CAD experts look to lesion characterization

Research conducted at the University of Chicago reported that using computer-aided detection with mathematical descriptors of lesions could provide measurable improvements in the sensitivity and specificity of lesion detection. The systems rely on margin, shape, and density to help characterize a lesion.

Such systems could report numeric estimates of the likelihood of cancer or predict its development. They could also help tie the CAD system findings to other patient data.

In one study reported at the 2002 SCAR meeting, a CAD system evaluated subjects with a genetic predisposition to breast cancer and compared them with low-risk subjects. Women at high risk tend to have dense breasts, and the parenchymal pattern was coarse and of low contrast. These findings could lead to a radiographic marker for risk assessment.