Advancing DTI Tractography Algorithms based on Qualitative and Quantitative Comparison of Algorithmic Performance
Poster Presentation

Stephen Meredith
School of Electrical, Electronic and Mechanical Engineering, University College Dublin, Ireland

Steve Crettenand
School of Electrical, Electronic and Mechanical Engineering, University College Dublin, Ireland

Matthew Hoptman
Division of Clinical Research, Nathan Kline Institute for Psychiatric Research, New York, USA

Richard Reilly
School of Electrical, Electronic and Mechanical Engineering, University College Dublin, Ireland

     Abstract ID Number: 134
     Full text: Not available
     Last modified: March 18, 2006
     Presentation date: 06/19/2006 10:00 AM in Hamilton Building, Foyer
     (View Schedule)

Abstract
Diffusion Tensor Imaging (DTI) is a relatively recent and intensely researched imaging modality based on the principles of MRI. Tracking of nerve fibre pathways within the brain using DTI is known as fibre tractography. Despite the importance of tractography and the number of proposed algorithms, few comparisons of these algorithms have been reported. A platform for qualitative and quantitative comparison of tractography algorithms has been developed. The motivation for this platform arises from the need for a feedback driven approach to the development of new and improved algorithms. Such an approach is of particular relevance in the absence of an anatomical gold standard. This research focuses on the development of a Level Sets based DTI Tractography algorithm, with the definition and optimisation of propagation conditions and speed functions being driven by the outputs of comparative studies using both a publicly available synthetic dataset and real datasets. Comparison of two existing tractography algorithms, Streamlines Tracking Techniques (STT) and Tensor-Deflection (TEND), has been made using parameters such as length, average FA, minimum FA, curvature and maximum angle. Results such as the fact that STT fibres exhibit higher curvature than those from TEND are applied in developing robust Level Set based tractography algorithms.