Muqing Lin1, Jeon-Hor Chen1,2, Ke Nie1, Orhan Nalcioglu1, Min-Ying Lydia Su1
1Tu & Yuen Center for Functional Onco-Imaging, University of California, Irvine, CA, USA; 2Department of Radiology, China Medical University, Taichung, Taiwan
We presented an algorithm based method for identification of blood vessels on dynamic contrast enhanced breast MRI. The algorithm is consisted of global histogram equalization, a filter bank based on Wavelet Transform and Hessian Matrix, and high-pass filtering to enhance vessels, then thresholding. The procedure was first performed on 2D MIP then extended to 3D. The detected vessels were converted to a vasculature skeleton by thinning, and compared to radiologists drawing. The mean correct detection rate was 85%, the incorrect detection rate (detected by algorithm not by radiologist) was 15%, and the missed-rate (identified by radiologist not by algorithm) was 21%.
Keywords