Computer-Aided Detection with Full-Field Digital Mammography

/Home/c. Diagnosing Breast Cancer/Breast Cancer Screening/Breast Cancer Screening - Mammograms/Full-field digital mammograms

Computer-Aided Detection with Full-Field Digital Mammography

Assessment Program
Volume 21, No. 3
May 2006

Executive Summary

Background
Computer-aided detection (CAD) with full-field digital mammography (FFDM) is proposed as an adjunct to radiologists' reading of digital mammograms. In the fall of 2005, the results of the American College of Radiology Imaging Network's (ACRIN) Digital Mammographic Imaging Screening Trial (DMIST) were released. The trial showed with reasonable certainty that there was no difference in accuracy between FFDM and screen-film mammography (SFM) for asymptomatic women in general. For three subgroups of women, however, FFDM performed better than SFM: women under age 50; pre- or perimenopausal women; and women with heterogeneously dense or extremely dense breasts.

An Assessment on CAD in SFM (Vol. 17, No. 17; December 2002) concluded that CAD improves the accuracy of SFM. Specifically, it improves net health outcome compared with single-reader radiologist interpretation by increasing the true-positive rate without a disproportionate increase in the false-positive rate. The Assessment also concluded that there was insufficient evidence on the use of CAD with FFDM because of the lack of conclusive evidence at the time on the benefits of FFDM in general, as well as the fact that the use of CAD with FFDM was sufficiently different from its use with SFM that evidence on the latter cannot be extrapolated to the former. Now that the performance of FFDM itself has been demonstrated, the question is whether the use of CAD can improve the sensitivity and specificity of digital mammography, which continues to be lower than optimal.

While there are conceptual similarities between the application of CAD to a digitized screen-film mammogram and to a directly acquired digital mammogram, there are some important differences. On the one hand, screen-film mammograms must be digitized before the CAD algorithms are applied. This process can introduce artifacts as a result, (e.g., of dust particles on the film). Digital mammograms are captured directly, and therefore one might expect them to be more accurate. On the other hand, there are more data in a raw digital mammogram than can be shown in a single display format. This difference may permit interaction between the CAD software and the digital mammography data being displayed. Whether this flexibility provides similar, improved, or worsened diagnostic performance will depend on how these interactions are optimized in commercially available products.

Objective
To evaluate the impact of using CAD vs. single reading on the sensitivity, specificity, and biopsy rates for FFDM.

Search Strategy
A literature search was conducted on MEDLINE covering the period December 2003 through March 2006, using the following search terms: "(mammography(mh) OR mammogra*) AND (radiographic image interpretation, computer-assisted(mh) OR diagnosis, computer-assisted(mh) OR computer-assisted OR computer-aided OR CAD," limited to English language. The search yielded 400 references, which were searched for studies on the impact on sensitivity, specificity, and/or biopsy rates of the use of CAD with digital mammography.

Selection Criteria
Articles had to compare the results of single reading of full-field digital mammography images with and without a subsequent review of the results of the CAD program. Of particular interest were articles that looked at the impact of CAD on the sensitivity, specificity, and biopsy rate associated with FFDM. Additionally, studies had to report on cancer detection based on histologic results for at least some of the patients in the sample.

Main Results
Unfortunately, no high-quality articles in peer-reviewed journals assessed the use of CAD as an adjunct to FFDM. Perhaps the strongest study was reported in an abstract from 2005, but there was not sufficient detail to assess the validity and generalizability of the results. For example, it is not clear how the sample of patients was amassed (and therefore how representative it is of the populations of interest) and whether the study was prospective or retrospective. A second study looked at the use of a commercial CAD system, but it did not report the incremental effect of CAD in addition to the initial reading by a radiologist. Several articles reported on the development of new CAD systems, but did not provide solid performance data. In addition, the primary focus of this Assessment is on commercially available CAD systems. A clinical trial on the use of CAD is just beginning in Taiwan, with plans to enroll 3,000 patients, although a focus on one ethnic group might limit its generalizability. It is not possible at this time to judge the impact of CAD on single reading of digital mammograms at this time. Therefore, the impact of CAD on cancer detection, treatment, and survival is unknown as well.

Author's Conclusions and Comments
While evidence shows that the use of CAD with screen-film mammography is equal to or better than single reading of the SFM images, there is scant information on the performance of CAD with FFDM. Logically, it might seem that CAD should play the same role with FFDM as with SFM, but the differences between film and digital mammography—which show up in the greater accuracy of FFDM in certain populations—preclude extrapolating from the impact of CAD with SFM to CAD with FFDM. The large increase in the magnitude of the data collected by FFDM, the ability to fine-tune the digitally acquired images, and the elimination of the digitization step make FFDM sufficiently different from SFM that separate studies on the impact of CAD on FFDM are needed. Until results from better studies focusing on the use of CAD with FFDM become available, the benefits of CAD with FFDM cannot be determined.

Based on the available evidence, the Blue Cross and Blue Shield Association Medical Advisory Panel made the following judgments about whether the use of computer-aided detection (CAD) as an adjunct to single reading of full-field digital mammography meets the Blue Cross and Blue Shield Association's Technology Evaluation Center (TEC) criteria.

1. The technology must have final approval from the appropriate governmental regulatory bodies.

Two CAD systems have received premarket application (PMA) approval by the U.S. Food and Drug Administration (FDA) for use with FFDM.

2. The scientific evidence must permit conclusions concerning the effect of the technology on health outcomes.

While there is a growing literature on the use of CAD with screen film mammography, there are very few articles that look at CAD with FFDM. Because the FDA approved the CAD devices for use with FFDM as a modification of their original approval of the CAD systems for use with SFM, whatever data were used to support the request for pre-market approval of these CAD systems with FFDM are not readily available. The articles on the use of CAD with FFDM are inadequate to determine the incremental impact on sensitivity, specificity, and biopsy rates.

In summary, the available evidence is considered insufficient to permit conclusions on the effect on relevant outcomes of using CAD after initial radiographic interpretation as a quality adjunct to single-reader mammography in patients having full-field digital mammography for screening or diagnostic purposes.

3. The technology must improve the net health outcome; and
4. The technology must be as beneficial as any established alternatives.

The available data from well-conducted studies are insufficient to make a determination of whether adding CAD to FFDM leads to diagnoses that are as or more accurate than reading the FFDM images alone. Given the lack of data on these intermediate outcomes, it also is not possible to determine the impact of CAD on health outcomes such as treatment success, recurrence rates, and survival. As a result, it is unclear whether health outcomes are the same or better with the use of CAD with FFDM compared to single reading of FFDM alone.

5. The improvement must be attainable outside the investigational settings.

Given the inability to determine whether the use of CAD improves health outcomes, considering the generalizability of the impact of CAD is premature.

For the above reasons, the use of CAD systems as an adjunct to single reading of full-field digital mammography images does not meet the TEC criteria.

© 2000-2007 BlueCross BlueShield Association - All Rights Reserved.