Homeomorphic Brain Image Segmentation

Pierre-Louis Bazin and Dzung L. Pham

Overview

We present TOADS, a segmentation framework based on both topological and statistical atlases of brain anatomy. Our method guarantees strict topological equivalence between the segmented image and the atlas, and relies only weakly on a statistical atlas of shape. The method is applied to the segmentation of cerebral and cerebellar gray and white matter, ventricles, basal ganglia and brainstem, and is freely available along with a topological and statistical atlas of these structures.

Introduction

Atlas-based segmentation techniques are often employed to encode anatomical information for the delineation of multiple structures in magnetic resonance images of the brain. One of the primary challenges of these approaches is to efficiently model qualitative and quantitative anatomical knowledge without introducing a strong bias toward certain anatomical preferences when segmenting new images.

We explored the use of topological information as a prior and propose a segmentation framework based on both topological and statistical atlases of brain anatomy. Topology can be used to describe continuity of structures, as well as the relationships between structures, and is often a critical component in cortical surface reconstruction and deformation-based morphometry. Our method guarantees a strict topological equivalence (a homeomorphism) between the segmented image and the atlas, and relies only weakly on a statistical atlas of shape.

Tissue classification and fast marching methods are used to provide a powerful and flexible framework to handle multiple image contrasts, high levels of noise and gain field inhomogeneities, and variable anatomies. The segmentation algorithm, called TOpology-preserving Anatomy-Driven Segmentation (TOADS), has been validated on simulated and real brain image data. Our experiments demonstrate the accuracy and robustness of the method and the limited influence of the statistical atlas.

Application

We applied the algorithm to the segmentation of MR brain images in normal aging, studies of the effects of drug abuse, patients with cerebellar ataxia and even hydrocephalus. An extension of the method to handle white matter lesions has also been created to analyze images from Multiple Sclerosis patients. The brains were first stripped of extra-cranial tissues, then segmented into the structures defined in the topological and statistical atlas. Additional validation studies on computational phantoms and canonical manually segmented data sets have demonstrated the robustness and accuracy of the method.

Example of brain segmentation with TOADS

Segmentation result with TOADS (from left to right, top to bottom: original image, hard segmentation, memberships for cerebral WM, GM, sulcal CSF, ventricles, caudate, thalamus,putamen, brainstem, cerebellar WM and GM, 3D rendering).

Software

The method has been released as a Mipav plug-in in the Download section of this site. The software is multi-platform, but requires at least 1GB of memory. A 1mm resolution image is usually segmented within 30 minutes to an hour.

User interface for the TOADS plug-in.

Publications

P.-L. Bazin and D.L. Pham, "Homeomorphic Brain Image Segmentation with topological and Statistical Atlases", Medical Image Analysis special issue on MICCAI 2007, 12(5):616-625, 2008.
P.-L. Bazin and D.L. Pham, ``Topology-Preserving Tissue Classification of Magnetic Resonance Brain Images", IEEE Transactions on Medical Imaging, special issue on Computational Neuroanatomy, 2007.
P.-L. Bazin, X. Han, D. Tosun, J.L. Prince, D.L. Pham, "Cortical Reconstruction using Topology Preserving Tissue Classification," Human Brain Mapping 2006.
P.-L. Bazin and D.L. Pham, ``TOADS: Topology-preserving, anatomy-driven segmentation,'' Proceedings of ISBI 2006.
P.-L. Bazin and D.L. Pham, ``Topology preserving tissue classification with fast marching and topology templates,'' Proceedings of IPMI 2005.