An Analysis of Algorithms for In Vivo Fibre Tractography

Diffusion tensor imaging (DTI) is a relatively new imaging technique which is used to measure the water diffusion in the brain. Since water diffusion is influenced by the micro-structure of tissue DTI can be used to identify nerve fibre tracts in the brain. Over recent years many fibre tracking techniques have been proposed, but little work has been done to evaluate and compare these techniques. In this paper we present a framework for designing virtual nerve fibre tracts and for simulating noisy DTI data sets for evaluating and comparing nerve fibre tracking techniques. We have implemented three classes of fibre tractography algorithms, streamlines, tensorlines and tensor deflection, and we devise several error metrics for comparing them. Our analysis shows that all methods are very sensitive to noise and that streamlines fail in regions where nerve fibres cross, whereas tensorlines and tensor deflection cope well with the tested tract topologies.