Computer-Aided Mammography Matches Accuracy of Double Reading
ABERDEEN, Scotland, Oct. 1 (MedPage Today) -- Computer-aided detection systems make single-reader mammography reading a reliable alternative to double reading, investigators here concluded.
In a large, randomized screening trial, single reading with computer-aided interpretation missed just one cancer detected by double reading, Fiona J. Gilbert, of the University of Aberdeen, and colleagues reported online in the New England Journal of Medicine.
Computer-aided single reading was associated with a slightly higher recall rate but otherwise did not differ significantly from double reading on measures of accuracy.
"The results of this study are applicable to programs in which double reading is standard practice," the authors said. "Where single reading is standard practice, computer-aided detection has the potential to improve cancer-detection rates to the level achieved by double reading."
Several studies have shown that double reading of mammograms improves the detection rate by an average of about 10% compared with a single reader.
Double reading is standard practice in a dozen European countries, the authors said, but is used in the interpretation of only 25% to 30% of all mammogram readings in the U.S.
The benefit of computer-aided detection, which employs algorithms to analyze digital mammograms, has been questioned and remains controversial, the authors continued.
So they compared computer-aided single reading with double reading in the multicenter Computer-Aided Detection Evaluation Trial II (CADET II), enrolling 31,057 women undergoing routine mammographic screening.
The patients were randomized to double reading, single reading with computer-aided detection, or both strategies at a ratio of 1:1:28.
For computer-aided detection, the software algorithm was set to operate at a sensitivity of 88% for breast masses and 95% for calcifications, settings associated with false image-marking rates of 1.5 and 1.0, respectively, per four-film examination.
Readers assigned to each strategy had no access to the outcome of the other strategy. Discordant results from double reading were arbitrated by a third reader, or in the case of one CADET II center, another pair of readers.
Double reading detected 199 of 227 (87.7%) of cancers compared with 198 of 227 (87.2%) for computer-aided single reading.
Double reading resulted in a sensitivity of 87.7%, specificity of 97.4%, and positive predictive value of 21.1%. Corresponding values for computer-aided single reading were 87.2, 96.9%, and 18%.
None of the values differed significantly.
Computer-aided single reading was associated with a recall rate of 3.9%, which differed slightly but significantly from the 3.4% rate for double reading (P<0.001).
Tumors detected by the two screening strategies did not differ significantly with respect to pathologic attributes.
The study was supported by Cancer Research UK and by the national Health Service Center Screening Program.
Several coauthors disclosed financial relationships with commercial imaging entities.
Primary source: New England Journal of Medicine Source reference: Gilbert FJ et al. "Single reading with computer-aided detection for screening mammography" N Engl J Med 2008; 359: 1675-1684.
What breastcancer.org says about this article…
Computer-Aided Mammography Matches Accuracy of Double Reading
The study reviewed here found that two different ways to read mammograms both were good at finding breast cancer:
- In computer-aided detection, one radiologist uses a computer program to help find breast cancer in a mammogram.
- In double reading, two different radiologists read the mammogram. They don't use a computer program.
More than 31,000 mammograms were evaluated in this large study. About half of the mammograms were evaluated using computer-aided detection. The other mammograms were evaluated using double reading. Both ways of reading mammograms found breast cancer about 87% of the time.
This study was done in Europe, where double reading is common. Double reading is less common in the United States. Other research has found that double reading improves breast cancer detection and reduces false positives. A false positive means a mammogram reading finds an abnormal area that looks like cancer but turns out to be normal. Besides worrying about being diagnosed with breast cancer, a false positive means more tests and follow-up doctor visits. The process can be extremely stressful and upsetting.
Computer-aided detection uses a computer program to highlight areas on the mammogram that may be abnormal. A radiologist looks at the highlighted areas and decides if more tests are needed. Some research has shown that computer-aided detection increases the risk of a false positive. Still, in this study the risk of a false positive was the same for both ways to read mammograms -- about 3.5%.
Because computer-aided detection requires only one radiologist rather than the two radiologists needed for double reading, computer-aided detection can be a less expensive way to improve mammogram accuracy.
Breast cancer screening isn't perfect, but it is the best way to find breast cancer early, when it's most treatable. As you and your doctor plan your breast cancer screening program, consider the following:
- Are any of the mammography centers you can go to better than another?
- Is digital mammography available? Research has shown that digital mammography, which uses a computer instead of x-ray film to make the breast image, can be more accurate than film-based mammography.
- Insist that your most recent mammogram is compared to your older mammograms when it's being read. This has been shown to improve the accuracy of mammogram readings.
- Ask if a second radiologist routinely reviews any suspicious mammograms before a final interpretation is made. Second readings have been shown to improve mammogram accuracy.
For more information on mammograms and other tests to detect breast cancer, visit the Breastcancer.org Screening and Testing section.
