| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Breast Imaging and Intervention, George Washington University
Medical Center, 2150 Pennsylvania Ave. N.W., Washington, DC 20037.
2 Department of Radiology, Beth Israel Deaconess Medical Center, 330 Brookline
Ave., CC0492, Boston, MA 02215.
3 Jane Brattain Breast Center, Park-Nicollett Clinic, St. Louis Park, MN
55416.
4 Strax Institute, 4300 N. University Dr., Lauderhill, FL 33332.
5 Department of Radiology, Kaiser Permanente Medical Center, 9961 Sierra Ave.,
Fontana, CA 92335.
6 Department of Radiology, Scott and White Clinic, 2401 S. 31st St., Temple, TX
76508.
7 CADx Systems, Ste. 100, 2689 Commons Blvd., Beavercreek, OH 45431.
Received December 17, 2002;
accepted after revision March 4, 2003.
Address correspondence to R. F. Brem.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Three radiologists reviewed 377 screening mammograms interpreted as showing normal or benign findings 9-24 months before cancer diagnosis from 17 of the 18 participating centers. In 313 cases, study radiologists recommended additional mammographic evaluation. In 177 cases, the area warranting additional workup precisely correlated with the subsequently diagnosed cancer. These 177 missed cancers were evaluated with computer-aided detection. The proportion of radiologists identifying the missed cancers was used to determine radiologist sensitivity without computer-aided detection.
RESULTS. The study radiologists determined that 123 of the 377 missed cancer cases warranted workup. Therefore, 123 additional cancers cases could have been found. The calculated radiologist sensitivity without computer-aided detection was therefore 75.4% (377 / [377 + 123]). Similarly, using the performance of the system on the missed cancers, we estimated that 80 (65.0%) of these 123 missed cancer cases would have been identified with the use of computer-aided detection. Consequently, the estimated sensitivity of radiologists using computer-aided detection was 91.4% ([377 + 80] / [377 + 123])—resulting in a 21.2% ([91.4% / 75.4%] - 1) increase in radiologist sensitivity with computer-aided detection.
CONCLUSION. Use of the computer-aided detection system significantly improved the detection of breast cancer by increasing radiologist sensitivity by 21.2%. Therefore, for every 100,000 women with breast cancer identified without the use of computer-aided detection, an estimated additional 21,200 cancers would be found with the use of computer-aided detection.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Although a recent analysis of eight randomized breast cancer screening trials questions the reduction in mortality from screening with mammography [3] and the Physician Data Query Panel sponsored by the National Cancer Institute concurs with this uncertainty [4], other United States and international organizations still affirm that screening with mammography, clinical breast examination, or both is effective in reducing breast cancer mortality. The United States Preventive Services Task Force, sponsored by the United States Department of Health and Human Services, commissioned an evaluation of eight mammography screening trials and concluded that mammography reduces breast cancer mortality by 16% [5]. The International Agency for Research on Cancer, which is part of the World Health Organization, concluded that these randomized trials showed that mammography reduces breast cancer mortality by 25-35% in women who are 50-69 years old [6].
The sensitivity of mammography ranges from approximately 70% to 90% [7-14]. Thus, for a woman with breast cancer, the probability that her cancer will be detected on mammography is 70-90% and the probability that it will not be detected is 10-30%. Therefore, even though mammography is an effective tool to use to detect breast cancer, thereby reducing mortality, further improvement in mammographic sensitivity is needed.
Studies have shown that the accuracy of mammographic interpretation increases when a mammogram is evaluated by two radiologists; mammographic interpretation by two radiologists improves breast cancer detection by 5-15% [15-26]. However, this approach is not currently advocated as a standard of care and requires substantial additional resources that are often not available [26]. A cancer can be missed because it is not mammographically visible, because it is not detected by the interpreting radiologist (i.e., an oversight), or because the radiologist misinterprets the finding. Clinical studies have shown that 30-70% of breast cancers diagnosed at screening mammography are visible in retrospect on prior examinations and that detection errors are responsible for approximately half of missed breast cancers, with interpretation errors accounting for the other half [10, 27-33].
In view of the frequency of missed breast cancers, computer-aided detection of breast lesions on mammograms has been evaluated as a method to improve mammographic sensitivity. In a study of the potential benefit of adding computer-aided detection to a single radiologist review [34], an analysis method was developed to estimate the increase in sensitivity a radiologist could expect with computer-aided detection. The results indicated that review by a single radiologist combined with computer-aided detection represents a possible alternative to mammographic interpretation by a single radiologist only or by two radiologists for reducing the rate of detection errors that lead to missed cancers. In fact, a prior study reported breast cancer detection is improved with the use of computer-aided detection—even when used by experienced mammographers [35].
The purpose of our study was to evaluate radiologist detection of breast cancer using a computer-aided detection system.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Study Design
Case characteristics.—This multicomponent study was
conducted between April 1999 and May 2001. None of the participating centers
enrolled more than 25% of the total study population for any case category.
All cases of breast cancer were in women 29 years old or older (age range,
29-93 years). Patients with normal findings on mammograms were approximately
10 years younger (mean age, 52.3 years) than those who had either potentially
or actually missed cancers (mean age, 62.3 and 64.4 years, respectively).
These findings reflect the fact that the risk of developing breast cancer
increases with patient age. All mammograms in this trial dated from less than
7 years before the start of the study so that only images of current quality
were included. All mammograms consisted of craniocaudal and mediolateral
oblique views for both the left and right breasts.
Missed cancers.—This study was designed to evaluate the efficacy of a computer-aided detection system in the detection of missed cancers. From a total of 930 screening-detected breast cancers accrued from 17 of the 18 sites, 377 potentially missed cancers were identified (Fig. 1). Potentially missed cancer cases were defined as mammographically detected breast cancer cases diagnosed 9-24 months after a mammogram with "normal" or "benign" findings (i.e., not requiring any additional workup). The interval of 9-24 months was chosen for efficiency of data collection because the most recent prior mammograms for women who follow typical breast cancer screening guidelines were generally expected to be within this interval. Mammograms known to have been reviewed by two radiologists were excluded.
|
Six radiologists were designated as the study radiologists. They reviewed the 377 potentially missed cancer cases to determine whether areas warranting additional workup were present on the prior mammograms in order to identify the (actually) missed cancer cases. Each case was reviewed by three of the six study radiologists. All study radiologists were board-certified mammographers, met the qualification standards of the Mammography Quality Standards Act [36], and had interpreted a minimum of 800 mammograms per year.
The study radiologists evaluated the potentially missed cancer cases without clinical information and without other images or additional prior mammograms. As stated in the protocol, in order to enhance blinding, the study radiologists were told that normal cases might be included as distracters among the cases they reviewed. In fact, no normal cases were included in the review of the prior screening mammograms. None of the study radiologists could review a case from his or her participating center. During the review of the potentially missed cancer cases, each study radiologist identified regions on the prior mammograms that warranted workup and characterized each region as mass, microcalcifications, or mixed (i.e., a mass lesion with microcalcifications), where "mass" lesions included architectural distortions and focal asymmetric densities. The study radiologists assigned an assessment recommendation on the basis of the Breast Imaging Reporting and Data System (BIRADS) categories of the American College of Radiology [37] and precisely identified the location of the region that warranted workup using coordinates from a scaled grid overlay. The coordinates from the scaled grid overlay were used to confirm that the mammographic abnormality correlated precisely with the location of recommended workup. Coordinates of the overlying grid had to be within 1 grid or 5 mm to be considered in the same location.
Any case from the 377 potentially missed cancer cases for which additional workup was recommended by at least one of the three study radiologists was forwarded for a "truthing" review to assess whether the region for which workup had been recommended corresponded precisely to the location of the pathologically proven cancer, again with the use of a scaled grid overlay. The characterization of the lesion recommended for workup was also compared with the mammographic appearance of the cancer. Different radiologists constituted the truthing radiologists than the study radiologists. All radiologists participating in this study had interpreted more than 800 mammograms per year, as per the protocol, and all truthing radiologists were board-certified and met the qualification standards of the Mammography Quality Standards Act.
The prior screening mammogram and radiology report for each case were independently reviewed by two truthing radiologists with the use of supporting clinical and pathology data. Additionally, the mammogram that showed the cancer was also used to identify the precise location and mammographic appearance of the cancer. The truthing radiologists confirmed that the mammographic abnormality was in precisely the same location as the cancer by using a scaled grid overlay. Lesions visible in only one view were eligible to be included in this study. Any discrepancies in interpretation between the two truthing radiologists were resolved by a consensus review during which both radiologists reviewed the case together.
If one or more of the three study radiologists recommended workup with the appropriate lesion characterization in the location that was confirmed to correlate precisely with the location of the cancer, the case was classified as a missed cancer and the prior screening mammogram was processed by the computer-aided detection system (Second Look [version 3.4], CADx Systems, Beavercreek, OH). The precise location of the region recommended for workup was correlated with the location of the cancer by comparing the grid coordinates of the region the study radiologist recommended for workup with the grid coordinates of the location of the cancer as determined by the truthing radiologists. For the study radiologists to correctly identify the missed cancer lesion, they also had to characterize the lesion appropriately, where a mixed characterization was also considered correct for lesions determined to be only a mass or only microcalcifications by the truthing radiologists.
Of the 377 potentially missed cancer cases, there were 177 cases in which the region recommended for workup by at least one of the three study radiologists was confirmed to correlate to the precise location of the cancer and the lesion was characterized correctly. These 177 cases are the missed cancer cases that were then evaluated by the computer-aided detection system. The areas of mammographic abnormalities marked by the computer-aided detection system were then assessed to determine whether the location and characterization of the computer-aided detection mark corresponded precisely to the mammographically detected cancer. This assessment was performed by two scoring radiologists with scaled grid overlays used on both the computer-aided detection printout and the prior mammogram. Any discrepancies in interpretation between the scoring radiologists were resolved by consensus.
Each case was determined to be a true-positive or a false-negative by comparing the grid coordinates of the marks from the computer-aided detection system with the grid coordinates of the confirmed mammographic location of the cancer (the "truth lesion"). If a case had a mass or microcalcification mark at the location of a retrospectively visible lesion of the same characterization (or mixed) on at least one mammogram, it was a true-positive. Otherwise, it was a false-negative.
Normal cases.—One hundred fifty-five mammograms with normal findings were obtained from eight of the 18 participating centers. Although all participating centers could accrue cases with normal findings, only eight of the 18 sites chose to do so. One of the centers submitted only cases with normal findings. A normal case was defined as a screening mammogram that was interpreted as showing normal findings, not requiring any further workup (BI-RADS category 1), and for which there were at least 3 subsequent years of screening mammography with normal findings (BI-RADS category 1 or 2). Mammograms known to have been reviewed by two reviewers were excluded. Because the normal cases contained no cancer lesions, all computer-aided detection marks in these cases were considered false-positives. All normal cases were accrued sequentially.
Analysis of Mammograms by the Computer-Aided Detection System
The 177 missed cancer cases and 155 mammograms with normal findings used in
this study were analyzed by the computer-aided detection system. The report
generated by the computer-aided detection system consists of the digitized
images with ellipses and rectangles highlighting potential areas of concern.
The ellipses mark potential masses (nonspiculated masses, spiculated masses,
architectural distortions, and asymmetric densities), and the rectangles mark
potential microcalcifications. Figure
2A is a computer-aided detection printout that shows potentially
suspicious areas highlighted, and Figure
2B is a computer-aided detection printout that a radiologist used
while reviewing the mammogram.
|
|
Statistical Methods
The sensitivity of the original interpreting radiologists at the study
centers was estimated as follows:
 |
The sensitivity of these radiologists using computer-aided detection (CAD)
was estimated as follows:
 |
Nmissed is the estimated number of missed cancer cases that could have been found by the original reviewing radiologists at the study centers at the time of the missed cancer mammogram. Nmissed was calculated using the proportion of study radiologists who correctly identified the truth lesion or lesions as a likelihood multiplier. The appropriate proportion of study radiologists (i.e., 0.3333, 0.6667, or 1.0) was multiplied by the number of missed cancers identified by one of three, two of three, and three of three study radiologists, respectively. The sum of these three quantities was calculated.
NCAD is the estimated number of cancer cases that a radiologist using computer-aided detection would identify in addition to Nfound. NCAD was calculated by multiplying the proportion of study radiologists (i.e., 0.3333, 0.6667, or 1.0) by the number of cases correctly marked by the computer-aided detection system in each of these three categories and then calculating the sum of these three quantities.
A 95% confidence interval (CI) for the relative risk of the estimated sensitivities was calculated to assess the statistical significance of the relative increase in sensitivity with the use of computer-aided detection as compared with the sensitivity without computer-aided detection. This interval was then expressed as a relative change by subtracting 1 from both the lower and upper bounds and multiplying by 100.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Malignant lesions at the locations that had been recommended for workup by the study radiologists with the correct characterization were confirmed in 177 of the 313 cases reviewed by the truthing radiologists. These 177 cases were the missed cancer cases that were processed by the computer-aided detection system and then forwarded to the scoring radiologists. Of these 177 cases, the lesions were correctly identified by one of the three reviewing radiologists in 58 cases, by two in 46 cases, and by all three radiologists in 73 cases.
In 71 of the 313 cases, the truthing radiologists found no malignant lesion. In 65 cases, the truthing radiologists found at least one malignant lesion but at a different location or with a different characterization from any region recommended for workup by the groups of three study radiologists. These 136 cases were not analyzed by the computer-aided detection system.
Tumor Stage
Information about tumor size from pathology and histology results was
collected if available at the participating site. Although there were 377
potentially missed cancer cases included in this study, pathologic cancer size
was available in 249. Similarly, of the 177 missed cancer cases, pathologic
size was available in 114. Table
1 refers to those cancers in which pathologic cancer size was
available.
|
Tumor staging data (TNM stage for tumor size, lymph nodes, and metastasis)
were obtained from sites with ready access to these data. These results are
summarized in Table 2. The mean
tumor size was 1.20 cm, and 76.9% of the tumors were invasive or microinvasive
cancers (Table 1). In 23 of the
177 missed cancer cases, primary tumor staging was unknown (all were invasive
or microinvasive tumors with size unavailable or unknown)
(Table 2). Among the 154 missed
cancer cases with known primary tumor staging, 102 (66.2%) were stage T1
(tumors 
2 cm in greatest diameter), and 41 (26.6%) were classified as Tis
(carcinoma in situ). Other stages (T2, T3, and T4) accounted for 11 (7.1%). In
90 of the 177 missed cancer cases, lymph node status could not be assessed
(NX) or was unknown. Among the 87 cases with known lymph node status, 75
(86.2%) had no regional lymph node metastasis (N0). None of the cases had
metastasis to the ipsilateral internal mammary lymph node (N3) or known
distant metastasis (M1) at the time of diagnosis.
|
Truthing Results
Missed cancer cases with mammographically visible malignant lesions were
classified as primarily a mass in 116 (65.5%) of the 177 cases and as
primarily microcalcifications in 61 (34.5%) of the 177 cases. In 170 (96.0%)
of the 177 cases, there was one malignant lesion; the remaining seven cases
(4.0%) consisted of two malignant lesions each. Lesion size ranged from 0.3 to
9.0 cm in largest diameter on mammograms, with an average diameter of 1.26 cm.
Approximately equal numbers of cancers were seen in the left and right
breasts. One patient had bilateral cancers.
Radiologist Sensitivity With and Without Computer-Aided
Detection
There were 177 missed cancer cases in this study. The number of study
radiologists who correctly identified the location of a lesion on the prior
mammogram with the appropriate characterization and recommended workup was
used to estimate the number of missed cancer cases that could have been found
by the original reviewing radiologists at the study centers. This number of
potentially detectable missed cancers was calculated by multiplying the
corresponding proportion of study radiologists (i.e., 33%, 67%, or 100%) by
the number of missed cancers identified by one of three, two of three, and
three of three study radiologists, respectively. The sum of these three
quantities was calculated (Table
3). The total result for Nmissed was 123 cases; this
value represents the estimated number of mammograms with retrospectively
visible cancers recommended for workup by the study radiologists conducting
the blinded review that were missed by the original interpreting
radiologists.
|
Because there were 377 (Nfound) potentially missed cancer cases with subsequent screening mammograms that led to a cancer diagnosis, the estimated sensitivity of a radiologist was 75.4% (377 / [377 + 123]). This calculated number of mammograms warranting workup (n = 123) was compared with the number of mammograms with retrospectively visible cancers that were correctly identified by the majority (two or three of three) of the study radiologists. The truth lesion or lesions were correctly identified by two (67%) of three study radiologists in 46 cases and by three (100%) of three study radiologists in 73 cases (Table 3). Thus, the majority of the study radiologists correctly identified and recommended 119 (46 + 73) cases for workup, which compares well to the calculated 123 potentially detectable missed cancer cases.
To estimate the number of additional cancers that would have been detected with the use of computer-aided detection, we multiplied the proportion of study radiologists recommending a workup by the number of cases that the computer-aided detection system correctly marked (Table 3). The computer-aided detection system detected 63% of the missed cancer cases that had a subsequent screening mammogram that led to a cancer diagnosis (111/177). Similar to the calculation of the 123 potentially detectable missed cancers, the performance of the system was used to estimate that 80 (NCAD) of these 123 missed cancer cases would have been identified by the original interpreting radiologists if they had used computer-aided detection. The estimated sensitivity of a radiologist using computer-aided detection was therefore 91.4% ([377 + 80] / [377 + 123]). This sensitivity of a radiologist using computer-aided detection is a 21.2% ([91.4% / 75.4%] - 1) increase (95% confidence limits = 14.5%, 28.3%) compared with the sensitivity of a single radiologist.
Normal Cases
In the 155 normal cases, the average number of mass and microcalcification
marks per mammogram was 1.00 and 0.25, respectively.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A prior study found that 36% of the most recent mammograms obtained 6-39 months (mean, 14 months) before the mammogram on which cancer was diagnosed were considered to have evidence of that cancer [30]. A recent study by Warren Burhenne et al. [34] indicated that 27% of prior mammograms (i.e., 115/427 mammograms obtained 9-24 months earlier) showed findings that warranted workup. Similarly, we found that mammograms obtained 9-24 months before the cancer diagnosis showed findings that warranted follow-up in 32% of the cancer cases in our study.
Researchers have inferred that detection or interpretation errors could be largely responsible for the incidence of missed breast cancers [10, 27, 33]. This may explain why breast cancer detection rates are higher with double interpretation [15-26]. Computer-aided detection is another way to reduce detection errors because it prompts the radiologist to reexamine the mammogram at the location of the computer-aided detection marks to determine whether further workup is necessary.
This study evaluated the benefit of using a computer-aided detection system by analyzing the mammograms obtained before a cancer diagnosis with computer-aided detection and estimating the increased sensitivity of breast cancer detection. The benefit derived from the computer-aided detection system was calculated as proportional to the number of radiologists who correctly identified the lesions at blinded review, so a more conservative estimate of computer-aided detection benefits could be obtained. For instance, if only one of three study radiologists identified a lesion, then only 33% of the cases correctly marked by computer-aided detection were considered in the calculation of computer-aided detection benefit (i.e., to determine that the original radiologist would have recommended workup for the cancer marked by computer-aided detection). It was only when all three study radiologists identified a lesion that 100% of the cases correctly marked by computer-aided detection were considered to calculate its benefit.
The estimated number of cases that would have benefited from the use of the computer-aided detection system (i.e., the cancer would have been detected on the prior mammogram if the computer-aided detection system had been used) was 80 (65%) of the 123 cancer cases found to warrant workup by study radiologists (i.e., cases that could have been detected by the radiologists who conducted the original mammography review). This value indicates that computer-aided detection could have helped reduce the rate of missed cancers by 65%.
To assess the impact of using computer-aided detection on radiologist sensitivity, we estimated that the sensitivity of a radiologist using computer-aided detection was 91.4%, which represented a 21.2% increase in radiologist sensitivity. Given that a 5-15% increase in breast cancer detection—as observed with double interpretation—is considered clinically significant [15-26], this 21.2% increase in radiologist sensitivity with the use of computer-aided detection is a clinically meaningful improvement in the detection of breast cancer and is similar to the improvement in sensitivity reported by Warren Burhenne et al. [34]. These estimates of sensitivity are based on a population of patients who had screening mammograms 9-24 months before cancer diagnosis. If it can be assumed that those cases without screening mammograms 9-24 months earlier are not appreciably different, similar results could be expected for the entire population. Thus, for every 100,000 women with breast cancer currently identified by a radiologist without using computer-aided detection, an estimated additional 21,200 cancers would be found with the use of computer-aided detection.
In conclusion, this study has shown that a computer-aided detection system can significantly improve radiologist sensitivity in detecting breast cancer during screening mammography. The computer-aided detection system could also be helpful in screening mammography for the earlier detection of breast cancer—a particularly important benefit because women with breast cancer that is detected earlier have a better prognosis. The results of this multiinstitutional trial indicate that computer-aided detection may be viewed as a potential tool to help reduce the number of missed breast cancers, assist in the earlier detection of breast cancers, and serve as an alternative to double interpretation.
Acknowledgments
We acknowledge all of the additional sites that contributed cases and all
the additional investigators of the study: Nagi Khouri, Johns Hopkins
University Hospital, Baltimore, MD; Lawrence Bassett, Iris Cantor Center for
Breast Imaging, UCLA School of Medicine, Los Angeles, CA; 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; G. Richard Locke, Department of Clinical Research,
Decatur Memorial Hospital, Decatur, IL; Jim Lynde, Mammography Department,
Virginia Baptist Hospital, Lynchberg, VA; Thomas Poulton, Radiology
Department, Aultman Hospital, Canton, OH; Robert Rosenberg, Department of
Radiology, University of New Mexico, Albuquerque, NM; Barbara Monsees,
Washington University, Mallinckrodt Institute of Radiology, St. Louis, MO;
Darrell Smith, Department of Radiology, Brigham and Women's Hospital, Boston,
MA; and Don Young, Department of Radiology, University of Iowa, Breast Imaging
and Diagnostic Center, Iowa City, IA.
We thank Nadege Chery for assistance in the preparation of the manuscript. We thank Glenn Pixton for assistance with statistical analysis.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
263b
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  F. J. Gilbert, S. M. Astley, M. G.C. Gillan, O. F. Agbaje, M. G. Wallis, J. James, C. R.M. Boggis, S. W. Duffy, and the CADET II Group Single Reading with Computer-Aided Detection for Screening Mammography N. Engl. J. Med., October 1, 2008; (2008) NEJMoa0803545v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Gromet Comparison of Computer-Aided Detection to Double Reading of Screening Mammograms: Review of 231,221 Mammograms Am. J. Roentgenol., April 1, 2008; 190(4): 854 - 859. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. F. Brem Blinded Comparison of Computer-Aided Detection with Human Second Reading in Screening Mammography: The Importance of the Question and the Critical Numbers Game Am. J. Roentgenol., November 1, 2007; 189(5): 1142 - 1144. [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]
|
|
|
|
|
|
J. J. Fenton, S. H. Taplin, P. A. Carney, L. Abraham, E. A. Sickles, C. D'Orsi, E. A. Berns, G. Cutter, R. E. Hendrick, W. E. Barlow, et al. Influence of Computer-Aided Detection on Performance of Screening Mammography N. Engl. J. Med., April 5, 2007; 356(14): 1399 - 1409. [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]
|
|
|
|
|
|
V. R. Pai, N. E. Gregory, A. E. Swinford, and M. Rebner Ductal Carcinoma in Situ: Computer-aided Detection in Screening Mammography Radiology, December 1, 2006; 241(3): 689 - 694. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. J. Gilbert, S. M. Astley, M. A. McGee, M. G. C. Gillan, C. R. M. Boggis, P. M. Griffiths, and S. W. Duffy Single Reading with Computer-aided Detection and Double Reading of Screening Mammograms in the United Kingdom National Breast Screening Program Radiology, October 1, 2006; 241(1): 47 - 53. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. C. Dean and C. C. Ilvento Improved cancer detection using computer-aided detection with diagnostic and screening mammography: prospective study of 104 cancers. Am. J. Roentgenol., July 1, 2006; 187(1): 20 - 28. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. K. Lindfors, M. C. McGahan, C. J. Rosenquist, and G. S. Hurlock Computer-aided Detection of Breast Cancer: A Cost-effectiveness Study Radiology, June 1, 2006; 239(3): 710 - 717. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. J. Morton, D. H. Whaley, K. R. Brandt, and K. K. Amrami Screening Mammograms: Interpretation with Computer-aided Detection--Prospective Evaluation Radiology, May 1, 2006; 239(2): 375 - 383. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. S. Greenberg An Appellate Court Ruling and Potential Implications for CAD Technology in the Courtroom Am. J. Roentgenol., January 1, 2006; 186(1): 52 - 53. [Full Text] [PDF]
|
|
|
|
|
|
L. A. L. Khoo, P. Taylor, and R. M. Given-Wilson Computer-aided Detection in the United Kingdom National Breast Screening Programme: Prospective Study Radiology, November 1, 2005; 237(2): 444 - 449. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. A. Baker, E. L. Rosen, M. M. Crockett, and J. Y. Lo Accuracy of Segmentation of a Commercial Computer-aided Detection System for Mammography Radiology, May 1, 2005; 235(2): 385 - 390. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. F. Brem, J. W. Hoffmeister, G. Zisman, M. P. DeSimio, and S. K. Rogers A Computer-Aided Detection System for the Evaluation of Breast Cancer by Mammographic Appearance and Lesion Size Am. J. Roentgenol., March 1, 2005; 184(3): 893 - 896. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. F. Brem, J. W. Hoffmeister, J. A. Rapelyea, G. Zisman, K. Mohtashemi, G. Jindal, M. P. DiSimio, and S. K. Rogers Impact of Breast Density on Computer-Aided Detection for Breast Cancer Am. J. Roentgenol., February 1, 2005; 184(2): 439 - 444. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K Doi Current status and future potential of computer-aided diagnosis in medical imaging Br. J. Radiol., January 1, 2005; 78(suppl_1): S3 - s19. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P Taylor and R M Given-Wilson Evaluation of computer-aided detection (CAD) devices Br. J. Radiol., January 1, 2005; 78(suppl_1): S26 - S30. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. Gur, J. S. Stalder, L. A. Hardesty, B. Zheng, J. H. Sumkin, D. M. Chough, B. E. Shindel, and H. E. Rockette Computer-aided Detection Performance in Mammographic Examination of Masses: Assessment Radiology, November 1, 2004; 233(2): 418 - 423. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. M. Hall and R. F. Brem Improved Sensitivity of Mammography with Computer-Aided Detection Am. J. Roentgenol., June 1, 2004; 182(6): 1598 - 1599. [Full Text] [PDF]
|
|
|
|
|
|
M. A. Guenin and R. F. Brem How Not to Assess Computer-Aided Detection for Mammography Am. J. Roentgenol., June 1, 2004; 182(6): 1599 - 1600. [Full Text] [PDF] |
