Bản chất của hình ảnh y sinh học (Phần 1)

Figure 1.2shows thermal images of a patient with benign brocystsand a patient with breast cancer the local increase in temperature due to atumor is evident in the latter case.. Example:

The Nature of Biomedical Images

The human body is composed of many systems, such as the cardiovascularsystem, the musculo-skeletal system, and the central nervous system Eachsystem is made up of several subsystems that carry on many physiological pro-cesses For example, the visual system performs the task of focusing visual orpictorial information on to the retina, transduction of the image informationinto neural signals, and encoding and transmission of the neural signals to thevisual cortex The visual cortex is responsible for interpretation of the imageinformation The cardiac system performs the important task of rhythmicpumping of blood through the arterial network of the body to facilitate thedelivery of nutrients, as well as pumping of blood through the pulmonary sys-tem for oxygenation of the blood itself The anatomical features of the organsrelated to a physiological system often demonstrate characteristics that reectthe functional aspects of its processes as well as the well-being or integrity ofthe system itself

Physiological processes are complex phenomena, including neural or monal stimulation and control inputs and outputs that could be in the form

hor-of physical material or information and action that could be mechanical,electrical, or biochemical Most physiological processes are accompanied by

or manifest themselves as signals that reect their nature and activities Suchsignals could be of many types, including biochemical in the form of hor-mones or neurotransmitters, electrical in the form of potential or current, andphysical in the form of pressure or temperature

Diseases or defects in a physiological system cause alterations in its health, and general well-being of the system A pathological process is typ-some respects from the corresponding normal patterns If we possess a goodunderstanding of a system of interest, it becomes possible to observe the cor-responding signals and features and assess the state of the system The task

nor-is not dicult when the signal nor-is simple and appears at the outer surface ofthe body However, most systems and organs are placed well within the bodyand enclosed in protective layers (for good reason!) Investigating or probingsuch systems typically requires the use of some form of penetrating radiation

or invasive procedure

2 Biomedical Image Analysis

1.1 Body Temperature as an Image

Most infections cause a rise in the temperature of the body, which may besensed easily, albeit in a relative and qualitative manner, via the palm ofone's hand Objective or quantitative measurement of temperature requires

an instrument, such as a thermometer

A single measurementf of temperature is a scalar, and represents the mal state of the body at a particular physical location in or on the bodydenoted by its spatial coordinates (x y z) and at a particular or single in-stant of timet If we record the temperature continuously in some form, such

ther-as a strip-chart record, we obtain a signal ther-as a one-dimensional (1D) tion of time, which may be expressed in the continuous-time or analog form

func-as f(t) The units applicable here are oC (degrees Celsius) for the ature variable, and s (seconds) for the temporal variable t If some meanswere available to measure the temperature of the body at every spatial posi-tion, we could obtain a three-dimensional (3D) distribution of temperature as

temper-f(x y z) Furthermore, if we were to perform the 3D measurement at everyinstant of time, we would obtain a 3D function of time as f(x y z t) thisentity may also be referred to as a four-dimensional (4D) function

When oral temperature, for example, is measured at discrete instants oftime, it may be expressed in discrete-time form as f(nT) or f(n), wheren

is the index or measurement sample number of the array of values, and T

represents the uniform interval between the time instants of measurement Adiscrete-time signal that can take amplitude values only from a limited list ofquantized levels is called a digital signal this distinction between discrete-timeand digital signals is often ignored

If one were to use a thermal camera and take a picture of a body, a dimensional (2D) representation of the heat radiated from the body would

two-be obtained Although the temperature distribution within the body (andeven on the surface of the body) is a 3D entity, the picture produced by thecamera is a 2D snapshot of the heat radiation eld We then have a 2D spatialfunction of temperature | an image | which could be represented asf(x y).The units applicable here areoCfor the temperature variable itself, andmm

(millimeters) for the spatial variables xand y If the image were to be pled in space and represented on a discrete spatial grid, the correspondingdata could be expressed as f(m$x n$y), where $x and $y are the sam-pling intervals along the horizontal and vertical axes, respectively (in spatialunits such asmm) It is common practice to represent a digital image simply

sam-as f(m n), which could be interpreted as a 2D array or a matrix of values

It should be noted at the outset that, while images are routinely treated asarrays, matrices, and related mathematical entities, they are almost alwaysrepresentative of physical or other measures of organs or of physiological pro-

cesses that impose practical limitations on the range, degrees of freedom, andother properties of the image data.

Examples: In intensive-care monitoring, the tympanic (ear drum) ature is often measured using an infrared sensor Occasionally, when cathetersare being used for other purposes, a temperature sensor may also be intro-duced into an artery or the heart to measure the core temperature of thebody It then becomes possible to obtain a continuous measurement of tem-perature, although only a few samples taken at intervals of a few minutesmay be stored for subsequent analysis Figure 1.1illustrates representations

temper-of temperature measurements as a scalar, an array, and a signal that is a tion of time It is obvious that the graphical representation facilitates easierand faster comprehension of trends in the temperature than the numericalformat Long-term recordings of temperature can facilitate the analysis oftemperature-regulation mechanisms 15, 16]

func-Infrared (with wavelength in the range 3 000;5 000nm) or thermal sensorsmay also be used to capture the heat radiated or emitted from a body or apart of a body as an image Thermal imaging has been investigated as apotential tool for the detection of breast cancer A tumor is expected to bemore vascularized than its neighboring tissues, and hence could be at a slightlyhigher temperature The skin surface near the tumor may also demonstrate ao

Chavebeen measured between surface regions near breast tumors and neighboringtissues Figure 1.2shows thermal images of a patient with benign brocystsand a patient with breast cancer the local increase in temperature due to atumor is evident in the latter case Thermography can help in the diagnosis

of advanced cancer, but has limited success in the detection of early breastcancer 17, 18] Recent improvements in detectors and imaging techniqueshave created a renewed interest in the application of thermography for thedetection of breast cancer 19, 20, 21, 22, 23]

Infrared imaging via a telethermographic camera has been applied to thedetection of varicocele, which is the most common cause of infertility inmen 24, 25, 26] In normal men, the testicular temperature is about 3;4oC

below the core body temperature In the case of varicocele, dilation of the ticular veins reduces the venous return from the scrotum, causes stagnation ofblood and edema, and leads to increased testicular temperature In the exper-iments conducted by Merla et al 25], a cold patch was applied to the subject'sscrotum, and the thermal recovery curves were analyzed The results obtainedshowed that the technique was successful in detecting subclinical varicocele.Vlaisavljevi*c 26] showed that telethermography can provide better diagnosticaccuracy in the detection of varicocele than contact thermography

tes-4 Biomedical Image Analysis

© 2005 by CRC Press LLC

6 Biomedical Image AnalysisThe thermal images shown in Figure 1.2 serve to illustrate an importantdistinction between two major categories of medical images:

 anatomical or physical images, and

 functional or physiological images

The images illustrate the notion of body temperature as a signal or image.Each point in the images in Figure 1.2 represents body temperature, which

is related to the ongoing physiological or pathological processes at the responding location in the body A thermal image is, therefore, a functionalimage An ordinary photograph obtained with reected light, on the otherhand, would be a purely anatomical or physical image More sophisticatedtechniques that provide functional images related to circulation and variousphysiological processes are described in the following sections

penetration of light through a large organ such as the breast and solid lesions however, the technique has had limited success in distin-guishing malignant tumors from benign masses 18, 28, 29]

Transillumina-1.3 Light Microscopy

Studies of the ne structure of biological cells and tissues require signi cantmagni cation for visualization of the details of interest Useful magni cation

of up to1 000 may be obtained via light microscopy by the use of tions of lenses However, the resolution of light microscopy is reduced by thefollowing factors 30]:

combina- Diraction: The bending of light at edges causes blurring the image

of a pinhole appears as a blurred disc known as the Airy disc

 Astigmatism: Due to nonuniformities in lenses, a point may appear

as an ellipse

 Chromatic aberration:

length or energy that compose the ordinarily used white light converge

Section 3.9for a description of confocal microscopy

 Spherical aberration: The rays of light arriving at the periphery

of a lens are refracted more than the rays along the axis of the lens.This causes the rays from the periphery and the axis not to arrive at areduced by using a small aperture

 Geometric distortion: Poorly crafted lenses may cause geometricWhereas the best resolution achievable by the human eye is of the order

of 0:1;0:2 mm, light microscopes can provide resolving power up to about

0:2

Example: Figure 1.3 shows a rabbit ventricular myocyte in its relaxedstate as seen through a light microscope at a magni cation of about 600.The experimental setup was used to study the contractility of the myocytewith the application of electrical stimuli 31]

Example: Figure 1.4shows images of three-week-old scar tissue and week-old healed tissue samples from rabbit ligaments at a magni cation ofabout300 The images demonstrate the alignment patterns of the nuclei of broblasts (stained to appear as the dark objects in the images): the three-tions, whereas the forty-week-old healed sample has fewer broblasts that arewell-aligned along the length of the ligament (the horizontal edge of the im-age) The appearance of the forty-week-old sample is closer to that of normalsamples than that of the three-week-old sample Images of this nature havebeen found to be useful in studying the healing and remodeling processes inligaments 32]

forty-8 Biomedical Image Analysis

FIGURE 1.3

A single ventricular myocyte (of a rabbit) in its relaxed state The width(thickness) of the myocyte is approximately 15 Image courtesy of R.Clark, Department of Physiology and Biophysics, University of Calgary

10 Biomedical Image Analysis

1.4 Electron Microscopy

Accelerated electrons possess EM wave properties, with the wavelength

given by = mvh , wherehis Planck's constant,mis the mass of the electron,and v is the electron's velocity this relationship reduces to = 1 : 23

p

V, where

V is the accelerating voltage 30] At a voltage of 60 kV, an electron beam

:005 nm, and a resolving power limit

of about 0:003 nm Imaging at a low kV provides high contrast but lowresolution, whereas imaging at a high kV provides high resolution due tosmaller wavelength but low contrast due to higher penetrating power Inaddition, a high-kV beam causes less damage to the specimen as the fasterelectrons pass through the specimen in less time than with a low-kV beam.Electron microscopes can provide useful magni cation of the order of 106,and may be used to reveal the ultrastructure of biological tissues Electronmicroscopy typically requires the specimen to be xed, dehydrated, dried,mounted, and coated with a metal

Transmission electron microscopy: A transmission electron microscope(TEM) consists of a high-voltage electron beam generator, a series of EMlenses, a specimen holding and changing system, and a screen- lm holder, allenclosed in vacuum In TEM, the electron beam passes through the specimen,through a screen- lm combination or viewed via a phosphorescent viewingscreen

Example: Figure 1.5 shows TEM images of collagen bers (in section) in rabbit ligament samples The images facilitate analysis of thediameter distribution of the bers 33] Scar samples have been observed tohave an almost uniform distribution of ber diameter in the range 60;70nm,whereas normal samples have an average diameter of about 150 nm over abroader distribution Methods for the detection and analysis of circular ob-jects are described inSections 5.6.1,5.6.3, and 5.8

cross-Example: In patients with hematuria, the glomerular basement membrane

of capillaries in the kidney is thinner (<200 nm) than the normal thickness

of the order of 300nm34] Investigation of this feature requires needle-corebiopsy of the kidney and TEM imaging Figure 1.6 shows a TEM image

of a capillary of a normal kidney in cross-section Figure 1.7 (a) shows animage of a sample with normal membrane thickness Figure 1.7 (b) shows

an image of a sample with reduced and variable thickness Although theranges of normal and abnormal membrane thickness have been established byseveral studies 34], the diagnostic decision process is subjective methods forobjective and quantitative analysis are desired in this application

(b)

FIGURE 1.5

TEM images of collagen bers in rabbit ligament samples at a magni cation

of approximately30 000 (a) Normal and (b) scar tissue Images courtesy

of C.B Frank, Department of Surgery, University of Calgary

12 Biomedical Image Analysis

FIGURE 1.6

TEM image of a kidney biopsy sample at a magni cation of approximately

3 500 The image shows the complete cross-section of a capillary with mal membrane thickness Image courtesy of H Benediktsson, Department ofPathology and Laboratory Medicine, University of Calgary

© 2005 by CRC Press LLC

14 Biomedical Image Analysis

Scanning electron microscopy: A scanning electron microscope (SEM)

is similar to a TEM in many ways, but uses a nely focused electron beamwith a diameter of the order of 2nmto scan the surface of the specimen Theelectron beam is not transmitted through the specimen, which could be fairlythick in SEM Instead, the beam is used to scan the surface of the specimen

in a raster pattern, and the secondary electrons that are emitted from thesurface of the sample are detected and ampli ed through a photomultipliertube (PMT), and used to form an image on a cathode-ray tube (CRT) Anfrom the sample, and may be used to obtain images with a depth of eld ofseveralmm

Example: Figure 1.8 illustrates SEM images of collagen bers in rabbitligament samples (freeze-fractured surfaces) 35] The images are useful in an-alyzing the angular distribution of bers and the realignment process duringhealing after injury It has been observed that collagen bers in a normal lig-ament are well aligned, that bers in scar tissue lack a preferred orientation,and that organization and alignment return toward their normal patterns dur-ing the course of healing 36, 37, 35] Image processing methods for directionalanalysis are described in Chapter 8

FIGURE 1.8

SEM images of collagen bers in rabbit ligament samples at a magni cation

of approximately4 000 (a) Normal and (b) scar tissue Reproduced withpermission from C.B Frank, B MacFarlane, P Edwards, R Rangayyan, Z.Q.Liu, S Walsh, and R Bray, \A quantitative analysis of matrix alignment

in ligament scars: A comparison of movement versus immobilization in animmature rabbit model", Journal of Orthopaedic Research, 9(2): 219 { 227,

1991 cOrthopaedic Research Society

1.5 X-ray Imaging

The medical diagnostic potential of X rays was realized soon after their covery by R
oentgen in 1895 (See Robb 38] for a review of the history ofX-ray imaging.) In the simplest form of X-ray imaging or radiography, a 2Dprojection (shadow or silhouette) of a 3D body is produced on lm by irra-diating the body with X-ray photons 4, 3, 5, 6] This mode of imaging isreferred to as projection or planar imaging Each ray of X-ray photons isattenuated by a factor depending upon the integral of the linear attenuationcoecient along the path of the ray, and produces a corresponding gray level(or signal) at the point hit on the lm or the detecting device used

dis-Considering the ray path marked as AB in Figure 1.9,let Ni denote thenumber of X-ray photons incident upon the body being imaged, within aspeci ed time interval Let us assume that the X rays are mutually parallel,with the X-ray source at a large distance from the subject or object beingimaged Let No be the corresponding number of photons exiting the body.Then, we have

No=Niexp



; Z rayAB

at an angle  with respect to the (x y) coordinates indicated in Figure 1.9,with the saxis being parallel to the rays Then, s=;xsin+ycos Thevariable of integrationdsrepresents the elemental distance along the ray, andthe integral is along the ray path AB from the X-ray source to the detector.(See Section 9.1for further details on this notation.) The quantities Ni and

No are Poisson variables it is assumed that their values are large for theequations above to be applicable The function (x y) represents the linearattenuation coecient at (x y) in the sectional plane PQRS The value of(x y) depends upon the density of the object or its constituents along theray path, as well as the frequency (or wavelength or energy) of the radiationused Equation 1.2 assumes the use of monochromatic or monoenergetic Xrays

A measurement of the exiting X rays (that is,No, andNifor reference) thusgives us only an integral of (x y) over the ray path The internal details ofthe body along the ray path are compressed onto a single point on the lm

or a single measurement Extending the same argument to all ray paths,

16 Biomedical Image Analysis

P

S P’

Q’

N i

3D object 2D projection

An X-ray image or a typical radiograph is a 2D projection or planar image

of a 3D object The entire object is irradiated with X rays The projection

of a 2D cross-sectional plane PQRS of the object is a 1D pro le P'Q' of the2D planar image See also Figures 1.19 and 9.1 Reproduced, with permis-sion, from R.M Rangayyan and A Kantzas, \Image reconstruction", WileyEncyclopedia of Electrical and Electronics Engineering,Supplement 1, Editor:John G Webster, Wiley, New York, NY, pp 249{268, 2000 cThis material

is used by permission of John Wiley & Sons, Inc

we see that the radiographic image so produced is a 2D planar image of the3D object, where the internal details are superimposed In the case that therays are parallel to the xaxis (as in Figure 1.9), we have  = 90o, s =;x,

ds=;dx, and the planar image

inte-uorescent (phosphor) screen made of compounds of rare-earth elements such

as lanthanum oxybromide or gadolinium oxysul de, where the X-ray photonsare converted into visible-light photons A light-sensitive lm that is placed incontact with the screen (in a light-tight cassette) records the result The lmcontains a layer of silver-halide emulsion with a thickness of about 10 The exposure or blackening of the lm depends upon the number of lightphotons that reach the lm

A thick screen provides a high eciency of conversion of X rays to light,but causes loss of resolution due to blurring (see Figure 1.10) The typicalthickness of the phosphor layer in screens is in the range 40;100 Some

receiving units make use of a lm with emulsion on both sides that is wiched between two screens: the second screen (located after the lm alongthe path of propagation of the X rays) converts the X-ray photons not af-fected by the rst screen into light, and thereby increases the eciency ofthe receiver Thin screens may be used in such dual-screen systems to achievehigher conversion eciency (and lower dose to the patient) without sacri cingresolution.

sand-A

B

X rays

screen film light

FIGURE 1.10

Blur caused by a thick screen Light emanating from point A in the screen isspread over a larger area on the lm than that from point B

A uoroscopy system uses an image intensi er and a video camera in place

of the lm to capture the image and display it on a monitor as a movie orvideo 5, 6] Images are acquired at a rate of 2;8 frames=s (fps), withthe X-ray beam pulsed at 30;100ms per frame In computed radiography(CR), a photo-stimulable phosphor plate (made of europium-activated barium

uorohalide) is used instead of lm to capture and temporarily hold the imagepattern The latent image pattern is then scanned using a laser and digitized

In digital radiography (DR), the lm or the entire screen- lm combination isreplaced with solid-state electronic detectors 39, 40, 41, 42]

Examples: Figures 1.11 (a)and (b) show the posterior-anterior (PA, that

is, back-to-front) and lateral (side-to-side) X-ray images of the chest of apatient Details of the ribs and lungs, as well as the outline of the heart,are visible in the images Images of this type are useful in visualizing anddiscriminating between the air- lled lungs, the uid- lled heart, the ribs, andvessels The size of the heart may be assessed in order to detect enlargement

of the heart The images may be used to detect lesions in the lungs andfracture of the ribs or the spinal column, and to exclude the presence of uid

in the thoracic cage The use of two views assists in localizing lesions: use

of the PA view only, for example, will not provide information to decide if atumor is located toward the posterior or anterior of the patient

The following paragraphs describe some of the physical and technical siderations in X-ray imaging 4, 5, 6, 43, 44].

con- Target and focal spot: An electron beam with energy in the range

of 20;140keV is used to produce X rays for diagnostic imaging Thetypical target materials used are tungsten and molybdenum The term

\focal spot" refers to the area of the target struck by the electron beam

to generate X rays however, the nominal focal spot is typically pressed in terms of its diameter inmmas observed in the imaging plane(on the lm) A small focal spot is desired in order to obtain a sharpimage, especially in magni cation imaging (See also Section 2.9 and

ex-Figure 2.18.) Typical focal spot sizes in radiography lie in the range of

0:1;2mm A focal spot size of 0:1;0:3mmis desired in phy

mammogra- Energy: The penetrating capability of an X-ray beam is mainly termined by the accelerating voltage applied to the electron beam thatimpinges the target in the X-ray generator The commonly used indi-cator of penetrating capability (often referred to as the \energy" of theX-ray beam) iskV p, standing for kilo-volt-peak The higher thekV p,the more penetrating the X-ray beam will be The actual unit of en-ergy of an X-ray photon is the electron volt oreV, which is the energygained by an electron when a potential of 1V is applied to it ThekV p

de-measure relates to the highest possible X-ray photon energy that may

be achieved at the voltage used

Low-energy X-ray photons are absorbed at or near the skin surface, and

do not contribute to the image In order to prevent such unwantedradiation, a lter is used at the X-ray source to absorb low-energy Xrays Typical lter materials are aluminum and molybdenum

Imaging of soft-tissue organs such as the breast is performed with energy X rays in the range of 25;32kV p45] The use of a higherkV p

low-ity or contrast A few other energy levels used in projection radiographyare, for imaging the abdomen: 60;100kV p chest: 80;120kV p andskull: 70;90 kV p The kV p to be used depends upon the distancebetween the X-ray source and the patient, the size (thickness) of thepatient, the type of grid used, and several other factors

 Exposure: For a given tube voltage (kV p), the total number of X-rayphotons released at the source is related to the product of the tube cur-rent (mA) and the exposure time (s), together expressed as the product

mAs As a result, for a given body being imaged, the number of tons that arrive at the lm is also related to themAsquantity A low

pho-mAsresults in an under-exposed lm (faint or light image), whereas ahighmAsresults in an over-exposed or dark image (as well as increased

20 Biomedical Image AnalysisX-ray dose to the patient) Typical exposure values lie in the range

of 2;120 mAs Most imaging systems determine automatically therequired exposure for a given mode of imaging, patient size, andkV p

setting Some systems use an initial exposure of the order of 5ms toestimate the penetration of the X rays through the body being imaged,and then determine the required exposure

 Beam hardening: The X rays used in radiographic imaging are cally not monoenergetic that is, they possess X-ray photons over a cer-tain band of frequencies or EM energy levels As the X rays propagatethrough a body, the lower-energy photons get absorbed preferentially,depending upon the length of the ray path through the body and theattenuation characteristics of the tissues along the path Thus, the Xrays that pass through the object at longer distances from the sourcewill possess relatively fewer photons at lower-energy levels than at thepoint of entry into the object (and hence a relatively higher concentra-and leads to incorrect estimation of the attenuation coecient in com-reduced by pre ltering or prehardening the X-ray beam and narrowingits spectrum The use of monoenergetic X rays from a synchrotron or alaser obviates this problem

typi- Scatter and the use of grids:As an X-ray beam propagates through

a body, photons are lost due to absorption and scattering at each point

in the body The angle of the scattered photon at a given point along theincoming beam is a random variable, and hence the scattered photoncontributes to noise at the point where it strikes the detector Fur-thermore, scattering results in the loss of contrast of the part of theobject where X-ray photons were scattered from the main beam Thesion imaging, and requires speci c methods to improve the quality ofgrids, collimation, or energy discrimination due to the fact that the scat-tered (or secondary) photons usually have lower energy levels than theprimary photons

A grid consists of an array of X-ray absorbing strips that are mutuallyparallel if the X rays are in a parallel beam, as in chest imaging (see

Figures 1.12and1.13),or are converging toward the X-ray source in thecase of a diverging beam (as in breast imaging, seeFigure 1.15) Lattice

or honeycomb grids with parallel strips in criss-cross patterns are alsoused in mammography X-ray photons that arrive via a path that is notaligned with the grids will be stopped from reaching the detector

A typical grid contains thin strips of lead or aluminum with a strip sity of 25;80lines=cmand a grid height:strip width ratio in the range

den-screen-film parallel grid

in the illustration

of 5:1 to 12:1 The space between the grids is lled with low-attenuationmaterial such as wood A stationary grid produces a line pattern that issuperimposed upon the image, which would be distracting Figure 1.13

(a) shows a part of an image of a phantom with the grid artifact clearlyvisible (An image of the complete phantom is shown inFigure 1.14.)

Grid artifact is prevented in a reciprocating grid, where the grid is movedabout 20 grid spacings during exposure: the movement smears the gridshadow and renders it invisible on the image Figure 1.13 (b) shows

an image of the same object as in part (a), but with no grid artifact.Low levels of grid artifact may appear in images if the bucky that holdsthe grid does not move at a uniform pace or starts moving late or endsmovement early with respect to the X-ray exposure interval A majordisadvantage of using grids is that it requires approximately two timesthe radiation dose required for imaging techniques without grids Fur-thermore, the contrast of ne details is reduced due to the smearedshadow of the grid

 Photon detection noise: The interaction between an X-ray beamand a detector is governed by the same rules as for interaction withany other matter: photons are lost due to scatter and absorption, andsize of the detectors in DR and CT imaging reduces their detection

22 Biomedical Image Analysiseciency Scattered and undetected photons cause noise in the mea-surement for detailed analysis of noise in X-ray detection, refer to Bar-rett and Swindell 3], Macovski 5], and Cho et al 4] More details onnoise in medical images and techniques to remove noise are presented in

Chapter 3

 Ray stopping by heavy implants:If the body being imaged containsextremely heavy parts or components, such as metal screws or pins inbones and surgical clips that are nearly X-ray-opaque and entirely stopthe incoming X-ray photons, no photons would be detected at the cor-responding point of exit from the body The attenuation coecient forthe corresponding path would be inde nite, or within the computationalcontext, in nity Then, a reconstruction algorithm would not be able

to redistribute the attenuation values over the points along the sponding ray path in the reconstructed image This leads to streakingartifacts in CT images

corre-Two special techniques for enhanced X-ray imaging | digital subtractionangiography (DSA) and dual-energy imaging | are described inSections 4.1

and 4.2, respectively

1.5.1 Breast cancer and mammography

Breast cancer: Cancer is caused when a single cell or a group of cellsescapes from the usual controls that regulate cellular growth, and begins tomultiply and spread This activity results in a mass, tumor, or neoplasm.Many masses are benign that is, the abnormal growth is restricted to a single,circumscribed, expanding mass of cells Some tumors are malignant that is,the abnormal growth invades the surrounding tissues and may spread, ormetastasize, to distant areas of the body Although benign masses may lead

to complications, malignant tumors are usually more serious, and it is forthese tumors that the term \cancer" is used The majority of breast tumorswill have metastasized before reaching a palpable size

Although curable, especially when detected at early stages, breast cancer

is a major cause of death in women An important factor in breast cancer

is that it tends to occur earlier in life than other types of cancer and othermajor diseases 47, 48] Although the cause of breast cancer has not yetbeen fully understood, early detection and removal of the primary tumor are

in time, only a few of the cells that departed from the primary tumor wouldhave succeeded in forming secondary tumors 49] When breast tumors aretumors would have metastasized 50]

If breast cancer can be detected by some means at an early stage, while it isclinically localized, the survival rate can be dramatically increased However,

24 Biomedical Image Analysis

FIGURE 1.14

X-ray image of the American College of Radiology (ACR) phantom for mography The pixel-value range 117 210] has been linearly stretched to thedisplay range 0 255] to show the details Image courtesy of S Bright, Sun-nybrook & Women's College Health Sciences Centre, Toronto, ON, Canada.See alsoFigure 1.13

mam-such early breast cancer is generally not amenable to detection by physical amination and breast self-examination The primary role of an imaging tech-nique is thus the detection of lesions in the breast 29] Currently, the mostphy Other modalities, such as ultrasonography, transillumination, thermog-raphy, CT, and magnetic resonance imaging (MRI) have been investigated forbreast cancer diagnosis, but mammography is the only reliable procedure fordetecting nonpalpable cancers and for detecting many minimal breast cancerswhen they appear to be curable 18, 28, 29, 51] Therefore, mammographyhas been recommended for periodic screening of asymptomatic women Mam-mography has gained recognition as the single most successful technique forthe detection of early, clinically occult breast cancer 52, 53, 54, 55, 56].

ex-X-ray imaging of the breast: The technique of using X rays to tain images of the breast was rst reported by Warren in 1930, after he hadexamined 100 women using sagital views 57] Because of the lack of a re-producible method for obtaining satisfactory images, this technique did notmake much progress until 1960, when Egan 58] reported on high-mA andlow-kV pX-ray sources that yielded reproducible images on industrial lm Itwas in the mid-1960s that the rst modern X-ray unit dedicated to mammog-raphy was developed Since then, remarkable advances have led to a strikingimprovement in image quality and a dramatic reduction in radiation dose

ob-A major characteristic of mammograms is low contrast, which is due tothe relatively homogeneous soft-tissue composition of the breast Many ef-forts have been focused on developing methods to enhance contrast In analternative imaging method known as xeromammography, a selenium-coatedaluminum plate is used as the detector 6] The plate is initially charged

to about 1 000 V Exposure to the X rays exiting the patient creates acharge pattern on the plate due to the liberation of electrons and ions Theplate is then sprayed with an ionized toner, the pattern of which is trans-ferred to plastic-coated paper Xeromammograms provide wide latitude andedge enhancement, which lead to improved images as compared to screen- lmmammography However, xeromammography results in a higher dose to thesubject, and has not been in much use since the 1980s

A typical mammographic imaging system is shown schematically in

Fig-ure 1.15 Mammography requires high X-ray beam quality (a narrow-band

or nearly monochromatic beam), which is controlled by the tube target compression is an important factor in reducing scattered radiation, creating

ma-as uniform a density distribution ma-as possible, eliminating motion, and rating mammary structures, thereby increasing the visibility of details in theimage The use of grids speci cally designed for mammography can furtherreduce scattered radiation and improve subject contrast, which is especiallysigni cant when imaging thick, dense breasts 59]

sepa-Generally, conventional screen- lm mammography is performed with thebreast directly in contact with the screen- lm cassette, producing essentially

26 Biomedical Image Analysis

X-ray source (target)

Breast compression paddle

Compressed breast

Focused grid Screen-film cassette

Filter Collimating diaphragm

FIGURE 1.15

A typical mammography setup

life-size images The magni cation technique, on the other hand, interposes

an air gap between the breast and the lm, so that the projected radiographicimage is enlarged Magni cation produces ne-detail images containing addi-tional anatomical information that may be useful in re ning mammographicdiagnosis, especially in cases where conventional imaging demonstrates un-certain or equivocal ndings 60] As in the grid method, the advantages

of magni cation imaging are achieved at the expense of increased radiationexposure Therefore, the magni cation technique is not used routinely.Screen- lm mammography is now the main tool for the detection of earlybreast cancer The risk of radiation is still a matter of concern, althoughthere is no direct evidence of breast cancer risk from the low-dose radiationexposure of mammography Regardless, technological advances in mammog-raphy continue to be directed toward minimizing radiation exposure whilemaintaining the high quality of the images

Examples: Figures 1.16 (a) and (b) show the cranio-caudal (CC) andmedio-lateral-oblique (MLO) views of the same breast of a subject The MLOview demonstrates architectural distortion due to a spiculated tumor near theupper right-hand corner edge

Mammograms are analyzed by radiologists specialized in mammography Anormal mammogram typically depicts converging patterns of broglandulartissues and vessels Any feature that causes a departure from or distortionwith reference to the normal pattern is viewed with suspicion and analyzedwith extra attention Features such as calci cations, masses, localized increase

in density, architectural distortion, and asymmetry between the left and rightbreast images are carefully analyzed

Several countries and states have instituted breast cancer screening grams where asymptomatic women within a certain age group are invited toparticipate in regular mammographic examinations Screen Test | AlbertaProgram for the Early Detection of Breast Cancer 61] is an example of such

pro-a progrpro-am Severpro-al pro-applicpro-ations of impro-age processing pro-and ppro-attern pro-anpro-alysistechniques for mammographic image analysis and breast cancer detection aredescribed in the chapters to follow

1.6 Tomography

The problem of visualizing the details of the interior of the human bodynoninvasively has always been of interest, and within a few years after thediscovery of X rays by R
ontgen in 1895, techniques were developed to imagesectional planes of the body The techniques of laminagraphy, planigraphy,

or \classical" tomography 38, 62] used synchronous movement of the X-raysource and lm in such a way as to produce a relatively sharp image of a

28 Biomedical Image Analysis

FIGURE 1.16

(a) Cranio-caudal (CC) and (b) medio-lateral oblique (MLO) mammograms

of the same breast of a subject Images courtesy of Screen Test | AlbertaProgram for the Early Detection of Breast Cancer 61]

single focal plane of the object, with the images of all other planes beingblurred Figure 1.17 illustrates a simple linear-motion system, where the X-ray source and lm cassette move along straight-line paths so as to maintainthe longitudinal (coronal) plane, indicated by the straight line AB, in focus.

It is seen that the X rays along the paths X1-A and X2-A strike the samephysical spot A1 = A2 on the lm, and that the rays along the paths X1-Band X2-B strike the same spot B1 = B2 On the other hand, for the point Cpoints C1 6= C2 on the lm Therefore, the details in the plane AB remain

in focus and cause a strong image, whereas the details in the other planes aresmeared all over the lm The smearing of information from the other planes

of the object causes loss of contrast in the plane of interest The development

of CT imaging rendered lm-based tomography obsolete

Example: Figure 1.18shows a tomographic image of a patient in a dinal (coronal) plane through the chest Images of this nature provided bettervisualization and localization of lesions than regular X-ray projection images,and permitted the detection of masses in bronchial tubes and air ducts

Patient Path of source movement

Path of film movement

Table Film cassette