MAP-based Brain Tissue Segmentation using Manifold Learning and Hierarchical Max-Flow regularization
Imaging Laboratories, Robarts Research Institute, London, ON
Grand Challenge on MR Brain Image Segmentation workshop, 2014
@article{rajchl2014map,
title={MAP-based Brain Tissue Segmentation using Manifold Learning and Hierarchical Max-Flow regularization},
author={Rajchl, Martin and Baxter, John SH and McLeod, A Jonathan and Yuan, Jing and Qiu, Wu and Peters, Terry M and White, James A and Khan, Ali R},
year={2014}
}
We developed a fully-automatic multi-atlas initialized segmentation algorithm for tissue segmentation using multi-sequence MR images. The Generalized Hierarchical Max-Flow (HMF) [1] framework proposed in [2] is employed to regularize a maximum a-posteriori data term with a linear label-ordering constraint [3]. The data term is derived from two probabilistic cost functions, i) an intensity model from learned Gaussian Mixture Models (GMM) via Kohonnen Self-organizing maps (KSOM) and ii) a shape prior from multi-atlas labeling. These costs are combined and subsequently regularized using the GHMF framework. The algorithm is fully automated and major components of the image processing pipeline are implemented using General-Purpose Programming on Graphics Processing Units (GPGPU) to achieve a substantial increase in computation speed.
April 12, 2014 by hgpu
