequency domain parameters, including a signal acquisition and storage module for acquiring and storing the heart rate variability signal to be analyzed; specifically, a single-lead electrocardiogram signal detection can be performed on the subject for 15 to 45 minutes to obtain the HRV RR interval signal;

The frequency domain calculation module performs frequency domain analysis and calculation on the RR interval of the heart rate variability digital signal, calculates the power values of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF by the average cycle method, and obtains the spectrum envelope diagram of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF of the ECG data to be analyzed by the AR model method;

Risk judgment module, if the power value of the frequency domain parameter LF/HF is within the first threshold range, it indicates that the risk level of latent atrial fibrillation is low risk, if the power value of the frequency domain parameter LF/HF is within the second threshold range, it indicates that the risk level of latent atrial fibrillation is medium risk, if the power value of the frequency domain parameter LF/HF is within the third threshold range, it indicates that the risk level of latent atrial fibrillation is high risk; or if the frequency spectrum envelope diagram of the heart rate variability signal to be analyzed is within the first standard group diagram, it indicates that the risk level of latent atrial fibrillation is low risk, if the frequency spectrum envelope diagram of the heart rate variability signal to be analyzed is within the second standard group diagram, it indicates that the risk level of latent atrial fibrillation is medium risk, if the frequency spectrum envelope diagram of the heart rate variability signal to be analyzed is within the third standard group diagram, it indicates that the risk level of latent atrial fibrillation is high risk.

The specific steps for calculating the power values of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF by the average cycle method in the risk judgment module are as follows:

The power spectral density PSD is calculated by performing fast Fourier transform on the RR interval data, where the power spectral density PSD can be expressed as a function of frequency, denoted as P(f), as follows:

In the formula, x(n) (n = 0, 1, ..., N-1) represents the RR interval sequence of the heart beat, N is the length of the RR interval sequence to be analyzed, usually N = 256 or N = 512; Δt is the average sampling interval of the RR interval sequence, that is, the average value of the RR interval sequence in this section X(f) is the discrete Fourier transform of x(n);

The power spectrum density PSD is divided into the following four frequency bands: ultra-low frequency band ULF: <0.003HZ, that is, PSD _ULF , very low frequency band VLF: 0.003～0.04HZ, that is, PSD _VLF , low frequency band LF: 0.04～0.15HZ, that is, PSD _LF , high frequency band HF0.15～0.40HZ, that is, PSD _HF , total frequency band: ≤0.4HZ, that is, PSD _total ;

The power spectral density of the total frequency band PSD _total = PSD _ULF + PSD _LF + PSD _HF + PSD _VLF ,

LFnorm＝100×PSD _LF /(PSD _total －PSD _VLF ),

HFnorm＝100×PSD _HF /(PSD _total －PSD _VLF ),

LF/HF＝PSD _LF /PSD _HF ;

The specific steps of obtaining the spectrum envelope diagram of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF by the AR model method in the risk judgment module are as follows:

The difference equation of the AR model, i.e. the autoregressive model, can be expressed as:

Where x(n) is the RR interval sequence of the heart beat, u(n) is a white noise sequence with a mean of zero and a variance of σ ² ; p is the model order, p = 15, a _k is the coefficient of the AR model, k = 1, 2, 3, ..., p;

According to the autocorrelation function of x(n), the canonical equation of the AR model, namely the Yule-Walker equation, is constructed. After solving it, the estimated values of each coefficient a _k and σ ² are obtained; then the final power spectrum P _AR (f) of x(n) can be expressed as:

Obtaining the envelope diagram of the ECG data to be analyzed, which is a power density spectrum curve of the RR interval every 5 minutes, is the spectrum envelope diagram of the autonomic nervous function frequency domain parameters LFnorm, HFnorm, and LF/HF;

The method for selecting the determined value of the model order p in the risk judgment module is as follows: the RR interval of the heart rate variability digital signal is analyzed and calculated in the frequency domain, the power values of the frequency domain parameters LFnorm, HFnorm, and LF/HF of the autonomic nervous function are calculated by the average period method, and the corresponding power density spectrum curve is drawn; the initial value of the model order p is set, and the envelope diagram of the power density spectrum curve of the RR interval is obtained by the AR model method, that is, the spectrum diagram of the frequency domain parameters LFnorm, HFnorm, and LF/HF of the autonomic nervous function; if the power density spectrum curve matches the spectrum envelope diagram, the model order p is the determined value; if the power density spectrum curve does not match the spectrum envelope diagram, the value of the model order p is adjusted, and the envelope diagram of the power density spectrum curve of the RR interval is obtained again by the AR model method until the power density spectrum curve matches the spectrum envelope diagram, and the determined value of the model order p is obtained.

The method for selecting the first standard group graph is as follows: a heart rate variability signal whose power value of the frequency domain parameter LF/HF is within the first threshold range is selected as the first standard heart rate variability digital signal, the standard heart rate variability digital signal is input into the AR model to obtain a cluster of standard AR model graphs, and then several graphs with a similarity greater than 0.7 to the standard AR model graphs are selected to form the first standard group graph.

The method for selecting the second standard group graph is as follows: a heart rate variability signal whose power value of the frequency domain parameter LF/HF is within the second threshold range is selected as the second standard heart rate variability digital signal, the standard heart rate variability digital signal is input into the AR model to obtain a cluster of standard AR model graphs, and then several graphs with a similarity greater than 0.7 to the standard AR model graph are selected to form the second standard group graph.

The method for selecting the third standard group graph is as follows: a heart rate variability signal whose power value of the frequency domain parameter LF/HF is within the third threshold range is selected as the third standard heart rate variability digital signal, the standard heart rate variability digital signal is input into the AR model to obtain a cluster of standard AR model graphs, and then several graphs with a similarity degree greater than 0.7 to the standard AR model graph are selected to form the third standard group graph.

Claims (5)

1. The hidden atrial fibrillation detection and analysis method based on RR interval frequency domain parameters is characterized by comprising the following steps of:

(1) Collecting heart rate variability signals to be analyzed and storing;

(2) Carrying out frequency domain analysis and calculation on the heart rate variability digital signal RR interval, calculating the power value of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF by an average period method, and obtaining a spectrum envelope diagram of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF of the electrocardio data to be analyzed by an AR model method;

The specific steps of calculating the power value of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF by the average cycle method in the step (2) are as follows:

by performing a fast fourier transform on the data of the RR intervals, a power spectral density PSD is calculated, which can be expressed as a function of frequency, recorded as The expression is as follows:，

in the method, in the process of the invention, (N=0, 1, … …, N-1) represents the RR interval sequence of the heart beat,For the RR interval sequence length to be analyzed,=256 Or=512；Is the average sampling interval of RR interval sequence, i.e. the average value of RR interval sequence of the section，Is thatIs a discrete fourier transform of (a);

The power spectral density PSD is divided into the following four bands: ultra low band ULF: <0.003HZ, i.e., PSD _ULF, very low band VLF: 0.003-0.04 HZ is PSD _VLF, low frequency band LF: 0.04-0.15 HZ, namely PSD _LF, and 0.15-0.40 HZ of high-frequency band HF, namely PSD _HF, and total frequency band: less than or equal to 0.4HZ, namely PSD _{Total (S)} ;

the power spectral density PSD _{Total (S)} ＝PSD_ULF＋PSD_LF＋PSD_HF＋PSD_VLF of the total frequency band,

LFnorm＝100×PSD_LF/(PSD _{Total (S)} －PSD_VLF)，

HFnorm＝100×PSD_HF/(PSD _{Total (S)} －PSD_VLF)，

LF/HF＝PSD_LF/PSD_HF；

The specific steps of obtaining the spectrum envelope map of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF by the AR model method in the step (2) are as follows:

the differential equation of the AR model, i.e., the autoregressive model, can be expressed as: ，

in the method, in the process of the invention, Is an RR interval sequence of the heart beat,Is a mean value of zero and a variance ofWhite noise sequences of (a); p is the model order, taking p=15,K=1, 2,3, …, p, coefficients of AR model; n is the sequence number n=0, 1, … …, N-1 of the RR interval sequence;

According to The autocorrelation function of (a) constructs a regular equation of an AR model, namely a Yule-Walker equation, and each coefficient is obtained after solvingAndIs a function of the estimated value of (2); then finallyPower spectrum of (2)Can be expressed as:，

Obtaining an envelope map of the electrocardio data to be analyzed of a power density spectrum curve every 5 minutes in the RR interval, namely a spectrum envelope map of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF;

The method for selecting the model order p determination value in the step (2) comprises the following steps: carrying out frequency domain analysis and calculation on the heart rate variability digital signal RR interval, calculating the power value of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF by an average period method, and drawing a corresponding power density spectrum graph; setting an initial value of a model order p, and obtaining an envelope graph of a power density spectrum curve of an RR interval, namely a frequency spectrum graph of an autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF by an AR model method; if the power density spectrum graph is matched with the spectrum envelope graph, the model order p is a determined value, if the power density spectrum graph is not matched with the spectrum envelope graph, the numerical value of the model order p is adjusted, the envelope graph of the power density spectrum graph at the RR interval is obtained again through an AR model method until the power density spectrum graph is matched with the spectrum envelope graph, and the determined value of the model order p is obtained;

(3) If the spectrum envelope of the heart rate variability signal to be analyzed is in the first standard population diagram, the risk level of the hidden atrial fibrillation is indicated to be low risk, if the spectrum envelope of the heart rate variability signal to be analyzed is in the second standard population diagram, the risk level of the hidden atrial fibrillation is indicated to be medium risk, and if the spectrum envelope of the heart rate variability signal to be analyzed is in the third standard population diagram, the risk level of the hidden atrial fibrillation is indicated to be high risk.

2. The method for detecting and analyzing the occult atrial fibrillation based on the RR interval frequency domain parameters according to claim 1, wherein the method comprises the following steps of: the selection method of the first standard group graph in the step (3) comprises the following steps: and selecting a heart rate variation signal with the power value of the frequency domain parameter LF/HF within a first threshold range as a first standard heart rate variation digital signal, inputting the standard heart rate variation digital signal into an AR model to obtain a cluster of standard AR model graphs, and selecting a plurality of graphs with the similarity degree with the standard AR model graphs being more than 0.7 to form a first standard group graph.

3. The method for detecting and analyzing the occult atrial fibrillation based on the RR interval frequency domain parameters as claimed in claim 2, wherein the method comprises the following steps: the selection method of the second standard population diagram in the step (3) comprises the following steps: and selecting a heart rate variation signal with the power value of the frequency domain parameter LF/HF within a second threshold range as a second standard heart rate variation digital signal, inputting the standard heart rate variation digital signal into an AR model to obtain a cluster of standard AR model graphs, and selecting a plurality of graphs with the similarity degree with the standard AR model graphs being more than 0.7 to form a second standard population graph.

4. The method for detecting and analyzing the occult atrial fibrillation based on the RR interval frequency domain parameters according to claim 3, wherein the method comprises the following steps of: the selection method of the third standard population diagram in the step (3) comprises the following steps: and selecting a heart rate variation signal with the power value of the frequency domain parameter LF/HF within a third threshold range as a third standard heart rate variation digital signal, inputting the standard heart rate variation digital signal into an AR model to obtain a cluster of standard AR model graphs, and selecting a plurality of graphs with the similarity degree with the standard AR model graphs being more than 0.7 to form a third standard population graph.

5. The hidden atrial fibrillation detection and analysis system based on RR interval frequency domain parameters is characterized by comprising a signal acquisition and storage module, a signal analysis module and a signal analysis module, wherein the signal acquisition and storage module is used for acquiring and storing heart rate variability signals to be analyzed;

the frequency domain calculation module is used for carrying out frequency domain analysis calculation on the heart rate variability digital signal RR interval, calculating the power value of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF by an average period method, and obtaining a frequency spectrum envelope diagram of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF of the electrocardio data to be analyzed by an AR model method;

the power value of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF is calculated by an average period method specifically comprises the following steps:

by performing a fast fourier transform on the data of the RR intervals, a power spectral density PSD is calculated, which can be expressed as a function of frequency, recorded as The expression is as follows:，

in the method, in the process of the invention, (N=0, 1, … …, N-1) represents the RR interval sequence of the heart beat,For the RR interval sequence length to be analyzed,=256 Or=512；Is the average sampling interval of RR interval sequence, i.e. the average value of RR interval sequence of the section，Is thatIs a discrete fourier transform of (a);

The power spectral density PSD is divided into the following four bands: ultra low band ULF: <0.003HZ, i.e., PSD _ULF, very low band VLF: 0.003-0.04 HZ is PSD _VLF, low frequency band LF: 0.04-0.15 HZ, namely PSD _LF, and 0.15-0.40 HZ of high-frequency band HF, namely PSD _HF, and total frequency band: less than or equal to 0.4HZ, namely PSD _{Total (S)} ;

the power spectral density PSD _{Total (S)} ＝PSD_ULF＋PSD_LF＋PSD_HF＋PSD_VLF of the total frequency band,

LFnorm＝100×PSD_LF/(PSD _{Total (S)} －PSD_VLF)，

HFnorm＝100×PSD_HF/(PSD _{Total (S)} －PSD_VLF)，

LF/HF＝PSD_LF/PSD_HF；

The spectrum envelope map of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF obtained by the AR model method is specifically:

the differential equation of the AR model, i.e., the autoregressive model, can be expressed as: ，

in the method, in the process of the invention, Is an RR interval sequence of the heart beat,Is a mean value of zero and a variance ofWhite noise sequences of (a); p is the model order, taking p=15,K=1, 2,3, …, p, coefficients of AR model; n is the sequence number n=0, 1, … …, N-1 of the RR interval sequence;

According to The autocorrelation function of (a) constructs a regular equation of an AR model, namely a Yule-Walker equation, and each coefficient is obtained after solvingAndIs a function of the estimated value of (2); then finallyPower spectrum of (2)Can be expressed as:，

Obtaining an envelope map of the electrocardio data to be analyzed of a power density spectrum curve every 5 minutes in the RR interval, namely a spectrum envelope map of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF;

The method for selecting the model order p determination value comprises the following steps: carrying out frequency domain analysis and calculation on the heart rate variability digital signal RR interval, calculating the power value of the autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF by an average period method, and drawing a corresponding power density spectrum graph; setting an initial value of a model order p, and obtaining an envelope graph of a power density spectrum curve of an RR interval, namely a frequency spectrum graph of an autonomic nerve function frequency domain parameter LFnorm, HFnorm, LF/HF by an AR model method; if the power density spectrum graph is matched with the spectrum envelope graph, the model order p is a determined value, if the power density spectrum graph is not matched with the spectrum envelope graph, the numerical value of the model order p is adjusted, the envelope graph of the power density spectrum graph at the RR interval is obtained again through an AR model method until the power density spectrum graph is matched with the spectrum envelope graph, and the determined value of the model order p is obtained;

The risk judging module is used for indicating that the risk level of the hidden atrial fibrillation is low if the spectrum envelope graph of the heart rate variability signal to be analyzed is in the first standard group graph, indicating that the risk level of the hidden atrial fibrillation is medium risk if the spectrum envelope graph of the heart rate variability signal to be analyzed is in the second standard group graph, and indicating that the risk level of the hidden atrial fibrillation is high risk if the spectrum envelope graph of the heart rate variability signal to be analyzed is in the third standard group graph.