DOI: 10.1148/radiol.2251011029
(Radiology 2002;225:182-189.)
© RSNA, 2002
Invasive Lobular Carcinoma of the Breast: Mammographic Characteristics and Computer-aided Detection1
W. Phil Evans, MD,
Linda J. Warren Burhenne, MD,
Louba Laurie, MD,
Kathryn F. O'Shaughnessy, PhD and
Ronald A. Castellino, MD
1 From the Susan G. Komen Breast Center, Baylor University Medical Center, Dallas, Tex (W.P.E., L.L.); the Department of Radiology, University of British Columbia, Vancouver, Canada (L.J.W.B.); and R2 Technology, Sunnyvale, Calif (K.F.O., R.A.C.). From the 2000 RSNA scientific assembly. Received June 13, 2001; revision requested August 6; final revision received March 15, 2002; accepted March 25. Address correspondence to W.P.E., UT Southwestern Center for Breast Care, 5323 Harry Hines Blvd, Dallas, TX 75390-8585 (e-mail: phil.evans@utsouthwestern.edu).
 |
ABSTRACT
|
|---|
PURPOSE: To characterize the mammographic appearance of invasive lobular carcinoma in a large series of screening-detected consecutive breast cancers and to evaluate the ability of a computer-aided detection system to mark these carcinomas.
MATERIALS AND METHODS: Investigators used the Breast Imaging Reporting and Data System lexicon to characterize lesions as part of a retrospective review of 90 screening mammographic examinations that led to biopsy-proved diagnosis of 94 invasive lobular carcinoma lesions. The 40 available prior mammographic examinations (obtained 9–24 months earlier) were also reviewed to characterize any visible findings. The results of a computer-aided detection analysis were compared with the images, and the sensitivity of the algorithm was calculated for correct detection of the lesions.
RESULTS: Fifty-six (60%) of 94 lesions manifested as masses, of which 40 (71%) were described as irregular and spiculated; 20 (21%) of 94, as architectural distortions; and the remainder, 18 (20%), as either asymmetric densities or calcifications. On the screening mammograms showing biopsy-proved cancers, the sensitivity of the computer-aided detection system was 86 (91%) of 94 lesions. Thirty-one of the 40 prior mammograms showed retrospectively visible findings, and 24 (77%) of 31 were marked by the computer-aided detection system.
CONCLUSION: Spiculated masses and architectural distortions are the predominant appearances of invasive lobular carcinoma, and a computer-aided detection system correctly marked a high percentage of invasive lobular carcinoma lesions.
© RSNA, 2002
Index terms: Breast neoplasms, diagnosis, 00.111, 00.1299 • Breast neoplasms, radiography, 00.111 • Breast radiography, 00.111 • Computers, diagnostic aid
|
INTRODUCTION
|
|---|
The detection of invasive lobular carcinoma (ILC) at screening mammography is known to be a difficult task. False-negative rates have been reported to range from 8% (1) to 19% (2). This most likely is related to the fact that ILC spreads through the breast parenchyma by means of diffuse infiltration of single rows of malignant cells in a linear fashion around nonneoplastic ducts. This infiltration causes little disruption of the underlying anatomic structures and generates little surrounding connective tissue reaction (3). A discrete mass is less common than that seen with other malignancies; instead, a higher incidence of subtle mammographic signs, such as asymmetric density (3%–25%) or architectural distortion (10%–25%) (1,2), has been reported. In addition, ILC is often less radiopaque than or as radiopaque as normal fibroglandular tissue (4–6). Reports of microcalcifications associated with ILC indicated that they are uncommon, with their presence noted only in 0%–10% (1,2,7) of cases, and frequently represent unrelated histologic lesions, such as invasive ductal carcinoma, ductal carcinoma in situ, and sclerosing adenosis. A large fraction of ILCs have been shown to be visible in one view only, most often the craniocaudal view (4,7).
This study, in which cases originally collected as part of a larger study (8) were used, represents an in-depth look at a specific histologic type of breast cancer, ILC, known to be difficult to detect on screening mammograms. We sought to determine whether a computer-aided detection (CAD) system is as sensitive to this challenging histologic group as to a large consecutive series comprising all histologic subtypes. The purposes of our study were to characterize the mammographic appearance of ILC in a consecutive series of screening-detected breast cancers and to evaluate the ability of a CAD system to mark these carcinomas.
|
MATERIALS AND METHODS
|
|---|
Case Collection
Investigators in an earlier study to evaluate the sensitivity of CAD in a large multiinstitutional (n = 13) data set of breast cancers (8) collected the screening mammograms of 1,083 consecutive patients that resulted in a biopsy-proved diagnosis of breast cancer between 1994 and 1996 (referred to as "current" mammograms). Institutional review board approval was obtained at all sites for review of patient records and image collection. All patients were asymptomatic, and their cancers were mammographically evident at screening. One radiologist from each site reviewed the screening mammographic examinations (typically craniocaudal and mediolateral oblique views of each breast) and work-up images to verify the location on the mammogram that corresponded to the biopsy-proved malignancy and provided a reference-standard outline of the lesion on an overlay. A pathology report was obtained for each biopsy-proved case to document the histologic findings for each lesion. Lobular carcinoma in situ was not considered malignant. In general, the specific subtype of ILC was not documented in the pathology reports.
The subset of 90 patients with 94 cancers with at least some component of ILC formed the basis of this study. There were 83 four-film bilateral mammographic examinations (four images on four mammograms) and seven two-film unilateral mammographic examinations. In the original study (8), the most recent prior mammograms of the screening-detected cancers, which had been acquired 9–24 months before the examination at which the cancer was first detected, were obtained. Of the 90 patients with ILC, 40 had prior mammograms available.
The mean age of the patients in this study was 64.8 years (range, 41–86 years). For 62 (66%) of the 94 cancers, only ILC was listed on the pathology report; for an additional 19 (20%), ILC was present with some other type of invasive carcinoma; and for 13 (14%), ILC was associated with ductal carcinoma in situ. The median size of 75 cancers, as documented in the pathology report, was 15 mm (range, 4–90 mm). No size was reported for 19 cancers.
Mammographic Findings
Two experienced mammographers (L.J.W.B., W.P.E.) jointly reviewed high-quality digital copies of the 90 current mammographic examinations. First, an assessment of breast density was given by using BI-RADS classification (9). The mammographic appearance of each lesion was then classified by using the BI-RADS lexicon with regard to lesion type (mass or calcification) and description (eg, mass margins and shape) and by determining whether there were calcifications associated with the mass. It was also noted in which views the lesion was visible. When available, the prior mammogram was then compared with the current mammogram, in particular with regard to the known location of the biopsy-proved cancer on the current mammogram. It was then determined whether the lesion was visible in retrospect (referred to as "visible prior mammogram"); if so, the lesion appearance was described in a fashion similar to that used for the current mammograms.
CAD Evaluation
The CAD system used to analyze the mammograms (ImageChecker M1000 version 2.2; R2 Technology, Sunnyvale, Calif) is composed of a film digitizer, a processing computer, and a motorized film viewer with video monitors. The digitizer provides a digital image of each film by using 50-µm resolution and 12 bits of gray scale. The processing computer uses proprietary signal processing software that highlights regions of interest with the following characteristics: clusters of bright spots (ie, regions suggestive of microcalcification clusters) and dense regions with radiating lines (ie, suggestive of masses or architectural distortions). The "mass" algorithm includes in its design the ability to look for features that are commonly associated with masses, namely, an area with central density and radiating lines. When no central density is found, the radiating lines must be more pronounced to be marked.
The CAD output is displayed on small video monitors located below the illuminator at which the images are viewed. The low-spatial-resolution mammographic images have markers superimposed on them, with a triangle used to indicate the location of a possible microcalcification cluster, and an asterisk used to indicate the location of a possible mass or architectural distortion.
The markers produced by the CAD algorithm were reviewed jointly by the two experienced mammographers to determine whether the location of the abnormality had been identified with the correct type of marker. If both mass and calcification features were noted for the abnormality, then either marker type was considered correct. The sensitivity of the CAD algorithm was calculated as the number of lesions correctly marked (in either view) divided by the total number of lesions. Note that for patients with more than one lesion, each lesion was treated individually. All other marks produced by the algorithm were also noted, and the average number of extra marks per image was calculated.
|
RESULTS
|
|---|
Mammographic Findings
Current mammograms.—Table 1 shows the distribution of breast density ratings in the 90 patients with ILC. In 62% (56 of 90) of the patients with ILC lesions, the lesions were seen in breasts with parenchymal patterns characterized as fatty or scattered fibroglandular. Table 2 describes mammographic lesion appearance for all 94 ILC lesions and the 62 pure ILC lesions. The distribution of mammographic lesion types was similar for all 94 ILCs and 62 pure ILCs. Ninety-six percent (90 of 94) of the lesions were judged as being visible in both views, with the remaining lesions visible in only the craniocaudal (n = 3) or mediolateral oblique (n = 1) view.
Sixty percent (56 of 94) of all ILCs manifested as masses at presentation, with 71% (40 of 56) noted as being irregular and spiculated (Fig 1, Table 3). Twenty (21%) lesions were architectural distortions (Fig 2). Five (9%) of the 56 lesions described as masses had associated microcalcifications. Three (15%) of the 20 lesions described as architectural distortions had associated microcalcifications.
View larger version (93K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1a. Current screening mammogram obtained in a 72-year-old woman. The radiopaque circle in the upper inner quadrant of the left breast is a skin marker. (a) Craniocaudal and (b) mediolateral oblique views of an asymmetric density (arrow) in the upper outer quadrant of the right breast. (c) Low-resolution digitized images of both breasts with CAD markers (*). The CAD system detects the spiculated mass (ILC) in both views. An additional mark (arrow) is present in the subareolar left breast in the mediolateral oblique view only. This area was classified by the interpreting radiologist as superimposed breast tissue. (d) Digitally magnified images (craniocaudal on the left, mediolateral oblique on the right) demonstrate a 1.5-cm irregular mass (arrows) with spiculation, best seen in the mediolateral oblique view.
|
|
View larger version (101K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1b. Current screening mammogram obtained in a 72-year-old woman. The radiopaque circle in the upper inner quadrant of the left breast is a skin marker. (a) Craniocaudal and (b) mediolateral oblique views of an asymmetric density (arrow) in the upper outer quadrant of the right breast. (c) Low-resolution digitized images of both breasts with CAD markers (*). The CAD system detects the spiculated mass (ILC) in both views. An additional mark (arrow) is present in the subareolar left breast in the mediolateral oblique view only. This area was classified by the interpreting radiologist as superimposed breast tissue. (d) Digitally magnified images (craniocaudal on the left, mediolateral oblique on the right) demonstrate a 1.5-cm irregular mass (arrows) with spiculation, best seen in the mediolateral oblique view.
|
|
View larger version (66K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1c. Current screening mammogram obtained in a 72-year-old woman. The radiopaque circle in the upper inner quadrant of the left breast is a skin marker. (a) Craniocaudal and (b) mediolateral oblique views of an asymmetric density (arrow) in the upper outer quadrant of the right breast. (c) Low-resolution digitized images of both breasts with CAD markers (*). The CAD system detects the spiculated mass (ILC) in both views. An additional mark (arrow) is present in the subareolar left breast in the mediolateral oblique view only. This area was classified by the interpreting radiologist as superimposed breast tissue. (d) Digitally magnified images (craniocaudal on the left, mediolateral oblique on the right) demonstrate a 1.5-cm irregular mass (arrows) with spiculation, best seen in the mediolateral oblique view.
|
|
View larger version (82K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1d. Current screening mammogram obtained in a 72-year-old woman. The radiopaque circle in the upper inner quadrant of the left breast is a skin marker. (a) Craniocaudal and (b) mediolateral oblique views of an asymmetric density (arrow) in the upper outer quadrant of the right breast. (c) Low-resolution digitized images of both breasts with CAD markers (*). The CAD system detects the spiculated mass (ILC) in both views. An additional mark (arrow) is present in the subareolar left breast in the mediolateral oblique view only. This area was classified by the interpreting radiologist as superimposed breast tissue. (d) Digitally magnified images (craniocaudal on the left, mediolateral oblique on the right) demonstrate a 1.5-cm irregular mass (arrows) with spiculation, best seen in the mediolateral oblique view.
|
|
View larger version (95K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2a. Current screening mammogram obtained in a 74-year-old woman. (a) Craniocaudal and (b) mediolateral oblique views show a subtle 1.5-cm area of architectural distortion and asymmetric density (arrow) in the posterior central portion of the left breast, seen best in a. (c) Low-resolution digitized images with CAD markers (*). Architectural distortion is identified in both views and was histologically related to ILC. An extra mark (arrow) is seen in the lateral left breast in the craniocaudal view only (left images). No major abnormality was found by the interpreting radiologist in this area. (d) Digitally magnified images (craniocaudal on the left, mediolateral oblique on the right) confirm parenchymal distortion (arrows) at the 12-o'clock position.
|
|
View larger version (115K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2b. Current screening mammogram obtained in a 74-year-old woman. (a) Craniocaudal and (b) mediolateral oblique views show a subtle 1.5-cm area of architectural distortion and asymmetric density (arrow) in the posterior central portion of the left breast, seen best in a. (c) Low-resolution digitized images with CAD markers (*). Architectural distortion is identified in both views and was histologically related to ILC. An extra mark (arrow) is seen in the lateral left breast in the craniocaudal view only (left images). No major abnormality was found by the interpreting radiologist in this area. (d) Digitally magnified images (craniocaudal on the left, mediolateral oblique on the right) confirm parenchymal distortion (arrows) at the 12-o'clock position.
|
|
View larger version (69K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2c. Current screening mammogram obtained in a 74-year-old woman. (a) Craniocaudal and (b) mediolateral oblique views show a subtle 1.5-cm area of architectural distortion and asymmetric density (arrow) in the posterior central portion of the left breast, seen best in a. (c) Low-resolution digitized images with CAD markers (*). Architectural distortion is identified in both views and was histologically related to ILC. An extra mark (arrow) is seen in the lateral left breast in the craniocaudal view only (left images). No major abnormality was found by the interpreting radiologist in this area. (d) Digitally magnified images (craniocaudal on the left, mediolateral oblique on the right) confirm parenchymal distortion (arrows) at the 12-o'clock position.
|
|
View larger version (73K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2d. Current screening mammogram obtained in a 74-year-old woman. (a) Craniocaudal and (b) mediolateral oblique views show a subtle 1.5-cm area of architectural distortion and asymmetric density (arrow) in the posterior central portion of the left breast, seen best in a. (c) Low-resolution digitized images with CAD markers (*). Architectural distortion is identified in both views and was histologically related to ILC. An extra mark (arrow) is seen in the lateral left breast in the craniocaudal view only (left images). No major abnormality was found by the interpreting radiologist in this area. (d) Digitally magnified images (craniocaudal on the left, mediolateral oblique on the right) confirm parenchymal distortion (arrows) at the 12-o'clock position.
|
|
Ten lesions manifested as microcalcifications alone, for which the reviewing radiologists found no evidence of infiltrating carcinoma on the mammogram. Seven of these 10 lesions were associated with ductal carcinoma in situ. Two others had associated sclerosing adenosis, and the tenth had fibrocystic changes at pathologic examination that were the most likely cause of the microcalcifications. Thus, at least 11% (10 of 90) of the patients received a diagnosis of ILC not because of any mammographic evidence of invasive cancer, but because there were microcalcifications in the same area that were due to another cause. Therefore, microcalcifications were not a mammographic sign of ILC in this series.
Prior mammograms.—Of the 40 prior mammograms, nine (23%) were judged as showing no evidence of cancer, even in retrospect. The mammographic lesion types for the 31 retrospectively visible prior mammograms are shown in Table 4. Similar to findings on the current mammograms, a large proportion (45%; 14 of 31) were masses. As compared with the lesions on the current mammograms, fewer lesions, 77% (24 of 31), were judged as being visible in both views, with five lesions visible in only the craniocaudal view, and with two lesions visible in only the mediolateral oblique view.
View this table:
[in this window]
[in a new window]
|
TABLE 4. Mammographic Lesion Appearance and CAD Sensitivity for ILCs Seen Retrospectively on the Visible Prior Mammograms
|
|
CAD Evaluation
Current mammograms.—The sensitivity of the CAD algorithm for all 94 ILC lesions and the 62 lesions of pure ILC is shown in Table 2 (Fig 3). For all ILCs, the algorithm correctly marked 91% (86 of 94) of the lesions. In the 62 pure ILCs, the CAD system marked 94% (58 of 62). The average number of false-positive marks, that is, marks not pointing to an ILC lesion, for both lesion types on all mammograms was 0.77 per image.
View larger version (103K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3a. Mammograms show an ILC detected at screening in a 67-year-old woman. (a) Craniocaudal and (b) mediolateral oblique views show the malignant lesion detected by the interpreting radiologist as a 1.2-cm asymmetric density (arrow) on the upper outer quadrant of the right breast. (c) This ILC is marked by CAD (*) in both views. Also marked are areas of benign calcification ( ) in both breasts and a benign mass (arrow) in the left breast. (d) Digitally magnified views (craniocaudal on the left, mediolateral oblique on the right) demonstrate density (arrows).
|
|
View larger version (103K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3b. Mammograms show an ILC detected at screening in a 67-year-old woman. (a) Craniocaudal and (b) mediolateral oblique views show the malignant lesion detected by the interpreting radiologist as a 1.2-cm asymmetric density (arrow) on the upper outer quadrant of the right breast. (c) This ILC is marked by CAD (*) in both views. Also marked are areas of benign calcification () in both breasts and a benign mass (arrow) in the left breast. (d) Digitally magnified views (craniocaudal on the left, mediolateral oblique on the right) demonstrate density (arrows).
|
|
View larger version (69K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3c. Mammograms show an ILC detected at screening in a 67-year-old woman. (a) Craniocaudal and (b) mediolateral oblique views show the malignant lesion detected by the interpreting radiologist as a 1.2-cm asymmetric density (arrow) on the upper outer quadrant of the right breast. (c) This ILC is marked by CAD (*) in both views. Also marked are areas of benign calcification () in both breasts and a benign mass (arrow) in the left breast. (d) Digitally magnified views (craniocaudal on the left, mediolateral oblique on the right) demonstrate density (arrows).
|
|
View larger version (68K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3d. Mammograms show an ILC detected at screening in a 67-year-old woman. (a) Craniocaudal and (b) mediolateral oblique views show the malignant lesion detected by the interpreting radiologist as a 1.2-cm asymmetric density (arrow) on the upper outer quadrant of the right breast. (c) This ILC is marked by CAD (*) in both views. Also marked are areas of benign calcification () in both breasts and a benign mass (arrow) in the left breast. (d) Digitally magnified views (craniocaudal on the left, mediolateral oblique on the right) demonstrate density (arrows).
|
|
Prior mammograms.—The CAD sensitivity for the different mammographic lesion types for the 31 visible prior mammograms is shown in the last column of Table 4. The algorithm correctly marked 77% (24 of 31) of all visible prior mammograms (Fig 4). The average number of false-positive marks on the prior mammograms was 0.66 per image.
View larger version (93K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4a. Mammograms obtained in the same patient as in Figure 3 14 months before detection of an ILC. In routine (a) craniocaudal and (b) mediolateral oblique views, the lesion (arrow) in the upper outer quadrant in the right breast is present but smaller than that in the later images (Fig 3a, 3b). (c) The lesion (*) is marked by the CAD system in both views. Additional marks (arrows pointing to *) on areas judged by the interpreting radiologist as not suspicious are noted in the left breast. = benign calcifications. (d) Digitally magnified views (craniocaudal on the left, mediolateral oblique on the right) reveal possible spiculation (arrow).
|
|
View larger version (97K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4b. Mammograms obtained in the same patient as in Figure 3 14 months before detection of an ILC. In routine (a) craniocaudal and (b) mediolateral oblique views, the lesion (arrow) in the upper outer quadrant in the right breast is present but smaller than that in the later images (Fig 3a, 3b). (c) The lesion (*) is marked by the CAD system in both views. Additional marks (arrows pointing to *) on areas judged by the interpreting radiologist as not suspicious are noted in the left breast. = benign calcifications. (d) Digitally magnified views (craniocaudal on the left, mediolateral oblique on the right) reveal possible spiculation (arrow).
|
|
View larger version (57K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4c. Mammograms obtained in the same patient as in Figure 3 14 months before detection of an ILC. In routine (a) craniocaudal and (b) mediolateral oblique views, the lesion (arrow) in the upper outer quadrant in the right breast is present but smaller than that in the later images (Fig 3a, 3b). (c) The lesion (*) is marked by the CAD system in both views. Additional marks (arrows pointing to *) on areas judged by the interpreting radiologist as not suspicious are noted in the left breast. = benign calcifications. (d) Digitally magnified views (craniocaudal on the left, mediolateral oblique on the right) reveal possible spiculation (arrow).
|
|
View larger version (55K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4d. Mammograms obtained in the same patient as in Figure 3 14 months before detection of an ILC. In routine (a) craniocaudal and (b) mediolateral oblique views, the lesion (arrow) in the upper outer quadrant in the right breast is present but smaller than that in the later images (Fig 3a, 3b). (c) The lesion (*) is marked by the CAD system in both views. Additional marks (arrows pointing to *) on areas judged by the interpreting radiologist as not suspicious are noted in the left breast. = benign calcifications. (d) Digitally magnified views (craniocaudal on the left, mediolateral oblique on the right) reveal possible spiculation (arrow).
|
|
|
DISCUSSION
|
|---|
In their original article on lobular carcinoma, Foote and Stewart (10) in 1941 established the term invasive lobular carcinoma. The histologic features described were an associated desmoplastic stromal reaction, a linear arrangement of carcinoma in cells—the so-called Indian file pattern, and the tendency of cells to grow circumferentially around ducts and lobules (targetoid growth)—all were considered specific characteristics by these authors. Subsequently, Newman (11) and Richter et al (12) differentiated between pure or "classical" ILC and other lesions with cytologic features of invasive carcinoma—the so-called variant forms. These include tubular, alveolar, and solid variants. These variants have been reported to constitute 23%–30% of all ILCs (12,13). The classic lobular neoplasms demonstrate the linear growth pattern and presumably are the most difficult to detect mammographically (14).
The CAD system assists the radiologist in breast cancer detection by minimizing observational lapses and identifying areas that may warrant a second review. It does so by digitizing a mammogram and analyzing the image by means of proprietary signal processing software, which highlights areas with features associated with cancer, particularly microcalcifications and masses with spiculated characteristics. It also marks regions of interest that often represent features that are not clinically important and that are often readily recognized as such by the radiologist at repeat review of the original mammogram.
According to the BI-RADS breast density classification, 62% (56 of 90) of the ILCs were detected in fatty tissue; the remainder (38%; 34 of 90) were in dense tissue. The fact that 21% (20 of 94) of all lesions manifested with architectural distortion alone and the remainder were primarily masses or calcifications explains the detectability of these lesions in primarily fatty tissue. It is important to emphasize that our series addresses only ILCs detected at screening mammography—8.3% (90 of 1,083) of all screening-detected cancers.
In 31 (78%) of the 40 cases in the current study for which prior mammograms were available, findings were present at retrospective review, and the proportion of these lesions in fatty tissue was 64% (20 of 31 cases). This compares with the 77% detectability of these lesions with CAD (Table 4). However, as our series addressed only screening-detected cancers, we did not have an opportunity to analyze tissue density for those cases not detected at screening that proved to be interval cancers at a later date.
The mammographic appearance of ILC in our study correlates with that in previous studies (1,2,4–7). Sixty-five percent (40 of 62) of the lesions with pure ILC had a mass at presentation, with 70% (28 of 40) of these masses being spiculated. Twenty-one percent (13 of 62) manifested with architectural distortion. Although 11% (10 of 94) of cases were diagnosed because of the presence of microcalcifications, in the patients with these lesions, there were associated lesions or benign processes that likely caused the microcalcifications. This again demonstrates the rarity of microcalcifications associated with ILC to aid in its detection.
Review of the 40 available prior mammograms in those patients who eventually received a diagnosis of ILC demonstrated that 78% (31 of 40) of the lesions were indeed visible in retrospect. The mammographic appearances of these lesions were similar to those of the eventually diagnosed ILCs, with 45% (14 of 31) appearing as masses and 19% (six of 31) as architectural distortions. Previous studies (8,15–17) have shown, however, that while a majority of breast carcinomas can be seen in retrospect, only a percentage of these are considered "actionable" (ie, requiring patient callback for additional evaluation with imaging or biopsy) in a blinded review.
In our case material, ILC was apparent in both views in 96% of cases. However, others (6,7,18,19) have reported that ILC was visible (or equally visible) on both views in only 50%–70% of cases. A possible explanation for this discrepancy might be the methods of analysis (retrospective versus prospective) or how the question was posed to the readers.
The ability of a CAD system to detect biopsy-proved cancers on screening mammograms has been previously evaluated (8). In that study, the CAD system (version 1.2; R2 Technology) correctly marked 84% of 1,083 carcinomas detected on current mammograms. The percentage of masses (spiculated and nonspiculated) correctly marked was 75% (506 of 677), and the percentage of microcalcifications correctly marked was 99% (400 of 406). The CAD system correctly marked 60% (171 of 286) of all carcinomas retrospectively visible on prior mammograms. With update of these results to the software version used in the current study (version 2.2), the CAD system correctly marked 90% (979 of 1,083) of all carcinomas in the series, with 86% (580 of 677) of masses and 98% (399 of 406) of microcalcifications correctly marked (20).
In the current study, the CAD system correctly marked 91% (86 of 94) of ILCs on the screening mammograms on which the cancer was detected by the original interpreting radiologist. It detected 100% (10 of 10) of cases described as primarily microcalcifications and 95% (53 of 56) of masses with or without calcifications. Its lowest sensitivity was for asymmetric densities (75%; six of eight cases). The CAD system correctly marked 77% (24 of 31) of the lesions visible on prior mammograms. The histopathologic data in our series did not allow us to report on the detectability of the various histologic subtypes of ILC with CAD. This will be an important topic for future study of the CAD system.
It is increasingly clear that use of the CAD system increases the likelihood of detecting breast cancer on screening mammograms (8,21). There is agreement that detection of ILC on screening mammograms is particularly difficult. We found that the ability of a CAD system to detect ILC (91% success rate) is similar to that for cancers of all histologic subtypes evaluated in the entire series (8,20). The CAD system is a useful prompting device for the radiologist that can direct attention to areas of interest with features that may be associated with breast carcinoma. The radiologist, however, remains the final interpreter.
|
FOOTNOTES
|
|---|
L.J.W.B. is a shareholder of R2 Technology.
Abbreviations: BI-RADS = Breast Imaging Reporting and Data System,
CAD = computer-aided detection,
ILC = invasive lobular carcinoma
Author contributions: Guarantors of integrity of entire study, all authors; study concepts and design, W.P.E., L.J.W.B., K.F.O., R.A.C.; literature research, W.P.E., L.J.W.B., L.L., K.F.O.; clinical studies, W.P.E., L.J.W.B., K.F.O.; data acquisition, W.P.E., L.J.W.B.; data analysis/interpretation, all authors; manuscript preparation, W.P.E., L.L., L.J.W.B.; manuscript definition of intellectual content, editing, revision/review, and manuscript final version approval, all authors.
|
REFERENCES
|
|---|
- Helvie MA, Paramagul C, Oberman HA, Adler DD. Invasive lobular carcinoma: imaging features and clinical detection. Invest Radiol 1993; 28:202-207.[Medline]
- Krecke KN, Gisvold JJ. Invasive lobular carcinoma of the breast: mammographic findings and extent of disease at diagnosis in 184 patients. AJR Am J Roentgenol 1993; 161:957-960.[Abstract/Free Full Text]
- Dixon JM, Anderson TJ, Page DL, Lee D, Duffy SW. Infiltrating lobular carcinoma of the breast. Histopathology 1982; 6:149-161.[Medline]
- Sickles EA. The subtle and atypical mammographic features of invasive lobular carcinoma. Radiology 1991; 178:25-26.[Free Full Text]
- Mendelson EB, Harris KM, Doshi N, Tobon H. Infiltrating lobular carcinoma: mammographic patterns with pathologic correlation. AJR Am J Roentgenol 1989; 153:265-271.[Abstract/Free Full Text]
- Hilleren DJ, Andersson IT, Lindholm K, Linnell FS. Invasive lobular carcinoma: mammographic findings in a 10-year experience. Radiology 1991; 178:149-154.[Abstract/Free Full Text]
- Newstead GM, Baute PB, Toth HK. Invasive lobular and ductal carcinoma: mammographic findings and stage at diagnosis. Radiology 1992; 184:623-627.[Abstract/Free Full Text]
- Warren Burhenne LJ, Wood SA, D'Orsi CJ, et al. Potential contribution of computer-aided detection to the sensitivity of screening mammography. Radiology 2000; 215:554-562.[Abstract/Free Full Text]
- American College of Radiology. Illustrated Breast Imaging Reporting and Data System 3rd ed. Reston, Va: American College of Radiology, 1998; 167-175.
- Foote FW, Stewart FW. A histologic classification of carcinoma of the breast. Surgery 1941; 19:74-99.
- Newman W. Lobular carcinoma of the female breast: report of 73 cases. Ann Surg 1966; 164:305-314.[Medline]
- Richter GO, Dockerty MB, Clagett OT. Diffuse infiltrating scirrhous carcinoma of the breast: special consideration of the single-filing phenomenon. Cancer 1967; 20:363-370.[CrossRef][Medline]
- DiCostanzo D, Rosen PP, Gareen I, Franklin S, Lesser M. Prognosis in infiltrating lobular carcinoma: an analysis of "classical" and variant tumors. Am J Surg Pathol 1990; 14:12-23.[Medline]
- Rosen PP. Invasive lobular carcinoma In: Rosen's breast pathology. Philadelphia, Pa: Lippincott-Raven, 1997; 545-566.
- Harvey JA, Fajardo LL, Innis CA. Previous mammograms in patients with impalpable breast carcinoma: retrospective vs blinded interpretation. AJR Am J Roentgenol 1993; 161:1167-1172.[Abstract/Free Full Text]
- van Dijck JA, Verbeek AL, Hendriks JH, Holland R. The current detectability of breast cancer in a mammographic screening program: a review of the previous mammograms of interval and screen-detected cancers. Cancer 1993; 72:1933-1938.[CrossRef][Medline]
- Ikeda DM, Birdwell RL, O'Shaughnessy KF, Sickles EA. Analysis of 172 subtle, undetected or non-recalled findings on prior "negative" mammograms in women with screening-detected breast cancer (abstr). Radiology 1999; 213(P):240.
- Sickles EA. Findings at mammographic screening on only one standard projection: outcomes analysis. Radiology 1998; 208:471-475.[Abstract/Free Full Text]
- Cornford EJ, Wilson AR, Athanassiou E, et al. Mammographic features of invasive lobular and invasive ductal carcinoma of the breast: a comparative analysis. Br J Radiol 1995; 68:450-453.[Abstract]
- Castellino RA, Roehrig J, Zhang W. Improved computer-aided detection (CAD) algorithms for screening mammography (abstr). Radiology 2000; 217(P):400.
- 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]
This article has been cited by other articles:
|
|
|
|
 
N. Biglia, L. Mariani, L. Sgro, P. Mininanni, G. Moggio, and P. Sismondi
Increased incidence of lobular breast cancer in women treated with hormone replacement therapy: implications for diagnosis, surgical and medical treatment
Endocr. Relat. Cancer,
September 1, 2007;
14(3):
549 - 567.
[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]
|
|
|
|
|
|
|
J. A. Harvey
Unusual Breast Cancers: Useful Clues to Expanding the Differential Diagnosis
Radiology,
March 1, 2007;
242(3):
683 - 694.
[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]
|
|
|
|
|
|
|
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]
|
|
|
|
|
|
|
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. A. Baker, J. Y. Lo, D. M. Delong, and C. E. Floyd
Computer-aided Detection in Screening Mammography: Variability in Cues
Radiology,
November 1, 2004;
233(2):
411 - 417.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
J. A. Baker, E. L. Rosen, J. Y. Lo, E. I. Gimenez, R. Walsh, and M. S. Soo
Computer-Aided Detection (CAD) in Screening Mammography: Sensitivity of Commercial CAD Systems for Detecting Architectural Distortion
Am. J. Roentgenol.,
October 1, 2003;
181(4):
1083 - 1088.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
