A multiscale active contour transformation based toolbox for the extraction of white matter from brain fMRI images

TitleA multiscale active contour transformation based toolbox for the extraction of white matter from brain fMRI images
Publication TypeConference Proceedings
Year of Conference2009
AuthorsSandesh JBananki., Gupta V., Lahoty A., Srinivasa G.
Conference NameInternational Conference on Electronic Design and Signal Processing
KeywordsDepartment of Computer Science and Engineering
Abstract

In this paper, we propose a toolbox for the extraction of white matter from brain fMRI images. The underlying algorithm is based on the multiscale active contour (MSAC) transform that includes adjustable parameters. We semiautomatically tune the parameters through a supervised learning approach. To obtain a desirable segmentation with minimal human intervention and hide the complexities involved in the algorithm from our end users (biologists and medical practitioners), we design an interactive toolbox that provides a user friendly interface. The backend of this package is designed to segment the fMRI image along three views viz., axial, coronal and sagittal and perform a majority voting to classify each voxel as white or nonwhite. We find this procedure yields an area similarity of 88% with the hand-segmented ground truth and outperforms the previous approach using active contour based segmentation algorithm.