Computer-aided analysis reduces false positives in breast MR

June 26, 2007

Breast MR may be as valuable in the diagnosis of breast cancer as it is difficult to interpret. Woven into contrast-enhanced MR images of the breast are data that can help identify cancerous lesions, reducing the need for subsequent biopsies. But the skill and time needed to make sense of these data can pose a formidable challenge.

Breast MR may be as valuable in the diagnosis of breast cancer as it is difficult to interpret. Woven into contrast-enhanced MR images of the breast are data that can help identify cancerous lesions, reducing the need for subsequent biopsies. But the skill and time needed to make sense of these data can pose a formidable challenge.

Kirkland, WA-based Confirma has been working to overcome this challenge with software that aids in these evaluations. Now research conducted by the Seattle Cancer Care Alliance at the University of Washington Medical Center indicate that Confirma's CADstream 3.0 can substantially reduce false positives in breast MR, although the thresholding software missed several cancerous lesions.

Dr. Teresa Williams, a fellow in pediatric radiology, and colleagues at Children's Hospital and Regional Medical Center in Seattle retrospectively compared results from 154 suspicious breast lesions found on MR with analyses processed using computer-aided enhancement (CAE) software by Confirma. The software enhances and evaluates the visible response to contrast agents absorbed by breast tissue.

False positives dropped by 23% when the software was set to its highest enhancement level, according to the results, which will be published in the July edition of Radiology. The authors noted, however, that the software failed to confirm that three of 41 lesions examined were cancerous.

The researchers warn in their article that, despite the benefits of CAE, the software should be used only to supplement radiologists' interpretations of breast MRI scans and should not be relied upon exclusively.

Regardless of the software's less than perfect performance, the authors conclude that the information provided by CADstream 3.0 has potential benefit in distinguishing between benign and malignant lesions.

"There are challenges associated with breast MR and one is the time it takes to process and evaluate the many images acquired," said Dr. Constance Lehman, director of radiology at the Seattle Cancer Care Alliance and one of the study's authors. "Computer software programs, such as the one evaluated in our study, can assist us in interpreting breast MRI scans more easily."

The use of MR as an adjunct to mammography is standard practice at the SCCA for women at high risk for breast cancer. In women who are newly diagnosed with breast cancer, MR is also used to examine the other, or contralateral, breast.

Particularly challenging is the interpretation of morphological and kinetic features that result as MR contrast washes in and out of breast tissue over time during the multiple images obtained during exams. These exams typically require three scans - one without contrast agent and two after its administration. CADstream 3.0 compares pixel signal intensity values on the precontrast and postcontrast series. If a pixel value increases above a user-specified minimum enhancement threshold, such as by 50% or 100%, the pixel is said to meet threshold enhancement.

One of three colors - blue, green, or red - is applied to pixels based on tissue enhancement due to the absorption of contrast. This results in a color overlay map displayed on MR images, indicating regions of threshold enhancement. Areas determined by the CAE software algorithm to be "connected" are summed to constitute a lesion.

One key analysis function performed by CAE is automatic kinetic assessment. The detailed kinetic information delivered by CADstream 3.0, according to the researchers, differed substantially from data obtained by conventional manual placement of a region of interest. This provided the basis for cutting false positives by 23%.

The lesions in the study, identified and biopsied between 2001 and 2004, were found in 125 women, ages 27 to 86. CAE threshold enhancement was sensitive for malignancy in 38 of the 41 malignant lesions examined using the software, according to the study.

"Given the presence of three false-negative lesions, a finding deemed suspicious by the radiologist should be further evaluated regardless of the enhancement features determined by CAE," the authors wrote.

Williams noted, however, that the software has been shown to be useful in reducing the false-positive rate of breast MRI.