Professional Documents
Culture Documents
Thoracic Radiology: # (149) # (149) # (149) Nestor L. Muller, MD, PHD
Thoracic Radiology: # (149) # (149) # (149) Nestor L. Muller, MD, PHD
The accuracies of chest radiography large number of chronic diseases most likely to yield a representative
and computed tomography (CT) in may result in diffuse infiltra- sample. Transbmonchial biopsy is the
the prediction of specific diagnoses tion of the lungs. The clinical and less invasive of these procedures, but
in 118 consecutive patients with functional features of most of these in chronic DILD it is limited almost
chronic diffuse infiltrative lung dis- entities are similar (1). Radiological- exclusively to the diagnosis of sam-
ease (DILD) were compared. The ra- ly, the differential diagnosis is based coidosis and lymphatic spread of tu-
diographs and CT scans were inde- on the type and distribution of opaci- mor (4). Open lung biopsy is me-
pendently assessed by three observ- ties. Traditionally, the radiographic quired in the diagnosis of most other
eN without knowledge of clinical or appearance has been divided into in- chronic diseases; it is an invasive pro-
pathologic data. The observers list- terstitial and air-space patterns of dis- cedure, and its sensitivity and speci-
ed the three most likely diagnoses ease. Felson showed, however, that ficity are limited by the small sample
in order of probability and recorded the prediction of microscopic distri- of lung parenchyma being assessed,
the degree of confidence they felt in bution from the radiographic pattern which may not be representative of
their first-choice diagnosis on a is unreliable (2). It is now recognized the diffuse process involving the
three-point scale. Confidence level 1 that it is better to assess the type and lungs (5).
(definite) was reached with 23% of distribution of opacities, to deter- A number of recent studies have
radiographic and 49% of CT scan mine the predominant pattern, and described the computed tomographic
readings, and the correct diagnosis then to try to predict the clinical di- (CT) appearances of various DILDs
was made with 77% and 93% of agnosis without trying to label the (6-18). It has been suggested that CT
those readings, respectively (P < pattern as air space or interstitial (2). of the chest is superior to chest madi-
.001). The correct first-choice diag- McLoud et al (3) devised a scheme ography because the decreased su-
nosis regardless of the level of con- for semiquantitative description of perimposition of structures on CT
fidence was made with 57% of ra- the radiographic appearances of dif- images allows a better assessment of
diographic and 76% of CT scan read- fuse infiltrative lung disease (DILD) the type, distribution, and severity of
ings (P < .001). The CT scan that was based on the International parenchymal abnormalities than is
interpretations were most accurate Labour Office (ILO) classification of possible on the radiograph (8). How-
in silicosis (93%), usual interstitial pneumoconioses. This scheme allows ever, to our knowledge the value of
pneumonia (89%),
lymphangitic car- for standardization of the method of CT in the prediction of pathologic di-
cinomatosis (85%),
and sarcoidosis analysis. In a review of the chest ma- agnosis has not been previously eva!-
(77%). Observers correctly predicted diographs of 365 patients, an experi- uated. The aim of the present study
whether a transbronchial or open enced chest radiologist was able to was to assess the accuracy of CT com-
lung biopsy was indicated with 65% include the correct histologic diagno- pared with that of chest radiography
of radiographs and 87% of CT scans sis in the first two radiologic diag- in the determination of specific diag-
(P < .001). It is recommended that nostic choices in 50% of cases and noses in patients with chronic DILD.
CT be performed before lung biopsy among the first three choices in 78% We also compared the accuracy of
in all patients with chronic DILD. of cases. Although certain radiologic both methods in the prediction of
patterns may be suggestive of a par- whether transbronchial biopsy was
Index terms: Computed tomography (CT), corn- ticular disease process, a confident likely to yield a diagnostic-quality
parative studies Computed
#{149} tomography (CT), diagnosis is rarely possible (3). In specimen.
clinical effectiveness Lung,
#{149} CT, 60.1211 most patients, lung biopsy is me-
B Lung, diseases, 60.213, 60.22, 60.331, 60.77 quired for definitive diagnosis. The
B Lung, fibrosis, 60.792 Lung neoplasms
#{149} sec- PATIENTS AND METHODS
radiograph is also of limited value in
ondary, 60.331 Pneurnoconiosis,
#{149} 60.77 #{149}Radi-
determining whether transbronchial All patients with chronic DILD me-
ography, comparative studies #{149}
Sarcoidosis,
or open lung biopsy should be per- femred to our hospital for CT of the chest
60.22
formed and which area of lung is between September 1983 and October
Radiology 1989; 171:111-116
1987 who had both undergone chest nadi-
ography and received a definitive diagno-
sis were included in the study. One hun-
I From the Department of Radiology, University of British Columbia and Vancouver General
dred eighteen patients met these criteria,
Hospital, 855 W 12th Ave. Vancouver, BC, Canada V5Z lM9. From the 1988 RSNA annual meeting.
Received August 3, 1988; revision requested October 17; revision received October 31; accepted
November 3. Address reprint requests to N.L.M. Abbreviations: DILD diffuse infiltrative
C RSNA, 1989 lung disease, ILO International Labour Of-
See also the editorial by Naidich (pp 22-24) in this issue. lice, UIP OR usual interstitial pneumonia.
111
73 men and 45 women with a mean age of
56.8 years (range, 24-84 years). The CT
Table 1
scans were obtained on an 8800 (24 pa- Summary of CT Appearances of Chronic Diffuse Infiltrative Lung Diseases
tients) or 9800 (94 patients) scanner (GE (References 3,6-23).
Medical Systems, Milwaukee). The medi- Disease CT Appearances
an time interval between chest radiogra-
phy and CT scanning was 2 days. The UIP Reticular pattern predominantly subpleunal; usually lower-zone
scanning routine initially consisted of 1- predominance; may show areas of haziness or air-space opacifi-
cation, but reticular changes predominate; may show honey-
cm-collimation scans at 1-cm intervals. combing
The scans were obtained at end-inspina- Silicosis Randomly distributed, well-defined nodules; upper-zone predomi-
tory lung volumes with the standard al- nance; may show posterior predominance; may show confluence
gomithm and photographed at window Sarcoidosis Nodules predominantly along bronchovascular bundles; middle-
and upper-zone predominance; may show reticulation/hazi-
levels and widths appropriate for lung ness/consolidation; bilateral hilar and mediastinal adenopathy
parenchyma (level = -600 to -700 HU, often present
width = 1,000-2,000 HU) and mediasti- Lymphangitic carcino- Pemibronchovascular nodules, thickened bronchovascular bundles;
num (level 30-50 HU, width 350-500
matosis often shows polygonal lines; may show lymphadenopathy;
sometimes unilateral
HU). After July 1986, additional 1.5-mm- Extrinsic allergic alveo- Small, randomly distributed nodules with poorly defined margins,
collimation scans were routinely obtained litis often associated with patchy haziness or consolidation and relic-
at the level of the aomtic arch, tracheal ca- ulation
Bronchiolitis obliterans Patchy peripheral consolidation, few reticular markings; dense
nina, and 1 cm above the right hemidia- consolidation versus haziness or mild air-space opacification of
organizing pneumo-
phragm (66 patients). These high-resolu- nia desquamative interstitial pneumonia and UIP; no lower-zone
tion CT images were obtained with 120 predominance; tends to be asymmetric
kVp and 360-420 mAs. Retrospective tar- Asbestosis Same as for UIP, plus bilateral pleural plaques
Desquamative intersti- Patchy subpleural haziness with mild or absent reticular changes
geting of the 1.5-mm-collimation scans tial pneumonia
was performed with a field of view of 20 All other conditions Specific CT appearances not yet described; well described in radi-
or 25 cm and the high-spatial-resolution ology and pathology literature (references 1-3,19,20,23,28)
algorithm (bone algorithm). This field of
view is best because it permits evaluation
of an entire lung.
The chest radiographs and CT scans Table 2
were separately reviewed in random on- Percentage of Correct Diagnoses by Disease Entity
den by three independent observers
(J.R.Mathieson, J.R.Mayo, C.A.S.). The ob- First-Choice Top Three
112 #{149}
Radiology April 1989
exclusively in lymphangitic carcino-
Table 3
matosis and sarcoidosis. Well-de-
Percentage of First-Choice Diagnoses Made with a High Level of Confidence fined, randomly distributed upper
(Level 1) That Were Correct
lobe nodules correlated highly with
Chest Radiography CT silicosis. The presence of peripheral-
ly distributed lower-zone meticulam
No. of Confident Confident
Disease Cases Interpretations Correct Interpretations Correct areas of increased attenuation and
honeycomb cysts was noted exc!u-
UIP 34 30 87 73 95 sively in UI? and in the two patients
Silicosis 20 37 100 72 100
Sarcoidosis 19 9 60 28 88 with asbestosis. CT scanning was
Lymphangitic most accurate in the diagnosis of sili-
carcinomatosis 18 20 64 54 93 cosis, UI?, lymphangitic carcinoma-
Extrinsic allergic
alveolitis 7 10 0 10 50 tosis, and sarcoidosis (Table 4). Both
radiography and CT were of limited
Total 118 23 77 49 93
value in the diagnosis of extrinsic al-
lergic alveolitis (Table 4). Typical ex-
amples of the CT appearances of
these five diseases are shown in Fig-
Table 4 ures 1-5.
Percentage of Selected CT Findings in Five Most Common DILDs A tnansbronchial biopsy was con-
mectly suggested by the radiographic
Extrinsic
Lymphangitic Allergic results in 65% of cases and by the CT
CT Finding UIP Silicosis Sancoidosis Carcinomatosis Alveolitis* results in 87% of cases (P < .001). An
open lung biopsy was correctly sug-
Small opacity
distribution gested by the radiographic results in
Upper 1.0 78.3 59.6 9.4 5.3 89% of cases and by the CT results in
Lower 58.4 1.7 10.5 45.2 47.4
99% of cases (P < .001).
All zones 39.6 20.0 26.3 41.5 47.4
Nodular densities 7.8 100.0 89.5 84.9 50.0 The concordance among the obser-
Reticular densities 97.1 43.3 73.7 94.3 65.0 vers’ first-choice diagnoses was good
Pleural fluid or
thickening 8.8 5.1 7.0 60.4 20.0 with the CT scans (Kendall W .089,
Septal lines 15.0 1.7 7.0 66.0 15.0 P < .001) and lower but still signifi-
Unilateral predom- cant with the chest radiographs (W
inan#{231}e
of
small opacities 3.0 1.7 8.8 38.5 5.3 .029, P < .05).
Peripheral
predominance 94.0 3.3 3.5 15.4 36.8
Penibronchovascular 4.0 6.7 70.2 59.6 5.3 DISCUSSION
Posterior
predominance 2.0 38.3 3.5 1.9 21.1 Accurate interpretation of chest ma-
Polygonal lines 1.0 1.7 3.5 50.9 5.0
diographs with diffuse abnormalities
Note-The percentages for the most important diagnostic features are in bold type. has been called “one of the most dif-
* In retrospect, the reticular densities, septal lines, and small pleural effusion seen on the radiograph ficult problems in diagnostic madiolo-
and CT scan in one patient with extrinsic allergic alveolitis were probably due to concomitant mild
congestive heart failure. Open lung biopsy showed that the patient did have mild interstitial disease gy” (23). The radiologic custom has
consistent with the clinical diagnosis of extrinsic allergic alveolitis. However, clinical and radiologic been to divide the diseases causing
follow-up suggested that most of the initial CT findings were due to congestive heart failure. diffuse pulmonary parenchymal ab-
nommalities into two groups: those
predominantly involving the inter-
among observers was established with el of confidence, the radiographic di- stitium and those predominantly in-
the Kendall test of concordance. The pen- agnosis was correct in 77% of cases as volving the terminal aim spaces. Fe!-
centage of correct diagnoses with the two compared with 93% for CT diagnosis son (19) described the radiographic
modalities represents the sum of the con-
(P < .001). characteristics of each of these
mect interpretations by the three observers
Five conditions accounted for 83% groups but later related his frustra-
divided by the totals of 354 radiographic
of the cases (Table 2). These included tion at being unable to teach others
and 354 CT readings.
UI? (29%), silicosis (17%), sarcoidosis to recognize these radiographic pat-
(16%), lymphangitic carcinomatosis temns (2). There are several funda-
(15%), and extrinsic allergic alveolitis mental problems with this approach.
RESULTS
(6%). The characteristic CT findings First, most disease processes involve
The 1 18 cases in the study included for these conditions are summarized both the aim spaces and the intersti-
18 different pathologic entities. The in Table 4. The number of patients tium (3). Second, the degree to which
correct first-choice diagnosis was within each of the other categories each compartment is involved varies
made in 57% of radiographs and 76% was too small to permit valid conclu- not only from patient to patient but
of CT scans. The correct diagnosis sions to be drawn with regard to over time in a given patient. Third, it
was among the top three choices in characteristic findings. The presence is not always possible to differentiate
73% of madiogmaphs and 89% of CT of polygonal lines, pleural fluid or between the radiographic features of
scans (Table 2). These differences thickening, septal lines, and a unilat- interstitial and aim space abnonmali-
were statistically significant (P < eral predominance of small opacities ties. The radiographic terminology
.001). A high confidence level (level all correlated highly with lymphan- has been called inconsistent (2), mis-
1) of diagnosis was reached in 23% of gitic carcinomatosis. A predominant- leading (3), and equivocal (25). Many
radiographic and 49% of CT scan in- ly pemibmonchovasculam distribution attempts have been made to improve
terpretations (Table 3). With this 1ev- of nodular densities was seen almost the diagnostic value of the chest ma-
2. 3.
Figures 2, 3. (2) A 1-cm-collimation CT scan
in a patient with silicosis shows multiple, small, well-defined nodular areas of attenuation that
are randomly distributed in the upper lung zones. The 1-cm-collimation scan allows easy distinction of small nodules from blood vessels.
(3) A 1-cm-collimation CT scan in a patient with sarcoidosis shows multiple, ill-defined, upper-zone nodular areas of attenuation, situated
predominantly along the bronchovascular bundles (arrows). The beaded appearance of the bronchovascular bundles, which is clearly seen
on 1-cm-collimation scans, is difficult to appreciate with high-resolution CT. Also notable is bilateral hilar lymphadenopathy.
114 #{149}
Radiology April 1989
may be present in only the acute or hances its usefulness. Nevertheless, High-resolution CT is superior to
subacute stage. we have shown that the CT diagnosis 1 -cm-collimation CT in demonstrat-
Our study was limited by the mela- is significantly more accurate. ing small cystic areas of honeycomb-
tively small number of cases of some We currently use CT in the initial ing (10). It is essential in the detec-
diseases and by the relatively small examination of all patients with tion of the characteristic polygonal
number of diseases included. We did, DILD. The methods and the advan- lines seen with lymphatic spread of
however, include a!! consecutive tages of the 1.5-mm collimation scans tumor (18) and in the assessment of
cases of DILD studied over a 4-year and retrospective targeting have been disease activity in idiopathic pulmo-
period, reflecting a mixture of the described by Mayo et a! (29). Target- nary fibrosis (13). However, the
prevalence and incidence of DILD in ing increases spatial resolution and beaded appearance of the broncho-
our community. The frequency and improves fine image detail despite an vascular bundles seen in lymphatic
range of diseases in our study were increase in visible noise (29). The spread of tumor and in sarcoidosis is
similar to those in the study by three high-resolution CT images we much easier to assess with 1-cm-colli-
McLoud et a! (3), with the exception obtained provide sections through mation scans. Small nodules can be
of a lower prevalence of desquama- each lobe of the lungs. These levels easily missed between high-resolu-
tive interstitial pneumonia in our were selected after consultation with tion CT sections and, when present,
study. Our observers, being deprived the two thoracic surgeons at our in- are difficult to distinguish from
of clinical information, were similar- stitution. They include virtually all blood vessels. Because high-nesolu-
ly at an unrealistic disadvantage open lung biopsy sites in patients tion CT is superior in the assessment
compared with what would be the with chronic DILD. Our routine is of diseases with predominantly imreg-
case in a real clinical situation. The also dictated in part by time con- ular linear opacities, whereas con-
accuracy of the radiographic diag- straints and scanning practice at our ventional CT is superior in the as-
noses in our series was similar to that hospital. Ideally, additional high-mes- sessment of small nodular opacities,
in the series of McLoud et a!, sup- olution CT sections should be ob- we believe that both should be used
porting their conclusion that a sys- tamed on the basis of the assessment in the initial examination of patients
tematic approach to the analysis of of the chest radiograph or the scano- with DILD. We do not routinely ob-
the chest radiograph greatly en- gram. tam supine and prone scans except in
.4
5.
Figures 4, 5. (4) A 1 .5-mm-collimation CT scan of the right lung in a patient with lymphangitic carcinomatosis shows irregular nodular
thickening of the bronchovascular bundles (curved arrow), polygonal lines (short arrows), and septal lines (long arrow). (5) A 1.5-mm-colli-
mation CT scan of the right lung in a patient with extrinsic allergic alveolitis shows areas of normal parenchyma; small, ill-defined, nodular
areas of attenuation that are randomly distributed (arrows); and areas of air-space opacification.
116 #{149}
Radiology April 1989