Computed tomography Principles design artifacts

and recent advances 2ed. Edition Hsieh J.

install download

https://ebookname.com/product/computed-tomography-principles-
design-artifacts-and-recent-advances-2ed-edition-hsieh-j/

Get the full ebook with Bonus Features for a Better Reading Experience on ebookname.com
Instant digital products (PDF, ePub, MOBI) available
Download now and explore formats that suit you...

Computed Aided Tomography Case Histories BRAIN 2nd

Edition D. Karthikeyan

https://ebookname.com/product/computed-aided-tomography-case-
histories-brain-2nd-edition-d-karthikeyan/

Cone Beam Computed Tomography in Orthodontics

Indications Insights and Innovations 1st Edition Sunil
Kapila

https://ebookname.com/product/cone-beam-computed-tomography-in-
orthodontics-indications-insights-and-innovations-1st-edition-
sunil-kapila/

Cardiac CT Made Easy An Introduction to Cardiovascular

Multidetector Computed Tomography 2nd Edition Paul
Schoenhagen

https://ebookname.com/product/cardiac-ct-made-easy-an-
introduction-to-cardiovascular-multidetector-computed-
tomography-2nd-edition-paul-schoenhagen/

Movement for Actors 1st Edition Nicole Potter

https://ebookname.com/product/movement-for-actors-1st-edition-
nicole-potter/
The Archaic The Past in the Present 1st Edition Paul
Bishop (Editor)

https://ebookname.com/product/the-archaic-the-past-in-the-
present-1st-edition-paul-bishop-editor/

Numerical and Analytical Methods with MATLAB 1st

Edition William Bober

https://ebookname.com/product/numerical-and-analytical-methods-
with-matlab-1st-edition-william-bober/

The Software IP Detective s Handbook Measurement

Comparison and Infringement Detection 1st Edition Bob
Zeidman

https://ebookname.com/product/the-software-ip-detective-s-
handbook-measurement-comparison-and-infringement-detection-1st-
edition-bob-zeidman/

Pearls and Pitfalls in Abdominal Imaging Pseudotumors

Variants and Other Difficult Diagnoses 1st Edition
Fergus V. Coakley Md

https://ebookname.com/product/pearls-and-pitfalls-in-abdominal-
imaging-pseudotumors-variants-and-other-difficult-diagnoses-1st-
edition-fergus-v-coakley-md/

Design Details for Health making the most of design s

healing potential 2e 2nd Edition Cynthia A. Leibrock

https://ebookname.com/product/design-details-for-health-making-
the-most-of-design-s-healing-potential-2e-2nd-edition-cynthia-a-
leibrock/
Cardiothoracic Surgery 3rd Edition Fritz J. Baumgartner

https://ebookname.com/product/cardiothoracic-surgery-3rd-edition-
fritz-j-baumgartner/
Bellingham, Washington USA
Library of Congress Cataloging-in-Publication Data
Hsieh, Jiang.
Computed tomography : principles, design, artifacts, and recent advances / Jiang Hsieh.
-- 2nd ed.
p. ; cm.
Includes bibliographical references and index.
ISBN 978-0-8194-7533-6
1. Tomography. I. SPIE (Society) II. Title.
[DNLM: 1. Tomography, X-Ray Computed. 2. Tomography Scanners, X-Ray
Computed--trends. 3. Tomography, X-Ray Computed--instrumentation. 4. Tomography,
X-Ray Computed--trends. WN 206 H873c 2009]
RC78.7.T6H757 2009
616.07'572--dc22
2009004797
Published by

SPIE
P.O. Box 10
Bellingham, Washington 98227-0010 USA
Phone: +1 360.676.3290
Fax: +1 360.647.1445
Email: [email protected]
Web: http://spie.org
and
John Wiley & Sons, Inc.
111 River Street
Hoboken, New Jersey 07030
Phone: +1 201.748.6000
Fax: +1 201.748.6088
ISBN: 9780470563533
Copyright © 2009 Society of Photo-Optical Instrumentation Engineers
All rights reserved. No part of this publication may be reproduced or distributed in any
form or by any means without written permission of the publisher.
The content of this book reflects the work and thought of the author(s).
Every effort has been made to publish reliable and accurate information herein, but the
publisher is not responsible for the validity of the information or for any outcomes
resulting from reliance thereon.
Printed in the United States of America.
Contents
Preface xi

Nomenclature and Abbreviations xiii

1 Introduction 1
1.1 Conventional X-ray Tomography 2
1.2 History of Computed Tomography 7
1.3 Different Generations of CT Scanners 14
1.4 Problems 19
References 19

2 Preliminaries 23
2.1 Mathematics Fundamentals 23
2.1.1 Fourier transform and convolution 23
2.1.2 Random variables 27
2.1.3 Linear algebra 30
2.2 Fundamentals of X-ray Physics 33
2.2.1 Production of x rays 33
2.2.2 Interaction of x rays with matter 36
2.3 Measurement of Line Integrals and Data Conditioning 42
2.4 Sampling Geometry and Sinogram 46
2.5 Problems 48
References 52

3 Image Reconstruction 55
3.1 Introduction 55
3.2 Several Approaches to Image Reconstruction 57
3.3 The Fourier Slice Theorem 61
3.4 The Filtered Backprojection Algorithm 65
3.4.1 Derivation of the filtered back-projection
formula 68
3.4.2 Computer implementation 71
3.4.3 Targeted reconstruction 85
3.5 Fan-Beam Reconstruction 88
3.5.1 Reconstruction formula for equiangular
sampling 89

v
vi Contents

3.5.2 Reconstruction formula for equal-spaced

sampling 95
3.5.3 Fan-beam to parallel-beam rebinning 97
3.6 Iterative Reconstruction 101
3.6.1 Mathematics verses reality 102
3.6.2 The general approach to iterative
reconstruction 103
3.6.3 Modeling of the scanner’s optics and physics 105
3.6.4 Updating strategy 109
3.7 Problems 112
References 114

4 Image Presentation 119

4.1 CT Image Display 119
4.2 Volume Visualization 123
4.2.1 Multiplanar reformation 123
4.2.2 MIP, minMIP, and volume rendering 128
4.2.3 Surface rendering 136
4.3 Impact of Visualization Tools 137
4.4 Problems 140
References 142

5 Key Performance Parameters of the CT Scanner 143

5.1 High-Contrast Spatial Resolution 143
5.1.1 In-plane resolution 144
5.1.2 Slice sensitivity profile 150
5.2 Low-Contrast Resolution 154
5.3 Temporal Resolution 160
5.4 CT Number Accuracy and Noise 167
5.5 Performance of the Scanogram 172
5.6 Problems 174
References 176

6 Major Components of the CT Scanner 179

6.1 System Overview 179
6.2 The X-ray Tube and High-Voltage Generator 180
6.3 The X-ray Detector and Data-Acquisition Electronics 190
6.4 The Gantry and Slip Ring 197
6.5 Collimation and Filtration 199
6.6 The Reconstruction Engine 202
6.7 Problems 203
References 205
Contents vii

7 Image Artifacts: Appearances, Causes, and Corrections 207

7.1 What Is an Image Artifact? 207
7.2 Different Appearances of Image Artifacts 209
7.3 Artifacts Related to System Design 214
7.3.1 Aliasing 214
7.3.2 Partial volume 226
7.3.3 Scatter 231
7.3.4 Noise-induced streaks 235
7.4 Artifacts Related to X-ray Tubes 239
7.4.1 Off-focal radiation 239
7.4.2 Tube arcing 242
7.4.3 Tube rotor wobble 244
7.5 Detector-induced Artifacts 244
7.5.1 Offset, gain, nonlinearity, and radiation
damage 244
7.5.2 Primary speed and afterglow 248
7.5.3 Detector response uniformity 253
7.6 Patient-induced Artifacts 258
7.6.1 Patient motion 258
7.6.2 Beam hardening 270
7.6.3 Metal artifacts 280
7.6.4 Incomplete projections 283
7.7 Operator-induced Artifacts 288
7.8 Problems 291
References 295

8 Computer Simulation and Analysis 301

8.1 What Is Computer Simulation? 301
8.2 Simulation Overview 303
8.3 Simulation of Optics 305
8.4 Computer Simulation of Physics-related Performance 316
8.5 Problems 323
References 324

9 Helical or Spiral CT 327

9.1 Introduction 327
9.1.1 Clinical needs 327
9.1.2 Enabling technology 331
9.2 Terminology and Reconstruction 332
9.2.1 Helical pitch 332
9.2.2 Basic reconstruction approaches 333
9.2.3 Selection of the interpolation algorithm and
reconstruction plane 339
9.2.4 Helical fan-to-parallel rebinning 343
9.3 Slice Sensitivity Profile and Noise 348
viii Contents

9.4 Helically Related Image Artifacts 355

9.4.1 High-pitch helical artifacts 355
9.4.2 Noise-induced artifacts 360
9.4.3 System-misalignment-induced artifacts 364
9.4.4 Helical artifacts caused by object slope 368
9.5 Problems 371
References 372

10 Multislice CT 375
10.1 The Need for Multislice CT 375
10.2 Detector Configurations of Multislice CT 378
10.3 Nonhelical Mode of Reconstruction 385
10.4 Multislice Helical Reconstruction 396
10.4.1 Selection of interpolation samples 398
10.4.2 Selection of region of reconstruction 402
10.4.3 Reconstruction algorithms with 3D
backprojection 405
10.5 Multislice Artifacts 410
10.5.1 General description 410
10.5.2 Multislice CT cone-beam effects 411
10.5.3 Interpolation-related image artifacts 413
10.5.4 Noise-induced multislice artifacts 416
10.5.5 Tilt artifacts in multislice helical CT 416
10.5.6 Distortion in step-and-shoot mode SSP 419
10.5.7 Artifacts due to geometric alignment 420
10.5.8 Comparison of multislice and single-slice
helical CT 422
10.6 Problems 422
References 425

11 X-ray Radiation and Dose-Reduction Techniques 433

11.1 Biological Effects of X-ray Radiation 434
11.2 Measurement of X-ray dose 436
11.2.1 Terminology and the measurement
standard 436
11.2.2 Other measurement units and methods 442
11.2.3 Issues with the current CTDI 443
11.3 Methodologies for Dose Reduction 445
11.3.1 Tube-current modulation 446
11.3.2 Umbra-penumbra and overbeam issues 448
11.3.3 Physiological gating 451
11.3.4 Organ-specific dose reduction 454
11.3.5 Protocol optimization and impact of the
operator 456
11.3.6 Postprocessing techniques 461
Contents ix

11.3.7 Advanced reconstruction 462

11.4 Problems 463
References 465

12 Advanced CT Applications 469

12.1 Introduction 469
12.2 Cardiac Imaging 471
12.2.1 Coronary artery calcification (CAC) 472
12.2.2 Coronary artery imaging (CAI) 476
12.2.2.1 Data acquisition and
reconstruction 478
12.2.2.2 Temporal resolution improvement 485
12.2.2.3 Spatial resolution improvement 492
12.2.2.4 Dose and coverage 493
12.3 CT Fluoroscopy 497
12.4 CT Perfusion 503
12.5 Screening and Quantitative CT 512
12.5.1 Lung cancer screening 512
12.5.2 Quantitative CT 516
12.5.3 CT colonography 519
12.6 Dual-Energy CT 522
12.7 Problems 532
References 534

Glossary 545

Index 551
Preface
Since the release of the first edition of this book in 2003, x-ray computed
tomography (CT) has experienced tremendous growth thanks to technological
advances and new clinical discoveries. Few could have predicted the speed and
magnitude of the progress, and even fewer could have predicted the diverse
nature of the technological advancement. The second edition of this book
attempts to capture these advances and reflect on their clinical impact.
The second edition provides significant changes and additions in several
areas. The first major addition is a new chapter on radiation dose. In the last few
years, significant attention has been paid to this subject by academic researchers,
radiologists, the general public, and the news media. An increased awareness of
the impact of radiation dose on human health has led to the gradual adoption of
the “as low as reasonably achievable” (ALARA) principle, the implementation of
American College of Radiology (ACR) accreditation and other dose reference
levels, and the development of many advanced dose-saving features for CT
scanners. The new Chapter 11 briefly describes some of the known biological
effects of radiation dose, then presents different dose definitions and
measurements, and concludes with an illustration of various dose-reduction
techniques.
At the time the first edition was published, the term “multislice” CT was an
accurate description of state-of-the-art scanners. Sixteen-slice scanners had just
been introduced commercially, and their clinical utilities and advantages had just
begun to be discovered. Since then, the “slice war” has continued, and now 64-,
128-, 256-, and 320-slice scanners are the new state of the art. These scanners
can be easily labeled as “cone-beam” CT. They require not only a detector with
wider coverage, but also other technologies such as new calibration techniques
and reconstruction algorithms. Chapter 10 has been significantly expanded to
discuss the technologies associated with these scanners and the new image
artifacts created by them.
Since the first edition, CT advancement has not been limited to the technology.
Advances also have been made in many areas of clinical applications, including the
rapid development of cardiac CT imaging and new applications inspired by the
reintroduction of dual-energy CT. Chapter 12 presents these advances and the
fundamental physics and technologies behind them.
Image artifacts have accompanied x-ray CT ever since its birth over 30 years
ago. Some artifacts are caused by the characteristics of the physics involved,
some are caused by technological limitations, some are created by new

xi
xii Preface

technologies, some are related to the patient, some result from suboptimal design,
and some are introduced by the operator. Chapter 7 has been expanded to reflect
the ever-evolving nature of these artifacts and various efforts to overcome them.
Historically, CT advances were driven by the development of new hardware.
However, it has become increasingly clear that hardware alone cannot solve all
of the technical and clinical problems that CT operators face. The second edition
includes significant updates to the section on statistical iterative reconstruction
technology and presents some of the exciting new developments in this area.
To enhance readers’ understanding of the material and to inspire creative
thinking about these subjects, a set of “problems” concludes each chapter. Many
problems are open-ended and may not have uniquely correct solutions. Hopefully
readers will find these problems useful and will develop new problems of their own.
At the time of this writing, the world is experiencing an unprecedented
financial crisis—that some call a financial “tsunami.” It is impossible to estimate
or predict the impact of this crisis on the market for x-ray CT. However, CT
technology is unlikely to remain stagnant. Many new exciting advances will take
place in both the technology and its clinical applications.

Acknowledgments
Many of the ideas, principles, results, and examples that appear in this book stem
from thoughts provoked by other books and research papers, and the author
would like to take this opportunity to acknowledge those sources. The author
would like to express his appreciation to Prof. Jeffrey A. Fessler of the
University of Michigan for his review of this text. His expert critical opinions
have significantly strengthened and enhanced the manuscript. The author owes a
debt to two people for supplying materials for both editions of this book: Dr.
Ting-Yim Lee of the Robarts Research Institute for providing reference materials
on CT perfusion, and Mr. Nick Keat of the ImPACT group in London for
supplying historical pictures on early CT development. The author would also
like to thank Dr. T. S. Pan of the M.D. Anderson Cancer Research Center for
providing some of the positron emission computed tomography (PET-CT)
images, and Dr. P. Kinahan of the University of Washington for providing
research results on patient motion artifacts. The author would like to thank many
current and former colleagues at GE Healthcare Technologies and the GE Global
Research Center for useful discussions, joint research projects, inspiration, and
many beautiful images. Finally, the most significant acknowledgment of all goes
to the author’s spouse, Lily J. Gong, for her unconditional support of the project,
and to his children, Christopher and Matthew, for their forgiveness of the missed
vacation.

Jiang Hsieh
August 2009
Nomenclature and Abbreviations
2D: two-dimensional
3D: three-dimensional
AAPM: American Association of Physicists in Medicine
ACR: American College of Radiology
ALARA: as low as reasonably achievable
ART: algebraic reconstruction technique
ASIC: application-specific integrated circuit
BMD: bone mineral density
bpm: beats per minute (heart rate)
CAC: coronary artery calcification
CAI: coronary artery imaging
CAT: computer-aided tomography
CBF: cerebral blood flow
CBV: cerebral blood volume
CDRH Center for Devices and Radiological Health (FDA)
CG: conjugate gradient
COPD: chronic obstructive pulmonary disease
CT: computed tomography
CTDI: CT dose index
DAS: data acquisition system
DECT: dual-energy CT
DFT: discrete Fourier transform
DLP: dose-length product
DSP: digital signal processing
EBCT: electron-beam computed tomography
EBT: electron-beam tomography
EC: European Commission
ECG / EKG: electrocardiogram
FBP: filtered backprojection
FDA: US Food and Drug Administration
FDK: Feldkamp-Davis-Kress (cone beam reconstruction algorithm)
FFT: fast Fourier transform
FOV: field of view
FWHM: full width at half maximum
FWTM: full width at tenth maximum
GDE: geometric detection efficiency

xiii
xiv Nomenclature and Abbreviations

GPU: graphic processor unit

HCT: helical computed tomography
HU: Hounsfield unit
IAC: inner auditory canal
IAEA: International Atomic Energy Agency
ICD: iterative coordinate decent
ICRP: International Commission on Radiological Protection
IFFT: inverse fast Fourier transform
IR: iterative reconstruction
LCD: low-contrast detectibility
LSF: line spread function
MIP: maximum intensity projection
minMIP: minimum intensity projection
ML: maximum likelihood
MPR: multiplanar reformation
MSAD: multiple-scan average dose
MTF: modulation transfer function
MTT: mean transit time
NPS: noise power spectrum
OS: ordered subset
PET: positron emission computed tomography
PSF: point spread function
QA: quality assurance
QDE: quantum detection efficiency
rad: radiation absorbed dose
RCA: right coronary artery
rem: Roentgen equivalent man
ROI: region of interest
SPECT: single-photon-emission computed tomography
SSP: slice sensitivity profile
Sv: sieverts
TAT: transverse axial tomography
VR: volume rendering
WL: (display) window level
WW: (display) window width
Chapter 1
Introduction
According to Webster’s dictionary, the word “tomography” is derived from the
Greek word “tomos” to describe “a technique of x-ray photography by which a
single plane is photographed, with the outline of structures in other planes
eliminated.”1 This concise definition illustrates the fundamental limitations of the
conventional radiograph: superposition and conspicuity due to overlapping
structures. In conventional radiography, the three-dimensional (3D) volume of a
human body is compressed along the direction of the x ray to a two-dimensional
(2D) image, as shown in Fig. 1.1(a). All underlying bony structures and tissues
are superimposed, which results in significantly reduced visibility of the object of
interest. Figure 1.1(b) shows an example of a chest x-ray study. The
superposition of the ribs, lungs, and heart is quite evident. Consequently, despite
the image’s superb spatial resolution (the ability to resolve closely placed high-
contrast objects), it suffers from poor low-contrast resolution (the ability to
differentiate a low-contrast object from its background). A recognition of this
limitation led to the development of conventional tomography.

Figure 1.1 Illustration of conventional x ray. (a) Acquisition setup and (b) example of a
chest study.
1
2 Chapter 1

1.1 Conventional X-ray Tomography

Conventional tomography is also known as planigraphy, stratigraphy,
laminagraphy, body section radiography, zonography, or noncomputed
tomography.2 One of the pioneers of conventional tomography was A. E. M.
Bocage.3 As early as 1921, Bocage described an apparatus to blur out structures
above and below a plane of interest. The major components of Bocage’s
invention consisted of an x-ray tube, an x-ray film, and a mechanical connection
to ensure synchronous movement of the tube and the film. The principle of
conventional tomography is illustrated in Fig. 1.2. For convenience, consider two
isolated points, A and B, located inside a patient. Point A is positioned on the
focal plane and point B is off the focal plane. The shadows cast on the x-ray film
by points A and B are labeled A1 and B1, respectively, as shown in Fig. 1.2(a).
The image produced on the film at this instant is not at all different from a
conventional radiograph. Next, we move both the x-ray source and the x-ray film
synchronously in opposite directions (for example, the x-ray source moves to the
left and the film moves to the right, as shown in the figure) to reach a second
location. We want to make sure that the shadow A2 produced by the stationary
point A overlaps with the shadow A1 produced by point A in the first position.
This can be easily accomplished by setting the distance traveled by the x-ray
source and the x-ray film to be proportional to their respective distances to point
A, as shown in Fig. 1.2(b). However, the shadow B2 produced by the stationary
point B at the second position does not overlap B1. This is due to the fact that
point B is off the focal plane, and the distance ratio from point B to the x-ray
source and to the film deviates significantly from that of point A. When the x-ray
tube and the film move continuously along a straight line (in opposite directions,
of course), the shadow produced by point B forms a line segment. This property
holds for any point located above or below the focal plane. Note that the
intensities of the shadows produced by the off-focus points are reduced, since the
shadows are distributed over an extended area. On the other hand, any point
located at the focal plane retains its image position on the film. Its shadow
remains a point and the corresponding intensity is not degraded.
Conventional tomography has several problems. Although the focal plane in
conventional tomography is theoretically a true plane, planes close to the focal
plane undergo little blurring. If we use the amount of blurring to judge whether a
point belongs to the focal plane, the slice thickness based on this definition
depends on the sweep angle , as shown in Fig. 1.3. In fact, the slice thickness is
inversely proportional to tan(). Clearly,  must be fairly large to obtain a
reasonable slice thickness.
Another problem associated with conventional tomography is the fact that
little blurring takes place in the direction perpendicular to the movement of the x-
ray source and the film. The net effect is that for structures parallel to the
direction of the source motion, the sharpness of the shadow boundaries is not
significantly reduced as desired. These structures appear to be elongated only
Introduction 3

Figure 1.2 Illustration of conventional tomography principle. (a) X-ray source and film
produce shadows A1 and B1 of points A and B at a first position. (b) X-ray source and film
are moved reciprocally such that shadow A2 of point A overlaps shadow A1, but shadow
B2 of point B does not overlap B1.

Figure 1.3 Illustration of slice thickness as a function of the scan angle.

4 Chapter 1

along the direction of motion. This effect is illustrated with a computer

simulation shown in Fig. 1.4. The imaged object was made of two long ellipsoids
and two spheres with a 2:1 density ratio in favor of the ellipsoids. The goal was
to enhance the visibility of the spheres. In the first simulation, the ellipsoids were
placed such that their long axes were perpendicular to the direction of the source
motion. For comparison, a conventional radiography image (stationary source
and detector) is shown in Fig. 1.4(a) and a conventional tomography image is
shown in Fig. 1.4(b). Compared to the conventional radiograph, the ellipsoids in
the conventional tomography image were blurred by the source motion because
the ellipsoids were placed away from the focal plane. (The spheres were located
on the focal plane.) The improvement in sphere visibility is obvious. When the
ellipsoids were rotated 90 deg so their long axes became parallel to the source
motion direction, little blurring took place, since the path lengths through the
ellipsoids at different source locations were virtually unchanged. Consequently,
no improvement in sphere visibility was obtained.
To partially compensate for the lack of tomographic effect in certain
directions, pluridirectional tomography has been proposed.2 For these devices,
the x-ray source and the film synchronously undergo more complicated motion

Figure 1.4 Simulated images of conventional tomography. The phantoms are made of two
long ellipsoids and two spheres. The top row depicts the scenario in which the long axes
of the ellipsoids are perpendicular to the direction of source-detector motion, and the
bottom row depicts the scenario in which the ellipsoids’ long axes are parallel to the
motion. (a) and (c) show conventional radiography images of the phantoms; (b) and (d)
show conventional tomography images of the phantoms.
Introduction 5

patterns, such as circular, ellipsoidal, sinusoidal, hypocycloidal, or spiral. Figure

1.5 depicts an example of an elliptical motion pattern that produces more uniform
blurring of the structures outside the focal plane. Disadvantages of
pluridirectional tomography include higher cost, increased procedure time, and a
larger x-ray dose to the patient.
Instead of forming a focal plane parallel to the patient long axis, axial
transverse tomography (also known as transverse axial tomography or TAT)
defines a cross-sectional plane that is perpendicular to the patient long axis, as
shown in Fig. 1.6. In this apparatus, the x-ray source is stationary and is oriented
at a shallow angle  with respect to the x-ray film. Both the patient and the film
rotate about their vertical axes synchronously at an identical direction and speed.
Because the geometric relationship between the x-ray source, the patient, and the
film remains unchanged, the magnification factor for each point located inside
the tomographic plane is constant (the magnification factor is defined as the
distance between the source and the shadow on the film over the distance
between the source and the point in the tomographic plane).
During the imaging process, structures inside the tomographic plane remain
in sharp focus, since structures inside the plane remain in the field of view (FOV)
at all times, and the shadow locations produced by these structures do not change
relative to the film. On the other hand, structures outside the tomographic plane
do not always stay inside the FOV, and their shadows move around relative to
the film during the scan. Thus, these shadows do not appear as sharp. Strictly
speaking, the tomographic plane is actually a volume. The thickness of the
volume decreases with the angle  between the center ray of the source and the
film. Since  is limited by many practical factors, the minimum thickness of the
tomographic volume is also limited. For example, for an extremely small , the

Figure 1.5 Illustration of pluridirectional tomography.

6 Chapter 1

amount of x-ray flux detected by the film is severely limited, and the image
quality is degraded by the quantum noise.
Although all of these tomographic techniques are somewhat successful in
producing images at the plane of interest, they all suffer from two fundamental
limitations: these techniques do not increase the object contrast, and they cannot
completely eliminate other structures outside the focal plane. Note that
conventional tomography blurs the overlying structure. The visibility of the
structures inside the focal plane may be enhanced, but the contrast between
different structures inside the focal plane is not enhanced. In addition, the blurred
overlying structures superimposed on the tomographic image significantly
degrade the quality of the image. Combined with the larger x-ray dose to the
patient, conventional tomography has found limited use in clinical applications.
With the development of digital flat-panel technology, the combination of
digital processing techniques and conventional tomographic acquisitions has
found renewed interest.4–8 The combined technique is often called tomosynthesis.
Compared to conventional tomography where the tomographic effect is achieved
through analog means, tomosynthesis has the advantage of producing
tomographic images at different depths with a single data acquisition and
additional improvements in image quality using image processing technologies.
It offers the potential of providing low-dose alternatives to x-ray CT for certain
clinical applications, such as breast cancer or lung cancer screening. Figure 1.7
shows an example of the improved visualization of a stone in a ureter (b)
compared to the conventional radiograph (a). Note that the intensities of the spine
and ribs in the reconstructed tomosynthesis image are significantly reduced but
not completely eliminated.

Figure 1.6 Schematic diagram of axial transverse tomography.

Introduction 7

Figure 1.7 Clinical example of tomosynthesis. (a) Conventional anterior-posterior (A-P)

radiography. (b) Tomosynthesis image showing a stone in the right ureter.

1.2 History of Computed Tomography

It is quite remarkable that image reconstruction from projections was attempted
as early as 1940.9 Needless to say, these attempts were made without the benefits
of modern computer technology. In a patent granted in 1940, Gabriel Frank
described the basic idea of today’s tomography.10 The patent includes drawings
of equipment to form sinograms (representations of measurement data as linear
samples versus view samples) and optical backprojection techniques to
reconstruct images. Backprojection can be roughly described as a process in
which the intensity of a sample is distributed uniformly along the path that
formed the sample. A more detailed discussion on the subject can be found in
Chapter 3. Although images generated from the proposed approach suffer from
blurring, the patent clearly envisioned the fundamental requirements for
tomographic devices.
Twenty-one years later, William H. Oldendorf, an American neurologist
from Los Angeles, performed a series of experiments based on principles similar
to those later used in CT.11 The objective of his work was to determine whether
internal structures within dense structures could be identified by transmission
measurements. His experimental apparatus is schematically diagrammed in Fig.
1.8. The phantom consists of two concentric rings of iron nails embedded in a
plastic block of 10 × 10 × 4 cm to represent the cranial vault. Another iron nail
and an aluminum nail were placed 1.5 cm apart near the center of the ring. The
8 Chapter 1

phantom was placed on a model train. The train was pulled by a clock motor
along the track at a slow speed (88 mm per hour). A relatively faster rotational
motion was provided by placing the entire apparatus on a turntable rotating at 16
revolutions per minute (rpm). A collimated radioiodine (131I) source provided a
pencil beam gamma ray to irradiate the phantom, and the gamma ray beam
always passed through the center of rotation. The signal was detected by a
sodium iodine scintillation crystal and a photomultiplier.
To understand the nature of the measurement, consider the intensity
modulation of the measured beam. Every nail in the phantom passed the gamma
ray exactly twice per rotation. At a rotation rate of 16 rpm, the relatively rapid
variation in the transmitted beam intensity caused by the peripheral nails was a
high-frequency signal. The translation of the central nails near the center of
rotation caused a relatively slow variation in the transmitted beam, resulting in
low-frequency signals. A low-frequency signal can be separated from a high-
frequency signal through filtering (Oldendorf used an electronic low-pass filter
with a 30-sec time constant). For this experiment, only one line passing through
the center of rotation was reconstructed. The reconstruction of additional lines
would require shifting the phantom relative to the center of rotation. Because
each scan required one hour to complete and there was no appropriate means of
storing the data, no attempt was made to reconstruct the 2D structure.

Figure 1.8 Simplified diagram of Oldendorf’s experiment.

Introduction 9

In 1963, David E. Kuhl and Roy Q. Edwards introduced transverse

tomography using radioisotopes, which was further developed and evolved into
today’s emission computed tomography.12 A sequence of scans was acquired at
uniform steps and regular angular intervals with two opposing radiation
detectors. At each angle, the film was exposed to a narrow line of light moving
across the face of a cathode ray tube with a location and orientation
corresponding to the detectors’ linear position. This is essentially the analog
version of the backprojection operation. The process was repeated at 15-deg
angular increments (the film was rotated accordingly to sum up the backprojected
views). In later experiments, the film was replaced by a computer-based
backprojection process. What was lacking in these attempts was an exact
reconstruction technique.
The mathematical formulation for reconstructing an object from multiple
projections dates back to the Austrian mathematician J. Radon. In 1917, Radon
demonstrated mathematically that an object could be replicated from an infinite
set of its projections. The concept was first applied by R. N. Bracewell in 1956 to
reconstruct a map of solar microwave emissions from a series of radiation
measurements across the solar surface.13 Between 1956 and 1958, several
Russian papers accurately formulated the tomographic reconstruction problem as
an inverse Radon transform.14–16 These papers discussed issues associated with
the implementation and proposed methodologies of performing reconstructions
with television-based systems. Although these algorithms were somewhat
inefficient, they offered a satisfactory performance.17
In 1963, Allan M. Cormack reported the findings from investigations of
perhaps the first CT scanner actually built.18 His work could be traced back to
1955 when he was asked to spend one and a half days a week at Groote Schuur
Hospital attending to the use of isotopes after the resignation of the hospital
physicist (Cormack was the only nuclear physicist in Capetown). While
observing the planning of radiotherapy treatments, Cormack came to realize the
importance of knowing the x-ray attenuation coefficient distribution inside the
body. He wanted to reconstruct attenuation coefficients of tissues to improve the
accuracy of radiation treatment. During a sabbatical at Harvard University in late
1956, he derived a mathematical theory for image reconstruction. When he
returned to South Africa in 1957, he tested his theory in a laboratory simulation
using a 5-cm-thick disk that was 20 cm in diameter. The disk consisted of a
central cylinder of pure aluminum (1.13 cm in diameter) surrounded by an
aluminum alloy annulus. This was surrounded in turn by an oak annulus. A
collimated 60Co isotope was used as the radiation source and a Geiger counter as
a detector. The disc was translated through the gamma ray in 5-mm steps to form
a linear scan. Because of circular symmetry, the disc was scanned at only one
angular position (the linear scans obtained from other angles would be identical).
The attenuation coefficients were calculated for aluminum and wood with the use
of his reconstruction technique.
In 1963, Cormack repeated his experiment at Tufts University (where he
became a member of the physics department in 1957) with a circularly
10 Chapter 1

unsymmetrical phantom of aluminum and plastic. The phantom consisted of an

outer ring of aluminum to represent the skull, a Lucite filling to represent soft
tissue, and two aluminum disks within the Lucite to represent tumors. Again, a
collimated gamma ray beam was used as the source. Twenty-five linear scans
were performed at 7.5-deg angular increments over a 180-deg angle. The results
of two experiments were published in 1963 and 1964, respectively.
Unfortunately, little attention was paid at the time to his work due to the time and
difficulty of performing the necessary calculations. Cormack remarked in his
1979 Nobel lecture that “There was virtually no response. The most interesting
request for a reprint came from the Swiss Center for Avalanche Research. The
method would work for deposits of snow on mountains if one could get either the
detector or the source into the mountain under the snow!” 19
The development of the first clinical CT scanner began in 1967 with Godfrey
N. Hounsfield at the Central Research Laboratories of EMI, Ltd. in England.
While investigating pattern recognition techniques, he deduced, independent of
Cormack, that x-ray measurements of a body taken from different directions
would allow the reconstruction of its internal structure.20 Preliminary calculations
by Hounsfield indicated that this approach could attain a 0.5% accuracy of the x-
ray attenuation coefficients in a slice. This is an improvement of nearly a factor
of 100 over the conventional radiograph. For their pioneering work in CT,
Cormack and Hounsfield shared the 1979 Nobel Prize in Physiology and
Medicine.
The first laboratory scanner was built in 1967, as shown in Fig. 1.9. Linear
scans were performed on a rotating specimen in 1-deg steps (the specimen
remained stationary during each scan). Because of the low-intensity americium
gamma source, it took nine days to complete the data acquisition and produce a
picture. Unlike the reconstruction method used by Cormack, a total of 28,000
simultaneous equations had to be solved by a computer in 2.5 hours (a more
detailed discussion on this reconstruction technique can be found in Section 3.4).
The use of a modified interpolation method, a higher-intensity x-ray tube, and a
crystal detector with a photomultiplier reduced the scan time to nine hours and
improved the accuracy from 4% to 0.5%.
The first clinically available CT device was installed at London’s Atkinson-
Morley Hospital in September 1971, after further refinement on the data
acquisition and reconstruction techniques. Images could be produced in 4.5
minutes. On October 4, 1971, the first patient, who had a large cyst, was scanned
and the pathology was clearly visible in the image.21 Figure 1.10 depicts a
patient’s head scan performed with the first clinical CT device, and Fig. 1.12(a)
depicts one of the first clinical head images obtained from a CT scanner.
Humans are not the sole beneficiaries of this wonderful invention. Over the
years, CT scanners have been used to scan trees, animals, industrial parts,
mummies, and just about everything that can fit inside a CT gantry. As an
example, Fig. 1.11 shows two images of scanning procedures on big cats from
zoos. Given the fact that CT is often called a CAT (computer-aided tomography)
scanner, scanning cats by CAT seems to be quite natural!
Introduction 11

Figure 1.9 Original lathe and scanner used in an early CT experiment by Hounsfield at
EMI Central Research Laboratory in 1967.

Figure 1.10 A patient head scan performed on the first clinical CT scanner. [Figure
supplied by ImPACT (www.impactscan.org), St George’s Hospital NHS Trust, London,
and reprinted with permission.]
12 Chapter 1

Figure 1.11 Another application of a medical CT: use of CAT to scan big cats.

Since the introduction of the first clinical scanner, tremendous advancement

has been made in CT technology. For illustration, Fig. 1.12(b) depicts a CT head
image obtained with a state-of-the-art scanner. Improvements in spatial and low-
contrast resolution are evident. As another simple illustration, we can examine
the progress made on one performance benchmark: time to acquire a single slice.
To cover a volume, consecutive slices must be acquired with an interslice
spacing that is roughly equal to the slice thickness. Therefore, the scan time of a
single slice includes the actual data acquisition time as well as interscan delays
between slices. Although other benchmarks such as gantry speed, reconstruction
time, or tube power can be considered, the scan time per slice is a more direct
indicator of the time to cover a fixed volume. Figure 1.13 plots the reported scan
time over 30 years of CT history. For clarity, a logarithmic scale is used. A
straight line (in logarithmic scale) was fitted through the samples depicted by the
Introduction 13

diamonds. The reduction in scan time (and therefore, the increase in volume
coverage speed) follows an exponential relationship over time. Based on the
slope of the fitted straight line, we can calculate the time it takes to reduce the
scan time to half; the result is 2.3 years. This means that CT scanners have
doubled their acquisition power every 2.3 years over the last 30 years! This
performance matches Moore’s law for electronics.

Figure 1.12 Comparison of CT head images acquired on (a) one of the first CT scanners
and (b) the GE LightSpeed VCT 2005.

Figure 1.13 Scan time per slice as a function of time (excluding electron-beam computed
tomography, discussed in Section 1.3). The scan time/slice decreased exponentially over
time and was reduced by a factor of 1.34 per year.
14 Chapter 1

1.3 Different Generations of CT Scanners

The previous section presented a brief review of CT history. This section
describes the evolution of CT technology over the last 30 years. Because of the
nature of the overview, detailed comparisons of the pros and cons of different
types of scanners are omitted. We will provide more in-depth coverage on this
subject in later sections of the book, wherever appropriate. For example, many
discussions can be found in Chapter 7 as part of the image artifacts analysis.
The type of scanner built by EMI in 1971 is called the first-generation CT. In
a first-generation scanner, only one pencil beam is measured at a time. In the
original EMI head scanner, the x-ray source was collimated to a narrow beam of
3 mm wide (along the scanning plane) and 13 mm long (across the scanning
plane). The x-ray source and detector were linearly translated to acquire
individual measurements. The original scanner collected 160 measurements
across the scan field. After the linear measurements were completed, both the x-
ray tube and the detector were rotated 1 deg to the next angular position to
acquire the next set of measurements, as shown in Fig. 1.14.
Although clinical results from the first-generation scanners were promising,
there remained a serious image quality issue associated with patient motion
during the 4.5-min data acquisition.22 The data acquisition time had to be

Figure 1.14 First-generation CT scanner geometry. At any time instant, a single

measurement is collected. The x-ray tube and detector translate linearly to cover the
entire object. The entire apparatus is then rotated 1 deg to repeat the scan.
Introduction 15

reduced. This need led to the development of the second-generation scanner

illustrated in Fig. 1.15. Although this was still a translation-rotation scanner, the
number of rotation steps was reduced by the use of multiple pencil beams. The
figure depicts a design in which six detector modules were used. The angle
between the pencil beams was 1 deg. Therefore, for each translation scan,
projections were acquired from six different angles. This allowed the x-ray tube
and detector to rotate 6 deg at a time for data acquisition, representing a
reduction factor of 6 in acquisition time. In late 1975, EMI introduced a 30-
detector scanner that was capable of acquiring a complete scan under 20 sec.9
This was an important milestone for body scanning, since the scan interval fell
within the breath-holding range for most patients.
One of the most popular scanner types is the third-generation CT scanner
illustrated in Fig. 1.16. In this configuration, many detector cells are located on
an arc concentric to the x-ray source. The size of each detector is sufficiently
large so that the entire object is within each detector’s FOV at all times. The x-
ray source and the detector remain stationary to each other while the entire
apparatus rotates about the patient. Linear motion is eliminated to significantly
reduce the data acquisition time.
In the early models of the third-generation scanners, both the x-ray tube
power and the detector signals were transmitted by cables. Limitations on the
length of the cables forced the gantry to rotate both clockwise and

Figure 1.15 Second-generation CT scanner geometry. At each time instant, measurements

from six different angles are collected. Although the x-ray source and detector still need to
be linearly translated, the x-ray tube and detector can rotate every 6 deg.
16 Chapter 1

counterclockwise to acquire adjacent slices. The acceleration and deceleration of

the gantry, which typically weighed several hundred kilograms, restricted the
scan speed to roughly 2 sec per rotation. Later models used slip rings for power
and data transmission. Since the gantry could rotate at a constant speed during
successive scans, the scan time was reduced to 0.5 sec or less. The introduction
of slip ring technology was also a key to the success of helical or spiral CT
(Chapter 9 is devoted to this topic). Because of the inherent advantages of the
third-generation technology, nearly all of the state-of-the-art scanners on the
market today are third generation.
Several technology challenges in the design of the third-generation CT,
including detector stability and aliasing, led to investigations of the fourth-
generation concept depicted in Fig. 1.17. In this design, the detector forms an
enclosed ring and remains stationary during the entire scan, while the x-ray tube
rotates about the patient. Unlike the third-generation scanner, a projection is
formed with signals measured on a single detector as the x-ray beam sweeps
across the object. The projection, therefore, forms a fan with its apex at the
detector, as shown by the shaded area in Fig. 1.17 (a projection in a third-
generation scanner forms a fan with the x-ray source as its apex). One advantage
of the fourth-generation scanner is the fact that the spacing between adjacent
samples in a projection is determined solely by the rate at which the
measurements are taken. This is in contrast to third-generation scanning in which

Figure 1.16 Third-generation CT scanner geometry. At any time instant, the entire object
is irradiated by the x-ray source. The x-ray tube and detector are stationary with respect to
each other while the entire apparatus rotates about the patient.
Introduction 17

the sample spacing is determined by the detector cell size. A higher sampling
density can eliminate potential aliasing artifacts. In addition, since at some point
during every rotation each detector cell is exposed directly to the x-ray source
without any attenuation, the detector can be recalibrated dynamically during the
scan. This significantly reduces the stability requirements of the detector.
A potential drawback of the fourth-generation design is scattered radiation.
Because each detector cell must receive x-ray photons over a wide angle, no
effective and practical scatter rejection can be performed by a post-patient
collimator. Although other scatter correction schemes, such as the use of a set of
reference detectors or software algorithms, are useful, the complexity of these
corrections is likely to increase significantly with the introduction of multislice or
volumetric CT.
A more difficult drawback to overcome is the number of detectors required
to form a complete ring. Since the detector must surround the patient at a fairly
large circumference (to maintain an acceptable source-to-patient distance), the
number of detector elements and the associated data acquisition electronics
become quite large. For example, a recent single-slice fourth-generation scanner
required 4800 detectors. The number would be much higher for multislice
scanners. Thus, for economical and practical reasons, fourth-generation scanners
are likely to be phased out.9

Figure 1.17 Geometry of a fourth-generation CT scanner. At any time instant, the x-ray
source irradiates the detectors in a fan-shaped x-ray beam, as shown by the solid lines. A
projection is formed using measurement samples from a single detector over time, as
depicted by the shaded fan-shaped region.
18 Chapter 1

The electron-beam scanner, sometimes called the fifth-generation scanner,

used in electron-beam computed tomography (EBCT), or electron-beam
tomography (EBT), was built between 1980 and 1984 for cardiac applications.23
To “freeze” cardiac motion, a complete set of projections must be collected
within 20 to 50 ms. This is clearly very challenging for conventional third- or
fourth-generation types of scanners due to the enormous centripetal force placed
on the x-ray tube and the detector. In the electron-beam scanner, the rotation of
the source is provided by the sweeping motion of the electron beam (instead of
the mechanical motion of the x-ray tube). Figure 1.18 shows a simplified
schematic diagram of an electron-beam scanner. The bottom arc (210 deg)
represents an anode with multiple target tracks. A high-speed electron beam is
focused and deflected by carefully designed coils to sweep along the target ring,
similar to a cathode ray tube. The entire assembly is sealed in a vacuum. Fan-
shaped x-ray beams are produced and collimated to a set of detectors, represented
by the top arc of 216 deg. The detector ring and the target ring are offset
(noncoplanar) to make room for the overlapped portion. When multiple target
tracks and detector rings are used, a coverage of 8 cm along the patient long axis
can be obtained for the heart. Since the system has no mechanical moving parts,
scan times as fast as 50 ms can be achieved. However, for noise considerations,
multiple scans are often averaged to produce the final image. A more complete
description of the electron-beam scanner can be found in Ref 24.

Figure 1.18 Geometry of an electron-beam scanner.

Introduction 19

1.4 Problems
1-1 Section 1.1 discussed conventional tomography and the many drawbacks
and shortcomings of this technique. However, recently many research
and development activities have combined conventional tomography
with advanced reconstruction techniques. This technology is called
tomosynthesis, as illustrated by a clinical example in Fig. 1.7. To
understand the rationale behind its comeback, list at least three
advantages of tomosynthesis over CT.
1-2 Figure 1.4 illustrates one limitation of conventional tomography.
Pluridirectional tomography was invented to overcome this shortcoming.
What are the limitations of this approach? Design a phantom to
demonstrate its limitations.
1-3 In a breast tomosythesis device similar to the configuration shown in Fig.
1.2, the film is replaced by a digital detector that remains stationary
during data acquisition. Only the x-ray tube translates during data
acquisition. What are the design considerations that ensure a good
resolving capability of such a device to suppress overlapping structures
across the entire volume?
1-4 In a tomosynthesis device similar to the one shown in Fig. 1.2, a digital
detector replaces the film and is capable of translational motion, as
shown in the figure. Is the focal plane uniquely defined by the speed of
the x-ray tube and the detector as in the case of conventional
tomography? If not, how is the focal plane defined, assuming the digital
detector is capable of high-speed sampling?
1-5 One application of x-ray CT technology is luggage scanning for airport
security. Luggage is transported over a conveyer belt, and cross-sectional
images of the luggage are generated and analyzed for security threats.
List at least three similarities and three differences between a medical CT
device and a luggage CT device in terms of technical design
considerations.
1-6 List three advantages and three disadvantages of the fourth-generation
CT scanner over the third-generation CT scanner.
1-7 In many micro-CT scanners designed for small-animal imaging, the CT
gantry remains stationary while the animal spins about its long axis,
while the gantry in a human clinical CT scanner rotates about the patient.
Describe the pros and cons of each design.

References
1. D. B. Guralnik, Ed., Webster’s New World Dictionary of the American
Language, 2nd college edition, William Collins + World Publishing Co.,
Cleveland (1974).
20 Chapter 1

2. C. L. Morgan, Basic Principles of Computed Tomography, University Park

Press, Baltimore (1983).
3. A. E. M. Bocage, “Procede et dispositifs de radiographie sur plaque en
mouvement,” French Patent No. 536,464 (1921).
4. J. T. Dobbins III and D. J. Godfrey, “Digital x-ray tomosynthesis: current
state of the art and clinical potential,” Phys. Med. Biol. 48, 65–106 (2003).
5. R. J. Warp, D. J. Godfrey, and J. T. Dobbins III, “Applications of matrix
inversion tomosynthesis,” Proc. SPIE 3977, 376–383 (2000).
6. G. M. Stevens, R. Fahrig, and N. J. Pelc, “Filtered backprojection for
modifying the impulse response of circular tomosynthesis,” Med. Phys.
28(3), 372–380 (2001).
7. B. Nett, S. Leng, and G.-H. Chen, “Plannar tomosynthesis reconstruction in a
parallel-beam framework via. Virtual object reconstruction,” Proc. SPIE
6510, 651028 (2007).
8. T. Deller, K. Jabri, J. Sabol, X. Ni, G. Avinash, R. Saunders, and R.
Uppaluri, “Effect of acquisition parameters on image quality in digital
tomosynthesis, Proc. SPIE 6510, 65101L (2007).
9. L. W. Goldman, “Principles of CT and the evolution of CT technology,” in
Categorical Course in Diagnostic Radiology Physics: CT and US Cross-
sectional Imaging, L. W. Goldman and J. B. Fowlkes, Eds., Radiological
Society of North America, Inc., Oak Brook, IL (2000).
10. S. Webb, “Historical experiments predating commercially available
computed tomography,” British J. of Radiology 65, 835–837 (1992).
11. W. H. Oldendorf, “Isolated flying spot detection of radiodensity
discontinuities: displaying the internal structural pattern of a complex
object,” IEEE Trans. Biomed. Electr. 8, 68–72 (1961).
12. D. E. Kuhl and R. Q. Edwards, “Reorganizing data from transverse section
scans of the brain using digital processing,” Radiology 91, 975–983 (1968).
13. R. N. Bracewell, “Strip integration in radiation astronomy,” Australian J. of
Physics 9, 198–217 (1956).
14. S. I. Tetel’baum, “About the problem of improvement of images obtained
with the help of optical and analog instruments,” Bull. Kiev Polytech. Inst.
21, 222 (1956) (in Russian).
15. S. I. Tetel’baum, “About a method of obtaining volume images with the help
of x-rays,” Bull. Kiev Polytech. Inst. 22, 154–160 (1957) (in Russian).
16. B. I. Korenblyum, S. I. Tetel’baum, and A. A. Tyutin, “About one scheme of
tomography,” Bull. Inst. Higher Educ. Radiophys. 1, 151–157 (1958).
17. H. H. Barrett, W. G. Hawkins, and M. L. G. Joy, “Historical note on
computed tomography,” Radiology 147, 172 (1983).
18. A. M. Cormack, “Representation of a function by its line integrals, with
some radiological applications,” J. Appl. Phys. 34, 2722–2727 (1963).
Introduction 21

19. A. M Cormack, “Early two-dimensional reconstruction and recent topics

stemming from it,” Nobel Lecture, December 9, 1979.
20. G. N. Hounsfield, “Historical notes on computerized axial tomography,” J. of
the Canadian Assoc. of Radiologists 27, 135–142 (1976).
21. J. Ambrose, “A brief review of the EMI scanner,” Proc. Br. Inst. Radiol. 48,
605–606 (1975).
22. D. Schellinger, G. Di Chiro, S. Axelbaum, H. L. Twigg, and R. S. Ledley,
“Early clinical experience with the ACTA scanner,” Radiology 114, 257–261
(1975).
23. D. P. Boyd, R. G. Gould, J. R. Quinn, R. Sparks, J. H. Stanley, and W. B.
Herrmannsfeldt, “A proposed dynamic cardiac 3D densitometer for early
detection and evaluation of heart disease,” IEEE Trans. Nucl. Sci. 26, 2724–
2727 (1979).
24. C. H. McCollough, “Principles and performance of electron beam CT,” in
Medical CT and Ultrasound: Current Technology and Applications, L. W.
Goldman and J. B. Fowlkes, Eds., Advanced Medical Publishing, Madison,
WI, 487–518 (1995).
Chapter 2
Preliminaries
This chapter covers two major topics: the fundamentals of the mathematics and
the fundamentals of the physics involved in x-ray physics. The objective is to
provide a general review of the important mathematics tools and background
knowledge of x-ray physics that are used throughout the book. The chapter
serves mainly as a refresher. Readers who have not been previously exposed to
these topics will find a list of recommended reference materials at the end of the
chapter so that detailed studies can be performed.
Although these topics can be integrated into other chapters when they are
encountered, many of the topics appear multiple times and at different locations.
The goal is to provide a convenient and quick reference point to the readers by
consolidating these subjects in a separate chapter.

2.1 Mathematics Fundamentals

2.1.1 Fourier transform and convolution
The one-dimensional (1D) Fourier transform of a function f(x) is defined as


F (u )   f ( x)e  j 2 ux dx , (2.1)


where j  1 . In this equation, e  j 2 ux  cos 2ux  j sin 2ux . Substituting this
expression into Eq. (2.1), the 1D Fourier transform can also be expressed as

 
F (u )   f ( x) cos 2uxdx  j  f ( x)sin 2uxdx . (2.2)
 

When f(x) is real, the real part of Eq. (2.2) is an even function of frequency u, and
the imaginary part is an odd function of frequency u. Based on the definition of a
complex conjugate, the Fourier transform of any real function possesses the
following property:

F (u )  F * (u ) , (2.3)

23
24 Chapter 2

where * denotes a complex conjugate. This property is often called Hermitian

symmetry. The Fourier transform of the measured CT projections always
possesses Hermitian symmetry, because the projection measurement p(x) is
always a real function. An interesting special case occurs when f(x) is a real and
even function of x. It can be shown that its Fourier transform F(u) is also a real
and even function of u.
The inverse Fourier transform is defined as


f ( x)   F (u )e j 2 ux du . (2.4)


One can show that Eqs. (2.1) and (2.4) form a transformation pair. Several
properties of the 1D Fourier transform pair can be readily obtained from these
definitions. Interested readers can refer to several textbooks for details.1–3 These
properties will be presented below without proof:
(1) Linearity: If f ( x)  af1 ( x)  bf 2 ( x) , then F (u )  aF1 (u )  bF2 (u ) . Here,
F1(u) and F2(u) are the Fourier transforms of f1(x) and f2(x), respectively.
This property indicates that the linear combination of two functions leads
to a linear combination of their Fourier transforms.
(2) Scaling: The Fourier transform of f(ax) is (1/a) F(u/a).
(3) This property states that scaling the size of a function leads to a
compression in frequency and an amplification in magnitude. A simple
example is shown in Fig. 2.1. In the discrete implementation of the
Fourier transform (called the discrete Fourier transform or DFT), one
technique that is often used to increase the sampling density of a spatial
(or time) domain signal is to perform zero padding in the Fourier
domain. The process is as follows: First, the Fourier transform of the
original signal is performed. When the original signal contains N
samples, its corresponding DFT also contains N samples. If the original
sampling density must be increased by a factor of K, we pad (K–1)N
zeroes to the Fourier transform before the inverse Fourier transform is
performed. Thus, the original signal is expanded to KN samples, as
shown in Fig. 2.2.
(4) Shift property: The Fourier transform of f(x–x0) is e  j 2 ux0 F (u ). This
indicates that a shift of a function in the spatial domain is equivalent to a
phase shift in the Fourier domain. This property is quite useful in data
resampling or interpolation. It is often preferred over spatial domain
interpolations. Figure 2.3 shows an example of shifting the original data
by 3.2 sampling intervals.
(5) Derivative: The Fourier transform of the derivative of f(x), df(x)/dx, is
j 2uF (u ).
 
 f ( x) dx  
2 2
(6) Conservation of energy: F (u ) du .
 
Preliminaries 25

Figure 2.1 Scaling the size of the original function in the spatial domain leads to the
compression of frequency and amplification of magnitude in the frequency domain.

Figure 2.2 Illustration of double sampling density in the spatial domain by zero padding in
the frequency domain.
26 Chapter 2

Figure 2.3 Data resampling with phase shift in the Fourier domain. Gray line: original
samples; black line: shifted by a 3.2 sampling interval.

Another useful property of the Fourier transform is its application to the

convolution process. The convolution of two functions, f1(x) and f2(x), is defined by


g ( x)  f1 ( x)  f 2 ( x)   f1 ( x ') f 2 ( x  x ')dx ' . (2.5)


The convolution operation can be conceptually performed by a sequence of

operations. First we flip the function f2(x′) at x′ = 0 to arrive at a function f2(x′),
as shown in Fig. 2.4. The function is then continuously shifted along the x axis by
an amount x to obtain f2(xx′). The area under the product of the two functions,
f1(x′)f2(xx′), is assigned as the convolution of the two functions at location x.
If we take the Fourier transform of Eq. (2.5), we obtain

G (u )     f1 ( x' ) f 2 ( x  x' )dx'  e j 2 ux dx  F1 (u ) F2 (u ) .

 
(2.6)
 
  

Equation (2.6) states that the Fourier transform of the convolution of two
functions is equal to the product of the Fourier transforms of the two functions
taken separately.
If the original spatial domain function f(x, y) is 2D, the Fourier transform can
be easily extended to 2D by the inclusion of the Fourier transform in the second
dimension:

 
F (u , v)    f ( x, y )e  j 2 ( ux  vy ) dxdy . (2.7)
 
Preliminaries 27

Figure 2.4 Illustration of the convolution process between functions f1(x) and f2(x). f2(x) is
first flipped and shifted with respect to its x axis. The resulting convolution represents the
area under the product of the two functions.

The inverse Fourier transform can be defined in a similar manner:

 
f ( x, y )    F (u, v)e j 2  ( ux  vy ) dudv . (2.8)
 

All properties presented for the 1D Fourier transform can be extended to the two
dimensional case.

2.1.2 Random variables

For a random variable x′, its cumulative distribution function P(x) is defined as a
function whose value at each point x is the probability that a random observation
of x′ will be less than or equal to x. Mathematically, it is represented by

Pr  x'  x  P ( x) . (2.9)

The probability density function p(x) is defined as

dP ( x)
p ( x)  . (2.10)
dx

Clearly, the cumulative distribution function P(x) is simply the integration of the
probability density function p(x) from  to x:
x
P ( x)   p( x') dx' . (2.11)


The probability that the random variable x′ will fall in any interval between a and
b is the area under the probability density function p(x) from a to b:
b
Pr(a  x'  b)   p ( x)dx  P (b)  P (a) . (2.12)
a
28 Chapter 2

The pictorial representation is shown in Fig. 2.5. The shaded area under the
probability density function [Fig. 2.5(a)] equals the difference between two
points on the cumulative distribution function [Fig. 2.5(b)]. Because a random
observation x must have some value, we have

P ()   p( x)dx  1 .


The expected value or the mean of a random variable x is defined as


E ( x)  x   xp( x)dx . (2.13)


The variance of a random variable 2 is defined as the second moment of the

probability density function with respect to the mean:

 2  E ( x  x ) 2   ( x  x ) 2 p ( x)dx . (2.14)


 is often called the standard deviation of x. One widely used distribution

function is the normal distribution (also called the Gaussian distribution), which
has the probability density function

1 2
/ 2 2
p ( x)  e  ( x  ) , (2.15)
 2

where  is the mean and  is the standard deviation. This distribution is of

particular interest in both applied and theoretical statistics. One application is the
estimation of confidence interval defined by the following probability:

Figure 2.5 Illustrations of (a) the probability that random variable x falls in an interval from
a to b using probability density function p(x), and (b) cumulative distribution function P(x).
Exploring the Variety of Random
Documents with Different Content
 14 TOOLS FOR THE WORK foot on a stool. This relieves the
back of much bending and back-aches are less frequent. TOOLS FOR
CUTTING For cutting ample provision should be made. The table
upon which material is laid preparatory to cutting should be perfectly
smooth and of sufficient dimensions to permit the largest patterns to
be laid out entirely. For such pieces as a trained or kilted skirt this is
often not practicable but the worker must then exercise her most
careful ingenuity and judgment. The home dressmaker is often led
into the most expensive mistakes by cutting out on the floor or bed.
We can not be too urgent against such a proceeding. Therefore in
our sewing-room there must be a table at least four feet long and
three feet wide for cutting out. The best table is the substantial
ordinary one of wood, with a smooth, even surface and square
corners. If this is not available, one of the folding tables of at least
that size is reasonably convenient. They are certainly entirely
satisfactory for cutting but a more substantial one is better for
pressing, and there is no reason why the same table should not be
used for both purposes. When working at the table, either cutting,
basting or pressing, one should sit, not stand. When sitting one can
easily reach across three feet of space and two feet on either side.
This saves much tiresome bend 
 TOOLS FOR THE WORK 15 ing of the body and wearied
feet and legs at the end of the season of sewing. This table in the
sewing-room will be used for all sorts of other purposes beside
cutting out the original garment. But an ordinary lap-board should
also be provided. It will often be used when cutting small pieces like
collars and facings and when putting flounces or pleatings on the
bottoms of skirts it will be found indispensable. Perhaps the most
important tool in the sewing-room is a pair of shears. In
dressmaking much depends in the beginning upon clean, evenly cut
edges. In basting or stitching seams the eye is easily, although often
unconsciously, influenced by the outlined edges and where they are
rough or uneven the seam is wavering and inaccurate. Clean cut and
even edges also influence the stitcher to finish the seams in a neater
manner. She will without thinking execute that part of the work with
greater precision. THE SHEARS Long, slender and sharp blades
should characterize the shears used. Never attempt a garment with
dull, rough or rusty ones with a loose rivet, nor with dainty little
embroidery scissors. Use shears of good metal not less than eight
inches long with bent handles, with well sharpened ends and riveted
just tight enough that no resistance will be noticeable when opening
and closing them. Take care of them when they are not in use. Keep
them from dampness and do not let
 1 6 TOOLS FOR THE WORK them fall as that will often
impair the nicety of their adjusted blades. TOOLS FOR SEWING
Ever}' sewing room should have a machine that is light running and
capable of sewing from the heaviest to the lightest fabrics. It should
be kept well oiled and in order. It should also be kept perfectly clean.
Do not let it become clogged up with dust or old oil. A little kerosene
will clean all this away, when it in turn must be wiped off and the
machine properly oiled with the best machine sperm oil. It is
pleasant to have all the attachments invented with the machine, but
for dressmaking one must have the hemmers, the tuckers and the
gatherer. At the side of the machine provide a scrap bag in which
can be stowed away useless pieces, and thus save the bother of
picking them off the floor later. Two bags are not too many ; one for
absolutely useless pieces, the other for larger scraps that may be
found useful later on. There are many minor details of the sewing-
room's furnishings which will gradually be provided and accumulated
as the sewer prosecutes her work. But her sewing basket must be
well stocked to commence. It should be a strong basket or box
sufficiently large to meet all ordinary requirements. It must contain
needles of all sizes and chosen from those of good quality. Those
with egg-shaped eyes are the easiest to thread. They should have
long taper points, as it is impossible to sew on stiff material with a
conical-pointed
 TOOLS FOR THE WORK 17 needle without pricking one's
fingers at every stitch. Tn every case the needle must be large
enough to draw the thread through the fabric without the least
effort. There must be pins in plenty, cotton thread and spool silk in
both white and black with a good large spool of coarse basting
cotton. There must be an emery bag, which should be home made,
as those bought in merchandise stores are generally filled with
anything rather than good filings. There should be a square of hard
white soap. A linen seam is a difficult seam to sew by hand or
machine. If you pass the soap over it before commencing, all the
difficulty is instantly removed. There must be in this basket a well
fitted thimble. Two thimbles are even better, as it is very provoking
to be forced to stop and hunt a thimble that has momentarily
disappeared just when you most need it. They must exactly fit the
finger. It is ver}' uncomfortable to work with a thimble which turns
on the finger; if the extra space is filled with paper or rag, it renders
the thimble too heavy and the thread is liable to catch. There must
be a lead pencil and a good tape line and a pair of button-hole
cutters with a gauge are a great convenience. Equally pleasant to
have at hand are a sharp steel punch or chisel and a perforated
bodkin for drawing a cord or tape through clings or hems. This
basket should be provided with a cover to keep its tools free from all
but ordinary dust. A piece of silk as long as the basket around its top
and about
 1 8 TOOLS FOR THE WORK six inches wide makes a good
cover. Join its ends and sew one of its edges to the basket top. Then
run a casing in the other edge of the silk and pass a drawing-string
through it. Thus the cover may be opened and closed at pleasure. A
medium size leaded pincushion is extremely convenient for use in
pinning the work. Pinning the cloth to the knee is very poor policy,
on account of the fatiguing stoop it causes. When the leaded
pincushion is at hand, the cloth is so easily attached, and a woman
who has become accustomed to one will never be without it. They
are easily made, the heavy piece of lead being securely hidden in
the sawdust used to fill the cushion. TOOLS FOR FITTING There is
an absolute necessity for a mirror in which the entire figure may be
surveyed. Even in fitting a bodice or short wrap the general effect
should be the thing considered. Their lengths can only be decided
correctly in reference to the entire length of the figure. The best
mirror is one that swings in a frame. Such a one in a dressing-case is
very convenient, and there are less expensive ones called easel
mirrors. If these are beyond the means at hand, place any ordinary
mirror on the floor at an angle Avhere a view of the entire figure can
be obtained. The gown must harmonize with the wearer. In other
words, you must adapt the materials to yourself, and this can only
be done by seeing yourself as others see you.
 TOOLS FOR THE WORK 19 The next best help one can
have for this purpose is an adjustable wire form. The forms that may
be adjusted to correspond with neck, waist and bust measure are
few and expensive, if they are of any value at all, but there are skirt
figures that may be bought for a couple of dollars — they are a
capital investment. Draperies may be adjusted with the greatest
ease when they are used. TOOLS FOR PRESSING Among the most
importanttools are aflatiron and some means of heating it. In this
day of steam radiators there is often no such means at hand. There
have been many inventions given an aggrieved and credulous public,
such as attachments to gas jets, alcohol burners, etc., but the best
thing is a little kerosene stove. There is no reason why it should be
dangerous ; nothing but the grossest carelessness makes it so, and
it heats a flatiron in a few moments. Remember there is everything
in the proper pressing of the garment. This applies to the skirt
seams, the hems, the bodice seams and facings and to the sleeves
as well. A good investment is a couple of press boards, one for skirts
and a smaller one for bodices and sleeves. Any carpenter will make
them and the cost is but a trifle, while the convenience will more
than repay an even greater expenditure. A skirt board should be
about forty-four inches long, the length of an ordinary skirt and nine
inches wide. The sleeve board should be five inches in width and
twenty-seven inches long.
 20 TOOLS FOR THE WORK Give them each at least one
thickness of flannel and add a cotton cover. With these and a couple
of hot flat-irons and plenty of strength, the homemade dress may be
made a very presentable affair 
 CHAPTER II DRESSMAKERS' FINDINGS LININGS FACINGS
— WADDING AND CANVAS EASTENINGS WHALEBONES AND
CASINGS — SHIELDS — YOKES LININGS Those who undertake to
learn the trade of dressmaking find that silesia, braid and canvas
represent the A B C's of the art. The novice will do well to try every
pattern or idea in the smooth, firm but inexpensive silesia. It is the
amateur dressmaker who frequently makes the mistake of plunging
at once into all the perplexities of silks, velvets and furbelows. Such
experiments are pretty sure to result disastrous!}'. She is liable to
waste a great deal of material and to expend so much time and
patience in several thousand times too many stitches that she gives
up trying to sew at all. How much better to begin with the anatomy
of the dress. Master the fit in the linings, which is reall}essential to
the successful fabrication, and then success awaits further along the
line of experience and a garment is achieved of which the maker
may well feel proud.
 22 DRESSMAKERS FINDINGS "But what kind of linings
should we use?" do you ask? In the first place, don't use old linings.
Such a course is not economy at all. For with linings that have lost
their firmness and body no waist can be made to fit, no skirt made
to hang properly. The same may be said of whalebones, hooks and
eyes, braids and sometimes of buttons, although the latter can again
be used more frequently. Yet cloth buttons are usually worn shiny
and metal ones are tarnished. Have plentiful and good lining
materials. Ail dress fabrics, except some cottons, require a
foundation to protect them from strain ; cloths and woolens stretch,
laces and sheer woolens tear and silks cut and split without a good
under foundation. Silk, cambric and silesia are each in turn used.
Each has its recommended qualities and each again is entirely
unsuited to certain purposes. Silk linings are by all odds the most
elegant and comfortable. To be sure they are a little expensive in the
first outlay but they wear so well and are so light in weight, perfect
in fit and generally elegant in appearance, they are favored by our
leading and best modistes. For dresses intended for general wear
the soft fine French cambric can not be too highly recommended.
Silesia is also an admirable lining material for almost any dress and
for all its parts. For the waist and its sleeves it is unsurpassed, and
for the skirts of dresses as well.
 DRESSMAKERS' FINDINGS 23 The purpose of a dress lining
is twofold. It is necessary as a neat finish and as a foundation. Some
ladies will tell you they use good, perhaps the best linings for the
waist, not quite so good for the sleeves, and that anything Avill do
for the skirt of a dress. These are mistaken economies. There is just
as much strain and wear on the sleeves of a dress as upon the waist
and the linings should be the same. The only difference which may
be made is for the skirt, where a lighter and less strong material can
be used, as there it is only the neat finish and protection, and really
no strain upon it. FACINGS Beside the linings proper for skirts there
are several accessories which must be provided, that must be
classed with them. The facing for skirts comes first among them.
There is quite a diversity of opinion as to which is preferable of some
three or four which are all in general use. Perhaps we are safe in
saying a majority of professional dressmakers face their skirts first
with crossbarred crinoline, afterward covering it with alpaca. This
certainly makes a soft finish to the skirt but also one which is thick
and clumsy and one which is extremely addicted to gathering and
holding dust. Equally objectionable for the same reason is the use of
canvas covered with alpaca. And there is yet another objection we
may urge, and that is such facings do not
 24 DRESSMAKERS' FINDINGS wear well and are very hard
upon the shoes of the wearer. Ladies who have their skirts finished
in this manner find themselves in a very short time forced to trim off
rags and tatters or look untidy with them hanging around their feet,
and, if the dress is at all durable, the facing must be renewed at
least twice during its existence. Very much better, for durabilit)^,
cleanliness and soft finish is the cotton padding. When it is used, the
work of facing a skirt is greatly simplified, it being easily put on as
will appear in our extended directions for facing a skirt in a following
chapter. Every skirt must be finished with a braid or a velveteen
band. Pleated braids are sometimes used when a little extra finish is
required, but when an ordinary braid is used it should be one of the
best and then it will not be a narrow one, but wide enough to cover
all edges. WADDING AND CANVAS Findings for the waists of dresses
are more complex. First is the lining proper, which as we have
suggested before should be either silesia, cambric or silk. Whichever
is used, let it be the color of the dress unless it be a black dress.
Black lining should never be used for waist or sleeves and dark
gra3MS better for black skirts too, still black may be sometimes
employed for them. It is quite likely to soil the underwear. There are
many good silesias woven black on one side and dark gray
 DRESSMAKERS' FINDINGS 25 on the other, which will be
found useful for some purposes. There are but few women — or
men either — whose forms do not require some "building up". Many
dressmakers place a laj^er of wadding between the lining and the
dress fabric reaching from the shoulders to the top of the darts. This
certainly gives a smoothness over the bust, that is desirable, still it
greatly increases the warmth. One thickness of light quality of
canvas accomplishes the same end and is cooler. The tops of the
sleeves, from the shoulder to the elbow should be given the same
treatment when the form is not plump and bones make
unevennesses. When the form is inclined to be too large below the
waist, one thickness of canvas placed between the lining and the
dress fabric below the tops of the darts assists in keeping a basque
or polonaise in shape. FASTENINGS When buttons are used, the
button-holes are a serious question for the dressmaker and must be
neatly worked with good twist, or the garment is not beautiful even
when handsome fabrics are used. In Chapter IX. will be found full
instructions concerning button-holes. When hooks and ej^es are
used for closing, the amateur dressmaker should ask for bent hooks,
as those slightly bent near the point stay fastened. Otherwise it is
necessaray to sew them on alternately, which makes them very
inconvenient for closing. Small rings used instead of eyes on the
outer part of dress waists
 26 DRESSMAKERS' FINDINGS should be covered with silk in
button-hole stitches. The very large hooks and eyes used as cloak
and wrap fastenings are also excellent for keeping up a heavy skirt :
four of the hooks being set on the waist just below the belt, — two
on the seam joining the back and side-forms and one on each
under-arm seam ; the eyes are placed on the skirt band to
correspond and the wearer hooks them before fastening the inside
belt of her dress. WHALEBONES AND CASINGS The use of
whalebones is an important item to consider. Most ladies require
every seam stayed. If the seams were curved absolutely perfectly it
would not be necessary, but this art is seldom encountered. When
stays are needed, use the best wht^lebones only. Nothing else
wears so well nor gives the proper elasticity. Horn, tin, steel and
rubber have all been used, and either rust, break or twist
unpleasantly. Galloon must be provided for casings in which to run
the stays. Casings of lining material make clumsy seams. Ribbon for
binding the edges of the waist and sleeves seams finish the waist in
the most acceptable manner. With loops to go in each armseye of
the same by which to hang up the waist, a neat finish is given.
However, this ribbon binding is repudiated by some ladies who
delight in the snuggest fits. They insist the binding of edges draws
the seams and demand rather they shall be loosely top sewed and
pressed.
 DRESSMAKERS' FINDINGS 27 In any case a sufficient
length of binding tor an inside belt must be provided to attach at the
back seams at the waist-line to take the strain off the front. It is
usually fastened in front by medium sized hooks and eyes. SHIELDS
Dress shields must be provided, large ones in the armseyes, and
ladies who perspire profusely use small ones in the sleeves at the
elbow curve. Some ladies abominate cheap shields for dresses. They
buy the best and take them out at intervals and wash them in clean
soap suds. This is not a bad idea. It is also good practice to
purchase cheaper ones and change then often. None are perfect and
neatness requires they should be changed as soon as the slightest
odor can be detected. In any case it is good policy to buy shields by
the half dozen pair and so have them always at hand. A RESUME For
a medium sized woman's ordinary costume, consisting of a walking
length skirt and a basque with coat sleeves, the following findings
will be found necessary : If silesia is used five and one-half yards for
the skirt, one and one-half yards for the waist and one yard for the
sleeves, or a total of eight yards. If ordinary silk is used ten yards
will be found sufficient. For the skirt facing one yard of canvas, with
one yard of alpaca, or one yard of padding alone, if the latter is
preferred. Add to these three long whalebones, one boltofbraid,
 28 DRESSMAKERS' FINDINGS one bolt of ribbon to bind
seams, one piece of galloon for whalebone casings, one card of
hooks and eyes or one and one-half dozen of medium sized button,
two spools of twist, one of sewing silk and one spool of basting
cotton. The findings required for jackets and outside wraps of all
kinds will be fully treated in chapters devoted to such garments. To
conclude and at the same time be explicit we would say do not buy
cheap findings. They do not pay. Do not use old linings or
whalebones. It is false economy. Findings do not show in one sense
of the word, but they tell every time in wear and general comfort.
YOKES A word as to keeping the whole gown in shape and doing
away with ' closet wrinkles. " Buy a wooden or wire yoke such as
tailors use for suspending coats, and after turning the dress wrong
side out fasten the waistband and slip the whole over the yoke. It
spreads the folds of the drapery, preventing them from being
crushed into an unshapely mass, and keeps the foundation from
stretching down at the seams. These yokes are inexpensive, and
may be found at any drygoods store.
 CHAPTER III HOW TO MAKE DRESS SKIRTS THE MODERN
GORED SKIRT — THE FOUNDATION SKIRT SKIRT DRAPERIES THE
KILT SKIRT THE TRAINED SKIRT REEDS THE MODERN GORED
SKIRT The size and st3de of skirts vary with each edict of fashion.
However, tliere are two general styles upon which the changes are
rung, the short walking skirt and the trained skirt. Walking skirts
may be divided into the round full skirt and the gored skirt ; the
former is a mere matter of straight seams, a hem, and a gathered
top on a band, tliat anyone can make. But the shapely gored skirt is
a different thing. The modern gored skirt is the work of an artist.
Some one has said "the making of one is like singing an old ballad. A
novice may sing a grand operatic aria but it takes a genius to sing
'Comin' thro' the Rye,' and to make a gored skirt. Both are most
simpte in design but most difficult of construction." There are three
things which go to make a perfect skirt ; first an accurate cut,
second a neat finish and third a thorough pressing. 29
 30 HOJV TO MAKE DRESS SKIRTS The walking skirt most
used is rather narrow in its proportions. The only skirt less ample
was the one which showed its back breadth gored at the top to fit as
close as the present front and side-gores do. Z^Inches. 7 Inches. 18
Fnches. < CD 18 Inches. THE FOUNDATION SKIRT Every skirt
should be made with a perfectly fitted foundation. It should be of
easy walking length, properly gored and not too wide. It is usually
cut with one front-gore, two side-gores and a straight back breadth.
For a lady of medium size who will measure twenty-four inches
around the waist the following are the correct measurements for
each part. The front gore will be fifteen inches wide at the top with a
dart two inches wide allowed for on each side of the mid 
 HOU^ TO MAKE DRESS SKIRTS 31 die of the front. It is
forty inches in length in front but is sloped to forty-one inches in
length at the sides and at the bottom is twenty-two inches wide. The
sidegores are each forty-one inches in length at their front sides and
forty-two where they are joined to their back breadth. They are
seven inches wide at the top with two inches allowed for darts and
are gracefully curved to sixteen inches in width at the bottom. The
back is just one yard in width its entire length, which is fortytwo at
its sides sloping to forty-four inches in the middle. When quarter-
inch-wide seams are taken, the back has all its fulness gathered into
five inches, which is the correct proportion. For home dressmakers it
is much the best plan to use a good pattern for this skirt, as no rule
given in figures can explain the graceful curves which each gore
should show to give the best effect. A skirt that hangs ill is always
dowdy looking. No matter how elaborately draped or trimmed, a
badly shaped foundation skirt ruins all. The materials used for the
foundation skirt vary with the fabrics employed for drapery. We will
for example suppose we require one for a dress of cashmere, serge,
cloth or silk. In which case the skirt should be of lining silk the same
color or a shade harmonizing with it. However, good silk linings
(poor ones are an abomination) are expensive, and there are other
materials which make admirable linings. By some sateen and silesia
are preferred and the latter cannot be
 32 HOW TO MAKE DRESS SKIRTS too highly recommended
except in the matter of weight. It is always heavier than any other
skirt lining, Yet it is not too weighty. French cambric is also an
excellent material for skirts. Whatever the material chosen, cut the
front and back breadths on a lengthwise fold of the goods and the
sidegores with their front edges on straight edges of it, The seams
of this foundation skirt are of course sewed up separately from the
outside or draped portions. They may be sewed so that the smooth
sides of the seams are on the underside of the skirt and their rough
edges next the draperies. The facings in that case should all be cut
to fit the skirt after its seams are closed. For heavy skirts it is better
to slash the front-gore for four inches at least on its lower edge to
give greater freedom in walking. This is a great saving to shoes,
whose leather over the instep is otherwise often worn through while
the other parts are intact. A tight braid is very wearing in that
respect. The foundation of most skirts is faced on the upper side
under the draperies, four inches with material of the drapery. This
facing is laid on each portion of the skirt with its upper edge turned
under and stitched down on the lining, before the skirt seams are
sewed. Begin to sew each seam (and there will be four in all) at the
top, allowing all unevenness to fall at the bottom. Be careful not to
stretch any bias edges. If you are an entire novice it is the best plan
to both pin and baste the seams before stitching them. It will often
save hours
 I/Of!^ TO MAKE DRESS SKIRTS 33 of worry and ripping.
When the seams are stitched, press each one down flat, turning
them alwa3S toward the back, this method is better than laying the
seams open. In either case the edges should be overcast or top-
sewed, and thoroughly pressed, with a warm iron. Then lay the skirt
folded down the middle at the front and back portions so that the
corresponding seams are together, on a table. Let them be even at
the top, and then pare off any unevenness at the bottom. The
matter of inside facing is a very important one and also one upon
which there is almost as much diversity of opinion as there are
dressmakers. However, when we resolve the question down to what
is really required, it is a simple thing to decide how to face your
walking skirts. A facing is required to neatly finish the bottom of a
skirt, and a facing is required to obviate any luipleasant clinging
about the limbs when walking. What will best cover both
requirements is what we want. Some dressmakers contend that this
is, first a fiveinch piece of crinoline or canvas and tlien a piece of
alpaca. Another will demand a hem lined for twelve inches with
horse hair cloth, still another demands canvas or buckram in like
width. In most things the simplest way is the best way, and we
believe it is especially true in putting on a skirt facing. Much practice
and experience convince us that skirt padding used alone is the best
thing and it certainly is the cheapest. In some localities this material
is
 34 HO]V TO MAKE DRESS SKIRTS called by other names.
But it is a moderate weight fabric glazed on one side and likecanton
flannel on the other. It possesses sufficient stiffness but at the same
time it is pliable and does not render even the lightest skirt
ungraceful. It is also a fabric which sheds the dust, and when it is
required may be wiped off with a damp cloth. Again, its durability is
a great recommendation: it will last as long as any skirt. The neatest
manner of putting on a facing is, after the lower edge of the skirt
has been properly pared, tocut the facing eight inches deep and to
fit it exactly. Then seam the skirt and facing's lower edges together
on the inside and turn. The upper edge of the facing should be cut
in fine notches and just below them stitch it down on the skirt lining.
Then finish the smooth edge with the customary braid. Always use
the best braid, it receives the hardest usage of any portion of the
skirt. It should be dipped in water and allowed to thoroughly dry
before it is stitched on the skirt. Otherwise, even the "warranted not
to shrink" braid will draw up on the skirt foundation. The above
directions are ample for finishing the bottom of a skirt when you
have it cut over a perfect pattern and the padding is used. When a
skirt design is used that has not the proper spring given its gores,
other resources must be called upon. Every woman knows how
disagreeable it is to walk, when at every step the foot is pulled back
by the skirt. This is obviated by
 HOll' TO MAKE DRESS SK/RTS 35 twice slashing for four
inches the foot of the skirt's front-gore and covering the slashes with
pleating. Some dainty imported dresses for wear in the house are
not bound with braid, but are simply faced with silk. Attached to this
facing inside the skirt is a pinked frill of silk instead of the»lace
balayeuse sometimes formerly employed. To protect the extreme
lower edges of skirts which are of extra length, many expedients are
resorted to. At the shops many new"protectors" are found and at
once recall those used for a similar purpose some years ago. The
new ones, of course, have the advantage of all the improvements of
progression. Several kinds are of buckram, which is widely bound
with rubber cloth and either pleated to a binding which curves the
protector to the shape of the train, or is sewed to a j^oke-like piece
shaped like the bottom of the skirt at the back. These protectors
extend across the sweep of the skirt only. Others, however, are in
the nature of a facing. The facing material is a strip of blacksilesia
orserge, and is long enough to pass completely round the skirt
foundation. For a sufficient distance to extend around the sweep at
the back this facing is again faced with a stirp of rubber cloth
securely stitched on, and as this comes next the surface on which
one is walking, it prevents the edges of the skirt and its foundation
becoming worn, soiled or damp. A braid is .also supplied which has a
rubber strip Stitched to it for a sufficient portion of its length to
 36 JJOIF TO MAKE DRESS SKIJRTS protect the sweep of
the skirt ; and folded rubber strips, which look like pipings, are made
to extend entirely round the skirt and are very satisfactory in the
capacity of protector. Most of these protectors may be found in gray,
brown and black. When the lower edge of the foundation skirt is
completed, it should also be finished at the top before its draperies
are ad justed. A placket-opening must be provided either at the back
or at one side. This is done by making an opening either in a seam
or by cutting the material the depth of ten inches from the top. Face
the upper or INSIDE OF FINISHED SKIRT Overlapping side with a
two-inch-widestrip of the material of the draperies. Then sew in a
seam to the opposite side a double flap of the same material and let
it extend under the faced side, tacking it fast at its lower end to the
opposite facing. A pocket should then be put in along the second
right-side seam. It may be made of either silk or siiesia and must be
faced with the material of the dress each side of its opening. When
these preliminaries are completed the entire skirt must be most
thoroughly pressed on the long skirt board with a hot iron.
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookname.com