| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Report |
1 Department of Radiology, The George Washington University, 2150 Pennsylvania
Ave. NW, Washington, DC 21117.
2 i-CAD, Inc., 2689 Commons Blvd., Beavercreek, OH.
Received June 24, 2003;
accepted after revision June 8, 2004.
Address correspondence to R. F. Brem.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CONCLUSION. The CAD system correctly marked most biopsy-proven breast cancers, with a greater sensitivity for microcalcification than for mass lesions but with no significant difference in performance based on cancer size. CAD was highly effective in detecting even the smallest lesions, with a sensitivity of 92% for lesions of 5 mm or less. CAD is a useful tool for the detection of breast cancer.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Double interpretation has been shown to increase radiologist sensitivity by 5–15% [12–15]. Although clinically effective, double interpretation is not advocated as a standard of care [16]. The use of a computer-aided system to assist a single radiologist is another strategy to further improve mammographic sensitivity.
Computer-aided detection (CAD) has the potential to significantly improve screening mammography [11, 17]. CAD can improve mammography sensitivity by marking cancerous regions that were initially missed by the radiologist [11, 17]. Consequently, it is important to assess the system's performance across variables that are correlated with the likelihood of a cancer being missed during mammographic interpretation. In addition, evaluating the system's performance across variables that affect patient prognosis is also important to determine the clinical usefulness of CAD in enhancing mammography sensitivity. Mammographic appearance and lesion size can affect the possibility that a radiologist may miss a cancer and the patient's prognosis [18, 19]. Specifically, subtle masses [6, 20, 21] and small lesions [6, 8, 20] are more apt to be missed. Similarly, for small invasive tumors, masses have a better prognosis than calcifications [22, 23], and smaller tumors have a better prognosis than larger ones [24].
In practice, CAD helps the radiologist detect potential areas of concern on screening mammograms after she or he performs an initial review of the films [11, 13, 17, 25]. The CAD system processes the films, uses propriety algorithms to detect potential areas of concern, and highlights potentially suspicious areas (e.g., masses, microcalcifications, architectural distortions, and asymmetric densities). With the information provided by the CAD system, the radiologist decides whether true areas of concern are present at the highlighted locations, and she or he retains the ability to make the final diagnosis.
The purpose of this study was to evaluate the performance of a CAD system in detecting breast cancer based on mammographic appearance (presentation of a mass or microcalcification) and cancer size.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Normal mammograms.—One hundred fifty-five screening mammograms with normal findings were randomly selected from a series of 494 consecutive normal screening mammograms. Normal mammograms were interpreted as Breast Imaging Reporting and Data System category 1 or 2 [26], and all had a 3-year follow-up showing no mammographic or clinical abnormality. All cases were retrospectively evaluated by the CAD system to determine the number of false-positive marks per image. The institutional review board approved this protocol.
Patients were also evaluated for age. The mean age of cancer patients was 60.6 years (range, 31–87 years). For 154 of the normal mammograms for which patient age was available, the mean age was 58.9 years (range, 38–80 years). Age data were not available for one woman with a normal normal mammogram.
Analysis of Mammograms by the CAD System
The Second Look CAD system (version 3.4; CADx Systems) was used in this
study. CAD sensitivity in the detection of cancer was evaluated by the number
of masses and microcalcifications identified, and by lesion size. The CAD
system consists of a film digitizer that uses 43-µm resolution with 12 bits
of gray-scale, a processing unit, and a CAD printout. The processing unit uses
proprietary artificial intelligence software, and the results of the analysis
are presented on a paper printout. The CAD printout 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, or asymmetric densities), and the rectangles mark
potential microcalcification clusters.
The precise mammographic location of the biopsy-proven cancers was correlated with the location of the CAD mark. The type of lesion marker (i.e., mass or calcification) had to correlate with the mammographic characteristic of the lesion. If the CAD system marked the correct type of lesion of the cancer in the precise mammographic location of the cancer in at least one view, the case was considered a true-positive. If both mass and microcalcification features were noted for the lesion, then either marker type was considered correct.
The sensitivity of the CAD algorithm was calculated as the number of lesions correctly marked divided by the total number of lesions.
Cancer sizes were based on pathologic size. Cancers were grouped into size intervals of 1–5 mm, 6–10 mm, 11–15 mm, 16–20 mm, and greater than 20 mm. CAD sensitivity correlated by cancer size was evaluated for all cancers, as well as for cancers that manifested mammographically as masses, as microcalcifications, and as masses with microcalcifications. Evaluation of statistically significant sensitivity differences of the CAD system for detection of cancers smaller than or equal to 1 cm and cancers greater than 1 cm was performed using a chi-square analysis.
Healthy subjects were used to determine the number of false-positive marks on the mammograms. False-positive marks were individually assessed for masses and microcalcifications.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
The size of cancers in the 201 patients in this study ranged from 0.5 to 70 mm. The smallest lesion, measured 0.5 mm pathologically although it was larger mammographically due to benign calcifications surrounding the cancer. The largest cancer was ductal carcinoma in situ, which measured 70 mm pathologically, was not palpable, and manifested mammographically as microcalcifications. Twenty-six of the cancers were 1–5 mm, 80 were 6–10 mm, 49 were 11–15 mm, 20 were 16–20 mm, and 26 were greater than 20 mm. (Table 2). The CAD system detected 24/26 of the cancers that measured 1–5 mm, 68/80 of the cancers that measured 6–10 mm, 46/49 of cancers that measured 11–15 mm, 16/20 of the cancers that measured 16–20 mm, and 24/26 of cancers that were greater than 20 mm (Table 2).
|
The prognosis of breast cancer depends on lesion size. Cancers smaller than or equal to 1 cm have a better prognosis than cancers greater than 1 cm. For this reason, we evaluated the performance of CAD for cancers of 1 cm or less and cancers greater than 1 cm using a chi-square test. No statistically significant difference was seen in the CAD sensitivity of cancers 1 cm or smaller and those greater than 1 cm (p = 0.41).
The cancers that manifested mammographically as masses were evaluated by size and by CAD sensitivity of detection. A total of 122 cancers presented mammographically as masses. Of these, 12, 54, 33, 15, and eight measured 1–5 mm, 6–10 mm, 11–15 mm, 16–20 mm, and greater than 20 mm, respectively. The CAD sensitivity for the detection of the masses was 83% for masses 1–5 mm, 81% for masses 6–10 mm, 91% for masses 11–15 mm, 80% for masses 16–20 mm, and 75% for masses greater than 20 mm (Table 3).
|
Similarly, cancers that manifested mammographically as microcalcifications were evaluated by size and by CAD sensitivity of detection. A total of 54 cancers presented mammographically as microcalcifications. Of these, 11, 19, 11, four, and nine measured 1–5 mm, 6–10 mm, 11–15 mm, 16–20 mm, and greater than 20 mm, respectively. The CAD sensitivity for cancers manifesting as microcalcifications measuring 1–5 mm, 6–10 mm, 11–15 mm, and greater than 20 mm was 100% in these four groups. CAD performance for cancers that manifested as microcalcifications measuring 16–20 mm was 75% (Table 4).
|
Finally, cancers that manifested mammographically as masses with microcalcifications were evaluated by size and by CAD sensitivity of detection. A total of 25 cancers presented mammographically as a mass with microcalcifications. Of these, three, seven, five, one, and nine measured 1–5 mm, 6–10 mm, 11–15 mm, 16–20 mm, and greater than 20 mm, respectively. The CAD sensitivity for the detection of microcalcifications was 100% for masses with microcalcifications 1–5 mm, 11–15 mm, 16–20 mm, and greater than 20 mm. The sensitivity of the CAD system for the detection of masses with microcalcifications that measured 6–10 mm was 71% (Table 5).
|
Normal Mammograms
Of the 155 normal cases evaluated with the CAD system, the false-positive
marks per image averaged 1.3, with 1.1 marks for masses and 0.2 marks for
calcifications.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Our study shows that the performance of the CAD system for cancers smaller than or equal to 5 mm is 92% and for cancers 6–10 mm is 85%. Interestingly, the performance of the CAD system for cancers 1–5 mm was greater than that for cancers measuring 6–10 mm, 16–20 mm, and greater than 20 mm. Only cancers measuring 11–15 mm had a slightly higher detection rate. No statistically significant difference was seen in CAD detection of cancer regardless of cancer size. In fact, our study shows that CAD performance does not depend on cancer size whether the cancer manifests as a mass, as microcalcifications, or as a mass with microcalcifications.
As previously reported, the performance of the CAD system was better for microcalcifications than for masses. This finding confirms multiple prior studies that have also shown this finding [11, 17, 27].
A large number of false-positive marks can significantly hinder the usefulness of CAD by distracting the interpreting radiologist. Therefore, the maximum number of marks per image is limited by the system. Our study showed 1.3 false-positive marks per image. Studies have shown that this false-positive rate is a reasonable number to not cause increased recall rates or time for interpretation [28]. This finding is further substantiated by the study of Zheng et al. [29], which showed that 2.0 false-positives per image are distracting. Although Zheng et al. did not evaluate 1.3 false-positives per image, they did show that 0.5 false-positives per image is not distracting. Newer software versions of the CAD system (version 6.0) have fewer false-positive marks per image. Recent studies at our institution show that version 6.0 has 0.5 false-positive marks per image. Future software development may result in even fewer false-positive marks per image.
The evaluation of the performance of CAD on the basis of cancer size may have prognostic implications as well. The prognosis of breast cancer worsens with increasing tumor size [24]. This study shows that CAD performs equally well, regardless of cancer size, with all mammographic manifestations of cancers (i.e., masses, microcalcifications, or masses with microcalcifications). CAD can detect the smallest cancers and it can detect larger cancers. Therefore, CAD may have an important impact not only on the reduction of the occurrence of missed cancers but also on breast cancer prognosis.
The use of CAD increases the mammographic detection of breast cancer [11, 17, 30]. The mammographic detection of small breast cancers is most challenging. Our study shows that cancer size does not affect the sensitivity of the CAD system, regardless of whether the cancer manifests as a mass, as microcalcifications, or as masses with microcalcifications. When CAD is used as a prompt for the radiologist to identify potential areas of abnormality on the mammogram, with the final decision being rendered by the radiologist, it is a useful tool for the improved detection of breast cancer, regardless of the size of the cancer.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  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]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
