Hi, I'm taking my first steps in DTI analysis with FSL and have been referring to this paper (http://www.nature.com/nprot/journal/v2/n3/full/nprot.2007.45.html) for guidance. It recommends checking the quality of the FA images, paying particular attention to the intensity histogram. While I did that, I found that my data look a whole lot worse than the ones displayed in Figure 1 of that paper (I'm attaching a screen shot for illustration). Even though the first tbss step removed a "halo" of hyperintense voxels around the brain (probably representing blood vessels in the dura?), there are still some random "hyperintense" voxels in certain places, and the histogram shows an upshot at 1. The image also looks generally blurrier. Part of that is probably that my voxels are larger (2.5 x 2.5 x 2.5) than the ones in the paper, but I'm worried that that may not be the only reason.
So, my questions are
1) What could be the cause of this bad image quality? The image I'm looking at is the FA_FA image in native space. The data were acquired on a Siemens Magnetom 3 T scanner with a 12-channel head coil and the following parameters: EPI images, TR = 7700 ms, TE = 100 ms, 5 b0 images in the beginning, followed by 30 b1000 images at different gradient directions, bandwidth 2264 Hz/Px, echo spacing 0.53 ms. I don't know about the SNR in the b0 image - how do I determine that? According to the printout I received from our MRI technician, fat saturation was done. The subjects were adults, and most of them did not move much, so I doubt that that is the problem.
2) Considering that the attached image is pretty representative of my entire dataset, is there something I can do to "save" the data, like removing hyperintense pixels and then rescaling? If not, is there even a point trying to work with these data, or are they too bad? And if so, what can I do better the next time I collect DTI data?
Any input you can offer would be greatly appreciated!
Thanks a lot,