A semi-automatic image segmentation method for extraction of brain volume from in vivo mouse head magnetic resonance imaging using Constraint Level Sets

Abstract

In vivo magnetic resonance imaging (MRI) of mouse brain has been widely used to non-invasively monitor disease progression and/or therapeutic effects in murine models of human neurodegenerative disease. Segmentation of MRI to differentiate brain from non-brain tissue (usually referred to as brain extraction) is required for many MRI data processing and analysis methods, including coregistration, statistical parametric analysis, and mapping to brain atlas and histology. This paper presents a semi-automatic brain extraction technique based on a level set method with the incorporation of user-defined constraints. The constraints are derived from the prior knowledge of brain anatomy by defining brain boundary on orthogonal planes of the MRI. Constraints are incorporated in the level set method by spatially varying the weighting factors of the internal and external forces and modifying the image gradient (edge) map. Both two-dimensional multislice and three-dimensional versions of the brain extraction technique were developed and applied to MRI data with minimal brain/non-brain contrast T 1-weighted (T 1-wt) FLASH and maximized contrast T 2-weighted (T 2-wt) RARE. Results were evaluated by calculating the overlap measure (OM) between the automatically segmented and manually traced brain volumes. Results demonstrate that this technique accurately extracts the brain volume (mean OM = 94%) and consistently outperformed the region growing method applied to the T 2-wt RARE MRI (mean OM = 81%). This method not only successfully extracts the mouse brain in low and high contrast MRI, but can also be used to segment other organs and tissues.