AJR 2001; 177:501-519
© American Roentgen Ray Society
High-Resolution CT of the Lungs
Ella A. Kazerooni1
1
Department of Radiology, 2910 Taubman Center, University of Michigan Medical
Center, 1500 E. Medical Center Dr., Ann Arbor, MI 48109-0326.
Received October 12, 2000;
accepted after revision March 19, 2001.
Address correspondence to E. A. Kazerooni.
Introduction
I first began to hear about high-resolution CT toward the end of my
radiology residency in 1991. I was new to radiology, and so was
high-resolution CT. In a way, that newness leveled the playing field between
residents and faculty; none of us were very sure what we were looking at. I
remember seeing new high-resolution CT images and scouring the literature to
figure out what we were seeing. When it was my turn to present at our weekly
Friday morning CT conference, I spent the entire hour presenting only
high-resolution CT cases to our group of thoracic and abdominal CT
radiologists and residents. Many squinted their eyes to see what was being
described. As a thoracic radiology fellow, I remember trying to figure out
what was a normal finding on high-resolution CT and what was not.
Now, a decade later, high-resolution CT continues to be an area of great
interest to me, and sometimes a source of considerable frustration. As I have
matured as a thoracic radiologist, so has the use of high-resolution CT. It is
part of my daily clinical work and research. My collaborations with my
pulmonary medicine and thoracic surgery colleagues in a specialized center of
research for interstitial lung disease, with our lung transplantation program
and lung volume reduction surgery, continue to raise questions about what
high-resolution CT can tell us about diffuse lung disease, and what more we
need to know to accurately diagnose and predict response to therapy and
survival in patients with diffuse lung disease. For many patients, their
disease will be the cause of their mortality, and both the disease and the
therapy itself, the source of morbidity.
The appearance of most lung diseases on high-resolution CT has already been
described, from asbestosis to Hermansky-Pudlak syndrome, and from sarcoidosis
to lysinuric protein intolerance
[1,2,3,4,5,6,7,8,9,10,11].
The accuracy of high-resolution CT for detecting disease and distinguishing
among diseases has been well documented, and the advantages over chest
radiography and conventional CT have been elucidated
[4,
12,13,14,15,16,17,18,19].
High-resolution CT is now being used extensively to evaluate the response of
lung disease to therapy and as a marker of underlying pathophysiology and
physiologic processes
[20,21,22,23].
New advances in CT technology may allow routine submillimeter scanning, total
lung volumetric high-resolution CT, dynamic CT throughout the respiratory
cycle (possible now only on ultrafast scanners), reproducible spirometric
gating of chest CT acquisitions to specific points in the respiratory cycle,
and computer-aided diagnosis
[24,25,26].
This article will review the history of high-resolution CT, technical
considerations in performing high-resolution CT, test characteristics of the
technique, and the clinical indications for high-resolution CT, and will
provide an overview of a pattern-based approach to interpretation. For readers
interested in more comprehensive resources on this subject, the third edition
of the textbook, High-Resolution CT of the Lung, by W. Richard Webb,
Nestor L. Müller, and David P. Naidich, is
recommended [27]; and the
second edition of the textbook, High-Resolution CT of the Chest:
Comprehensive Atlas, by Eric J. Stern and Stephen J. Swensen, is a richly
illustrated complementary resource
[28]. New editions of both
textbooks were published in early 2001.
Background
During the last two decades, high-resolution CT of the lungs has developed
into a mature technique for the evaluation of diffuse pulmonary parenchymal
abnormality. At its simplest, high-resolution CT is a sampling tool that
combines 1- to 2-mm thin-collimation CT images with a high-spatial-frequency
reconstruction algorithm to generate images that show exquisite lung detail.
The foundation for high-resolution CT began in 1975 with
radiologic—pathologic correlative studies of postmortem lungs, resulting
in a 1978 publication on small nodules, with special reference to
peribronchial nodules, in the American Journal of Roentgenology
[29]. The technique of
high-resolution CT for diffuse lung disease was initially described by Todo et
al. [30] from Kyoto
University, in 1982, for 21 patients with either diffuse panbronchiolitis,
lymphangitic spread of cancer, sarcoidosis, or tuberculosis. Their report in
the Japanese Journal of Clinical Imaging presented careful
correlation of the abnormalities seen on high-resolution CT images with the
abnormalities seen on corresponding inflation-fixed lung specimens. From this
beginning, the foundation of high-resolution CT interpretation has been, and
remains, radiologic—pathologic correlation and the relationship of
abnormalities to the architecture of the secondary pulmonary lobule.
Great interest in the technique followed the 1985 publication of an article
by Zerhouni el al. [31] in the
inaugural issue of the Journal of Thoracic Imaging, in which the
researchers described their 3-year experience with the technique and presented
for the first time an attempt to define patterns of abnormality that could be
used to classify diffuse pulmonary diseases. To this day, a pattern-based
approach to interpretation, coupled with the distribution of abnormality
throughout the pulmonary parenchyma, remains the key to differential
diagnosis. In 1990, the radiologist—pathologist team of Nestor L.
Müller and Roberta R. Miller published a
two-part manuscript [32,
33] on the use of CT in
chronic diffuse infiltrative lung diseases in the American Review of
Respiratory Disease; and by 1993, sufficient interest existed in
high-resolution CT that the Journal of Thoracic Imaging dedicated two
consecutive issues to the topic, guest-edited by
Müller
[34]. The material in those
two issues remains some of the best available. For readers interested in
becoming more familiar with the terminology and definitions applied to
high-resolution interpretations, the "Standardized Terms For
High-Resolution Computed Tomography of the Lung: A Proposed Glossary"
[35] appeared in the
Journal of Thoracic Imaging in 1993, and an illustrated glossary of
high-resolution CT terms appears in the textbook High-Resolution CT of the
Lung [27].
Anatomic and Technical Considerations
Anatomy
The interstitium of the lung can be divided into a central (or axial)
compartment that surrounds the bronchovascular bundles, and the peripheral (or
septal) interstitium that includes the interlobular septa and the subpleural
interstitium [36]. The
smallest anatomic unit visible on high-resolution CT is the secondary
pulmonary lobule (Fig. 1). The
walls of the lobules are the interlobular septa; they correspond to the Kerley
B lines seen on chest radiographs of patients with left heart failure and
interstitial edema. The interlobular septa are not usually visible unless
abnormal; at 0.1 mm thick, they are at the lower limit of high-resolution CT
resolution [37]. The
occasional visible septa may be normal. Structures of 0.2-0.3 mm can be
routinely identified on high-resolution CT when they are perpendicular to the
plane of imaging. The diameter of the pulmonary artery supplying each lobule
is 1 mm, and the diameter of the intralobular acinar arteries is 0.5 mm; both
are readily seen on high-resolution CT. Bronchi are visible, depending on the
thickness of their walls. The 1.0-mm-diameter bronchiole supplying the lobule
has an approximately 0.15-mm wall, just at the limit of high-resolution CT
resolution, and barely, if at all, visible on high-resolution CT images
[37,
38]. The major types of
abnormality involving the secondary pulmonary lobule are illustrated in
Figure 2.
View larger version (27K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1. —Line drawing of a secondary pulmonary lobule. Borders of
lobule are interlobular septa. At center of each lobule is a bronchiole and a
pulmonary artery (blue). Pulmonary vein (red) run in
interlobular septa. Lymphatics (green) are found in interlobular
septa and in central or axial interstitium that surrounds bronchovascular
bundles.
|
|
Thin Collimation and High-Spatial-Frequency Reconstruction
Algorithm
The two consistent components of high-resolution CT technique include the
use of thin collimation, usually 1-2 mm, coupled with a high-spatial-frequency
reconstruction algorithm. These technical adaptations to conventional chest CT
are designed to improve spatial resolution and thereby improve the ability to
detect small structures and subtle abnormalities such as thick interlobular
septa, cyst walls, small nodules, ground-glass opacity, and bronchiectasis
[39] (Figs.
3A,3B
and
4A,4B).
Thin collimation reduces partial volume averaging from adjacent structures and
in particular from adjacent aerated lung tissue; and the sharp algorithm
reduces the image smoothing that is characteristic of standard or soft-tissue
reconstruction algorithms in order to increase spatial resolution, at the
expense of increased image noise
[39,
40]. In larger patients, the
noise can usually, but not always, be countered by increasing the tube current
used for scanning (Fig.
5A,5B).
View larger version (122K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3B. —63-year-old man with asbestosis and pleural plaques resulting
from exposure to asbestos. 1.5-mm collimation high-resolution CT scan
reformatted using high-spatial-frequency reconstruction algorithm obtained at
same level shows pleural plaques. However, thickened inter-and intralobular
septa of asbestosis (arrowheads) are more clearly seen on B.
On A, it is difficult to distinguish partial volume averaging adjacent
to pleural plaques from lung abnormality.
|
|
View larger version (104K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 4A. —57-year-old man with obliterative bronchiolitis of chronic
lung transplant rejection with normal chest radiograph. Conventional CT scan
through lung bases shows subtle areas of ground-glass opacity
(arrows), representing partial volume averaging of bronchial
walls.
|
|
View larger version (109K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 4B. —57-year-old man with obliterative bronchiolitis of chronic
lung transplant rejection with normal chest radiograph. High-resolution CT
scan at same anatomic level as A shows diffuse cylindrical
bronchiectasis. Signet ring sign of bronchiectasis is illustrated
(arrowheads).
|
|
View larger version (161K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 5A. —59-year-old obese woman who underwent high-resolution CT that
was nondiagnostic because of patient's size. High-resolution CT scan is
degraded by extensive noise and is uninterpretable.
|
|
View larger version (121K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 5B. —59-year-old obese woman who underwent high-resolution CT that
was nondiagnostic because of patient's size. Scout topogram from CT
examinations reveals patient's body size. Although in most obese patients
increasing scanning technique can improve image quality, in very obese
patients to do so is not possible.
|
|
Image Spacing
High-resolution CT is a sampling examination of the lung, in which thin
sections are taken at staggered intervals, revealing both the pattern and the
distribution of abnormality, so that a differential diagnosis—or
sometimes a single diagnosis—can be rendered
[33]. There is a tendency to
think that with greater sampling or more images, more diagnostic information
is provided. However, there is little consistency in the number of images
obtained at different centers. Sampling ranges from one or two images at set
anatomic levels, such as the aortic arch, the carina, and just above the
diaphragm; to six to eight images evenly spaced throughout the lungs; to
images at 1-cm intervals throughout the entire lung
[41].
Few studies have actually looked at the appropriate sampling frequency. In
a report by Leung et al. [19]
comparing the accuracy of high-resolution CT and conventional CT in 75
consecutive patients with chronic diffuse infiltrative lung disease, two
observers interpreted three separate sets of CT scans in each patient in
random order. These sets included three high-resolution CT scans at the level
of the aortic arch, the tracheal carina, and 1 cm above the right
hemidiaphragm; three 10-mm collimation CT scans obtained at the same levels as
the high-resolution CT scans; and a complete conventional CT scan. The correct
diagnosis was made in 71% of the high-resolution CT scans and in 72% of the
three-level 10-mm and complete conventional CT scans. A definite confidence
level was reached with 49% of high-resolution CT scans, 31% of the three-level
10-mm scans, and 43% of complete conventional CT examinations, with the
correct diagnosis made in 92%, 96%, and 94% of these cases, respectively,
suggesting that a specific diagnosis can be made with a limited number of
high-resolution CT scans in many patients. Similarly, Kazerooni et al.
[41] scored the severity and
profusion of ground-glass opacity and reticular abnormality in 25 consecutive
patients with idiopathic fibrosis on both limited three-level high-resolution
CT and high-resolution CT obtained at 1-cm intervals throughout the entire
lungs. The scores from both sets correlated equally with the abnormalities
shown on open lung biopsy specimens. Other investigators concluded that the
gains in high-resolution CT over conventional CT in visualization of small
parenchymal structures that allow confident evaluation of diffuse interstitial
lung diseases is only possible when the entire lung is studied
[13]. Henschke
[42] reported a methodologic
framework for selecting the appropriate number of high-resolution CT images
using simple and stratified random sampling that could reduce the number of
high-resolution CT images given prior knowledge of the disease that can be
obtained from chest radiographs, pulmonary function tests, radionuclide
studies, and clinical parameters.
Patient Position
Most high-resolution CT images are obtained with the patient in the supine
position. When the lung abnormality is diffuse in distribution or severe in
profusion, inspiratory images alone are usually sufficient. Images with the
patient prone may be useful when the only abnormality is in the dependent
portion of the lungs; on supine images alone it may be difficult to determine
if the findings represent true lung disease or dependent atelectasis
[3,
43]. The latter occurs more
often in current and former smokers (34-43%) than in nonsmokers (12%), and
with increasing age [44].
Because dependent atelectasis occurs in the most dependent portion of the
lungs, when a patient is placed prone atelectasis shifts from the anatomically
posterior lung to the anatomically anterior aspect of the lung that is now the
most dependent lung (Fig.
6A,6B).
In contrast, with real lung disease the opacities persist; note the opacity
may be less than when the patient is supine, because some, but not all, of the
opacity may have been dependent atelectasis (Fig.
7A,7B).
If high-resolution CT scans are obtained according to a protocol and not
checked routinely before the patient leaves the radiology department, prone
images should be included in the routine scanning protocol. A less-recognized
site of focal atelectasis that may mimic true interstitial lung disease is the
lung immediately anterior to the spine in the azygoesophageal recess, which is
particularly common if large osteophytes are present. Similarly, atelectasis
may occur adjacent to a large hiatal hernia or bulky callus resulting from a
rib fracture.
View larger version (99K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 6A. —61-year-old man with dependent opacity mimicking lung
disease. High-resolution CT scan through lung bases with patient supine
reveals bilateral ill-defined ground-glass and faint reticular opacity
confined to dependent portion of lungs.
|
|
View larger version (82K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 6B. —61-year-old man with dependent opacity mimicking lung
disease. High-resolution CT scan at same anatomic level as A and with
patient prone reveals that opacity completely clears, indicating opacity shown
on A was atelectasis.
|
|
View larger version (79K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 7A. —29-year-old woman with dependent opacity representing usual
interstitial pneumonitis. High-resolution CT scan through lung bases with
patient supine reveals bilateral ill-defined ground-glass and reticular
opacity confined to dependent portion of lungs.
|
|
View larger version (99K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 7B. —29-year-old woman with dependent opacity representing usual
interstitial pneumonitis. High-resolution CT scan at same anatomic level as
A and with patient prone reveals that opacity persists, confirming lung
parenchyma is abnormal.
|
|
Level of Respiration
High-resolution CT is typically performed at full inspiration. Dynamic or
ultrafast high-resolution CT can be performed throughout the respiratory cycle
on an electron beam CT scanner
[45]. Some of the same
information can be obtained by comparing end-inspiration and end-expiration
high-resolution CT images
[46,47,48].
Expiratory images are usually obtained at maximum expiration. Although
inspiratory high-resolution CT may be obtained at 1-cm spacing, usually fewer
expiratory high-resolution CT images are obtained, perhaps at 2-cm spacing or
less. Normal lung should increase in attenuation at end-expiration, similar to
the increased lung opacity seen on end-expiratory chest radiographs. The lungs
should also become smaller, and the posterior membranous wall of the trachea
appears concave, in contrast to the flat or convex appearance at inspiration.
Failure of the lung parenchyma to increase in attenuation on expiration
indicates air trapping and suggests small airways disease. In some disease
processes, such as bronchiolitis obliterans, a mosaic attenuation pattern of
air trapping on expiratory high-resolution CT images may be the only evidence
of abnormality, because the lungs may appear entirely normal or near normal on
inspiratory images (Fig.
8A,8B).
Expiratory images may be particularly helpful when trying to determine if a
pattern of mosaic attenuation is primarily caused by airway disease, vascular
disease, or infiltrative lung disease (Fig.
9A,9B).
Air trapping may also be seen with asthma, hypersensitivity pneumonitis,
emphysema, and cystic lung diseases, including Langerhans' cell histiocytosis
and lymphangioleiomyomatosis
[48,49,50,51].
View larger version (147K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 8A. —55-year-old woman with hypersensitivity pneumonitis.
Inspiratory high-resolution CT scan shows a few scattered thickened
interlobular septa and very faint pattern of mosaic attenuation.
|
|
View larger version (145K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 8B. —55-year-old woman with hypersensitivity pneumonitis.
Expiratory high-resolution CT scan at same anatomic level as A reveals
multifocal bilateral air trapping represented by low-attenuation lung
parenchyma. High-attenuation areas represent normal lung that has developed
atelectasis with expiration. Note internal bowing of posterior wall of
bronchus intermedius as evidence that scan was taken at expiration.
|
|
View larger version (136K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 9A. —54-year-old woman with idiopathic bronchiolitis obliterans.
Inspiratory high-resolution CT scan shows diffuse cylindric bronchiectasis,
with bronchi larger than adjacent arteries; signet ring sign of bronchiectasis
(arrows); and subtle mosaic attenuation. All are findings of small
airways disease.
|
|
View larger version (142K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 9B. —54-year-old woman with idiopathic bronchiolitis obliterans.
Expiratory high-resolution CT scan at same anatomic level as A reveals
that expected decrease in lung size is absent, and lungs remain low in
attenuation, indicating severe diffuse air trapping, with only normal lung
parenchyma found as a few individual secondary pulmonary lobules that
increased in attenuation (arrowheads).
|
|
Reduced-Dose Technique
Low-dose high-resolution CT refers to the use of a reduced tube current, as
low as 40 mAs, to obtain high-resolution CT images of the lungs
[52]. Although the images are
noisier than standard-dose high-resolution CT scans and the image quality is
poorer, some researchers have reported that the anatomic detail is equivalent
between low-dose and standard-dose high-resolution CT
[52,
53]. Other researchers have
reported that a minimum of 160 mAs is necessary to reliably identify
ground-glass opacity and subpleural lines, and that although ground-glass
opacity and emphysema can be identified on the low-dose technique, they may be
very subtle [52,
54,
55]. In general, this
technique should be avoided in obese patients.
Photography
Although there is no single correct window width and level combination, the
appropriate window width is between 100 and 2000 H, and window level is
between -500 and -700 H. The exact combination used is generally a matter of
personal preference. It is important that consistent photography is used to
accustom the observers to the normal appearance of the lung, and to avoid
misinterpretations of disease improvement or progression that are created by
an artifact of photography. When interpreting hard-copy images, in general
fewer images are photographed per sheet of film than for standard chest CT, to
make small lines and nodules bigger and easier to see. This may mean six or
nine images per film. Retrospective targeted image reconstruction can be used
to reduce image pixel size and to further increase spatial resolution. In
general, this targeted image reconstruction is not usually done because of the
additional time required to reconstruct the images to each lung, and the
preference of many radiologists to see both lungs on the same image.
Pitfalls
Recognizing artifacts and potential interpretive and cognitive pitfalls in
the approach to high-resolution CT images is important to avoid confusion of
artifacts with real lung disease and other misinterpretation. The pitfalls are
summarized in Appendix 1.
Motion
Movement during high-resolution CT image acquisition creates false-positive
findings on high-resolution CT images, particularly pseudo—ground-glass
opacity, pseudobronchiectasis, and double fissures
[56,57,58].
Pseudobronchiectasis occurs because of the motion of a pulmonary blood vessel
during the acquisition of an image, creating two parallel vessels on the image
that simulate the walls of a bronchus. This effect is most commonly seen
adjacent to the left ventricle and the aortic arch. It is particularly
problematic because bronchiectasis is commonly a focal disease.
Pseudobronchiectasis may skip images, unlike a true dilated bronchus that can
usually be followed on consecutive images.
Vascular pulsation artifact most commonly occurs adjacent to the left
ventricle and aortic arch. If two parallel opaque lines that are identical in
morphology are seen adjacent to the left ventricle, look carefully at the left
ventricle border with the lung. If the border of the left ventricle is seen in
two places, the distance separating these borders is the distance it moved
during the acquisition of the CT image. The distance separating the parallel
opaque lines should be the similar to this distance if it was caused by this
motion. Care should be taken not to consider that subtle abnormality in these
areas is true lung disease when the rest of the lungs are normal.
Respiratory motion creates false opacity throughout the image and may be
more difficult to recognize as artificial. A clue to recognizing subtle
respiratory motion artifacts that occur throughout the image is that the
attenuation of the lung varies diffusely every few images as the lungs move
during several respiratory cycles while scanning.
A star-shaped artifact representing movement of a blood vessel
perpendicular to the axial plane of imaging is another clue to identifying
motion-related artifacts.
Improper Viewing Width and Level
A window width that is too narrow or a level that is too low falsely
thickens bronchial walls and creates false ground-glass opacity by making
normal parenchymal structures appear too opaque
[56]. A window width that is
too wide may mask diseases that are characterized by reduced attenuation,
including emphysema and cystic lung disease. Serial examinations should be
viewed at the same window width and level combinations. A high-resolution CT
examination photographed at one window—level combination should not be
compared with a high-resolution CT examination photographed at another
window—level combination without recognizing this difference, because
doing so may lead to the incorrect conclusion that the lung disease has either
improved or progressed.
Left Heart Failure
The population of patients with interstitial lung disease, particularly
pulmonary fibrosis, is similar in age to patients with ischemic heart disease.
In some patients with ischemic heart disease, it may be difficult to determine
whether the cardiac disease is the only cause of shortness of breath,
particularly in the setting of a low diffusing capacity and mild restrictive
pulmonary function tests. The same may also be seen in patients with fluid
overload, such as patients with renal failure. Pulmonary edema resulting from
left heart failure may mimic interstitial lung disease on high-resolution CT,
with smoothly thickened interlobular septa and ground-glass opacity
(Fig. 10). Perihilar and
dependent ground-glass attenuation, enlarged nondependent blood vessels
(cephalization), and smoothly thickened septa that are gravity-dependent are
findings that support left heart failure and edema over infiltrative lung
disease [59]. When in doubt,
the high-resolution CT examination can be repeated in a few days when the
patient's fluid status or heart failure has been corrected.
View larger version (112K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 10. —68-year-old woman with interstitial edema resulting from left
heart failure. High-resolution CT scan through upper lobes shows smooth septal
thickening in a gravity-dependent distribution, with no honeycombing or septal
nodularity. Mild centrilobular emphysema is shown as small areas of abnormally
low attenuation.
|
|
Vessels Versus Miliary Nodules
Distinguishing small vessels from miliary nodules may be difficult on
noncontiguous high-resolution CT images. Some nodules, particularly nodules of
small size and low density, are best seen on high-resolution CT images,
whereas obtaining a cluster of contiguous high-resolution CT images or
obtaining several thicker 5- to 10-mm contiguous images may be useful if
discrete dense nodules are suspected
[13].
Pulmonary Vascular Disease
Acute and chronic pulmonary thromboembolic disease may alter the appearance
of the lungs and be confused with interstitial lung disease and small airways
disease
[60,61,62,63,64,65].
The typical high-resolution CT appearance of pulmonary vascular disease is a
mosaic pattern of alternating geographic areas of lung attenuation, with
reduced attenuation and decrease in size of pulmonary vessels as a result of
reduced perfusion in the distribution of the occluded pulmonary vessels, which
is referred to as mosaic perfusion or mosaic attenuation
(Fig. 11). The remainder of
the lung is either normal in attenuation or higher in attenuation than normal
because of a relative increase in blood flow away from the areas of pulmonary
artery occlusion. Mosaic perfusion also occurs with small airways disease, in
which the abnormally low-attenuation lung is created both by trapped air and
shunting of blood flow to the more normal lung to optimize matching of
ventilation and perfusion
[66]. The greatest difficulty
occurs in separating mosaic attenuation caused by patchy infiltrative lung
disease from mosaic attenuation caused by pulmonary vascular disease
[64,
67]. Expiratory images are
useful in this setting [63].
With small airway disease, the low-attenuation areas of the lung remain low
attenuation with expiration, whereas in pulmonary thromboembolic disease the
lungs become increased in attenuation with expiration compared with
inspiration images. In infiltrative lung disease with patchy ground-glass
opacity, the pulmonary blood vessels are usually the same caliber throughout
the lungs.
Normal High-Resolution CT and Suspected Interstitial Lung
Disease
High-resolution CT may show normal findings in a small percentage of
patients with biopsy-proven interstitial lung disease
[14,
68,
69]. A normal examination
occurs much less frequently with high-resolution CT than with chest
radiography. Occasionally, a patient with a good-quality high-resolution CT
examination with normal findings has a convincing clinical presentation for
interstitial lung disease, including progressive shortness of breath,
nonproductive cough, restrictive abnormality on pulmonary function tests, and
negative findings for infection or malignancy on bronchoscopy. With a
convincing clinical picture and normal high-resolution CT, open or
video-assisted thoracoscopic lung biopsy may still be indicated. Some lung
disease is occult on high-resolution CT, and normal findings on
high-resolution CT should not preclude further workup if the remainder of the
clinical picture strongly suggests lung disease.
Clinical Indications for High-Resolution CT
The clinical indications for performing high-resolution CT are to detect
and evaluate bronchiectasis, to evaluate suspected lung disease when chest
radiography findings are normal, to clarify the pattern of abnormality from
chest radiography in order to narrow the differential diagnosis, to evaluate
disease activity, to predict response to therapy and the likelihood of
survival, to guide selection of the type and location of biopsy, and to
evaluate the effectiveness of medical therapy.
Bronchiectasis
High-resolution CT, having replaced more invasive bronchography, is the
technique of choice for identifying and defining the extent of bronchiectasis.
Patients with suspected bronchiectasis usually have chronic respiratory
symptoms, including cough, recurrent pneumonia, and abundant sputum
production. Chest radiographs are notoriously insensitive for detecting
bronchiectasis, particularly if it is mild; radiographs show normal findings
in as many as 50% of patients with bronchiectasis. Naidich et al.
[70] first reported the use of
CT to detect bronchiectasis in 1982. Although early investigations showed poor
test performance using standard CT compared with bronchography and pathology
[71,
72], subsequent reports of
high-resolution CT, including the initial work by Grenier et al.
[73], have shown consistently
high sensitivity of 84-95% and specificity of 93-100% for the detection of
bronchiectasis
[73,74,75,76,77]
(Fig.
4A,4B).
Normal or Equivocal Chest Radiographic Findings, Suspected Lung
Disease, and Pattern Clarification
Chest radiographs lack both sensitivity and specificity in the evaluation
of diffuse lung disease. To begin with, chest radiographic findings may be
normal in patients with suspected interstitial lung disease (Fig.
12A,12B).
Three large series of infiltrative lung disease totaling more than 1000
patients showed normal findings on chest radiographs in an average of 13% of
patients with open lung biopsy—confirmed disease
[14,
78,
79]. Approximately 10% of
immunocompromised patients with acute diffuse lung disease in one series
[80] and 10% of patients with
Pneumocystis carinii pneumonia in another series were reported to
have normal chest radiographic findings
[81]; and in another series of
112 neutropenic patients with fever of unknown origin and normal chest
radiographic findings who had a total of 188 high-resolution CT scans, 60% of
the high-resolution CT scans showed abnormal findings and indicated the source
of infection [82]. The
high-resolution CT findings were abnormal, on average, 5 days before abnormal
chest radiographic findings developed, allowing more rapid diagnosis and
treatment in this high-risk population, many of whom were bone marrow
transplant recipients. In patients with asbestos exposure, high-resolution CT
often shows abnormal findings in the setting of normal chest radiographic
findings, as reported in 57 (45%) of 169 patients with a score of less than
1/0 using the International Labour Office interpretation scheme
[83], and these individuals
had significantly lower diffusing capacity (p = 0.024) and vital
capacity (p = 0.005) than those with normal high-resolution CT
findings [84]. Similarly, in
patients with scleroderma or progressive systemic sclerosis, and in patients
with rheumatoid arthritis, chest radiographic findings have been reported to
be abnormal in 9-59% of published series, whereas corresponding
high-resolution CT examinations showed abnormal findings in 71-100%, depending
on the population being studied
[85,86,87,88].
In as many as 50% of patients with pulmonary lymphangitic carcinomatosis, the
chest radiographic findings may be normal; therefore, high-resolution CT
should be considered in cancer patients with unexplained pulmonary symptoms
and normal chest radiographic findings
[89].
View larger version (132K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 12A. —56-year-old man with hypersensitivity pneumonitis resulting
from bird-fancier's lung. Posteroanterior chest radiograph, originally
interpreted as showing normal findings, shows subtle hazy opacity in mid and
lower lungs.
|
|
View larger version (120K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 12B. —56-year-old man with hypersensitivity pneumonitis resulting
from bird-fancier's lung. High-resolution CT scan obtained 1 hr after chest
radiograph reveals diffuse ground-glass opacity with faint centrilobular
nodules in less confluent areas, in addition to air trapping in scattered
secondary pulmonary lobules (arrow).
|
|
High-resolution CT is also more sensitive than conventional CT for the
detection of ground-glass opacity. A series by Remy-Jardin et al.
[13] of 150 consecutive
patients compared conventional CT and high-resolution CT. The only technique
to reveal ground-glass opacity was high-resolution CT, and both fine bronchial
and parenchymal lesions were better seen on high-resolution CT than on
conventional CT.
When abnormalities are identified on chest radiographs, they are less
specific than on high-resolution CT scans. For example, Padley et al.
[14] reported a specificity of
82% for identifying normal findings using chest radiography compared with 100%
using high-resolution CT in a series of 100 patients, 86 with biopsyproven
interstitial lung disease and 14 normal control subjects. Not only is a
correct first-choice diagnosis of interstitial lung disease made more often
with high-resolution CT than with chest radiography, but the first-choice
diagnosis is also made with greater confidence. For example, in a series of
118 consecutive patients with chronic infiltrative lung disease reported by
Mathieson et al. [18], three
observers were asked to rank their top three diagnoses and the degree of
confidence in their first-choice diagnosis. Chest radiography rendered a
confident first-choice diagnosis in 23% of cases compared with more than twice
as many—49%—with high-resolution CT. High-resolution CT
interpretations were correct in 93% of cases compared with 77% for the
interpretations of chest radiographs. Similarly, Grenier et al.
[90] reported a comparison in
140 consecutive patients with chronic infiltrative lung disease, with three
observers also listing their top three diagnoses and confidence. Correct
confident diagnoses were made by the three observers on 19-34% of chest
radiographs each, compared with 47-57% for the high-resolution CT
interpretations (p < 0.001). Nishimura et al.
[91] also reported a more
frequent correct first-choice diagnosis for high-resolution
CT—46%—than for chest radiography—38%—in a series of
134 patients with diffuse infiltrative lung disease (p <
0.01).
In addition to the gains in sensitivity, specificity, and diagnostic
confidence over chest radiography, high-resolution CT also reduces
interobserver variability in interpretation when compared with chest
radiography [92,
93]. For example, in a series
of 61 consecutive individuals exposed to asbestos who had chest radiography
major category scores of 0 or 1 using the International Labor Organization
interpretation scheme [83],
mixed with five normal control subjects, Begin et al.
[93] used kappa statistics to
show significantly better interobserver agreement for CT interpretations than
for chest radiography interpretations (p < 0.001) and reduced the
frequency of indeterminate interpretations. Collins et al.
[92] tested inter- and
intraobserver variability of high-resolution CT and chest radiography for
determining pattern type and extent of disease in fibrosing alveolitis among
two experienced and two inexperienced observers on a total of 126
high-resolution CT examinations and 108 chest radiographs scored on two
occasions at least 8 weeks apart. Three of four observers agreed on pattern
type in 81% of high-resolution CT examinations (
= 0.48) compared with
54% for chest radiographs ( = 0.16). The greater the observer
confidence in identifying pattern type on high-resolution CT, the lower the
interobserver variability and the more extensive the disease. Intraobserver
variability for high-resolution CT pattern was less for the experienced
( = 0.78 and 0.70) than inexperienced observers ( = 0.50 and
0.37). Interobserver variability for extent of disease was significantly less
on CT than on chest radiography (p < 0.001).
Disease Activity and Predicting Response to Therapy and Survival
Honeycombing has been well documented by high-resolution
CT—pathologic correlation to represent irreversible end-stage pulmonary
fibrosis (Fig. 13), as
described in the initial report by Müller et al.
[94] of
radiologic—pathologic correlation in nine patients with established
interstitial fibrosis in 1986. Later, in 1987,
Müller et al.
[95] reported the first
evidence that the high-resolution CT appearance was an indicator of disease
activity in a series of 12 patients with idiopathic pulmonary fibrosis who
underwent open lung biopsy. Disease activity, measured by a pathologic grading
system, identified seven patients with mild disease activity and five with
moderate to marked activity. Disease activity was independently scored 0-3 on
CT scans on the basis of the presence and density of air-space consolidation.
The pathologic score was significantly greater in patients with higher CT
scores (p = 0.001), and CT correctly identified five patients with
marked disease activity and five of the seven with mild activity.
View larger version (144K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 13. —58-year-old man with usual interstitial pneumonitis.
High-resolution CT scan through lung bases shows extensive honeycombing,
indicating severe irreversible fibrosis. When this pattern is subpleural and
lower-lobepredominant, it is characteristic of usual interstitial
pneumonitis.
|
|
Ground-glass opacity is a less specific finding and may represent active
inflammation or fibrosis of alveolar or intralobular septa that is below the
resolution of high-resolution CT. In 1993, Remy-Jardin et al.
[96] reported a series of 26
patients with extensive ground-glass attenuation as the predominant or
exclusive abnormality and no honeycombing on high-resolution CT, with
histologic correlation at 37 biopsy sites. The ground-glass opacity
corresponded to inflammation in 24 sites (65%) and to fibrosis in 13 sites
(35%); in 85% of the cases with fibrosis, associated traction bronchiectasis
or bronchiolectasis was seen. Therefore, ground-glass opacity in the absence
of bronchiectasis usually represents active inflammation, whereas ground-glass
opacity with traction bronchiectasis usually represents fibrosis.
In general, patients with more ground glass respond better to therapy, and
patients with greater fibrosis have both poorer response and poorer survival
(Fig.
14A,14B).
Wells et al. [97] showed
significantly greater survival in a retrospective 4-year review of patients
with fibrosing alveolitis with predominantly ground-glass opacity (100%;
n = 8) compared with patients with a predominantly reticular (15%;
n = 50) or mixed pattern (45%; n = 18) (p <
0.001), and this prediction of survival was better than for existing
functional measurements and dyspnea duration. Later, using a semiquantitative
visual scoring system for the alveolar (ground-glass) and fibrosis (lines and
honeycombing) components of interstitial lung disease, Gay et al.
[98] showed that
high-resolution CT alveolar scores were higher and fibrosis scores lower in
patients responding to corticosteroid therapy than in nonrespondents in a
prospective study. Using receiver operating curve analysis to identify
pretreatment features of longer term survival, these researchers found that
only the high-resolution CT fibrosis score (p = 0.009) and the open
lung biopsy fibrosis score (p = 0.03) were able to predict mortality,
and high-resolution CT did so non-invasively. Pulmonary function and chest
radiographic features were not helpful.
View larger version (127K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 14B. —53-year-old woman with desquamative interstitial pneumonitis.
High-resolution CT scan after 6 months of medical therapy with azathioprine
reveals that abnormality has almost completely resolved.
|
|
Guide Type of Biopsy and Location of Biopsy
Transbronchial and open or thoracoscopic lung biopsy are the two main
techniques for the histologic diagnosis of chronic infiltrative lung disease.
Transbronchial biopsy is most effective in patients with sarcoidosis and
lymphangitic spread of cancer, both disease processes that characteristically
involve the central or axial interstitial compartment
[99]. Transbronchial biopsy is
insufficient to diagnose interstitial pneumonitis and pulmonary fibrosis, but
it may be useful to exclude other diagnoses, such as infection, in these
patients. The accuracy of transbronchial biopsy has improved for the diagnosis
of other less common diseases, such as Langerhans' cell histiocytosis,
alveolar proteinosis, and eosinophilic lung disease. Open lung biopsy is more
than 90% accurate and has largely been replaced by video-assisted
thoracoscopic surgery, a technique associated with less morbidity, faster
postoperative recovery, and lower cost
[100]. Samples are usually
obtained from all lobes on the side undergoing biopsy, three from the right or
two from the left. Because the tactile sensation of open lung biopsy is lost
with the use of video-assisted thoracoscopic surgery, and the field of view is
smaller, high-resolution CT has become more important for directing the
surgeon to the areas of ground-glass opacity and nodularity and away from the
honeycombing that may yield fibrosis of indeterminate cause at pathology
(Figs. 13 and
14A,14B).
Because the subtypes of interstitial pneumonitis (desquamative, usual, and
nonspecific) are associated with a different prognosis for the patient, it is
important to sample the areas of lung parenchyma that maximize the
differentiation among the subtypes
[101,
102]. CT has been shown to be
more accurate than chest radiography for determining whether transbronchial or
open lung biopsy is needed for accurate diagnosis
[18].
Evaluate the Effectiveness of Medical Therapy
High-resolution CT is used to monitor the response of lung disease to
medical therapy, which is particularly important given the morbidity
associated with many of the cytotoxic drugs used to treat interstitial lung
disease [10,
103,104,105,106,107].
With treatment, ground-glass opacity may regress, as seen in 18 (32%) of 56
patients with fibrosing alveolitis reported by Wells et al.
[108]. Ground-glass opacity
progressed in five patients, and reticular abnormality progressed in nine
patients. In no patient did the reticular abnormality regress. Although
ground-glass opacity significantly correlates with improvement in pulmonary
function after steroid treatment for pulmonary fibrosis or usual interstitial
pneumonitis, areas of ground-glass opacity on high-resolution CT have been
shown to precede and predict the development of honeycombing in the same
location over time [103,
105,
107]. In sarcoidosis,
ground-glass opacity, alveolar or pseudoalveolar consolidation, nodular and
irregular linear opacities, and interlobular septal thickening have been shown
to represent potentially reversible disease, whereas cystic air spaces,
architectural distortion, and septal lines may remain unchanged and may be
irreversible [10,
104].
Specific Diagnoses Possible with High-Resolution CT
A full description of the pattern-based approach to high-resolution CT is
beyond the confines of this article. The predominant patterns of
abnormality—reticular or interstitial abnormality, nodular abnormality,
and altered attenuation—are listed in Appendix 2. Some patterns or
combinations of patterns together with the anatomic distribution of the
abnormality from lung apex to base, and peripheral subpleural versus central
bronchovascular, can lead the interpreter to a specific diagnosis.
High-resolution CT is particularly accurate in the diagnosis of cystic lung
diseases and their distinction from emphysema, in cases of usual interstitial
pneumonitis when subpleural honeycombing predominates, in sarcoidosis, and in
lymphangitic carcinomatosis
[16,
109]. The diseases and
patterns that can lead to a single diagnosis include bronchiectasis,
emphysema, Langerhans' cell histiocytosis, lymphangioleiomyomatosis, usual
interstitial pneumonitis, hypersensitivity pneumonitis, lymphangitic
carcinomatosis, pneumoconiosis, and sarcoidosis. The diagnosis in these cases
may be sufficiently specific from high-resolution CT to obviate tissue
confirmation.
Bronchiectasis
Bronchiectasis has a very specific appearance on high-resolution CT; it
appears as dilated bronchi, larger in cross-section than the adjacent
pulmonary artery, with or without bronchial wall thickening. The signs of
bronchiectasis on CT include the signet ring sign (the artery is the diamond
and the bronchus is the ring in cross-section), nontapering of bronchi,
dilated bronchi within 1 cm of the visceral pleura, linear and cystic
air-filled structures, air—fluid levels in distended bronchi, and
bronchial wall thickening caused by peribronchial fibrosis
[70,
75] (Figs.
4A,4B
and
9A,9B).
Emphysema
Emphysema appears on high-resolution CT as areas of abnormally low
attenuation without definable walls, which distinguishes emphysema from the
cystic lung diseases [16,
35]. The most common form,
centrilobular emphysema, typically occurs as a result of cigarette smoking and
is upper-lobe-predominant in distribution. Early centrilobular emphysema is
characterized by round black holes that may appear in the central portion of
the secondary pulmonary nodule around the centrilobular artery
[110]. As emphysema
progresses, the low-attenuation areas become confluent and inseparable (Fig.
15A,15B).
The pulmonary vessels in areas of severe emphysema are small, with shunting of
blood flow to lung parenchyma that can better exchange air to maintain matched
ventilation and perfusion. Panlobular emphysema is lower-lobe-predominant,
typically occurs as a result of 
1 antiprotease deficiency,
and progresses more rapidly if associated with cigarette smoking
[111].
View larger version (111K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 15A. —72-year-old woman with severe centrilobular emphysema.
High-resolution CT scan at level of aortic arch shows severe emphysema, with
normal lung parenchyma almost completely replaced by abnormally low
attenuation.
|
|
View larger version (123K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 15B. —72-year-old woman with severe centrilobular emphysema.
High-resolution CT scan at lung bases shows mild emphysema, appearing as small
round areas of low attenuation, often abutting centrilobular artery
(arrows).
|
|
Langerhans' Cell Histiocytosis
Langerhans' cell histiocytosis, otherwise known as histiocytosis X or
eosinophilic granuloma, is considered a smoking-related lung disease that may
improve or even resolve with smoking cessation. A history of cigarette smoking
is reported in 90-100% of cases. When a combination of cysts and irregular
nodules is identified in a cigarette smoker, more severely involving the upper
lungs than the lung bases, this diagnosis can be made with confidence
[112,113,114]
(Fig. 16). The abnormality is
more nodular early in the disease course and evolves to be more cystic with
time [115] (Figs.
16 and
17). In a series of 48
patients with Langerhans' cell histiocytosis reported by Travis et al.
[116], all patients had a
history of cigarette smoking, two had pituitary involvement, and four had bone
lesions.
View larger version (131K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 16. —65-year-old man with Langerhans' cell histiocytosis and a
3-year history of progressive dyspnea. High-resolution CT scan at level of
aortic arch shows mixed pattern of irregular nodules and cysts that was less
severe at lung bases.
|
|
View larger version (117K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 17. —54-year-old woman with 20-year history of Langerhans' cell
histiocytosis. High-resolution CT scan at level of aortic arch shows
predominant pattern of cysts that was less severe at lung bases. Irregular
nodules are relatively minor component.
|
|
Lymphangioleiomyomatosis
Lymphangioleiomyomatosis is a rare cystic lung disease with classic, if not
specific, radiologic findings. This disease is characterized by uniformity.
Uniformly sized cysts, uniformly distributed from lung apex to base and center
to periphery, uniformly occurring in women of childbearing age. Early in the
disease the cysts are surrounded by normal lung parenchyma
(Fig. 18). Additional findings
include chylous pleural effusions, low-attenuation lymph nodes containing
dilated spaces filled with chylous material, and spontaneous pneumothorax
(Fig. 18). Over time, no
normal parenchyma is visualized
[16,
117]
(Fig. 19). The severity of the
cysts measured visually or using attenuation-based quantitative CT directly
corresponds to the severity of the associated obstructive pulmonary function
abnormalities, the impairment in gas exchange, and the reduction in exercise
performance in these patients
[118,
119].
View larger version (125K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 18. —39-year-old woman with lymphangioleiomyomatosis.
High-resolution CT scan at level of carina displayed at lung window on left
and soft-tissue window on right. In addition to large bilateral pleural
effusions, note small round low-attenuation areas with faint walls,
representing cysts, that were uniformly distributed throughout lung
parenchyma.
|
|
View larger version (113K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 19. —40-year-old woman lymphangioleiomyomatosis. High-resolution
CT scan at level of aortopulmonary window shows severe lung destruction, with
almost complete replacement of normal lung parenchyma by cysts that were
uniformly distributed throughout lungs.
|
|
Usual Interstitial Pneumonitis
Usual interstitial pneumonitis may appear on high-resolution CT as a
spectrum of abnormalities ranging from ground-glass opacity early in the
disease to the honeycombing of end-stage disease. The appearance of
subpleural, lower-lobe-predominant honeycombing on high-resolution CT is
characteristic of and highly specific for usual interstitial pneumonitis
[97,
109]
(Fig. 13). Usual interstitial
pneumonitis may be idiopathic or related to collagen vascular disease or drug
toxicity [120,
121]. When ground-glass
opacity is the predominant pattern on high-resolution CT, the differential
diagnosis is long and includes most of the idiopathic interstitial pneumonias,
interstitial infections such as cytomegalovirus and P. carinii
pneumonia, hypersensitivity pneumonitis, edema, alveolar proteinosis, and
diffuse bronchoalveolar cell carcinoma
[122]. Most patients with
usual interstitial pneumonitis present fairly late in their disease course
with honeycombing, compared with patients with either desquamative or
nonspecific interstitial pneumonitis who present when their disease is
characterized by more active inflammation and ground-glass opacity and is more
amenable to therapy
[123,124,125].
In 1997, Müller and Colby
[101] reviewed the idiopathic
interstitial pneumonias, a topic that alone could occupy this entire
article.
Hypersensitivity Pneumonitis
Hypersensitivity pneumonitis presents in the subacute stage with diffuse
centrilobular nodules, with or without patchy ground-glass opacity and air
trapping [126] (Figs.
12A,12B
and 20). Clinical history and
serologic tests combined with this high-resolution CT appearance are
sufficient to make the diagnosis. The CT findings represent a mononuclear cell
bronchiolitis and cellular interstitial infiltrate, with poorly defined,
scattered nonnecrotizing granulomas
[126]. Centrilobular nodules
are reported in 40-100% of patients with subacute hypersensitivity
pneumonitis, and patchy ground-glass opacity in 52-100%
[126,127,128].
The abnormality may be more severe in the mid and lower lungs than in the lung
apices. In a population-based study by Lynch et al.
[69] of 31 symptomatic
recreation center employees referred because of possible hypersensitivity
pneumonitis, 11 were diagnosed with hypersensitivity pneumonitis. Chest
radiographic findings were abnormal in only one (9%) of 11 patients.
High-resolution CT findings were abnormal in five (45%) of 11 patients, and in
each of these five patients appeared as poorly defined centrilobular nodules.
The appearance of chronic hypersensitivity pneumonitis is less specific,
appearing as fibrosis superimposed on patchy ground-glass opacity and
centrilobular nodules. Hypersensitivity pneumonitis often spares the lung
bases, a distinction that can be used to distinguish chronic hypersensitivity
pneumonitis from usual interstitial pneumonitis
[129,
130]
(Fig. 21). After the antigen
that has provoked the lung injury is withdrawn from the environment of the
patient with hypersensitivity pneumonitis, the ground-glass opacity and
centrilobular nodules may improve or resolve; whereas in patients with
persistent antigen exposure, the high-resolution CT findings persist or
progress [127].
View larger version (120K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 21. —69-year-old woman with 12-year history of chronic
hypersensitivity pneumonitis. High-resolution CT scan through mid lungs shows
traction bronchiectasis, reticular abnormality superimposed on patchy
ground-glass opacity, and a few centrilobular nodules. Unlike usual
interstitial pneumonitis, distribution of abnormality is not predominantly
subpleural.
|
|
Lymphangitic Carcinomatosis
Lymphangitic carcinomatosis can be recognized by the characteristic
appearance of irregular, nodular, or "beaded" interlobular septa
forming pol
Top
Introduction
Background
