ABSTRACT
This chapter presents a noninvasive approach for early diagnosis of prostate cancer from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). In order to precisely analyze the complex 3D DCE-MRI of the prostate, a novel processing framework that consists of four basic steps is proposed. The first step isolates the prostate region from the surrounding anatomical tissue based on a maximum a posteriori (MAP) estimate of a log-likelihood function that accounts for three image descriptors: the shape priori of the prostate, the spatial interaction between the prostate tissue, and the current appearance of the prostate tissue and its background. In the second step, a nonrigid registration approach based on the solution of the Laplace equation is employed to account for any local deformations that could occur in the prostate during the scanning process due to patient breathing and local motion. In the third step, the contrast agent kinetics are obtained from the segmented prostate of the whole image sequence of the patient. Then, two perfusion-related features are collected from these curves and a kn -nearest neighbor classifier is used to distinguish between malignant and benign detected tumors. Finally, parametric perfusion maps that illustrate the propagation of the contrast agent (CA) into the prostate tissues are constructed. This is achieved based on the analysis of the 3D spatial interaction of the change of the gray-level values of prostate voxels using a generalized Gauss–Markov random field (GGMRF) image model. Moreover, the tumor boundaries are determined using a level-set deformable model controlled by the perfusion information and the spatial interaction between the prostate voxels. Experimental results on 30 clinical prostate DCE-MRI datasets yielded promising diagnostic results.
