DOI:10.2214/AJR.07.2870
AJR 2008; 191:W80-W88
© American Roentgen Ray Society
Image Quality of Digital Direct Flat-Panel Mammography Versus an Analog Screen-Film Technique Using a Low-Contrast Phantom
Kathrin Barbara Krug1,
Hartmut Stützer2,
René Schröder1,
Julia Boecker1,
Jörg Poggenborg1 and
Klaus Lackner1
1 Department of Radiology, University of Cologne Medical School,
Kerpenerstraße 62, Cologne NRW 50924, Germany.
2 Institute of Medical Statistics, Informatics and Epidemiology, University of
Cologne, Cologne, Germany.
Received July 15, 2007;
accepted after revision March 28, 2008.
Address correspondence to K. B. Krug
(barbara.krug{at}uk-koeln.de).
WEB
This is a Web exclusive article.
Abstract
OBJECTIVE. The objective of our study was to compare the
detectability and distinguishability of simulated soft-tissue opacities of 50
variants of an anthropomorphic breast phantom in mammograms acquired with a
digital direct flat-panel detector versus an analog system; we also compared
the image settings "analog film," "digital film," and
"digital monitor."
MATERIALS AND METHODS. The studies were performed on digital (Lorad
Selenia) and analog (Mammomat 3) mammography systems. Four hundred fifty
silicone cubes devised with different randomly distributed columns, holes, or
both columns and holes (diameter, 3–7 mm; height, 0.5–4.0 mm) were
used as test bodies. One experimental series was performed with a silicone
scatter body and one with a silicone and an anthropomorphic ground-meat
scatter body. All x-rays were obtained at identical settings and exposures.
Four radiologists rated the films and monitor-displayed images independently
of each other in randomized order on a standardized electronic
questionnaire.
RESULTS. The digital monitor technique generally scored better than
digital film viewing and analog readings. The McNemar test for multiple paired
comparisons mostly yielded a p value of < 0.0005. The smallest
volume category counted as the most valid test scenario for all raters, where
the percentage of correct positive findings ranged between 30% and 58% (analog
technique), 43% and 68% (digital film viewing), and 55% and 66% (monitor
viewing). The corresponding accuracy rates were 77–93%, 75–95%,
and 81–85%, respectively, with kappa values of 0.2–0.5 (analog)
and 0.3–0.6 (digital) for comparing the gold standard with raters'
evaluations.
CONCLUSION. Digital flat-panel mammography is superior to the analog
screen-film method for the detection of simulated opacities.
Keywords: breast cancer screening • digital flat-panel mammography • mammography • screen-film imaging
Introduction
The technologic demands placed on mammographic x-ray imaging systems are
particularly great. First, the dynamic range must be broad enough to enable
simultaneous visualization of both structures with strong x-ray absorption,
such as calcifications, and those with poor absorption, such as fatty tissue.
Second, a comparatively high local image resolution is required to allow
detection and morphologic characterization of microcalcifications. The most
important of the currently commercially available mammography technologies
include computed radiography systems based on digital flat-panel detector
radiography that use indirect conversion, systems based on digital flat-panel
radiography with direct conversion and photostimulable phosphor plates, and
systems using linear quantum-counting digital x-ray detectors
[1–8].
Digital flat-panel detector radiography using indirect conversion takes
place in two steps. In the first step, a scintillator layer made of gadolinium
oxysulfide or cesium iodide captures x-ray quanta and converts them to light.
In the second step, the light is converted to electrical charges in
photodiodes made of amorphous silicon. In digital flat-panel radiography using
direct conversion, x-ray quanta are converted directly to electrical charges
in a layer of amorphous selenium. Direct flat-panel detectors are one of the
more recent technologic approaches and, compared with other methods, offer the
advantages of a higher quantum efficiency because x-ray conversion takes place
in just one step. Thanks to the chemical properties of amorphous selenium,
imaging of a larger area is possible as well.
The results of recent experimental studies on digital mammography have
shown that the latest advances in the aforementioned technologies make them at
least equivalent to analog screen-film mammography in terms of image quality
and diagnostic accuracy
[8–12].
However, in these studies investigators mainly focused on evaluating
visualization of microcalcifications. Our study aimed to go a step further by
determining on a larger sample and using an anthropomorphic model the
following: first, whether the detection rate of varying round opacities
achievable with direct digital flat-panel mammography (test method) is
comparable with the diagnostic profile of an analog screen-film technique
(reference method); second, whether workstation viewing is diagnostically
equivalent with or superior to film reading in the low-contrast range.
Materials and Methods
Model
The test device consisted of a square box made of acrylic plastic sheets
that was filled with nine silicone cubes with a square surface area of 3
x 3 cm edge length and 1.5 cm in height to create a matrix of three rows
(
–
) times three columns (1–3)
(Fig. 1). For loading the
acrylic boxes, 84 different cubes were provided, from which 450 combinations
of nine were created. Each cube was conceptually divided into four subcubes
(a, b, c, d). Seventy-five cubes contained round silicone columns or drilled
holes that were devised to simulate masses of high and low density that
contrasted with the surrounding structures. Nine cubes were kept empty. The
diameter of the columns or holes varied between 3 and 7 mm in 1-mm increments;
the heights were staggered in 0.5-mm increments of 0.5–4.0 mm. The
acrylic box was packed with different randomly selected cubes a total of 50
times. Data on the number, localization, diameter, and height of the columns
and holes were compiled as the gold standard in a data-processing system.
During scanning, each of the nine cubes was assigned to one of the nine main
fields (1–3) within the electronic data matrix in accordance
with its localization. Each main field was subdivided again into small squares
(a–d) so that each column or hole was unequivocally assignable to one of
the subcubes (1a–3d).
View larger version (132K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1 —Schematic shows box made of square acrylic plastic sheets
that was packed in 50 variations with nine different silicone cubes and that
served as phantom for study. Each cube had surface area of 3 x 3 cm edge
length and 1.5 cm in height. Some cubes contained either round silicone
columns or holes devised to simulate masses with elevated or low density to
contrast with surroundings. Some cubes had no simulated masses.
|
|
Mammograms of each of the 50 different box arrangements were obtained on
digital (Lorad Selenia, Lorad-Hologic) and analog (Mammomat 3, Siemens Medical
Solutions) mammography systems. A 3-mm-thick silicone disk, which was placed
under the acrylic box before x-ray expo sure, was used as a scatter body in
all of the x-ray exposures. This phantom setup was used in the first test
series designed to simulate fatty breasts. In the second test series, designed
to imitate dense breasts, the identical examination and exposure settings as
in series 1 were used, but an additional 3-mm layer of ground meat was applied
around the test device as a so-called ground-meat scatter body.
Because the two x-ray systems were located in one spatial unit, the digital
and analog images could be obtained in direct succession without changing the
configuration of the scatter body. To produce the most realistic absorption
differences in x-ray images, the acrylic box was filled to the brim with
liquid cream (series A–C) or water (series D–F); we ensured that
no air bubbles remained that might have mimicked indentations. In summary,
series A–C, which were supposed to simulate fatty breasts, were given a
scatter body of silicone and were filled with cream. Series D–F,
supposed to simulate dense breasts, included scatter bodies of silicone and of
meat and were filled with water.
Examination Technique
The full-field digital mammography system (Lorad Selenia, Lorad/Hologic)
was equipped with a double-focus, bimetal anode with a 25-µm molybdenum
filter. The nominal focal spot size was 0.3 mm for the survey images, and the
focus–film distance was 65 cm. In the honeycomb-shaped grid, air was
used as the interseptum material to suitably absorb scatter radiation more
strongly than utilizable radiation in all transverse directions. The
flat-panel detector consisted of a semiconductor layer of amorphous selenium
that was placed under a direct current before x-ray exposure. The absorption
of x-rays causes local equalization of the charges that are captured on an
array of electrodes, storage capacitors, and transistors lo cated behind the
selenium layer.
After electronic enhancement and analog-to-digital signal conversion, the
digital image data were transmitted to the imaging PC. Edge enhancement was
applied to the digital images, as is routinely done for reading clinical
images. The active field of view was 24 x 29 cm; the matrix had 3,328
x 4,096 pixels; and the pixel edge length was 70 x 70 µm,
corresponding to a nominal local resolution of 7.2 lp/mm. Data acquired in a
format of 18 x 24 cm and 2,560 x 3,328 pixels of the active field
of view were used for imaging. Before the study was initiated, the
manufacturer's image-processing settings were optimized by systematic
variation and then were not changed during the examinations.
The digital imaging data sets were printed on a dry laser printer that
operates on the principle of direct thermography (DryView 8610, Kodak). The
geometric print resolution was 655 dots per inch, the effective pixel edge
length was 36 x 39 µm in a matrix of 5,025 x 6,200 pixels, and
the contrast resolution was 12-bit gray-scale depth (4,096 gray levels). An
infrared-sensitive photothermographic film with a blue film base coating on
one side was used in a 18 x 24 cm format (DryView Laser Imaging Film,
Kodak).
The system was connected to a viewing monitor (Selenia Softcopy
Workstation, Lorad/Hologic, with MeVis Breast Care software package,
Lorad/Hologic). The graphic controllers (model 5MP1H, Barco) operated with an
internal 12-bit gray-scale depth (4,096 gray levels). For monitor viewing, a
digital-to-analog converter converted the signal for a display depth
equivalent of 10 for the gray-scale depth (1,024 gray levels). The two
monitors had an image field size of 30 x 40 cm in diameter with a line
resolution of 2,048 x 2,560 pixels (effective pixel edge length, 147
x 156 µm). Whenever a scan with a format of 18 x 24 cm was
completely displayed on the monitor, the geometric resolution was reduced by a
factor of 0.8, down to 7.0 lp/mm. For a 1:1 reproduction of the digital image
data set, the image was cut off slightly on one edge, producing an effective
display resolution of approximately 8 lp/mm. Through bicubic inter polation,
in addition, interactive zooming (factor of 2) of the digital image data set
was possible using the optical magnifying glass option.
The analog mammography unit (Mammomat 3, Siemens Medical Solutions) was
equipped with a double-focus molybdenum–tungsten bimetal anode. The
nominal focal spot size for survey images was 0.3 mm, the focus–film
distance was 60 cm, and a molybdenum filter (up to 40 kV, 0.03 mm) was used.
The images were reproduced with single-coated MIN-R 2190 screens (Eastman
Kodak) and MIN-R 2000 mammography films (Eastman Kodak) with a blue base
support. The films had a format of 18 x 24 cm and were developed on a
daylight load system (Miniloader Model 2 Plus, Eastman Kodak). The nominal
local resolution of the films was 12 lp/mm.
The two systems were serviced before the study. The specimens were imaged
with the unit positioned for a craniocaudal exposure. The distance between the
focus and the detection level was the same in all x-ray exposures. Preliminary
tests were conducted by systematically changing the settings between 20 and 32
kV using the automatic exposure control. Afterward, all analog and digital
scans were obtained at a tube voltage of 29 kV and tube current of 18 mAs.
The entrance skin absorption was 1.925 mGy for the Selenia and 1.875 mGy
for the Mammomat 3, and the average glandular dose was 0.593 mGy for the
Selenia and 0.588 mGy for the Mammomat 3. The doses were measured with a
calibrated Solidose 300 (RTI Electronics). The manufacturer states a measuring
error for the dosimeter of ± 5%. This test proved that any minor
differences in exposure doses observed between the analog and digital methods
were within the range of error. Thus, the two machines were comparable in
terms of overall exposure levels.
Image Evaluation
Independently of one another, four raters, radiologists with 2–18
years' experience in ana log mammography (raters 1–4: 18, 4, 6, and 2
years, respectively) and 2–3 years' experience in digital mammography
(raters 1–3, 3 years; rater 4, 2 years), evaluated the analog films,
digital im ages, and digital monitor images in a randomized order. Before
interpreting the digital images, the raters received a brief introduction in
the use of the image-processing program and were given time to practice.
The raters' notations were entered in a structured electronic
questionnaire. The electronic screen consisted of three rows
(–) and three columns (1–3) as a schematic of the
phantom. Each of the nine fields thus defined was subdivided into another four
small squares (a–d) so that 36 interpretations (fields
1a–3d) had to be rated per image. A total of six test series
(A, analog with just a silicone scatter body; B, digital film with a silicone
scatter body; C, viewing workstation with a silicone scatter body; D, analog
film with a silicone and a ground-meat scatter body; E, digital film with a
silicone and a ground-meat scatter body; F, viewing workstation with a
silicone and a ground-meat scatter body) with 50 images per test series and
1,800 individual items were available. In each small square a–d, the
rater stated whether a "mass" was present in the field in question
and, if so, whether the mass had a positive (column) or negative (hole)
contrast to its surroundings. If a mass was thought to be present, the rater
was required to enter the diameter of the lesion (3, 4, 5, 6, or 7 mm) and the
rater's confidence with regard to the finding (1, finding certainly present;
2, finding probably present; and 3, finding questionably present).
Statistical Analysis
The notations made by the four raters were classified separately according
to the exposures with a silicone scatter body only (series A, B, C simu lating
fatty breasts) and with a silicone scatter body and an additional ground-meat
scatter body (series D, E, F simulating dense breasts) and according to the
imaging modes—analog (series A and D), digital film (series B and E),
and digital monitor (series C and F)—for the dichotomous and rank-scaled
parameters positive versus neg ative contrast; the number, diameter, and
height of the columns and holes; and rater confidence. The raters' notations
were compared with experi mentally predefined reference values (i.e., the gold
standard). With a given density of a mass and its surroundings, the
detectability of a lesion is dictated by its diameter and height. For that
reason, the volumes of the individual columns and holes were calculated from
the radius r and height h of the simulated masses using the
following formula for straight circular cylinders:
 |
Then, each cube was assigned to one of five volume categories, where category
0 was defined as 0.0 mm3 (no silicone column or drill hole);
category I as 0.1–< 3.8 mm3; category II, 3.8–<
7.7 mm3; category III, 7.7–< 13.1 mm3; and
category IV, 
13.1 mm3. There were 1,212 cubes in category 0,
139 in category I, 159 in category II, 143 in category III, and 147 in
category IV.
View larger version (153K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 2A —Radiographs of phantom shown in
Figure 1 containing columns
(positive contrast) and holes (negative contrast) between 3 and 7 mm in
diameter and 1 and 4 mm in height. Analog image of phantom without ground-meat
scatter body.
|
|
View larger version (140K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 2B —Radiographs of phantom shown in
Figure 1 containing columns
(positive contrast) and holes (negative contrast) between 3 and 7 mm in
diameter and 1 and 4 mm in height. Digital image of phantom without
ground-meat scatter body.
|
|
View larger version (22K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3A —Bar graphs of number of correct positive findings (ordinate)
according to raters 1–4 and images in series A (analog film), B (digital
film), and C (digital monitor viewing) without scatter body made of ground
meat (abscissa). Black solid horizontal lines indicate respective number of
cubes with simulated masses. Each rater's stated degree of confidence is given
on 3-item ordinal scale: Finding is certainly present (black),
probably present (gray), or questionably present (white).
Raters 1–4 had 18, 4, 6, and 2 years of experience, respectively, with
analog mammography; raters 1–3 had 3 years of experience with digital
mammography and rater 4, 2 years. Category I (volume, 0.1–< 3.8
mm3, 139 cubes).
|
|
View larger version (28K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3B —Bar graphs of number of correct positive findings (ordinate)
according to raters 1–4 and images in series A (analog film), B (digital
film), and C (digital monitor viewing) without scatter body made of ground
meat (abscissa). Black solid horizontal lines indicate respective number of
cubes with simulated masses. Each rater's stated degree of confidence is given
on 3-item ordinal scale: Finding is certainly present (black),
probably present (gray), or questionably present (white).
Raters 1–4 had 18, 4, 6, and 2 years of experience, respectively, with
analog mammography; raters 1–3 had 3 years of experience with digital
mammography and rater 4, 2 years. Category II (volume, 3.8–< 7.7
mm3, 159 cubes).
|
|
View larger version (30K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3C —Bar graphs of number of correct positive findings (ordinate)
according to raters 1–4 and images in series A (analog film), B (digital
film), and C (digital monitor viewing) without scatter body made of ground
meat (abscissa). Black solid horizontal lines indicate respective number of
cubes with simulated masses. Each rater's stated degree of confidence is given
on 3-item ordinal scale: Finding is certainly present (black),
probably present (gray), or questionably present (white).
Raters 1–4 had 18, 4, 6, and 2 years of experience, respectively, with
analog mammography; raters 1–3 had 3 years of experience with digital
mammography and rater 4, 2 years. Category III (volume, 7.7–< 13.1
mm3, 143 cubes).
|
|
View larger version (29K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3D —Bar graphs of number of correct positive findings (ordinate)
according to raters 1–4 and images in series A (analog film), B (digital
film), and C (digital monitor viewing) without scatter body made of ground
meat (abscissa). Black solid horizontal lines indicate respective number of
cubes with simulated masses. Each rater's stated degree of confidence is given
on 3-item ordinal scale: Finding is certainly present (black),
probably present (gray), or questionably present (white).
Raters 1–4 had 18, 4, 6, and 2 years of experience, respectively, with
analog mammography; raters 1–3 had 3 years of experience with digital
mammography and rater 4, 2 years. Category IV (volume, 13.1
mm3, 147 cubes).
|
|
The raters' diagnostic accuracy was plotted for all 1,800 individual items
and for the individual volume categories per series. The subsequent
evaluations counted only yes–no scores on the presence of a column or
hole irrespective of the raters' stated degree of confidence with regard to
positive classifications. First, the sensitivity, specificity, and accuracy
rate (percentage of correct ratings across all classifications) were
determined for all images of each series as well as separately according to
rater and volume category. To obtain a general impression of the results, each
rater's correct positive and correct negative findings were presented in
comparative graphs for each series as well as separately according to the
volume categories.
Kappa values were calculated as a measure of the agreement of all raters'
classifications per series with the reference standard. A negative kappa value
or a kappa value equaling zero indicated a lack of agreement or a purely
coincidental agreement, respectively, and a kappa equaling 1 indicated 100%
agreement with the reference standard. The odds ratios (with 95% CIs) of the
diagnostic relevance of each individual rating were calculated from the data
on the test and reference series. An odds ratio of 1 meant that the likelihood
of a positive rating (i.e., lesion detected) of an actually existing column or
hole was just as great as for an actually nonexistent lesion. If a lesion
actually existed, an odds ratio of > 1 meant that the likelihood of a
positive rater score was greater than if a lesion actually did not exist, and
an odds ratio of < 1 meant that the likelihood of a positive rating was
smaller than if a lesion actually did not exist. Accordingly, the target odds
ratios ought to consistently be > 1 for all raters in one test
procedure.
Results
In the images acquired with only a silicone scatter body (series A–C
simulating fatty breasts) (Figs.
2A and
2B), the detectability of the
columns and holes was higher than for the images acquired with a silicone
scatter body and an additional ground-meat scatter body (series D–F
simulating dense breasts) (Figs.
2C and
2D). Diagnostic accuracy was
superior using the digital technique compared with analog films through the
series A–C and the series D–F. Rater diagnostic accuracy and the
number of correct positive findings increased proportionately with increasing
volume of columns and holes (Fig.
3A,
3B,
3C,
3D). The contrast type
(positive contrast = column, negative contrast = hole) did not prove to have
any relevant impact on diagnostic accuracy
(Fig. 4).
View larger version (38K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 4 —Sensitivity as function of contrast (positive contrast =
column on silicone cube, negative contrast = hole in silicone cube; compare
with Fig. 1). Raters 1–4
had 18, 4, 6, and 2 years of experience, respectively, with analog
mammography; raters 1–3 had 3 years of experience with digital
mammography and rater 4, 2 years.
|
|
Table 1 presents the
percentages of correct positive findings (sensitivity); the percentages of
false-negative findings are the difference to 100%. In the series without a
ground-meat scatter body, the raters correctly identified 57–77% of all
588 columns and holes when evaluating analog images (series A), 58–89%
of all 588 simulated masses when evaluating digital films (B), and
69–84% of all 588 artifacts when evaluating images on the viewing
workstation (C). The digital technique proved diagnostically superior to the
analog technique in the most valid test scenario—that is, volume
categories I and II. In category I, the percentage of correct positive
findings ranged between 30% and 58% (series A), 43% and 68% (B), and 55% and
66% (C); in category II, the percentage of correct positive findings ranged
between 43% and 67% (series A), between 51% and 93% (B), and between 64% and
81% (C). The percentages of correct positive findings were 69–91%
(series A), 67–99% (series B), and 82–100% (series C) for category
III; and 84–97% (A), 72–100% (B), and 86–100% (C) for
category II. The balanced results obtained for the larger volumes are most
likely attributable to inter- and intrarater variability rather than to the
examination technique because of the comparably easy detectability of
simulated lesions.
View this table:
[in this window]
[in a new window]
|
TABLE 1: Percentage of Correct Positive Findings (Sensitivity) Among the 588
Subcubes with Simulated Lesions (Columns and Drill Holes) Presented According
to
Ratera
and Volume Category
|
|
Under the more difficult reading conditions for the series with ground-meat
scatter body (series D–F), the examination technique affected the
percentage of correctly identified columns and holes in only volume categories
I and II (Table 1). In volume
category I, the percentages of correct positive findings were 12–22% for
analog images (series C), 15–39% for digital films (D), and 15–18%
for monitor readings (E). In category II, the percentages of correct positive
findings were 30–46% (series D), 40–58% (E), and 31–58% (F).
The results were balanced when the methods were compared in volume categories
III (series D, 54–72%; E, 59–73%; F, 54–70%) and IV (series
D, 86–95%; E, 85–95%; F, 78–95%), which meant that digital
film reading was moderately superior with regard to all 588 images with
experimentally simulated lesions (series D, 46–57%; E, 50–65%; F,
54–57%).
Evaluation of the 1,212 cubes without columns or holes showed that the
percentage of false-positive findings was dependent on the decision-making
level of the individual raters and not on the examination technique in the
series both without (A–C) and with (D–F) ground-meat scatter body
(Table 2). Hence, the raters'
spans were consistent for the most part—that is, 0.5–18.5% (series
A), 0.1–23.8% (B), 0.5–16.9% (C), 1.3–10.0% (D),
0.2–28.2% (E), and 0.6–5.9% (F).
View this table:
[in this window]
[in a new window]
|
TABLE 2: Percentage of False-Positive Findings (1 – Specificity) Among the
1,212 Subcubes Without Simulated Lesions (Columns and Drill Holes) by
Ratera
|
|
Across all volumes, the accuracy rate according to rater was 80–89%
(series A), 80–93% (B), and 81–94% (C)
(Table 3). In series A, the
kappa values increased from 0.6–0.7 to 0.5–0.8 (B) and
0.6–0.9 (C). As anticipated, the values were lower in volume category I,
for which the accuracy rate was 77–93% for the analog technique,
75–95% for digital films, and 81–85% for viewing workstations,
with kappa values of 0.2–0.5 (series A) and 0.3–0.6 (B and C).
View this table:
[in this window]
[in a new window]
|
TABLE 3: Accuracy Rate and Kappa Values as Measures of How Well Ratings of Films
Obtained in a Series Agreed with the Reference Standard According to
Ratera
|
|
The diagnostic accuracy of evaluations of images acquired with a
ground-meat scatter body declined compared with the exposures without a
ground-meat scatter body across all methods. In the sum of all images, the
corresponding rates were comparable in 79–85% for series D, 70–86%
for series E, and 78–85% for series F; the kappa values were
0.5–0.6 (series D) and 0.4–0.6 (E and F). The same applied to
volume category I. Accuracy was 83–89% (series D), 68–91% (E), and
86–91% (F), corresponding to kappa values of 0.1–0.2 (series D)
and 0.1–0.2 (E and F).
In general, the odds ratios were > 1; in other words, the individual
classifications of each rater basically correlated with the reference standard
(Table 4). The odds ratios were
less at the low volumes than at the high volumes. For the images acquired
without a ground-meat scatter body (series A–C), there was a trend
toward higher odds ratios than in the images acquired without a ground-meat
scatter body (D–F). The exceptions were rater 3 for all subanalyses of
series A–C, rater 1 for the sum of all volume categories of series
D–F, and raters 3 and 4 for the volume categories III and IV with a
ground-meat scatter body. No direct pattern was identifiable when the odds
ratios in the subclass analyses for digital film and monitor reading were
compared.
Discussion
In an imaging system, the variables contrast and local resolution are
mutually dependent: Lowering the number of x-ray quanta received per pixel
improves local resolution while causing the signal-to-noise ratio to worsen,
thereby leading to a lower contrast resolution and vice versa. The metric used
to describe this phenomenon is the detective quantum efficiency (DQE). The DQE
states the percentage of photons hitting the detector that are converted into
electrical signals but does not account for the process steps involved:
conversion of the electrical signal from analog to digital, image rendering,
and then image reproduction on film or on a monitor.
Compared with analog screen-film systems, digital mammography techniques
are distinguished by a higher DQE (
20–40% vs 50–80%,
respectively) and a broader breadth of exposure. The latter is essentially
achieved through the linear course of the digital gradation curve over a wide
dose range and by the ability to modify the gray-scale and contrast of the
digital imaging data set. This contrasts with a lower local resolution
(digital mammography, 7–10 lp/mm; screen-film mammography, 15–20
lp/mm). Earlier digital systems were limited by too low a local resolution,
particularly in the high-contrast range—that is, in the detection and
morphologic characterization of microcalcifications (diameter
100–1,500 µm). More recent phantom measurements
[8,
9,
11–13]
and comparative clinical studies
[1,
6,
10,
14,
15] have shown that the local
resolution of modern digital mammography systems has improved to such an
extent that digital mammography is at least diagnostically equivalent to the
conventional screen-film technique in the high-contrast range. For this
reason, the use of digital mammography for curative diagnostics and breast
cancer screening is on the rise.
By contrast, until now, relatively little attention has been paid to
comparing diagnostic methods in the low-contrast range. The mammographic
detectability of a lesion depends on its size, the noise in the object and
background, and the difference in mean optical density between the lesion and
its surroundings [16]. In
clinical practice, unlike microcalcifications, noncalcified masses, which
frequently have the same density as or a density similar to the healthy
parenchyma, are usually not identified by virtue of their asymmetric
localization or shape and contours until they are larger than 3–6 mm in
diameter. This observation particularly applies to dense breasts.
In 2004, one in vitro study indicated that the direct flat-panel detector
technique is superior to analog screen-film mammography, digital film-screen
radiography, and direct flat-panel radiography in its ability to detect
fibrous structures, although all methods identified microcalcifications and
round opacities with similar accuracy
[8]. A model consisting of 16
cubes was investigated: five cubes containing aluminum oxide granules
(diameter, 200–740 µm) devised to reproduce microcalcifications, six
cubes containing nylon threads (diameter, 0.4–1.6 mm) to imitate fibrous
structures, and four cubes with tumorlike masses (thickness, 5–14 mm).
One cube had no structure. The 16 cubes were evaluated in three different
configurations on one analog and five different digital mammography systems.
Using a structured questionnaire, three radiologists rated the 18 images that
were generated.
In the present study, conducted on a larger sample size than previous
studies, the digital technique proved superior to the analog technique,
particularly in detecting small simulated masses. Their detectability was
higher in the images acquired without a ground-meat scatter body (simulating
fatty breasts) than in the images acquired with a ground-meat scatter body
(simulating dense breasts). These findings are attributable to the fact that
the geometric structure and the differences in density within the ground-meat
layer were so similar to the test bodies (columns and holes) in the imaging
series with a ground-meat scatter body that the test bodies were often
indistinguishable from the simulated masses. The diagnostic accuracy increased
with increasing simulated tumor volume. In contrast, the percentage of
false-positive findings was governed by the raters' decision-making strategy
and not by the examination technique.
Overall, digital film and monitor viewing scored better than the analog
technique. The trend suggested that monitor viewing was diagnostically
superior to digital film reading in the series without an anthropomorphic
ground-meat scatter body and that digital film reading was superior to monitor
viewing in the series with a ground-meat scatter body. The results of other
studies primarily dealing with the high-contrast range suggest that the
density and superimposed structures of the anthropomorphic scatter body used
for monitor reproduction in the present low-contrast study were visually
emphasized to such an extent that they hampered accurate detection of the test
bodies [12,
15,
17–19].
Particularly the number of correct positive findings declined for this reason.
Additional studies are required to more profoundly analyze the reasons monitor
reading scored comparatively poorly when an anthropomorphic scatter body was
used; these studies should also establish a suitable digital imaging
processing algorithm for image reading in the low-contrast range.
Our study has a limitation that should be noted. Although the same exposure
settings (kV, mAs) and comparable surface doses were used, the detector doses
presumably deviated slightly from one another because of the difference in
focus–film distances (60 cm for analog and 65 cm for digital) and the
different analog and digital grid configurations. This difference might have
influenced image quality by affecting the signal-to-noise ratio.
Another inherent limitation of the study is that the readers could not be
blinded to what type of imaging they were reading. They knew if they were
looking at films versus images on a monitor and if they were evaluating analog
or digital films. This knowledge might have had an unavoidable effect on their
ratings, and thus is a potential bias in the study design.
In conclusion, we were able to show experimentally that direct digital
flat-panel detector mammography is comparable to or superior to analog
screen-film mammography with respect to the detection and morphologic
characterization of masses in the low-contrast range. The answers to the
question whether monitor reading should be preferred over digital film viewing
in the low-contrast range and the question of which image-generation
algorithms are most suitable for monitor-based detection of small masses with
ground meat in contrast to healthy mammary parenchyma will have to be provided
by additional studies. When our results are considered together with the
results of in vitro studies conducted on high-contrast ranges, it can
certainly be said that digital mammography in conjunction with a viewing
workstation should be the method preferred for clinical diagnostics and breast
cancer screening in the future.
Acknowledgments
We wish to thank Volker Hesselmann for the evaluation of the analog films,
the digital scans, and the monitor-displayed digital images.
References
- Bick U, Diekmann F. Digital mammography: what do we and what don't
we know? Eur Radiol 2007;17
:1931
–1942[CrossRef][Medline]
- Haus AG, Yaffe MJ. Screen-film and digital mammography.
Radiol Clin North Am 2000;38
: 871–898[CrossRef][Medline]
- Funke M, Obenauer S, Hermann K-P, Fischer U, Grabbe E. Soft copy
versus hard copy findings in digital mammography [in German].
Radiologe 2002;42
: 256–259[CrossRef][Medline]
- James JJ. The current status of digital mammography.
Clin Radiol 2004;59
: 1–10[CrossRef][Medline]
- Pisano ED, Yaffee MJ. Digital
mammography.Radiology 2005;234
: 353–362[Abstract/Free Full Text]
- Pisano ED, Gatsonis C, Hendrick E, et al.; Digital Mammographic
Imaging Screening Trial (DMIST) Investigators Group. Diagnostic performance of
digital versus film mammography for breast-cancer screening. N Engl
J Med 2005; 353:1773
–1783 [Erratum in N Engl J Med
2005; 355:1840][Abstract/Free Full Text]
- Rong XJ, Shaw CC, Johnston DA, et al. Microcalcification
detectability for four mammographic detectors: flat-panel, CCD, CR, and
screen-film. Med Phys 2002;29
:2052
–2061[CrossRef][Medline]
- Schulz-Wendtland R, Hermann K-P, Böhner C, et al. Phantom
study for the detection of simulated lesions in five different digital and one
conventional mammography system [in German]. Rofo2004; 176:1127
–1132[Medline]
- Diekmann F, Diekmann S, Bick U, et al. Comparing the visualization
of microcalcifications with direct magnification in digital full-field
mammography vs. film-screen mammography [in German].
Rofo 2002; 174:297
–300[Medline]
- Diekmann S, Bick U, von Heyden H, Diekmann F. Visualization of
microcalcifications on mammographies obtained by digital full-field
mammography in comparison to conventional film-screen mammography [in German].
Rofo 2003; 175:775
–779[Medline]
- Obenauer S, Hermann K-P, Schorn C, Funke M, Fischer U, Grabbe E.
Full-field digital mammography: a phantom study for the detectability of
simulated microcalcification [in German]. Rofo2000; 172:646
–650[Medline]
- Krug KB, Stützer H, Girnus R, et al. Image quality of digital
direct flat-panel mammography versus an analog screen-film technique using a
phantom model. AJR 2007;188
: 399–407[Abstract/Free Full Text]
- Yip WM, Pang SY, Yim WS, Kwok CS. ROC analysis of lesion
detectability on phantoms: comparison of digital spot film mammography with
conventional spot film mammography. Br J Radiol2001; 74:621
–628[Abstract/Free Full Text]
- Heddson B, Röonnow K, Olsson M, Miller D. Digital versus
screen-film mammography: a retrospective comparison in a population-based
screening program. Eur J Radiol 2007;64
: 419–425; Epub 2007Mar
26[CrossRef][Medline]
- Skaane P, Balleyguier C, Diekmann F, et al. Breast lesion detection
and classification: comparison of screen-film mammography and full-field
digital mammography with soft-copy reading—observer performance study.
Radiology 2005;237
: 37–44[Abstract/Free Full Text]
- Perry N, Broeders M, de Wolf C, et al., eds. European
guidelines for quality assurance in mammography screening, 4th
ed. Luxembourg: Office for Official Publications of the European Communities,2001
: 130–133
- Pisano ED, Cole EB, Major S, et al. Radiologists' preferences for
digital mammographic display. Radiology2000; 216:820
–830[Abstract/Free Full Text]
- Pisano ED, Cole EB, Kistner EO, et al. Interpretation of digital
mammograms: comparison of speed and accuracy of soft-copy versus printed-film
display. Radiology 2002;223
: 483–488[Abstract/Free Full Text]
- Zuley ML, Willison KM, Bonaccio E, et al. Full-field digital
mammography on LCD versus CRT monitor. AJR2006; 187:1492
–1498[Abstract/Free Full Text]
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
Top
Abstract
Introduction
