November 30, 2005

Computer aided evaluation may improve accuracy in breast MRI interpretation

The Seattle Cancer Care Alliance and Confirma, a Kirkland, Washington manufacturer of computer aided detection software, today announced that research being presented at the 91st Annual Radiological Society of North America (RSNA) meeting demonstrates that computer aided evaluation for breast images done by magnetic resonance imaging (MRI) may improve accuracy in diagnostic interpretation.

In an abstract entitled "Analysis of computer aided evaluation for breast MRI in discriminating benign and malignant lesions," SCCA physicians compared the accuracy of breast MRI interpretation without and with commercially available computer aided evaluation (CAE) in discriminating benign from malignant lesions. The CAE's automated assessment improved accuracy compared to MRI interpretation without computer assistance. The researchers believe that if these results are validated in larger studies with varied imaging protocols, the number of unnecessary MRI-guided biopsies may be reduced.

"Our study shows that computer aided evaluation can significantly assist radiologists in more accurately interpreting breast MRI studies," said Wendy B. DeMartini, M.D., a radiologist at the SCCA and an associate professor at the University of Washington School of Medicine.

DeMartini and colleagues used a product called CADstreamâ“ž¢ developed by Confirma that provides automated assessment of image enhancement parameters.

The doctors evaluated 154 consecutive, suspicious breast lesions detectable only on MRI (41 malignant, 113 benign). The lesions were subsequently biopsied under MRI-guidance and then evaluated with and without CADstreamâ“ž¢.

"The presence of significant enhancement was highly sensitive for predicting malignancy, with 38/41 (93%) malignant lesions demonstrating CAE-detected enhancement at both the 50% and 100% thresholds," the authors said. "The absence of CAE significant enhancement improved specificity when compared to MRI interpretation without CAE. False positive rates were reduced by 8.8% at a 50% enhancement threshold (NS), and by 23.0% at a 100% enhancement threshold (p=0.02). There were no significant differences between enhancement patterns of benign and malignant lesions, with all lesions demonstrating a wide range of signal intensity peaks and a wide range of washout, plateau, and persistent patterns of enhancement."

The authors concluded that automated assessment of the presence or absence of significant enhancement by CAE for breast MRI "demonstrates high sensitivity and improved specificity compared to MRI interpretation without computer assistance."

On the World Wide Web:

Fred Hutchinson Cancer Research Center