Fang Liu1,2,
Anat Kornecki3, Olga Shmuilovich3, Neil Gelman1,2
1Lawson Imaging, Lawson Health Research
Institute, London, Ontario, Canada; 2Department of Medical
Biophysics, University of Western Ontario, London, Ontario, Canada; 3Department
of Diagnostic Imaging, St. Joseph's Health Center, London, Ontario, Canada
Most
previous quantitative studies using dynamic contrast enhanced MRI of the
breast have analyzed either signal time evolution features or morphological
features as predictors of malignancy. However, combined use of both types of
features should provide stronger predictors. In this study, a time evolution
feature (time to peak (TTP)) and a 3D morphological feature (spherical shape
index (SSI)), both obtained from tumor regions segmented automatically with
K-means clustering, were investigated as independent and combined predictors
of malignancy. Results suggest that the combination provides stronger
discrimination of malignant versus benign lesions than either feature alone.