CAD Detects Breast Cancers That Are Most Challenging to Detect

A computer-aided detection system (CAD) can effectively detect breast cancers a radiologist is more apt to initially miss, a new study shows.

Small lesions are very challenging for the radiologist to detect, said Rachel F. Brem, MD, director of breast imaging at George Washington University in Washington, DC, and lead author of the study. The study of 201 women found that CAD was “highly effective in detecting even the smallest lesions, with a sensitivity of 92% for lesions of 5 mm or less,” said Dr. Brem. CAD was most effective in detecting cancers measuring 11-15 mm (94%) and least effective in detecting cancers measuring 16-20 mm (80%), Dr. Brem said. Lesion size is important; “cancers smaller than or equal to 1 cm have a better prognosis than cancers greater than 1 cm,” she said.

Subtle masses are also difficult to detect, Dr. Brem said. The study found that CAD had a sensitivity of 83% for the smallest masses (those measuring 1-5 mm). It was most sensitive for masses 11-15 mm (91%) and least sensitive for larger masses measuring greater than 20 mm (75%).

CAD systems assist the radiologist by “processing the breast films, using algorithms to detect suspicious areas and highlighting those areas,” said Dr. Brem. This study shows that CAD can help in the most difficult-to-detect cases. CAD prompts the radiologist to re-examine the films, and the radiologist decides whether “true areas of concern are present at the highlighted locations before making the final diagnosis,” she said.

The study also looked at the false-positive mark rate, i.e. the number of areas CAD highlighted that were not cancers. “A large number of false-positive marks can significantly hinder the usefulness of CAD by distracting the interpreting radiologist,” Dr. Brem said. However, “our study showed 1.3 false-positive marks per image, a reasonable number that should not affect the radiologists interpretation, nor should it mean that more women would be called back for additional mammograms,” Dr. Brem said.

CAD is “a useful tool for the improved detection of breast cancer. Based on this study, CAD could reduce the number of missed cancers, as well as positively impact breast cancer prognosis, without increasing the patient recall rate or the physician’s workload,” Dr. Brem said.

The study appears in the March 2005 issue of the American Journal of Roentgenology.