没有合适的资源?快使用搜索试试~ 我知道了~
资源推荐
资源详情
资源评论
摘 要
I
摘 要
《中国心血管健康与疾病报告 2020》中统计数据表明,心血管疾病的致死率高于
肿瘤等其它疾病,位于现阶段国内疾病死亡率的首位。充血性心力衰竭(Congestive
heart failure,CHF)患者在没有及时得到治疗的情况下其死亡概率会大幅度提升,若能
实现 CHF 的早期诊断,使病患及时得到治疗,可有效阻止病情进一步的恶化,从而降
低 CHF 疾病的死亡率。CHF 的临床诊断方式包括查看病史、分析病症和体征、心脏磁
共振、血液生化分析以及超声心动图等,由于心电信号在临床检测中具有高效、无创、
经济、简便等特性,因此得到广泛应用。心电信号自动分析技术可以辅助医疗人员提
高 CHF 的诊断效率,实现早诊断、早治疗,继而降低患者的死亡风险,同时也减轻了
医护人员的工作负担。
针对不同 CHF 患者心电信号形态多样、个体间差异性大和存在导联相关性等问题,
本文构建了基于心率变异性(Heart Rate Variability,HRV)多域特征融合的 CHF 自动
检测模型,充分利用 HRV 信号的导联无关性特征,从多个特征域对 HRV 信号进行表
征,实现了 CHF 的自动检测。针对 HRV 信号缺乏心动周期内的细节信息而导致 CHF
的检测精度较差的问题,提出了融合 HRV 和心电信号波形信息的 CHF 的自动检测模
型,从而实现了更为精准的 CHF 自动检测。具体研究内容如下:
(1)提取了 HRV 信号的时频域、混沌域和熵的特征,从不同的特征域角度表征
了 CHF 患者的 HRV 变化,通过多域特征的有效融合实现了 CHF 的精准检测。利用
Physionet 网站上的心电数据进行了实验测试,经五折交叉验证,正常人和 CHF 患者的
分类准确率、特异度和敏感度分别达到 96.00%、96.39%、95.52%。
(2)同步提取了心电信号的波形信息和 HRV 特征,将 HRV 特征拼接到波形信息
矩阵中。构建深度回声状态网络,自动提取融合信号中的多尺度特征,通过多尺度特
征的综合分析,实现了 CHF 的精准检测。经五折交叉验证,正常人和 CHF 患者的分类
准确率、特异度和敏感度分别达到 99.19%、98.64%、99.84%。进一步提升了 CHF 的检
测精度。
河北大学硕士学位论文
关键词 充血性心力衰竭 心电信号 心率变异性 特征融合 自动检测
Abstract
III
Abstract
The statistics in "China Cardiovascular Health and Disease report 2020" show that the
fatality rate of cardiovascular disease is higher than that of other diseases such as tumors, and
ranks first in the mortality rate of diseases in China at the present stage. Congestive heart failure
(CHF) patients are significantly more likely to die if they are not treated in a timely manner.
Early diagnosis of CHF and timely treatment of patients can effectively prevent further
deterioration of the disease, thus reducing the mortality rate of CHF disease. The clinical
diagnosis of CHF includes review of medical history, analysis of signs and symptoms, cardiac
magnetic resonance, blood biochemical analysis, and echocardiography. Because ECG is
efficient, non-invasive, economical and simple in clinical detection, it has been widely used.
The automatic analysis technology of ECG signals by computer can assist medical personnel
in early diagnosis and early treatment of congestive heart failure, which plays an important role
in reducing the burden of doctors and reducing the risk of death.
To address the problems of diverse ECG signal patterns, inter-individual variability and
lead correlation in different CHF patients, this paper constructs an automatic CHF detection
model based on the fusion of Heart Rate Variability (HRV) multi-domain features, which makes
full use of the lead-independent characteristics of HRV signals and characterizes HRV signals
from multiple feature domains to achieve The automatic detection of CHF was achieved by
making full use of the lead-independent feature of HRV signal and characterizing HRV signal
from multiple feature domains. To address the problem of poor CHF detection accuracy due to
the lack of detailed information within the cardiac cycle of HRV signals, an automatic CHF
detection model that fuses HRV feature and ECG signal waveform information is proposed to
achieve more accurate automatic CHF detection. The details of the study are as follows.
(1) The features of time-frequency domain, chaotic domain and entropy of HRV signal
were extracted to characterize the HRV changes of CHF patients from different feature domain
河北大学硕士学位论文
IV
perspectives, and the accurate detection of CHF was achieved by the effective fusion of multi-
domain features. Experimental tests were conducted using ECG data from the Physionet
website, and the classification accuracy, specificity and sensitivity reached 96.00%, 96.39%
and 95.52% for normal subjects and CHF patients, respectively, after a five-fold cross-
validation.
(2) The waveform information and HRV features of the ECG signal were extracted
simultaneously, and the HRV features were stitched into the waveform information matrix. A
deep echo state network was constructed to automatically extract the multiscale features in the
fused signals, and the accurate detection of CHF was achieved through the comprehensive
analysis of multiscale features. The classification accuracy, specificity and sensitivity of normal
subjects and CHF patients reached 99.19%, 98.64% and 99.84%, respectively, after a five-fold
cross-validation. Further improves the detection accuracy of CHF.
Keywords Congestive heart failure Electrocardiographic signals Heart rate variability
Feature fusion Automatic detection
目 录
V
目 录
第一章 绪论 ............................................................................................................................ 1
1.1 课题研究背景及意义 ............................................................................................... 1
1.2 国内外研究现状 ....................................................................................................... 2
1.2.1 基于 HRV 分析的 CHF 自动检测研究 ........................................................ 2
1.2.2 基于心电信号波形分析的 CHF 检测研究 .................................................. 4
1.3 论文主要工作及章节安排 ....................................................................................... 5
第二章 CHF 的心电信号及数据库 ....................................................................................... 7
2.1 CHF 的定义 .............................................................................................................. 7
2.2 正常人和 CHF 患者的心电信号特点 ..................................................................... 7
2.2.1 正常心电信号 ................................................................................................ 7
2.2.2 CHF 心电信号 ............................................................................................... 9
2.3 心电信号数据库 ..................................................................................................... 11
2.4 本章小结 ................................................................................................................. 11
第三章 基于 HRV 多域特征融合的 CHF 自动检测 .......................................................... 13
3.1 引言 ......................................................................................................................... 13
3.2 数据预处理 ............................................................................................................. 13
3.2.1 RR 间期提取 ................................................................................................ 14
3.2.2 数据归一化 .................................................................................................. 15
3.3 基于 HRV 特征融合的 CHF 自动检测 ................................................................. 15
3.3.1 HRV 多域特征提取 .................................................................................... 15
3.3.2 SVM 分类器 ................................................................................................ 21
3.4 算法验证和结果分析 ............................................................................................. 22
3.4.1 实验数据 ...................................................................................................... 23
3.4.2 特征参数的选择 .......................................................................................... 24
剩余56页未读,继续阅读
资源评论
2201_75761617
- 粉丝: 21
- 资源: 7339
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功