EDUCATION EXHIBIT 881

RadioGraphics
Missed Breast Carcino-
ma: Pitfalls and Pearls1
Aneesa S. Majid, MD ● Ellen Shaw de Paredes, MD ● Richard D.
Doherty, MD ● Neil R. Sharma ● Xavier Salvador

Mammography is the standard of reference for the detection of breast

carcinoma, yet 10%–30% of breast cancers may be missed at mam-
mography. Possible causes for missed breast cancers include dense pa-
renchyma obscuring a lesion, poor positioning or technique, percep-
tion error, incorrect interpretation of a suspect finding, subtle features
of malignancy, and slow growth of a lesion. Recent studies have em-
phasized the use of alternative imaging modalities to detect and diag-
nose breast carcinoma, including ultrasonography (US), magnetic
resonance imaging, and nuclear medicine studies. However, the radi-
ologist can take a number of steps that will significantly enhance the
accuracy of image interpretation at mammography and decrease the
false-negative rate. These steps include performing diagnostic as well
as screening mammography, reviewing clinical data and using US to
help assess a palpable or mammographically detected mass, strictly
adhering to positioning and technical requirements, being alert to
subtle features of breast cancers, comparing recent images with earlier
mammograms to look for subtle increases in lesion size, looking for
additional lesions when one abnormality is seen, and judging a lesion
by its most malignant features.
©
RSNA, 2003

Abbreviation: CAD ⫽ computer-aided detection

Index terms: Breast neoplasms, 00.32 ● Breast neoplasms, radiography, 00.11 ● Diagnostic radiology, observer performance

RadioGraphics 2003; 23:881– 895 ● Published online 10.1148/rg.234025083

1From the Department of Radiology, Medical College of Virginia, Virginia Commonwealth University, 401 N 12th St, Richmond, VA 23298. Recipi-
ent of a Certificate of Merit award and an Excellence in Design award for an education exhibit at the 2001 RSNA scientific assembly. Received April
22, 2002; revision requested May 23; final revision received April 25, 2003; accepted April 25. Address correspondence to E.S.d.P. (e-mail:
[email protected]).
©
RSNA, 2003
 882 July-August 2003 RG f Volume 23 ● Number 4

RadioGraphics

Figure 1. Invasive ductal carcinoma in a 36-year-old woman with dense breasts and a pal-
pable mass. (a) Left mediolateral oblique mammogram demonstrates no finding that corre-
sponds to a palpable mass (arrow). (b) US image obtained in the area of the palpable abnor-
mality reveals a heterogeneous, hypoechoic mass with irregular margins. Although there is
no acoustic shadowing and the mass is wider than it is tall, the hypoechogenicity and irregu-
lar margins are suspect for malignancy. Pathologic analysis demonstrated invasive ductal car-
cinoma.

Introduction nuclear medicine studies. However, high-quality

Mammography is the standard of reference for mammography performed with meticulous atten-
the early detection of breast cancer. Screening tion to detail and positioning can significantly
mammography is performed to detect an abnor- enhance the accuracy of image interpretation.
mality, whereas diagnostic mammography is used Breast cancers may be missed because of dense
to further evaluate the abnormality or a clinical parenchyma that obscures a lesion (4), poor posi-
problem. tioning or technique, lesion location outside the
The purpose of screening mammography is field of view, lack of perception of an abnormality
simply to detect a potential cancer; therefore, the that is present, incorrect interpretation of a sus-
radiologist should not try to make a diagnosis on pect finding, subtle features of malignancy, or a
the basis of screening findings alone. Additional slowly changing malignancy. Breast cancers are
views are important in further assessing an identi- easily missed when they appear as focal areas of
fied abnormality and suggesting appropriate pa- asymmetry or distortion (eg, invasive lobular car-
tient treatment. According to data from the cinoma) or when their appearance suggests a be-
Breast Cancer Detection Demonstration Project, nign cause (eg, medullary and mucinous [colloid]
the false-negative rate of mammography is ap- invasive ductal carcinomas, which usually mani-
proximately 8%–10% (1). After evaluating retro- fest as mostly circumscribed masses) (5). Bird et
spective versus blinded interpretations of mam- al (6) found that 77 of 320 cancers (24%) in a
mograms, others have concluded that the rate of screening population were missed, primarily due
missed breast cancers is as high as 35% (2). In a to dense breasts and a developing density that was
series of 150 mammograms (27 cancers) read by not identified by the radiologist. Goergen et al (7)
10 radiologists, immediate work-up of the true found that cancers missed at screening mammog-
cancers was recommended in 74%–96% of cases raphy were significantly lower in density and were
(3). Recent studies have emphasized the use of more often seen on only one of two views than
alternative imaging modalities to detect and diag- were detected cancers. In a review of interval can-
nose breast carcinoma, including ultrasonography cers in the Malmo Screening Trial, Ikeda et al (8)
(US), magnetic resonance (MR) imaging, and found that 10 of 94 cases were missed due to ob-
server error and 21 of 94 showed subtle signs of
malignancy.
 RG f Volume 23 ● Number 4 Majid et al 883

RadioGraphics

Figure 2. Invasive lobular carcinoma

in a 40-year-old woman with dense
breasts. (a) Right mediolateral oblique
screening mammogram shows a small,
oval obscured mass superiorly (arrow)
that was not seen on the craniocaudal
view. (b) US image of the mass demon-
strates a simple cyst. (c) US image re-
veals an incidentally detected irregular
mass with acoustic shadowing in the
lower outer quadrant. Pathologic analy-
sis demonstrated invasive lobular carci-
noma.

of normal parenchymal elements. Unless it is

documented as a postsurgical scar, an area of ar-
chitectural distortion must be further evaluated
with additional views (eg, spot compression, mag-
In this article, we discuss and illustrate the nification, off-angle). US may also be helpful in
aforementioned pitfalls that can lead to missed determining the presence of a solid mass that cor-
breast cancers and provide guidelines to help re- responds to an area of distortion.
duce the false-negative rate of mammography. Any patient with dense breast parenchyma, a
palpable mass, and negative mammographic find-
Causes of Missed Breast Cancers ings should undergo US for further evaluation of
the mass (Fig 1). US is very important in the
Dense Parenchyma evaluation of mammographic abnormalities, be-
Breast parenchyma that is inherently dense com- ing useful in characterizing palpable masses in
promises the ability to detect a mass, especially a dense tissue and circumscribed isodense masses
noncalcified, nondistorting lesion. The radiologist (Fig 2). US can be especially helpful in the evalu-
must be particularly attentive in searching for ar- ation of asymmetric densities seen at mammogra-
eas of architectural distortion or faint microcalci- phy because it can help identify the density as ei-
fications. Magnification views are used to evaluate ther breast tissue or a true mass. Soo et al (9) and
the morphologic features of suspect or faint mi- Skaane (10) found the negative predictive value
crocalcifications. Because architectural distortion of US with mammography for a palpable lesion to
may be the only sign of malignancy in a dense
breast, the tissue must be intensely evaluated for
any areas of tethering or disruption of orientation
 884 July-August 2003 RG f Volume 23 ● Number 4

RadioGraphics

Figure 3. Proper positioning. (a) Left mediolateral oblique (left) and craniocaudal (right) mam-
mograms obtained with improper positioning demonstrate poor visualization of the posterior tis-
sue. The margin of a mass is barely perceptible at the edge of the mediolateral oblique image (ar-
row). (b) On a left mediolateral oblique mammogram obtained with improved positioning, a can-
cer is seen near the chest wall. An exaggerated craniocaudal view may also help demonstrate such a
mass.

be 99.8% and 100%, respectively. Moy et al (11) rant, which demonstrates the greatest proportion
found the negative predictive value of US with of breast cancers, is necessary. However, the tech-
mammography for a palpable mass to be 97.4%. nologist must use the craniocaudal view as a
However, a palpable mass that appears solid at complement to the mediolateral oblique view to
US warrants further evaluation with biopsy. visualize the medial tissue as well.
Creative positioning may be necessary to in-
Poor Positioning clude areas of palpable abnormalities on the im-
Proper positioning and image contrast are abso- ages. Radiopaque markers should be placed on
lutely necessary in all aspects of radiology, but palpable areas, with repositioning of the marker
especially in mammography. The technologist between projections as needed to keep the marker
must adhere to the positioning standards to maxi- superimposed on the palpable finding. In addi-
mize the amount of tissue included on the image tion, a spot compression view obtained over a
(12). Findings on the mediolateral oblique view palpable mass with the skin in tangent can reveal
that indicate proper positioning include visualiza- an underlying mass and demonstrate overlying
tion of the pectoralis muscle to the level of the skin thickening or retraction. Creative positioning
nipple, a convex appearance of the pectoralis ma- may also be helpful in patients who are tense, who
jor muscle, complete visualization of posterior have suffered a stroke, or who have shoulder
breast tissue, breast tissue that is well compressed problems or other debilitating factors that limit
and positioned in an up-and-out orientation, and visualization of the posterior breast on standard
an open inframammary fold (Fig 3). At cranio- mediolateral oblique views.
caudal imaging, the technologist should verify Off-angle or step oblique views are very helpful
that the breast is pulled straight forward and not in the evaluation of densities or abnormalities
exaggerated laterally, and that the breast tissue is seen in only one projection (13). Densities seen
well compressed. The difference between the pos- on the craniocaudal view alone may be further
terior nipple line measurement on the mediolat- characterized and localized with use of spot com-
eral oblique and craniocaudal views should not pression and rolled craniocaudal views. If a lesion
exceed 1 cm. Emphasis on the upper outer quad- rolls medially when the top of the breast is rolled
medially, it is located superiorly; if it rolls later-
ally, it is located inferiorly. The technologist
should label the image with the orientation in
 RG f Volume 23 ● Number 4 Majid et al 885

RadioGraphics

Figure 4. Creative positioning for lesion

detection. (a) Bilateral mediolateral oblique
mammograms show dense parenchyma with
well-defined masses (arrows) and a focal ir-
regular density superoposteriorly on the right
side (arrowheads). The well-defined masses
proved to be cysts at US. (b) On a right lat-
eromedial mammogram, the irregular density
(arrow) has moved upward, a finding that in-
dicates a medial location. At lateromedial
mammography, the medial aspect of the
breast is closer to the film and can therefore
be better evaluated. (c) Spot magnification
mammogram (right cleavage view) demon-
strates a spiculated mass. Pathologic analysis
revealed invasive ductal carcinoma.

which the top of the breast was rolled (eg, cranio- be helpful in verifying the location of a mass that
caudal RL ⫽ “craniocaudal rolled laterally”). If a is clearly seen on only one view.
density is seen only on the mediolateral oblique
view, a mediolateral view is required to locate and Poor Technique
further evaluate the lesion (Fig 4). In such a case, The technologist must optimize image contrast to
a medial lesion will move superiorly on the lateral avoid obtaining over- or underpenetrated images.
view, whereas a lateral lesion will move inferiorly. Proper positioning of the photocell is necessary
This concept of triangulation is extremely impor- to achieve correct optical density on the image.
tant in identifying the actual position of a lesion. Careful attention to daily processor quality con-
Off-angle or step oblique views, like standard trol is also necessary to optimize contrast. The
views, are most helpful when the lesion is super- technologist should always review the images
imposed over fat and not dense tissue. Exagger-
ated craniocaudal views may be helpful in demon-
strating a posteriorly located lesion that is seen on
the mediolateral oblique view only. US may also
 886 July-August 2003 RG f Volume 23 ● Number 4

RadioGraphics

Figure 5. Proper imaging technique. (a) Right

craniocaudal screening mammogram obtained in a 65-
year-old woman demonstrates underpenetration.
(b) Right mediolateral oblique mammogram reveals an
irregular density (arrow) that was obscured on the
craniocaudal view. (c) Right craniocaudal spot magni-
fication mammogram demonstrates an irregular mass
with microcalcifications. At pathologic analysis, the
mass proved to be invasive ductal carcinoma.

under proper mammographic viewing conditions curs when the lesion is included in the field of
to assess the adequacy of imaging technique (Fig view and is evident but is not recognized by the
5). Image blur is problematic, particularly in the radiologist. The lesion may or may not have
assessment of microcalcifications. Rosen et al subtle features of malignancy that cause it to be
(14) found that in 62% of cancers that manifested less visible. Small nonspiculated masses, areas of
as microcalcifications and were incorrectly fol- architectural distortion and asymmetry, and small
lowed up with imaging rather than biopsy, image clusters of amorphous or faint microcalcifications
blur on magnification views compromised image may all be difficult to perceive.
quality. To avoid perception error, images should be
reviewed as mirror images, with mediolateral
Lack of Perception oblique images placed together and craniocaudal
Two major causes of missed breast cancers are images placed together (Figs 6, 7). The radiolo-
related to radiologist error. The first of these gist should compare like areas on the side-by-side
causes is lack of perception. Perception error oc- images to identify any focal asymmetric density or
low-density mass. Identification of a focal density
should prompt a search for this density on the
corresponding view in the same arc from the
 RG f Volume 23 ● Number 4 Majid et al 887

RadioGraphics

Figure 6. Drawings illus-

trate useful search patterns in
mirror image interpretation.
CC ⫽ craniocaudal, MLO ⫽
mediolateral oblique.

Figure 7. Mirror image in-

terpretation. (a) Bilateral me-
diolateral oblique mammo-
grams reveal an irregular mass
posteriorly on the left side with
a highly suspect appearance
(arrow). In addition, a subtle
distortion is noted more inferi-
orly (arrowhead), a finding
that becomes more evident
with mirror image interpreta-
tion. (b, c) On left craniocau-
dal spot compression mam-
mograms, the posterior (b)
and anterior (c) lesions dem-
onstrate a spiculated appear-
ance (arrowhead in c). Patho-
logic analysis demonstrated
multicentric invasive ductal
carcinoma.
 888 July-August 2003 RG f Volume 23 ● Number 4

RadioGraphics

Figure 8. Diagrams illustrate nipple-to-

lesion arc measurements used to determine
lesion depth. a ⫽ distance from nipple to
anterior lesion, b ⫽ distance from nipple to
posterior lesion, CC ⫽ craniocaudal,
MLO ⫽ mediolateral oblique.

Figure 9. Multicentric breast can-

cer in a 63-year-old woman. Right
mediolateral oblique (a) and right
exaggerated craniocaudal lateral (b)
screening mammograms show a
prominent area of architectural dis-
tortion at the 10 o’clock position
(solid arrow). Note also the two
small, indistinct masses in the axil-
lary tail (arrowheads) and the lin-
early arranged microcalcifications at
the 7 o’clock position (open arrow).
An indistinct high-density node is
also seen in the axilla and proved to
be malignant at surgery. Pathologic
analysis demonstrated multicentric
invasive ductal carcinoma and duc-
tal carcinoma in situ.
 RG f Volume 23 ● Number 4 Majid et al 889

RadioGraphics

Figure 10. Satisfaction of search. Right mediolateral oblique (a) and craniocaudal (b) mammo-
grams demonstrate subtle architectural distortion (arrow) behind an obvious calcified fibroad-
enoma. The first interpreting radiologist noted the fibroadenoma but missed the distortion, which
proved to be invasive ductal carcinoma.

nipple (Fig 8). Additional views may be needed to of patients (15) and may actually be seen in 9%–
verify the presence of a true lesion. 10% of patients at MR imaging (16). Satisfaction
Failure to diagnose multifocal and multicentric of search can also occur in cases of an obvious
breast cancers can directly affect patient treat- benign lesion with a subtle cancer. The radiolo-
ment. Multifocal breast cancer is defined as two gist must not be satisfied with finding just one
or more cancers in the same quadrant, whereas lesion, but must search carefully for others,
multicentric breast cancer is defined as two or whether benign or malignant.
more cancers in different quadrants (Fig 9). In Another special circumstance that can present
multicentric disease, breast conservation therapy a perception problem involves a patient with a
is contraindicated. These disease entities may not palpable node in the axilla that is evaluated with
be perceived owing to “satisfaction of search,” in biopsy and represents metastatic adenocarci-
which observation of an obvious finding misleads noma, likely of breast origin. The primary breast
the radiologist into not looking carefully for other cancer may be occult and either not observed or
lesions (Fig 10). Careful attention must also be very subtle at mammography. Careful attention
paid to the contralateral breast after observation
of a suspect lesion because contralateral synchro-
nous cancers have been reported in 0.19%–2.0%
 890 July-August 2003 RG f Volume 23 ● Number 4
RadioGraphics

Figure 11. Occult cancer with metastases in a 36-year-old woman. (a) Right mediolateral oblique mammo-
gram that was thought to be otherwise negative reveals an enlarged axillary node (arrow) that was palpable.
(b) On a right mediolateral oblique mammogram obtained 3 months later while the patient was being evalu-
ated for adenopathy, the previously occult cancer in the 11 o’clock position (arrowhead) became visible. Patho-
logic analysis demonstrated invasive ductal carcinoma with metastasis to the axilla.

to mirror image abnormalities or focal asymmet- Incorrect Interpretation

ric densities is important in identifying the pri- The second major cause of missed breast cancers
mary lesion (Fig 11). MR imaging has been use- that is related to radiologist error is incorrect in-
ful in identifying the primary carcinoma when a terpretation of a lesion, which occurs when an
metastatic node is found in the axilla and mam- abnormality with suspect features is observed but
mographic findings are negative (17). is misinterpreted as being definitely or at least
US may be helpful (like MR imaging and, oc- probably benign. Several factors may lead to mis-
casionally, scintimammography) in the search interpretation, such as lack of experience, fatigue,
for occult breast malignancy in special circum- or inattention. Misinterpretation may also occur
stances, such as those involving patients with if the radiologist fails to obtain all the views
multicentric cancer or with metastases to the ax- needed to assess the characteristics of a lesion or
illa and no obvious breast lesion (18,19). Berg if the lesion is slow growing and prior images are
and Gilbreath (20) found preoperative whole not used for comparison. The radiologist may
breast US to be complementary to mammogra- erroneously judge the abnormality by its most
phy in patients with known breast cancer and in benign features and miss important malignant
whom breast conservation was planned. MR im- features that necessitate biopsy (Fig 12).
aging is becoming increasingly important in dem- The margins of masses are best evaluated with
onstrating the local extent of disease in patients spot compression imaging. A mass that appears
with breast cancer. relatively smooth may be indistinct or microlobu-
lated on spot compression images. Therefore,
margins should not be characterized on the basis
 RG f Volume 23 ● Number 4 Majid et al 891

RadioGraphics

Figure 12. Apparent lesion thinning at spot compression mammography. (a, b) Right craniocaudal (a) and
mediolateral oblique (b) mammograms demonstrate focal architectural distortion (arrow in a) that may corre-
spond to a superiorly located lesion (arrowhead in b). (c, d) Mediolateral oblique (c) and craniocaudal (d)
spot compression mammograms show a persistent but less prominent area of distortion. At 6-month follow-up
mammography, the area appeared more prominent, and biopsy was performed. Pathologic analysis demon-
strated invasive ductal carcinoma. Rolled craniocaudal views were also obtained and helped confirm the persis-
tence of the lesion.
 892 July-August 2003 RG f Volume 23 ● Number 4

RadioGraphics of a screening study alone. Any areas of microcal-

cifications should be evaluated with magnification
views to accurately define their morphologic fea-
tures as well as their number and distribution.
Characterization of a lesion that is identified at
screening mammography should be based on di-
agnostic mammographic findings and not on
screening findings alone.

Subtle Signs of Malignancy

The cancers that are the most challenging to diag-
nose and that most often lead to interpretation
errors are those with subtle or indistinct features
of malignancy. These features include areas of
architectural distortion, small groups of amor-
phous or punctate microcalcifications, focal
asymmetric densities, dilated ducts, and relatively
well circumscribed masses. In a study of nonpal-
pable cancers, Sickles (21) found that only 39%
manifested with classic signs, including spiculated
masses and linear microcalcifications.
Although well-circumscribed cancers are rela-
tively uncommon, they do exist (22). Medullary,
colloid (mucinous), and papillary carcinoma
commonly manifest as well-circumscribed masses
Figure 13. Circumscribed cancer in a
(Fig 13). Invasive ductal carcinoma not otherwise
63-year-old woman. Right exaggerated
specified is usually not circumscribed; however, craniocaudal lateral mammogram dem-
because it occurs frequently, it accounts for the onstrates a nonpalpable mass in the ax-
majority of circumscribed cancers. Spot compres- illary tail. The mass is lobulated and
sion magnification of a seemingly circumscribed circumscribed and has high density.
mass that proves to be a cancer will often demon- Spot compression mammography
strate some area of indistinctness or microlobula- would help verify the characteristics of
the margins. Pathologic analysis dem-
tion of the margin.
onstrated mucinous carcinoma.
US is helpful in predicting the likelihood of
malignancy in a circumscribed mass. Simple cysts
seen at US constitute a benign finding. Solid le- review of interval cancers, Ikeda et al (8) found
sions that are smooth, elliptic, and wider than that 21 of 94 cases (22%) showed subtle signs of
they are tall are probably benign. However, malignancy, mostly asymmetric densities. Other
masses that have irregular or angulated margins, worrisome features associated with focal asym-
are markedly hypoechogenic, and are taller than metric densities include interval enlargement, a
they are wide are probably malignant (23). A new asymmetric density, a nonhormonal finding
nonpalpable circumscribed mass at mammogra- at mammography, and a palpable mass. Clinical
phy that demonstrates what are likely benign solid history is important in evaluating focal areas of
features at US may be reevaluated at an early in- asymmetry. In the absence of tumor or infection,
terval (24). If, however, the mass is seen at US as focal developing densities should prompt further
a solid lesion with worrisome features such as a assessment and, usually, biopsy. Rosen et al (14)
“taller-than-wide” shape or irregular margins found that 10 of 12 malignant areas of asymmetry
(23), biopsy is indicated. Any increase in the (83%) were new, yet were incorrectly followed up
size of a circumscribed, noncystic mass should by the radiologist. Hormonal changes are typi-
prompt further evaluation with biopsy. cally diffuse and bilateral, although a focal de-
Asymmetric densities are frequently seen at veloping density can result from hormone re-
mammography. These findings in isolation have a placement therapy. A developing density that is
low positive predictive value for malignancy; how- thought to be hormonally related calls for discon-
ever, when they are associated with microcalcifi- tinuation of therapy for 3– 4 weeks, followed by
cations or architectural distortion, the risk of ma- repeat mammography.
lignancy is increased (Fig 14). In a retrospective Invasive lobular carcinoma accounts for ap-
proximately 8%–10% of breast cancers and is eas-
ily missed because common manifestations in-
 RG f Volume 23 ● Number 4 Majid et al 893

RadioGraphics

Figure 14. Asymmetric density. Bilateral mediolateral oblique (a) and craniocaudal (b) mammograms demon-
strate a new focal asymmetric area in the left axillary tail (arrow), a finding that becomes more evident with mirror
image interpretation. Biopsy revealed infiltrating lobular carcinoma.

Figure 15. Slow-growing cancer.

(a) Right mediolateral collimated
mammogram shows focal architec-
tural distortion superiorly (arrow).
The area was not noted on sub-
sequent images because it had
changed imperceptibly. (b) Right
mediolateral collimated mammo-
gram obtained 8 years later demon-
strates interval growth of the lesion.
Biopsy was performed, and patho-
logic analysis demonstrated tubular
carcinoma.

clude a focal asymmetric density, an area of archi- Slow-growing Cancers

tectural distortion, and negative mammographic The doubling time for breast cancers has been
findings (25). US may demonstrate prominent reported to range from 44 to 1,869 days (27).
focal shadowing. However, malignant calcifications have been re-
The work-up of a focal asymmetric density ported to be stable at mammography for as long
includes a clinical examination; additional mam- as 63 months (28). Low-grade malignancies may
mography (spot compression and off-angle views) not undergo obvious change between annual in-
and US may also be helpful. However, negative terval screenings. Therefore, a slowly changing
US findings at the site of a suspect asymmetric cancer may go undetected if the radiologist fails
density should not preclude biopsy. Dilated ducts to compare findings with those on older images
are infrequently associated with malignancy. Pat- (Fig 15). A lesion with features that strongly
terns of ductal dilatation that suggest malignancy
include a unilateral solitary dilated duct (21) and
dilated ducts associated with microcalcifications
or in a nonsubareolar location (26).
 894 July-August 2003 RG f Volume 23 ● Number 4

RadioGraphics suggest malignancy but that has been stable for radiologist should take the following steps when
1–2 years still requires biopsy because it may rep- interpreting mammographic findings:
resent a slowly changing cancer. In particular, 1. Do not rely on screening views alone to di-
caution should be used in evaluating stable agnose a detected abnormality; complete the
masses or lesions with suspect morphologic fea- evaluation with diagnostic mammography.
tures that decrease in size in patients who are re- 2. Review clinical data and use US to help as-
ceiving tamoxifen. Tamoxifen is used to treat sess a palpable or mammographically detected
breast cancers and to prevent the development of mass.
breast cancer in high-risk women, but it can also 3. Be strict about positioning and technical
be used to check the growth of occult malignan- requirements to optimize image quality.
cies. 4. Be alert to subtle features of breast cancers.
5. Compare current images with multiple
Role of Double Reading prior studies to look for subtle increases in lesion
Double reading of mammograms has been shown size.
to increase the detection rate for breast cancer by 6. Look for other lesions when one abnormal-
up to 15% (29,30). Computer-aided detection ity is seen.
(CAD) represents a relatively new technology 7. Judge a lesion by its most malignant fea-
that has been implemented in some mammogra- tures.
phy facilities for double reading. Clinical studies
have shown that CAD increases the sensitivity of Acknowledgment: The authors wish to thank Louise
Logan for manuscript preparation.
breast cancer detection by radiologists by up to
20% (31,32). The sensitivity of the CAD systems
is greater for detecting calcifications than for de-
References
1. Baker LH. Breast Cancer Detection Demonstra-
tecting masses (33). In a study of 115 cancers tion Project: five-year summary report. CA Cancer
retrospectively judged to merit recall on the J Clin 1982; 32:194 –225.
screening mammogram prior to the mammogram 2. Harvey JA, Fajardo LL, Innis CA. Preview mam-
on which they were diagnosed, 77% of lesions mograms on patients with impalpable breast carci-
were identified with CAD (34). In all, 86% of 35 nomas: retrospective vs blind interpretation. AJR
Am J Roentgenol 1993; 161:1167–1172.
missed areas of calcifications and 73% of 80 3. Elmore JG, Wells CK, Lee CH, et al. Variability in
missed malignant masses were detected with radiologists’ interpretations of mammograms.
CAD. We may continue to see increasing use of N Engl J Med 1994; 331:1493–1499.
both CAD and a second radiologist for double 4. Patel MR, Whitman GJ. Negative mammograms
reading of screening mammograms. in symptomatic patients with breast cancer. Acad
Radiol 1998; 5:26 –33.
5. Burrell HC, Sibbering DM, Wilson AR, et al.
Conclusions Screening interval breast cancers: mammographic
Although mammography is the standard of refer- features and prognosis factors. Radiology 1996;
ence for the detection of early breast cancer, as 199:811– 817.
many as 30% of breast cancers may be missed. 6. Bird RE, Wallace TW, Yankaskas BC. Analysis of
cancers missed at screening mammography. Radi-
To reduce the possibility of missing a cancer, the ology 1992; 184:613– 617.
7. Goergen SK, Evans J, Colen GPB, Macmillan JH.
Characteristics of breast carcinomas missed by
radiologist. Radiology 1997; 204:131–135.
 RG f Volume 23 ● Number 4 Majid et al 895

RadioGraphics 8. Ikeda DM, Andersson I, Wattsgard C, et al. Inter- 22. Swann CA, Kopans DB, Koerner FC, McCarthy
val carcinomas in the Malmo Mammographic KA, White G, Hall DA. The halo sign and malig-
Screening Trial: radiologic appearance and prog- nant breast lesions. AJR Am J Roentgenol 1987;
nostic considerations. AJR Am J Roentgenol 1992; 149:1145–1147.
159:287–294. 23. Stavros AT, Thickman D, Rapp CL, et al. Solid
9. Soo MS, Rosen EL, Baker JA, et al. Negative pre- breast nodules: use of sonography to distinguish
dictive value of sonography with mammography in between benign and malignant lesions. Radiology
patients with palpable breast lesions. AJR Am J 1995; 196:123–124.
Roentgenol 2001; 177:1167–1170. 24. Sickles EA. Nonpalpable circumscribed noncalci-
10. Skaane P. Ultrasonography as adjunct to mam- fied solid breast masses: likelihood of malignancy
mography in the evaluation of breast tumors. Acta based on lesion size and age of patient. Radiology
Radiol 1999; 420(suppl):1– 47. 1994; 192:439 – 442.
11. Moy L, Slanetz P, Moore R, et al. Specificity of 25. Mendelson EB, Harris KM, Doshi N, Tobon H.
mammography and ultrasound in the evaluation of Infiltrating lobular carcinoma: mammographic
a palpable abnormality: retrospective review. Radi- patterns with pathologic correlation. AJR Am J
ology 2002; 225:176 –181. Roentgenol 1989; 153:265–271.
12. Hendrick RE, Bassett L, Botsco MA, et al. Mam- 26. Huynh PT, Parellada JA, Shaw de Paredes E, et
mography quality control manual. Reston, Va: al. Dilated duct pattern at mammography. Radiol-
American College of Radiology, 1999. ogy 1997; 204:137–141.
13. Pearson K, Sickles E, Frakel S. Efficacy of step- 27. Fournier DV, Weber E, Hoeffken W, et al.
oblique mammography for confirmation and local- Growth rate of 147 mammary carcinomas. Cancer
ization of densities seen on only one standard 1980; 45:2198 –2207.
mammographic view. AJR Am J Roentgenol 2000; 28. Lev-Toaff AS, Feig SA, Saitas VL, et al. Stability
174:745–752. of malignant breast microcalcifications. Radiology
14. Rosen EL, Baker JA, Soo MS. Malignant lesions 1994; 198:153–156.
initially subjected to short-term mammographic 29. Anderson ED, Muir BB, Walsh JS, et al. The effi-
follow-up. Radiology 2002; 223:221–228. cacy of double reading mammograms in breast
15. Kinne DW. Management of the contralateral screening. Clin Radiol 1994; 49:248 –251.
breast. In: Harris JR, Hellman S, Henderson P, et 30. Thurjell EL, Lernevall KA, Taube AAS. Benefit of
al, eds. Breast diseases. Philadelphia, Pa: Lippin- independent double reading in a population-based
cott, 1987; 620 – 621. mammography screening program. Radiology
16. Liberman L, Morris E, Kim C, et al. MR imaging 1994; 191:241–244.
findings in the contralateral breast in women with 31. Burhenne LJ, Wood SA, D’Orsi CJ, et al. Poten-
recently diagnosed breast cancer. AJR Am J tial contribution of computer-aided detection to
Roentgenol 2003; 180:333–341. the sensitivity of screening mammography. Radiol-
17. Orel S, Weinstein SP, Schnall MD, et al. Breast ogy 2001; 56:150 –154.
MR imaging in patients with axillary node metas- 32. Brem RF, Schoonjans JM. Radiologists detection
tases and unknown primary malignancy. Radiol- of microcalcifications with and without computer-
ogy 1999; 212:543–549. aided detection: a comparative study. Clin Radiol
18. Khalkhali I, Vargas HI. The role of nuclear medi- 2001; 56:150 –154.
cine in breast cancer detection: functional breast 33. Lechner M, Nelson M, Elvecrog E. Comparison of
imaging. Radiol Clin North Am 2001; 39:1053– two commercially available computer-aided detec-
1068. tion (CAD) systems. Appl Radiol 2002; 31:31–35.
19. Orel SG. MR imaging of the breast. Radiol Clin 34. Birdwell RL, Ikeda DM, O’Shaugnessy KF, Sick-
North Am 2000; 38:899 –913. les EA. Mammographic characteristics of 115
20. Berg WA, Gilbreath PL. Multicentric and multifo- missed cancers later detected with screening mam-
cal cancer: whole breast ultrasound in preoperative mography and the potential utility of computer-
evaluation. Radiology 2000; 214:59 – 66. aided detection. Radiology 2001; 219:192–202.
21. Sickles EA. Mammographic features of 300 con-
secutive nonpalpable breast cancers. AJR Am J
Roentgenol 1986; 146:661– 663.