Main Paths of Brain Fibers in Diffusion Images Mixed with a Noise to Improve Performance of Tractography Algorithm-Evaluation in Phantom

Abstract

Background

Some voxels may alter the tractography results due to unintentional alteration of noises and other unwanted factors.
Objective: This study aimed to investigate the effect of local phase features on tractography results providing data are mixed by a Gaussian or random distribution noise.

Material and Methods

In this simulation study, a mask was firstly designed based on the local phase features to decrease false-negative and -positive tractography results. The local phase features are calculated according to the local structures of images, which can be zero-dimensional, meaning just one point (equivalent to noise in tractography algorithm), a line (equivalent to a simple fiber), or an edge (equivalent to structures more complex than a simple fiber). A digital phantom evaluated the feasibility current model with the maximum complexities of configurations in fibers, including crossing fibers. In this paper, the diffusion images were mixed separately by a Gaussian or random distribution noise in 2 forms: a zero-mean noise and a noise with a mean of data.

Results

The local mask eliminates the pixels of unfitted values with the main structures of images, due to noise or other interferer factors.

Conclusion

The local phase features of diffusion images are an innovative solution to determine principal diffusion directions