| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, The George Washington University, 2150 Pennsylvania
Ave. NW, Washington, DC 21117.
2 i-CAD, Inc., Beavercreek, OH.
Received March 22, 2004;
accepted after revision July 14, 2004.
Address correspondence to R. F. Brem
(rbrem{at}mfa.gwu.edu)
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Nine hundred six sequential mammographically detected breast cancers and 147 normal screening mammograms from 18 facilities were classified by mammographic density. BI-RADS 1 and 2 density cases were classified as nondense breasts; BI-RADS 3 and 4 density cases were classified as dense breasts. Cancers were classified as either masses or microcalcifications. All mammograms from the cancer and normal cases were evaluated by the CAD system. The sensitivity and false-positive rates from CAD in dense and nondense breasts were evaluated and compared.
RESULTS. Overall, 809 (89%) of 906 cancer cases were detected by CAD; 455/505 (90%) cancers in nondense breasts and 354/401 (88%) cancers in dense breasts were detected. CAD sensitivity was not affected by breast density (p = 0.38). Across both breast density categories, 280/296 (95%) microcalcification cases and 529/610 (87%) mass cases were detected. One hundred fourteen (93%) of the 122 microcalcifications in nondense breasts and 166 (95%) of 174 microcalcifications in dense breasts were detected, showing that CAD sensitivity to microcalcifications is not dependent on breast density (p = 0.46). Three hundred forty-one (89%) of 383 masses in nondense breasts, and 188 (83%) of 227 masses in dense breasts were detected—that is, CAD sensitivity to masses is affected by breast density (p = 0.03). There were more false-positive marks on dense versus nondense mammograms (p = 0.04).
CONCLUSION. Breast density does not impact overall CAD detection of breast cancer. There is no statistically significant difference in breast cancer detection in dense and nondense breasts. However, the detection of breast cancer manifesting as masses is impacted by breast density. The false-positive rate is lower in nondense versus dense breasts. CAD may be particularly advantageous in patients with dense breasts, in which mammography is most challenging.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Computer-aided detection (CAD) has been developed to improve mammographic detection of breast cancer. Recent studies evaluating CAD have shown a greater than 20% improvement in sensitivity with the use of CAD for mammographic detection of breast cancer [4, 5].
A significant factor limiting the sensitivity of mammography is increased breast density. Analysis of characteristics of cancers missed on screening mammography show that greater breast density is associated with missed breast cancers [6–8]. Therefore, additional methods to improve the detection of breast cancer, particularly in mammographically dense breasts, are needed. With the recent data supporting the positive impact of CAD for the mammographic detection of breast cancer, the question of the performance of the CAD in women with dense breasts was raised—that is, was the performance of CAD impacted by breast density?
The purpose of this study was to determine whether breast density affects the performance of a CAD system for the mammographic detection of breast cancer.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The mammographic characteristics of the cancer cases were classified as microcalcifications or masses. Masses included spiculated and circumscribed masses, architectural distortions, and focal asymmetric densities. If a cancer was a mass with calcifications, the lesion was classified by the primary characteristic of the tumor. A mass with calcification was classified by the primary characteristic—that is, if the lesion was a mass with a few calcifications, it was classified as a mass. If a lesion was predominantly calcifications with an associated density, it was characterized as microcalcifications.
One hundred forty-seven randomly selected normal cases were included to evaluate the number of false-positive marks per mammogram. Initially, 150 normal cases were targeted to be included in this study. However, there were incomplete data for three cases; therefore, 147 cases were included. A normal case was defined as a screening mammogram for which the findings were interpreted as normal, not requiring any further workup (BI-RADS 1) and for which there had been at least 3 subsequent years of normal findings on screening mammography (BI-RADS 1 or 2) with no clinical finding to suggest malignancy. Mammograms with normal findings were obtained between 1992 and 1997.
The 906 mammograms with findings of cancer and 147 with normal findings were evaluated by the Second Look CAD system. Mammograms were digitized using a CCD digitizer with 12 bits of gray-scale and 43 µ resolution. The CAD system used proprietary algorithms to detect potential areas of concern. The CAD report consists of the digitized images with ellipses and rectangles highlighting potential areas of concern. The ellipses mark potential masses (circumscribed masses, spiculated masses, architectural distortions, and asymmetric densities), and rectangles mark potential microcalcifications.
The locations of cancers on the mammogram and CAD printouts were specified with the use of a transparent grid-template overlay using 1-cm squares. The 1-cm square grid overlay was used on the mammogram to determine the precise location of the cancer. Similarly, an appropriately scaled transparent grid was overlaid on the paper CAD reports to assess the marking of the cancers by the CAD system. CAD-marked lesions were within the same 1-cm grid square on the CAD output and the mammogram, marked on at least one view by the CAD system, and correlated by lesion feature—that is, mass or microcalcifications were considered true-positive detections.
The overall sensitivity of the CAD system for the detection of breast cancer in nondense and dense breasts was evaluated for differences in performance in these two groups. In addition, CAD detection of microcalcification cancer cases and mass cancer cases in dense and nondense cases was evaluated. Differences in CAD detection rate of cancer in dense and nondense breasts were compared using chi-square analysis.
The total number of false-positive marks per case was determined by adding the number of microcalcification and mass marks in each normal case. To determine whether more false-positive marks are found in dense or nondense breasts, the number of CAD marks in normal dense and nondense mammograms were calculated and compared using chi-square analysis.
The design of the CAD system limits the total number of marks per case. Therefore, theoretically, in the cancer cases, the maximum number of false-positive marks could not be achieved because of the required marks for the cancer. In practice, the maximum number of marks per case is rarely reached. Nevertheless, the design of the study used normal mammograms to assess the false-positive marks per case to allow the maximum number of false-positive marks per case. Statistical analysis showing p values of less than 0.05 were considered statistically significant.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Overall, 809 (89%) of 906 cancer cases were detected by CAD. Two hundred eighty (95%) of 296 cancers that manifested mammographically as microcalcifications, and 529 (87%) of 610 cancers that manifested mammographically as masses were detected by CAD. It detected 455 (90%) of the 505 cancer cases in nondense breasts, and 354 (88%) of 401 cancer cases in dense breasts. There was no statistically significant difference in CAD of cancer in dense versus nondense breasts (p = 0.38).
One hundred fourteen (93%) of the 122 microcalcification cancer cases in nondense breasts and 166 (95%) of 174 microcalcification cancer cases in dense breasts were detected by the CAD system. There was no statistically significant difference in detection of cancers that manifested mammographically as microcalcifications in nondense versus dense breasts (p = 0.46). Examples of these images are seen in Figures 1A, 1B and 2A, 2B.
|
|
|
|
Of the cancers that mammographically manifested as masses, 341 (89%) of 383 cases in nondense breasts and 188 (83%) of 227 cases in dense breasts were detected. There was a statistically significant difference in the CAD performance for the detection of mass cancer cases in nondense versus dense breasts (p = 0.03). Examples of these images are seen in Figures 3A, 3B and 4A, 4B.
|
|
|
|
Of the 147 normal cases, the average number of false-positive marks per case was 2.95, with 0.65 microcalcification markers and 2.30 mass markers per case. There were 0.59 microcalcification false-positives and 2.09 mass false-positives per case in nondense normal breasts for a total of 2.68 false-positives per case in nondense normal breasts. The mean number of false-positive per case in dense normal breasts was 3.35, made up of 0.75 microcalcification false-positives and 2.60 mass false-positives per case.
Further analysis of the number of cases with no false-positive microcalcification marks revealed that 35 (58%) of 60 cases with dense breast tissue had zero microcalcification false-positives, whereas the remaining 25 (42%) had one or more. In nondense cases, 55 (63%) of 87 had no microcalcification false-positives, whereas the remaining 32 (37%) had one or more. The distribution of microcalcification false-positive marks did not differ significantly between nondense and dense breast tissue cases (p = 0.55).
Regarding mass false-positives, 25 (42%) of 60 cases with dense breasts had two or fewer mass false-positives, whereas the remaining 35 (58%) had three or more false-positives. In nondense breasts, 52 (60%) of 87 cases had two or fewer mass false-positives, whereas the remaining 35 (40%) had three or more. There were statistically significantly fewer mass false-positive marks in nondense than in dense breast cases (p = 0.03).
In 18 (30%) of 60 cases with dense tissue there were two or fewer total false-positive marks, whereas the remaining 42 (70%) had three or more. Of the 87 nondense cases, 41 (47%) had two or fewer total false-positives, whereas 46 (53%) had three or more. There were significantly more total false-positive marks in dense breasts than in nondense breasts (p = 0.04).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Multiple studies have shown the improvement in breast cancer detection with the implementation of CAD [4, 5, 12]. Evaluations of CAD in both the academic [4, 5] and private [12] practice settings have shown a similar improvement in breast cancer detection. Warren Burhenne et al. [4] found that the radiologists' false-negative rate of 21% potentially could have been reduced by 77% with CAD prompting. A study by Brem et al. [5] showed similar results with a false-negative rate decrease of 65% with CAD.
Our study shows that overall CAD performance for the detection of breast cancer was not impacted by breast density. Similarly, no statistically significant difference in CAD detection of cancers manifesting as microcalcifications was found. However, a statistically significant difference in CAD detection of malignant masses occurred in nondense versus dense breasts. CAD detection of masses is lower in dense breasts because of the greater "background noise" and similar density of masses to the surrounding parenchymal density. This difference, however, is still a smaller impact of breast density on mammographic sensitivity of breast cancer detection than has been reported for mammographic detection when CAD is not used. Few studies have evaluated the sensitivity of CAD in breasts of different densities. Ho and Lam [13] showed a decrease in sensitivity of CAD from 93.9% in women with nondense breasts to 64.3% in women with markedly dense breasts. This study included a total of 264 mammograms with 108 cancer and 156 normal cases. Although Ho and Lam's findings are in contrast to ours, the difference may be due to the variance in the size of the study population in the two studies, with our study containing a nearly ninefold larger sample size of cancer cases. Another possible explanation for the difference in findings may be the proportion of mass versus microcalcification cases included in the study populations. Our study showed a statistically greater detection of mass cancer cases in nondense breasts. Therefore, it is possible that the proportion of mass versus microcalcification cases included could impact the statistical analysis. However, this study, which included 33% microcalcification cases, reflects the proportion of cancer cases that manifest as microcalcifications in clinical practice.
Similar to our findings, a recent study by Birdwell et al. [11], which included 110 positive cancer-screening mammograms, showed no significant difference in CAD performance in dense and nondense breasts. During screening mammography without the use of a CAD system, Birdwell et al. found that the second most frequently suggested reason for a missed cancer was greater breast density, second only to another distracting lesion. The dense breast was cited as a factor for lesion misses more often in cases of missed calcifications (34%) than for masses (14%), a finding that is different from that reported in our study. Birdwell et al. showed that CAD performed similarly in the detection of calcifications (83%) and masses (82%) in dense breasts and in nondense breasts. The results of our study, in conjunction Birdwell et al.'s findings, show that CAD has a high sensitivity for the detection of microcalcifications equally throughout all breast densities. Although our study shows a difference in the sensitivity of CAD of masses in nondense (89%) and dense (83%) breasts, the sensitivity remains high in mammograms depicting both nondense and dense breast parenchyma.
Our study again confirmed multiple prior studies reporting significantly higher sensitivity of CAD in detecting microcalcifications versus masses [4, 5, 12]. Our results indicate that CAD may be particularly helpful in patients with dense breasts. Overall CAD sensitivity for the detection of breast cancer in dense breasts is 88%, with a 95% sensitivity for microcalcifications. When compared with other detection techniques such as sonography, physical examination, and mammography alone, mammography with CAD showed superior sensitivity at breast cancer detection in BI-RADS 3 and 4 density breasts [8].
The need for improved diagnosis of breast cancer in women with increased breast density is further emphasized by the greater risk of breast cancer in this population. Boyd et al. [14] showed that women with mammographically dense breasts have a 1.8- to sixfold increased risk of breast cancer. A possible explanation for this finding is that the increased glandular tissue in a dense breast may result in greater propensity for malignant transformation by virtue of the increased amount of tissue. However, it is unclear what mechanism or genetic predisposition results in increased breast density or the resultant increased risk for breast cancer [15]. Regardless, the impact of using CAD for the detection of breast cancer may lead to a reduction in breast cancer mortality rates in these higher risk women. Additional studies are needed to further investigate the impact of CAD in women with dense breast tissue and breast cancer mortality.
Baum et al. [16] have shown that CAD can be used without significantly increasing workup rate or radiologist interpretation time, resulting in improved breast cancer detection without decreased radiologist productivity. Although our study shows additional false-positives in patients with dense breasts, this finding should not impact workup rate. The increased false-positives without additional workup rate can be explained by the fact that not all false-positives result in patient recall. The function of CAD is to point out potential areas of concern to the radiologist. The radiologist makes the final decision. The results of the findings by Baum et al. imply that the false-positives are noted not to be of clinical concern and therefore do not result in patient recall. Furthermore, a recent study by Freer and Ulissey [12] showed that the use of CAD does not decrease the positive predictive value for biopsy. These studies show that CAD does not result in a negative impact on recall rate or positive predictive rate for breast biopsy.
In conclusion, our data suggest that increased breast density does not affect the performance of a CAD system for the detection of breast cancer. The use of CAD can improve breast cancer detection, regardless of breast density. The importance of this finding becomes more significant when considering the intrinsic increased risk of women with dense breasts. In addition, dense breast tissue is generally found in predominantly young and premenopausal women [14]. CAD proved to be particularly adept at marking microcalcifications in dense breasts but was nevertheless effective in detecting mass lesions in dense breast tissue. The decreased performance of CAD for detecting mass lesions in women with dense breasts supports further development of CAD algorithms to enhance the detection of malignant masses in patients with dense breast tissue.
Acknowledgments
We wish to acknowledge all the sites that contributed cases and all the
investigators of the study: Lawrence Bassett, School of Medicine, Iris Cantor
Center for Breast Imaging, UCLA, Los Angeles, CA; Janet K. Baum, Beth Israel
Deaconess (Harvard), Boston, MA; Wendie Berg, Department of Radiology,
University of Maryland, Baltimore, MD; Richard Bird, Presbyterian Breast
Center, Charlotte, NC; Steven Edell, Delaware SPECT Imaging Center, Newark,
DE; Stuart Kaplan, Strax Institute, Fort Lauderdale, FL; Nagi Khouri, Green
Springs Clinic (Johns Hopkins), Baltimore, MD; Mary Lechner, Park Nicollet
Hospital, St. Louis Park, MN; G. Richard Locke, Department of Clinical
Research, Decatur Memorial Hospital, Decatur, IL; Jim Lynde, Department of
Mammography, Virginia Baptist Hospital, Lynchburg, VA; Barbara Monsees,
Washington University, Mallinckrodt Institute of Radiology, St. Louis, MO; L.
Gil Naul, Scott and White Clinic, Temple, TX; Thomas Poulton, Department of
Radiology, Aultman Hospital, Canton, OH; Robert Rosenberg, Department of
Radiology, University of New Mexico, Albuquerque, NM; Darrell Smith,
Department of Radiology, Brigham and Women's Hospital (Harvard), Boston, MA;
Stuart Sauders, Kaiser Permanente Medical Center, Fontana, CA; Don Young,
Department of Radiology, University of Iowa, Breast Imaging and Diagnostic
Center, Iowa City, IA.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  J. S. The, K. J. Schilling, J. W. Hoffmeister, E. Friedmann, R. McGinnis, and R. G. Holcomb Detection of Breast Cancer with Full-Field Digital Mammography and Computer-Aided Detection Am. J. Roentgenol., February 1, 2009; 192(2): 337 - 340. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. J. Kim, W. K. Moon, N. Cho, J. H. Cha, S. M. Kim, and J.-G. Im Computer-aided Detection in Full-Field Digital Mammography: Sensitivity and Reproducibility in Serial Examinations Radiology, December 1, 2007; 246(1): 71 - 80. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. K. Yang, W. K. Moon, N. Cho, J. S. Park, J. H. Cha, S. M. Kim, S. J. Kim, and J.-G. Im Screening Mammography-detected Cancers: Sensitivity of a Computer-aided Detection System Applied to Full-Field Digital Mammograms Radiology, July 1, 2007; 244(1): 104 - 111. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. J. Kim, W. K. Moon, N. Cho, J. H. Cha, S. M. Kim, and J.-G. Im Computer-aided Detection in Digital Mammography: Comparison of Craniocaudal, Mediolateral Oblique, and Mediolateral Views Radiology, December 1, 2006; 241(3): 695 - 701. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. M. Ko, M. J. Nicholas, J. B. Mendel, and P. J. Slanetz Prospective assessment of computer-aided detection in interpretation of screening mammography. Am. J. Roentgenol., December 1, 2006; 187(6): 1483 - 1491. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
