Professional Documents
Culture Documents
AJR 2003;181:10831088
0361803X/03/18141083
American Roentgen Ray Society
1083
Baker et al.
Materials and Methods
Downloaded from www.ajronline.org by 190.179.191.48 on 07/07/13 from IP address 190.179.191.48. Copyright ARRS. For personal use only; all rights reserved
1084
Results
Downloaded from www.ajronline.org by 190.179.191.48 on 07/07/13 from IP address 190.179.191.48. Copyright ARRS. For personal use only; all rights reserved
Fig. 1.59-year-old woman with invasive ductal carcinoma of right breast. Architectural distortion seen on images was successfully detected and marked by only one of
two computer-aided detection systems tested.
A, Craniocaudal mammogram shows typical appearance of architectural distortion (box): radiating lines without central density.
B, Mediolateral oblique mammogram also shows architectural distortion (box).
C, Spot compression magnification image of right breast shows lack of central density at site of distortion (arrow) more clearly than do mammograms A and B.
Fig. 2.55-year-old woman with ductal carcinoma in situ of left breast. Architectural distortion was successfully identified by interpreting radiologist but was not detected
by either computer-aided detection system.
A, Craniocaudal mammogram shows subtle architectural distortion (box) in lateral aspect of breast. Note radiating lines without central mass in dense breast parenchyma.
B, Mediolateral oblique mammogram shows more conspicuous architectural distortion (box) than craniocaudal view (A). No central density is present.
C, Specimen radiograph from wire-localized surgical excision confirms architectural distortion (arrow) centered on middle of thickened wire. No mass or associated calcifications are seen.
1085
Downloaded from www.ajronline.org by 190.179.191.48 on 07/07/13 from IP address 190.179.191.48. Copyright ARRS. For personal use only; all rights reserved
Baker et al.
of benign proliferative change, and one case of
a surgically confirmed postoperative scar.
The R2 ImageChecker system correctly
marked at least one of the two screening mammographic views as containing a possible
mass at the correct location in 22 of the 45
cases of architectural distortion (case sensitivity, 49%). The CADx SecondLook system correctly marked 15 cases of architectural
distortion (case sensitivity, 33%). We found a
trend toward better case sensitivity for the ImageChecker system for detection of both benign and malignant causes of architectural
distortion, but the difference between the ImageChecker and the SecondLook systems did
not reach the level of statistical significance
( p = 0.10).
The ImageChecker system had virtually
identical sensitivity for detecting malignant
causes of architectural distortion as it did for
detecting all cases of distortion. In malignant
cases, the ImageChecker system successfully
identified 13 (48%) of 27 cases of malignant
distortion. The rate of detection for malignant
cases for the ImageChecker system was significantly higher than that for the SecondLook
system, which identified only five (19%) of the
27 malignant lesions ( p = 0.027).
Architectural distortion without associated
findings was seen on 80 mammographic
views35 cases in which it was visible in
both craniocaudal and mediolateral views and
10 cases in which it was visible in only one of
the two views. The focus of distortion was correctly identified in 30 of the 80 views (image
sensitivity, 38%) by the ImageChecker system,
significantly better than the 17 of 80 views correctly marked by the SecondLook system (image sensitivity, 21%) ( p = 0.01).
The ImageChecker system was also significantly more successful at detecting malignant
foci of distortion on each image (view) in which
it was deemed actionable by the panel of radiologists. This system successfully detected the
malignancy in 16 of the 51 images in which the
distortion represented breast cancer (image sensitivity, 31%) compared with five of the 51 images (image sensitivity, 10%) for the
SecondLook system (p = 0.01).
The two CAD systems marked different
subsets of lesions (Fig. 1) as possible malignancies. Of the 45 cases, nine cases (20%)
were successfully identified on at least one
view by both CAD systems. The ImageChecker system identified 13 cases (29%)
of architectural distortion that were not identified by the SecondLook system. In comparison, the SecondLook system identified five
1086
cases (11%) that were not identified by the ImageChecker system. Eighteen (40%) of the 45
cases of architectural distortion were not detected by either CAD system (Fig. 2).
Each CAD system also displayed a number
of false-positive marks. On average, the ImageChecker system displayed 0.70 false-positive marks per image. This rate was
statistically less than the 1.27 false-positive
marks per image displayed by the SecondLook
system ( p < 0.0001).
Discussion
Downloaded from www.ajronline.org by 190.179.191.48 on 07/07/13 from IP address 190.179.191.48. Copyright ARRS. For personal use only; all rights reserved
[4] had lower sensitivity for detecting architectural distortion than for detecting other lesions
may be explained in part by the researchers
statement that [w]hen no central density is
found, the radiating lines must be more pronounced to be marked. This approach limits the
number of false-positive marks for each case because normal overlapping tissue (e.g., Coopers
ligaments) can mimic the radiating lines of architectural distortion, deceiving both a radiologist and a CAD algorithm. Each of the other
commercial systems approved by the United
States Food and Drug Administration, including
SecondLook and the more recently approved
MammoReader (iCAD, Nashua, NH), must
make similar trade-offs between sensitivity and
false-positive marks.
In our study, the more sensitive of the two
systems (ImageChecker) also had a significantly lower number of false-positive marks
per image. Systems that generate many falsepositive marks may result in a true-positive
mark being ignored by a radiologist overwhelmed by distracting prompts. Therefore,
the false-positive rate of a CAD system must
be considered along with its sensitivity.
The purpose of our investigation was to test
the sensitivity of increasingly available CAD
systems to determine whether such systems
are as successful in detecting worrisome foci
of architectural distortion as they are in detecting more common breast masses and calcifications. Although one of the CAD systems was
significantly more sensitive than the other for
detecting architectural distortion, a study by
Nelson et al. [36] found that the three commercially available mammography CAD systemsthe R2 ImageChecker, the CADx
SecondLook, and the iCAD MammoReader
all performed with nearly identical sensitivity
in a study of 128 malignant masses and clusters of calcifications. We found that both systems had substantially lower rates for
identifying architectural distortion than the
previously reported rates of those systems for
detecting more common masses and calcifications. Clearly, both systems need to be improved, given that one half to two thirds of the
cases of architectural distortion were not identified by the two most widely available commercial CAD systems.
Because of the similarity between architectural distortion and overlapping fibroglandular
tissue, improvement in detection may prove
difficult without a concomitantand perhaps
unacceptableincrease in the number of
false-positive marks per image. Nevertheless,
now that CAD systems can successfully iden-
1087
Downloaded from www.ajronline.org by 190.179.191.48 on 07/07/13 from IP address 190.179.191.48. Copyright ARRS. For personal use only; all rights reserved
Baker et al.
12. Burrell H, Evans A, Wilson A, Pinder S. Falsenegative breast screening assessment: what lessons can we learn? Clin Radiol 2001;56:385388
13. Bird RE, Wallace TW, Yankaskas BC. Analysis of
cancers missed at screening mammography. Radiology 1992;184:613617
14. Baker JA, Kornguth PJ, Floyd CE Jr. Breast imaging reporting and data system standardized
mammography lexicon: observer variability in lesion description. AJR 1996;166:773778
15. Vyborny C, Doi T, OShaughnessy K, Romsdahl
H, Schneider A, Stein A. Breast cancer: importance of spiculation in computer-aided detection.
Radiology 2000;215:703707
16. Kallergi M, Carney GM, Gaviria J. Evaluating the
performance of detection algorithms in digital
mammography. Med Phys 1999;26:267275
17. Winsberg F, Elkin M, Macy J, Bordaz V, Weymouth
W. Detection of radiographic abnormalities in mammograms by means of optical scanning and computer analysis. Radiology 1967;89:211215
18. American College of Radiology. Breast imaging reporting and data system (BI-RADS), 3rd ed. Reston,
VA: American College of Radiology, 1998
19. Gajdos C, Tartter P, Bleiweiss I, et al. Mammographic appearance of nonpalpable breast cancer
reflects pathologic characteristics. Ann Surg
2002;235:246251
20. Ciatto S, Morrone D, Catarzi S, et al. Radial scars
of the breast: review of 38 consecutive mammographic diagnoses. Radiology 1993;187:757760
21. Mitnick J, Vazquez M, Harris M, Roses D. Differ-
1088
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
Downloaded from www.ajronline.org by 190.179.191.48 on 07/07/13 from IP address 190.179.191.48. Copyright ARRS. For personal use only; all rights reserved