LEVEL SET BRIAN SEGMENTATION WITH AGENT CLUSTERING FOR INITIALISATION - Fast Level Set Based MRI Tissue Segmentation with Termite-Like Agent Clustering for Parameter Initialization
David Feltell, Li Bai · 2008
This paper presents a novel 3D brain segmentation method based on level sets and bio-inspired methodologies. Level set segmentation methods, although highly promising, require manual selection of seed positions and thereshold parameters, along with manual reinitialisation to a new level set surface for each candidate region. Here, the use of swarm intelligent mechanisms is used to provide all the statistical data and sample points required, allowing automatic initialisation of multiple level set solvers. This is shown by segmentation of white matter, grey matter and cerebro-spinal fluid in a simulated T1 MRI scan, followed by direct comparison between a commercial level application FMRIB’s FAST and the ground truth anatomical model.