Format

Send to

Choose Destination
Acta Neurochir Suppl. 2006;98:33-41.

Uncertainty in diffusion tensor based fibre tracking.

Author information

1
MeVis, Center for Medical Diagnostic Systems and Visualization, Bremen, Germany. hahn@mevis.de

Abstract

BACKGROUND:

Diffusion tensor imaging and related fibre tracking techniques have the potential to identify major white matter tracts afflicted by an individual pathology or tracts at risk for a given surgical approach. However, the reliability of these techniques is known to be limited by image distortions, image noise, low spatial resolution, and the problem of identifying crossing fibres. This paper intends to bridge the gap between the requirements of neurosurgical applications and basic research on fibre tracking uncertainty.

METHOD:

We acquired echo planar diffusion tensor data from both 1.5 T and 3.0 T scanners. For fibre tracking, an extended deflection-based algorithm is employed with enhanced robustness to impaired fibre integrity such as caused by diffuse or infiltrating pathological processes. Moreover, we present a method to assess and visualize the uncertainty of fibre reconstructions based on variational complex Gaussian noise, which provides an alternative to the bootstrap method. We compare fibre tracking results with and without variational noise as well as with artificially decreased image resolution and signal-to-noise.

FINDINGS:

Using our fibre tracking technique, we found a high robustness to decreased image resolution and signal-to-noise. Still, the effects of image quality on the tracking result will depend on the employed fibre tracking algorithm and must be handled with care, especially when being used for neurosurgical planning or resection guidance. An advantage of the variational noise approach over the bootstrap technique is that it is applicable to any given set of diffusion tensor images.

CONCLUSIONS:

We conclude that the presented approach allows for investigating the uncertainty of diffusion tensor imaging based fibre tracking and might offer a perspective to overcome the problem of size underestimation observed by existing techniques.

PMID:
17009699
[Indexed for MEDLINE]

Supplemental Content

Loading ...
Support Center