IPMI 2013 - LNCS 7917

Efficient 3D Multi-region Prostate MRI Segmentation Using Dual Optimization

Wu Qiu¹, Jing Yuan¹, Eranga Ukwatta^{1, 2}, Yue Sun^{1, 2}, Martin Rajchl^{1, 2}, and Aaron Fenster^{1, 2}

¹Robarts Research Institue, Western University, London, Ontario, Canada

²Department of Biomedical Engineering, Western University, London, Ontario, Canada

Abstract. Efficient and accurate extraction of the prostate, in particular its clinically meaningful sub-regions from 3D MR images, is of great interest in image-guided prostate interventions and diagnosis of prostate cancer. In this work, we propose a novel multi-region segmentation approach to simultaneously locating the boundaries of the prostate and its two major sub-regions: the central gland and the peripheral zone. The proposed method utilizes the prior knowledge of the spatial region consistency and employs a customized prostate appearance model to simultaneously segment multiple clinically meaningful regions. We solve the resulted challenging combinatorial optimization problem by means of convex relaxation, for which we introduce a novel spatially continuous flow-maximization model and demonstrate its duality to the investigated convex relaxed optimization problem with the region consistency constraint. Moreover, the proposed continuous max-flow model naturally leads to a new and efficient continuous max-flow based algorithm, which enjoys great advantages in numerics and can be readily implemented on GPUs. Experiments using 15 T2-weighted 3D prostate MR images, by inter- and intra-operator variability, demonstrate the promising performance of the proposed approach.

Keywords: 3D Prostate MRI, Zonal Segmentation, Convex Optimization

LNCS 7917, p. 304 ff.

Full article in PDF | BibTeX