ArticlePDF AvailableLiterature Review

Recent Advances in Chest Radiography 1

January 2007
Radiology 241(3):663-83

January 2007
241(3):663-83

DOI:10.1148/radiol.2413051535

Source
PubMed

Authors:

Holman Page Mcadams

Duke University

Ehsan Samei

Duke University

James Dobbins

Duke University

Georgia D Tourassi

Oak Ridge National Laboratory

Show all 5 authorsHide

There have been many remarkable advances in conventional thoracic imaging over the past decade. Perhaps the most remarkable is the rapid conversion from film-based to digital radiographic systems. Computed radiography is now the preferred imaging modality for bedside chest imaging. Direct radiography is rapidly replacing film-based chest units for in-department posteroanterior and lateral examinations. An exciting aspect of the conversion to digital radiography is the ability to enhance the diagnostic capabilities and influence of chest radiography. Opportunities for direct computer-aided detection of various lesions may enhance the radiologist's accuracy and improve efficiency. Newer techniques such as dual-energy and temporal subtraction radiography show promise for improved detection of subtle and often obscured or overlooked lung lesions. Digital tomosynthesis is a particularly promising technique that allows reconstruction of multisection images from a short acquisition at very low patient dose. Preliminary data suggest that, compared with conventional radiography, tomosynthesis may also improve detection of subtle lung lesions. The ultimate influence of these new technologies will, of course, depend on the outcome of rigorous scientific validation.

…

Figures - uploaded by Holman Page Mcadams

Content may be subject to copyright.

Content uploaded by Holman Page Mcadams

Content may be subject to copyright.

Recent Advances in Chest

Radiography

H. Page McAdams, MD

Ehsan Samei, PhD

James Dobbins III, PhD

Georgia D. Tourassi, PhD

Carl E. Ravin, MD

There have been many remarkable advances in conven-

tional thoracic imaging over the past decade. Perhaps the

most remarkable is the rapid conversion from ﬁlm-based

to digital radiographic systems. Computed radiography is

now the preferred imaging modality for bedside chest

imaging. Direct radiography is rapidly replacing ﬁlm-based

chest units for in-department posteroanterior and lateral

examinations. An exciting aspect of the conversion to dig-

ital radiography is the ability to enhance the diagnostic

capabilities and inﬂuence of chest radiography. Opportuni-

ties for direct computer-aided detection of various lesions

may enhance the radiologist’s accuracy and improve efﬁ-

ciency. Newer techniques such as dual-energy and tempo-

ral subtraction radiography show promise for improved

detection of subtle and often obscured or overlooked lung

lesions. Digital tomosynthesis is a particularly promising

technique that allows reconstruction of multisection im-

ages from a short acquisition at very low patient dose.

Preliminary data suggest that, compared with conven-

tional radiography, tomosynthesis may also improve de-

tection of subtle lung lesions. The ultimate inﬂuence of

these new technologies will, of course, depend on the

outcome of rigorous scientiﬁc validation.

娀 RSNA, 2006

From the Department of Radiology, Duke Advanced Im-

aging Laboratories, Duke University Medical Center, Box

3808, Durham, NC 27710. Received September 14,

2005; revision requested October 24; revision received

November 15; accepted January 2, 2006; ﬁnal version

accepted January 6. Supported by National Institutes of

Health grant R01 CA80490 and a research agreement

with from GE Healthcare. Address correspondence to

H.P.M. (e-mail: Page.mcadams@duke.edu).

姝 RSNA, 2006

REVIEWS AND COMMENTARY

䡲

STATE OF THE ART

Radiology: Volume 241: Number 3—December 2006 663

Note: This copy is for your personal non-commercial use only. To order presentation-ready

copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.

espite recent advances in cross-

sectional imaging of the thorax,

chest radiography remains the

mainstay for diagnosis of many pulmo-

nary diseases. In most instances, it is

the ﬁrst—and frequently the only—di-

agnostic imaging test performed in pa-

tients known to have or suspected of

having a thoracic abnormality. In the

United States, and very likely in the

world, chest radiography remains the

most commonly performed diagnostic

imaging test overall.

In the more than 100 years since

the discovery of the x-ray, technologic

advances have resulted in many im-

provements in chest radiography (1).

Progress in ﬁlm-based imaging led to

the development of excellent screen-

ﬁlm systems designed speciﬁcally for

chest radiography (1). More recently,

advances in electronics and computer

technology have resulted in rapid devel-

opment in digital image receptors and

displays. Further, rapid development of

image-processing techniques and of ad-

vanced applications such as dual-energy

and temporal subtraction radiography,

digital tomosynthesis, and computer-

aided detection (CAD) and diagnosis

(CADx) promise to substantially im-

prove the way chest radiography is

practiced in the future.

In this article, we will discuss the

inherent challenges of chest radiogra-

phy and the speciﬁc advances made to

address these challenges, including new

digital detector and image display tech-

nologies, developments in image-pro-

cessing techniques and CAD and CADx

applications for chest radiography, ad-

vanced applications such as dual-energy

and temporal subtraction radiography,

and chest tomosynthesis.

Challenges

To better appreciate the recent ad-

vances in chest radiography, we must

ﬁrst review some of the inherent chal-

lenges of the technique, because these

challenges have been the prime motiva-

tors behind most of the developments.

These challenges include but are not

limited to issues related to image area

and patient body habitus, latitude and

dynamic range of x-ray transmission

through the chest, scattered radiation,

overlap of anatomic structures, and per-

ceptual limitations.

Image Area and Body Habitus

The chest is one of largest and thickest

body parts that is imaged on a routine

basis. The typical imaging ﬁeld in an

adult readily exceeds 40-cm, particu-

larly for patients with a large body hab-

itus. This large ﬁeld of view imposes

challenging constraints on the size of the

image receptor, given that the receptor

must also provide consistent and uni-

form response over the entire ﬁeld. This

large ﬁeld of view also increases the

contribution of scattered radiation to

the image, degrading the image’s inher-

ent quality. Given the marked increase

in obesity in the United States over the

past 20 years (2) among both children

and adults, these issues will continue to

pose considerable challenges to chest

radiography in the future (Fig 1). For

example, the authors of a recent study

(3) found that the number of chest ra-

diographs that were considered limited

because of body habitus had more than

doubled during a 14-year period at one

hospital.

Latitude and Dynamic Range

The wide latitude of x-ray transmission

through the thorax imposes a funda-

mental limit on the visualization of sub-

tle abnormalities on conventional chest

radiographs. For a typical x-ray beam

used in chest radiography, the regional

variations in transmission through the

thorax can extend over two orders of

magnitude (4) (Fig 2). Ideally, an imag-

ing system should have enough latitude

to capture and effectively display this

wide range, or at least the diagnostically

meaningful part, of the x-ray transmis-

sion. However, coverage of such wide

latitude can limit depiction of minute

differences associated with subtle low-

contrast lesions. Maintaining wide lati-

tude while preserving visualization of

low-contrast features in the image is a

particular challenge for chest radiogra-

phy (1). This large variation in x-ray

opacity of anatomic structures is the

primary reason that the modality uses

the highest range of x-ray energy set-

tings (typically 100–150 kVp) of all di-

agnostic imaging procedures, at the cost

of increased scatter and reduced inher-

ent contrast.

Scattered Radiation

The combined use of high x-ray photon

energy in conjunction with a thick body

part and a large ﬁeld of view results in a

large amount of scattered radiation in

chest radiography compared with that

in other x-ray imaging modalities. Scat-

tered x-rays can account for 95% of the

detected x-ray ﬂux in the mediastinum

and up to 70% in the lung in radio-

graphs acquired without a grid (Fig 3)

Essentials

䡲

Chest radiography remains the

mainstay for diagnosis of many

thoracic diseases.

䡲 Recent developments in chest ra-

diography have been primarily

driven by recognized challenges

and limitations of the technique.

䡲 Computed radiographic, and now

full-ﬁeld ﬂat-panel detector, sys-

tems have replaced or are rapidly

replacing conventional ﬁlm-based

systems for chest image acquisi-

tion.

䡲 Selection of image-processing pa-

rameters and display devices has

a profound and fundamental ef-

fect on the appearance of the digi-

tal chest image.

䡲 The rapid conversion from

screen-ﬁlm to digital chest image

acquisition has greatly facilitated

advances such as computer-aided

detection and diagnosis, dual-en-

ergy or temporal subtraction im-

aging, and tomosynthesis.

Published online

10.1148/radiol.2413051535

Radiology 2006; 241:663–683

Abbreviations:

CAD ⫽ computer-aided detection

CADx ⫽ computer-aided diagnosis

CCD ⫽ charge-coupled device

CMOS ⫽ complementary metal-oxide semiconductor

CR ⫽ computed radiography

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

664 Radiology: Volume 241: Number 3—December 2006

(6,7). Scattered radiation has the dele-

terious effects of degrading contrast and

increasing image noise. While this re-

duction in contrast can be remedied

with postprocessing, at least in digital

radiography, the contribution of scat-

tered radiation to image noise may not

be readily correctable. Further, in-

creased obesity in the U.S. population,

as noted earlier, has had a notable ef-

fect on the increased level of scattered

radiation in chest radiography.

Overlay of Anatomic Structures

In its conventional form, chest radiogra-

phy involves the projection of a three-

dimensional structure onto a two-dimen-

sional image. In this process, anatomic

features such as ribs, lung vessels, heart,

mediastinum, and diaphragm overlay

each other in a pattern that can be re-

ferred to as anatomic noise (8). Anatomic

Figure 1

Figure 1: Chest radiographs obtained with computed radiography (CR) technique and grid in morbidly obese patient with signs and symptoms of pulmonary edema.

(a) Anteroposterior image is limited by body habitus and was interpreted as showing possible edema. (b) Posteroanterior image obtained 30 minutes later shows no evi-

dence of edema.

Figure 2

Figure 2: Graph shows dy-

namic range of an image as func-

tion of thickness of soft tissue

penetrated and peak voltage (5).

Max ⫽ maximum, Min ⫽ mini-

mum.

Figure 3

Figure 3: Graph shows scatter fractions in

lung, mediastinum, and diaphragm regions with-

out a grid (black bars), with a grid (white bars), and

with a slot-scan device (gray bars). Note marked

reduction in scatter with slot-scan device.

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

Radiology: Volume 241: Number 3—December 2006 665

noise can have a detrimental effect on the

ability of the observer to detect abnor-

malities of concern. The projection of

ribs is of particular concern for detec-

tion of lung nodules, because the ribs

overlay about 75% of the area of the

lungs (9). Furthermore, a substantial

portion of the lungs is projected over

the heart and diaphragm regions of the

image (1), affecting the contrast of po-

tential lung lesions on the image.

The appearance of lung nodules on

chest radiographs is of particular con-

cern. The similarity of a lesion’s appear-

ance to that of the background anatomy

in which it is located causes the conspi-

cuity of the lesion to be poor. This phe-

nomenon can result in up to 30% of

pulmonary nodules being missed on ini-

tial chest radiographs (10), even though

the nodules could be observed retro-

spectively. Purely on the basis of inher-

ent contrast, a nodule as small as 3 mm

in diameter should be visible on chest

radiographs, even in the presence of

scattered radiation; however, it is rare

to detect nodules smaller than about 8

mm on chest radiographs, owing to the

inﬂuence of anatomic noise. A great

deal of work has been performed on the

effect of anatomic noise on nodule de-

tectability, beginning with the work of

Kundel and colleagues (11,12) several

decades ago. More recently, Samei et al

(13) demonstrated that anatomic back-

ground is far more important than

quantum noise in limiting the detectabil-

ity of lung nodules (Fig 4).

Perceptual Limitations

Medical images are generally of little

value until an expert reader interprets

them. Thus, the perceptual and cog-

nitive processes involved in reading a

medical image have a direct bearing

on its clinical utility and effectiveness.

These processes are of particular im-

portance in chest radiography because

of the complexity of the tasks and the

confounding effect of anatomic noise,

as noted earlier in this article (14).

Anatomic noise can hinder detection

through two processes: (a) local inﬂu-

ence, or “camouﬂaging,” and (b) global

inﬂuence, or “confusion” (15). The local

effect obscures the deﬁnition of an ab-

normality by means of patterns immedi-

ately surrounding or overlaying it. For

the global effect, however, the detect-

ability of an abnormality (eg, a nodule)

is inﬂuenced by the degree of its distinc-

tiveness from similar structures created

by the global noise characteristics of the

background.

Perceptual errors can occur at both

the visual and the cognitive level. In-

completeness of the search task may

lead to about 55% of the missed lesions.

These errors occur when the observer

fails to look at the territory of the lesion

(30%) (11,16) or when he or she does

not ﬁx his or her eyes on the territory

for a dwell time of at least 0.3 second

(25%) (17). Cognitive errors (account-

ing for 45% of missed lesions) can occur

when the ﬁxation time on an abnormal-

ity candidate exceeds the above limit

(eg, a nodule is recognized), but the

observer commits a decision-making er-

ror by calling the case negative (11).

Results of prior studies suggest that

perceptual errors can be reduced by pro-

viding radiologists with ﬁxation feedback

(18) and by using systematic search strate-

gies, coning devices, and double reading of

chest radiographs (19). CAD (further dis-

cussed later), with its ability to offer a com-

plete search of the image data, has the po-

tential to reduce certain types of perceptual

error. The most effective method by far,

however, is the reduction or elimination of

anatomic noise, which has been shown to

be the main factor limiting the detection of

subtle lung nodules (13).

Detector Developments

The imaging receptor (or detector) is a

key component of chest radiography. In

the past few decades, changes in recep-

tor technology have brought about one

of the most important advances in chest

radiography, leading to improved image

quality and new image acquisition tech-

niques. This section will summarize the

most important developments in this

area. Because most advances have been

focused on digital technologies, they will

be the main focus of this section. The

Table provides a summary of the current

commercially available digital receptor

technologies for chest radiography.

Analog Chest Radiography

Chest radiography is conventionally per-

formed with analog screen-ﬁlm recep-

tors. In the most common implementa-

tion, these receptors are made of double-

emulsion light-sensitive ﬁlm sandwiched

between two layers of phosphor screens.

The screens serve as the primary medium

to convert x-ray photons to light photons,

which are subsequently detected by the

ﬁlm emulsion, where the ionic silver of

the emulsion is converted to metallic

silver. Subsequent chemical processing

washes the desensitized chemicals away,

leaving the residual metallic silver behind

to form the x-ray image (20).

While screen-ﬁlm receptors have

Figure 4

Figure 4: Images in middle-aged woman with

history of right partial mastectomy for breast can-

cer who presented for routine follow-up. (a) Pos-

teroanterior chest radiograph was interpreted as

normal. (b) Transverse computed tomographic

(CT) scan shows 8-mm nodule (arrowhead) in

right lower lobe that was obscured by overlying

breast implant, diaphragm, and rib shadows on a.

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

666 Radiology: Volume 241: Number 3—December 2006

been in continual use for many years,

their inherent characteristics have im-

posed certain limitations on the practice

of chest radiography. Owing to the lim-

ited light range sensitivity of ﬁlm, repre-

sented by the Hurter and Drifﬁeld re-

sponse of the receptor (Fig 5), the range

of x-ray exposures that can be recorded

by a screen-ﬁlm detector is limited.

Only a narrow range of exposures can

be reliably represented on the ﬁlm with

acceptable contrast; exposures above

or below this range will be represented

suboptimally. Given the large variability

in radiopacity of organs in the thoracic

cavity, this characteristic leads to two

limitations of analog chest radiography:

(a) suboptimal contrast renditions at

the extremes of the attenuation range of

the thorax and (b) susceptibility to sub-

optimal image acquisition due to under-

or overexposure.

In recent years, most technical de-

velopments have focused on digital tech-

nologies (discussed later). The latest de-

velopment in analog chest radiographic

receptors, now over 13 years old, is an

improvement on traditional screen-ﬁlm

chest techniques; this development in-

volves combining two different emul-

sions and screens with an anticrossover

layer that is opaque until the ﬁlm is de-

veloped. This asymmetric screen-ﬁlm

technology, developed and marketed by

Eastman Kodak (Rochester, NY), im-

proves the contrast of ﬁne-detail anat-

omy by using a thin front screen while

maintaining good overall image noise

through improved x-ray detection in a

thick rear screen (4,21). The overall re-

ceptor has a markedly wider acceptable

exposure window and contrast sensitiv-

ity that provide a better balance be-

tween contrast sensitivity and latitude

while maintaining good quantum mottle

properties.

Figure 5 shows the Hurter and Drif-

ﬁeld curves for a typical chest radiogra-

phy screen-ﬁlm system in comparison

with a typical digital detector.

Digital Chest Radiography

While screen-ﬁlm receptors have been

the main technology for acquiring chest

Summary of Selected Commercial Digital Chest Radiography Systems

Manufacturer or Provider Product Family Size (cm)

Pixel Pitch

(␮m) X-ray Absorbing Material*

Agfa Healthcare CR 25.0, 75.0, DX-S 36 ⫻ 43 ⱖ100 Ba halide

Eastman Kodak Directview CR 35 ⫻ 43 ⱖ97 Ba halide

Fujiﬁlm Medical Clearview, Velocity, SmartCR, XG5000 35 ⫻ 43,

43 ⫻ 43

ⱖ100 Ba halide

Konika Minolta XpressCR, IQue CR 35 ⫻ 43 175 Ba halide

Orex PcCR 1417 ACL, ACLxy 35 ⫻ 43 100 Ba halide

Philips Medical Systems Compano, Corado, Cosima X 35 ⫻ 43 ⱖ100 Ba halide

Indirect ﬂat panel

GE/Perkin Elmer Revolution XQ/i, XR/d 41 ⫻ 41 200 CsI (Tl), undisclosed thickness

Trixell Pixium 4600TM; Siemens: Thorax/Vertix/Multix FD, Axoim Aristos;

Philips: Digital Diagnost Inﬁmed and Listem: Stingray Quantum:

Q-RAD (CsI), QV (CsI)

43 ⫻ 43 143 CsI (Tl), ⬃550 ␮m thick

Canon Canon: CXDI; Agfa: ADR Thorax; Quantum: Q-RAD (GOS), QV

(GOS); Siemens: Axiom Multix M

43 ⫻ 43 160 Gd

S (Tb), ⬃200 ␮m thick

Direct ﬂat panel

Direct Ray Hologic: Epex, Radex; Eastman Kodak: DirectViewDR; DEL: EPEX,

RADEX

35 ⫻ 43 139 Amorphous Se, ⬃500 ␮m

thick

Anrad Toshiba: DynaDirect 35 ⫻ 35 150 Amorphous Se, ⬃1000 ␮m

thick

Edge Medical Quix DR systems 43 ⫻ 43 120 Amorphous Se, undisclosed

thickness

CCD- and CMOS-based

†

SwissRay ddR 35 ⫻ 43 167 CsI, undisclosed doping

Wuestec DX2000 35 ⫻ 43 120 Gd

S(Tb)

Imaging Dynamics Xplorer 43 ⫻ 43 108 Gd

S(Tb)

Oy ImixAB Imix 2000 40 ⫻ 40 200 Gd

S(Eu)

Delft Diagnostic Imaging Thorascan 44 ⫻ 44 162 CsI(Tl), ⬃500 ␮m thick

Cares Built Cares Built: Clarity 7000; Lodox: Statscan 43 ⫻ 43 60 Gd

S(Tb)

Star V-Ray Tradix 4000 43 ⫻ 43 140 Gd

S(Tb)

*Ba⫽ barium, CsI ⫽ cesium iodide, Eu ⫽ europium, Gd ⫽ gadolinium, O ⫽ oxygen, S ⫽ sulfur, Se ⫽ selenium, Tl ⫽ thallium, Tb ⫽ terbium.

†

CCD ⫽ charge-coupled device, CMOS ⫽ complementary metal-oxide semiconductor.

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

Radiology: Volume 241: Number 3—December 2006 667

radiographs until recently, the use of

digital radiography has been on the rise

and is expected to replace analog tech-

nology in the near future. Digital radiog-

raphy has also enabled the implementa-

tion of picture archiving and communi-

cation systems, which have their own

associated advantages. There have been

six main forces driving the transition to

digital:

1. Decoupling of acquisition and dis-

play functions of the acquisition device

in digital radiography makes it possible

to optimize either of those functions in-

dependently—for example, by optimiz-

ing display contrast independently of

the exposure level.

2. Availability of the image data in

an electronic form makes it possible to

postprocess the image for optimal display

and to display the images on the more

ﬂexible soft-copy viewing workstations.

3. Availability of the image data in

an electronic form makes it possible to

archive the data electronically, which

uses less space and requires less labor

for storage.

4. Availability of the image data in

an electronic form makes it possible to

distribute images widely and to make

copies of images available concurrently

to multiple users.

5. Most digital radiography systems

have integrated acquisition and pro-

cessing units in the same physical sys-

tem, eliminating the need to handle the

cassettes and thus improving the work-

ﬂow for image acquisition in the depart-

ment.

6. Most digital radiography systems

are able to acquire multiple images in a

rapid sequence, enabling new advanced

imaging applications that require the ac-

quisition of multiple images from the

chest, as will be discussed in a later

section of this article.

Because of these advantages, digital

radiography has been a main focus of

new developments in chest radiogra-

phy. Differing technologies have been

the basis of the development of various

commercial products, which are sum-

marized below.

In comparing various chest radiog-

raphy systems, a common metric, de-

tective quantum efﬁciency, has been

used as a basis of comparison. Detective

quantum efﬁciency describes the inher-

ent signal-to-noise performance of an

imaging system per unit incident expo-

sure to the detector (22). Most com-

monly reduced to its value at zero spa-

tial frequency, various chest radiogra-

phy systems offer a detective quantum

efﬁciency in the range of 20%–70%

(23,24). These values are reﬂective of

the system performance in the absence

of scattered radiation. Most recently, a

new metric, termed effective detective

quantum efﬁciency, has been intro-

duced that further accounts for the

presence of scattered radiation (25), re-

ﬂecting the inherent signal-to-noise per-

formance of an imaging system in actual

clinical use.

CR Systems

CR was the ﬁrst commercial digital im-

aging modality widely used in chest im-

aging (26) and currently is still the most

common technology for acquiring digital

chest radiographs, especially for bed-

side applications. The technology is

based on photostimulable properties of

barium halide phosphors. With a cas-

sette similar to that used in screen-ﬁlm

radiography, the phosphor screen is ex-

posed to x-rays. After exposure, the

cassette is transported to a computed

radiography reader device, which sub-

jects the screen to a scanning laser beam.

The laser releases the energy locally de-

posited by x-rays on the screen, causing

the screen to ﬂuoresce. The released light

is used to form the image after it is col-

lected by a light guide, digitized, and asso-

ciated with the geometric location of the

laser beam at the time of stimulation. The

digital image data are then processed for

presentation.

Since its commercial introduction

more than 2 decades ago, CR has been

under continuous improvement. The

most recent advances include the collec-

tion of photostimulated light from both

sides of the screen, leading to improved

detective quantum efﬁciency (currently

available only for mammography) (27),

the use of line scanning and light collec-

tion technology for improved speed

(28), and the use of structured phos-

phor for improved detection efﬁciency

without associated loss in resolution

(29).

In the most common implementa-

tion, CR devices produce image quality

that tends to be lower than that for

other digital radiography systems (23),

but CR systems possess certain opera-

tional and economic advantages that

cause them to remain competitive with

other digital modalities. Those advan-

tages can be summarized as follows:

(a) In the most common implementa-

tion, CR cassettes are identical in size to

screen-ﬁlm cassettes, making it possible

to retroﬁt existing analog radiography

rooms; (b) the fact that the CR reader

and cassettes are separate entities

makes the technology extremely conve-

nient for bedside applications; and

sette makes it further possible to have

one (higher-cost) CR reader used in

reading multiple (lower-cost) CR cas-

settes, which allows one CR reader to

serve multiple radiography rooms and

reduces the up-front cost of the transi-

tion to digital radiography.

Flat-Panel Detectors

While CR technology currently has the

largest number of installations in the

digital radiography market, with its no-

table advantages in terms of cost and

utility for bedside applications, its dis-

advantages in terms of image quality per

unit dose and suboptimal workﬂow (ow-

Figure 5

Figure 5: Hurter and Drifﬁeld curves of analog

screen-ﬁlm (InSight; Eastman Kodak) and digital

(CR) systems. Detector signal is optical density for

screen-ﬁlm system and relative digital value for

the digital system.

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

668 Radiology: Volume 241: Number 3—December 2006

ing to the physical separation of the ac-

quisition and processing functions) pro-

vide opportunities for another notewor-

thy technology, the ﬂat-panel detector.

Made possible by advances in the

fabrication of ﬂat-panel displays for the

computer industry, ﬂat-panel detectors

are made of thin layers of amorphous

silicon thin-ﬁlm transistors (TFTs) de-

posited on a piece of glass. The TFT

layer is coupled with an x-ray absorptive

layer. Depending on the material, there

are two types of ﬂat-panel detector. In-

direct ﬂat-panel detectors use a phos-

phor screen, most commonly cesium io-

dide, to convert the x-rays to light pho-

tons, which are subsequently detected

by the photodiode array associated with

the TFT layer and are converted to

charge deposited in the capacitors asso-

ciated with each TFT (24,30). Direct

ﬂat-panel detectors use instead a photo-

conductor layer, most commonly amor-

phous selenium that converts the x-ray

energy directly to charge, which is

subsequently directed to the collecting

TFT-capacitor array through the appli-

cation of a strong electric ﬁeld (31,32).

After exposure, the charge on the ca-

pacitors is collected line by line and pixel

by pixel by using the associated gate and

data lines, forming the raw digital image

data for processing and display.

Flat-panel detectors are relatively

new, and there continue to be a number

of new developments in the technology.

Most noteworthy are the use of x-ray cap-

ture materials with higher inherent ab-

sorption efﬁciency, improvement in the

electronics to increase the frame rate and

bit-depth resolution of the images, and

use of ﬂexible substrates in place of glass

to enable more rugged and damage-resis-

tant detectors, which can perhaps make

these detectors practical for bedside ap-

plications (33).

Flat-panel detectors have three note-

worthy advantages compared with other

digital radiographic technologies. These

advantages can be summarized as fol-

lows:

1. With the use of structured phos-

phor in indirect detectors and the appli-

cation of an electric potential in direct

detectors, the x-ray–sensitive layer of

these detectors can be notably thicker

than that of competing technologies.

This makes it possible to have increased

x-ray detection efﬁciency with minimal

loss of resolution. Consequently, patient

dose can be reduced without degrada-

tion of image quality (34).

2. Flat-panel detectors can acquire

multiple images in a short time. Current

frame rates of 30 frames per second are

available in the most recent models.

This makes it possible to acquire multi-

ple images of the patient with different

techniques or from different directions

with minimal motion blur. This feature

facilitates the use of these detectors for

a host of advanced applications such as

tomosynthesis and dual-energy imaging.

3. With integration of the acquisi-

tion and processing units, ﬂat-panel de-

tectors offer improved workﬂow com-

pared with that of CR.

CCD- and CMOS-based Detectors

In the past few years, CCD and CMOS

cameras have provided an alternative

technology for the acquisition of digital

chest radiographs. With these detec-

tors, the x-ray energy is ﬁrst converted

to light within a phosphor layer. The

light is then directed to a single or a

multitude of CCD or CMOS cameras

that detect the light image and form the

radiograph (35). An important compo-

nent of these detectors is the coupling of

the phosphor layer and the camera.

Since most CCD and CMOS sensors are

limited in size, it is necessary to demag-

nify the original light image generated

on the phosphor screen so that it can be

entirely captured by the camera. This is

accomplished by using either a ﬁberop-

tic coupler or a lens system. In either

case there is a loss of efﬁciency, since

only a small fraction of the light photons

generated by the phosphor are detected

by the camera(s). Consequently, the in-

herent efﬁciency of these detectors is

limited (34).

New developments in these types of

detectors have been focused on more

efﬁcient coupling between the phosphor

and the camera, the use of larger sen-

sors to minimize demagniﬁcation, and

improvements in phosphor efﬁciency.

These advances have led to the develop-

ment of detectors with detection efﬁ-

ciencies notably higher than those of

previous-generation CCD or CMOS de-

vices. These detectors are generally less

expensive than other digital technolo-

gies.

Slot-Scan Technology

As noted earlier, scattered radiation is

of notable concern in chest radiogra-

phy. Over the years, various techniques

have been developed to reduce the con-

tribution of scattered photons to the x-

ray image, including the use of antiscat-

ter grids (36,37), which are used in

most current chest radiography sys-

tems, and air gaps (38). While these

techniques can reduce scatter substan-

tially (36), they also lead to increased

patient dose, as well as a reduced ﬁeld

of view in cases of an air gap. These

fundamental limitations can be over-

come with the use of scanning beam and

slit devices (37,39–42).

A recent commercial product has

taken advantage of slot-scan technology

to reduce the amount of scattered radi-

ation on digital chest radiographs (43).

The detector consists of a cesium iodide

scintillation layer ﬁberoptically coupled

to a series of linear CCDs. With no anti-

scatter grid in place, a narrow-fan x-ray

beam synchronized with the movement

of the detector assembly scans the pa-

tient. The image data are continuously

read from the CCDs as the patient is

scanned by using the time-integration

method (44,45). After scanning, the im-

age data are processed for optimal dis-

play.

The main advantage of this technol-

ogy is superior scatter rejection with

little effect on the detection of primary

radiation. This can notably enhance the

effective detection efﬁciency of the im-

aging system. Results from a recent

study (25) show that this enhancement

can more than compensate for the in-

herent efﬁciency limitation of CCD-

based detectors, leading to improved

image quality at reduced dose.

The technologies described in the

preceding paragraphs represent the

bulk of commercial offerings for digital

chest radiography. There are a number

of other technologies that are currently

under development, however, including

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

Radiology: Volume 241: Number 3—December 2006 669

those related to new sensor materials

and new detector designs such as photon-

counting devices. Commercial implemen-

tation of these technologies awaits further

development.

Image-processing Developments

Image data acquired with radiographic

detectors cannot be viewed without ad-

ditional processing and proper display.

In the case of screen-ﬁlm images, the

ﬁlm must be chemically processed and

is eventually displayed on a light box. In

the case of digital images, digital pro-

cessing is required before a clinician can

view the images. Since most of the re-

cent developments in chest radiography

have focused on digital modalities, this

section will outline the current state of

digital image-processing methods for

optimal presentation of digital images,

as well as hardware components of soft-

copy display devices. Readers inter-

ested in the current state of ﬁlm-pro-

cessing techniques are advised to con-

sult two publications focused on that

topic (20,46).

In recent years, the increasingly dig-

ital nature of chest radiography, as well

as advances in computer technology,

have facilitated the application of com-

puterized image analysis for improv-

ing the clinical efﬁcacy of chest radio-

graphs. As a general rule, computerized

medical image analysis aims to reduce

both the perceptual and the cognitive

errors that burden the diagnostic inter-

pretation process. Consequently, the

latest image-processing developments in

chest radiography fall into two general

application areas: (a) techniques to im-

prove visual presentation and soft-copy

reading of radiographs and (b) techniques

to automate diagnostic interpretation of

such images. For the most part, these

techniques are application dependent.

Prior to display, each digital image

commonly undergoes a series of process-

ing steps (Fig 6). Broadly speaking, these

processes can be divided into two parts:

preprocessing and postprocessing.

Image Preprocessing

Image preprocessing consists of two com-

ponents: correction and scaling (47). Ne-

cessitated by the intrinsic nonuniformity

of digital detectors, the ﬁrst type of pro-

cessing includes image corrections for de-

tector defects or nonuniformities often

present on raw digital images. The non-

uniformities include stationary patterns

of the detector, thickness nonuniformities

of the capture element of the detector,

pixel dropouts, “dead” (inactive) pixels,

dead columns and rows, and dark current

variations. The corrections often require

periodic ﬂat-ﬁeld calibration processes

for the detector. These include one-di-

mensional shading corrections in the

case of CR and linear CCD-based detec-

tors and two-dimensional ﬂat-ﬁeld cor-

rection for ﬂat-panel detectors and two-

dimensional CCD- and CMOS-based

systems. In either method, an averaged,

normalized ﬂat-ﬁeld image (map) is

used to normalize the original image to

remove the patterns.

The second type of preprocessing is

scaling. One of the advantages of digital

radiography is the wide range of expo-

sures within which the detector is able

to provide a consistent response. This

enables the acquisition of high-latitude

chest radiographs with less susceptibil-

ity to over- and underexposure. How-

ever, the dynamic range of the detector

is beyond the perception capability of

the human visual system, and full-data

presentation leads to considerable con-

trast reduction. Identiﬁcation of the an-

atomically relevant range of exposures

is essential for optimal display of the

image.

This task is typically achieved through

two steps. First, the collimated area is

segmented to identify the exposed area

and thus exclude unexposed areas from

further analysis. Second, the anatomi-

cally relevant range of exposures in the

exposed area is identiﬁed. The most

common technique for this task is histo-

gram analysis of the image data. On the

basis of the expected general distribu-

tion of pixel values in the anatomic area

of interest, the image system marks the

range of the detector signal of interest

to be used for postprocessing.

Image Postprocessing

Once the raw image data are corrected

for inherent “ﬂaws” and the useful range

of image data is identiﬁed, the data un-

dergo image postprocessing. Digital ra-

diography has utilized various digital

postprocessing algorithms for enhanced

image display of chest radiographs.

Broadly speaking, these algorithms can

be categorized into three types: gray-

scale processing, edge enhancement,

and multifrequency processing (48,49).

Gray-scale processing.—This pro-

cess involves the conversion of detector

signal values to display values. In this

process, the display intensities of an im-

age are changed by means of either a

look-up table or windowing and level-

ing. Most systems employ a response

look-up table similar to the Hurter and

Drifﬁeld response of screen-ﬁlm sys-

tems, so that digital chest radiographs

look similar to conventional ﬁlm images.

Edge enhancement.—This process

aims to enhance ﬁne details within the

image by manipulating the high-fre-

quency content of the radiograph, most

commonly by using a variant of the un-

sharp masking technique in which a

blurred version of the image is formed,

and a fraction of the resultant image is

subtracted from to the original image.

The process is commonly used to com-

pensate for the lower inherent resolu-

tion performance of CR images. How-

ever, its use is often balanced against

the implied enhancement of quantum

mottle within the image and the unusu-

ally textured appearance of the lungs.

Multifrequency processing.—While

edge enhancement offers only a simplis-

tic modiﬁcation of spatial frequencies

on the radiograph, multifrequency pro-

cessing involves a more ﬂexible manipu-

lation of multiple portions of the fre-

quency spectrum. The image is initially

decomposed into multiple frequency

components. The component images

are then weighted and added back to-

gether. If the processing parameters are

set optimally, the resultant image can

compress the overall dynamic range of

the image while at the same time en-

hance local contrast. This is of particu-

lar utility in chest radiography, where

the details of opaque regions of the im-

age are made visible with no compro-

mise in the contrast in the lung regions.

The widely used MUSICA multiscale im-

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

670 Radiology: Volume 241: Number 3—December 2006

age-processing package offered by Agfa

(Greenville, SC) is an example of this

type of processing (50).

While these algorithms have en-

abled ﬂexibility in the presentation qual-

ity of chest radiographs, most chest ra-

diography systems are set up to provide

digital chest radiographs that mimic the

appearance of screen-ﬁlm images, with

varying degrees of success. While it is

possible to use the appearance ﬂexibility

of digital radiographs to optimize the

system in terms of diagnostic perfor-

mance, that task has remained largely

untackled. One reason, perhaps, has

been the large number of abnormalities

of interest on chest radiographs, each of

which might need to be optimized inde-

pendently. Furthermore, there is uncer-

tainty as to the extent that image post-

processing algorithms can improve the

presentation of subtle lesions without

similar enhancement of anatomic pat-

terns and increased false-positive rates.

Display Developments

Once processed, the digital chest radio-

graph must be viewed by a radiologist.

Soft-copy display is an essential element

of contemporary digital chest radiogra-

phy, because many advantages of digital

imaging cannot be realized without soft-

Figure 6

Figure 6: Effects of various (a– c) gray-scale and (d–f) equalization postprocessing schemes on appearance of a posteroanterior chest radiograph. Gray-scale pro-

cessing improves contrast of image features with a corresponding reduction in latitude (ie, range of data that is properly displayed). Equalization, on the other hand, pro-

vides improved visualization of details without a corresponding reduction in latitude. (Image courtesy of M. J. Flynn, PhD, Henry Ford Hospital, Detroit, Mich.)

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

Radiology: Volume 241: Number 3—December 2006 671

copy display. The conventional method

to display digital radiographs has been

on cathode-ray tubes, which currently

still dominate the market (51). More

recently, however, active-matrix liquid-

crystal displays are rapidly replacing

cathode-ray tubes in many facilities (52–

54). The advantages of liquid-crystal dis-

plays include improved resolution, re-

duced weight, smaller form factor, re-

duced reﬂection, improved bit depth, and

improved luminance range (55), although

their disadvantages in terms of angular

response and structured noise have not

yet been fully addressed (56–58).

Another recent trend in soft-copy

display has been the increased accep-

tance of color monitors, some of which

have shown acceptable technical perfor-

mance (59) for radiographic applica-

tions. With cathode-ray tubes, a mono-

chrome monitor offers important ad-

vantages over a color monitor in terms

of image quality due to improved bright-

ness, reduced glare, reduced reﬂection,

and improved resolution (55). There-

fore, color cathode-ray tubes have not

demonstrated adequate performance

for clinical use. Current color liquid-

crystal displays, however, do not have

the same drawbacks as do color cathode-

ray tubes, other than reduced brightness,

which can be tolerated given the fact that

most current liquid-crystal displays have

better luminance response. The use of

color monitors offers the advantage of be-

ing able to accommodate applications

other than image viewing on the same

device, with workﬂow and multitasking

advantages. Color monitors would also

make it possible to take advantage of

color for viewing multidimensional chest

images on the same display. Current

commercial offerings also have enabled

the luminance calibration of color dis-

plays to the gray-scale standard display

function (60). It is thus expected that

color liquid-crystal displays will gradually

replace the monochrome monitors in

clinical practice.

Application Developments

As noted earlier, discerning subtle le-

sions in chest radiographs is made difﬁ-

cult by the conﬂuence of anatomy that

overlies the lesion. An early attempt at

overcoming poor lesion conspicuity

was equalization radiography (61,62).

Equalization was developed to improve

the visibility of lesions in dense regions

of the chest such as the mediastinum

and the retrocardiac and retrodiaphrag-

matic areas. With screen-ﬁlm radiogra-

phy, these areas of the chest were typi-

cally underexposed and showed poor le-

sion contrast owing to the shape of the

Hurter and Drifﬁeld curve of the ﬁlm.

Equalization modulated the incident in-

tensity onto the patient in such a way

that a higher exposure was delivered to

the dense regions, and a lower exposure

was delivered to the unobscured lungs.

This process forced all of the thoracic

regions into the higher-contrast “linear”

portion of the ﬁlm response curve. Stud-

ies indicated signiﬁcantly improved de-

tection of lesions when equalization was

used (62,63). Despite its promise, equal-

ization has largely disappeared from clin-

ical use because of the edge-enhanced

appearance imposed on the images,

which some found difﬁcult to interpret,

and also because of the more recent

transition to digital imaging and away

from screen-ﬁlm radiography.

The advent of digital chest radiogra-

phy in the 1980s enabled the develop-

ment of new techniques to improve the

detection of subtle lesions. These tech-

niques included algorithms, typically

coupled with methodological innova-

tions that used some aspect of imaging

physics to improve conspicuity, and/or

image subtraction strategies. Three no-

table techniques are dual-energy imag-

Figure 7

Figure 7: Dual-energy subtraction radiography in healthy middle-aged woman. Conventional (a) posteroanterior, (b) bone-subtracted, and (c) soft-tissue–subtracted

images are normal except for scoliosis. Note improved depiction of vascular anatomy in b.

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

672 Radiology: Volume 241: Number 3—December 2006

ing, temporal subtraction imaging, and

digital tomosynthesis. All of these tech-

niques are implemented by using a con-

ventional chest radiography system cou-

pled with a digital imaging receptor.

Dual-Energy Subtraction Imaging

Dual-energy subtraction imaging can be

used to generate images of two indepen-

dent tissue types, most commonly bone

and soft tissue. The dual-energy tech-

nique distinguishes bone from soft tis-

sue by using the known energy depen-

dence of x-ray attenuation in soft tissue

and in bone. Calciﬁed structures attenu-

ate far more heavily, by means of photo-

electric absorption, than do soft-tissue

structures; thus, the contrast of calcium

diminishes with increases in beam en-

ergy much faster than does the contrast

of soft tissue. Thus, if two images are

acquired at different beam energies, the

image obtained at the lower energy will

show a larger fraction of contrast from

bone than from soft tissue. These two

images may be combined in such a way

that the soft-tissue or calcium compo-

nents can be exactly isolated. Typically,

an image containing only calciﬁed struc-

tures and an image containing only soft-

tissue structures are generated (Fig 7).

Thus, dual-energy subtraction radiogra-

phy can improve lung nodule conspicu-

ity by eliminating overlying anatomic

noise from the bones. The technique

can also be used to better demonstrate

calcium in lesions (64–68).

Although the dual-energy technique

was proposed as early as the 1950s (69), it

was not until practical digital radiographic

detectors became available in the 1980s

that dual-energy imaging was clinically fea-

sible for chest applications. There cur-

rently are two commercially available

methods for accomplishing dual-energy

imaging. These methods are based on dif-

ferent types of detectors. The ﬁrst

method, based on CR storage-phosphor

plates, was developed in the late 1980s

(70–72). It generates low- and high-en-

ergy images by exposing a sandwich of

two CR plates with a copper ﬁlter be-

tween. The ﬁrst plate records the stan-

dard chest image at a spectral beam

energy typical of conventional chest ra-

diography. The second plate records a

higher mean spectral beam energy, due

to beam hardening that occurs in the

intervening copper ﬁlter; this second

plate has worse image noise, however,

due to the loss of beam ﬂux from the

copper ﬁlter. The low- and high-energy

images are recorded simultaneously

with the CR dual-energy method, with a

single x-ray exposure to the patient.

The second commercially available

dual-energy method uses ﬂat-panel de-

tectors (73). In this embodiment, a

sandwich detector conﬁguration is not

practical, so two separate x-ray expo-

sures are made of the patient. The ﬁrst

exposure is at a standard voltage for

chest radiography (typically, 120 kV),

and the second is acquired at a lower

voltage (typically, 60 kV). These images

are separated in time by up to several

hundred milliseconds, so there is the

potential for some motion of patient

anatomy between exposures, thus giv-

ing rise to some potential edge artifacts

in the resulting tissue and bone im-

ages. However, the image quality of the

ﬂat-panel dual-energy images is much

higher than that of the CR embodiment,

owing to better energy separation be-

tween the low- and high-energy beams,

higher x-ray ﬂux in the high-energy im-

age, and better detective quantum efﬁ-

ciency of the ﬂat-panel detector, relative

to those of CR. Image postprocessing is

accomplished with the ﬂat-panel dual-

energy images to mitigate anatomic mis-

registration between the low- and high-

energy images, but some slight edge ef-

fects still persist (Fig 8).

Considerable attention has been

given to optimizing the acquisition and

processing methods for dual-energy im-

aging. The effects of scattered radiation

have been considered (71,74–76), as

has the importance of accurate account-

ing for beam hardening in the separa-

tion of tissue and bone (77,78). It is also

important to include some type of post-

processing for noise suppression, re-

gardless of which dual-energy method is

used, because of the inherent decrease

in signal-to-noise ratio with dual-energy

relative to the ratio with conventional

chest radiography (79). Some of these

processing methods have been summa-

rized in a recent publication (80).

Dual-energy methods have been in

clinical use for several years, and there

are data in the literature that indicate

that dual-energy imaging can improve

detection and classiﬁcation of pulmo-

nary nodules. Early reports (65,67,

70,81) indicated statistically signiﬁcant

improvements in the detection of pul-

monary nodules and patterns of calciﬁ-

cation (Fig 9). Recent studies in which a

ﬂat-panel dual-exposure system (Revo-

lution XQ/i; GE Healthcare, Milwaukee,

Wis) was used showed the clinical efﬁ-

cacy of dual-energy subtraction for the

detection of calciﬁed chest abnormali-

ties (82) and noncalciﬁed pulmonary

nodules (83). For 37 calciﬁed chest le-

sions, radiologists’ sensitivity signiﬁ-

cantly increased from 36% to 66%,

while speciﬁcity remained constant at

73% (82). Statistically signiﬁcant but

smaller improvements in sensitivity,

speciﬁcity, and conﬁdence were also ob-

served for 59 noncalciﬁed pulmonary

nodules ranging from 0.2 to 2.5 cm in

diameter (83).

Dual-energy imaging has also been

investigated for other applications. It

has been shown (84), for example, to

improve detection of coronary artery

Figure 8

Figure 8: Soft-tissue subtracted image from

dual-energy chest radiograph in patient with right

aortic arch. Excess cardiac motion results in in-

complete subtraction of soft-tissue components

and considerable artifact (arrow) at left cardiac

border.

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

Radiology: Volume 241: Number 3—December 2006 673

and other cardiac calciﬁcations (Fig 10).

On the other hand, dual-energy imaging

does not seem to improve detection of

interstitial disease (85).

The clinical use of dual-energy tech-

niques in chest imaging is still evolving.

Some institutions advocate dual-energy

radiography for all chest examinations,

precisely because one does not know, a

priori, which patients will have previ-

ously undiagnosed pulmonary nodules.

Other institutions, however, have ap-

plied dual-energy imaging more selec-

tively to subsets of patients in whom

nodules are more likely. Additional clin-

ical experience will likely be necessary

before a consensus is reached on the

best practices for the use of dual-energy

techniques in chest imaging.

Temporal Subtraction Imaging

Another way to improve the visual as-

sessment of chest radiographs is with

temporal image subtraction. Temporal

Figure 9

Figure 9: Dual-energy subtraction radiography in a healthy middle-aged man. (a) Conventional posteroanterior radiograph shows possible lung nodule (arrow) over-

lying right upper lobe. (b) Bone-subtracted image shows small soft-tissue nodule (arrow) in left lung apex, not seen on a. No nodule is seen in right upper lobe. (c) Soft-

tissue–subtracted image conﬁrms that nodule seen on a represents calciﬁcation at ﬁrst costochondral junction (arrow).

Figure 10

Figure 10: Dual-energy subtraction radiography in an elderly man. (a) Soft-tissue–subtracted posteroanterior image shows coronary artery calciﬁcation (arrow) not

clearly identiﬁed on (b) conventional posteroanterior image. (Image courtesy of R. C. Gilkeson, MD, University Hospital, Cleveland, Ohio.)

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

674 Radiology: Volume 241: Number 3—December 2006

subtraction techniques aim to selec-

tively enhance areas of interval change

by subtracting the patient’s previous ra-

diograph from the current one (86).

The quality of the difference image

strongly depends on the success of the

two-dimensional registration and warp-

ing of the two radiographs so that the

effect of patient positioning variation is

minimized (87–89). Generally, the dif-

ference image appears uniformly gray in

areas of no change. Areas that stand out

in the uniform gray background indicate

interval change (Fig 11). Several studies

(91–98) have shown that temporal sub-

traction improves the visual perception

of subtle abnormalities such as pulmo-

nary nodules, inﬁltrative opacities, and

diffuse lung disease. A 20% reduction in

the average reading time when tempo-

ral subtraction is used was noted as well

(91). Currently, temporal subtraction is

commercially available in Japan (99).

Digital Tomosynthesis

Section imaging is another method for im-

proving detection of subtle lesions such as

pulmonary nodules. Traditional geomet-

ric tomography, which produces a single

section image at a time, has been known

since the 1930s (100). In recent years,

however, with the advent of multisection

CT and other three-dimensional imaging

modalities, conventional geometric to-

mography has fallen out of favor and is

used today only in limited cases of excre-

tory urography and some skeletal imaging

applications. The difﬁculties with conven-

tional geometric tomography were that

only a single section could be acquired at

a time, and, if multiple sections were de-

sired, considerable positioning time and

patient dose were required. Digital tomo-

synthesis is a technique that has evolved

from conventional tomography and solves

many of the problems associated with the

earlier technique.

Digital tomosynthesis can produce

an unlimited number of section images

at arbitrary depths from a single set of

acquisition images. A digital detector,

conventional x-ray tube, and computer-

controlled apparatus to move the x-ray

tube are used; during motion of the

tube, a series of projection radiographs

are acquired, and the anatomy at differ-

ent depths in the patient changes orien-

tation in the projection images owing to

parallax. These projection images are

then shifted and added to bring into fo-

cus objects in a given plane. By varying

the amount of shift, different plane

depths can be reconstructed. Objects

outside of the focus plane are rendered

with varying amounts of blur.

A variety of reconstruction ap-

proaches have been investigated for to-

mosynthesis, although the simple shift-

and-add approach is the most common.

It is important to use a deblurring algo-

Figure 11

Figure 11: Temporal subtraction radiography in a

middle-aged patient. (a) Initial posteroanterior radio-

graph was interpreted as normal, even when com-

pared with (b) that from examination performed 1 year

previously. (c) Subtraction image suggests new

masses in both upper lobes (arrows), conﬁrmed on

(d, e) transverse CT scans. Biopsy revealed synchro-

nous lung cancers. (Reprinted, with permission, from

reference 90.)

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

Radiology: Volume 241: Number 3—December 2006 675

rithm to eliminate the residual blur

from structures overlying the planes of

interest. Various approaches for deblur-

ring have been investigated, including ma-

trix-inversion tomosynthesis (101–105),

ﬁltered back projection (106,107), and it-

erative restoration (108). These deblur-

ring methods produce section images

with excellent rendition of anatomy and

effective elimination of structures outside

the section of interest.

Tomosynthesis has been applied

to such diverse applications as angio-

graphic, dental, orthopedic, breast, and

chest imaging (104); currently, the ar-

eas receiving the most clinical and re-

search interest are breast and chest im-

aging. Investigators in our laboratory

are currently conducting an National In-

stitutes of Health–funded clinical trial of

the efﬁcacy of tomosynthesis for im-

proving the detection of pulmonary nod-

ules. As part of this effort, a series of

optimization studies were conducted to

determine the best acquisition and re-

construction parameters for chest imag-

ing applications. It was found that 20° of

vertical motion, 71 projection images,

and 69 reconstructed planes are best

(109). Spacing of reconstructed sec-

tions is typically 3–5 mm. Several adja-

cent sections are averaged to produce a

slab image with much better image qual-

ity than a single thin section, but the

spacing of these sliding-average slabs

remains about 5 mm, sufﬁcient to easily

depict nodules of several millimeters in

diameter. Total x-ray exposure for ac-

quisition of the tomosynthesis projec-

tion images is about equal to that for

one lateral screen-ﬁlm radiograph.

In a pilot study of 20 human subjects

(110), tomosynthesis provided subjec-

tively far superior visualization of nod-

ules, vasculature, airways, and ribs than

did conventional posteroanterior chest

radiographs (Figs 12–14). In the pilot

study, preliminary evaluation revealed

considerable improvement in the detec-

tion of CT-conﬁrmed nodules with digi-

tal tomosynthesis compared with detec-

tion with conventional posteroanterior

radiographs, but these ﬁndings must be

conﬁrmed quantitatively in the larger

ongoing National Institutes of Health

trial. Early indications are that tomo-

synthesis may offer an improvement

over conventional radiography in the vi-

sualization of chest anatomy, particu-

larly pulmonary nodules, at a lower ra-

diation dose than with CT. This tech-

nique may also prove to be less expensive

than CT because of the anticipated cost of

soon-to-be-available commercial equip-

ment. Tomosynthesis is not envisioned as

a replacement for CT (owing to its re-

duced resolution in the depth direction),

but it does appear to offer potential for

improved diagnosis when used as an ad-

junct to conventional chest radiography.

A commercial chest tomosynthesis prod-

uct is being readied for market.

CAD and CADx Systems

Digital image data manipulation paved

the way for the automated diagnostic

interpretation of chest radiographs

as well. CAD and CADx technology

emerged 15 years ago and has slowly but

steadily made its way into the clinical

arena. This rapidly expanding research

ﬁeld shifts the focus from human percep-

tion to automated decision making. The

clinical role of CAD and CADx technology

is highly debated and continuously evolv-

ing. Currently, CAD technology has a

complementary role in clinical practice

as a second opinion, as long as radiolo-

gists understand the capabilities and

limitations of CAD (99). Any image-

based decision (human or computer-

ized) is always limited by the fundamen-

tal constraints of the imaging modality.

From an engineering point of

view, CAD and CADx systems typically

rely on a carefully selected combination

of elaborate image-processing, pattern-

recognition, and artiﬁcial intelligence

techniques. Thus far, the application of

CAD and CADx analysis in chest radiog-

raphy has followed a traditional hierar-

chic model of ﬁrst detecting and then

characterizing potential abnormalities

(Fig 15) (111). Initially, image-process-

ing algorithms are applied to identify

regions of interest that appear suspi-

cious according to predeﬁned clinical

expectations. Subsequently, detailed im-

age feature analysis seeks to capture the

morphologic and textural characteristics

of the candidate regions. Finally, feature-

based decision analysis is implemented

to provide a deﬁnitive assessment of the

candidate regions. The most important

among the advances in decision-making

analysis is the application of artiﬁcial

intelligence techniques such as artiﬁcial

neural networks and knowledge-based

systems.

The overwhelming majority of chest

radiography CAD applications involve

the detection of pulmonary nodules.

There are numerous publications on the

topic (112). The proposed CAD tech-

niques involve painstakingly optimized

combinations of image-processing algo-

rithms (ie, gray-level thresholding, mul-

tiresolution analysis, spatial ﬁltering,

template matching, morphologic and

textural analysis, model-based analysis)

with statistical or artiﬁcial intelligence–

based decision models. Typically, mor-

phology-based image processing is ap-

plied to detect nodular-appearing struc-

tures, while more detailed morphologic

and texture analyses follow to eliminate

false-positive nodule candidates. The ﬁ-

nal decision is made by applying a linear

classiﬁer, a neural network, or a rule-

base algorithm that carefully merges the

image ﬁndings into a ﬁnal binary deci-

sion regarding the presence of a nodule

at a particular image location.

Owing to lack of benchmark image

databases, the reported results vary

substantially, and direct comparison is

impossible because of differences in the

size and difﬁculty level of the private

image data sets. Regardless, for all tech-

niques it is a struggle to maintain a clin-

ically acceptable sensitivity level while

reducing the number of false-positive

detections generated because of over-

lapping ribs or vessels. Several labora-

tory observer studies have tested the

clinical potential of CAD in a comple-

mentary role for lung nodule detection.

Reported results (97,99,113–118) show

that CAD can assist physicians in im-

proving their overall detection rate for

lung nodules.

There is only one commercially

available CAD system in the United

States for the detection of pulmonary

nodules in both digitized and digitally

acquired chest radiographs. The Rapid-

Screen system (Riverain Medical, Mi-

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

676 Radiology: Volume 241: Number 3—December 2006

amisburg, Ohio) was initially approved

for clinical use by the Food and Drug

Administration (FDA) in July 2001. In

evaluation studies presented to the FDA

for premarket approval (119), it was

shown that the system helped radiolo-

gists improve their detection rate for

small lung nodules (9–14.5-mm diame-

ter) by 21%. Other commercial CAD

systems for lung nodule detection are

thus far available only in Japan (120).

Although not clinically available,

several investigators have explored the

feasibility of computerized image analy-

sis for determining the malignancy sta-

tus of pulmonary nodules. The overall

aim is to develop CADx systems that

could potentially reduce the number of

unnecessary diagnostic CT studies.

Some laboratory observer studies have

shown promising results for such an ap-

plication (96,121).

Another popular application of CAD

in chest radiography is in the detection

and differential diagnosis of interstitial

lung disease (ILD). Both tasks are con-

sidered clinically difﬁcult and are often

burdened with subjective assessment

and lack of quantiﬁable descriptions.

Unlike the case with pulmonary nod-

ules, the radiographic manifestations of

ILD are diverse, and its visual percep-

tion and differentiation rely more

heavily on texture than on morphology.

As a consequence, researchers have fo-

cused their attention on applying so-

Figure 12

Figure 12: Images in middle-aged woman with history of right partial mastec-

tomy for breast cancer who presented for routine follow-up (same patient as in Fig

4). (a) Digital tomosynthesis section image of whole chest and (b) magniﬁed sec-

tion images of right lower lobe clearly show right lower lobe nodule (arrows).

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

Radiology: Volume 241: Number 3—December 2006 677

phisticated texture-analysis techniques

as the foundation of CAD and CADx

systems targeting ILD (112). Results of

limited observer studies suggest the

clinical potential of these systems for

substantial improvement in radiologists’

performance (96,122–124). However,

more work is needed to establish the

clinical usefulness of these CAD sys-

tems. Finally, other less-well-studied

CAD applications in chest radiography

involve the detection of cardiomegaly

(125), pneumothorax (126,127), inter-

val changes (128), and tuberculosis

(129).

Given the increasingly digital nature

of chest radiography, CAD will most

certainly be an integral part of clinical

practice. However, the existing para-

digm of CAD playing a complementary

role for speciﬁc clinical tasks is rather

limited. Beyond improving diagnostic

accuracy, there is high demand regard-

ing general diagnostic tasks (130), in-

terpretive capabilities, interactive na-

ture, and individualized guidance in pa-

tient care (eg, optimal timing for follow-

Figure 13

Figure 13: Images in middle-aged man. (a) Conventional posteroanterior radio-

graph, (b) digital tomosynthesis section image of whole chest, and (c) magniﬁed

section images of right hilum show 15-mm nodule (arrows) overlying right hilum.

Despite relatively large size of this nodule, it cannot be seen on a owing to superim-

position of hilar vessels.

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

678 Radiology: Volume 241: Number 3—December 2006

up). Consequently, CAD is expected to

keep evolving to meet the increasing

new challenges of chest radiography.

Clinical Perspective

Over the past decade there have been

remarkable advances in the technology

applied to conventional thoracic imag-

ing. CR has rapidly become the stan-

dard for bedside chest imaging in the

United States. Direct radiography has

rapidly replaced ﬁxed ﬁlm-based chest

units in most U.S. academic radiology

departments and is rapidly replacing

such units in the community setting as

well. The impetus for these changes can

be largely attributed to the advantages

inherent in digital imaging. Consistently

high-quality images and a marked re-

duction in the number of repeat exami-

nations required because of technical

causes have made digital imaging a fa-

vorite of physicians and technologists

alike. The ability to readily incorporate

digital images into a picture archiving

and communication system (PACS) al-

lows for much improved communica-

tion between radiologists and referring

physicians and has eliminated the prob-

lem of the “lost ﬁlm.” The improved

availability of critical images and en-

hanced communication provided through

Figure 14

Figure 14: Images in middle-aged man. (a) Conventional posteroanterior radiograph shows possible nodule (arrow) overlying the right ﬁrst rib. (b) Digital tomosyn-

thesis section images clearly show that this opacity is due to calciﬁcation (arrows) at ﬁrst costochondral junction.

Figure 15

Figure 15: Flow diagram illus-

trates operation of CAD system.

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

Radiology: Volume 241: Number 3—December 2006 679

PACS has markedly improved the efﬁ-

ciency of both inpatient and outpatient

health care operations. These improved

efﬁciencies, combined with the improved

satisfaction on the part of both physicians

and patients, have more than justiﬁed the

increased costs associated with these

newer technologies.

Perhaps the most exciting feature of

these new techniques is the evolving op-

portunity to further enhance the diag-

nostic capabilities and impact of digital

chest examinations. In particular, the

opportunities for direct CAD of various

lesions through sophisticated computer

programs, as discussed in this article,

offer the possibility of enhancing the ra-

diologist’s accuracy while at the same

time improving efﬁciency.

The introduction of newer, hereto-

fore unavailable techniques such as dig-

ital tomosynthesis again hold the prom-

ise of further revolutionizing what has

now become conventional digital chest

imaging. As discussed in this article, the

opportunity to provide tomographic im-

ages of the chest easily and routinely,

thus removing overlying and frequently

confusing background structures, again

holds the promise of substantial im-

provements in the diagnostic accuracy

of conventional chest imaging. Com-

bined with other digitally driven tech-

niques such as dual-energy imaging and

temporal subtraction imaging, the op-

portunities for dramatic improvement

in the diagnostic capabilities of conven-

tional digital chest imaging seem prom-

ising. Obviously, the effect of these new

technologies and of those still to come

will need rigorous scientiﬁc validation to

ensure that the reality truly fulﬁlls the

promise of these exciting new advance-

ments.

References

1. Ravin CE, Chotas HG. Chest radiography.

Radiology 1997;204(3):593– 600.

2. Centers for Disease Control and Prevention.

Overweight and obesity: obesity trends. At-

lanta, Ga: Centers for Disease Control and

Prevention, 2005.

3. Uppot R, Sahani D, Hahn P, Kalra M, Saini

S, Mueller P. Limited by body habitus: eco-

nomic and quality control issues in the abil-

ity of a radiology department to provide

diagnostic imaging to a fattening population

[abstract]. In: Radiological Society of North

America scientiﬁc assembly and annual

meeting program. Oak Brook, Ill: Radiolog-

ical Society of North America, 2004; 593.

4. Van Metter R. Describing the signal-trans-

fer characteristics of asymmetrical radio-

graphic screen-ﬁlm systems. Med Phys

1992;19:53–58.

5. Samei E, Saunders RS Jr. The nature of the

digital image. In: Goldman L, Yester MV,

eds. Speciﬁcations, performance evaluation

and quality assurance of radiographic and

ﬂuoroscopic systems in the digital era.

Madison, Wis: Medical Physics, 2004; 103–

123.

6. Floyd CE Jr, Baker JA, Lo JY, Ravin CE.

Measurement of scatter fractions in clinical

bedside radiography. Radiology 1992;183:

857–861.

7. Niklason LT, Sorenson JA, Nelson JA. Scat-

tered radiation in chest radiography. Med

Phys 1981;8:677– 681.

8. Samei E. Efects of anatomical structure on

signal detection. In: Beutel J, Kundel HL,

Van Metter RL, eds. Handbook of medical

imaging. Bellingham, Wash: International

Society for Optical Engineering, 2000; 655–

682.

9. Waring JJ, Wasson WW. The imperfec-

tions of the stereoscopic manoeuvre in ra-

diography of the chest. Radiology 1926;6:

198–203.

10. Stitik FP, Tockman MS. Radiographic

screening in the early detection of lung can-

cer. Radiol Clin North Am 1978;16:347–

366.

11. Kundel HL, Nodine CF, Carmody D. Visual

scanning, pattern recognition and decision-

making in pulmonary nodule detection. In-

vest Radiol 1978;13:175–181.

12. Kundel HL, Revesz G, Toto L. Contrast gra-

dient and the detection of lung nodules.

Invest Radiol 1979;14:18 –22.

13. Samei E, Flynn MJ, Eyler WR. Detection of

subtle lung nodules: relative inﬂuence of

quantum and anatomic noise on chest ra-

diographs. Radiology 1999;213:727–734.

14. Revesz G, Kundel HL, Graber MA. The in-

ﬂuence of structured noise on the detection

of radiologic abnormalities. Invest Radiol

1974;9:479– 486.

15. Samei E, Flynn MJ, Peterson E, Eyler WR.

Subtle lung nodules: inﬂuence of local ana-

tomic variations on detection. Radiology

2003;228:76– 84.

16. Kundel HL. Peripheral vision, structured

noise and ﬁlm reader error. Radiology

1975;114:269–273.

17. Carmody DP, Nodine CF, Kundel HL. An

analysis of perceptual and cognitive factors

in radiographic interpretation. Perception

1980;9:339–344.

18. Nodine CF, Krupinski E, Kundel HL. A per-

ceptually-based algorithm provides effec-

tive visual feedback to radiologists search-

ing for lung nodules. IEEE Proc Vis Biomed

Comput 1990;1:202–207.

19. Hessel SJ, Herman PG, Swensson RG. Im-

proving performance by multiple interpre-

tations of chest radiographs: effectiveness

and cost. Radiology 1978;127:589 –594.

20. Haus AG, Jaskulski S. Basics of ﬁlm pro-

cessing in medical imaging. Madison, Wis:

Medical Physics, 1997.

21. Van Metter R, Dickerson R. Objective per-

formance characteristics of a new asym-

metric screen-ﬁlm system. Med Phys 1994;

21:1483–1490.

22. Dobbins J. Image quality metrics for digital

systems. In: Beutel HKJ, Metter RV, eds.

Handbook of medical imaging. Washing-

ton, DC: International Society for Optical

Engineering, 2000; 163–222.

23. Samei E, Flynn MJ. An experimental com-

parison of detector performance for com-

puted radiography systems. Med Phys 2002;

29:447–459.

24. Samei E, Flynn MJ. An experimental com-

parison of detector performance for direct

and indirect digital radiography systems.

Med Phys 2003;30:608 –622.

25. Samei E, Lo JY, Yoshizumi TT, et al. Com-

parative scatter and dose performance of

slot-scan and full-ﬁeld digital chest radiog-

raphy systems. Radiology 2005;235:940 –

949.

26. Sonoda M, Takano M, Miyahara J, Kato H.

Computed radiography utilizing scanning

laser stimulated luminescence. Radiology

1983;148:833–838.

27. Fetterly KA, Schueler BA. Performance

evaluation of a “dual-side read” dedicated

mammography computed radiography sys-

tem. Med Phys 2003;30:1843–1854.

28. Schaetzing R. Computed radiography tech-

nology. In: Samei E, Flynn MJ, eds. 2003

Syllabus: categorical course in diagnostic

radiology physics—advances in digital radi-

ography. Oak Brook, Ill: Radiological Soci-

ety of North America, 2003; 7–22.

29. Leblans P, Struye L, Willems P. A new nee-

dle-crystalline computed radiography de-

tector. J Digit Imaging 2000;13:117–120.

30. Siewerdsen JH, Antonuk LE, el-Mohri Y, et

al. Empirical and theoretical investigation

of the noise performance of indirect detec-

tion, active matrix ﬂat-panel imagers

(AMFPIs) for diagnostic radiology. Med

Phys 1997;24:71– 89.

31. Lee DL, Cheung LK, Rodricks B, Powell

GF. Improved imaging performance of a

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

680 Radiology: Volume 241: Number 3—December 2006

14 ⫻ 17-inch direct radiography system us-

ing Se/TFT detector. In: Medical imaging

1998: physics of medical imaging, Feb

22–24 1998. San Diego, Calif: International

Society for Optical Engineering, 1998; 14 –

23.

32. Zhao W, Blevis I, Germann S, Rowlands

JA, Waechter D, Huang Z. Digital radiology

using active matrix readout of amorphous

selenium: construction and evaluation of a

prototype real-time detector. Med Phys

1997;24:1834–1843.

33. Yorkston J. Digital radiographic technol-

ogy. In: Samei E, Flynn MJ, eds. 2003

Syllabus: categorical course in diagnostic

radiology physics—advances in digital radi-

ography. Oak Brook, Ill: Radiological Soci-

ety of North America, 2003; 23–36.

34. Samei E. Performance of digital radiogra-

phy detectors: factors affecting sharpness

and noise. In: Samei E, Flynn MJ, eds. 2003

Syllabus: categorical course in diagnostic

radiology physics—advances in digital radi-

ography. Oak Brook, Ill: Radiological Soci-

ety of North America, 2003; 49– 61.

35. Yaffe MJ, Rowlands JA. X-ray detectors for

digital radiography. Phys Med Biol 1997;

42:1–39.

36. Sorenson JA, Niklason LT, Knutti DF. Per-

formance characteristics of improved anti-

scatter grids. Med Phys 1980;7:525–528.

37. Barnes GT, Moreland RF, Yester MV, Wit-

ten DM. The scanning grid: a novel and

effective bucky movement. Radiology 1980;

135:765–767.

38. Sorenson JA, Floch J. Scatter rejection by

air gaps: an empirical model. Med Phys

1985;12:308–316.

39. Sorenson JA, Nelson JA, Niklason LT, Ja-

cobsen SC. Rotating disk device for slit ra-

diography of the chest. Radiology 1980;

134:227–231.

40. Plenkovich D, Sorenson JA, Kruger RA.

Scatter rejection by electronic collimation.

Med Phys 1986;13:158 –163.

41. Barnes GT, Brezovich IA, Witten DM.

Scanning multiple slit assembly: a practical

and efﬁcient device to reduce scatter. AJR

Am J Roentgenol 1977;129:497–501.

42. Chotas HG, Floyd CE Jr, Dobbins JT 3rd,

Lo JY, Ravin CE. Scatter fractions in AM-

BER imaging. Radiology 1990;177:879 –

880.

43. Samei E, Saunders RS, Lo JY, et al. Funda-

mental imaging characteristics of a slot-

scan digital chest radiographic system. Med

Phys 2004;31:2687–2698.

44. Holdsworth DW, Gerson RK, Fenster A. A

time-delay integration charge-coupled de-

vice camera for slot-scanned digital radiog-

raphy. Med Phys 1990;17:876 – 886.

45. Mainprize JG, Ford NL, Yin S, Tumer T,

Yaffe MJ. A slot-scanned photodiode-ar-

ray/CCD hybrid detector for digital mam-

mography. Med Phys 2002;29:214 –225.

46. Haus AG. Advances in ﬁlm processing sys-

tems technology and quality control in med-

ical imaging. Madison, Wis: Medical Phys-

ics, 2001.

47. Seibert JA. Digital radiographic image

presentation: preprocessing methods. In:

Samei E, Flynn MJ, eds. 2003 Syllabus: cat-

egorical course in diagnostic radiology

physics—advances in digital radiography.

Oak Brook, Ill: Radiological Society of

North America, 2003; 53–70.

48. Prokop M, Neitzel U, Schaefer-Prokop C.

Principles of image processing in digital

chest radiography. J Thorac Imaging 2003;

18:148–164.

49. Flynn MJ. Processing digital radiographs of

speciﬁc body parts. In: Samei E, Flynn MJ,

eds. 2003 Syllabus: categorical course in

diagnostic radiology physics—advances in

digital radiography. Oak Brook, Ill: Radio-

logical Society of North America, 2003; 71–

78.

50. Vuylsteke P, Schoeters EP. Multiscale im-

age contrast ampliﬁcation (MUSICA). In:

Loew, MH, ed. Proceedings of SPIE: medi-

cal imaging 1994 —image processing. Vol

2167. Bellingham, Wash: International So-

ciety for Optical Engineering, 2001; 551–

560.

51. Uffmann M, Prokop M, Kupper W, Mang

T, Fiedler V, Schaefer-Prokop C. Soft-copy

reading of digital chest radiographs: effect

of ambient light and automatic optimization

of monitor luminance. Invest Radiol 2005;

40:180–185.

52. Oschatz E, Prokop M, Scharitzer M, Weber

M, Balassy C, Schaefer-Prokop C. Compar-

ison of liquid crystal versus cathode ray

tube display for the detection of simulated

chest lesions. Eur Radiol 2005;15:1472–

1476.

53. Scharitzer M, Prokop M, Weber M, Fuch-

sjager M, Oschatz E, Schaefer-Prokop C.

Detectability of catheters on bedside chest

radiographs: comparison between liquid

crystal display and high-resolution cathode-

ray tube monitors. Radiology 2005;234(2):

611–616.

54. Balassy C, Prokop M, Weber M, Sailer J,

Herold CJ, Schaefer-Prokop C. Flat-panel

display (LCD) versus high-resolution gray-

scale display (CRT) for chest radiography:

an observer preference study. AJR Am J

Roentgenol 2005;184:752–756.

55. Samei E, Badano A, Chakraborty D, et al.

Assessment of display performance for

medical imaging systems. Madison, Wis:

Medical Physics, 2005.

56. Badano A, Gagne RM, Jennings RJ, Drilling

SE, Imhoff BR, Muka E. Noise in ﬂat-panel

displays with subpixel structure. Med Phys

2004;31:715–723.

57. Badano A, Flynn MJ, Martin S, Kanicki J.

Angular dependence of the luminance and

contrast in medical monochrome liquid

crystal displays. Med Phys 2003;30:2602–

2613.

58. Samei E, Wright SL. Effect of viewing angle

response on DICOM compliance of liquid

crystal displays. In: Ratib OM, Huang HK,

eds. Proceedings of SPIE: medical imaging

2004—PACS and imaging informatics. Vol

5371. Bellingham, Wash: International So-

ciety for Optical Engineering, 2004; 170 –

177.

59. Averbukh AN, Channin DS, Flynn MJ. As-

sessment of a novel, high-resolution, color,

AMLCD for diagnostic medical image

display: luminance performance and

DICOM calibration. J Digit Imaging 2003;

16:270–279.

60. Samei E, Badano A, Chakraborty D, et al.

Assessment of display performance for

medical imaging systems: executive sum-

mary of AAPM TG18 report. Med Phys

2005;32:1205–1225.

61. Plewes DB, Wandtke JC. Improved lung

nodule detection with an equalized image.

In: Gray JE, Hendee WR, eds. Proceedings

of SPIE: application of optical instrumenta-

tion in medicine VIII. Vol 233. Bellingham,

Wash: International Society for Optical En-

gineering, 1980; 183–189.

62. Kool LJ, Busscher DL, Vlasbloem H, et al.

Advanced multiple-beam equalization radi-

ography in chest radiology: a simulated

nodule detection study. Radiology 1988;

169:35–39.

63. Wandtke JC, Plewes DB, McFaul JA. Im-

proved pulmonary nodule detection with

scanning equalization radiography. Radiol-

ogy 1988;169:23–27.

64. Kamimura R, Takashima T. Clinical appli-

cation of single dual-energy subtraction

technique with digital storage-phosphor ra-

diography. J Digit Imaging 1995;8:21–24.

65. Kelcz F, Zink FE, Peppler WW, Kruger

DG, Ergun DL, Mistretta CA. Conventional

chest radiography vs dual-energy computed

radiography in the detection and character-

ization of pulmonary nodules. AJR Am J

Roentgenol 1994;162:271–278.

66. Kido S, Ikezoe J, Naito H, et al. Clinical

evaluation of pulmonary nodules with sin-

gle-exposure dual-energy subtraction chest

radiography with an iterative noise-reduc-

tion algorithm. Radiology 1995;194:407–

412.

67. Oestmann JW, Greene R, Rhea JT, et al.

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

Radiology: Volume 241: Number 3—December 2006 681

“Single exposure” dual-energy digital radi-

ography in the detection of pulmonary nod-

ules and calciﬁcations. Invest Radiol 1989;

24(7):517–521.

68. Hartman TE. Dual-energy radiography. Se-

min Roentgenol 1997;32:45– 49.

69. Jacobson B. Dichromatic absorption radi-

ography: dichromatography. Acta Radiol

1953;39:437–452.

70. Ishigaki T, Sakuma S, Ikeda M. One-shot

dual-energy subtraction chest imaging with

computed radiography: clinical evaluation

of ﬁlm images. Radiology 1988;168:67–72.

71. Ergun DL, Mistretta CA, Brown DE, et al.

Single-exposure dual-energy computed ra-

diography: improved detection and pro-

cessing. Radiology 1990;174:243–249.

72. Stewart BK, Huang HK. Single-exposure

dual-energy computed radiography. Med

Phys 1990;17:866 –875.

73. Sabol JM, Avinash GB, Nicolas F, Claus BE,

Zhao J, Dobbins JT III. Development and

characterization of a dual-energy subtrac-

tion imaging system for chest radiography

based on CsI:Tl amorphous silicon ﬂat-

panel technology. In: Antonuk LE, Yaffe

MJ, eds. Proceedings of SPIE: medical im-

aging 2001—physics of medical imaging.

Vol 4320. Bellingham, Wash: International

Society for Optical Engineering, 2001;

399– 408.

74. Shaw CG, Plewes DB. Effects of scattered

radiation and veiling glare in dual-energy

tissue-bone imaging: a theoretical analysis.

Med Phys 1987;14:956 –967.

75. Kruger DG, Zink F, Peppler WW, Ergun

DL, Mistretta CA. A regional convolution

kernel algorithm for scatter correction in

dual-energy images: comparison to single-

kernel algorithms. Med Phys 1994;21:175–

184.

76. Hinshaw DA, Dobbins JT III. Recent

progress in noise reduction and scatter cor-

rection in dual-energy imaging. In: Van

Metter RL, Beutel J, eds. Proceedings of

SPIE: medical imaging 1995—physics of

medical imaging. Vol 2432. Bellingham,

Wash: International Society for Optical En-

gineering, 1995; 134 –142.

77. Alvarez RE, Macovski A. Energy-selective

reconstructions in x-ray computerized to-

mography. Phys Med Biol 1976;21:733–

744.

78. Cardinal HN, Fenster A. An accurate

method for direct dual-energy calibration

and decomposition. Med Phys 1990;17:

327–341.

79. Warp RJ, Dobbins JT 3rd. Quantitative

evaluation of noise reduction strategies in

dual-energy imaging. Med Phys 2003;30:

190–198.

80. Dobbins JT III, Warp RJ. Dual-energy

methods for tissue discrimination in chest

radiography. In: Samei E, Flynn MJ, eds.

2003 Syllabus: categorical course in diag-

nostic radiology physics—advances in digi-

tal radiography. Oak Brook, Ill: Radiologi-

cal Society of North America, 2003; 173–

180.

81. Fraser RG, Hickey NM, Niklason LT, et al.

Calciﬁcation in pulmonary nodules: detec-

tion with dual-energy digital radiography.

Radiology 1986;160:595– 601.

82. Fischbach F, Freund T, Rottgen R, Engert

U, Felix R, Ricke J. Dual-energy chest radi-

ography with a ﬂat-panel digital detector:

revealing calciﬁed chest abnormalities. AJR

Am J Roentgenol 2003;181:1519 –1524.

83. Ricke J, Fischbach F, Freund T, et al. Clin-

ical results of CsI-detector-based dual-ex-

posure dual energy in chest radiography.

Eur Radiol 2003;13:2577–2582.

84. Gilkeson RC, Novak RD, Sachs P. Digital

radiography with dual-energy subtraction:

improved evaluation of cardiac calciﬁca-

tion. AJR Am J Roentgenol 2004;183:

1233–1238.

85. Barnes GT, Sabbagh EA, Chakraborty DP,

et al. A comparison of dual-energy digital

radiography and screen-ﬁlm imaging in the

detection of subtle interstitial pulmonary

disease. Invest Radiol 1989;24:585–591.

86. Kano A, Doi K, MacMahon H, Hassell D,

Giger M. Digital image subtraction of tem-

porally sequential chest images for detec-

tion of interval change. Med Phys 1994;21:

453–461.

87. Ishida T, Ashizawa K, Engelmann R, Kat-

suragawa S, MacMahon H, Doi K. Applica-

tion of temporal subtraction for detection

of interval changes on chest radiographs:

improvement of subtraction images using

automated initial image matching. J Digit

Imaging 1999;12:77– 86.

88. Ishida T, Katsuragawa S, Nakamura K,

MacMahon H, Doi K. Iterative image warp-

ing technique for temporal subtraction of

sequential chest radiographs to detect in-

terval change. Med Phys 1999;26:1320 –

1329.

89. Zhao H, Lo SCB, Freedman MT, Mun SK.

Automatic temporal subtraction of chest

radiographs and its enhancement for lung

cancers. In: Sonka M, Hanson KM, eds.

Proceedings of SPIE: medical imaging

2001—image processing. Vol 4322. Belling-

ham, Wash: International Society for Opti-

cal Engineering, 2001; 1867–1872.

90. MacMahon H. Dual-energy and temporal

subtraction digital chest radiography. In:

2003 Syllabus: categorical course in diag-

nostic radiology physics—advances in digi-

tal radiography. Oak Brook, Ill: Radiologi-

cal Society of North America, 2003; 181–

188.

91. Difazio MC, Macmahon H, Xu XW, et al.

Digital chest radiography: effect of tempo-

ral subtraction images on detection accu-

racy. Radiology 1997;202:447– 452.

92. MacMahon H, Engelmann RM, Roe C,

Behlen FM, Hoffmann KR, Doi K. Use of

temporal subtraction for detection of early

lung cancer: results of observer test [ab-

stract]. Radiology 1998;209(P):543.

93. Nakagawa K, Oosawa A, Tanaka H, Oh-

tomo K. Clinical effectiveness of improved

temporal subtraction for digital chest radio-

graphs. In: Chakraborty D, Krupinski EA,

eds. Proceedings of SPIE: medical imaging

2002—image perception, observer perfor-

mance, and technology assessment. Vol

4686. Bellingham, Wash: International So-

ciety for Optical Engineering, 2002; 319 –

325.

94. Kakeda S, Nakamura K, Kamada K, et al.

Improved detection of lung nodules by us-

ing a temporal subtraction technique. Radi-

ology 2002;224:145–151.

95. Uozumi T, Nakamura K, Watanabe H, et al.

ROC analysis of detection of metastatic pul-

monary nodules on digital chest radio-

graphs with temporal subtraction. Acad Ra-

diol 2001;8:871– 878.

96. Abe H, MacMahon H, Engelmann R, et al.

Computer-aided diagnosis in chest radiog-

raphy: results of large-scale observer tests at

the 1996–2001 RSNA scientiﬁc assemblies.

RadioGraphics 2003;23:255–265.

97. Tsubamoto M, Johkoh T, Kozuka T, et al.

Temporal subtraction for the detection of

hazy pulmonary opacities on chest radiog-

raphy. AJR Am J Roentgenol 2002;179:

467–471.

98. Okazaki H, Nakamura K, Watanabe H, et

al. Improved detection of lung cancer aris-

ing in diffuse lung diseases on chest radio-

graphs using temporal subtraction. Acad

Radiol 2004;11:498 –505.

99. Johkoh T, Kozuka T, Tomiyama N, et al.

Temporal subtraction for detection of solitary

pulmonary nodules on chest radiographs:

evaluation of a commercially available com-

puter-aided diagnosis system. Radiology

2002;223:806– 811.

100. Ziedses des Plantes BG. Eine neue methode

zur differenzierung in der roentgenogra-

phie (planigraphie). Acta Radiol 1932;000:

182–192.

101. Warp RJ, Godfrey DJ, Dobbins JT III. Ap-

plications of matrix inverse tomosynthesis.

In: Dobbins JT III, Boone JM, eds. Proceed-

ings of SPIE: medical imaging 2000—phys-

ics of medical imaging. Vol 3977. Belling-

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

682 Radiology: Volume 241: Number 3—December 2006

ham, Wash: International Society for Opti-

cal Engineering, 2000; 376 –383.

102. Godfrey DJ, Warp RJ, Dobbins JT III. Opti-

mization of matrix inverse tomosynthesis.

In: Antonuk LE, Yaffe MJ, eds. Proceedings

of SPIE: medical imaging 2001—physics of

medical imaging. Vol 4320. Bellingham,

Wash: International Society for Optical En-

gineering, 2001; 696 –704.

103. Godfrey DJ, Rader A, Dobbins JT III. Prac-

tical strategies for the clinical implementa-

tion of matrix inversion tomosynthesis

(MITS). In: Yaffe MJ, Antonuk LE, eds. Pro-

ceedings of SPIE: medical imaging 2003—

physics of medical imaging. Vol 5030. Belling-

ham, Wash: International Society for Optical

Engineering, 2003; 379 –390.

104. Dobbins JT 3rd, Godfrey DJ. Digital x-ray

tomosynthesis: current state of the art and

clinical potential. Phys Med Biol 2003;48:

R65–R106.

105. Dobbins JT III, Godfrey DJ, McAdams HP.

Chest tomosynthesis. In: Samei E, Flynn

MJ, eds. 2003 Syllabus: categorical course

in diagnostic radiology physics—advances

in digital radiography. Oak Brook, Ill: Ra-

diological Society of North America, 2003;

211–226.

106. Lauritsch G, Harer WH. Theoretical frame-

work for ﬁltered back projection in tomo-

synthesis. In: Hanson KM, ed. Proceedings

of SPIE: medical imaging 1998—image pro-

cessing. Vol 3338. Bellingham, Wash: In-

ternational Society for Optical Engineering,

1998; 1127–1137.

107. Stevens GM, Fahrig R, Pelc NJ. Filtered

backprojection for modifying the impulse

response of circular tomosynthesis. Med

Phys 2001;28:372–380.

108. Ruttimann UE, Groenhuis RA, Webber RL.

Restoration of digital multiplane tomosyn-

thesis by a constrained iteration method.

IEEE Trans Med Imaging 1984;3:141–148.

109. Godfrey DJ, McAdams HP, Dobbins JT

3rd. Optimization of the matrix inversion

tomosynthesis (MITS) impulse response

and modulation transfer function charac-

teristics for chest imaging. Med Phys 2006;

33:655–667.

110. McAdams HP, Godfrey DJ, Dobbins JT III.

Digital tomosynthesis for improved lung

nodule detection: initial clinical experience

[abstract]. In: Radiological Society of North

America scientiﬁc assembly and annual

meeting program. Oak Brook, Ill: Radiolog-

ical Society of North America, 2003; 524.

111. Doi K. Overview on research and develop-

ment of computer-aided diagnostic schemes.

Semin Ultrasound CT MR 2004;25:404 –410.

112. van Ginneken B, ter Haar Romeny BM,

Viergever MA. Computer-aided diagnosis

in chest radiography: a survey. IEEE Trans

Med Imaging 2001;20:1228 –1241.

113. MacMahon H, Engelmann R, Behlen FM, et

al. Computer-aided diagnosis of pulmonary

nodules: results of a large-scale observer

test. Radiology 1999;213:723–726.

114. Kobayashi T, Xu XW, MacMahon H, Metz

CE, Doi K. Effect of a computer-aided diag-

nosis scheme on radiologists’ performance

in detection of lung nodules on radiographs.

Radiology 1996;199:843– 848.

115. Shiraishi J, Abe H, Engelmann R, Doi K.

Effect of high sensitivity in a computerized

scheme for detecting extremely subtle soli-

tary pulmonary nodules in chest radiographs:

observer performance study. Acad Radiol

2003;10:1302–1311.

116. Marten K, Seyfarth T, Auer F, et al. Com-

puter-assisted detection of pulmonary

nodules: performance evaluation of an ex-

pert knowledge-based detection system in

consensus reading with experienced and in-

experienced chest radiologists. Eur Radiol

2004;14(10):1930–1938.

117. Freedman MT, Osicka T, Lo SCB, et al.

Computer-aided detection of lung cancer

on chest radiographs: effect of machine CAD

true positive/false negative detections on ra-

diologists’ conﬁdence level. In: Chakraborty

D, Eckstein MP, eds. Proceedings of SPIE:

medical imaging 2004—image perception,

observer performance, and technology as-

sessment. Vol 5372. Bellingham, Wash: In-

ternational Society for Optical Engineering,

2004; 185–191.

118. Osicka T, Freedman MT, Lo SCB, et al.

Computer-aided detection of lung cancer

on chest radiographs: differences in the in-

terpretation time of radiologist’s showing

vs. not showing improvement with CAD.

In: Chakraborty D, Krupinski EA, eds. Pro-

ceedings of SPIE: medical imaging 2003—

image perception, observer performance,

and technology assessment. Vol 5034. Bel-

lingham, Wash: International Society for

Optical Engineering, 2003; 483– 494.

119. Freedman M. Improved small volume lung

cancer detection with computer-aided detec-

tion: database characteristics and imaging of

response to breast cancer risk reduction

strategies. Ann N Y Acad Sci 2004;1020:175–

189.

120. Kakeda S, Moriya J, Sato H, et al. Im-

proved detection of lung nodules on chest

radiographs using a commercial computer-

aided diagnosis system. AJR Am J Roent-

genol 2004;182:505–510.

121. Shiraishi J, Abe H, Engelmann R, Aoyama

M, MacMahon H, Doi K. Computer-aided

diagnosis to distinguish benign from malig-

nant solitary pulmonary nodules on ra-

diographs: ROC analysis of radiologists’

performance—initial experience. Radiology

2003;227:469– 474.

122. Monnier-Cholley L, MacMahon H, Katsura-

gawa S, Morishita J, Ishida T, Doi K. Com-

puter-aided diagnosis for detection of inter-

stitial opacities on chest radiographs. AJR

Am J Roentgenol 1998;171:1651–1656.

123. Ashizawa K, MacMahon H, Ishida T, et al.

Effect of an artiﬁcial neural network on ra-

diologists’ performance in the differential

diagnosis of interstitial lung disease using

chest radiographs. AJR Am J Roentgenol

1999;172:1311–1315.

124. Abe H, Ashizawa K, Li F, et al. Artiﬁcial

neural networks (ANNs) for differential di-

agnosis of interstitial lung disease: results

of a simulation test with actual clinical

cases. Acad Radiol 2004;11:29 –37.

125. Nakamori N, Doi K, MacMahon H, Sasaki

Y, Montner S. Effect of heart-size parame-

ters computed from digital chest radio-

graphs on detection of cardiomegaly: po-

tential usefulness for computer-aided diag-

nosis. Invest Radiol 1991;26:546 –550.

126. Abe K, Doi K, MacMahon H, et al. Comput-

er-aided diagnosis in chest radiography: pre-

liminary experience. Invest Radiol 1993;28:

987–993.

127. Sanada S, Doi K, MacMahon H. Image fea-

ture analysis and computer-aided diagnosis

in digital radiography: automated detection

of pneumothorax in chest images. Med

Phys 1992;19:1153–1160.

128. Ko JP, Betke M. Chest CT: automated nod-

ule detection and assessment of change

over time—preliminary experience. Radiol-

ogy 2001;218:267–273.

129. van Ginneken B, Katsuragawa S, Romeny

BM, Doi K. Automatic detection of abnor-

malities in chest radiographs using local

texture analysis. IEEE Trans Med Imaging

2002;21:139–149.

130. Summers RM. Road maps for advancement

of radiologic computer-aided detection in the

21st century [editorial]. Radiology 2003;229:

11–13.

STATE OF THE ART: Recent Advances in Chest Radiography McAdams et al

Radiology: Volume 241: Number 3—December 2006 683

Postoperative Extremity Tomosynthesis-A Superimposition-Free Alternative to Standard Radiography?

Article

Full-text available

May 2024
INVEST RADIOL

Rationale and Objectives This study investigates the performance of tomosynthesis in the presence of osteosynthetic implants, aiming to overcome superimposition-induced limitations in conventional radiograms. Materials and Methods After surgical fracture induction and subsequent osteosynthesis, 8 cadaveric fracture models (wrist, metacarpus, ankle, metatarsus) were scanned with the prototypical tomosynthesis mode of a multiuse x-ray system. Tomosynthesis protocols at 60, 80, and 116 kV (sweep angle 10°, 13 FPS) were compared with standard radiograms. Five radiologists independently rated diagnostic assessability based on an equidistant 7-point scale focusing on fracture delineation, intra-articular screw placement, and implant positioning. The intraclass correlation coefficient (ICC) was calculated to analyze interrater agreement. Results Radiation dose in radiography was 0.48 ± 0.26 dGy·cm ² versus 0.12 ± 0.01, 0.36 ± 0.02, and 1.95 ± 0.11 dGy·cm ² for tomosynthesis scans at 60, 80, and 116 kV. Delineation of fracture lines was superior for 80/116 kV tomosynthesis compared with radiograms ( P ≤ 0.003). Assessability of intra-articular screw placement was deemed favorable for all tomosynthesis protocols ( P ≤ 0.004), whereas superiority for evaluation of implant positioning could not be ascertained (all P 's ≥ 0.599). Diagnostic confidence was higher for 80/116 kV tomosynthesis versus radiograms and 60 kV tomosynthesis ( P ≤ 0.002). Interrater agreement was good for fracture delineation (ICC, 0.803; 95% confidence interval [CI], 0.598–0.904), intra-articular screw placement (ICC, 0.802; 95% CI, 0.599–0.903), implant positioning (ICC, 0.855; 95% CI, 0.729–0.926), and diagnostic confidence (ICC, 0.842; 95% CI, 0.556–0.934). Conclusions In the postoperative workup of extremity fractures, tomosynthesis allows for superior assessment of fracture lines and intra-articular screw positioning with greater diagnostic confidence at radiation doses comparable to conventional radiograms.

Correction for X-ray Scatter and Detector Crosstalk in Dark-field Radiography

Article

Full-text available

Mar 2024
IEEE T MED IMAGING

Dark-field radiography, a new X-ray imaging method, has recently been applied to human chest imaging for the first time. It employs conventional X-ray devices in combination with a Talbot-Lau interferometer with a large field of view, providing both attenuation and dark-field radiographs. It is well known that sample scatter creates artifacts in both modalities. Here, we demonstrate that also X-ray scatter generated by the interferometer as well as detector crosstalk create artifacts in the dark-field radiographs, in addition to the expected loss of spatial resolution. We propose deconvolution-based correction methods for the induced artifacts. The kernel for detector crosstalk is measured and fitted to a model, while the kernel for scatter from the analyzer grating is calculated by a Monte-Carlo simulation. To correct for scatter from the sample, we adapt an algorithm used for scatter correction in conventional radiography. We validate the obtained corrections with a water phantom. Finally, we show the impact of detector crosstalk, scatter from the analyzer grating and scatter from the sample and their successful correction on dark-field images of a human thorax.

Tomosynthesis of the Appendicular Skeleton on a Twin Robotic X-ray System: A Cadaveric Fracture Study

Article

Full-text available

Mar 2024
ACAD RADIOL

Rationale and objectives: Aiming to offset image quality limitations in radiographs due to superimposition, this study investigates the diagnostic potential of appendicular skeleton tomosynthesis. Materials and methods: Eight cadaveric extremities (four hands and feet) were examined employing the prototypical tomosynthesis mode of a twin robotic X-ray scanner. 12 protocols with varying sweep angles (10, 20 vs. 40°), frame rates (13 vs. 26 fps), and tube voltages (60 vs. 80 kV) were compared to radiographs. Four radiologists separately evaluated cortical and trabecular bone visualization and fracture patterns. Interreader reliability was assessed based on the intraclass correlation coefficient (ICC). Results: Radiation dose in radiography was 0.59 ± 0.20 dGy * cm2 versus 0.11 ± 0.00 to 2.46 ± 0.17 dGy * cm2 for tomosynthesis. Cortical bone display was inferior for radiographs compared to 40° and 20° tomosynthesis. Best results were ascertained for the 80 kV/40°/26 fps protocol. Trabecular bone depiction was also superior in tomosynthesis (p ≤ 0.009) and best with the 80 kV/10°/26 fps setting. Interreader reliability was moderate for cortical bone display (ICC 0.521, 95% confidence interval 0.356-0.641) and good for trabecular bone (0.759, 0.697-0.810). Diagnostic accuracy for articular involvement and multifragment situations was higher in tomosynthesis (93.8-100%/92.2-100%) vs. radiography (85.9%/82.8%). Diagnostic confidence was also better in tomosynthesis (p ≤ 0.003). Conclusion: Compared to radiography, tomosynthesis allows for superior assessability of cortical and trabecular bone and fracture morphology, especially at high framerates. Operating on a multipurpose X-ray system, tomosynthesis of the appendicular skeleton can be performed without additional scanner hardware.

Artificial Intelligence in Chest Radiology: Advancements and Applications for Improved Global Health Outcomes

Article

Full-text available

Jan 2024

Purpose of Review In this review, we explored the most recent innovations in artificial intelligence and their applications in chest radiology to promote population health. Recent Findings AI algorithms have revolutionized imaging modalities essential for the diagnosis and management of thoracic pathologies such as pneumonia, lung cancer, interstitial lung disease, and cardiovascular disease. AI implementation in low- and middle-income countries has the potential to significantly improve chest radiology services but restricted access to adequate resources and infrastructure must be addressed to ensure globalization of AI in underserved areas. Summary The advancement of AI models enhanced the field of chest radiology and improved performance of health systems. Efforts to expand AI accessibility globally will further improve health care.

A Novel Artificial Intelligence System for the Prediction of Interstitial Lung Diseases

Article

Full-text available

Jan 2024

Interstitial lung disease (ILD) encompasses a spectrum of more than 200 fatal lung disorders affecting the interstitium, contributing to substantial mortality rates. The intricate process of diagnosing ILDs is compounded by their diverse symptomatology and resemblance to other pulmonary conditions. High-resolution computed tomography (HRCT) assumes the role of the primary diagnostic tool for ILD, playing a pivotal role in the medical landscape. In response, this study introduces a computational framework powered by artificial intelligence (AI) to support medical professionals in the identification and classification of ILD from HRCT images. Our dataset comprises 3045 HRCT images sourced from distinct patient cases. The proposed framework presents a novel approach to predicting ILD categories using a two-tier ensemble strategy that integrates outcomes from convolutional neural networks (CNNs), transfer learning, and machine learning (ML) models. This approach outperforms existing methods when evaluated on previously unseen data. Initially, ML models, including Logistic Regression, BayesNet, Stochastic Gradient Descent (SGD), RandomForest, and J48, are deployed to detect ILD based on statistical measures derived from HRCT images. Notably, the J48 model achieves a notable accuracy of 93.08%, with the diagnostic significance of diagonal-wise standard deviation emphasized through feature analysis. Further refinement is achieved through the application of Marker-controlled Watershed Transformation Segmentation and Morphological Masking techniques to HRCT images, elevating accuracy to 95.73% with the J48 model. The computational framework also embraces deep learning techniques, introducing three innovative CNN models that achieve test accuracies of 94.08%, 92.04%, and 93.72%. Additionally, we evaluate five full-training and transfer learning models (InceptionV3, VGG16, MobileNetV2, VGG19, and ResNet50), with the InceptionV3 model achieving peak accuracy at 78.41% for full training and 92.48% for transfer learning. In the concluding phase, a soft-voting ensemble mechanism amplifies training outcomes, yielding ensemble test accuracies of 76.56% for full-training models and 92.81% for transfer learning models. Notably, the ensemble comprising the three newly introduced CNN models attains the pinnacle of test accuracy at 97.42%. This research is poised to drive advancements in ILD diagnosis, presenting a resilient computational framework that enhances accuracy and ultimately betters patient outcomes within the medical domain.

Diagnostic performance of chest CT average intensity projection (AIP) reconstruction for the assessment of pleuro-parenchymal abnormalities

Article

Apr 2024
CLIN RADIOL

Design optimization of a triple-layer flat-panel detector for three-material decomposition

Conference Paper

Apr 2024

Beyond the Norm: A Modified VGG-16 Model for COVID-19 Detection

Article

Full-text available

Jan 2023

Correlation Between Contrast-Detail Analysis and Clinical Image Quality Assessment of Intrapulmonary Lesions in Dual-Energy Subtraction Chest Radiography Using the Two-Shot Method: A Phantom Study

Article

Full-text available

Nov 2023
ACAD RADIOL

The Importance of Positioning in Left Lateral Chest X-Ray Examination

Article

Aug 2023

Pyong-Kon Cho

Performance of digital radiography detectors: Factors affecting sharpness and noise

Article

Jan 2003

E. Samei

Advanced multiple beam equalization radiography (AMBER) in chest radiography: Comparison with conventional chest images by means of a simulated nodule detection study

Article

Jan 1988
RADIOLOGY

Digital radiography technology

Article

Mar 2005

Digital radiography has broad field of application and various X-ray detector systems have become available for medical and industrial applications. This paper briefly introduces the recent digital radiography technology. Two direct and indirect methods of detection processing as well as their advantages and disadvantages for some digital radiography systems are also discussed.

A regional convolution kernel algorithm for scatter correction in dual-energy images: Comparison to single-kernel algorithms

Article

Jan 1999

David G. Kruger

Single kernel scatter correction algorithms are based on the model that the scatter field can be predicted by convolution of the primary intensity (I prim) with a spatially invariant scatter point‐spread function (PSF). Practical limitations (I prim unknown) suggest the substitution of the total detected intensity (I det) for I prim as the source image in the convolution. In regions of high scatter fraction (SF), I det is a poor approximation of I prim, thereby causing an overestimation of scatter originating in the region. This contributes to errors in estimating detected scatter in the mediastinum and neighboring regions. A technique using a regionally variable point‐spread function that significantly reduces RMS error in estimation of the primary image as compared to the single PSF method is investigated. The regionally variable convolution method employs a larger PSF in the mediastinum and a smaller PSF in the lungs to reduce the error in estimating the scatter throughout the image. The method to allow for patient differences has also been expanded and various implementations of these methods have been compared. Results show that the dual‐kernel algorithm is always more effective than an equivalent single‐kernel algorithm. The dual‐kernel algorithm using a predicted scatter fraction curve gives an overall RMS error in the primary of as low as 20.8% which is equivalent to 8.7% RMS error in the scatter. The dual‐kernel method using a predicted scatter fraction curve approaches the accuracy of the single‐kernel method using patient specific scatter measurements. Because using individual scatter measurements is a less desirable method for clinical use, we feel that the dual‐kernel algorithm which uses two region specific convolution kernels and a variable scatter fraction curve is the preferable method.

Assessment of a novel, high resolution, color, AMLCD for diagnostic medical image display: Luminance performance and DICOM calibration

Conference Paper

May 2003
Proceedings of SPIE

This paper documents the results of the first in a series of experiments designed to evaluate the suitability of a novel, high resolution, color, digital, LCD panel for diagnostic quality, gray scale image display. The goal of this experiment was to measure the performance of this display especially with respect to luminance. A DICOM Part 14 calibration of the panel was done using both a pure gray look up table and a color look up table. The panel evaluated was the IBM T221 22.2" backlit AMLCD display with native resolution of 3840 x 2400 pixels. Taking advantage of the color capabilities of the workstation, we were able to create a 256 entry grayscale calibration look up table derived from a palette of 1786 nearly gray luminance values. We also constructed a 256 entry gray-scale calibration look up table derived from a palette of 256 true gray values for which the red, green, and blue values were equal. For the DICOM calibration derived from the 256 palette, 45 of the 256 shades of gray were redundant. For the DICOM calibration derived from a 1786 palette, all shades of gray Were different and the luminance change between each gray level accurately agreed with the DICOM part 14 standard. These calibrations will now be used in our subsequent evaluation of human contrast-detail perception on this panel.

Image feature analysis and computer-aided diagnosis in digital radiography: Automated detection of pneumothorax in chest images

Article

Sep 1992

Shigeru Sanada

In order to aid radiologists in the diagnosis of pneumothorax from chest radiographs, an automated method for detection of subtle pneumothorax is being developed. The computerized method is based on the detection of a fine curved‐line pattern, which is a unique feature of radiographic findings of pneumothorax. Initially, regions of interest (ROIs) are determined in each upper lung area, where subtle pneumothoraces commonly appear. The pneumothorax pattern is enhanced by the selection of edge gradients within a limited range of orientations. Rib edges included in this edge‐enhanced image are removed, based on the locations of posterior ribs that are determined separately. A subtle curved line due to pneumothorax is then detected by means of the Hough transform. The detected pneumothorax pattern is marked on the chest image displayed on a CRT monitor. With the present computer method applied to 50 chest images (28 normals and 22 abnormals with pneumothorax), we were able to detect 77% of pneumothoraces, with 0.44 false‐positives per image.

Improved Lung Nodule Detection With An Equalized Image

Conference Paper

Aug 1980

A custom designed equalizing filter was used to test the potential for improving nodule detectability in chest radiographs. The filter was constructed to produce a radiograph so that the film densities over the lung field and mediastinum were approximately equal. Eighteen radiographs were prepared with small lesions placed over the lung field and mediastinal areas of a chest phantom and the films were read by four radiologists. The films prepared with the equalizing filter showed an increased true positive rate of 18% in the retrocardiac, retrodiaphragmatic and mediastinal areas and only a slight decrease of the true positive rate in the lung field (4%). The average false positive rate for films prepared with and without the filter were virtually identical.

Digital Tomosynthesis for Improved Lung Nodule Detection: Initial Clinical Experience

Conference Paper

Dec 2003

Holman Page Mcadams

HTML Purpose: The ability of conventional chest radiography to depict small lung nodules remains limited despite recent advances. Digital tomosynthesis allows reconstruction of multiple image planes from a limited set of projection data. It could replace or supplement standard chest radiography as it is comparable in exam time and dose, can be done using available digital equipment, and is 10-100 times lower dose than CT. We report initial clinical results from a NIH-sponsored trial to evaluate the ability of digital tomosynthesis to improve small lung nodule detection. Methods and Materials: All patients scheduled for CT to follow-up lung nodules were eligible. Exclusion criteria included > 5 nodules, large (>1.5-cm) or small nodules (< 3-mm), pregnancy and inability to give consent. Five subjects have been enrolled to date. All received a digital PA chest radiograph. Digital tomosynthesis was performed with a commercial CsI/a-Si flat-panel detector (GE Medical Systems) installed with a custom-built computer-controlled tube mover. Images were reconstructed using an algorithm (matrix inversion tomosynthesis) developed in our lab and were acquired with various parameters. Total angle of tube movement varied from 10 - 20 degrees. The number of projection and reconstructed images varied from 21 - 61 and 19 - 59, respectively. Reconstruction with 59 images gave plane spacing of about 5-mm. All projection images were acquired in about 10 sec. Total dose per study was about equal to that of a screen-film lateral chest radiograph. Results: CT was used to confirm size and position of all nodules as well as presence of calcification. Tomosynthesis data from one subject was unusable due to computer malfunction. In the remaining four subjects, all calcified nodules � 1-2 mm and all non-calcified nodules � 4-mm were seen by tomosynthesis. In some cases, these nodules were not visible on the PA view. Non-calcified nodules < 4-mm were not seen by tomosynthesis. In one subject, a subtle nodule (7-mm lung cancer) was seen on the PA view, but was identified with greater clarity by tomosynthesis. This subject also had a second nodular opacity seen on the PA view that was correctly shown by tomosynthesis to be a vessel crossing. Tomosynthesis obtained with the widest tube angle (20 degrees) also showed superb visualization of pulmonary vasculature, particularly when viewed dynamically in a movie loop. Conclusion: Digital tomosynthesis shows promise for improving sensitivity and specificity of detection of small lung nodules, when compared to conventional chest radiography. (J.T.D. received a grant from General Electric; H.P.M. has a research agreement with GE Medical Systems.) Questions about this event email: mcada003@mc.duke.edu

Limited by Body Habitus: Economic and Quality Control Issues in the Ability of a Radiology Department to Provide Diagnostic Imaging to a Fattening Population

Conference Paper

Dec 2004

PURPOSE Obesity is a growing medical problem which can influence the ability of radiology department to provide optimum image quality and accurate diagnosis. This paper will address the economic and quality care impact of obesity on the ability of radiology departments to provide quality diagnostic imaging. METHOD AND MATERIALS A fourteen year retrospective review of dictated radiology reports with the disclaimer "limited due to body habitus" between the years 1989-2003 was performed from the electronic medical records.Percentages of "limited" radiographic studies were calculated per year. "Limited studies" were also classified based on modality and percentages per modality per year were calculated. Comparisons were made of the calculated percentages between modalities within a year and within each modality across the 5 years.Economic impact was assessed by calculating the cost estimates for the limited studies on a per modality basis, per year. Quality control issues were accessed by examining what technical factors and patient factors resulted in the limitations for each modality. RESULTS Overall, 0.15% of all studies were limited by body habitus. For all studies from 1989 through 2003, there has been an increase at a rate of 0.010% (95% CI 0.007-0.013%) per year (PThe study most dictated as "limited" is the abdominal ultrasound, followed by chest radiograph. 1.5% of ultrasounds were reported limted by body habitus. The rate of US cases limited by body habitus for 1989 through 2003 increased at a rate of 0.090% (95% CI 0.045-0.134) per year (PAlthough the direct economic impact of the "limited studies" is small, the steady rise over 14 years and the additional diagnostic tests and increased hospital stay to make up for the "limited study", also has an economic impact. A review of quality control shows technical factors available for each modality to improve image quality in obese patients. CONCLUSIONS Changes in the American body habitus over 14 years has increasingly affected the ability of radiology departments to provide quality images and accurate image interpretations.

Improved imaging performance of a 14"x17" direct radiography system using a Se/TFT detector

Article

Jul 1998
Proceedings of SPIE

Progress is discussed on the improvement of a Direct RadiographyTM solid state, flat panel, digital detector designed for use in general radiographic applications. This detector, now known as DirectRayTM, operates on the principle of direct detection of X-ray photons with a selenium photoconductor and consists of 500 micrometer thick amorphous selenium coupled to an amorphous silicon thin-film-transistor (TFT) readout array. This device is fabricated with a 14 X 17-inch (35 X 43-cm) active imaging area, corresponding to 2560 X 3072 pixels having dimensions of 139 micrometer X 139 micrometer and a geometrical fill factor of 86%. Improvements include a TFT array design upgrade with reduced noise characteristic, lower-noise readout electronics, and improved interfaces. Clinical radiographic images are currently being generated with the DirectRay detector using an X-ray exposure level equivalent to that of a 400 speed screen- film combination while maintaining the superior spatial resolution that is inherent in the direct conversion method. An effective sensor restoration technique has been implemented that eliminates the potential for selenium memory artifacts after a high dose. New results on NPS, MTF, DQE and signal linearity are presented. Detectability of low contrast objects using FAXiL test objects as well as the results of clinical studies are discussed.

Recent Advances in Chest Radiography 1

Abstract and Figures

Recommended publications

Artifacts in Digital Radiography

Quantitative analysis of epical root resorption by means of digital subtraction radiography

Novel display technique for reference images for visibility of temporal change on radiographs: Color...

Comparison of clinical and physical measures of image quality in chest and pelvis computed radiograp...