The British Journal of Radiology, 74 (2001), 3241

2001 The British Institute of Radiology

Visibility of small peripheral lung cancers on chest

radiographs: inuence of densitometric parameters,
CT values and tumour type
1

Z-G YANG, MD, 1S SONE, MD, 1F LI, MD, 1S TAKASHIMA, MD, 1Y MARUYAMA, MD,
2
T HONDA, MD and 1M HASEGAWA, MD
Departments of 1Radiology and 2Laboratory Medicine, Shinshu University School of Medicine, Asahi,
Matsumoto, 390 8621, Japan

Abstract. The purpose of this study was to determine the effects of tumour density and tumour
type on the visibility of small peripheral lung cancers on chest radiographs. We retrospectively
evaluated the visibility of 63 small (20 mm) peripheral lung cancers on chest radiographs. 48
(76%) were detected in our low dose CT screening for lung cancer and 15 (24%) in routine clinical
examination. Analysis was based on tumour optical contrast, gradient at the tumour margin, CT
values and tumour type. There were 31 (49%) visible cancers and 32 (51%) invisible cancers on
chest radiographs. Visible tumours had an optical density of 0.10.3 OD and a gradient of
0.030.11 OD mm21. The mean size of visible tumour (14.3 mm) was larger than that of invisible
tumour (11.1 mm; p,0.001). The mean CT value (2140 HU) of visible tumour was higher than
that of invisible tumour (2490 HU; p,0.001). The detection rates of adenocarcinomas with
lepidic growth (0% for type A, 29% for type B and 68% for type C) were less than those with hilic
growth (100% for types DF). All squamous and small cell carcinomas with hilic growth were
visible on chest radiographs, but the numbers of each were small. In summary, tumour type
inuenced the contrast, gradient, CT values and margin of the tumour. Small adenocarcinomas
with a lepidic tumour growth were less well seen on chest radiographs compared with small lung
cancers with hilic tumour growth.

The low cure rate of lung cancer is mainly

owing to delayed diagnosis based on the poor
diagnostic value of the conventional chest radiograph (CXR) in detecting small lung cancers
[13]. Several recent studies from Japan and USA
have shown that low radiation dose spiral CT
(low dose CT) can greatly improve the detection
of lung cancer [46]. Diederich et al [7] reported
that pulmonary nodules were reliably detected at
low dose CT with 50 mA or 25 mA. In our
screening trial for lung cancer by low dose CT,
many small peripheral lung cancers were found
among the general population, and most of these
tumours were invisible on CXRs even when the
lms were examined retrospectively. Specically,
we noticed that the majority of adenocarcinomas
that were 20 mm were difcult to recognize on
CXRs, even when they were located in well
penetrated lung elds and were not concealed by
Received 7 July 2000 and accepted 31 August 2000.
Address correspondence to Prof. S Sone, MD, Azumi
General Hospital, Ikeda, Nagano 399-8695, Japan.
32

the mediastinum, heart, hilum, hemidiaphragm or

clavicle. Further analysis suggested that tumour
growth pattern as evidenced by histopathological
examination was a potentially important factor
that could determine tumour visibility, although
these conclusions have not appeared in recent
studies [13, 4, 8].
In the present study, we compared densitometric parameters, CT values and histopathological features of 63 small peripheral lung cancers
that were visible or invisible on CXRs. The aim of
the study was to determine the effects of tumour
density and histopathological tumour type on the
visibility of small peripheral lung cancer on CXRs
and to dene the relationships between optical
contrast, gradient at the tumour margin, CT
values and histopathological tumour type.

Materials and methods

Patients
The subjects in this study consisted of
two groups of patients. The rst group were
The British Journal of Radiology, January 2001

Visibility of small peripheral lung cancers on CXRs

consecutive patients identied in our screening

trial for lung cancer using low dose CT, which
was performed for a general population in a rural
area of Japan, and comprised 47 patients with 48
small (20 mm in diameter) peripheral lung
cancers. The second group consisted of 15
consecutive clinical cases of small peripheral
lung cancers measuring 20 mm treated surgically at our hospital. All tumours in the second
group were detected on initial CXRs taken during
routine clinical examination.
CT screening group
Between May 1996 and December 1998 we
conducted a population-based mass screening trial
for lung cancer using a low dose CT scanner
(model CT W950SR; Hitachi Medical, Tokyo,
Japan). Details of the trial have been described
previously [5, 9]. In brief, scanning parameters
were set at 10 mm collimation, 10 mm s21 table
speed, 120 kVp, 50 mA (in 1996) or 25 mA (in
1997 and 1998), and 2 s per rotation of the X-ray
tube. The reconstruction interval of the CT image
was 10 mm using a standard reconstruction
algorithm with 180 linear interpolation. We
interpreted CT images on two high resolution
cathode ray tube (CRT) monitors (230061700
lines). Each CT examination was rst read by one
of ve radiologists, three with more than 10 years
experience and two with more than 5 years
experience in this eld, to evaluate the presence
or absence of any abnormality and to clarify
tentatively the abnormality into one of ve
categories. One radiologist, a specialist in thoracic
radiology, re-evaluated the CT images of the
abnormality. When any abnormality in the lung
was suggestive of a probable or possible lung
cancer, we performed further diagnostic workups, including chest radiography and diagnostic
CT (including thin section CT) scan.
In this trial we identied 70 patients with 71
surgically conrmed lung cancers 20 mm in
diameter. Among these, 48 lung cancers in 47
patients were located in well penetrated lung elds
on CXRs. These elds were not concealed by the
mediastinum, heart, hilum, hemidiaphragm or
clavicle when the location of tumour shown on
CT images was used as reference. These cases
comprised the subjects of the present study. The
remaining 23 patients were excluded from further
analysis: ve patients did not undergo work-up
examinations at our hospital and thin section CT
images were not available for review; and 18
patients had tumours concealed by the mediastinum, heart, hilum or hemidiaphragm (n515) or
by the clavicle (n53), and were not visible on
CXRs, regardless of tumour density and histopathological tumour type.
The British Journal of Radiology, January 2001

Clinical group
To compare the image features of small lung
cancers detected in the CT screening trial with
those cancers detected in clinical work, we studied
15 small peripheral lung cancers located in well
penetrated lung elds on CXRs. All of these cases
were detected on initial CXRs taken during
routine clinical examination, and were diagnosed
and treated at our hospital within the period of
the CT screening trial (May 1996 to December
1998).
Final study group
The present series included 62 patients with 63
small peripheral lung cancers of 20 mm in
diameter. It consisted of 29 males and 33 females,
ranging in age from 3376 years (mean 65.6
years). The size of each of the lung cancers was
10 mm in diameter (n522), 1115 mm (n527)
and 1620 mm (n514). 20 tumours were located
in the upper lobe, 9 in the middle lobe and 12 in
the lower lobe of the right lung, while 11 tumours
were located in the upper lobe and 11 in the lower
lobe of the left lung. Histological diagnosis
was well differentiated adenocarcinoma (n550),
moderately differentiated adenocarcinoma (n57),
poorly differentiated adenocarcinoma (n52),
squamous cell carcinoma (SCC) (n53) and
small cell lung carcinoma (SCLC) (n51).

Chest radiography and thin section CT

techniques
Chest radiography and CT examinations
(including thin section CT) were performed in
all patients. Standard posteroanterior (14617
inch) and lateral (14614 inch) CXRs were
obtained with a radiographic system (KXO 80G
Toshiba Co., Tokyo, Japan). The technical
parameters for chest radiography were:
135 kVp, 14:1 grid ratio and 180 cm focus-tolm distance, with a compensatory lter to
provide an adequate lm blackening in mediastinal and diaphragmatic areas. A rare earth lm
screen combination (screen, XGS; lm, ES-G;
Konica, Tokyo, Japan) of a standard system
speed (speed 250) and contrast was employed.
Non-enhanced CT (including thin section CT) for
all patients was performed with a state-of-the-art
CT scanner (HiSpeed Advantage, GE Medical
Systems, Milwaukee, WI). Non-enhanced CT
scans were taken from the lung apex to the
base at end inspiration, with technical parameters
of 120 kVp, 200 mA, 1 s scanning time, 10 mm
collimation, 10 mm s21 table speed and 32 cm
eld of view. Additional thin section CT scans
were taken through the nodule, with technical
scan parameters of 120 kVp, 200 mA, 1 mm
collimation, 1 mm s21 table speed, 1 s per rotation,
33

Z-G Yang, S Sone, F Li et al

pitch 1, reconstruction interval of 0.5 mm by

employing a bone algorithm, and eld of view of
20 cm.

Image interpretation
The CXRs of 62 patients with 63 small lung
cancers, together with the CXRs of 58 control
patients who had demonstrated no lung lesion on
CT images were evaluated independently by two
chest radiologists (Z-GY, FL) who were not told
of the presence or absence of any lesion on CXRs.
Interpretation was conducted while the radiograph was placed on a standard view box in a dim
lit room, without controlling the viewing distance.
Observers were asked to detect and to locate
nodular opacities in the lung parenchyma, using
four condence levels: (1) no lesion present;
(2) nodule present with 50% certainty; (3) nodule probably present; and (4) nodule certainly
present. Discrepancies in interpretation between
observers were resolved by a third radiologist
(SS). When a radiologist rated a lung eld at a
condence level of 2, 3 or 4 to indicate nodule
presence, and this was evidenced by CT nding of
tumour, the interpretation was classied as truepositive (visible tumour). With a condence level
rating of 1 (a lung eld without a cancerous
nodule), but a tumour visible on CT images, the
interpretation was designated as false-negative
(invisible tumour). Regarding the border of visible
tumour on CXRs, we considered it as distinct
when the tumour contour could be convincingly
identied, and as indistinct when it was difcult to
recognize a continuous boundary between the
tumour and the surrounding lung.

Measured parameters
31 visible tumours on the posteroanterior
CXRs were scanned using a densitometer
(Sakura PDM5, Konica, Tokyo, Japan) set at
an aperture size of 1.25 mm60.125 mm and a
scanning speed of 0.25 mm s21. The contrast of
the nodule (the difference in optical density of the
tumour and that of the surrounding lung
parenchyma) and the density gradient across the
tumour border (the rate of change of density as a
function of distance through which the density
changes at the nodule margin) were computed
according to the method described by Revesz et al
[10, 11].
For 63 tumours including visible and invisible
tumours, tumour size was determined by averaging the long and short axes measured at the
largest tumour diameter level, using a window
width of 1000 HU and a window level of
2700 HU on thin section CT images. The mean
CT value of tumour shown on the thin section CT
34

images was measured inside the region of interest

(ROI) determined on the CRT by manual tracing
using a light pen just along the interior edge of the
tumour. The CT value of the surrounding lung
parenchyma, excluding juxtatumoral pulmonary
emphysema seen in one patient, was also measured using ROIs of the same size as described
above. The difference in CT values (DCT)
between the tumour and its surrounding lung
parenchyma was then calculated.

Histopathological examination
All surgical specimens were xed at an inated
state by transbronchial infusion of formalin.
Surgical specimens were sectioned transversely
at 1.01.5 cm distances in almost the same
transverse plane as the CT scan, and stained
with hematoxylin-eosin (H&E) or elastic van
Gieson stain. A representative slice with the
largest tumour diameter was selected for macroscopic and microscopic examination of the
tumour. For each case of 63 cancers, two
pathologists independently evaluated histopathological tumour type. Discrepancies in interpretation between the pathologists were resolved by
consensus. Adenocarcinomas (n559) were classied into six types according to Noguchi's
classication [12]: Type A, tumour growth of
the alveolar lining; type B, alveolar lining tumour
growth with scattered brotic foci due to alveolar
collapse; type C, alveolar lining tumour growth
with foci of active broblastic proliferation; and
type DF, solid tumour growth, which showed
frequent expansive tumour growth; type D
indicates poorly differentiated adenocarcinoma,
E, tubular adenocarcinoma, and F, papillary
adenocarcinoma with compressive and destructive
growth. Types A, B and C were lepidic tumour
growth; whereas types D, E, F, and SCC and
SCLC were hilic (solid) tumour growth [12, 13].

Statistical analysis
Data analysis was performed with the SPSS
analysis program (SPSS Inc., Chicago, IL). The
pooled-variance t-test was used to compare the
mean values of size and CT values between visible
and invisible tumours, or between the CT screening group and the clinical group, as well as the
mean values of contrast and gradient between
tumours with distinct and indistinct borders on
CXRs. One-way analysis of variance followed by
Bonferroni method of multiple comparisons was
used to compare the mean values of the size,
contrast, gradient and CT values among different
tumour types. x2 test of association was used to
assess the signicance of difference in proportion
between the CT screening group and the clinical
The British Journal of Radiology, January 2001

Visibility of small peripheral lung cancers on CXRs

group. A statistically signicant difference was

considered to be present when the p-value was less
than 0.05.

Results
We examined 63 small peripheral lung cancers
located in well penetrated lung elds on CXRs. 31
(49%) were visible and 32 (51%) were invisible on
CXRs. Examples of these tumours are illustrated
in Figures 13.

Visibility of tumours on CXRs based on

tumour size, contrast, gradient, CT values
and position
Table 1 shows the factors that inuenced
visibility of tumours on CXRs. The size of visible
tumours was larger than that of invisible tumours;
the minimum tumour size detected on CXRs was
9 mm. In general, tumours 10 mm in diameter
were difcult to detect on CXRs, with a detection
rate of only 18% (4/22). In contrast, the detection
rate of tumours .10 mm was 66% (27/41)
(Figure 4; p,0.001). Visible tumours had
an optical density of 0.100.30 OD and a
gradient of 0.030.11 OD mm21. The mean CT
value of visible tumours was higher than that
of invisible tumours. Furthermore, DCT between
the tumour and surrounding lung parenchyma
was greater for visible tumours than invisible
tumours. Tumours 400 HU DCT were difcult
to detect on CXRs, with a detection rate of 8% (2/
24), whereas tumours .400 HU DCT had an
improved detection rate of 74% (29/39) (Figure 4;
p,0.001).
The detection rate of tumours on CXRs was
40% (8/20) for the upper lobe, 56% (5/9) for the
middle lobe and 58% (7/12) for the lower lobe
of the right lung; and 36% (4/11) for the upper
lobe and 64% (7/11) for the lower lobe of the left
lung.

Size, contrast, gradient and CT values of

tumours based on histopathological tumour
type
Table 3 summarizes the size, contrast, gradient
and CT values of tumours according to histopathological type. The size of type A tumours was
lower than type B, type C, type DF, and SCC/
SCLC (Bonferroni method of multiple comparisons, p50.039, p,0.001, p,0.013 and p,0.015,
respectively; Table 3). The contrast of type B
tumours was lower than type C, type DF, and
SCC/SCLC (Bonferroni method of multiple
comparisons, p50.005, p50.017 and p50.017,
respectively; Table 3). Analysis of the gradient
of tumours based on their histopathological type
showed a lower gradient for type B tumours
relative to type DF, and SCC/SCLC (Bonferroni
method of multiple comparisons, p50.013 and
p50.013, respectively; Table 3).
Regarding CT values of tumours based on their
histopathological type, CT values of type A or
type B tumours were lower than type C, type DF,
and SCC/SCLC, and values of type C tumours
were lower than type DF, and SCC/SCLC.
Analysis of the DCT of tumours based on their
histopathological type showed a smaller DCT for
type A or type B tumours relative to type C, type
DF, and SCC/SCLC, and a smaller DCT for
type C tumours relative to type DF, and SCC/
SCLC (Table 3).

Border of 31 visible tumours on CXRs

The entire margin of 9 of 31 tumours was
indistinct; that of 18 was partially indistinct and
partially distinct; and the entire margin was
distinct in 4. The mean contrast (0.21 OD) and
gradient (0.08 OD mm21) of tumours with a
distinct border were higher than those
(0.14 OD, 0.05 OD mm21) of tumours with an
indistinct border (p,0.001).

Visibility of tumours on CXRs based on

histopathological tumour type

Characteristics of tumours detected in CT

screening group vs clinical group

Table 2 summarizes the visibility of 63 tumours

according to histopathological type. 31 visible
tumours consisted of 27 adenocarcinomas (4 of
type B, 17 of type C, 6 of type DF), 3 SCCs
(Figure 1) and 1 SCLC. On the other hand, 32
invisible tumours consisted of type A (Figure 2),
type B (Figure 3) and type C adenocarcinoma.
Specically, types DF adenocarcinoma, SCC and
SCLC were the most easily detectable on CXRs,
while type A was the least detectable.

We also compared the characteristics of

tumours detected in the CT screening group and
the clinical group (Table 4). The mean CT values
of tumours in the CT screening group were lower
than those of the tumours in the clinical group,
while the mean DCT of tumours in the CT
screening group was smaller than that of tumours
in the clinical group. Furthermore, tumours with
lepidic growth were detected more often in the CT
screening group than the clinical group (Table 4).

The British Journal of Radiology, January 2001

35

Z-G Yang, S Sone, F Li et al

(a)

(b)

(c)

(d)

(e)

36

Figure 1. 67-year-old male smoker with squamous cell

carcinoma (16 mm616 mm) in the apical segment of
the right lower lobe, with hilic growth pattern.
(a) Low dose CT shows a small nodule (arrow).
(b) High resolution CT shows a homogeneous solid
tissue density nodule (arrow). (c) Histopathological
examination of the tumour shows hilic (solid) tumour
growth (H&E stain; original magnication61.25).
(d) Close-up view of the posteroanterior chest radiograph shows the lung nodule (arrow). (e) Optical densitometric measurement of visible tumour on the
chest radiograph. The contrast and gradient values
are 0.21 OD and 0.11 OD mm21, respectively, which
corresponds to the histopathological ndings of hilic
tumour growth.
The British Journal of Radiology, January 2001

Visibility of small peripheral lung cancers on CXRs

(a)

(b)

(c)

(d)

Figure 2. 56-year-old male smoker with a well differentiated type A adenocarcinoma (11 mm612 mm) in the
apical segment of the right lower lobe, with lepidic growth pattern. (a) Low dose CT shows a small faint lesion
(arrow). (b) High resolution CT shows a ground-glass attenuation nodule through which the small vessels are visible (arrow). (c) Histopathological examination shows alveolar lining tumour growth without alveolar collapse
(H&E stain; original magnication61.25). (d) Close-up view of the posteroanterior chest radiograph shows no
evidence of the nodule in the middle zone of the right lung eld.

Discussion
Perception of the pulmonary lesion is inuenced by various factors, including the observer,
lesion characteristics, and lm or technical parameters. The focus of this study was to determine the
effects of nodule characteristics, particularly
tumour density and tumour type, on the visibility
of small peripheral lung cancer on CXRs.
Our results showed that the minimum optical
The British Journal of Radiology, January 2001

contrast of small visible tumours on CXRs was at

least 0.1 OD. This is consistent with previous
studies [14], but is different from the minimum
contrast of 0.03 OD that has been cited as a
threshold density for a nodule placed on a
homogeneous background to be detected by a
naked eye [15]. This phenomenon may be
explained by the presence of complex surrounding
anatomical structures, such as pulmonary vessels
and/or ribs, even if the nodule is located in a well
37

Z-G Yang, S Sone, F Li et al

(a)

(b)

(c)

(d)

Figure 3. 33-year-old male non-smoker with a well differentiated type B adenocarcinoma (8 mm69 mm) in the
apical segment of the left lower lobe, with lepidic growth pattern. (a) Low dose CT shows a small, faint lesion
(arrow). (b) High resolution CT shows a heterogeneous, low attenuation nodule (arrow). (c) Histopathological
examination shows alveolar lining tumour growth with scattered foci of alveolar collapse (H&E stain; original
magnication61.25). (d) Close-up view of the posteroanterior chest radiograph shows no evidence of the nodule
in the middle zone of the left lung eld.

penetrated lung eld on the CXR, as described

previously [14, 16]. However, to our knowledge,
our report is the rst to describe measurement of
contrast and gradient of visible tumours, as well
38

as corresponding CT values, and compares these

values with those of invisible tumours, and with
histopathological tumour type.
Our data showed that the contrast of tumour
The British Journal of Radiology, January 2001

Visibility of small peripheral lung cancers on CXRs

Table 1. Factors inuencing the visibility of 63 tumours on chest radiographs
Factor

Visible (n531)a

Invisible (n532)a

p-valueb

Tumour size (mm)

Contrast (OD)
Gradient (OD mm21)
CT values (HU)
DCT (HU)

14.32.8 (919)
0.170.04 (0.100.30)
0.070.02 (0.030.11)
2140180 (2540 to 30)
700170 (330890)

11.13.6 (618)

2490180 (2690 to 2140)

380140 (170690)

,0.001

,0.001
,0.001

a
b

Data are meanSD. The numbers in parentheses are the range.

Pooled-variance t-test.

nodules correlated with tumour growth pattern.

Tumour growth of peripheral lung cancer has
traditionally been divided into two patterns,
lepidic growth and hilic growth. Lepidic growth
is basically alveolar lining tumour cell growth,
which maintains more or less air-lled alveolae in
the tumour; while hilic growth is characteristically
a solid tumour growth, which frequently enlarges,
displacing the surrounding lung parenchyma [13].
Lepidic tumour growth pattern, frequently seen in
adenocarcinoma, is often associated with a
variable degree of alveolar aeration within the
tumour, yielding a low density lesion and subsequently resulting in a low detection rate. On the
other hand, hilic tumour growth pattern is
associated with growth of solid nodules with
little air spaces, subsequently resulting in higher
contrast and a higher detection rate. In the
present study, we classied adenocarcinoma
according to Noguchi's classication into types
A, B, C, D, E and F. Type AC adenocarcinomas
represent tumours with a lepidic growth pattern,

while type DF adenocarcinomas exhibit a hilic

growth pattern [12, 13]. The retained aeration
within lepidic tumours results in a faint opacity
(ground-glass attenuation) nodule, and thus low
contrast and a very low detection rate on CXRs.
Type DF adenocarcinomas are hilic tumours and
characteristically appear as homogeneous solid,
soft tissue attenuation nodules; all such tumours
were detected on CXRs when located in well
penetrated lung eld. Our results indicated that
the gradient was also inuenced by tumour type.
For example, type B adenocarcinomas, which
showed the least degree of alveolar lining, had
vague borders and lowest gradient (mean gradient
0.04 OD mm21), whereas type DF adenocarcinomas, which exhibited a hilic growth pattern,
had distinct borders and a high gradient (mean
gradient, 0.08 OD mm21). This factor may result
in a poor detection rate for low density tumours.
It is well known that tumour size is an
important determinant of visibility of nodular
lesions. Previous studies have shown that approximately 50% of lesions (mostly simulated nodules)
of 810 mm in diameter were visible on CXRs [3,
16]. In our study, the detection rate of small lung
cancers measuring 10 mm in diameter was only
18%, while the detection rate of those measuring
1120 mm was 66%. The inconsistent results
between our study and those of previous reports
could be owing to different morphological
features of the tumours in terms of density and
border of the nodule. The simulated nodule that
Table 2. Visibility of 63 tumours on chest radiographs according to histological type

Figure 4. Relationship between tumour size and the

difference in CT values between the tumour and its
surrounding lung parenchyma (DCT) in visible and
invisible tumours in cases detected on CT screening
and those detected clinically. The detection
rates on chest radiographs were 18% (4/22) for
tumours 10 mm in diameter and 66% (27/41) for
tumours 1120 mm in diameter, and 8% (2/24)
for tumours 400 HU DCT and 74% (29/39)
for tumours .400 HU DCT. n, Invisible tumour in
CT screening group; m, visible tumour in CT screening group; #, invisible tumour in clinical group;
$, visible tumour in clinical group.
The British Journal of Radiology, January 2001

Histological typea Visible Invisible Detection rate

(%)
Adenocarcinoma
Type A
Type B
Type C
Type DF
SCC/SCLC

0
4
17
6
4

14
10
8
0
0

0
29
68
100
100

Total

31

32

49

Type A, B, C, D, E and F according to Noguchi's

classication for adenocarcinoma [12]; SCC, squamous cell
carcinoma; SCLC, small cell lung carcinoma.

39

Z-G Yang, S Sone, F Li et al

Table 3. Size, contrast, gradient and CT values of tumours according to histological type
Histological Typea
Adenocarcinoma
Type A
Type B
Type C
Type DF
SCC/SCLC

Size (mm)
(n563)

Contrast (OD)
(n531)

Gradient (OD mm21)

(n531)

CT values (HU)
(n563)

DCT (HU)
(n563)

9.82.9
12.63.9*
13.83.0*
13.72.9*
16.54.0*

0.110.01
0.180.05{
0.180.05{
0.230.15{

0.040.02
0.060.02
0.080.02"
0.080.02"

2550130
2470130
2210200{
1010{
2010{

330130
400110
640170{
83020{
84020{

All data are meanSD.

a
Type A, B, C, D, E and F according to Noguchi's classication for adenocarcinoma [12]; SCC, squamous cell carcinoma; SCLC,
small cell lung carcinoma.
* Signicant difference (p,0.05) beween type A and type B, type C, type DF or SCC/SCLC.
{ Signicant difference (p,0.05) beween type B and type C, type DF or SCC/SCLC.
{ Signicant difference (p,0.05) beween type A or type B and type C, type DF or SCC/SCLC, and between type C and type DF or
SCC/SCLC.
" Signicant difference (p,0.05) beween type B and type DF or SCC/SCLC.

Table 4. Characteristics of tumours in CT screening group vs clinical group

Factor

CT screening group
(n548)

Clinical group
(n515)a

p-value

Tumour size (mm)a

CT values (HU)a
DCT (HU)a
Tumour growthb
Lepidic
Hilic

12.43.7 (619)
2390230 (2690 to 30)
480210 (170880)

13.62.9 (919)
2120150 (2390 to 30)
710140 (440890)

0.254
,0.001
,0.001
0.003

a
b

44
4

9
6

Data are meanSD. Numbers in parentheses are the range.

Lepidic growth included type A, B and C adenocarcinoma according to Noguchi's classication [12]; hilic growth included type D
F adenocarcinoma, squamous cell carcinoma and small cell lung carcinoma.

frequently showed well marginated and dense

characteristics was different from the small
peripheral lung cancer; the majority of the latter
were often of inhomogeneous density and had an
indistinct border.
According to a previous report, position of the
tumour was one of the factors inuencing
visibility of the lesion, the upper lobe being the
leading site for missed lung cancer [1]. In this
study, we found a preference of the upper lobe of
the right lung for lung cancer, with lower
detection rates than the lower or middle lobes,
indicating that the position of the tumour in the
well penetrated lung eld also inuenced visibility
of the tumour on CXRs.
In general, tumour size, contrast and gradient
directly inuenced the visibility of the tumour.
Small lung cancers become visible on CXRs when
their contrast exceeds the cut-off value of
0.10 OD. Furthermore, histopathological tumour
types inuenced the contrast, gradient, CT values
and margin of the tumour, and their visibility
varied on CXRs. Based on these ndings, we
propose that small tumours of the alveolar lining
growth type are difcult to detect on CXRs owing
to their low contrast and gradient.
40

This study showed that tumours detected in the

CT screening group had lower CT values and
smaller DCT compared with tumours detected in
the clinical group, and the majority of the former
were invisible on CXRs. These differences were
attributable to the fact that low dose CT could
demonstrate small-sized, low density adenocarcinoma with lepidic growth, particularly type A
adenocarcinoma, which was difcult to detect on
CXRs taken during routine clinical examination
or chest radiography screening for lung cancer.

Acknowledgments
The authors are indebted to Dr Takeshi
Yamanda and Dr Masayuki Haniuda, from the
Department of Surgery, and Dr Keishi Kubo,
from the Department of Internal Medicine, for
their collaboration in the present study. We also
thank Kazuhisa Hanamura, BS, and Kazuhiro
Asakura, EE, from the Telecommunications
Advancement
Organization
of
Japan
Matsumoto Research Center for their technical
support.
The British Journal of Radiology, January 2001

Visibility of small peripheral lung cancers on CXRs

References
1. Austin JHM, Romney BM, Goldsmith LS. Missed
bronchogenic carcinoma: radiographic ndings in
27 patients with a potentially resectable lesion
evident in retrospect. Radiology 1992;182:11522.
2. Quekel LGBA, Kessels AGH, Goei R, Engelshoven
JMA. Miss rate of lung cancer on the chest
radiograph
in
clinical
practice.
Chest
1999;115:7204.
3. White CS, Salis AI, Meyer CA. Missed lung cancer
on chest radiography and computed tomography:
imaging and medicolegal issues. J Thorac Imaging
1999;14:638.
4. Kaneko M, Eguchi K, Ohmatsu H, et al. Peripheral
lung cancer: screening and detection with low-dose
spiral CT versus radiography. Radiology 1996;
201:798802.
5. Sone S, Takashima S, Li F, et al. Mass screening
for lung cancer with mobile spiral computed
tomography scanner. Lancet 1998;351:12425.
6. Henschke CI, McCauley DI, Yankelevitz DF, et al.
Early lung cancer action project: overall design and
ndings from baseline screening. Lancet 1999;354:
99105.
7. Diederich S, Lenzen H, Windmann R, et al.
Pulmonary nodules: experimental and clinical
studies at low-dose CT. Radiology 1999;213:28998.
8. Woodring JH. Pitfalls in the radiologic diagnosis of
lung cancer. AJR 1990;154:116575.

The British Journal of Radiology, January 2001

9. Yang ZG, Sone S, Takashima S, Li F, Honda T,

Yamanda T. Small peripheral carcinomas of lung:
thin-section CT and pathologic correlation. Eur
Radiol 1999;9:181925.
10. Revesz G, Kundel HL, Toto LC. Densitometric
measurements of lung nodules on chest
radiographs. Invest Radiol 1981;16:2015.
11. Revesz G. Conspicuity and uncertainty in the
radiographic detection of lesions. Radiology 1985;
154:6258.
12. Noguchi M, Morikawa A, Kawasaki M, et al. Small
adenocarcinoma of the lung: histologic characteristics
and
prognosis.
Cancer
1995;75:
284452.
13. Heitzman ER, Markarian B, Raasch BN, Carsky
EW, Lane EJ, Berlow ME. Pathways of tumor
spread through the lung: radiologic correlations
with anatomy and pathology. Radiology 1982;144:
314.
14. Kundel HL, Revesz G, Toto L. Contrast gradient
and the detection of lung nodules. Invest Radiol
1979;14:1822.
15. Dyke WP, Barbour JP, Charbonnier FM. Depth
resolution: a mechanism by which high kilovoltage
improves visibility in chest lms. Radiology 1975;
117:15964.
16. Brogdon BG, Kelsey CA, Moseley RD Jr. Factors
affecting perception of pulmonary lesions. Radiol
Clin North Am 1983;21:63354.

41