Biosignal and Medical Image Processing 3rd Edition John L. Semmlow 2025 Scribd Download
Biosignal and Medical Image Processing 3rd Edition John L. Semmlow 2025 Scribd Download
com
https://ebookgate.com/product/biosignal-and-medical-image-
processing-3rd-edition-john-l-semmlow/
OR CLICK HERE
DOWLOAD NOW
https://ebookgate.com/product/manufacturing-design-production-
automation-and-integration-1st-edition-john-l-semmlow/
ebookgate.com
https://ebookgate.com/product/medical-image-processing-reconstruction-
and-restoration-concepts-and-methods-1st-edition-jiri-jan/
ebookgate.com
https://ebookgate.com/product/the-image-processing-handbook-5th-
edition-john-c-russ/
ebookgate.com
https://ebookgate.com/product/the-image-processing-handbook-7th-
edition-john-c-russ/
ebookgate.com
Information Fusion in Signal and Image Processing Digital
Signal and Image Processing 1st Edition Isabelle Bloch
https://ebookgate.com/product/information-fusion-in-signal-and-image-
processing-digital-signal-and-image-processing-1st-edition-isabelle-
bloch/
ebookgate.com
https://ebookgate.com/product/the-image-processing-handbook-seventh-
edition-russ/
ebookgate.com
https://ebookgate.com/product/image-processing-principles-and-
applications-1st-edition-tinku-acharya/
ebookgate.com
https://ebookgate.com/product/digital-signal-and-image-processing-1st-
edition-tamal-bose/
ebookgate.com
https://ebookgate.com/product/logic-based-nonlinear-image-processing-
stephen-marshall/
ebookgate.com
BIOSIGNAL
and
MEDICAL
IMAGE
PROCESSING
Third Edition
JOHN L. SEMMLOW
BENJAMIN GRIFFEL
Biosignal
and
Medical
iMage
Processing
Third Edition
BIOSIGNAL
and
MEDICAL
IMAGE
PROCESSING
Third Edition
JOHN L. SEMMLOW
BENJAMIN GRIFFEL
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2014 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been
made to publish reliable data and information, but the author and publisher cannot assume responsibility for the valid-
ity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright
holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may
rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or uti-
lized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopy-
ing, microfilming, and recording, or in any information storage or retrieval system, without written permission from the
publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://
www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For
organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
This book is dedicated to the memory of Larry Stark, who showed
me the many possibilities and encouraged me to take them.
Contents
Preface........................................................................................................................................ xv
Acknowledgments................................................................................................................... xix
Authors..................................................................................................................................... xxi
Chapter 1 Introduction...........................................................................................................1
1.1 Biosignals.....................................................................................................1
1.2 Biosignal Measurement Systems...............................................................3
1.3 Transducers.................................................................................................4
1.4 Amplifier/Detector.....................................................................................6
1.5 Analog Signal Processing and Filters.......................................................7
1.5.1 Filter Types.....................................................................................8
1.5.2 Filter Bandwidth............................................................................9
1.5.3 Filter Order.....................................................................................9
1.5.4 Filter Initial Sharpness................................................................12
1.6 ADC Conversion.......................................................................................13
1.6.1 Amplitude Slicing........................................................................15
1.6.2 Time Slicing..................................................................................18
1.6.3 Edge Effects.................................................................................. 22
1.6.4 Buffering and Real-Time Data Processing.............................. 24
1.7 Data Banks................................................................................................ 24
1.8 Summary................................................................................................... 24
Problems.............................................................................................................. 25
vii
Contents
2.3.1.2 Orthogonality.............................................................. 48
2.3.1.3 Basis Functions............................................................ 50
2.3.2 Correlation-Based Analyses.......................................................52
2.3.2.1 Correlation and Covariance.......................................53
2.3.2.2 Matrix of Correlations................................................ 54
2.3.2.3 Cross-Correlation....................................................... 56
2.3.2.4 Autocorrelation............................................................59
2.3.2.5 Autocovariance and Cross-Covariance....................62
2.3.3 Convolution and the Impulse Response.................................. 64
2.4 Summary....................................................................................................70
Problems...............................................................................................................71
viii
Contents
ix
Contents
x
Contents
10.7 Summary..................................................................................................353
Exercises..............................................................................................................353
xi
Contents
Bibliography............................................................................................................................ 589
Index........................................................................................................................................ 593
xiii
Preface
Signal processing can be broadly defined as the application of analog or digital techniques to
improve the utility of data. In biomedical engineering applications, improved utility usually
means that the data provide better diagnostic information. Analog techniques are sometimes
applied to data represented by a time-varying electrical signal, but most signal-processing meth-
ods take place in the digital domain where data are represented as discrete numbers. These
numbers may represent a time-varying signal, or an image. This book deals exclusively with
signal processing of digital data, although the Introduction briefly describes analog processes
commonly found in medical devices.
This book should be of interest to a broad spectrum of engineers, but it is written specifically
for biomedical engineers (called bioengineers in some circles). Although the applications are
different, the signal-processing methodology used by biomedical engineers is identical to that
used by electrical and computer engineers. The difference for biomedical engineers is in the
approach and motivation for learning signal-processing technology. A computer engineer may
be required to develop or modify signal-processing tools, but for biomedical engineers these
techniques are tools to be used. For the biomedical engineer, a detailed understanding of the
underlying theory, while always of value, is not usually needed. What is needed is an under-
standing of what these tools do and how they can be employed to meet biomedical challenges.
Considering the broad range of knowledge required to be effective in biomedical engineering,
encompassing both medical and engineering domains, an in-depth understanding of all of the
useful technology is not realistic. It is most important to know what tools are available, have a
good understanding of what they do (if not how they do it), be aware of the most likely pitfalls
and misapplications, and know how to implement these tools given available software packages.
The basic concept of this book is that just as the cardiologist can benefit from an oscilloscope-
type display of the ECG without a deep understanding of electronics, so a biomedical engineer
can benefit from advanced signal-processing tools without always understanding the details of
the underlying mathematics.
In line with this philosophy, most of the concepts covered in this book can be broken
down into two components. The first component is a presentation of a general understand-
ing of the approach sufficient to allow intelligent application of the concepts. Sometimes
this includes a description of the underlying mathematical principles. The second section
describes how these tools can be implemented using the MATLAB ® software package and
several of its toolboxes.
This book was originally written for a single-semester course combining signal and image
processing. Classroom experience indicates that this ambitious objective is possible for most
graduate formats, although eliminating a few topics may be desirable. For example, some of the
introductory or basic material covered in Chapters 1 and 2 could be skipped or treated lightly
for students with the appropriate prerequisites. In addition, topics such as Advanced Spectral
Methods (Chapter 5), Time–Frequency Analysis (Chapter 6), Wavelets (Chapter 7), Advanced
Filters (Chapter 8), and Multivariate Analysis (Chapter 9) are pedagogically independent and
can be covered as desired without affecting the other material. With the inclusion of Chapters
xv
Preface
10 and 11 on nonlinear signal processing, the material on image processing could be deleted and
the course restricted to one-dimensional signal processing.
Nonlinear signal analysis is an emerging area within signal processing that recognizes that
real biomedical signals do not always conform to our assumptions of linearity. For such signals,
nonlinear analysis can reveal additional information that linear analysis cannot. New tech-
niques and algorithms are being continuously added, but Chapters 10 and 11 provide a founda-
tion for those interested in nonlinear signal analyses. These techniques have gained a foothold
in the area of heart rate and EEG analysis, where there is evidence of nonlinear behavior. Other
new biomedical applications include DNA sequence analysis and speech processing. Nonlinear
signal processing can only grow in importance as the techniques are improved offering greater
speed and better immunity to noise, and as new biomedical applications are discovered.
This third edition of this textbook was undertaken to improve clarity and understanding by
expanding both examples and problems. Although much of the material covered here will be new
to most students, the book is not intended as an “introductory” text since the goal is to provide
a working knowledge of the major signal-processing methods without the need for additional
course work. The challenge of covering a broad range of topics at a useful, working depth is moti-
vated by current trends in biomedical engineering education, particularly at the graduate level
where a comprehensive education must be attained with a minimum number of courses. This has
led to the development of “core” courses to be taken by all students. This book was written for just
such a core course in the Graduate Program of Biomedical Engineering of Rutgers University. It is
also quite suitable for an upper-level undergraduate course and would also be of value for students
in other disciplines that would benefit from a working knowledge of signal and image processing.
It is not possible to cover such a broad spectrum of material to a depth that enables productive
application without heavy reliance on MATLAB-based examples and problems. In this regard,
the book assumes that the student has some knowledge of MATLAB programming and has the
basic MATLAB software package at hand, including the Signal Processing and Image Processing
Toolboxes. (MATLAB also has toolboxes for wavelets and adaptive neural nets, but these sections
are written so as not to require additional toolboxes, primarily to keep the number of required
toolboxes to a minimum.) The problems are an essential part of this book and often provide a dis-
covery-like experience regarding the associated topic. The code used for all examples is provided in
the software package that accompanies this book, available at http://www.crcpress.com/product/
isbn/9781466567368. Since many of the problems are extensions or modifications of examples
given in the book, time can be saved by starting with the code of a related example. The package
also includes support routines and data files used in the examples and problems, and the code used
to generate many of the figures.
For instructors, there is an Education Support Package available that contains the prob-
lem solutions and PowerPoint® presentations for each of the chapters. These PowerPoint slides
include figures, equations, and text and can be modified by the instructor as desired.
In addition to heavy reliance on MATLAB problems and examples, this book makes exten-
sive use of simulated data. Examples involving biological signals are only occasionally used. In
our view, examples using biological signals provide motivation, but they are not generally very
instructive. Given the wide range of material to be presented at a working depth, emphasis is
placed on learning the tools of signal processing; motivation is left to the reader (or the instructor).
Organization of the book is straightforward. Chapters 1 through 4 are fairly basic. Chapter 1
covers topics related to analog signal processing and data acquisition, while Chapter 2 includes
topics that are basic to all aspects of signal and image processing. Chapters 3 and 4 cover clas-
sical spectral analysis and basic digital filtering, topics fundamental to any signal-processing
course. Advanced spectral methods are covered in Chapter 5 and are important due to their
widespread use in biomedical engineering. Chapter 6 and the first part of Chapter 7 cover top-
ics related to spectral analysis when the signal’s spectrum is varying in time, a condition often
found in biological signals. Chapter 7 addresses both continuous and discrete wavelets, another
popular technique used in the analysis of biomedical signals. Chapters 8 and 9 feature advanced
xvi
Preface
topics. In Chapter 8, optimal and adaptive filters are covered: the latter’s inclusion is also moti-
vated by the time-varying nature of many biological signals. Chapter 9 introduces multivariate
techniques, specifically Principal Component Analysis and Independent Component Analysis,
two analysis approaches that are experiencing rapid growth with regard to biomedical applica-
tions. Chapters 10 and 11 cover nonlinear signal-processing methods. The next three chapters
are about image processing with the first of these, Chapter 12, covering the conventions used
by MATLAB’s Imaging Processing Toolbox. Image processing is a vast area and the material
included here is limited primarily to areas associated with medical imaging: image acquisition
(Chapter 15); image filtering, enhancement, and transformation (Chapter 13); and segmentation
and registration (Chapter 14). The last two chapters concern classification: linear discriminators
and support vector machines in Chapter 16, and adaptive neural nets in Chapter 17.
Some chapters cover topics that may require an entire book for thorough presentation. Such
chapters involve a serious compromise and omissions are extensive. Our excuse is that class-
room experience with this approach seems to work: students end up with a working knowledge
of a vast array of signal- and image-processing tools. A few of the classic or major books on these
topics are cited in a bibliography at the end of the book. No effort has been made to construct an
extensive bibliography or reference list since more current lists are readily available off the Web.
Textbook Protocols
Some early examples of MATLAB code are presented in full, but in most examples some of the
routine code (such as for plotting, display, and labeling operation) is omitted. Nevertheless, we
recommend that students carefully label (and scale when appropriate) all graphs done in the
xvii
Preface
problems. Some effort has been made to use consistent notation as described in Table I. In gen-
eral, lower-case letters n and k are used as data indices, and capital letters (i.e., N or L) are used to
indicate the signal length (or maximum subscript value) or other constants. In two-dimensional
data sets, lower-case letters m and n are used to indicate the row and column subscripts of an
array, while capital letters M and N are used to indicate vertical and horizontal dimensions,
respectively. The letter m is also used as the index of a variable produced by a transformation,
or as an index indicating a particular member of a family of related functions.* As is common,
analog or continuous time variables are denoted by a “t” within parentheses (i.e., x(t)), whereas
brackets enclose subscripts of digitized signals composed of discrete variables (i.e., x[n]). Other
notation follows either standard or MATLAB conventions.
Italics is used to introduce important new terms that should be incorporated into the reader’s
vocabulary. If the meaning of these terms is not obvious from their use, they are explained
where they are introduced. All MATLAB commands, routines, variables, and code are shown in
courier typeface. Single quotes are used for string variables following MATLAB conven-
tions. These and other protocols are summarized in Table I.
MATLAB® and Simulink® are registered trademarks of The MathWorks, Inc. For product infor-
mation, please contact:
* For example, m would be used to indicate the harmonic number of a family of harmonically related sine functions;
that is, f m(t) = sin(2pmt).
xviii
Acknowledgments
John L. Semmlow wishes to thank Susanne Oldham who, from the quaint little village of Yellow
Springs, Ohio, managed to edit this book, and provided strong, continuing encouragement and
support. He would also like to acknowledge the patience and support of Peggy Christ and Lynn
Hutchings. Benjamin Griffel would like to particularly thank Professor Tony Shinbrot for dis-
cussions on chaos and nonlinear dynamics that contributed greatly to Chapters 10 and 11. Both
authors are deeply indebted to Raymond Muzic of Case Western University for his thoughtful
comments, corrections, and insights regarding many of the topics in this book. Finally, we both
would like to thank our students who provided suggestions and whose enthusiasm for the mate-
rial provided much-needed motivation.
xix
Authors
John L. Semmlow was born in Chicago, in 1942, and received BSEE from the University of
Illinois in Champaign in 1964. Following several years as a design engineer for Motorola, Inc.,
he entered the Bioengineering Program at the University of Illinois Medical Center in Chicago,
and received his PhD in 1970. He has held faculty positions at the University of California,
Berkeley, the University of Illinois, Chicago, and a joint Professorship in Surgery, Robert Wood
Johnson Medical School and Biomedical Engineering, School of Engineering, Rutgers University
in New Jersey. In 1985 he was a NSF/CNRS (National Science Foundation/Centre National de
la Recherche Scientifique) Fellow in the Sensorimotor Control Laboratory of the University of
Provence, Marseille, France. He was appointed a Fellow of the IEEE (Institute of Electrical and
Electronics Engineers) in 1994 in recognition of his work on acoustic detection of coronary
artery disease. He was elected Fellow of AIMBE (American Institute for Medical and Biological
Engineering) in 2003 and Fellow of BMES (Biomedical Engineering Society) in 2005. He was
founding chair of the International Conference on Vision and Movement in Man and Machines,
first held in Berkeley in 1995. His primary research interests include discovering strategies for
how the brain controls human motor behavior such as eye movements and the development of
medical instrumentation for noninvasive detection of coronary artery disease.
He is an avid, if somewhat dangerous, skier, and an enthusiastic, but mediocre, tennis player.
He teaches folk dance and is president of the Board of the New Brunswick Chamber Orchestra.
He especially enjoys travel and the slow, but nonetheless expensive, sport of sailing. He recently
purchased a sailboat (currently in the British Virgin Islands) and he soon hopes to launch his
second career in computer-controlled kinetic art.
Benjamin Griffel was born in Brooklyn, New York. He received bachelor’s and PhD degrees
from Rutgers University in biomedical engineering in 2005 and 2011, respectively, and MS from
Drexel University in 2007. His thesis topic explored the use of chaotic and entropy-based signal
analysis to diagnose coronary artery disease using acoustic cardiovascular recordings. He is
currently an INSPIRE (IRACDA New Jersey/New York for Science Partnerships in Research and
Education) fellow at Rutgers University and a teaching partner at New Jersey City University.
INSPIRE is an NIH (National Institutes of Health) fellowship designed to help young research-
ers develop teaching skills and to increase diversity within the areas of STEM (science, technol-
ogy, engineering, and mathematics). His current research examines the use of nonlinear and
chaotic signal analysis to analyze the effects of inflammation on heart rate recordings. In his
spare time, Benjamin enjoys guitar playing, arm chair philosophizing, and spending time with
his wife.
xxi
1
Introduction
1.1 Biosignals
Much of the activity in biomedical engineering, be it clinical or research, involves the measure-
ment, processing, analysis, display, and/or generation of biosignals. Signals are variations in
energy that carry information, and the search for information from living systems has been a
preoccupation of medicine since its beginnings (Figure 1.1). This chapter is about the modifica-
tion of such signals to make them more useful or more informative. To this end, this chapter
presents tools and strategies that use digital signal processing to enhance the information con-
tent or interpretation of biosignals.
The variable that carries information (the specific energy fluctuation) depends on the type of
energy involved. Biological signals are usually encoded into variations of electrical, chemical,
or mechanical energy, although, occasionally, variations in thermal energy are of interest. For
communication within the body, signals are primarily encoded as variations in electrical or
chemical energy. When chemical energy is used, encoding is usually done by varying the con-
centration of the chemical within a “physiological compartment,” for example, the concentra-
tion of a hormone in blood. Bioelectric signals use the flow or concentration of ions, the primary
charge carriers within the body, to transmit information. Speech, the primary form of commu-
nication between humans, encodes information as variations in pressure. Table 1.1 summarizes
the different types of energy that can be used to carry information and the associated variables
that encode this information. Table 1.1 also shows the physiological measurements that involve
these energy forms.
Outside the body, information is commonly transmitted and processed as variations in elec-
trical energy, although mechanical energy was used in the seventeenth and early eighteenth
centuries to send messages. The semaphore telegraph used the position of one or more large
arms placed on a tower or high point to encode letters of the alphabet. These arm positions could
be observed at some distance (on a clear day) and relayed onward if necessary. Information pro-
cessing can also be accomplished mechanically, as in the early numerical processors constructed
by Babbage. Even mechanically based digital components have been attempted using variations
in fluid flow. Modern electronics provides numerous techniques for modifying electrical sig-
nals at very high speeds. The body also uses electrical energy to carry information when speed
is important. Since the body does not have many free electrons, it relies on ions, notably Na+,
K+, and Cl−, as the primary charge carriers. Outside the body, electrically based signals are so
useful that signals carried by other energy forms are usually converted into electrical energy
when significant transmission or processing tasks are required. The conversion of physiological
1
Biosignal and Medical Image Processing
Figure 1.1 Information about internal states of the body can be obtained through the acquisition
and interpretation of biosignals. Expanding such information is an ongoing endeavor of medicine
and medical research. It is also the primary motivation of this chapter.
2
1.2 Biosignal Measurement Systems
energy into an electric signal is an important step, often the first step, in gathering information
for clinical or research use. The energy conversion task is done by a device termed a transducer,*
specifically, a biotransducer. The biotransducer is usually the most critical component in systems
designed to measure biosignals.
With the exception of this chapter, this book is limited to topics on digital signal and image
processing. To the extent possible, each topic is introduced with the minimum amount of infor-
mation required to use and understand the approach along with enough information to apply
the methodology in an intelligent manner. Strengths and weaknesses of the various methods are
also explored, particularly through discovery in the problems at the end of the chapter. Hence,
the problems at the end of each chapter, most of which utilize the MATLAB® software package
(Waltham, MA), constitute an integral part of the book and a few topics are introduced only in
the problems.
A fundamental assumption of this book is that an in-depth mathematical treatment of sig-
nal-processing methodology is not essential for effective and appropriate application of these
tools. This book is designed to develop skills in the application of signal- and image-processing
technology, but it may not provide the skills necessary to develop new techniques and algo-
rithms. References are provided for those who need to move beyond the application of sig-
nal- and image-processing tools to the design and development of new methodology. In the
subsequent chapters, the major topics include sections on implementation using the MATLAB
software package. Fluency with the MATLAB language is assumed and is essential for the use of
this book. Where appropriate, a topic area may also include a more in-depth treatment, includ-
ing some of the underlying mathematics.
Physiol. Analog
Amplifier/
or Transducer signal
detector
External processing
energy
Classification
Analog- Signal
to-digital storage Display
or Signal
converter processing
buffering
Computer
* Learning the vocabulary is an important part of mastering a discipline. In this book, the commonly used terms are
highlighted using italics. Sometimes, the highlighted term is described when it is introduced, but occasionally, the
determination of its definition is the responsibility of the reader.
3
Biosignal and Medical Image Processing
and the acquisition of different types of diagnostic information. Information from the biological
process of interest must first be converted into an electric signal via the transducer. Some analog
signal processing is usually required, often including amplification and lowpass (or bandpass)
filtering. Since most signal processing is easier to implement using digital methods, the analog
signal is converted into a digital format using an analog-to-digital converter (ADC). Once con-
verted, the signal is often stored or buffered in the memory to facilitate subsequent signal pro-
cessing. Alternatively, in real-time applications, the incoming data are processed as they come
in, often with minimal buffering, and may not be permanently stored. Digital signal-processing
algorithms can then be applied to the digitized signal. These signal-processing techniques can
take on a wide variety of forms with varying levels of sophistication and they make up the major
topic areas of this book. In some applications such as diagnosis, a classification algorithm may
be applied to the processed data to determine the state of a disease or the class of a tumor or
tissue. A wide variety of classification techniques exist and the most popular techniques are dis-
cussed in Chapters 16 and 17. Finally, some sort of output is necessary in any useful system. This
usually takes the form of a display, as in imaging systems, but it may be some type of effector
mechanism such as in an automated drug delivery system. The basic elements shown in Figure
1.2 are discussed in greater detail in the next section.
1.3 Transducers
A transducer is a device that converts energy from one form to another. By this definition, a
lightbulb or a motor is a transducer. In signal-processing applications, energy is used to carry
information; the purpose of energy conversion is to transfer that information, not to transform
energy as with a lightbulb or a motor. In measurement systems, all transducers are so-called
input transducers: they convert nonelectrical energy into an electronic signal. An exception to
this is the electrode, a transducer that converts ionic electrical energy into electronic electrical
energy. Usually, the output of a biomedical transducer is voltage (or current) whose amplitude is
proportional to the measured energy (Figure 1.3).
Input transducers use one of the two different fundamental approaches: the input energy
causes the transducer element to generate voltage or current, or the input energy creates a change
in an electrical property (the resistance, inductance, or capacitance) of the transducer material.
Most optical transducers use the first approach. Photons strike a photo sensitive material pro-
ducing free electrons (or holes) that can then be detected as external current flow. Piezoelectric
devices used in ultrasound also generate a charge when under mechanical stress. Many exam-
ples can be found in the use of the second category, a change in some electrical property. For
example, metals (and semiconductors) undergo a consistent change in resistance with changes
in temperature and most temperature transducers utilize this feature. Other examples include
the strain gage, which measures mechanical deformation using the small change in resistance
that occurs when the sensing material is stretched.
The energy that is converted by the input transducer may be generated by the physiological
process itself, it may be energy that is indirectly related to the physiological process, or it may be
energy produced by an external source. In the last case, the externally generated energy inter-
acts with, and is modified by, the physiological process and it is this alteration that produces
Varying
Energy voltage or
variation current
Biotransducer
4
1.3 Transducers
the measurement. For example, when externally generated x-rays are transmitted through the
body, they are absorbed by the intervening tissue and a measurement of this absorption is used
to construct an image. Many diagnostically useful imaging systems are based on this external
energy approach.
In addition to images created by external energy that is passed through the body, some images
are generated using the energy of radioactive emissions of radioisotopes injected into the body.
These techniques make use of the fact that selected or “tagged,” molecules will collect in the
specific tissue. The areas where these radioisotopes collect can be mapped using a γ-camera or
with certain short-lived isotopes, which are localized better using positron emission tomography
(PET).
Many physiological processes produce energy that can be detected directly. For example,
cardiac internal pressures are usually measured using a pressure transducer placed on the tip
of a catheter introduced into the appropriate chamber of the heart. The measurement of elec-
trical activity in the heart, muscles, or brain provides other examples of the direct measure-
ment of physiological energy. For these measurements, the energy is already electrical and only
needs to be converted from ionic into electronic current using an “electrode.” These sources
are usually given the term ExG, where “x” represents the physiological process that produces
electrical energy, ECG—electrocardiogram, EEG—electroencephalogram, EMG—electromyo
gram, EOG—electrooculogram, ERG—electroretinogram, and EGG—electrogastrogram. An
exception to this terminology is the galvanic skin response (GSR), the electrical activity gener-
ated by the skin. The typical physiological measurements that involve the conversion of other
energy forms into electrical energy are shown again in Table 1.2. Figure 1.4 shows the early ECG
machine where the interface between the body and the electrical monitoring equipment was
buckets filled with saline (“E” in Figure 1.4).
The biotransducer is often the most critical element in the system since it is the interface
between the subject or life process and the rest of the system. The transducer establishes the
risk or “noninvasiveness,” of the overall system. For example, an imaging system based on dif-
ferential absorption of x-rays, such as a computed tomography (CT) scanner, is considered more
invasive than an imaging system based on ultrasonic reflection since CT uses ionizing radiation
that may have an associated risk. (The actual risk of ionizing radiation is still an open ques-
tion and imaging systems based on x-ray absorption are considered minimally invasive.) Both
ultrasound and x-ray imaging would be considered less invasive than, for example, monitoring
internal cardiac pressures through cardiac catheterization, where a small catheter is threaded
into the heart chambers.
Many critical problems in medical diagnosis await the development of new approaches and
new transducers. Indeed, many of the outstanding problems in biomedical measurement, such as
noninvasive measurement of internal cardiac pressures or the noninvasive measurement of intra-
cranial pressure, await an appropriate, and undoubtedly clever, transducer mechanism.
5
Biosignal and Medical Image Processing
Figure 1.4 An early ECG machine using buckets filled with a saline solution as electrodes (marked
“E” in this figure).
1.4 Amplifier/Detector
The design of the first analog stage in a biomedical measurement system depends on the basic
transducer operation. If the transducer is based on a variation in electrical property, the first
stage must convert that variation into a variation in voltage. If there is only one sensing element
in the transducer, the transducer signal is said to be single ended and a constant current source
can be used as a detector (Figure 1.5). A typical transducer that would use this type of detector
circuit is a temperature-measurement device based on a thermistor, a resistor-like element that
changes its resistance with temperature. Thermistors have a well-defined relationship between
temperature and resistance. The detector equation for the circuit in Figure 1.5 follows Ohm’s law:
If the transducer can be configured differentially so that one element increases with increas-
ing input energy while the other element decreases, the bridge circuit is commonly used as
a detector. Figure 1.6 shows a device made to measure intestinal motility using strain gages.
Vout volts
Transducer
I amps element
R + ΔR
Figure 1.5 A constant current source is used in the detector circuit for a transducer based on an
energy-induced variation in resistance. One example is a temperature-measuring transducer based
on a thermistor, a semiconductor that changes resistance with temperature in a well-defined manner.
6
1.5 Analog Signal Processing and Filters
R
Cable to stimulator
Spring clamp Vout
Strain gages Vs
Electrodes
R
Support member
Figure 1.6 A strain gage probe used to measure motility of the intestine. The bridge circuit is used
to convert the differential chance in resistance from a pair of strain gages into a change in voltage.
Figure 1.7 (a) A cardiac microphone that uses a piezoelectric element to generate a small voltage
when deformed by sound pressure. (b) The detector circuit for this transducer is just a low-noise
amplifier.
A bridge circuit detector is used in conjunction with a pair of differentially configured strain
gages: when the intestine contracts, the end of the cantilever beam moves downward and the
upper strain gage (visible) is stretched and increases in resistance whereas the lower strain gage
(not visible) compresses and decreases in resistance. The output of the bridge circuit can be
found from simple circuit analysis to be Vout = VS R/2, where VS is the value of the source voltage.
If the transducer operates based on a change in inductance or capacitance, the above techniques
are still useful except a sinusoidal voltage source must be used.
If the transducer element is a voltage generator, the first stage is usually an amplifier. Figure 1.7a
shows a cardiac microphone used to monitor the sounds of the heart that is based on a piezoelectric
element. The piezoelectric element generates a small voltage when sound pressure deforms this
element; so, the detector stage is just a low-noise amplifier (Figure 1.7b). If the transducer pro
duces a current output, as is the case in many electromagnetic detectors, then a current-to-voltage
amplifier (also termed a transconductance amplifier) is used to produce a voltage output. In some
circumstances, additional amplification beyond the first stage may be required.
7
Biosignal and Medical Image Processing
analog signal processing restricts the frequency range or bandwidth of the signal using analog
filters. It is this filtering that usually sets the bandwidth of the overall measurement system.
Since signal bandwidth has a direct impact on the process of converting an analog signal into an
equivalent (or nearly equivalent) digital signal, it is often an essential element in any biomedi-
cal measurement system. Filters are defined by several properties: filter type, bandwidth, and
attenuation characteristics. The latter can be divided into initial and final characteristics. Each
of these properties is described and discussed in the next section.
The lowpass filter has a filter gain of 1.0 for the lower frequencies (Figure 1.8). This means
that the output equals the input at those frequencies. However, as frequency increases, the gain
drops, indicating that the output signal also drops for a given input signal. The highpass filter
has exactly the opposite behavior with respect to frequency (Figure 1.8). As frequency increases,
the gain and output signal increase so that at higher frequency, the gain is 1.0 and the output
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 1 2 3 0 1 2 3
Frequency (relative) Frequency (relative)
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 1 2 3 0 1 2 3
Frequency (relative) Frequency (relative)
Figure 1.8 Influence on signal frequency of the four basic filter types.
8
1.5 Analog Signal Processing and Filters
equals the input. The bandpass filter is a combination of these two filters; so, the gain and output
increase with frequency up to a certain frequency range where the gain is constant at 1.0, then
gain decreases with further increases of frequency (Figure 1.8). The spectrum of the band-stop
filter is the inverse of the bandpass filter (Figure 1.8).
Within each class, filters are also defined by the frequency ranges that they pass, termed
the filter bandwidth, and the sharpness with which they increase (or decrease) attenuation as
frequency varies. (Again, band-stop filters are an exception as their bandwidth is defined by
the frequencies they reject.) Spectral sharpness is specified in two ways: as an initial sharpness
in the region where attenuation first begins, and as a slope further along the attenuation curve.
These various filter properties are best described graphically in the form of a frequency plot
(sometimes referred to as a Bode plot), a plot of filter gain against frequency. Filter gain is simply
the ratio of the output voltage divided by the input voltage, Vout/Vin, often taken in dB. (The dB
operation is defined in Section 2.1.4, but is simply a scaled log operation.) Technically, this ratio
should be defined for all frequencies for which it is nonzero, but practically, it is usually stated
only for the frequency range of interest. To simplify the shape of the resultant curves, frequency
plots sometimes plot gain in dB against the log of frequency.* When the output/input ratio is
given analytically as a function of frequency, it is termed the transfer function. Hence, the fre-
quency plot of a filter’s output/input relationship can be viewed as a graphical representation of
its transfer function (Figure 1.8).
* When gain is plotted in dB, it is in logarithmic form, since the dB operation involves taking the log (see Section 2.1.4).
Plotting gain in dB against log frequency puts the two variables in similar metrics and results in more straight-line
plots.
† This defining point is not entirely arbitrary because when the signal is attenuated at 3 dB, its amplitude is 0.707
(10 −3/20) of what it was in the passband region and it has half the power of the unattenuated signal since 0.7072 = 1/2.
Accordingly, this point is also known as the half-power point.
9
Biosignal and Medical Image Processing
(a) (b)
1 Stopband 1 Stopband
Passband Passband
0.8 0.8
Gain
Gain
0.6 0.6
0.707
0.4 0.4 (–3 dB)
0.2 0.2
fc fc
0 0
100 102 104 100 102 104
Frequency (Hz) Frequency (Hz)
Gain
0.6 0.6
0.4 0.4
Stopband Stopband
0.2 0.2
fc fl fh
0 0 0 0
10 102 104 10 102 104
Frequency (Hz) Frequency (Hz)
Figure 1.9 Frequency plots of ideal and realistic filters. Each of the frequency plots shown here
has a linear vertical axis, but often, the vertical axis is plotted in dB. The horizontal axis is in log
frequency. (a) Ideal lowpass filter. (b) Realistic lowpass filter with a gentle attenuation characteris-
tic. (c) Realistic lowpass filter with a sharp attenuation characteristic. (d) Bandpass filter.
referred to as the characteristic equation because it defines the basic characteristics of the related
system.) As with any polynomial equation, the number of roots of this equation will depend on
the order of the equation; hence, filter complexity (i.e., the number of energy storage devices)
is equivalent to the number of roots in the denominator of the transfer function. In electrical
engineering, it has long been common to call the roots of the denominator equation poles. Thus,
the complexity of a filter is also equivalent to the number of poles in the transfer function. For
example, a second-order or two-pole filter has a transfer function with a second-order polyno-
mial in the denominator and would contain two independent energy storage elements (very
likely two capacitors).
Applying an asymptote analysis to the transfer function, it can be shown that for frequen-
cies much greater than the cutoff frequency, fc, the slope of most real-world filters is linear if
it is plotted on a log-versus-log plot. Figure 1.10 shows the frequency characteristics of the
transfer function of a first-order (single-pole) filter with a cutoff frequency, fc, of 5 Hz plotted
in both linear (Figure 1.10a) and dB versus log frequency (Figure 1.10b) format. Converting
the vertical axis to dB involves taking the log (see Section 2.1.4); so Figure 1.10b is a log–log
plot. At the cutoff frequency of 5 Hz, the frequency characteristic is curved, but at higher fre-
quencies, above 10 Hz, the curve straightens out to become a downward slope that decreases
20 dB for each order of magnitude, or decade, increase in frequency. For example, at 50 Hz,
the frequency characteristic has a value of −20 dB and at 500 Hz the value would be -40 dB
although this is not shown in the graph. Plotting dB versus log frequency leads to the unusual
units for the slope of dB/decade. Nonetheless, this type of plot is often used because it gener-
ates straight-line segments for the frequency characteristics of real-world filters and because
taking logs extends the range of values presented on both axes. Both linear and dB versus
10
1.5 Analog Signal Processing and Filters
(a) (b) 10
1 5
0
0.8
Gain (dB)
–5
Gain
0.6
–10
0.4
–15
0.2
–20
0 –25 –1
0 20 40 60 80 100 10 100 101 102
Frequency (Hz) Frequency (Hz)
Figure 1.10 Two representations of the gain characteristics (i.e., transfer function) for a first-order
filter. (a) A linear plot of gain against frequency. (b) The same curve is plotted with gain in dB, a
log function, against log frequency. The attenuation slope above the cutoff frequency becomes a
straight line with a slope of 20 dB/decade.
log frequency plotting is used in this book; the axes will describe which type of plot is being
presented.
The downward slope of 20 dB/decade seen for the first-order filter shown in Figure 1.10b gen-
eralizes, in that for each additional filter pole added (i.e., each increase in filter order), the slope
is increased by 20 dB/decade. (In a lowpass filter, the downward slope is sometimes referred
to as the filter’s roll-off.) Figure 1.11 shows the frequency plot of a second-order, two-pole filter
10
–5
–10
Gain (dB)
–15
–20
–25
–30
–40
101 102 103
Frequency (Hz)
Figure 1.11 A dB versus log frequency plot of a second-order (two-pole) and a 12th-order (12-pole)
lowpass filter having the same cutoff frequency. The higher-order filter more closely approaches the
sharpness of an ideal filter.
11
Biosignal and Medical Image Processing
10 10
Butterworth Chebyshev
0 0
–10 –10
–20 –20
Gain (dB)
Gain (dB)
–30 –30
–40 –40
–50 –50
–60 –60
101 102 103 101 102 103
Frequency (Hz) Frequency (Hz)
Figure 1.12 Two filters that have the same cutoff frequency (100 Hz) and the same order (four-
pole), but differing in the sharpness of the initial slope. The filter labeled Chebyshev has a steeper
initial slope, but contains ripples in the passband region.
with a slope of 40 dB/decade and a 12th-order lowpass filter. Both filters have the same cutoff
frequency, fc, hence the same bandwidth. The steeper slope or roll-off of the 12-pole filter is
apparent. In principle, a 12-pole lowpass filter would have a slope of 240 dB/decade (12 × 20 dB/
decade). In fact, this frequency characteristic is theoretical because in real analog filters, para-
sitic components and inaccuracies in the circuit elements limit the actual attenuation that can be
obtained. The same rationale applies to highpass filters, except that the frequency plot decreases
with decreasing frequency at a rate of 20 dB/decade for each highpass filter pole.
Example 1.1
An ECG signal of 1 V peak to peak has a bandwidth from 0.01 to 100 Hz. (Note that this fre-
quency range has been established by an official standard and is meant to be conservative.) It is
desired to reduce any noise in the signal by at least 80 dB for frequencies above 1000 Hz. What
is the order of analog filter required to achieve this goal?
Solution
Since the signal bandwidth must be at least 100 Hz, the filter’s cutoff frequency, fc, must be not
less than 100 Hz, but the filter must reduce the signal by 80 dB within 1 decade. Since typical
12
1.6 ADC Conversion
analog filters reduce the gain by 20 dB/decade for each pole, and an 80-dB reduction is desired,
the filter must have at least 80/20 = 4 poles.
Similar, but more involved, problems of this sort are explored in Chapter 4 that covers digital
filters.
1
fs = Hz (1.3)
Ts
To convert a continuous waveform into a digital format requires slicing the signal in two ways:
slicing in time and in slicing amplitude (Figure 1.13). So, the continuous signal x(t) becomes just
a series of numbers, x[1], x[2], x[3],. . .,x[n] that are the signal values at times 1Ts, 2Ts, 3Ts, and
nTs. In addition, if the waveform is longer than the computer memory, only a portion of the ana-
log waveform can be converted into digital format. Segmenting a waveform to fit in computer
memory is an operation termed windowing. The consequences of this operation are discussed
in Chapter 3. Note that if a waveform is sampled at Ts seconds for a total time, TT seconds, the
number of values stored in the computer will be
TT
N = points (1.4)
Ts
The relationship between a sample stored in the computer and the time at which it was sam-
pled is determined by its sample index, n, and the sample interval or sampling frequency:
n
t = nTs = (1.5)
fs
Each of these modifications, time slicing and amplitude slicing, has an impact on the result-
ing digital signal. Comparing the analog and digital signal in Figure 1.13 (left upper and lower
graphs), it is evident that the digitized signal is not the same as the original. The question is:
Does the difference matter, and if so, how does it matter? Intuitively, we might expect that if
the time and amplitude slices were very small, then “for all practical purposes” (that great
engineering expression), the digital and analog signals are essentially the same. The ques-
tion then becomes: how small must we make the time and amplitude slices? This question is
addressed separately for the two slicings in the following sections. Before approaching these
questions, we present a simple MATLAB example of the generation and display of a digitized
waveform.
* It is common to use brackets to represent discrete or digital functions whereas parentheses are used in continuous or
analog functions. See Section 1.3.1.
13
Biosignal and Medical Image Processing
Digitization grid
10 10
Continuous
8
9
6
x(t) 8
4
2 7
x(t) andx[n]
0 6
0 2 4 6 8 10
Time (s) 5
10
Discrete 4
8
3
6
x[n]
4 2
2 1
0 0
0 2 4 6 8 10 0 2 4 6 8 10
Sample number Sample number
Figure 1.13 Digitizing a continuous signal, upper left, requires slicing the signal both in time and
amplitude, right side. The result is a series of discrete numbers (squares) that approximate the
original signal. The resultant digitized signal, lower left, consists of a series of discrete numerical
values sampled at discrete intervals of time. In this example, x[n] = 2, 3, 2, 1, 0, 2, 3, 5, and 8.
Example 1.2
Generate a discrete 2-Hz sine wave using MATLAB. Assume a sample time of 0.01 s and use
enough points to make the sine wave 1-s long, that is, the total time, TT, should be 1 s. Plot the
result in seconds. Label the time axis correctly (as is always expected in MATLAB problems).
Solution
In this example, we are given sample intervals and the total time; so, the number of discrete
points that will be needed is fixed. From Equation 1.5, the number of points in the array will be
In many MATLAB problems, the number of points is arbitrary and we need to pick some
reasonable value for N. The best way to solve this problem in MATLAB is first to generate a time
vector* that is 1-s long with increments of Ts seconds: t = 0:Ts:1. We do not even have to find
the number of points explicitly, although it turns out to be 100. We can use this time vector in
conjunction with the MATLAB sin function to generate the sine wave.
% Example 1.2 Generate a discrete 2 Hz sine wave using MATLAB.
% Assume a sample time of 0.01 sec. and use enough points to make
% the sine wave 1 sec. long; i.e., the total time, TT should be 1 sec.
%
clear all; close all;
Ts = .01; % Define Ts = .01 sec
TT = 1; % Define total time = 1 sec
f = 2; % Define frequency = 2 Hz
t = 0:Ts:1; % Generate time vector
x = sin(2*pi*f*t); % Generate desired sine wave
plot(t,x,'.k'); % Plot sine wave as discrete points
xlabel('Time (sec)'); % and label
ylabel('x(t)');
* The reason a MATLAB sequence of numbers or an array is sometimes called a vector is discussed in Section 2.3.1.1.
14
1.6 ADC Conversion
1
0.8
0.6
0.4
x (t) 0.2
0
–0.2
–0.4
–0.6
–0.8
–1
0 0.2 0.4 0.6 0.8 1
Time (s)
Figure 1.14 Sequence of points generated by the code in Example 1.2. The sine wave pattern of
these points is evident. If these points were connected by straight lines, the graph would look very
much like a sine wave.
Results
The program produces the sequence of points shown in Figure 1.14. The sequence of points
clearly varies in a sinusoidal manner; if the points were connected by lines (as done by the plot
unless an alternative is requested), the plot would look just like a sine wave. This is examined in
Problem 1.5. Note how the MATLAB command that produces the sine wave looks the same as
an equivalent mathematical statement: x = sin (2πft).
VMAX
q= V (1.6)
2b − 1
where VMAX is the range of the ADC and b is the number of bits converted.
Example 1.3
The specifications (specs) of a 12-bit ADC advertise an accuracy of ± the least significant bit (LSB).
If the input range of the ADC is 0–10 V, what is the resolution of the ADC in analog voltage?
15
Biosignal and Medical Image Processing
Solution
If the input range is 10 V, then the analog voltage represented by the LSB can be found using
Equation 1.6:
VMAX 10 10
VLSB = = 12 = = 0.0024 V
(2 − 1)
bits
(2 − 1) 4095
6
x (t)
1
n (t)
–1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s)
Figure 1.15 The effect of quantization on the original signal can be viewed as noise added to the
signal. The amplitude of this noise depends on the quantization level, q, which in turn depends on
the number of bytes employed by the ADC. The variance of this noise, which is approximately the
same as the RMS value, is given in Equation 1.8.
16
1.6 ADC Conversion
σ 2 = e2 = ∫ e PDF(e) de
−∞
2
(1.7)
where PDF(e) is the uniform probability density function and e is the error voltage (the bottom
trace in Figure 1.13). The PDF(e) for a uniform distribution between ±q is simply 1/q. Substituting
into Equation 1.7:
(1/q) e 3
q /2 q /2
q2
σ =
2
∫ e (1/q)
− q /2
2
de =
3
− q /2
=
12
(1.8)
Example 1.4
Write a MATLAB program to evaluate Equation 1.8 through simulation. Generate a 4-Hz sine
wave in a 1000-point array (N = 1000). Assume a sample interval of Ts = 0.002. Quantize this
sine wave array into a 6-bit binary number using the routine quantization.m.* The calling
structure for this routine is
signal_out = quantization(signal_in,bits)
where signal_in is the original signal, bits is the number of bits for quantization (bits = 6
in this case), and signal _out is the quantized signal. Find the noise signal by subtracting
the original signal from the quantized signal. Plot this signal and take the variance to quantify
the noise. Then evaluate the theoretical noise using Equations 1.6 and 1.8 to find the theoretical
variance and compare the two results.
Solution
Since the number of points desired is given, it is easier to generate a 1000-point time vector and
then multiply it by Ts, that is, t = (0:999)*Ts. Then use that vector to generate the desired
4-Hz sine wave signal as in Example 1.2. (Note that almost any waveform and sample interval
will work as long as it includes a fair number of points: the sine wave is just a handy waveform.)
Quantize the signal into 6 bits using the routine quantization.m. Subtract the original sig-
nal from the quantized signal and take the variance of the result as the simulated noise variance.
Then evaluate Equation 1.6 with bits = 6 to find the quantization level q, and use that value of
q in conjunction with Equation 1.8 to find the theoretical value of noise variance. Compare the
two results. Use a formatted output to ensure that the results are displayed in an adequate num-
ber of decimal places.
% Example 1.4 Evaluate the quantization equation, Eq. 1.8 using simulated
data.
%
f = 4; % Desired frequency
N = 1000; % Number of points
Ts = 0.002; % Ts
bits = 6; % Quantization level
t = (0:N-1)*Ts; % Vector used to generate 1-cycle sine wave
signal_in = sin(2*pi*f*t); % Generate signal
signal_out = quantization(signal_in,bits); % Quantize signal
noise_signal = signal_out - signal_in; % Determine quantization error
* The MATLAB routine quantization.m, like all auxiliary routines, can be found in this book’s website. For clar-
ity, MATLAB variables, files, instructions, and routines are shown in courier typeface throughout the book.
17
Biosignal and Medical Image Processing
Results
The results produced by this program are shown in Table 1.3. The noise variance determined
empirically by the program is quite close to the theoretical value determined from Equations 1.6
and 1.8. This evaluation is extended to four different bit values in Problem 1.7.
It is relatively easy and common to convert between the analog and digital domains using
electronic circuits specially designed for this purpose. Many medical devices acquire the physi-
ological information as an analog signal and convert it to a digital format using an ADC for
subsequent computer processing. For example, the electrical activity produced by the heart can
be detected using properly placed electrodes and, after amplification of the resulting signal, the
ECG is an analog-encoded signal. This signal might undergo some preprocessing or conditioning
using analog electronics such as those described above. The signal is then converted into a digital
signal using an ADC for more complex, computer-based processing and storage. In fact, conver-
sion to digital format would usually be done even if the data are only to be stored for later use.
Conversion from the digital to the analog domain is also possible using a digital-to-analog con-
verter (DAC). Most personal computers (PCs) include both ADCs and DACs as part of a sound
card. This circuitry is specifically designed for the conversion of audio signals, but can be used for
other analog signals in some situations. Universal serial bus (USB)-compatible data-transforma-
tion devices designed as general-purpose ADCs and DACs are readily available; they offer greater
flexibility than sound cards with regard to sampling rates and conversion gains. These cards pro-
vide multichannel ADCs (usually 8–16 channels) and several channels of DAC. MATLAB has a
toolbox that will interface directly with either a PC sound card or a number of popular converters.
* A sinusoid has a straightforward frequency domain representation: it is defined only by a single magnitude and phase
(or a single complex point) at the frequency of the sinusoid. The classical methods of frequency analysis described in
Chapter 3 make use of this fact.
18
1.6 ADC Conversion
1
0.8
0.6
0.4
0.2
x (t)
0
–0.2
–0.4
–0.6
–0.8
–1
Figure 1.16 A sine wave (solid line) is sampled at two locations within one period. According to
the Shannon sampling theorem, there are no other sine waves that pass through these two points
at a lower frequency. The lowest-frequency sine wave is uniquely defined. However, there are an
infinite number of higher-frequency sine waves that could pass through these two points. Two of
these higher-frequency sine waves are shown: a sine wave at double the frequency (dotted line) and
one at triple the frequency of the original sine wave.
can be represented by a series of sinusoids (possibly an infinite series); so, if we can determine
the appropriate sampling frequency for a single sinusoid, we have also solved the more general
problem. The Shannon sampling theorem states that any sinusoidal waveform can be uniquely
reconstructed provided it is sampled at least twice in one period. (Equally spaced samples are
assumed.) That is, the sampling frequency, fs, must be >2 fsinusoid. In other words, only two equally
spaced samples are required to uniquely specify a sinusoid and these can be taken anywhere
over the cycle. Figure 1.16 shows a sine wave (solid line) defined by two points (“*”) covering
a time period slightly <1 cycle of the sine wave. According to the Shannon sampling theorem,
there are no other sinusoids of a lower frequency that can pass through these two points; hence,
these two points uniquely define a lowest-frequency sine wave. Unfortunately, there are a lot
of higher-frequency sine waves that are also defined by these two points, for example, all those
points at frequencies that are multiples of the original, two of which are shown in Figure 1.16. If
you continue to go up in frequency, there are, in theory, an infinite number of high-frequency
sine waves that can be defined by these two points.
Since an infinite number of other higher frequencies can be represented by the two sample
points in Figure 1.16, the sampling process can be thought of as actually generating all these
additional sinusoids. In other words, you cannot rule out the possibility that these two points
also represent all those other sine waves. A more comprehensive picture of the influence of sam-
pling on the original signal can be found by looking at the frequency characteristics of a hypo-
thetical signal before and after sampling. Figure 1.17 presents an example of a signal spectrum
before (Figure 1.17a) and after (Figure 1.17b) sampling. As discussed in Chapter 3, an individual
point in a spectrum represents a single sinusoidal* waveform; so, before sampling, the signal is a
mixture of seven sinusoids at frequencies of 1, 2, 3, 4, 5, 6, and 7 Hz. After sampling, the original
spectrum is still there, but in addition, it is now found in many other places in the spectrum,
specifically, on either side of the sampling frequency, fs. Additional sine waves are found reflected
about multiples of fs and added sine waves are even found at negative frequencies. Negative
* A sinusoidal waveform is a sine wave, or cosine wave, or a mixture of the two at the same frequency.
19
Biosignal and Medical Image Processing
1 1
0.8 0.8
0.6
0.6
|X (n)|
|X (f )|
0.4
0.4
0.2
0.2
0
–f fs
0 –0.2
0 1 2 3 4 5 6 7 8 9 10 –40 –30 –20 –10 0 10 20 30 40
Frequency (Hz) Frequency (Hz)
Figure 1.17 The effect of sampling viewed in the frequency domain. (a) An original spectrum
before sampling. (b) The spectrum after sampling. The original spectrum has been duplicated and
reflected around fs and at all multiples of fs including negative frequencies.
frequencies are generated by mathematics that describes the sampling operation. Although they
are mathematical constructs, they do have an influence on the sampled signal because they gen-
erate the added sine wave reflected to the left of fs (and to the left of multiples of fs). Moreover,
Figure 1.17b only shows a portion of the sampled signal’s spectrum because, theoretically, there
are an infinite number of higher frequencies and so the spectrum continues to infinity.
The comparison of frequency characteristics presented in Figure 1.17 clearly demonstrates
that the sampled signal is not the same as the original, but the critical question is: Can we pos-
sibly recover the original signal from its sampled version? This is an extremely important ques-
tion because if we cannot get an accurate digital version of an analog signal, then digital signal
processing is a lost cause. This crucial question is best answered by looking closely at the fre-
quency characteristics of a signal before and after sampling. If we can reconstruct the original
unsampled spectrum from the sampled spectrum, then the digital signal is an adequate repre-
sentation of the original.
Figure 1.18 shows a blowup of just one segment of the spectrum shown in Figure 1.17b, the
period between 0 and fs Hz. Comparing this spectrum to the spectrum of the original signal
(Figure 1.17a), we see that the two spectra are the same for the first half of the spectrum up to
fs/2 and the second half is just the mirror image (a manifestation of these theoretical negative
frequencies). It would appear that we could obtain a frequency spectrum that was identical to
the original if we somehow get rid of all frequencies above fs/2. In fact, we can get rid of the fre-
quencies above fs/2 by filtering as described previously. As long as we can get back to the original
spectrum, our sampled computer data are a useful reflection of the original data. The frequency
fs/2 is so important that it has its own name: the Nyquist frequency.*
This strategy of just ignoring all frequencies above the Nyquist frequency (fs/2) works well and
is the approach that is commonly adopted. But it can only be used if the original signal does not
have spectral components at or above fs/2. Consider a situation where four sinusoids with frequen-
cies of 100, 200, 300, and 400 Hz are sampled at a sampling frequency of 1000 Hz. The spectrum
produced after sampling actually contains eight frequencies (Figure 1.19a): the four original fre-
quencies plus the four mirror image frequencies reflected about fs/2 (500 Hz). As long as we know,
in advance, that the sampled signal does not contain any frequencies above the Nyquist frequency
* Nyquist was one of the most prominent engineers to hone his skills at the former Bell Laboratories during the first
half of the twentieth century. He was born in Sweden, but received his education in the United States.
20
1.6 ADC Conversion
0.8
|X (n)|
0.6
0.4
0.2
0
0 2 4 6 8 10 12 14 16 18 20
Frequency (Hz)
Figure 1.18 A portion of the spectrum of the sampled signal is shown in Figure 1.17b. Note that
the added frequencies above fs /2 (10 Hz in this example) are distinct from, and do not overlap, the
original frequencies. If we were to eliminate those frequencies with a lowpass filter, we would have
the original spectrum back.
(500 Hz), we will not have a problem: we know that the first four frequencies are those of the sig-
nal and the second four frequencies, above the Nyquist frequency, are the reflections that can be
ignored. However, a problem occurs if the signal contains frequencies higher than the Nyquist
frequency. The reflections of these high-frequency components will be reflected back into the lower
half of the spectrum. This is shown in Figure 1.19b, where the signal now contains two additional
(a) 5 (b) 5
New original
4 Original Added 4 frequencies
?
frequencies frequencies
3 3
X (f )
X (f )
2 2
1 1
0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Frequency (Hz) Frequency (Hz)
Figure 1.19 (a) Four sine waves between 100 and 400 Hz are sampled at 1 kHz. Only one period
of the sampled spectrum is shown, the period between 0 and fs Hz. Sampling essentially produces
new frequencies that are not in the original signal. Because of the periodicity and even symmetry
of the sampled spectrum, the additional frequencies are a mirror image reflection around fs /2, the
Nyquist frequency. If the frequency components of the sampled signal are all below the Nyquist fre-
quency as shown here, then the upper frequencies do not interfere with the lower spectrum and can
be filtered out or simply ignored. (b) If the sampled signal contains frequencies above the Nyquist
frequency, they are reflected into the lower half of the spectrum (circles with “*”). It is no longer
possible to determine which frequencies belong where, an example of aliasing.
21
Biosignal and Medical Image Processing
frequencies at 650 and 850 Hz. These frequency components have their reflections in the lower half
of the spectrum at 350 and 150 Hz, respectively. Now, it is no longer possible to determine if the
350- and 150-Hz signals are part of the true spectrum of the signal (i.e., the spectrum of the sig-
nal before it was sampled) or whether these are reflections of signals with frequency components
greater than fs/2 (which, in fact, they are). Both halves of the spectrum now contain mixtures of
frequencies above and below the Nyquist frequency, and it is impossible to know where they really
belong to. This confusing condition is known as aliasing. The only way to resolve this ambiguity is
to ensure that all frequencies in the original signal are less than the Nyquist frequency.
If the original signal contains frequencies above the Nyquist frequency, then the digital signal
in the computer is hopelessly corrupted. Fortunately, the converse is also true. If there are no
corrupting frequency components in the original signal (i.e., the signal contains no frequencies
above half the sampling frequency), the spectrum in the computer can be altered by filtering to
match the original signal spectrum. This leads to a common representation of the famous sam-
pling theorem of Shannon: the original signal can be recovered from a sampled signal provided
the sampling frequency is more than twice the maximum frequency contained in the original:
In practical situations, fmax is taken as the frequency above which negligible energy exists in
the analog waveform. The sampling frequency is generally under software control and it is up
to the biomedical engineer doing the data acquisition to ensure that fs is high enough. To make
elimination of the unwanted higher frequencies easier, it is common to sample at three to five
times fmax. This increases the spacing between the frequencies in the original signal and those
generated by the sampling process (Figure 1.20). The temptation to set fs higher than really nec-
essary is strong, and it is a strategy often pursued. However, excessive sampling frequencies lead
to larger data storage and signal-processing requirements that could unnecessarily burden the
computer system.
(a) 5 (b) 5
fs = 500 Hz fs = 1000 Hz
4 4
3 3
X (f )
X (f )
2 2
1 1
0 0
0 200 400 600 800 1000 0 500 1000 1500 2000
Frequency (Hz) Frequency (Hz)
Figure 1.20 The same signal sampled at two different sampling frequencies. (a) A spectrum of a
signal consisting of four sinusoids sampled at 500 Hz. (b) The same signal sampled at 1000 Hz.
The higher sampling frequency provides greater separation between the original signal spectrum
and the spectral components added by sampling.
22
1.6 ADC Conversion
(a)
Zero padding
20
10
0
–5 0 5 10 15 20 25
n
(b)
Periodic extension
20
10
0
–5 0 5 10 15 20 25
n
(c)
Symmetric extension
20
10
0
–5 0 5 10 15 20 25
n
Figure 1.21 Three strategies for extending the length of data: (a) zero padding: zeros are added at
the ends of the data set, (b) periodic or wraparound: the waveform is assumed to be periodic; so,
the end points are added at the beginning and beginning points are added at the end, and (c) sym-
metric: points are added at the ends in reverse order. With this last strategy, the edge points may,
or may not, be repeated at the beginning and end of the data set.
when an end point is encountered. There are three common strategies for dealing with situations
where you come to the end of the data set yet the algorithm needs additional samples: extending
with zeros (or a constant) termed zero padding, using periodicity or wraparound, and extending by
reflection also known as symmetric extension. These options are illustrated in Figure 1.21.
In the zero-padding approach, zeros are added at the end or beginning of the data sequence
(Figure 1.21a). This approach is frequently used in spectral analysis and is justified by the
implicit assumption that the waveform is zero outside the sample period anyway. A variant of
zero padding is constant padding where the data sequence is extended using a constant value,
often the last (or first) value in the sequence. If the waveform can be reasonably thought of as
1 cycle of a periodic function, then the wraparound approach is justified (Figure 1.21b). Here,
the data are extended by tacking on the initial data sequence at the end of the data set and vice
versa. These two approaches will, in general, produce a discontinuity at the beginning or end of
the data set; this can lead to an artifact for some algorithms. The symmetric-reflection approach
eliminates this discontinuity by tacking on the end points in reverse order (or beginning points
if the extending is done at the beginning of the data sequence) (Figure 1.21c).*
To reduce the number of points in cases in which an operation has generated additional data,
two strategies are common: simply eliminate the additional points at the end of the data set, or
eliminate data from both ends of the data set, usually symmetrically. The latter is used when the
* When using this extension, there is a question as to whether or not to repeat the last point in the extension: either
strategy will produce a smooth extension. The answer to this question will depend on the type of operation being
performed and the number of data points involved; determining the best approach may require empirical evaluation.
23
Biosignal and Medical Image Processing
data are considered periodic, and it is desired to retain the same timing or when other similar
concerns are involved. An example of this strategy is found in periodic convolution often used
in wavelet analysis as described in Chapter 7.
1.8 Summary
From a traditional reductionist viewpoint, living things are described in terms of component
systems. Many traditional physiological systems such as the cardiovascular, endocrine, and
nervous systems are quite extensive and are composed of many smaller subsystems. Biosignals
provide communication between systems and subsystems, and are our primary source of infor-
mation on the behavior of these systems. Interpretation and transformation of signals are a
major focus of this chapter.
24
1.8 Summary
Biosignals, like all signals, must be “carried” by some form of energy. The common biologi-
cal energy sources include chemical, mechanical, and electrical energy (in the body, electrical
signals are carried by ions, not electrons). In many diagnostic approaches, an external energy
source is used to probe the body, as in ultrasound and CT scanners. Signals can be encoded in a
number of formats, but most biosignals are generally analog in nature: they vary continuously
over time and the desired information is embedded in the amplitude or instantaneous value of
the signal.
Measuring biosignals involves complex multicomponent systems, but the lead element is
the biotransducer, which provides the interface between the living system and the measure-
ment system. It converts biological energy (or external energy in some systems) into an elec-
tric signal compatible with analog and digital processing. The biotransducer usually sets the
noise level of the measurement and often limits the ultimate utility of the system. For these
reasons, the biotransducer can be the most critical component of a measurement system.
Since most signal processing is performed in the digital domain, measurement systems usu-
ally include an ADC to transform the continuously time-varying biosignals into a sequence
of numbers suitable for digital processing. Analog signal processing is common before con-
version, including amplification (boosting the signal values) and filtering. The latter is used
to reduce noise and, as shown in Chapter 3, to ensure accurate conversion to digital format.
Filters vary in function, but most filters can be fully defined by their frequency characteristics
or spectrum. Specifically, three frequency attributes describe the important features of a fil-
ter: the filter type (lowpass, highpass, bandpass, and bandstop), attenuation slope, and initial
sharpness. The frequency characteristics of filters (and other analog system components) are
often displayed as a plot in dB versus log frequency since these log–log-type plots are often
easier to interpret.
The conversion of analog signals to digital (or discrete time) signals requires slicing the con-
tinuous signal into discrete levels of both amplitude and time. Amplitude slicing adds noise to
the signal; the amount of noise added is dependent on the size of the slice. Typically, the slices
are so small that the noise from amplitude slicing is usually ignored. Not so for time slicing,
which produces such a complicated transformation that it adds additional frequencies to the
digitized version of the signal. While the digitized signal is quite different in spectral content
from the original analog signal, the added frequencies will all be above those in the original,
provided the sampling frequency is greater than twice the maximum frequency in the original
signal. This rule is known as Shannon’s sampling theorem.
In addition to time and amplitude slicing, many real-world signals must be truncated if they
are to be stored in computer memory. The details of truncation or data windowing are discussed
in Chapter 3, but one consequence of finite data is often the need for a strategy for dealing with
the end points. Three popular methods exist: zero padding (or constant padding) that adds zeros
at the end, periodic extension which wraps data around using beginning points at the end and
ending points at the beginning, and symmetric extension that reflects the last few points around
the end point in a symmetric manner.
The multitudinous advantages of the World Wide Web extend to signal processers as
well; many resources including algorithms and data banks are available and continually
expanding.
Problems
1.1 A lowpass filter is desired with a cutoff frequency of 10 Hz. This filter should attenu-
ate a 100-Hz signal by a factor of at least 78. What should be the order of this filter?
1.2 Since sine waves have energy at only one frequency (see Chapter 3), they can be used
to probe the frequency characteristics of a filter (or other linear systems). Here, sine
waves at several different frequencies are used as input to an analog filter. The input
25
Biosignal and Medical Image Processing
sine waves have a value of 1.0 V root mean square (RMS) and the output measured
at each frequency is given below:
Use MATLAB to plot the filter’s frequency characteristic. (As in all plots, label
the axes correctly.) What type of filter is it? What is the cutoff frequency and order
of this filter? Use the MATLAB grid function (grid on;) to help find the slope and
cutoff frequency from the plot. [Hint: Since sine waves have energy at only one fre-
quency, the frequency characteristics of the filter can be determined from the ratio
of output to input (Equation 1.2). Take 20 log(Vout/Vin) where Vin = 1.0 to get the
gain in dB and use MATLAB’s semilogx to get a plot gain against log frequency.
This will allow you to estimate the spectral slope in dB/decade better.]
1.3. The MATLAB routine analog _ filter1.m found in the support material
associated with this chapter is a simulated analog filter. (Calling structure: out-
put = analog _ filter1(input);). Repeat the strategy used in Problem 1.2,
but generate the input sine waves in MATLAB and plot the output. Make Ts = 0.001 s
and the number of points N = 1000. To generate sine waves, define the time vector
t = (0:999)*Ts as in Example 1.3. Use this vector to generate the input sine waves
at various frequencies: input = sin(2*pi*f*t);. Use frequencies of 2, 10, 15,
20, 30, 40, 50, 60, 80, 90, 100, 150, 200, 300, and 400 Hz.
Find the amplitude of the output of analog _ filter1.m using MATLAB’s
max operator and plot the resulting values in dB versus log frequency. This gives an
approximation of the filter’s frequency characteristics or spectrum. Use this spec-
trum with the grid function enabled to determine the type, bandwidth, and attenu-
ation slope of the filter. [Hint: Put the desired input frequencies in an array and use
a for-loop to generate the input sine waves. Store the maximum values of the filter’s
output in an array for plotting. Plot the 20 log of the output values against the fre-
quency array using the semilogx routine.]
1.4 Repeat Problem 1.3 but use the routine analog _ filter2.m for the filter (same
calling structure). In this problem, you are to select the frequencies. Use the mini-
mum number of frequencies that allow you to accurately measure the filter’s cutoff
frequency and slope. Limit frequencies to be between 2 and 400 Hz.
1.5 Generate a discrete 2-Hz sine wave using MATLAB as in Example 1.2. Use sample
intervals of 0.05, 0.01, and 0.001 s and again use enough points to make the sine
wave 1-s long, that is, the total time, TT, should be 1 s. Plot with lines connecting
the points and on the same plot, the individual points superimposed on the curves
as in Example 1.2. Plot the sine waves using the three sample intervals separately,
but use subplot to keep the three plots together. Note that even the plot with very
few points looks sinusoidal. Chapter 3 discusses the minimum number of points
required to accurately represent a signal.
1.6 Write a MATLAB problem to test Equation 1.6 through simulation. Generate a 4-Hz,
1000-point sine wave as in Example 1.4 assuming a sample interval of Ts = 0.002.
Use quantization.m to digitize it using 4-, 8-, 12-, and 16-bit ADC. Then, as in
Example 1.4, subtract the original signal from the quantized signal to find the error
signal. The amplitude of the error signal should be equal to the quantization level, q
26
1.8 Summary
in Equation 1.6. Use MATLAB’s max and min functions to find this amplitude and
compare it with the value predicted by Equation 1.6. Put this code in a for-loop and
repeat the evaluations for the four different bit levels requested. Be sure to display
the results to at least four decimal places to make an accurate comparison. [Hint:
Most of the code needed for this problem will be similar to that in Example 1.4.]
1.7 Extend the code in Example 1.4 to include quantization levels involving 4, 8, 10, and
12 bits. Compare the theoretical and simulated noise for these four different quan-
tization levels. Present the output in a format with sufficient resolution to illustrate
the small differences between the simulated and theoretical noise.
1.8 Write a MATLAB program to generate 1 s of a 5-Hz sine wave in a 1000-point array.
(Use Equation 1.4 to determine the equivalent Ts.) Plot the sine wave. Simulate the
sampling of this waveform at 7 Hz by plotting a point (such as an “*”) at intervals
of Ts = 1/fs = 1/7 s. (Use a for-loop to generate the sample time, Ts, insert it into the
equation for the 5-Hz sine wave, and plot the resulting point.) Aliasing predicts that
these sampled points should fall on a 2-Hz (7–5 Hz) sine wave. Can you find a 2-Hz
sine wave that includes all seven points? [Hint: The sine wave could be phase shifted
by 180°.]
1.9 Repeat Problem 1.8 using a simulated sample interval of 9 Hz (Ts = 1/fs = 1/9 s) and
find the appropriate sine wave produced by aliasing.
27
Exploring the Variety of Random
Documents with Different Content
Behalve de Echte Bladwespen en de van haar arbeid profiteerende, doch zelf geen gallen
voortbrengende B a s t a a r d g a l w e s p e n (Cynipidae inquilinae), waarvan hierboven
eenige (Synergus, Aulax) genoemd zijn, die als commensalen zich met de bestanddeelen
der gallen voeden, kent men ook een aantal P a r a s i e t g a l w e s p e n (Cynipidae
parasitae). Deze gelijken door haar lichaamsbouw zoozeer op de overige leden der
familie, dat zij hierin een plaats verdienen, maar hebben aan de vorming van gallen geen
schuld; zij ontwikkelen zich geheel op dezelfde wijze als de Sluipwespen in het lichaam
van andere Insecten. Zoo leven b.v. de kleine leden van het geslacht Allotria (ongeveer
40 soorten) in Bladluizen. Door het korte, bijna ronde, zittende of nagenoeg ongesteelde
achterlijf, welks grootste deel door het eerste segment wordt gevormd, en door het
maaksel der vleugels komen zij geheel met de Echte Galwespen overeen, van welke zij
echter verschillen door de gladheid van haar als ’t ware gepolijste huid en de dunheid der
sprieten, die het lichaam meestal in lengte overtreffen. Door een meer langwerpige
gedaante en een korten, ringvormigen achterlijfssteel kenmerkt zich het geslacht Figites,
welks leden, naar het schijnt, alle in larven van Vliegen hunne beide eerste
levenstijdperken doorbrengen. Nog slanker zijn de Ibalia’s. Het grootste lid van de
geheele familie is de 14 mM. lange Ibalia cultellator, die op een Sluipwesp gelijkt, in
larven van Houtwespen (Sirex) hare eieren legt en een als het lemmet van een scheermes
samengedrukt achterlijf heeft.
Een niet gering aantal kleine, parasitisch levende Wespen worden onder den naam van
P r o c t o t r u p i d e n tot een familie vereenigd, die in sommige opzichten aan de vorige,
in andere aan de volgende herinnert en dus tusschen beide een overgang vormt; het is
daarom moeielijk deze groep als zoodanig te kenschetsen. Vele soorten naderen door de
gedaante van het vleugeladerstelsel tot de Galwespen, van welke zij echter verschillen
door het bezit van de randstip en door den algemeenen vorm van het lichaam. Ook zijn
er soorten, die (zooals de hierna afgebeelde) met de Chalcididen de gebroken sprieten en
het volkomen gemis van cellen in den vleugel en van alle aders (behalve de
onderrandader) gemeen hebben, hoewel zij niet, zooals deze, den legboor aan de
buikzijde van het achterlijf, maar aan de spits naar buiten laten treden. De meeste
Proctotrupiden zijn zeer klein van gestalte en zwart van kleur, langwerpig van vorm en
toch niet slank gebouwd, langzaam en log van beweging, maar toch niet traag van aard.
Een soortgelijk verschil als tusschen de plompe, met buitengewone volharding bezielde
Hommels en de hartstochtelijke, ongedurige, zeer bewegelijke Zandbijen merkt men ook
op tusschen de Proctotrupiden en de Chalcididen. Gene, niet in staat om een naderenden
vijand reeds op eenigen afstand waar te nemen, geven zich geen moeite om hem door
een snelle vlucht te ontkomen, vestigen zich bij voorkeur op vochtige plaatsen, onder
afgevallen bladen of in de onmiddellijke nabijheid van den grond in dichte omheiningen.
De Chalcididen daarentegen zijn onophoudelijk in beweging, laten de sprieten nooit
rusten; de wijfjes houden zich voortdurend bezig met het opsporen van geschikte
legplaatsen; alle houden gaarne verblijf in een zonnige omgeving en gaan tusschen
g r o e n e bladen zitten, als zij schaduw verlangen; eerst door de nadering van den
winter worden zij van hier verdreven en genoodzaakt om haar teer lichaam tegen de
guurheid van het klimaat te beschutten door te midden van het rottende afval op den
bodem een schuilplaats te zoeken.
Het wijfje van Torymus regius zagen wij op een galnoot bezig met haar legboor een ei te
plaatsen in de hier levende Galwesp-larve; de made, die uit dit ei ontstaat, voedt zich met
de sappen van den galbewoner en brengt diens dood teweeg, zoodra zij hem niet meer
noodig heeft. Uit een kleinere opening dan de rechtmatige eigenares van de woning
geboord zou hebben, indien zij in ’t leven was gebleven, ontsnapt ten slotte de goudgele,
op den rug blauw iriseerende, met roodachtig gele pootjes uitgeruste parasiet.
Op oude leemmuren, die den verzamelaar van Vliesvleugeligen steeds een rijken buit
beloven, ziet men in den zomer, te midden van een groot aantal andere bewoners van
zulke oorden, een slank diertje zich op zulk een zonderlinge wijze bewegen, dat het den
opmerkzamen toeschouwer onmogelijk ontgaan kan. Als ’t ware dreigend wordt het
achterlijf opgeheven, dat knotsvormig is, evenals de ver uiteenwijkende achterscheenen;
zoo dartelt het Insect, flauw gekromde booglijnen volgend, dicht bij den muur langs en
schijnt onvermoeibaar te zijn, daar men het slechts zeer zelden met opgeheven vleugels
op hooge pooten eenige schreden ziet doen. Dit is de 10 mM. lange J i c h t w e s p
(Foenus affectator), een parasiet van de Vliesvleugeligen, die in den muur wonen. Haar
borst en haar achterlijf zijn zijdelings samengedrukt; de kleur is zwart, met roode
vlekken op het achterlijf en aan de kniestreek van de achterpooten; de lengte van den
legboor is ongeveer gelijk aan het vierde gedeelte van die van ’t achterlijf.—Een tweede,
zeldzamere soort, de 14 mM. lange P ij l d r a g e r (Foenus jaculator), kenmerkt zich
door de witte kleur van het wortelgedeelte van den scheen en den voet, althans van die
der achterpooten; het achterlijf is in het midden rood en wordt door den legboor ver
overtroffen in lengte.
Naar den bouw der monddeelen heeft men deze soortenrijke familie in drie groepen
verdeeld. Bij de G e s l o t e n m o n d i g e n (Clidostomi) is het kopschild afgerond,
toegespitst of ondiep uitgesneden en kruisen de beide bovenkaken elkander, zoodat de
mondopening er door bedekt wordt, of zich hoogstens als een smalle spleet vertoont.—
Bij de K r i n g m o n d i g e n (Cyclostomi) blijft tusschen het van onderen diep
uitgesneden kopschild (met naar binnen teruggeslagen bovenlip) en de bovenkaken (die
zoo kort zijn, dat alleen de spitsen elkander aanraken) een half-kringvormige ruimte
over.—Bij de B u i t e n t a n d i g e n (Exodontes) zijn de bovenkaken zoo kort, dat zij
elkander in ’t geheel niet raken, en bovendien als ’t ware verdraaid, daar beide haar
gewelfde zijde naar den onderrand van den kop, de holle zijde naar buiten richten. Deze
zoo slecht gewapende Braconiden leggen eieren in larven van Vliegen en Kevers.
Aan de boogvormig naar onderen gerichte sprieten, het lancetvormige achterlijf met
duidelijken steel, waarvan het 2e en het 3e lid onderling n i e t vergroeid zijn, herkent
men de kleine, hoogstens 2.4 mM. lange B l a d l u i s d o o d e r s (Aphidius), die, evenals
de vroeger genoemde Parasiet-galwespen van het geslacht Allotria, alle in Bladluizen
leven en daarom het best door kweeking verkregen kunnen worden. De Bladluis, die een
Aphidius-larve bevat, bezwijkt, zoodra deze tot rijpheid is gekomen en zit dan met
zijwaarts gerichte pooten en met een bolvormig gezwollen, metaalachtig glinsterend
achterlijf te midden van hare gave, ongevleugelde zusters. Een gat in haar lichaam, niet
grooter dan een speldeprik, levert het bewijs, dat de Wesp de ledige huid van haar
gastheer verlaten heeft.
Als voorbeeld noemen wij Bracon palpebrator, waarvan wij een groot aantal
exemplaren verkregen uit dennenstammetjes, die sterk aangetast waren door de Rood-
en-wit-bonte Dennensnuittor. De borst is op den rug glad en glanzig; zwart is de
hoofdkleur; rood zijn de pooten, met uitzondering van de achterste, en het achterlijf, met
uitzondering van een zwarte vlek op den eersten ring.
Spathius clavatus is een medebewoner van ons huis, voor zoover hier bepaalde soorten
van Kevers voorkomen. Zijn larve parasiteert n.l. bij de Klopkevers, die in oud
houtwerk, o.a. in oude meubels gangen boren, vooral bij Anobium striatum en misschien
ook bij het Bonttorretje. Doe hem dus geen kwaad, wanneer hij zich tusschen Juni en
Augustus op de vensterruiten vertoont!—Zijn grootte wisselt af tusschen 4½ en 8¾ mM.
De kleur is grootendeels bruinachtig rood; de tasters, de heupen en een breede ring aan
den wortel van den scheen zijn bruinachtig wit, de vleugels donker gevlekt. De legboor
is langer dan het lichaam.
Als de kern van de familie, de vereeniging van hare edelste vertegenwoordigers, kan
men de I c h n e u m o n e n (Ichneumones) beschouwen. De steel, die het van boven naar
beneden platgedrukte, lancetvormige achterlijf met het borststuk verbindt, is op zulk een
wijze gebogen, dat het achtereinde van den eersten ring en alle volgende leden van het
achterlijf hooger geplaatst zijn dan de wortel van den steel. De ademgaten van het eerste
segment zijn a c h t e r het midden van den steel en n i e t nader bij elkander dan bij het
achterste uiteinde van den ring gelegen. De legboor kan meestal volkomen in het
lichaam verborgen worden. De spiegelcel is vijfhoekig. De sprieten zijn bij het mannetje
altijd borstelvormig, bij het wijfje soms draadvormig. Tot de Ichneumonen behooren de
bontst gekleurde Sluipwespen; de wijfjes tooien zich met rood, zwart en wit (of geel); in
geen familie vindt men even zuivere kleuren als in deze en een grooter verschil van kleur
tusschen de mannetjes en wijfjes. De larven, voor zoover bekend, onderscheiden zich
door een zekere fletschheid; zij spinnen geen cocon, daar groote vlinderpoppen haar tot
woonplaats dienen. Door deze te bewaren, kan men de Wesp verkrijgen, die zich een
weg baant naar buiten door het bovenste deel van de pophuid weg te knagen. Daarom
legt het wijfje in iedere r u p s slechts é é n e i .
De C r y p t i d e n (Cryptidae) kunnen den legboor niet geheel terugtrekken; een deel
van dit orgaan treedt nog in den rusttoestand door een spleet aan de buikzijde van ’t
achterlijf naar buiten; de ademgaten van het eerste achterlijfssegment zijn dichter bij
elkander dan bij het uiteinde van dit segment gelegen. De eenige Echte Sluipwespen
zonder of met rudimentaire vleugels vormen het geslacht Pezomachus; voor ’t
meerendeel zijn zij zeer klein en parasiteeren bij andere Sluipwespen.
Een der meest gewone soorten van Tryphoniden is de 11 mM. lange Tryphon (Exenterus)
marginatorius, kenbaar aan den gelen achterrand der achterlijfsringen, aan de
veranderlijke gele teekening op den kop en het borststuk (welker overigens zwarte
oppervlakte door rimpeltjes oneffen is) en aan het volkomen o n t b r e k e n van een
e i n d d o o r n aan den gelen, van onderen zwarten achterscheen. Deze Sluipwesp vliegt
bij voorkeur in dennebosschen rond, en legt op de bastaardrups van de Gewone
Dennenbladwesp (Lophyrus pini) een ei, dat aan de huid door tusschenkomst van een
haakje bevestigd wordt. Het slachtoffer, dat zich tevergeefs door krachtige bewegingen
tegen den vijand heeft verzet, spint weldra een tonvormige cocon om hierin te
overwinteren, alsof er niets gebeurd is. De parasitische made, die nu het ei verlaat, blijft
buiten op de rups zitten en zuigt haar volkomen uit; zij laat van haar gastheer niets
anders over dan de verschrompelde huid, die een klein hoekje beslaat van de cocon,
waarbinnen de indringer zelf een spinsel vervaardigt, dat slechts de helft van de
beschikbare ruimte vult.
Het geslacht Bassus onderscheidt zich door den nagenoeg vierzijdigen vorm van het
segment, waardoor het korte, eivormige achterlijf aan het borststuk bevestigd is.
Bassus albosignatus heeft geen lange ontdekkingsreis te doen voor het opsporen van een
geschikte legplaats voor hare eieren, maar vindt deze bij het azen op het zoete vocht, dat
de Bladluizen uitscheiden, in de Bladluizen-etende, op kleine Bloedzuigers gelijkende
maden van Zweefvliegen (Syrphus). De made, die zulk een ei met zich omdraagt, schijnt
zich hierover niet veel te bekommeren; zij blijft vreten, totdat zij groot genoeg geworden
is om zich in te spinnen. Uit haar peervormige cocon, die in overlangsche richting aan
een denne- of sparrenaald, aan een breed blad, of aan een ander plantendeel is
vastgehecht, komt echter geen Vlieg, maar een Wesp te voorschijn. Deze is 5.2 à 8.6
mM. lang en op zwarten grond met vele witte vlekken en ringen geteekend.
Een andere soort heeft ongeveer dezelfde levenswijze, daar zij eieren legt in larven van
Lievenheersbeestjes (Coccinella), die zich eveneens met Bladluizen voeden.
Een groot aantal leden van dezelfde onderfamilie komen door houding en kleur met de
genoemde soorten zoo zeer overeen, dat een ongeoefende allicht geneigd zal zijn ze voor
gelijksoortig te houden; overal ontmoet men leemkleurige Sikkelwespen, die met
bovenwaarts gerichte vleugels op struiken, heggen en bloemen rondwandelen of
langzaam, als ’t ware tuimelend, soms met hoorbaar gonzenden vleugelslag, wegvliegen
om in de onmiddellijke nabijheid op eenigszins plompe wijze weder neer te strijken en
hier te zoeken naar hetgeen zij ginds niet vonden. Hoewel deze Insecten in vorm met
Anomalon overeenstemmen, blijkt het bij nader onderzoek, dat zij tot andere geslachten
behooren en wel vooral tot Ophion (fign. 7 en 8) en Paniscus, die o.a. merkwaardig zijn,
doordat hare eieren een steel bezitten. Dikwijls ziet men één van deze eieren of
verscheidene, die trosvormig dicht opeengedrongen zijn, aan de spits van het achterlijf
van de vrouwelijke Wesp hangen; door haar vorm en glanzig zwarte kleur gelijken zij
wel eenigszins op zaden van sommige planten, o.a. op die van den bekenden kattestaart
(Amarantus).
In 1829 gaf Gravenhorst een beschrijving van 274 soorten van het geslacht Ichneumon,
die in Europa, en voor ’t meerendeel in Duitschland, gevonden worden; hierbij zijn er
echter niet weinige waarvan men alleen het wijfje of alleen het mannetje kent. De
grootste en schitterendst gekleurde soorten van Sluipwespen behooren tot deze groep:
rood, geel, wit en zwart prijken op haar kleed. Door samenvoeging van deze weinige
kleuren ontstaat de groote verscheidenheid van teekening, die men bij de Ichneumonen
opmerkt; in den regel is het wijfje bonter uitgedost dan het mannetje; dit verschil maakt
het soms zeer moeielijk te ontdekken, dat zij tot één soort behooren. De wijfjes kunnen
als zoodanig gemakkelijk herkend worden aan hare min of meer knobbelige, na den
dood altijd eenigszins gekronkelde, draad- of borstelvormige sprieten, slechts in enkele
gevallen aan de nauwelijks merkbare scheede van den legboor. Met uitzondering van
eenige onder mos of in vermolmd hout overwinterende soorten, die zich reeds in ’t
voorjaar vertoonen, krijgt men de Ichneumonen niet vóór Juni te zien. Met plat op den
rug gelegde vleugels loopen zij dan snuffelend rond op bladen van struiken, waar men ze
soms in grooten getale bijeenvindt, als de Bladluizen er haar zoet uitscheidingsproduct
hebben achtergelaten, of als hier de woonplaats is van rupsen, waarin zich de
Sluipwespmaden ontwikkelen. Een klepperend en rammelend gedruisch kan men hooren
op zulke vergaderplaatsen van allerlei Ichneumonen en andere Sluipwespen, waar men
gewoonlijk ook Graafwespen en niet minder veelvuldig Vliegen en dergelijk gespuis ziet
verschijnen. Een merkwaardig schouwspel verschaft dit bonte mengelmoes van
lekkerbekken en roovers, zoo verschillend van aard en beweging, waarvan sommige
licht en vlug, andere plomp en log, deze schroomvallig, gene onvervaard hun deel
trachten te verkrijgen van den hier aanwezigen buit.
De afbeelding van Ichneumon pisorius (fig. 1), een van de grootste soorten, kan een
voorstelling geven van het voorkomen van het geheele geslacht, terwijl men uit de
daaronder liggende, van haar kruin beroofde, ledige pophuid van den Dennenpijlstaart
kan zien, op welke wijze deze Insecten hun geboorteplaats verlaten. Tot nadere
aanduiding van de genoemde soort kan dienen, dat de achterlijfssteel niet breeder is dan
hoog; het achterlijf, dat uit 7 leden bestaat, loopt bij het wijfje spits toe; de buikplaat van
het laatste segment is op eenigen afstand van de aanhechtingsplaats van den legboor
gelegen; de ademgaten van het achterborststuk zijn langwerpig spleetvormig; de rugzijde
van kop en borststuk hebben geen bijzondere eigenaardigheden. Bij het wijfje zijn het
schildje en een streepje aan iederen vleugelwortel geel; het met putjes bezette en
hierdoor doffe achterlijf is, met uitzondering van het bruine steellid, bleek roestrood. Bij
het mannetje zijn het geheele aangezicht en het grootste deel van de pooten geel, bij het
wijfje alleen de bovenrand van het oog en het midden van de scheen; bovendien hebben
hare sprieten een witten ring. De niet genoemde lichaamsdeelen zijn zwart. Yan Juni af
zwerft Ichneumon pisorius in gemengde naaldhoutbosschen rond. Door opgewektheid en
vroolijkheid toont dit Insect zich bewust van het overwicht, dat zijn meerdere grootte
hem boven zijne verwanten verschaft. Gedurende het vliegen veroorzaken zijne wijngele
vleugels een duidelijk waarneembaar, gonzend gedruisch. Het wijfje steekt met haar
legboor groote pijlstaartrupsen, vooral die van Dennenpijlstaarten, welke in het door
haar bewoonde gebied meestal niet zeldzaam zijn; in iedere rups legt zij echter slechts
é é n ei.
*
1) Ichneumon pisorius, mannetje; daaronder de ledige pophuid van den Dennenpijlstaart (Sphinx pinastri),
waaruit deze Sluipwesp voortgekomen is.—2) Cryptus tarsoleucus, mannetje.—3) Mesostenus gladiator,
wijfje.—4) Ephialtes manifestator, mannetje en eierleggend wijfje. Ware grootte.
Mesostenus gladiator (fig. 3) heeft een zeer langen legboor; het rugschild van het
achterborststuk is ongedoornd en door ineenvloeiende putjes zeer oneffen. Deze groote
Wesp vliegt in Juni en wordt veelvuldig op oude muren aangetroffen; vermoedelijk legt
zij hare eieren in de larven van de hier wonende Graafwespen of Bijen.
*
Men zou de Cryptiden “Staartwespen” met gesteeld achterlijf kunnen noemen; in dit
geval komt aan de P i m p l a r i ë n den naam “Staartwespen” met z i t t e n d achterlijf
toe. De legboor van het wijfje, die den indruk wekt van een staart, komt bij eenige
geslachten uit een spleet aan de buikzijde, bij andere uit de spits van het achterlijf te
voorschijn en bereikt bij deze soms het drievoud van de lichaamslengte. Het geslacht
Rhyssa overtreft in dit opzicht alle leden der onderfamilie, terwijl het, wat grootte
betreft, aan de spits van de geheele familie staat. Bij eenige Noord-Amerikaansche
soorten heeft het wijfje een lichaamslengte van 35 mM. en bovendien een legboor, zoo
dik als een paardehaar, die 104 mM. lang is; de totale lengte van dit Insect komt dus
overeen met ⅔ van de lengte van het bedrukte deel dezer bladzijde. In onze
naaldhoutbosschen komt een soort voor, die bij de bedoelde Noord-Amerikaansche
weinig achterstaat in lichaamslengte, daar zij 28 cM. lang is, zonder den dubbel zoo
langen legboor. Deze “Pijpdoorsteker”, zooals een insectenverzamelaar Rhyssa
persuasoria placht te noemen, heeft den kop, het borststuk en het middengedeelte van
het achterlijf met witte vlekken op zwarten grond geteekend; de pooten zijn roodachtig
geel. Hare maden leven parasitisch in de larven van Houtwespen (Sirex), die diepe
gangen boren in stammen van naaldboomen. Het eierenleggende wijfje kan de legboor
tot aan den wortel, dus ongeveer 6 cM. diep, in gaaf hout steken en de hier verborgen
larve treffen.
Het soortenrijke geslacht Ephialtes heeft niet, gelijk Rhyssa, het middelste, maar het
achterste borstsegment aan de rugzijde van dwarse rimpels voorzien, het andere
daarentegen glad. Beide geslachten komen overeen door den langwerpigen vorm der
achterlijfsleden, de groote lengte van den legboor en de kleur der pooten. Evenals bij
vele andere Sluipwespen, verschilt de lengte bij leden van dezelfde soort naar gelang van
de grootte der larve, waarin zij parasiteerden. Ephialtes manifestator (fig. 4) kan 35 mM.
lang worden; de legboor steekt ongeveer even ver achter de spits van het achterlijf uit,
maar is langer dan ’t lichaam, daar zijn wortel in een spleet aan den buik verborgen is.
Na den langsten dag ziet men de soorten van dit geslacht in de bosschen rondzwerven en
op boomstammen zitten, vooral op zulke, waarin boorgaten voorkomen; deze moet het
wijfje, naar het schijnt, met den legboor volgen. Waarschijnlijk is het haar niet mogelijk
dit werktuig in het harde hout te doen doordringen en aan de hierin levende larven de
verzorging van hare jongen op te dragen; terwijl de Rhyssa-wijfjes, die zachtere
houtsoorten opzoeken, geen boorgat noodig hebben om haar doel te bereiken. Zeer
oplettend onderzoekt het Ephialtes-wijfje elk plekje met de vooruitgestoken sprieten,
welker spits zij boogvormig naar beneden kromt, besnuffelt iedere opening en verdiept
zich zoozeer in dezen arbeid, dat de aanwezigheid van een toeschouwer het schuwe dier
niet op de vlucht drijft. Wanneer eindelijk de geschikte plaats gevonden is, wordt het
achterlijf hoog opgeheven, zoodat het dier letterlijk op den kop staat, de legboor in de
gang gestoken en voorzichtig tot aan de larve voortgeschoven, waarbij de spits van het
achterlijf langzamerhand daalt en de scheede van den legboor loodrecht omhoog gericht
blijft. Dezen stand behoudt de Wesp, tot het ei gelegd is; zoolang dit duurt, verkeert zij
in volkomen weerloozen toestand, daar zij zichzelf vastgehecht heeft. In ’t volgende jaar
is haar larve volwassen en spint een rolronde, zwarte cocon; de Wesp doorknaagt deze
en gebruikt het boorgat van zijn gastheer om buiten te komen.
Een van de algemeenste Sluipwespen en tevens een van de grootste inheemsche leden
der onderfamilie, wanneer zij gedurende haar ontwikkeling over een overvloed van
voedsel kon beschikken, is Pimpla instigator, zwart van kleur, behoudens de scheen en
de voet van de 4 voorste pooten, en de scheen van de achterpooten, die helder geelrood
zijn. De reden van de algemeenheid dezer soort en van het afwisselen der grootte harer
leden van 11 tot 19.5 mM., is gelegen in de gewoonte van het wijfje om met de
verzorging van haar kroost een groot aantal rupsen van zeer verschillende soorten van
Vlinders te belasten; in den regel prest zij rupsen van Spinners voor deze taak. Daar zij
alle rupsen van dit slag, die in onze tuinen schade aanrichten, en vele van de meest
beruchte boschbedervers, zooals de rups van den Nonvlinder (Ocneria monacha), de
Processie-rups en de Dennenspinner-rups voor haar doel geschikt acht, ontmoet men
deze zwervelinge overal. In den regel ziet men haar met eenigszins opgeheven vleugels
rondwandelen op boomstammen, heggen, leemwanden, kortom op alle plaatsen, waar zij
een buit kan vinden. Voordat de rustig grazende rups gevaar vermoedt, treft haar de
legboor; in den kortst mogelijken tijd glijdt, ondanks alle afwerende bewegingen, het ei
door den korten eileider en dringt in haar lichaam door. Half springend, half vliegend is
de onheilstichter in ’t zelfde oogenblik al weer verdwenen, met het doel om haar snood
bedrijf in de onmiddellijke nabijheid voort te zetten. Zelfs de spinne-eieren zijn in hare
zijden nestjes niet veilig voor de aanslagen van deze Wespen.
Het geslacht Pimpla, dat (evenals Ephialtes) in tal van soorten over de geheele wereld
verbreid is, onderscheidt zich door een meer gedrongen lichaamsbouw: de
achterlijfsleden zijn, althans bij ’t wijfje, steeds breeder dan lang; de legboor bereikt
slechts zelden de lengte van het achterlijf.
Niet zelden merkt men aan de toppen der twijgen van jonge dennen een uitzweeting van
hars op. Ten onrechte heeft men deze producten “harsgallen” genoemd, daar zij niet door
woekering van het celweefsel van de plant ontstaan, maar door de werkzaamheid van
een in ’t jonge hout borende rups, die een harsachtig vocht doet uitvloeien, dat, aan de
lucht blootgesteld, verhardt. De hierdoor gevormde knobbels, welke mettertijd de grootte
van walnoten bereiken, bevatten de rups en later de pop van een sierlijk vlindertje
(Retinia resinella), dat tot de familie der Bladrollers (Tortricina) behoort. In de lente van
het tweede jaar van hun bestaan moet men deze “gallen” inzamelen om het Insect op te
kweeken; dikwijls echter ziet men in plaats van den Vlinder een zwarte Pimplariër van
nauwelijks 8.75 mM. lengte—Glypta resinanae—uit den knobbel te voorschijn komen.
In den zomer klautert dit wespje tusschen de dennenaalden rond; het behoeft
waarschijnlijk geen andere verblijfplaats op te zoeken, daar hier, zooals bekend is,
genoeg Bladluizen voorkomen, welker uitwerpselen gretig worden opgelekt. Het wijfje
onderzoekt nauwkeurig iedere jonge “harsgal” en weet zeer goed de hierin verborgen
rups te treffen. Deze blijft den geheelen winter door leven met den kiem des doods in ’t
lichaam; in de lente echter bevat haar woonplaats geen zwart vlinderpopje, maar een
cocon van lichte kleur, waaruit weldra de Sluipwesp ontwijkt.
De kop staat in den regel dicht voor de borst en is (behoudens eenige uitzonderingen)
van bijoogen, 6- (of 7-) ledige kaaktasters en 4-ledige liptasters voorzien. De sprieten
zijn niet gebroken, bij de meeste draad- of borstelvormig, gelijk in de geheele orde regel
is; vooral als tooi van de mannetjes treft men ook andere vormen aan. Het achterlijf is
gemiddeld dubbel zoo lang als de borst, bij het mannetje een weinig platgedrukt, bij het
wijfje meestal rolvormig. Aan de buikzijde ziet men de tweekleppige scheede van den
legboor, die op een mes, een schrobzaag, een vijl of een rasp gelijkt. Met alle overige
legboordragende Vliesvleugeligen hebben deze Wespen den tweeledigen dijring gemeen.
Verreweg de meeste larven van Bladwespen leven vrij op bladen, zelfstandiger dus dan
alle overige larven van Vliesvleugeligen. Daar zij door hare bonte kleuren op rupsen van
Vlinders gelijken, en er door onkundigen dikwijls mede verward worden, geeft men haar
den naam van b a s t a a r d r u p s e n . Meestal gezellig zitten zij in den toestand van rust
spiraalsgewijs opgerold op of onder een blad. Het eten geschiedt schrijlings zittend op
den rand van het blad, die door haar op een zeer eigenaardige wijze omzoomd wordt,
wanneer verscheidene zich op één blad bevinden. Ook hebben vele de zonderlinge
gewoonte om het achterlijf in den vorm van een vraagteeken op te heffen en regelmatig
op en neer te bewegen, zoodra één van haar begonnen is de maat te slaan. Met
uitzondering van het 4e en dikwijls ook van het voorlaatste lid van den romp, draagt elk
segment één paar korte pootjes; alleen de 3 voorste paren (de borstpooten) zijn
hoornachtig, in leden verdeeld en van een klauw voorzien; de overige gelijken op
vleeschknobbeltjes of wratten, die ingetrokken en uitgestulpt kunnen worden. Door de
genoemde gewoonten en door het aantal pooten (20 of 22) verschillen de bastaardrupsen
van de rupsen der Vlinders, die nooit meer dan 16 pooten hebben. Bij oppervlakkige
beschouwing schijnt haar huid naakt; hoewel men er bij nader onderzoek een dun
haarkleed en dikwijls ook duidelijke doorntjes op ziet, is zij echter nooit zoo dicht
behaard als bij vele echte rupsen. Zij is getooid met heldere kleuren; deze zijn niet
talrijk; de gewone teekening bestaat uit donkere vlekken op lichten grond. De
bastaardrupsen hebben enkelvoudige oogen en kleine sprieten; zij verwisselen
herhaaldelijk van huid, en veranderen daarbij soms niet slechts van kleur, maar ook van
vorm.
De larven van de meeste soorten verlaten, zoodra zij volwassen zijn, haar voederplant en
spinnen in of op den grond, onder afgevallen bladen of te midden van het mos, soms ook
aan den stengel van andere planten een tonvormige, meestal perkamentachtige, soms
echter zachtere cocon, waarin zij ineengekrompen en bewegingloos den winter
doorbrengen en eerst kort voor het uitvliegen van de Wesp in een vrije pop veranderen.
Sommige soorten brengen ieder jaar 2 of meer generaties voort; de ontwikkelingsduur
van de zomergeneratie is in dit geval zeer kort; bij andere soorten wordt voor de
volledige reeks van gedaantewisselingen een vol jaar of meer vereischt.
1) G e w o n e H o u t w e s p (Sirex juvencus); wijfje, larve en pop; ware grootte.—2) G e w o n e
H a l m w e s p (Cephus pygmaeus); twee roggehalmen zijn geopend voorgesteld, om de larve te toonen.—
3) Pachymerus calcitrator, een in de Halmwesp parasiteerende Sikkelwesp.—4) Larve van Cephus
vergroot volgens den daaronder aangeduiden maatstaf; daarnaast een pop (ware grootte) op haar gewone
rustplaats.
Beide Wespen leven slechts korten tijd en verschijnen nu eens vroeger, dan weer later,
niet dikwijls echter vóór het einde van Juni. Uitgezonderd de jaren waarin zij
buitengewoon veelvuldig zijn, krijgt men ze nagenoeg niet te zien, want zij houden zich
tamelijk goed verborgen op den stam of in de kroon der door hen bezochte boomen. Hun
luid gebrom gedurende het vliegen gelijkt eenigszins op dat van een Paardewesp en staat
naar alle waarschijnlijkheid in nauw verband met de twee spleten op het achterborststuk.
Het ei wordt 18 cM. diep in een gaven boomstam gelegd. De gekronkelde gangen, die de
voortdurend dieper doordringende larve in het hout knaagt, worden, daar zij in omvang
toeneemt, steeds wijder en kunnen ten slotte een middellijn van meer dan 4.5 mM.
bereiken. Zij zijn met houtspaantjes en uitwerpselen gevuld. Het is gebleken, dat de
larve soms eerst na verscheidene jaren volwassen is. Zij knaagt, om zich voor den
poptoestand een rustplaats te bereiden, het uiteinde van haar gang een weinig dieper uit
en vervaardigt een hier aanvangende gang, die zich tot dicht bij de oppervlakte van den
stam uitstrekt, om aan de Wesp het verlaten van het hout gemakkelijk te maken. Vooral
de larve van de Sparrenhoutwesp komt dikwijls met het timmerhout in woningen.
Bechstein verhaalt, dat in Juli 1798 in de boekdrukkerij te Schnepfenthal gedurende 10
opeenvolgende dagen iederen morgen een groot aantal gele Houtwespen uit de pas
gelegde vloer te voorschijn kwamen en bij de vensterruiten langs fladderden. In het huis
van een koopman te Schleusingen vertoonden zich in Juli 1843 een menigte van dezelfde
Wespen; deze waren echter afkomstig uit een balklaag onder de vloer, die in ’t vorige
jaar vernieuwd was; zij hadden zich dus ook door de planken een weg moeten banen. Te
Bautzen kwamen in Augustus 1856 eveneens uit de balklaag onder een vloer 60 à 80
exemplaren van de Gewone Houtwesp te voorschijn. Daar het huis al sedert 2½ jaar
voltooid was en de eieren in het hout gelegd moeten zijn, toen de balken in de open lucht
lagen, heeft het waarschijnlijk ongeveer 3 jaren geduurd, voordat deze kiemen als
Wespen de planken doorboorden. Dergelijke feiten heeft men ook waargenomen in
mijnen; hier speelden de gevleugelde Insecten vervolgens de rol van berggeesten en
doofden de lampen der mijnwerkers uit. Het is voorgekomen, dat deze dieren, om hun
streven naar vrijheid te bevredigen, behalve het hout, ook de looden platen, waarmede
het bekleed was, doorknaagden. Te Nuszdorf hebben gele Houtwespen herhaaldelijk
gaten gemaakt in de looden kamers van de zwavelzuurfabrieken; later geschiedde dit
ook te Freiberg door de staalblauwe soort. Uit de genoemde voorbeelden blijkt, hoe
lastig deze insecten soms kunnen zijn, hoewel zij in den gevleugelden toestand in hun
gewone omgeving aan de boomen geen schade meer toebrengen.
Een tweede, eveneens op dennen levende, veel schadelijker soort is die, welke door de
Duitsche houttelers G r o o t e D e n n e n b l a d w e s p (Lyda stellata, L. pratensis)
wordt genoemd, hoewel zij als imago kleiner, als larve slechts weinig grooter is dan de
vorige soort, van welke zij verschilt door de roodbruine zijranden van het (op den rug
bruinzwarte, aan de buikzijde vuilgele) achterlijf. Dezelfde woonplaats heeft de larve
van de R o o d k o p p i g e S p i n s e l b l a d w e s p (Lyda erythrocephala) als imago
kenbaar aan den staalblauwen romp en den rooden kop van het wijfje.
In het geslacht Selandria heeft men een groot aantal soorten van Bladwespen van kort-
eivormige gedaante vereenigd, waarbij ook de kleinste leden van de geheele familie zijn.
Zij kenmerken zich door het bezit van 2 rand- en 4 onderrandcellen, waarvan de 2e en de
3e de terugloopende aders opnemen; de 9-ledige, meestal draadvormige sprieten zijn zoo
lang als de kop en het borststuk te zamen genomen. Van de lente tot in den zomer vindt
men deze Wespen meestal op struiken. Bij ruw weer zitten zij stil en roerloos, maar zijn
steeds bereid om zich dood te houden, als men haar te na komt. Zeer vlug en vroolijk
vliegen zij daarentegen rond, als de zon haar warm op ’t lichaam schijnt.
Door knotsvormige sprieten en een plompen lichaamsbouw onderscheidt zich een laatste
groep van Bladwespen, waarvan de B e r k e n k n o p h o o r n w e s p (Cimbex betulae)
een voorbeeld levert. Door de breedte en de logge gedaante van het lichaam herinnert
haar geslacht aan dat der Hommels. De kop, het borststuk en de pooten zijn bij haar
zwart of geel behaard, hoewel niet zoo dicht, dat de zwarte kleur en de glans van de huid
hierdoor onzichtbaar worden. Het achterlijf is bij het wijfje lichter van kleur. De larve
heeft 22 pooten en is in volwassen toestand fraai groen van kleur; haar huid vertoont
talrijke, fijne dwarsplooien en is met witte wratjes onregelmatig bezaaid, vooral aan de
zijden. Zij heeft midden op den rug een niet geheel tot aan den kop reikende,
overlangsche, zwarte, geel gezoomde streep; ook de kop is geel. Men ziet deze
bastaardrups eenzaam op bladen van berken zitten. Evenals hare verwanten, heeft zij de
gewoonte uit de zijden van het lichaam een groenachtig vocht te laten ontwijken,
wanneer men haar aanvat.
Welcome to Our Bookstore - The Ultimate Destination for Book Lovers
Are you passionate about books and eager to explore new worlds of
knowledge? At our website, we offer a vast collection of books that
cater to every interest and age group. From classic literature to
specialized publications, self-help books, and children’s stories, we
have it all! Each book is a gateway to new adventures, helping you
expand your knowledge and nourish your soul
Experience Convenient and Enjoyable Book Shopping Our website is more
than just an online bookstore—it’s a bridge connecting readers to the
timeless values of culture and wisdom. With a sleek and user-friendly
interface and a smart search system, you can find your favorite books
quickly and easily. Enjoy special promotions, fast home delivery, and
a seamless shopping experience that saves you time and enhances your
love for reading.
Let us accompany you on the journey of exploring knowledge and
personal growth!
ebookgate.com