[go: up one dir, main page]

WO2024250308A1 - 一种脉搏分析方法及装置 - Google Patents

一种脉搏分析方法及装置 Download PDF

Info

Publication number
WO2024250308A1
WO2024250308A1 PCT/CN2023/099577 CN2023099577W WO2024250308A1 WO 2024250308 A1 WO2024250308 A1 WO 2024250308A1 CN 2023099577 W CN2023099577 W CN 2023099577W WO 2024250308 A1 WO2024250308 A1 WO 2024250308A1
Authority
WO
WIPO (PCT)
Prior art keywords
pulse
features
domain features
gaussian mixture
frequency domain
Prior art date
Application number
PCT/CN2023/099577
Other languages
English (en)
French (fr)
Inventor
唐延斌
龙子鑫
蒋鑫
谢子成
刘志成
孟祯
彭长虹
Original Assignee
湖南敬凯投资管理有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 湖南敬凯投资管理有限公司 filed Critical 湖南敬凯投资管理有限公司
Publication of WO2024250308A1 publication Critical patent/WO2024250308A1/zh

Links

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 for evaluating the cardiovascular system, e.g. pulse, heart rate, blood pressure or blood flow
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/48—Other medical applications
    • A61B5/4854—Diagnosis based on concepts of alternative medicine, e.g. homeopathy or non-orthodox
    • 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
    • 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
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/10—Pre-processing; Data cleansing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2123/00—Data types
    • G06F2123/02—Data types in the time domain, e.g. time-series data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E30/00—Energy generation of nuclear origin
    • Y02E30/30—Nuclear fission reactors

Definitions

  • the present invention relates to the technical field of pulse analysis, and in particular to a pulse analysis method and device.
  • pulse diagnosis a traditional Chinese medicine diagnostic method in my country, the doctor presses the radial artery at the wrist to feel the periodic expansion changes of the blood vessels, and divides the pulsation patterns felt into different pulse patterns.
  • the pulse pattern can be used to judge the nature of the disease such as cold, heat, deficiency and excess, and can also be used to judge the prognosis of the disease. It is an important objective indicator for identifying diseases in Chinese medicine.
  • pulse diagnosis is easy to diagnose, low-cost, and effective, and has a relatively complete evaluation and intervention system.
  • different doctors have different medical skills and different standards for pulse strength. Therefore, relying solely on the doctor's subjective feelings and experience, it is often impossible to objectively and accurately analyze the patient's pulse. Therefore, it is urgent to develop a pulse analysis method and device that can objectively and accurately analyze the patient's pulse.
  • the present application provides a pulse analysis method and device to solve the problem in the prior art that different doctors have different medical skills and different standards for pulse strength, making it impossible to objectively and accurately analyze the patient's pulse.
  • a pulse analysis method comprising the following steps:
  • Gaussian mixture model features are the weight, mean and covariance matrix of each Gaussian distribution function in the Gaussian mixture model
  • the above pulse analysis method can realize the problem of objectively and accurately analyzing the patient's pulse.
  • each set of pulse data includes multiple pulse data acquired by multiple pulse acquisition units, this multi-channel pulse data can improve the prediction accuracy compared with single-channel pulse data.
  • the introduction of Gaussian mixture model features can further improve the model accuracy.
  • a step of using principal component analysis to eliminate redundant frequency domain features, time domain features and Gaussian mixture model features is also included between steps S2 and S3.
  • the trained neural network model is obtained according to the following steps:
  • Obtain a pulse sample data set annotate the pulse sample data set to obtain pulse classification, divide the pulse sample data set into a training set and a test set, obtain frequency domain features, time domain features and Gaussian mixture model features of each pulse sample data in the training set and the test set, and use principal component analysis to eliminate redundant frequency domain features, time domain features and Gaussian mixture models;
  • the neural network model is trained using the frequency domain features, time domain features and Gaussian mixture model features of all pulse sample data in the training set, and the accuracy of the neural network model is evaluated using the frequency domain features, time domain features and Gaussian mixture model features of all pulse sample data in the test set to obtain a trained neural network model.
  • the pulse data includes pulse width; in step S3, the frequency domain features, time domain features, Gaussian mixture model features and pulse width of each pulse data are input into a pre-trained neural network model, and the predicted pulse classification result is output.
  • a pulse sample data set is obtained, pulse condition annotation is performed on the pulse sample data set to obtain pulse condition classification, the pulse sample data set is divided into a training set and a test set, frequency domain features, time domain features and Gaussian mixture model features of each pulse sample data in the training set and the test set are obtained, and the principal component analysis method is used to eliminate redundant frequency domain features, time domain features and Gaussian mixture models;
  • the neural network model is trained using the frequency domain features, time domain features, Gaussian mixture model features and pulse width of all pulse sample data in the training set, and the accuracy of the neural network model is evaluated using the frequency domain features, time domain features, Gaussian mixture model features and pulse width of all pulse sample data in the test set to obtain a trained neural network model.
  • a pulse analysis device comprising:
  • a first acquisition module is used to acquire three groups of pulse data sets corresponding to the three parts of Cun, Guan and Chi, each group of pulse data sets includes multiple pulse data acquired by multiple pulse acquisition units;
  • a feature extraction module used to obtain frequency domain features, time domain features and Gaussian mixture model features of each pulse data; wherein the Gaussian mixture model features are the weight, mean and covariance matrix of each Gaussian distribution function in the Gaussian mixture model; and
  • the prediction module is used to input the frequency domain features, time domain features and Gaussian mixture model features of each pulse data into a pre-trained neural network model, and output the predicted pulse classification results.
  • the pulse analysis device further includes a feature elimination module for eliminating redundant frequency domain features, time domain features, and Gaussian mixture model features using principal component analysis.
  • the pulse analysis device further includes a model training module for obtaining a trained model according to the following steps: Neural Network Model:
  • Obtain a pulse sample data set annotate the pulse sample data set to obtain pulse classification, divide the pulse sample data set into a training set and a test set, obtain frequency domain features, time domain features and Gaussian mixture model features of each pulse sample data in the training set and the test set, and use principal component analysis to eliminate redundant frequency domain features, time domain features and Gaussian mixture models;
  • the neural network model is trained using the frequency domain features, time domain features and Gaussian mixture model features of all pulse sample data in the training set, and the accuracy of the neural network model is evaluated using the frequency domain features, time domain features and Gaussian mixture model features of all pulse sample data in the test set to obtain a trained neural network model.
  • the pulse data includes pulse width; the prediction module is used to input the frequency domain characteristics, time domain characteristics, Gaussian mixture model characteristics and pulse width of each pulse data into a pre-trained neural network model, and output a predicted pulse classification result.
  • a model training module is further included, which is used to obtain a trained neural network model according to the following steps:
  • Obtain a pulse sample data set annotate the pulse sample data set to obtain pulse classification, divide the pulse sample data set into a training set and a test set, obtain frequency domain features, time domain features and Gaussian mixture model features of each pulse sample data in the training set and the test set, and use principal component analysis to eliminate redundant frequency domain features, time domain features and Gaussian mixture models;
  • FIG1 is a flow chart of a pulse analysis method in one embodiment of the present application.
  • FIG2 is a comparison diagram of an original pulse signal and a reconstructed signal after denoising in an embodiment of the present application
  • FIG. 3 is a schematic diagram of a pulse wave in an embodiment of the present application.
  • a pulse analysis method comprises the following steps:
  • each set of pulse data sets includes multiple pulse A plurality of pulse data acquired by a pulse acquisition unit;
  • the pulse data set is a collection of preprocessed pulse data.
  • wavelet transform is used to preprocess the pulse data. The specific steps are: multi-scale wavelet decomposition of the original signal, threshold quantization processing based on the obtained high-frequency detail components and low-frequency approximate components, and then wavelet reconstruction to obtain the denoised signal.
  • Figure 2 is a comparison diagram of the original signal and the denoised reconstructed signal.
  • the time domain feature includes 6 amplitude ratio features (h2/h1 (descending branch wave height/main wave height), h3/h1 (tidal wave height/main wave height), h4/h1 (descending middle gorge height/main wave height), h5/h1 (dibotomy wave height/main wave height) and the abscissa width w at 2/3 of the main wave height), 6 time ratio features (t1 (main wave time)/T, t2 (descending branch wave time)/T, t3 (tidal wave time)/T, t4 (descending middle gorge time)/T, t5 (dibotomy wave time)/T), as well as mean, variance, root mean square and skewness of pulse pressure, waveform factor, peak factor, pulse factor, margin factor, and a total of 20 time domain features
  • Frequency domain features need to be extracted from the filtered multi-period signal, and the power spectrum of the reconstructed pulse signal sequence is estimated.
  • the extracted frequency domain features include frequency maximum value, frequency domain variance, frequency domain range, rectified mean (the average value of the integral of the absolute value of the signal), frequency domain skewness, and frequency domain root mean square, a total of 6 frequency domain features.
  • a single-cycle pulse wave can be modeled as a Gaussian mixture signal.
  • ⁇ , ⁇ ) is the probability density function of the multidimensional Gaussian distribution, which represents the probability density value of x in the Gaussian distribution with mean ⁇ and covariance matrix ⁇ .
  • the new Gaussian weight, Gaussian mean and Gaussian covariance matrix of each Gaussian distribution can be calculated:
  • the estimated value of Gaussian weight ⁇ k is:
  • the estimated value of the Gaussian mean ⁇ k is:
  • the estimated value of the Gaussian covariance matrix ⁇ k is:
  • the weighted covariance matrix is obtained by subtracting the corresponding mean ⁇ k from each sample xi and then multiplying it by the weight of its responsiveness to the Kth Gaussian distribution.
  • Two Gaussian functions can be used to model a single-cycle pulse waveform, including Gaussian weight, Gaussian mean, Gaussian covariance, and a total of 6 Gaussian mixture model parameter features.
  • the trained neural network model can be obtained according to the following steps:
  • the pulse sample data set should include data of all pulse conditions, and the number of samples corresponding to each pulse condition should not be less than one thousand people.
  • the pulse sample data set is annotated to obtain pulse classification (the pulse categories include 28 pulse types, including floating, sinking, slow, rapid, slippery, astringent, weak, solid, long, short, surging, weak, tight, slow, stringy, wiry, leathery, firm, moist, weak, scattered, thin, hidden, moving, hurried, knotted, alternating, and large).
  • the pulse sample data set is divided into a training set and a test set.
  • the frequency domain features, time domain features, and Gaussian mixture model features of each pulse sample data in the training set and the test set are obtained, and the principal component analysis method is used to eliminate the redundant frequency domain features, time domain features, and Gaussian mixture models.
  • the neural network model is trained using the frequency domain features, time domain features and Gaussian mixture model features of all pulse sample data in the training set, and the accuracy of the neural network model is evaluated using the frequency domain features, time domain features and Gaussian mixture model features of all pulse sample data in the test set to obtain a trained neural network model.
  • the parameters of the neural network model are updated using the Adam optimizer according to the first-order moment and second-order moment estimates of the gradient of the loss function of the neural network model until the accuracy reaches a preset threshold (e.g., 95%).
  • pulse data includes pulse width in addition to pulse wave data.
  • the pulse width is also input as one of the features into the neural network model (including training, testing and prediction).

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Theoretical Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Medical Informatics (AREA)
  • Evolutionary Computation (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Pathology (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Physiology (AREA)
  • Signal Processing (AREA)
  • Psychiatry (AREA)
  • Computing Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Software Systems (AREA)
  • Fuzzy Systems (AREA)
  • Probability & Statistics with Applications (AREA)
  • Alternative & Traditional Medicine (AREA)
  • Cardiology (AREA)
  • Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

一种脉搏分析方法及装置，包括如下步骤：S1、获取与寸、关、尺三个部位相对应的三组脉搏数据集；S2、获取各脉搏数据的频域特征、时域特征和高斯混合模型特征；S3、将各脉搏数据的频域特征、时域特征和高斯混合模型特征输入预先训练好的神经网络模型中，输出得到预测的脉象分类结果。由于每组脉搏数据集中包括由多个脉搏采集单元获取的多个脉搏数据，这种多通道脉搏数据相对于单通道脉搏数据能提高预测的准确度。高斯混合模型特征的引入可进一步提高模型的准确度。

Description

一种脉搏分析方法及装置 技术领域
本发明涉及脉搏分析技术领域，具体涉及一种脉搏分析方法及装置。
背景技术
在心血管内的血液循环过程中，血液在心脏的收缩作用下从左心室涌入主动脉，使得动脉近端管壁扩张、内部压力升高；而当心室舒张时，射血活动暂时停止，动脉在血管弹性的作用下恢复收缩。在收缩舒张的过程中，血液在压力差的作用下以心脏为起点沿着血管快速向远端传播，使得远端动脉亦出现相似的搏动，形成脉搏波，该信号携带了大量与心血管状态相关的信息。在我国的传统中医诊断方法脉诊中，医师通过按压腕部桡动脉感受这种血管周期性的扩张变化，并将感受到的搏动形态分为不同脉象。脉象可以用来判断寒热虚实等疾病性质，也可以用来判断疾病预后，在中医中是辨识疾病的重要客观指标。
脉诊作为一种无创的辨诊方法，其具有诊断方便、成本低、结果有效等特点，拥有较完善的评估和干预体系。然而，不同医师的医术水平参差不齐，对脉搏的强度也有不同的标准，因此单纯依靠医师的主观感觉和经验，往往无法对患者的脉搏进行客观、准确的分析。因此，亟待开发一种能够客观、准确地分析患者脉搏的脉搏分析方法及装置。
发明内容
本申请通过提供一种脉搏分析方法及装置，解决了现有技术中因不同医师的医术水平参差不齐，对脉搏的强度也有不同的标准，使得无法对患者的脉搏进行客观、准确的分析的问题。
为达到上述目的，本申请实施例采用如下技术方案。
一方面，提供一种脉搏分析方法，包括如下步骤：
S1、获取与寸、关、尺三个部位相对应的三组脉搏数据集，每组脉搏数据集中包括由多个脉搏采集单元获取的多个脉搏数据；
S2、获取各脉搏数据的频域特征、时域特征和高斯混合模型特征；其中所述高斯混合模型特征为高斯混合模型中每个高斯分布函数的权重、均值和协方差矩阵；
S3、将各脉搏数据的频域特征、时域特征和高斯混合模型特征输入预先训练好的神经网络模型中，输出得到预测的脉象分类结果。
上述脉搏分析方法可实现对患者的脉搏进行客观、准确的分析的问题。另一方面，由于每组脉搏数据集中包括由多个脉搏采集单元获取的多个脉搏数据，这种多通道脉搏数据相对于单通道脉搏数据能提高预测的准确度。其次，高斯混合模型特征的引入可进一步提高模型 的准确度。
为加快模型计算速度，在一些实施例中，步骤S2与S3之间还包括采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型特征的步骤。
在一些实施例中，根据如下步骤获取训练好的神经网络模型：
获取脉搏样本数据集，对脉搏样本数据集进行脉象标注得到脉象分类，将脉搏样本数据集分为训练集和测试集，获取训练集和测试集中每个脉搏样本数据的频域特征、时域特征和高斯混合模型特征，并采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型；
利用训练集中所有脉搏样本数据的频域特征、时域特征和高斯混合模型特征对神经网络模型进行训练，利用测试集中所有脉搏样本数据的频域特征、时域特征和高斯混合模型特征对神经网络模型的准确度进行评估，得到训练好的神经网络模型。
在一些实施例中，所述脉搏数据包括脉搏宽度；步骤S3中，将各脉搏数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度输入预先训练好的神经网络模型中，输出得到预测的脉象分类结果。
在一些实施例中，获取脉搏样本数据集，对脉搏样本数据集进行脉象标注得到脉象分类，将脉搏样本数据集分为训练集和测试集，获取训练集和测试集中每个脉搏样本数据的频域特征、时域特征和高斯混合模型特征，并采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型；
利用训练集中所有脉搏样本数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度对神经网络模型进行训练，利用测试集中所有脉搏样本数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度对神经网络模型的准确度进行评估，得到训练好的神经网络模型。
又一方面，提供一种脉搏分析装置，包括：
第一获取模块，用于获取与寸、关、尺三个部位相对应的三组脉搏数据集，每组脉搏数据集中包括由多个脉搏采集单元获取的多个脉搏数据；
特征提取模块，用于获取各脉搏数据的频域特征、时域特征和高斯混合模型特征；其中所述高斯混合模型特征为高斯混合模型中每个高斯分布函数的权重、均值和协方差矩阵；和
预测模块，用于将各脉搏数据的频域特征、时域特征和高斯混合模型特征输入预先训练好的神经网络模型中，输出得到预测的脉象分类结果。
在一些实施例中，脉搏分析装置还包括特征剔除模块，用于采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型特征。
在一些实施例中，脉搏分析装置还包括模型训练模块，用于根据如下步骤获取训练好的 神经网络模型：
获取脉搏样本数据集，对脉搏样本数据集进行脉象标注得到脉象分类，将脉搏样本数据集分为训练集和测试集，获取训练集和测试集中每个脉搏样本数据的频域特征、时域特征和高斯混合模型特征，并采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型；
利用训练集中所有脉搏样本数据的频域特征、时域特征和高斯混合模型特征对神经网络模型进行训练，利用测试集中所有脉搏样本数据的频域特征、时域特征和高斯混合模型特征对神经网络模型的准确度进行评估，得到训练好的神经网络模型。
在一些实施例中，所述脉搏数据包括脉搏宽度；所述预测模块用于将各脉搏数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度输入预先训练好的神经网络模型中，输出得到预测的脉象分类结果。
在一些实施例中，还包括模型训练模块，用于根据如下步骤获取训练好的神经网络模型：
获取脉搏样本数据集，对脉搏样本数据集进行脉象标注得到脉象分类，将脉搏样本数据集分为训练集和测试集，获取训练集和测试集中每个脉搏样本数据的频域特征、时域特征和高斯混合模型特征，并采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型；
利用训练集中所有脉搏样本数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度对神经网络模型进行训练，利用测试集中所有脉搏样本数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度对神经网络模型的准确度进行评估，得到训练好的神经网络模型。
本申请实施例中提供的一个或多个技术方案，至少具有如下技术效果或优点：可实现对患者的脉搏进行客观、准确的分析的问题。另一方面，由于每组脉搏数据集中包括由多个脉搏采集单元获取的多个脉搏数据，这种多通道脉搏数据相对于单通道脉搏数据能提高预测的准确度。其次，高斯混合模型特征的引入可进一步提高模型的准确度。
附图说明
图1为本申请一实施例中一种脉搏分析方法的流程示意图；
图2为本申请一实施例中脉搏的原始信号和去噪后重构信号的对比图；
图3为本申请一实施例中脉搏波的示意图。
具体实施方式
为了更好的理解上述技术方案，下面将结合说明书附图以及具体的实施方式对上述技术方案进行详细的说明。
参见图1，一种脉搏分析方法，包括如下步骤：
S1、获取与寸、关、尺三个部位相对应的三组脉搏数据集，每组脉搏数据集中包括由多个脉 搏采集单元获取的多个脉搏数据；
需要说明的是，脉搏数据集为经预处理后的脉搏数据的集合。本实施例中采用小波变换对脉搏数据进行预处理，具体步骤为：将原始信号进行多尺度小波分解，根据得到的高频细节分量和低频近似分量进行阈值量化处理，再经小波重构得到去噪后的信号。图2为原始信号和去噪后重构信号的对比图。
S2、获取各脉搏数据的频域特征、时域特征和高斯混合模型特征，并采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型特征。其中所述高斯混合模型特征为高斯混合模型中每个高斯分布函数的权重、均值和协方差矩阵。
脉搏信号数据经过预处理后，对每个样本经过重采样和平均后的单周期脉象信号进行时域特征提取。如图3所示，时域特征包括6个幅值比例特征(h2/h1(降支波高度/主波高度)，h3/h1(潮波高度/主波高度)，h4/h1(降中峡高度/主波高度)，h5/h1(重博波高度/主波高度)以及主波高度2/3处的横坐标宽度w)，6个时间比例特征(t1(主波时间)/T，t2(降支波时间)/T，t3(潮波时间)/T，t4(降中峡时间)/T，t5(重博波时间)/T)，还有脉搏压力的均值、方差、均方根以及偏度，波形因子，峰值因子，脉冲因子，裕度因子，时域共20个时域特征。
频域特征需要从滤波后的多周期信号中提取，对重构之后的脉搏信号序列进行功率谱估计，提取的频域特征包括频率最大值，频域方差，频域极差，整流平均值(信号绝对值积分的平均值)，频域偏度，频域均方根共6个频域特征。
单周期脉搏波可以建模为高斯混合信号，假设单周期脉搏波数据集为X＝x1,x2,x3,...xN,单周期脉搏波有2个高斯分布，每个高斯分布的参数为θk＝(πk,μk,∑k)，其中πk是该高斯分布的权重(满足)，μk是该高斯分布的均值，∑k是该高斯分布的协方差矩阵。
则对于每个样本xi，它对第k个高斯分布的响应度为：
其中，N(x|μ,∑)是多维高斯分布的概率密度函数，表示x在均值为μ，协方差矩阵为∑的高斯分布中的概率密度值。
根据上述响应度，可以计算出每个高斯分布的新的高斯权重、高斯均值和高斯协方差矩阵：
1.高斯权重πk的估计值为：
即为每个样本xi对第K个高斯分布的响应度的平均值。
2.高斯均值μk的估计值为：
即为每个样本xi乘上它对第K个高斯分布的响应度的权重后的加权平均值。
3.高斯协方差矩阵∑k的估计值为：
即为每个样本xi减去对应的均值μk后再乘以它对第K个高斯分布的响应度的权重后的加权协方差矩阵。
用二个高斯函数即可对单周期脉搏波形建模，包括高斯权重，高斯均值，高斯协方差共6个高斯混合模型参数特征。
S3、将各脉搏数据的频域特征、时域特征和高斯混合模型特征输入预先训练好的神经网络模型中，输出得到预测的脉象分类结果。
其中，训练好的神经网络模型可根据如下步骤获取：
获取脉搏样本数据集，脉搏样本数据集也是经预处理后的数据集，预处理的方法与脉搏数据集的预处理方法一致。脉搏样本数据集应囊括所有脉象的数据，每种脉象对应的样本数量不低于一千人。
对脉搏样本数据集进行脉象标注得到脉象分类(脉象类别包括浮、沉、迟、数、滑、涩、虚、实、长、短、洪、微、紧、缓、弦、芤、革、牢、濡、弱、散、细、伏、动、促、结、代、大共28种脉象)，将脉搏样本数据集分为训练集和测试集，获取训练集和测试集中每个脉搏样本数据的频域特征、时域特征和高斯混合模型特征，并采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型。
利用训练集中所有脉搏样本数据的频域特征、时域特征和高斯混合模型特征对神经网络模型进行训练，利用测试集中所有脉搏样本数据的频域特征、时域特征和高斯混合模型特征对神经网络模型的准确度进行评估，得到训练好的神经网络模型。对神经网络模型的准确度进行评估时，利用Adam优化器根据神经网络模型的损失函数梯度的一阶矩和二阶矩估计来更新神经网络模型的参数，直至准确度达到预设阈值(比如95％)。
实施例二
本实施例与实施例一基本一致，区别在于脉搏数据除了脉搏波数据之外，还包括脉搏宽 度，且脉搏宽度也作为特征之一输入到神经网络模型中(包括训练、测试和预测)。
最后应说明的是：以上实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的精神和范围。

Claims (10)

  1. 一种脉搏分析方法，包括如下步骤：
    S1、获取与寸、关、尺三个部位相对应的三组脉搏数据集，每组脉搏数据集中包括由多个脉搏采集单元获取的多个脉搏数据；
    S2、获取各脉搏数据的频域特征、时域特征和高斯混合模型特征；其中所述高斯混合模型特征为高斯混合模型中每个高斯分布函数的权重、均值和协方差矩阵；
    S3、将各脉搏数据的频域特征、时域特征和高斯混合模型特征输入预先训练好的神经网络模型中，输出得到预测的脉象分类结果。
  2. 根据权利要求1所述的脉搏分析方法，其特征在于，步骤S2与S3之间还包括采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型特征的步骤。
  3. 根据权利要求2所述的脉搏分析方法，其特征在于：根据如下步骤获取训练好的神经网络模型：
    获取脉搏样本数据集，对脉搏样本数据集进行脉象标注得到脉象分类，将脉搏样本数据集分为训练集和测试集，获取训练集和测试集中每个脉搏样本数据的频域特征、时域特征和高斯混合模型特征，并采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型；
    利用训练集中所有脉搏样本数据的频域特征、时域特征和高斯混合模型特征对神经网络模型进行训练，利用测试集中所有脉搏样本数据的频域特征、时域特征和高斯混合模型特征对神经网络模型的准确度进行评估，得到训练好的神经网络模型。
  4. 根据权利要求2所述的脉搏分析方法，其特征在于：所述脉搏数据包括脉搏宽度；步骤S3中，将各脉搏数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度输入预先训练好的神经网络模型中，输出得到预测的脉象分类结果。
  5. 根据权利要求4所述的脉搏分析方法，其特征在于，根据如下步骤获取训练好的神经网络模型：
    获取脉搏样本数据集，对脉搏样本数据集进行脉象标注得到脉象分类，将脉搏样本数据集分为训练集和测试集，获取训练集和测试集中每个脉搏样本数据的频域特征、时域特征和高斯混合模型特征，并采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型；
    利用训练集中所有脉搏样本数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度对神经网络模型进行训练，利用测试集中所有脉搏样本数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度对神经网络模型的准确度进行评估，得到训练好的神经网络模型。
  6. 一种脉搏分析装置，其特征在于，包括：
    第一获取模块，用于获取与寸、关、尺三个部位相对应的三组脉搏数据集，每组脉搏数 据集中包括由多个脉搏采集单元获取的多个脉搏数据；
    特征提取模块，用于获取各脉搏数据的频域特征、时域特征和高斯混合模型特征；其中所述高斯混合模型特征为高斯混合模型中每个高斯分布函数的权重、均值和协方差矩阵；和
    预测模块，用于将各脉搏数据的频域特征、时域特征和高斯混合模型特征输入预先训练好的神经网络模型中，输出得到预测的脉象分类结果。
  7. 根据权利要求6所述的脉搏分析装置，其特征在于：还包括特征剔除模块，用于采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型特征。
  8. 根据权利要求7所述的脉搏分析装置，其特征在于，还包括模型训练模块，用于根据如下步骤获取训练好的神经网络模型：
    获取脉搏样本数据集，对脉搏样本数据集进行脉象标注得到脉象分类，将脉搏样本数据集分为训练集和测试集，获取训练集和测试集中每个脉搏样本数据的频域特征、时域特征和高斯混合模型特征，并采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型；
    利用训练集中所有脉搏样本数据的频域特征、时域特征和高斯混合模型特征对神经网络模型进行训练，利用测试集中所有脉搏样本数据的频域特征、时域特征和高斯混合模型特征对神经网络模型的准确度进行评估，得到训练好的神经网络模型。
  9. 根据权利要求6所述的脉搏分析装置，其特征在于：所述脉搏数据包括脉搏宽度；所述预测模块用于将各脉搏数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度输入预先训练好的神经网络模型中，输出得到预测的脉象分类结果。
  10. 根据权利要求9所述的脉搏分析装置，其特征在于：还包括模型训练模块，用于根据如下步骤获取训练好的神经网络模型：
    获取脉搏样本数据集，对脉搏样本数据集进行脉象标注得到脉象分类，将脉搏样本数据集分为训练集和测试集，获取训练集和测试集中每个脉搏样本数据的频域特征、时域特征和高斯混合模型特征，并采用主成分分析法剔除冗余的频域特征、时域特征和高斯混合模型；
    利用训练集中所有脉搏样本数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度对神经网络模型进行训练，利用测试集中所有脉搏样本数据的频域特征、时域特征、高斯混合模型特征和脉搏宽度对神经网络模型的准确度进行评估，得到训练好的神经网络模型。
PCT/CN2023/099577 2023-06-06 2023-06-12 一种脉搏分析方法及装置 WO2024250308A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202310661630.2 2023-06-06
CN202310661630.2A CN116649924A (zh) 2023-06-06 2023-06-06 一种脉搏分析方法及装置

Publications (1)

Publication Number Publication Date
WO2024250308A1 true WO2024250308A1 (zh) 2024-12-12

Family

ID=87718680

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/099577 WO2024250308A1 (zh) 2023-06-06 2023-06-12 一种脉搏分析方法及装置

Country Status (2)

Country Link
CN (1) CN116649924A (zh)
WO (1) WO2024250308A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117224092B (zh) * 2023-11-16 2024-02-09 常熟理工学院 基于决策树的光电容积脉搏波干扰段实时检测方法及系统

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107049269A (zh) * 2017-03-06 2017-08-18 武汉嫦娥医学抗衰机器人股份有限公司 一种中医脉象信号处理分析系统
CN107822608A (zh) * 2017-10-26 2018-03-23 中国民航大学 基于高斯混合模型的脉搏波特征提取方法
CN108670209A (zh) * 2018-03-29 2018-10-19 中国科学院微电子研究所 一种自动识别中医脉象的方法和系统
CN112487945A (zh) * 2020-11-26 2021-03-12 上海贝业斯健康科技有限公司 一种基于双路卷积神经网络融合的脉象识别方法
CN116089797A (zh) * 2023-02-08 2023-05-09 苏州大学 一种基于卷积神经网络的脉象识别方法和系统
CN116616727A (zh) * 2023-06-06 2023-08-22 湖南敬凯投资管理有限公司 一种脉搏定位、采集、分析方法及装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107049269A (zh) * 2017-03-06 2017-08-18 武汉嫦娥医学抗衰机器人股份有限公司 一种中医脉象信号处理分析系统
CN107822608A (zh) * 2017-10-26 2018-03-23 中国民航大学 基于高斯混合模型的脉搏波特征提取方法
CN108670209A (zh) * 2018-03-29 2018-10-19 中国科学院微电子研究所 一种自动识别中医脉象的方法和系统
CN112487945A (zh) * 2020-11-26 2021-03-12 上海贝业斯健康科技有限公司 一种基于双路卷积神经网络融合的脉象识别方法
CN116089797A (zh) * 2023-02-08 2023-05-09 苏州大学 一种基于卷积神经网络的脉象识别方法和系统
CN116616727A (zh) * 2023-06-06 2023-08-22 湖南敬凯投资管理有限公司 一种脉搏定位、采集、分析方法及装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZHANG SHIYU, YANG KE; XIA CHUNMING; JIN CHENLING; YAN HAIXIA; WANG YIQIN: "Research on Feature Reduction and Classification of Pulse Signal Based on Random Forest", SHI JIE KE XUE JI SHU ZA ZHI, CN, vol. 22, no. 7, 1 January 2020 (2020-01-01), CN , pages 2418 - 2426, XP093246116, ISSN: 1674-3849, DOI: 10.11842/wst.20190730010 *

Also Published As

Publication number Publication date
CN116649924A (zh) 2023-08-29

Similar Documents

Publication Publication Date Title
Zhang et al. A sensor-based wrist pulse signal processing and lung cancer recognition
Gupta et al. Neural network classification of homomorphic segmented heart sounds
CN103034837B (zh) 特征参数与脉象要素的关联
CN109561841A (zh) 用于测量心肌缺血、狭窄识别、定位和血流储备分数估计的非侵入式方法和系统
Milani et al. Abnormal heart sound classification using phonocardiography signals
CN107468221B (zh) 基于金氏脉学的脉搏波信号处理方法及肺癌检测系统
CN109948396B (zh) 一种心拍分类方法、心拍分类装置及电子设备
CN109864714A (zh) 一种基于深度学习的心电信号分析方法
Zhang et al. Application of deep neural network for congestive heart failure detection using ECG signals
CN114027804A (zh) 一种脉象诊断的方法、装置及可读存储介质
CN116451110A (zh) 基于信号能量特征和脉搏周期的血糖预测模型构建方法
Taşkıran et al. A deep learning based decision support system for diagnosis of Temporomandibular joint disorder
CN113749658A (zh) 基于集成学习的心肺耦合抑郁状态识别方法及系统
CN113116300A (zh) 一种基于模型融合的生理信号分类方法
WO2024250308A1 (zh) 一种脉搏分析方法及装置
Wang et al. IMSF-Net: An improved multi-scale information fusion network for PPG-based blood pressure estimation
CN111370120A (zh) 一种基于心音信号的心脏舒张功能障碍的检测方法
Pan et al. Improved blood pressure estimation using photoplethysmography based on ensemble method
CN113243902B (zh) 一种基于光电容积脉搏波的特征提取方法
CN114145725A (zh) 一种基于无创连续血压测量的ppg采样率估算方法
CN110491504B (zh) 一种心音信号医学指标数据的获取方法
CN113729648A (zh) 一种基于多脉搏传感器的穿戴式脉诊手环系统
WO2024250309A1 (zh) 一种脉搏定位、采集、分析方法及装置
CN117281479A (zh) 一种基于表面肌电信号多维度特征融合的人体下肢慢性疼痛判别方法、存储介质及装置
CN114864095B (zh) 一种结合多运动强度下的狭窄冠脉血运变化的分析方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23940190

Country of ref document: EP

Kind code of ref document: A1