Dear All:

Some of my FA maps show a lot of randomly distributed high intensity voxels. Please see attached FA map for example. My first impression is that these high intensity voxels indicate high level noises. Recently, I used the Local Noise Estimate Tool developed by Dr. Landman to measure the noise level of each DTI scan. Most subjects showing this type of "noisy" FA map also showed high level noise, but some of them also showed noise level lower than subjects showing good FA maps. Could any one comments on this inconsistent finding? Are these high intensity voxels caused by high level noise or any artifacts?

I noticed that the "Noise Field Estimation" tool developed by Dr. Landman did not ask for bval and bvec data. Am I right?

Those are definitely issues with negative eigenvalues in your tensor estimates. From the relatively incoherent pattern (not localized), I'd suspect its more likely due to noise than specific artifacts.

This paper has a nice discussion of the impact of SNR on tensor estimation:
http://www3.interscience.wiley.com/journ...

-Bennett