| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 All authors: Department of Radiology, Duke University Medical Center, Box 3808, DUMC, 2nd Fl., Red Zone, South Hospital, Durham, NC 27710.
Received March 25, 2004;
accepted after revision July 23, 2004.
Address correspondence to M. S. Soo.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. A commercially available CAD system evaluated mammograms of 82 patients with 85 mammographically detected and histologically sampled groups of amorphous calcifications (21 malignant, 14 high risk, and 50 benign). The sensitivity of the system for detecting the calcifications on at least one image of the two-view mammographic examination (case sensitivity) and on each individual mammographic image (image sensitivity) was determined. Findings were correlated with results from large core needle biopsy or surgical excision in each case.
RESULTS. The CAD system detected amorphous calcifications in 43 of 85 cases (case sensitivity, 51%) and in 59 of 146 mammographic images (image sensitivity, 40%). The case sensitivities by histologic outcome were 57% for malignant calcifications, 29% for high-risk calcifications, and 54% for benign calcifications. An average of 2.0 false-positive marks were displayed per case.
CONCLUSION. The CAD sensitivity for malignant amorphous calcifications is markedly lower than previously reported for all malignant calcifications. Breast imaging radiologists who use CAD systems should continue to search diligently for these difficult-to-detect lesions.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Amorphous calcifications are small, indistinct yet significant calcifications that reportedly represent 34% of the calcifications recommended for biopsy and that have a 20% rate of malignancy and an additional 20% rate of high-risk disease [4]. These calcifications have been described by Berg et al. [4] as indistinct calcifications not falling into definitely benign categories (skin calcifications, fat necrosis, popcorn, secretory, vascular, scattered) or higher probability of malignancy categories (fine linear or branching, pleomorphic) because even on spot magnification views their small size and lack of characteristic morphology do not allow definitive characterization. These calcifications are frequently difficult to identify on routine mammograms, with 78% seen retrospectively but not prospectively by Berg et al. in their series. Because amorphous calcifications without associated findings can be easily overlooked on routine screening mammograms, the use of CAD systems to aid in detection could be beneficial to the radiologist. Therefore, the purpose of our study was to determine the sensitivity of one commercially available CAD system for detecting amorphous calcifications unassociated with other mammographic findings.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The radiologist evaluated the routine craniocaudal and mediolateral oblique mammograms plus any spot compression magnification mammograms for more accurate characterization of the calcifications. During the review process, only calcifications strictly adhering to the criteria for amorphous calcifications were included. Groups that also contained calcifications meeting the description of pleomorphic, round, or dystrophic calcifications or groups associated with a mammographic mass, architectural distortion, or another abnormality were excluded from the study.
After image review, 41 of the 117 lesions were excluded from further evaluation. In nine cases, the mammograms were missing or only poor quality copy film was available for review, and in four cases the amorphous calcifications were seen only on magnification mammograms, not on routine craniocaudal and mediolateral oblique mammograms. In 24 excluded cases, the calcifications did not meet criteria for amorphous calcifications or had associated findings such as a mass or architectural distortion. Four lesions were excluded because biopsy was not performed. The remaining 76 lesions subsequently underwent histologic evaluation with either core needle biopsy or open surgical biopsy, resulting in 12 malignant lesions (16%) (three invasive ductal carcinomas; nine ductal carcinomas in situ [DCIS]) and 14 high-risk lesions (18%) (eight atypical ductal hyperplasia, atypical lobular hyperplasia, or lesions with atypia; one lobular carcinoma in situ [LCIS]; five complex sclerosing lesions or radial scars). Diagnosis of each of the 12 malignant and 14 high-risk lesions was confirmed at open surgical biopsy. The remaining 50 lesions (66%) were benign.
Nine additional consecutive cases of malignant amorphous calcifications (two invasive ductal carcinomas and seven DCIS) were identified from a separate biopsy database of malignant lesions to enrich the population with malignant lesions. As with cases from the first database, these cases were also reviewed to confirm that these lesions met the criteria for amorphous calcifications and were then added to the 76 existing study cases. These 85 cases of amorphous calcifications without an associated mammographic abnormality in 82 patients (age range, 36–78 years; mean, 56 years) constituted the study population for evaluation with the CAD system.
The 85 cases in the study population included 21 malignant lesions (25%) (16 low- to intermediategrade DCIS, five invasive ductal carcinomas with high-grade DCIS), 14 high-risk lesions (16%), and 50 benign lesions (59%). Core biopsy was performed initially in 74 cases, and open surgical biopsy was initially performed in 11. Of the 50 benign lesions, six were diagnosed at open surgical biopsy and the remaining 44 were diagnosed at core needle biopsy. Imaging follow-up (mean follow-up time, 26 months; range, 9–43 months) of these 44 lesions sampled with core biopsy showed stable or resolved findings in 38. Six cases were lost to follow-up.
At the time of film review, the BI-RADS categorization for parenchymal density was recorded for each case [3]. In addition, the size and BI-RADS distribution (clustered, segmental, linear, regional) of the amorphous calcifications were recorded. Statistical analyses using Wilcoxon's two-sample test and Fisher's exact test were performed to determine whether any significant difference existed between the benign, high-risk, and malignant lesions based on these characteristics.
CAD Analysis and Case Evaluation
Craniocaudal and mediolateral oblique mammograms from each patient were
analyzed using a commercially available CAD system (ImageChecker MI000,
version 3.2, R2 Technology). Digital images of each mammographic film were
made using the digitizer that is included with the CAD system (50-µm
resolution). The digital images were then analyzed by the software included in
the CAD system, which was designed to identify breast cancers presenting as
calcifications or mass lesions. The system searches for clusters of bright
spots on the digitized image that could represent calcifications, then the
output of the CAD system is displayed on two small videotape monitors or one
larger flatpanel display [1].
The monitors display miniature versions of the digitized mammograms, over
which the system places a small triangle to indicate the site of possible
calcifications and an asterisk to mark areas of possible masses.
The location of the amorphous calcifications was determined on both craniocaudal and mediolateral oblique mammograms by the radiologist who had access to all diagnostic images and biopsy and wire localization images. This location was then compared with the location marked by the CAD system to determine whether the calcifications on each mammogram were correctly identified by the CAD system. A true-positive mark was established if the system placed either a triangle or an asterisk over the specific calcific particles as shown by the CAD system. Although an asterisk indicates a possible mass, this mark was included as a true-positive because both the triangle and asterisk would prompt the radiologist to carefully inspect the area that would include the calcifications.
The sensitivity of the CAD system for identifying the amorphous calcifications in the 85 cases was determined. The case sensitivity was determined by dividing the number of cases in which the amorphous calcifications were correctly marked on the craniocaudal or mediolateral oblique mammograms or on both by the total number of cases of amorphous calcifications. The image sensitivity was determined by dividing the number of mammogram films on which amorphous calcifications were detected by the total number of mammogram films on which this finding was visible. Similarly, the case sensitivities and image sensitivities for the histologic subsets of lesions (malignant, high risk, and benign) in the population were determined.
In addition to evaluating the sensitivity of CAD for detecting malignant amorphous calcifications in the overall study population of 21 malignant lesions, we also determined the sensitivity of CAD for detecting malignant amorphous calcifications in the subpopulation of 12 consecutive malignant lesions identified from the larger database series alone to ensure that the addition of malignant cases from a second database introduced no selection bias into the analysis. False-positive lesions marked by the CAD system were recorded, and the average number of false-positive marks was determined. True-positive marks were considered only those marks that correctly indicated the location of the amorphous calcifications. Coincident lesions other than amorphous calcifications seen on the mammograms that had been recommended for biopsy initially were excluded from evaluation.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The amorphous calcifications were seen by the study radiologist on both the routine craniocaudal and mediolateral oblique films in 61 cases (72%), and on only one of the two projections in the remaining 24 cases (28%). Therefore, 146 of the 170 images showed findings that were considered actionable. Of the subset of 21 malignant lesions, the amorphous calcifications were seen by the radiologist on both routine craniocaudal and mediolateral oblique projections in 15 cases and on only one projection in six cases. These findings were therefore considered actionable on 36 (86%) of 42 films. For the 14 high-risk lesions, the amorphous calcifications were seen by the radiologist on both routine projections in eight cases and on only one projection in six cases. These high-risk lesions were therefore actionable on 22 (79%) of 28 films. Image sensitivities were calculated based on only those films with findings deemed actionable.
The mean size of the groups of amorphous calcifications was 9 mm (range, 2–40 mm). The calcification distribution was either clustered (n = 77) or segmental (n = 8). No linear or regional distributions were identified. Malignant groups of amorphous calcifications ranged from 3 to 40 mm in greatest size (mean greatest size, 10 mm) and were either clustered (n = 17) or segmental (n = 4). High-risk lesions ranged from 2 to 11 mm (mean size, 5 mm) and were all clustered in distribution. Benign lesions ranged in size from 2 to 35 mm (mean size, 9 mm) and were clustered (n = 46) or segmental (n = 4) in distribution. Calcifications in the high-risk group tended to be smaller than those in the benign and malignant groups (p = 0.04). No significant difference was noted in the distribution of the lesions between the benign, high-risk, and malignant groups.
Tables 1, 2, 3 show the accuracy of CAD marks per case as related to breast density, lesion size, and distribution, respectively. The CAD system correctly marked amorphous calcifications on at least one of two mammographic views in 43 of 85 cases, for an overall case sensitivity of 51%. The system placed triangles indicating calcifications in 41 (95%) of 43 cases and asterisks in two cases (5%). Of the malignant subset of lesions within the overall study population (five invasive ductal carcinomas and 16 DCIS), the CAD system had a case sensitivity of 57%, marking 12 of 21 lesions. Figures 1A, 1B, and 1C shows a malignant cluster of calcifications overlooked by the CAD system. Of the nine missed malignant lesions, all were DCIS (Table 4 and Figs. 1A, 1B, and 1C).
|
|
|
|
|
|
|
All invasive carcinomas were detected by the CAD system, each being marked
by a triangle. In the subset of 12 consecutive malignant amorphous
calcifications identified from a single database, the CAD system correctly
identified eight of 12 as malignant lesions (case sensitivity, 67%). For the
high-risk lesions, the case sensitivity was 29%, marking only four of 14
lesions. The system also marked 27 (54%) of 50 benign lesions. There was no
significant difference in the CAD system's ability to detect calcifications
between the benign and malignant groups (p 
0.59).
In the population overall, the CAD system successfully identified the amorphous calcifications on 59 of the 146 images on which amorphous calcifications were detected by the radiologist, for an image sensitivity of 40%. For malignant lesions, the system correctly marked 19 (53%) of 36 films, and for high-risk lesions, it correctly marked five (23%) of 22 films. The system marked 35 (40%) of 88 films showing amorphous calcifications that proved benign histologically.
Overall, the CAD system displayed 171 false-positive marks (78 calcification marks and 93 mass marks), with a mean of 2.0 falsepositive marks (range, 0–7 marks) per case and 0.6 per image (Figs. 2A, 2B, and 2C).
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The use of CAD systems in this manner reportedly results in approximately 20% improvement in cancer detection beyond the interpretation of the radiologist alone, and most initially overlooked malignant lesions detected by the CAD system are calcifications [5, 6]. Results of other studies also support the use of these systems by showing the system's high sensitivity for cancer detection on screening mammography [1, 2]. One study reported a 99% sensitivity for the CAD system in detecting 406 malignant calcifications on films taken at the time of diagnosis and detection of 79% of 110 malignant calcifications originally missed on the previous year's study in those cases [1]. A second study found CAD useful in detecting 86% (30/35) of missed cancers represented as calcifications [2]. These studies have supported the use of CAD systems for aiding the radiologist as a second reviewer in evaluating screening mammograms, and the systems are rapidly gaining acceptance in the breast imaging community.
Based on the data from these previous studies, a negative interpretation by the radiologist and the absence of CAD marks likely increases the radiologist's confidence that no malignancy represented by calcifications is present on the mammographic images. Some radiologists might even be tempted to reduce the time spent searching for calcifications or even to omit this step of the evaluation completely, given the reported success of the CAD system for identifying malignant calcifications; however, this is not the intended use of this CAD system.
Amorphous calcifications are small, faint, difficult-to-detect calcifications that are significant because they represent 33% of malignant calcifications and 34% of all calcification lesions recommended for biopsy [4]. Twenty percent of amorphous calcifications prove to be malignant, and another 20% represent high-risk lesions [4]. In our consecutive cases of 76 amorphous calcifications without other mammographic features, we found percentages of malignant and high-risk lesions— 16% and 18%, respectively—similar to those reported in the literature. Unfortunately, as confirmed in our study, mammography does not adequately discriminate among benign, high-risk, and malignant disease represented by amorphous calcifications. We found no significant difference in lesion size or distribution between the benign and malignant subsets of amorphous calcifications, although the high-risk lesions tended to be smaller. Therefore, identification and biopsy of all cases of amorphous calcifications are necessary.
Amorphous calcifications are frequently overlooked because they are at the threshold of visibility on routine mammographic images [4]. In a study by Berg et al. [4], 78% (52/67) of amorphous calcifications could be seen retrospectively on prior films but had not been detected prospectively. Spot magnification mammograms allow proper characterization of these lesions; however, even minimal motion can result in blurring of the image, causing the calcifications to be overlooked. In addition, malignant amorphous calcifications can be stable in appearance for many years, likely contributing to the difficulty and variability in managing these lesions [4]. These small faint calcifications are also difficult to target for stereotactic biopsy procedures because the resolution of digital units is limited and procedures are frequently aborted or avoided initially because of limited conspicuity.
Given the difficulty in detection of amorphous calcifications, assistance in identifying these lesions by a CAD system would be welcomed by radiologists. Overall, the CAD system in our study successfully detected only 51% of amorphous calcifications that were prospectively recommended for biopsy. In the subset of malignant amorphous calcifications, the system had the highest case sensitivity (57%) compared with the benign group (54%) and high-risk group (29%). Although the techniques used by the commercial CAD systems are proprietary, the lower detection rate of amorphous calcifications in our study compared with calcifications in previous studies probably relates to the small size of calcific particles and decreased contrast resolution of these faint calcifications. In general, the CAD system was most successful in marking the larger, more extensive groups of malignant and benign calcifications, compared with the smaller clustered high-risk lesions (Table 2). Interestingly, however, the system (4/8) only marked half of the larger segmentally distributed calcifications, three of which were malignant and one benign.
Another factor in detecting these calcifications may relate to breast density. A previous report suggests that CAD sensitivity is inversely related to BI-RADS parenchymal density and therefore that dense breast tissue would limit detection of some lesions [6]. It follows that amorphous calcifications would then be more difficult to detect in breasts that have higher density on mammography because of further decrease in contrast resolution between the calcifications and background. Although a trend to this effect was seen in our study, the statistical evidence was not considered very strong, and the two cases of amorphous calcifications in fatty breasts were not marked by the CAD system.
A recent report discussing reproducibility of prompts in CAD systems suggested that slight changes in a film position between sequential digitizations could result in variations of individual pixel values, thereby causing a CAD mark to move from just above to just below the threshold for detection on repeated scans [7]. Reproducibility was not evaluated in this study; however, amorphous calcifications may be especially susceptible to this problem given their small size and low contrast resolution.
Although the CAD system's detection of 57% of the malignant cases of amorphous calcifications may prove beneficial to the radiologist in a clinical setting for any given case, the detection rate in this subset of malignant calcifications is far less than that reported by Warren Burhenne et al. [1] for malignant calcifications (99%). Although the calcification descriptors were not reported in that study, amorphous calcifications could have been underrepresented because the mammograms evaluated were obtained between 1994 and 1996 [1], a time when there was great variability in the management of amorphous calcifications [4, 8]. A second report by Birdwell et al. [2] using a subset of films (those with prior mammograms for comparison) from the study by Warren Burhenne et al. described only 6% (2/35) of the calcification lesions as amorphous, which is lower than the 33% in a subsequent consecutive series of biopsied malignant calcifications by Berg et al. [4].
Given that amorphous calcifications represent approximately one third of calcification lesions that undergo biopsy, the radiologist must be cautious not to become overly reliant on the CAD system for detecting calcifications. Rather than assuming the CAD system will detect nearly 100% of malignant calcifications and therefore bypassing inspection of the films for calcifications, the radiologist must continue to search diligently in every case for these subtle yet important signs of breast carcinoma that can be missed by the CAD system.
As with prior studies, our study is limited in that it could evaluate the sensitivity of the CAD system only for those groups of amorphous calcifications that were detected by the human observer [9]. It is possible that calcifications overlooked by the radiologist and therefore not included in the study could have been detected by the CAD system, resulting in a higher percentage of detected calcifications than found in our study. A prospective study with long-term mammographic follow-up of each CAD mark would be required to further investigate this issue.
In conclusion, although CAD systems have been successful in detecting many malignant calcifications seen mammographically and can improve the cancer detection rate during screening, the subset of amorphous calcifications are much less reliably detected. Given that these lesions are challenging for the radiologist alone to detect, the 57% detection rate suggests that the system may be useful in certain cases. However, radiologists certainly should not become overly reliant on the CAD system for detection of these significant lesions but should continue to search the films diligently for these subtle yet important signs of breast carcinoma.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  D. Georgian-Smith, R. H. Moore, E. Halpern, E. D. Yeh, E. A. Rafferty, H. A. D'Alessandro, M. Staffa, D. A. Hall, K. A. McCarthy, and D. B. Kopans Blinded Comparison of Computer-Aided Detection with Human Second Reading in Screening Mammography Am. J. Roentgenol., November 1, 2007; 189(5): 1135 - 1141. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. L. Ellis, A. A. Meade, M. A. Mathiason, K. M. Willison, and W. Logan-Young Evaluation of Computer-aided Detection Systems in the Detection of Small Invasive Breast Carcinoma Radiology, October 1, 2007; 245(1): 88 - 94. [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]
|
|
|
|
|
|
F. M. Hall Computer-Aided Detection (CAD) of Amorphous Calcifications Am. J. Roentgenol., March 1, 2006; 186(3): 902 - 902. [Full Text] [PDF]
|
|
|
|
|
|
M. S. Soo, S. Ghate, J. A. Baker, and E. Rosen Reply Am. J. Roentgenol., March 1, 2006; 186(3): 902 - 903. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
