Computer-aided Detection Improves Early Breast Cancer Detection

The study was conducted to evaluate the impact of CAD in a non-academic setting, most notably its effect on cancer detection in both screening and diagnostic patients. The positive predictive value (PPV) of biopsy recommendations, biopsy rate, and recall rate before and after the introduction of CAD were compared. Then, size, stage, and histology of cancers detected with and without CAD findings were evaluated.

"Early detection is essential in preventing breast cancer deaths, so as a mammography specialist, I'm always looking for tools to improve cancer detection," said Judy Dean, MD, and lead author of the study. "This study began as acceptance testing for the CAD system I purchased for my mammography practice in 2002. I scanned 50 known cancer cases and 50 normal mammograms to see how the software would perform on films from my own practice and was astounded to find that it marked 90% of the cancers. I was really surprised that a computer program could achieve that level of sensitivity, so I decided to see what effect it would have when put into day-to-day use in my practice," said Dr. Dean.

"We collected data until more than a hundred cancers had been found, and then analyzed the results to compare what types of cancers were found with and without CAD assistance," said Dr. Dean. "We used the CAD system for every mammogram performed during the study, not just screening patients, and even in patients with prior surgery, breast implants, or other findings, CAD was helpful in finding more cancers."

During a 28 month period, 9,520 film-screen mammograms were interpreted. According to the study screening-detected cancers increased 13.3% with CAD assistance and 9.5% in diagnostic exams. The greatest impact was on ductal carcinoma in situ (DCIS), the most common type of non-invasive breast cancer in women, for which CAD increased cancer detection by 14.2%.

"Cancer detections were higher in both screening studies and symptomatic patients. This means that CAD should be used for all mammograms, not just routine screening examinations," said Dr. Dean. "According to the ACR, there are currently 8,881 accredited mammography facilities in the United States but only 4,000 are using CAD."

"The principal barrier with CAD is the cost, and a proposed 50% reduction in Medicare reimbursement will make the technology out of reach for most small and medium sized facilities," said Dr. Dean. "These may be the very places that need CAD the most, as prior studies have shown that CAD has the most impact for non-specialist radiologists," she said.

This study appears in the July issue of the American Journal of Roentgenology.