Table Of ContentDavid Zhang · Wangmeng Zuo
Peng Wang
Computational
Pulse Signal
Analysis
Computational Pulse Signal Analysis
David Zhang • Wangmeng Zuo • Peng Wang
Computational Pulse Signal
Analysis
David Zhang Wangmeng Zuo
School of Science and Engineering Harbin Institute of Technology
The Chinese University of Hong Kong Harbin, China
Shenzhen, Guangdong, China
Peng Wang
Northeast Agricultural University
Harbin, China
ISBN 978-981-10-4043-6 ISBN 978-981-10-4044-3 (eBook)
https://doi.org/10.1007/978-981-10-4044-3
Library of Congress Control Number: 2018955291
© Springer Nature Singapore Pte Ltd. 2018
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, express or implied, with respect to the material contained herein or for any errors
or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims
in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Preface
Traditional Chinese diagnostics is a fundamental component in traditional Chinese
medicine (TCM). In general, there are four major diagnostic methods of TCM, i.e.,
looking, listening, asking, and feeling the pulse. Among them, pulse diagnosis (i.e.,
feeling the pulse) is operated by placing the three fingers of the practitioner at the
wrist radial artery of the patient for analyzing the health condition. For thousands of
years, pulse diagnosis has played an indispensable role in TCM and traditional
Ayurvedic medicine (TAM). Due to its convenient, inexpensive, and noninvasive
properties, even nowadays pulse diagnosis is still very competitive for disease
diagnosis.
Recent studies have revealed that wrist pulse signal is a kind of bloodstream
signal influenced by many physiological or pathological factors and can be applied
for disease analyses. However, the practice of traditional Chinese pulse diagnosis
(TCPD) extremely depends on the experience of the practitioners. The measure-
ment and interpretation in TCPD generally require years of training of the practitio-
ners. It is also difficult for different practitioners to share their feelings on the pulse
signal. All these restrict its development and applications in contemporary clinical
practice.
Fortunately, with the development on sensors, signal processing, and pattern rec-
ognition, considerable progresses have been achieved in computational pulse signal
analysis. With the advances in sensor technologies, three types of sensors, e.g., pres-
sure, photoelectric, and ultrasonic sensors, have been developed for pulse signal
acquisition. To simulate the practitioners in analyzing the pulse signal, signal pro-
cessing and pattern recognition methods have been developed. By far, pulse signals
have been investigated for pulse waveform classification and the diagnosis of many
diseases, such as cholecystitis, nephrotic syndrome, diabetes, etc.
In this book, we intend to provide an in-depth summary to the latest advances in
pulse signal acquisition, processing, and applications in classification and diagno-
sis. The system design, model and algorithm implementation, experimental evalua-
tion, and underlying rationales are also given in the book. Following the pipeline of
computational pulse signal analysis, the book is organized into six parts. In the first
part, Chap. 1 introduces the connection between wrist pulse signal and cardiac
v
vi Preface
electrical activity, which lays a physiological foundation for pulse diagnosis.
Subsequently, we provide an overview on the practice of TCPD and the pipeline of
computational pulse analysis.
In the second part, pulse acquisition systems are introduced to capture pulse
signals at representative positions, under various pressures, and from different types
of sensors. In Chap. 2, we introduce a compound multiple-channel pressure signal
acquisition system. By equipping with sensor array design and pressure adjustment,
the system can capture multichannel pulse signals and is effective in measuring the
width of the pulse. Chapter 3 integrates a pressure sensor with a photoelectric sensor
to acquire more pulse information. The photoelectric sensor array is used to detect
the pulse width and the center of radial artery, while the pressure sensor measures
the pulsations with high resolution.
In the third part, several representative preprocessing methods are described for
baseline wander correction and detection of low-quality pulse signal. In Chap. 4, we
present an energy ratio-based criterion to evaluate the level of baseline drift and a
wavelet-based cascaded adaptive filter to remove baseline drift. In Chap. 5, we con-
sider two types of corruption, i.e., saturation and artifact. For the detection of satura-
tion, we use two criteria based on its definition. For the artifact detection, we suggest
a complex network-based scheme by measuring the network connectivity. Finally,
Chap. 6 presents an optimal preprocessing framework by integrating frequency-
dependent analysis, curve fitting, period segmentation, and normalization.
The fourth part introduces the feature extraction of wrist pulse signal. In Chap.
7, the Lempel-Ziv complexity analysis is adopted to detect arrhythmic pulses. In
Chap. 8, the spatial features and spectrum feature are extracted from blood flow
velocity signal. In Chap. 9, generalized 2-D matrix feature is extracted to character-
ize the periodic and nonperiodic information. In Chap. 10, complex network is
introduced to transform the pulse signal from time domain to network domain, and
multi-scale entropy is used to measure the inter- and intra-cycle variations of pulse
signal.
The fifth part presents several representative classification methods for the rec-
ognition and diagnosis of pulse signal. In Chap. 11, the ERP-based KNN classifiers
are developed for pulse waveform classification. In Chap. 12, a modified Gaussian
model is used for modeling pulse signal and a fuzzy C-means (FCM) classifier is
adopted for computational pulse diagnosis. In Chap. 13, the residual error of auto-
regressive (AR) model is utilized for disease diagnosis. In Chap. 14, we present a
multiple kernel learning model for the integration of heterogeneous features for
pulse classification and diagnosis.
Finally, in the sixth part, some discussions are provided to reveal the relationship
between different types of pulse signals. In Chap. 15, we analyze the physical mean-
ings and sensitivities of signals acquired by different types of pulse signal acquisi-
tion systems to guide the sensor selection for computational pulse diagnosis. In
Chap. 16, a comparative study on pulse and ECG signals is conducted to reveal their
complementarities. Finally, Chap. 17 provides a brief recapitulation on the main
content of this book.
Preface vii
The book is based on our years of researches on computational pulse signal anal-
ysis. Since 2003, under the grant support from the National Natural Science
Foundation of China (NSFC), we have published our first chapter on computational
pulse signal analysis. Since then, more and more researches have been conducted in
this ever-growing field, and we have systematically studied the acquisition, prepro-
cessing, feature extraction, and classification of pulse signals. With several typical
diseases such as gallbladder diseases and diabetes, we also show the feasibility of
pulse signal for disease diagnosis. We would like to express our special thanks to
Mr. Zhaotian Zhang, Mr. Ke Liu, and Ms. Xiaoyun Xiong from NSFC, who consis-
tently supported our research work for decades.
We would like to express our gratitude to our colleagues and PhD students, i.e.,
Prof. Naimin Li, Prof. Kuanquan Wang, Prof. Jie Zhou, Prof. Lisheng Xu, Prof.
Guangming Lu, Prof. Yong Xu, Prof. Jane You, Prof. Lei Zhang, Dr. Hongzhi Zhang,
Dr. Yinghui Chen, Dr. Dongyu Zhang, Dr. Lei Liu, and Dr. Dimin Wang, for their
contributions to the research achievements on this topic. It is our great honor to
work with them in this inspiring topic in the previous years. The authors owe a debt
of gratitude to Mr. Pengju Liu for his careful reading and for checking the draft of
the manuscript. We are also hugely indebted to Ms. Celine L. Chang and Ms. Jane
Li of Springer for their consistent help and encouragement. Finally, the work in this
book was mainly sponsored by the NSFC Program under Grant Nos. 61332011,
61271093, and 61471146.
The Chinese University of Hong Kong David Zhang
Shenzhen, Guangdong, China
July, 2017
Contents
Part I Background
1 Introduction: Computational Pulse Diagnosis ..................................... 3
1.1 Principle of Pulse Signal ................................................................ 3
1.2 Traditional Pulse Diagnosis ........................................................... 4
1.3 Computational Pulse Signal Analysis ............................................ 5
1.4 Summary ........................................................................................ 10
References ................................................................................................. 10
Part II Pulse Signal Acquisition
2 Compound Pressure Signal Acquisition ................................................ 13
2.1 Introduction .................................................................................... 13
2.2 Application Scenario and Requirement Analysis .......................... 15
2.3 System Architecture ....................................................................... 16
2.3.1 Mechanical Structure ....................................................... 16
2.3.2 Sensor .............................................................................. 18
2.3.3 Circuit .............................................................................. 20
2.3.4 Summary .......................................................................... 23
2.4 System Evaluation ......................................................................... 24
2.4.1 Sampled Pulse Signals ..................................................... 25
2.4.2 Computational Pulse Diagnosis ....................................... 28
2.4.3 Comparisons with Other Pulse Sampling Systems .......... 31
2.5 Summary ........................................................................................ 32
References ................................................................................................. 32
3 Pulse Signal Acquisition Using Multi-sensors ...................................... 35
3.1 Introduction .................................................................................... 35
3.2 Framework of the Proposed System .............................................. 37
3.2.1 Pulse Collecting ............................................................... 38
3.2.2 Pulse Processing and Interaction Design ......................... 39
3.3 Design of the Different Sensor Arrays ........................................... 40
ix
x Contents
3.3.1 Pressure Sensor ................................................................ 41
3.3.2 Photoelectric Sensor Array .............................................. 44
3.3.3 Combination of Pressure and Photoelectric
Sensor Arrays ................................................................... 45
3.4 Multichannel Optimization ............................................................ 47
3.4.1 Selection of Base Channel ............................................... 49
3.4.2 Multichannel Selection .................................................... 53
3.5 The Optimization of Different Sensors Fusion .............................. 56
3.6 Experimental Results ..................................................................... 57
3.6.1 Experiment 1 .................................................................... 5 8
3.6.2 Experiment 2 .................................................................... 5 9
3.7 Summary ........................................................................................ 60
References ................................................................................................. 61
Part III Pulse Signal Preprocessing
4 Baseline Wander Correction in Pulse Waveforms Using
Wavelet-Based Cascaded Adaptive Filter ............................................. 65
4.1 Introduction .................................................................................... 65
4.1.1 Pulse Waveform Analysis ................................................ 65
4.1.2 Related Works on Baseline Drift Removal ...................... 68
4.2 The Proposed CAF ........................................................................ 69
4.2.1 The Design of CAF .......................................................... 69
4.2.2 Detection Level of Baseline Drift Using ER ................... 71
4.2.3 The Discrete Meyer Wavelet Filter .................................. 75
4.2.4 Cubic Spline Estimation Filter ......................................... 78
4.3 Simulated Signals: Experimental Results and Analysis ................ 81
4.3.1 Experimental Results of the CAF for Different
Baseline Drifts ................................................................. 81
4.3.2 Experimental Results for Different ER Thresholds ......... 85
4.3.3 Experimental Results for Several Typical Pulses ............ 86
4.4 Experimental Results for Actual Pulse Records ............................ 87
4.5 Summary ........................................................................................ 88
References ................................................................................................. 89
5 Detection of Saturation and Artifact ..................................................... 91
5.1 Introduction .................................................................................... 91
5.2 Saturation and Artifact ................................................................... 92
5.2.1 Saturation ......................................................................... 92
5.2.2 Artifact ............................................................................. 93
5.3 The Detection of Saturation and Artifact ....................................... 94
5.3.1 The Preprocessing and the Priority .................................. 94
5.3.2 Saturation Detection ........................................................ 97
5.3.3 Artifact Detection ............................................................ 98
5.4 Experimental Results ..................................................................... 102
Contents xi
5.4.1 Saturation Detection ........................................................ 102
5.4.2 Artifact Detection ............................................................ 102
5.5 Summary ........................................................................................ 103
References ................................................................................................. 106
6 Optimized Preprocessing Framework for Wrist Pulse Analysis ........ 109
6.1 Introduction .................................................................................... 109
6.2 Description of Pulse Database ....................................................... 111
6.2.1 Data Acquisition .............................................................. 111
6.2.2 Time Domain Characteristic ............................................ 112
6.2.3 Frequency Domain Characteristic ................................... 113
6.3 Proposed Pulse Preprocessing Method .......................................... 116
6.3.1 Pulse Denoising ............................................................... 117
6.3.2 Interval Selection ............................................................. 118
6.3.3 Baseline Drift Removal.................................................... 119
6.3.4 Period Segmentation and Normalization ......................... 122
6.4 Experiments on Actual Pulse Database ......................................... 124
6.4.1 Comparison of Pulse Denoising ...................................... 124
6.4.2 Optimal Segmentation Strategy ....................................... 125
6.4.3 Preprocessing for Pulse Diagnosis ................................... 128
6.5 Summary ........................................................................................ 130
References ................................................................................................. 131
Part IV Pulse Signal Feature Extraction
7 Arrhythmic Pulse Detection ................................................................... 135
7.1 Introduction .................................................................................... 135
7.2 Clinical Value of Pulse Rhythm Analysis ...................................... 136
7.3 The Approach to Automatic Recognition of Pulse Rhythms ......... 136
7.3.1 Lempel–Ziv Complexity Analysis ................................... 138
7.3.2 Definitions and Basic Facts.............................................. 138
7.3.3 Automatic Recognition of Pulse Patterns Distinctive
in Rhythm ........................................................................ 142
7.4 Experiments ................................................................................... 151
7.5 Summary ........................................................................................ 154
References ................................................................................................. 154
8 Spatial and Spectrum Feature Extraction ............................................ 157
8.1 Introduction .................................................................................... 157
8.2 Data Acquisition and Preprocessing .............................................. 159
8.3 Feature Extraction .......................................................................... 160
8.3.1 Spatial Feature Extraction of Blood Flow
Velocity Signal ................................................................. 160
8.3.2 EMD-Based Spectrum Feature Extraction ...................... 161
8.4 Experimental Result and Discussion ............................................. 163
8.5 Summary ........................................................................................ 166
References ................................................................................................. 166
