A BREAST MRI BIOMARKER FOR CYSTS AND INFILTRATING DUCTAL CARCINOMAS

Parametric MR information from tissues is known to increase the specificity of segmentation resulting from contrast enhanced MR studies. Rapid signal acquisition methods, such as FSPGR, prevent the extraction of complete parametric information; this in turn limits the specificity that can be achieved from dynamic breast MR studies. We introduce a biomarker k, which combines pseudo-proton density and T₂* information. Following tissue segmentation by two experts and removal of the bias field, we compute k for each tissue type for 83 patients. A Gaussian distribution model for k for each tissue is used to detect cysts and infiltrating ductal carcinomas (IDCs) using a 95% confidence interval in new cases. The method yields encouraging results in both cases