Sang Ho Lee1,2, Jong Hyo Kim2,3, In Chan Song2,3, Yun Sub Jung1,2, Jeong Seon Park4, Woo Kyung Moon3
1Interdisciplinary Program in Radiation Applied Life Science, Seoul National University College of Medicine, Seoul, Korea; 2Institute of Radiation Medicine, Seoul National University Medical Research Center, Seoul, Korea; 3Department of Radiology, Seoul National University College of Medicine, Seoul, Korea; 4Department of Radiology, Hanyang University College of Medicine, Seoul, Korea
This study demonstrates the importance of quantitatively capturing spatio-temporal properties of intratumoral enhancement patterns for MR-based breast tumor diagnosis. MR-time-series images were registered for motion compensation and tumor regions were segmented semi-automatically from our proposed perfusion index map enhancing tumor contrast. Eigenvalues were obtained for voxel-wise temporal enhancement curves within tumor by using singular value decomposition (SVD), generating eigenvalue maps. The spatial variations of eigenvalues within tumor were captured by 3D geometric moment invariants (GMIs). The potential of our SVD-based GMI features in differentiation of benign and malignant tumors is validated by classification performance using least square support vector machine.