Method for processing and analyzing pulse condition information of atrial fibrillation patient

Abstract

The invention relates to a method for processing and analyzing pulse condition information of an atrial fibrillation patient in the technical field of pulse condition analysis and recognition, which comprises the following steps of: S1, acquiring a pulse signal and converting the pulse signal into a digital signal; s2, noise reduction preprocessing is conducted on the digital pulse signals through an empirical mode decomposition method in combination with self-adaptive threshold value processing; s3, Fourier transform FT is carried out on the preprocessed pulse signal, the pulse signal is converted into a frequency domain signal, Hilbert-Huang transform HHT is combined to process a non-stationary signal, frequency components of local instantaneous transform are extracted, and frequency spectrum features of the pulse are comprehensively obtained; s4, key characteristic parameters including frequency, amplitude, period and fluctuation amplitude are extracted from the frequency spectrum, and matching analysis is carried out on the key characteristic parameters and a predefined pulse condition database; according to the processing and analyzing method, automation, digitization and quantification of pulse condition analysis are achieved through the modern signal processing technology combining FT-HHT, and then objectivity, accuracy and wide applicability of pulse diagnosis are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of pulse analysis and recognition, in particular to a method for processing and analyzing pulse information of atrial fibrillation patients. Background Art

[0002] As is known to all, arrhythmia is a common heart disease that affects millions of people worldwide, among which atrial fibrillation is one of the most common and clinically significant forms. Early and accurate detection of atrial fibrillation is crucial for preventing complications such as stroke and heart failure. Figure 1 and 2 As mentioned above, since the pulse signal of a normal person is regular and orderly, while the spectrum of atrial fibrillation is usually irregular low-frequency components and lacks obvious P-wave components, it can be distinguished by measuring the pulse signal; currently common pulse signal acquisition methods include pressure sensors, photoelectric volume pulse wave sensors, electronic blood pressure monitors, etc. These devices can convert the pulse image of traditional manual diagnosis into a digital signal for subsequent analysis and processing; since the pulse signal is affected by multiple factors such as the patient's physiological state, measurement environment, sensor type, etc., the signal quality may fluctuate, especially when the pulse is weak or irregular, the collected signal is easily affected by interference and noise, and errors may occur when analyzing these low-quality signals, resulting in inaccurate analysis results. Summary of the Invention

[0003] In order to overcome the deficiencies in the background technology and solve the problems of subjectivity, inaccurate diagnosis and non-standardized data in traditional pulse diagnosis analysis, the present invention discloses a method for processing and analyzing pulse information of patients with atrial fibrillation. By combining modern signal processing technology with FT-HHT, the pulse analysis is automated, digitized and quantified, thereby improving the objectivity, accuracy and wide applicability of pulse diagnosis.

[0004] To achieve the above object, the present invention adopts the following technical solutions:

[0005] A method for processing and analyzing pulse information of atrial fibrillation patients comprises the following steps: S1, collecting pulse signals and converting them into digital signals; S2, performing noise reduction preprocessing on the digitized pulse signals by combining an empirical mode decomposition method with adaptive threshold processing; S3, performing Fourier transform (FT) on the preprocessed pulse signals to convert them into frequency domain signals, and then combining Hilbert-Huang transform (HHT) to process non-stationary signals, extracting the frequency components of local instantaneous transformations, and comprehensively obtaining the spectrum characteristics of the pulse; S4, extracting key characteristic parameters from the spectrum, including frequency, amplitude, period, and fluctuation amplitude, and performing matching analysis with a predefined pulse database.

[0006] Furthermore, in S1, a photoplethysmography sensor or a pressure sensor is used to collect the patient's pulse to obtain an electrical signal reflecting the pulse fluctuation, and the electrical signal is converted into a digital signal through a signal conditioning circuit.

[0007] Furthermore, in S2, the empirical mode decomposition method can decompose non-stationary signals into multiple intrinsic mode functions (IMFs); first, the pulse digital signal is detected for extreme points, upper and lower envelopes are generated through the extreme points, the mean of the envelopes is calculated, and the first IMF is separated. After removing the first IMF, the remaining signal continues to undergo the same processing until a group of IMFs is obtained, each of which represents a different frequency component in the signal. The above process is repeated until the remaining signal is monotonic and cannot be further decomposed into IMFs; then, energy analysis and correlation analysis are performed on all extracted IMFs to identify components related to noise.

[0008] Furthermore, in S2, adaptive threshold processing is used to remove the identified noise-related IMFs and retain the main components of the signal; high-frequency IMFs are processed by an adaptive threshold based on kurtosis to remove noise. The kurtosis calculation formula is: Where μ is the mean, σ is the standard deviation, and N is the number of data points; the adaptive threshold is calculated using the following formula: The threshold calculated by this formula will be dynamically adjusted according to the size of the IMF, retaining the more significant components in the signal and removing irrelevant noise components; the IMFs with noise removed are summed to reconstruct the clean part of the signal to restore the denoised pulse signal.

[0009] Furthermore, in S2, the noise-reduced signal is amplified.

[0010] Furthermore, in S3, discrete Fourier transform DFT is selected for calculation, and fast Fourier transform FFT is introduced to shorten the calculation time.

[0011] Furthermore, in S3, if the problem of modal aliasing occurs when HHT performs empirical mode decomposition, it can in turn use FT to filter the signal, separate the different frequency components, and use HHT to operate again.

[0012] Furthermore, in S4, the predefined pulse condition database includes frequency spectrum information of floating pulse, sinking pulse, rapid pulse and slow pulse.

[0013] Furthermore, in S4, a deep learning model is established, and different pulse data are collected to train the deep learning model, so that the trained model can be used to automatically match and analyze the input pulse characteristics and output the pulse classification results.

[0014] Furthermore, in S4, based on the spectrum analysis result, a statistical method is used to describe the frequency components.

[0015] Due to the adoption of the above-mentioned technical solution, the present invention has the following beneficial effects:

[0016] The method for processing and analyzing pulse information of atrial fibrillation patients disclosed in the present invention reduces the noise of the pulse signal and performs FT-HHT joint transform analysis to convert the pulse signal from the time domain to the frequency domain. This can extract the specific spectral characteristics of the pulse, eliminate the influence of human factors, and provide objective pulse analysis data, thereby realizing the quantification and automation of pulse analysis. In addition, the method can make subsequent pulse diagnosis more accurate and objective, reduce the complexity and workload of pulse diagnosis, and expand the application scenarios and scope of pulse diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a diagram of a normal person's pulse signal;

[0018] Figure 2 This is a schematic diagram of the pulse signal of a patient with atrial fibrillation;

[0019] Figure 3 It is a schematic diagram of the time domain pulse signal;

[0020] Figure 4 This is a schematic diagram of the spectrum of the pulse signal after filtering;

[0021] Figure 5 This is a diagram showing the fitting effects of nine types of pulse data. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be described below with reference to the accompanying drawings in the embodiments of the present invention.

[0023] A method for processing and analyzing pulse information of a patient with atrial fibrillation, comprising the following steps:

[0024] Step 1: Collect the pulse signal, obtain the original data of the atrial fibrillation patient's pulse information, and convert the signal into a digital signal; the specific operation can use a photoelectric volumetric pulse wave sensor or a pressure sensor to collect the patient's pulse. The sensor converts the pulse fluctuations, such as pressure or light reflection changes, into an electrical signal. The electrical signal is processed by a signal conditioning circuit, such as an amplifier and a filter, to convert the analog signal into a digital signal.

[0025] Step 2: In order to optimize the collected pulse signal and reduce noise interference, the collected pulse signal is subjected to noise filtering to remove environmental interference or noise from the sensor itself. This is mainly done by combining empirical mode decomposition with adaptive threshold processing to perform noise reduction preprocessing on the digitized pulse signal.

[0026] The empirical mode decomposition method can decompose non-stationary signals into multiple intrinsic mode functions (IMFs), each representing a different frequency component in the signal. First, the pulse digital signal is subjected to extreme point detection, and upper and lower envelopes are generated from the extreme points. The envelope mean is calculated to isolate the first IMF. After removing the first IMF, the remaining signal is processed in the same way until a set of IMFs is obtained. This process is repeated until the remaining signal is monotonic and cannot be further decomposed into IMFs. The advantage is that it does not require a preset basis function and can adaptively decompose non-stationary signals, which is suitable for the nonlinear characteristics of the pulse wave.

[0027] Then, energy analysis and correlation analysis are performed on all extracted IMFs to identify components related to noise. IMF energy analysis requires calculating the energy contribution of each IMF. High-frequency IMFs usually correspond to noise. IMF correlation analysis requires calculating the correlation between the IMF and the original signal to filter out IMFs related to noise, such as high-frequency IMFs with low correlation.

[0028] Adaptive threshold processing is to remove the identified noise-related IMF and retain the main components of the signal; the high-frequency IMF is processed by an adaptive threshold based on kurtosis to remove noise and retain significant features. The kurtosis calculation formula is: Where μ is the mean, σ is the standard deviation, and N is the number of data points; the adaptive threshold is calculated using the following formula: The threshold calculated by this formula is dynamically adjusted according to the size of the IMF, retaining the more significant components of the signal and removing irrelevant noise components. The IMFs after noise removal are summed to reconstruct the clean part of the signal to restore the denoised pulse signal.

[0029] If the signal after noise reduction is weak, signal amplification is required to ensure that the signal is clear enough for subsequent analysis, and finally a clear, noise-free pulse signal can be output.

[0030] Step 3: Perform Fourier transform (FT) on the pre-processed pulse signal to convert it into a frequency domain signal. Then, combine it with Hilbert-Huang transform (HHT) to process the non-stationary signal, extract the frequency components of the local instantaneous transformation, and comprehensively obtain the spectral characteristics of the pulse. Specifically, the pre-processed pulse signal is input into the FT-HHT module; the discrete Fourier transform (DFT) is selected for calculation, and the fast Fourier transform (FFT) is introduced to shorten the calculation time.

[0031] Discrete Fourier transform DFT is applicable to discrete signals of finite length, and the formula is: Where x[n] is a discrete time signal and X[k] is a frequency domain signal. The time domain signal is Fourier transformed using the Fast Fourier Transform (FFT) formula and algorithm to convert it into a frequency domain signal. The collected time domain pulse signal is converted into an image of the sum of five sinusoidal functions. The amplitude and phase of the image are analyzed to construct a time domain function formula:

[0032]

[0033] The amplitude A i represents the amplitude of the i-th (i=1,2,3,4,5) function, represents the phase of the i-th function;

[0034] FT is suitable for global analysis, converting signals from the time domain to the frequency domain. Pulse time domain signals are relatively complex, so FT is first used to perform frequency domain analysis on the signal to identify the main frequency components of the signal. Local signals change rapidly, and FT cannot identify the local time domain characteristics of the signal. Therefore, HHT is introduced to process non-stationary signals and extract the frequency components of local instantaneous changes.

[0035] If the problem of modal aliasing occurs when HHT performs empirical mode decomposition, the signal can be filtered with the help of FT to separate the different frequency components and then use HHT to operate again;

[0036] The spectrum signal diagram is obtained after the joint transformation analysis combining FT and HHT

[0037]

[0038] Here f(t) is the original time domain function, which represents the change of the signal with time t, α is the angular frequency, and F(α) is the spectrum of the signal f(t). Draw the spectrum diagram, as shown in the attached figure. Figure 3 and 4 As shown in the figure, the key frequency components in the frequency domain signal diagram are extracted, such as the pulse fluctuation frequency, amplitude, period and other information.

[0039] Step 4: Extract key characteristic parameters from the spectrum, including frequency, amplitude, period and fluctuation range, and perform matching analysis with the predefined pulse database; the predefined pulse database contains the spectrum information of floating pulse, sinking pulse, rapid pulse and slow pulse, as shown in the attached Figure 5 As shown, the pulse condition database can contain nine different pulse conditions.

[0040] According to needs, a deep learning model can be established, and different pulse data can be collected to train the deep learning model, so that the trained model can be used to automatically match and analyze the input pulse characteristics and output the pulse classification results. The purpose is to automatically diagnose the pulse type through machine learning or pattern recognition methods; specifically, it is necessary to select a model first, traditional machine learning models: such as SVM (support vector machine), random forest, KNN (K-nearest neighbor), etc., deep learning models: such as CNN (convolutional neural network), LSTM (long short-term memory network), etc., which are suitable for processing high-dimensional time series data; then conduct model training, divide the data set, and use the training set-validation set-test set method: training set: 70%, used for model training; validation set: 15%, used for hyperparameter tuning; test set: 15%, used for final performance evaluation; if the amount of data is small, cross-validation (such as 5-fold cross-validation) can be used to improve generalization ability; then model evaluation, use the test set to evaluate model performance; evaluation The estimation indicators include: Accuracy: the proportion of correct overall classification; Recall: the ability to detect abnormal pulse; Precision: the probability that the detected abnormal pulse is a true abnormality; F1-score: a comprehensive indicator that balances precision and recall; AUC-ROC curve: measures the model performance in binary classification tasks; finally, model deployment, exporting the trained model into a deployable format, integrating the model into the pulse intelligent diagnosis system to provide real-time diagnosis; when applied, the extracted pulse features are input into the intelligent diagnosis module, such as machine learning algorithms and pattern recognition systems; based on the trained model, the system automatically analyzes and calculates the probability that the sample belongs to each pulse category, and uses the maximum probability classification method to select the most likely category. For example, the pulse of atrial fibrillation patients and normal people, the system can use cosine similarity or Euclidean distance to calculate the similarity based on historical data and current pulse features, perform similarity matching, and give a diagnosis result.

[0041] Based on the results of spectrum analysis, statistical methods such as mean, standard deviation, skewness, kurtosis, etc. in SPSS software are used to describe the frequency components. These statistical features can help identify the pulse characteristics of patients with atrial fibrillation.

[0042] The parts of the present invention that are not described in detail are prior art. It is obvious to those skilled in the art that the present invention is not limited to the details of the above-mentioned exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, no matter from which point of view, the above-mentioned embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is limited by the appended claims rather than the above description. Therefore, it is intended that all changes that fall within the meaning and scope of the equivalent elements of the claims are included in the present invention, and any figure marks in the claims should not be regarded as limiting the content of the claims involved.

Claims

1. A method for processing and analyzing pulse information of a patient with atrial fibrillation, characterized in that: The following steps are involved: S1, collect pulse signals and convert them into digital signals; S2. Preprocessing the digitized pulse signal for noise reduction by using the empirical mode decomposition method combined with adaptive threshold processing; S3, performing Fourier transform (FT) on the pre-processed pulse signal to convert it into a frequency domain signal, then combining it with Hilbert-Huang transform (HHT) to process the non-stationary signal, extract the frequency components of the local instantaneous transformation, and comprehensively obtain the spectrum characteristics of the pulse; S4. Extract key characteristic parameters from the spectrum, including frequency, amplitude, period and fluctuation amplitude, and perform matching analysis with the predefined pulse database.

2. The method for processing and analyzing pulse information of atrial fibrillation patients according to claim 1, wherein: In S1, a photoelectric capacitance pulse wave sensor or a pressure sensor is used to collect the patient's pulse to obtain an electrical signal reflecting the pulse fluctuation. The electrical signal is converted into a digital signal through a signal conditioning circuit.

3. The method for processing and analyzing pulse information of atrial fibrillation patients according to claim 1, wherein: In S2, the empirical mode decomposition method can decompose non-stationary signals into multiple intrinsic mode functions (IMFs). First, the pulse digital signal is detected for extreme points, and upper and lower envelopes are generated through the extreme points. The mean of the envelope is calculated to separate the first IMF. After removing the first IMF, the remaining signal is processed in the same way until a group of IMFs is obtained. Each IMF represents a different frequency component in the signal. The above process is repeated until the remaining signal is monotonic and cannot be further decomposed into IMFs. Then, energy analysis and correlation analysis are performed on all extracted IMFs to identify components related to noise.

4. The method for processing and analyzing pulse information of atrial fibrillation patients according to claim 3, wherein: In S2, adaptive threshold processing is used to remove the identified noise-related IMFs and retain the main components of the signal. The high-frequency IMFs are processed by an adaptive threshold based on kurtosis to remove noise. The kurtosis calculation formula is: Where μ is the mean, σ is the standard deviation, and N is the number of data points; The adaptive threshold is calculated using the following formula: The threshold calculated by this formula will be dynamically adjusted according to the size of the IMF, retaining the more significant components in the signal and removing irrelevant noise components; the IMFs with noise removed are summed to reconstruct the clean part of the signal to restore the denoised pulse signal.

5. The method for processing and analyzing pulse information of atrial fibrillation patients according to claim 1, wherein: In S2, the signal after noise reduction is amplified.

6. The method for processing and analyzing pulse information of atrial fibrillation patients according to claim 1, wherein: In S3, discrete Fourier transform (DFT) is selected for calculation, and fast Fourier transform (FFT) is introduced to shorten the calculation time.

7. The method for processing and analyzing pulse information of atrial fibrillation patients according to claim 1, wherein: In S3, if the problem of modal aliasing occurs when HHT performs empirical mode decomposition, it can in turn use FT to filter the signal, separate the different frequency components, and use HHT to operate again.

8. The method for processing and analyzing pulse information of atrial fibrillation patients according to claim 1, wherein: In S4, the predefined pulse database contains frequency spectrum information of floating pulse, sinking pulse, rapid pulse and slow pulse.

9. The method for processing and analyzing pulse information of atrial fibrillation patients according to claim 1, wherein: In S4, a deep learning model is established, and different pulse data are collected to train the deep learning model, so that the trained model can be used to automatically match and analyze the input pulse characteristics and output the pulse classification results.

10. The method for processing and analyzing pulse information of atrial fibrillation patients according to claim 1, wherein: In S4, based on the spectrum analysis results, the frequency components are described using statistical methods.

Images