DOI: 10.1148/radiol.2413051145
(Radiology 2006;241:695-701.)
© RSNA, 2006
Computer-aided Detection in Digital Mammography: Comparison of Craniocaudal, Mediolateral Oblique, and Mediolateral Views1
Seung Ja Kim, MD2,
Woo Kyung Moon, MD,
Nariya Cho, MD,
Joo Hee Cha, MD,
Sun Mi Kim, MD and
Jung-Gi Im, MD
1 From the Department of Radiology and Clinical Research Institute, Seoul National University Hospital and the Institute of Radiation Medicine, Seoul National University Medical Research Center, 28, Yongon-dong, Chongno-gu, Seoul, 100-744, Korea (S.J.K., W.K.M., N.C., S.M.K., J.G.I.); and Department of Radiology, Boramae Municipal Hospital, Seoul, Korea (J.H.C.). Received July 8, 2005; revision requested September 12; revision received October 19; accepted November 17; final version accepted February 1, 2006. Supported by KISTEP, Ministry of Science and Technology, Korea.
Address correspondence to W.K.M. (e-mail: moonwk{at}radcom.snu.ac.kr).
 |
ABSTRACT
|
|---|
Purpose: To retrospectively compare the sensitivity of a computer-aided detection (CAD) system for depicting breast cancer in three digital mammographic views.
Materials and Methods: This study was conducted with institutional review board approval; informed consent was waived. A commercially available CAD system was applied to the craniocaudal, mediolateral oblique, and mediolateral digital mammographic views of 83 women (mean age, 48 years; range, 30–66 years) with 83 histologically proved breast cancers. Findings were 59 masses and 41 microcalcifications (17 lesions showed both findings; 42 lesions, mass only; and 24 lesions, microcalcification only). The paired t test was used to analyze sensitivity of the CAD system for the detection of cancer in these three mammographic views and in combinations of the views.
Results: The sensitivities of the CAD system were 92% (76 of 83) in the craniocaudal view, 83% (69 of 83) in the mediolateral oblique view, and 86% (71 of 83) in the mediolateral view; the differences were not significant (P = .07–.62). Sensitivity increased to 96% (80 of 83) in the craniocaudal plus mediolateral oblique views and to 99% (82 of 83) in the craniocaudal plus mediolateral oblique plus mediolateral views. For masses, the sensitivity of the CAD system was 76% (45 of 59) in the craniocaudal view and 75% (44 of 59) in the mediolateral oblique view and increased to 93% (55 of 59) when mediolateral oblique and craniocaudal views were combined (P < .001). For microcalcifications, sensitivity was 98% (40 of 41) in the craniocaudal view and 95% (39 of 41) in the mediolateral oblique view, and this increased to 100% (41 of 41) when the mediolateral oblique and craniocaudal views were combined (P = .31).
Conclusion: The sensitivities of the CAD system were not significantly different among these three digital mammographic views. Sensitivity for depicting masses was significantly increased (P < .001) when the craniocaudal view was added to the mediolateral oblique view.
© RSNA, 2006
|
INTRODUCTION
|
|---|
A computer-aided detection (CAD) system can assist the radiologist in the early detection of breast cancer by highlighting suspicious areas at screening mammography. In a retrospective study by Warren Burhenne et al (1), the CAD system correctly marked 77% (89 of 115) of missed breast cancers. Findings from the prospective study of Freer and Ulissey (2) showed that the use of CAD for 12 860 screening mammograms resulted in a 20% (eight of 41) increase in the number of cancers detected, with an increase in recall rate from 6.5% to 7.5%. The sensitivity of the CAD algorithm is greater for the detection of microcalcifications (86%–99%) than for the detection of masses (75%–86%) (1,3–5). Recently, the CAD system was applied to full-field digital mammography, and results for breast cancer detection were reported to be similar to those obtained with an analog system (6).
Three views (craniocaudal, mediolateral oblique, and mediolateral) are most commonly used for screening and diagnostic mammographic examinations. The mediolateral oblique view is the single view that includes most of the breast. A combination of craniocaudal and mediolateral oblique views is accepted as the standard examination because this combination is better at depicting breast cancer and has a lower false-positive rate than does the mediolateral oblique view alone (7–10). The mediolateral view is usually obtained for localizing a lesion in three dimensions, and the value of adding this view in high-risk patients has been addressed (8,9,11). Commercially available CAD systems can be applied to these images. Application of CAD to full-field digital mammography is limited, however, and no report, to our knowledge, has been issued about the sensitivity of the CAD system for these mammographic views. Thus, the purpose of our study was to retrospectively compare the sensitivity of a CAD system for the detection of breast cancer in these three digital mammographic views.
|
MATERIALS AND METHODS
|
|---|
Patient Selection
From November 2003 to April 2005, 191 women suspected of having breast cancer underwent digital mammography in the three views (craniocaudal, mediolateral oblique, and mediolateral) with a full-field digital mammographic unit (Senographe 2000D; GE Medical Systems, Milwaukee, Wis). Of these 191 patients, we selected 83 women (mean age, 48 years; range, 30–66 years) with 83 histologically proved breast cancers and visible mammographic signs of malignancy on all three views. All other patients (n = 108) were excluded because of mammographic signs of malignancy on only one or two views (n = 87), more than one mammographically visible suspicious lesion per image (n = 19), or bilateral cancers (n = 2). Mediolateral views were obtained while magnification views were obtained. The interval between the performance of the standard examination and the acquisition of the mediolateral view ranged from 0 to 33 days (mean, 1.6 days). No intervention, such as percutaneous biopsy, was performed during this interval. This study was conducted with institutional review board approval; informed consent was waived.
Mammographic and Histologic Findings
Of the 83 patients, seven (8%) had homogeneously fatty breasts (American College of Radiology Breast Imaging Reporting and Data System [BI-RADS] category 1 density), 13 (16%) had scattered fibroglandular tissues in fatty breasts (BI-RADS category 2 density), 37 (45%) had heterogeneously dense breasts (BI-RADS category 3 density), and 26 (31%) had extremely dense breasts (BI-RADS category 4 density) (12). Forty-seven tumors were in the upper outer quadrant, 22 in the upper inner quadrant, four in the lower outer quadrant, and seven in the lower inner quadrant; three were subareolar. Fifty-nine suspicious lesions were described as masses, and 41 were described as microcalcifications. Seventeen lesions had both signs of malignancy, 42 were described only as masses, and 24 were described only as microcalcifications. The masses were 10 mm or smaller (n = 4), 11–20 mm (n = 29), and 21–30 mm (n = 26), and the microcalcifications were 10 mm or smaller (n = 16), 11–20 mm (n = 9), 21–30 mm (n = 8), 31–40 mm (n = 5), and 41 mm or larger (n = 3).
Final histopathologic diagnosis of malignant lesions at surgery included noncomedo-type ductal carcinoma in situ (n = 3); comedo-type ductal carcinoma in situ (n = 9); invasive ductal carcinoma, usual type (n = 68); mucinous carcinoma (n = 1); invasive micropapillary carcinoma (n = 1); and mixed lobular and ductal carcinoma (n = 1). For treatment, mastectomy was performed in 21 patients, and breast conservation therapy was performed in 62 patients.
CAD Evaluation
A commercially available CAD system (ImageChecker M1000-DM, version 3.1; R2 Technology, Sunnyvale, Calif) developed for full-field digital mammographic images was applied to the three mammographic views by one of the three attending radiologists (S.J.K., N.C., J.H.C.) who interpreted the initial images. The three radiologists had 3–6 years of experience with breast imaging. In practice, the system was used by the radiologist after initial review of digital mammograms. It took 1 second to activate and display the CAD marks at the workstation (GE Medical Systems). The CAD system marks regions suspicious for a mass or a microcalcification cluster by superimposing a small asterisk or triangle, respectively, on the image. Images with CAD marks were saved in the review workstation and then sent to a picture archiving and communication system for data analysis.
Two radiologists (S.J.K., N.C.) together used all available mammographic views, including magnification views, to determine the outline (reference standard) of the actual breast cancer (ie, margins of a mass or extent of microcalcifications) for the 83 patients in our study. The two radiologists had 3 and 6 years, respectively, of experience with breast imaging. The radiologists determined in consensus whether each CAD mark indicated the location of the malignancy. If the asterisk was located anywhere inside a true-positive mass, this mass was considered to have been identified correctly by the CAD system. Similarly, as long as a triangle was overlapping any of the microcalcification areas, the CAD marks were considered to represent true-positive detection. All CAD marks that did not denote the known malignancy were defined as false-positive marks. When the CAD system marked typical benign calcifications or crossing lines, we considered them false-positive calcification marks.
Data and Statistical Analysis
The sensitivities and number of false-positive marks allocated by the CAD system, as discussed in the preceding paragraph, were analyzed for the ipsilateral breast. Sensitivities of the CAD system were calculated in two ways. First, sensitivity was calculated for masses and for microcalcifications. In the case of a mass with microcalcifications, each component was considered separately. Because the CAD system marks the mass component and the calcification component separately, we counted both the mass mark and the calcification mark for the malignancies that presented both as a mass and calcification. Second, case-based sensitivity was calculated for the involved breast. In the case of a mass with microcalcifications, any CAD mark denoting either component was regarded as an identification. We analyzed sensitivities for the craniocaudal, mediolateral oblique, and mediolateral views and for three combinations (craniocaudal and mediolateral oblique views; craniocaudal and mediolateral views; and the combination of craniocaudal, mediolateral oblique, and mediolateral views). The number of false-positive marks per image for masses or microcalcifications were calculated for each view and for the combinations. We classified each breast into one of the two following groups according to its composition: fatty (BI-RADS density category of 1 and 2) and dense (BI-RADS density category of 3 and 4). We also determined whether the CAD sensitivities and false-positive rates were related to breast composition.
We applied the paired t test to compare differences between sensitivities for the three views and the combinations of them. We used the unpaired t test to compare differences in sensitivities in the fatty and dense breast groups and to compare differences between the numbers of false-positive marks for the three views and the combinations of them. A P value of less than .05 was considered to indicate a statistically significant difference. Statistical analyses were performed with software (SPSS, version 10 for Windows; SPSS, Chicago, Ill).
|
RESULTS
|
|---|
CAD Marks
Of the 83 breast cancers, 58 were marked by the CAD system on all three views, one was not marked on any view, four were marked on the craniocaudal view only, none were marked on the mediolateral oblique view only, and two were marked on the mediolateral view only (Figure). The remaining lesions were marked on various combinations of views. One invasive cancer that was not marked by the CAD system was a 1.6-cm mass without calcifications in the upper outer quadrant.
View larger version (94K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure a: Mammograms in a 44-year-old woman with invasive ductal carcinoma detected as a 9-mm ill-defined mass (arrow) at screening mammography. (a) Craniocaudal (left), mediolateral oblique (center), and mediolateral (right) digital views. (b) Screen-capture images of computer monitor display of CAD system output in the three views. The CAD system marked the mass (*) only in the craniocaudal view.
|
|
View larger version (96K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure b: Mammograms in a 44-year-old woman with invasive ductal carcinoma detected as a 9-mm ill-defined mass (arrow) at screening mammography. (a) Craniocaudal (left), mediolateral oblique (center), and mediolateral (right) digital views. (b) Screen-capture images of computer monitor display of CAD system output in the three views. The CAD system marked the mass (*) only in the craniocaudal view.
|
|
Sensitivities
The case-based sensitivities of the CAD system for the three views and combinations of the three views are summarized in Table 1. Sensitivity was 92% (76 of 83) in the craniocaudal view, 83% (69 of 83) in the mediolateral oblique view, and 86% (71 of 83) in the mediolateral view; the differences were not significant (P = .07–.62). Sensitivity increased to 96% (80 of 83) for the craniocaudal view plus the mediolateral oblique view and to 99% (82 of 83) for both the craniocaudal view plus the mediolateral view and the craniocaudal view plus the mediolateral oblique view plus the mediolateral view; these findings were not significantly different (P = .16).
For masses, sensitivities of the CAD system were 76% (45 of 59) for the craniocaudal view, 75% (44 of 59) for the mediolateral oblique view, and 76% (45 of 59) for the mediolateral view; for microcalcifications, sensitivities were 98% (40 of 41), 95% (39 of 41), and 95% (39 of 41), respectively (Table 2). Differences were not significant (P = .57–.83). For masses, sensitivity for the mediolateral oblique view was 75% (44 of 59) but increased to 93% (55 of 59) for the craniocaudal view plus the mediolateral oblique view; this finding was significant (P < .001) (Table 3). When the mediolateral view was added to the craniocaudal view plus the mediolateral oblique view, sensitivity increased to 97% (57 of 59) (P = .16). For microcalcifications, sensitivity for the mediolateral oblique view was 95% (39 of 41), and sensitivities for the craniocaudal view plus the mediolateral oblique view and the craniocaudal view plus the mediolateral oblique view plus the mediolateral view were both 100% (41 of 41). However, the differences were not significant (P = .31).
View this table:
[in this window]
[in a new window]
|
Table 3. Performance of CAD System for Mass and Microcalcification Detection with Combinations of Three Mammographic Views
|
|
False-Positive Marks
The numbers of false-positive marks were 0.28 per image for the craniocaudal view, 0.13 per image for the mediolateral oblique view, and 0.27 per image for the mediolateral view (Table 4); these differences were not significant (P = .05–.89). The corresponding numbers of false-positive marks for masses were 0.17, 0.11, and 0.22 per image, respectively, and those for microcalcifications were 0.11, 0.02, and 0.05 per image, respectively. In addition, false-positive marks per image for the combinations of views were 0.20 for the craniocaudal view plus the mediolateral oblique view, 0.27 for the craniocaudal view plus the mediolateral view, and 0.22 for the craniocaudal view plus the mediolateral oblique view plus the mediolateral view; these findings were not significantly different (P = .28–.71).
Breast Groups
The sensitivities of the CAD system in the fatty breast group were 100% (20 of 20) for the craniocaudal view, 80% (16 of 20) for the mediolateral oblique view, and 85% (17 of 20) for the mediolateral view; in the dense breast group, sensitivities were 89% (56 of 63) for the craniocaudal view, 84% (53 of 63) for the mediolateral oblique view, and 86% (54 of 63) for the mediolateral view (Table 1). Differences were not significantly different (P = .12–.94) except for mass lesions in the craniocaudal view (P = .004) (Table 2). The numbers of false-positive marks per image were similar in the fatty and dense breast groups (Table 4).
|
DISCUSSION
|
|---|
Cancers can be missed in one or two mammographic projections for two main reasons: (a) juxtathoracic position, which results in the exclusion of the tumor from the radiograph, and (b) superimposition of dense tissues. Because we selected cases in which lesions were visible in all three views, superimposition of dense tissues is probably the main reason for different sensitivities of the CAD system for these three projections. In our study, the sensitivities of the CAD system for the mediolateral oblique view (83%, 69 of 83) were notably lower than those for the craniocaudal view (92%, 76 of 83), but the difference was not significant. These findings are concordant with findings of previous clinical studies in which the three mammographic views were compared for breast cancer detection (8,9). In one study (8) with 491 breast cancers, the numbers of malignancies that were not visible because of superimposed dense tissue were 22 in the mediolateral oblique view, 14 in the mediolateral view, and 13 in the craniocaudal view. The mediolateral oblique view, however, was the best in which tumors were visualized in the juxtathoracic part of the breast.
Current CAD systems generally detect and help characterize suspicious abnormal structures in individual mammographic images. The clinical experiences of radiologists indicate that screening that involves two mammographic views improves the accuracy of detection of abnormalities in the breast (7–10). A preliminary study by Paquerault et al (13) showed that the fusion of information from different mammographic views could improve the performance of CAD systems. In terms of the detection of malignant masses, detection sensitivity of the CAD system was found to improve from 62% with a one-view detection scheme to 73% with a new two-view scheme, with a false-positive rate of one per image in a data set of 169 pairs of mammograms with the craniocaudal view plus the mediolateral oblique view. Our study results also suggest that the combination of the craniocaudal view and the mediolateral oblique view can increase the capability of the CAD system to detect breast cancers that manifest as mass lesions, with an acceptable increase in the false-positive rate.
In our study, the CAD system was applied to full-field digital mammography and detected 96% (80 of 83) of cancers by using a two-view standard examination. This result is better than the 87% (55 of 63) sensitivity quoted by Baum et al (6), who first reported on the use of a commercially available CAD system for full-field digital mammography. In their study, however, single-view mammograms were used in 15 cases. They also used an older version of the same CAD system we used. Higher detection sensitivities for lesions that manifest as microcalcifications rather than as masses were found in our study and in the study by Baum et al. In our study, sensitivities for masses were higher in the fatty breast group than in the dense breast group. This result is consistent with the findings of a previous study (14) in which the researchers found that the sensitivity of the CAD system for detection of lesions that manifest as masses was influenced by breast parenchymal density.
Primary digital CAD systems have some advantages compared with CAD systems based on analog images. Primary digital CAD systems do not require digitization, which is essential for the application of CAD in screen-film mammography. Digitization can be a source of variability and of false-positive marks in film-based CAD systems (4,5). These variabilities are probably caused by digitization inconsistencies and electronic noise introduced during the digitization process. In our study, 0.20 false-positive mark was observed per image in the craniocaudal and mediolateral oblique views; these results were notably lower than the 0.7–1.0 false-positive mark achieved with a system based on analog images (2,4,15). The false-positive rate, however, was underestimated in our study because the selection criteria included exclusively malignant lesions. Primary digital CAD systems also allow CAD marks to be displayed rapidly after image acquisition. It takes 1 second to activate and display the CAD marks at the workstation.
The limitation of our study was that we selected cases in which lesions were visible in all three views; thus, the overall sensitivity for the CAD system was probably overestimated, and the sensitivity for the mediolateral oblique view was underestimated. Moreover, we did not evaluate the performance of the radiologists who used the CAD system embedded in full-field digital mammography. Nawano et al (16) performed a CAD study by using digital mammograms in 86 women, and the average area under the receiver operating characteristic curve increased significantly (P < .02) when the radiologists used the CAD system. It remains to be shown whether CAD implemented with a full-field digital mammographic system can improve the radiologist's performance for detection of breast cancer.
In conclusion, the sensitivities of the CAD system were not significantly different for the three different digital mammographic views. Sensitivity for mass detection was increased when the craniocaudal and mediolateral views were added to the mediolateral oblique view, whereas combinations of the views had little effect on microcalcification detection.
|
ADVANCES IN KNOWLEDGE
|
|---|
- The sensitivities of the CAD system were not significantly different for three digital mammographic views (P = .07–.62).
- A combination of the craniocaudal and mediolateral oblique views significantly increased the detection of breast cancer that manifested as a mass (P < .001), whereas combinations of the views had little effect on microcalcification detection (P = .31).
|
FOOTNOTES
|
|---|
Abbreviations: BI-RADS = Breast Imaging Reporting and Data System • CAD = computer-aided detection
2 Current address: Department of Radiology, Konkuk University Hospital, Seoul, Korea. 
Authors stated no financial relationship to disclose.
Author contributions: Guarantors of integrity of entire study, S.J.K., W.K.M.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, S.J.K., W.K.M., N.C.; clinical studies, S.J.K., N.C.; statistical analysis, S.J.K., W.K.M., J.H.C., S.M.K.; and manuscript editing, S.J.K., W.K.M., J.G.I.
|
References
|
|---|
- Warren Burhenne LJ, Wood S, D'Orsi C, et al. Potential contribution of computer-aided detection to the sensitivity of screening mammography. Radiology 2000;215:554–562.[Abstract/Free Full Text]
- Freer TW, Ulissey MJ. Screening mammography with computer-aided detection: prospective study of 12,860 patients in a community breast center. Radiology 2001;220:781–786.[Abstract/Free Full Text]
- Helvie MA, Hadjiiski L, Makariou E, et al. Sensitivity of noncommercial computer-aided detection system for mammographic breast cancer detection: pilot clinical trial. Radiology 2004;231:208–214.[Abstract/Free Full Text]
- Baker JA, Lo JY, Delong DM, Floyd CE. Computer-aided detection in screening mammography: variability in cues. Radiology 2004;233:411–417.[Abstract/Free Full Text]
- Zheng B, Hardesty LA, Poller WR, Sumkin JH, Golla S. Mammography with computer-aided detection: reproducibility assessment initial experience. Radiology 2003;228:58–62.[Abstract/Free Full Text]
- Baum F, Fischer U, Obenauer S, Grabbe E. Computer-aided detection in direct digital full-field mammography: initial results. Eur Radiol 2002;12:3015–3017.[Medline]
- Ikeda DM, Sickles EA. Second-screening mammography: one versus two views per breast. Radiology 1988;168:651–656.[Abstract/Free Full Text]
- Andersson I, Hildell J, Muhlow A, Pettersson H. Number of projections in mammography: influence on detection of breast disease. AJR Am J Roentgenol 1978;130:349–351.[Abstract]
- Bassett LW, Bunnell DH, Gold RH, Jahanshahi R. Film-screen mammography: comparison of views. J Natl Med Assoc 1989;81:391–394.[Medline]
- Sickles EA, Weber WN, Galvin HB, Ominsky SH, Sollitto RA. Baseline screening mammography: one vs two views per breast. AJR Am J Roentgenol 1986;147:1149–1153.[Abstract/Free Full Text]
- Kreager JA, Kornguth PJ. Mammography for screening high-risk patients for cancer: value of including a lateral projection. AJR Am J Roentgenol 1994;162:295–297.[Abstract/Free Full Text]
- American College of Radiology. Breast imaging reporting and data system: BI-RADS atlas. 4th ed. Reston, Va: American College of Radiology, 2003.
- Paquerault S, Petrick N, Chan HP, Sahiner B, Helvie MA. Improvement of computerized mass detection on mammograms: fusion of two-view information. Med Phys 2002;29:238–247.[CrossRef][Medline]
- Brem RF, Hoffmeister JW, Rapelyea JA, et al. Impact of breast density on computer-aided detection for breast cancer. AJR Am J Roentgenol 2005;184:439–444.[Abstract/Free Full Text]
- Malich A, Sauner D, Marx C, et al. Influence of breast lesion size and histologic findings on tumor detection rate of a computer-aided detection system. Radiology 2003;228:851–856.[Abstract/Free Full Text]
- Nawano S, Murakami K, Moriyama N, Kobatake H, Takeo H, Shimura K. Computer-aided diagnosis in full digital mammography. Invest Radiol 1999;34:310–316.[CrossRef][Medline]
This article has been cited by other articles:
|
|
|
|
 
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]
|
|
|
TOP
ABSTRACT
INTRODUCTION
