I've been using FLASH sequences based on the 5- and 30-degree FLASH from
MGH (http://www.nmr.mgh.harvard.edu/~andre/MGH_Morphometry_3T.pdf) with the
Siemens 32-channel head coil. Image reconstruction fails for this sequence
on a fairly regular basis, possibly due to the additional data (8 echoes x
32 channels) overwhelming the MR acquisition computer. As far as I can
tell remote reconstruction is only an option for EPI sequences.
The simplest option might be dropping a few echoes from the sequence. When
I do this the inner skull surface produced by a slightly-modified
mne_flash_bem looks comparable for 8 echoes and for 5 echoes (i.e., the
echoes at 11.85, 13,85, and 15.85 ms were dropped). I've only been able to
test this on a couple of subjects. Has anyone done something similar or
anticipate any problems with BEM construction using this approach?
This is off the topic of your question about using only 5 echoes, but I wanted to mention that we experienced precisely what you describe on our Trio. Incidentally, a spanking new Skyra can handle those 8 echoes on 32 channels just fine. Our work-around was to go with a 12-channel birdcage instead. I figured we'd get good SNR anyway, which appears to be the case.
I would verify that the BEM shell surfaces (inner/outer skull and scalp)
look decent e.g. in tkmedit. If they look good, then you should be okay. I
*think* I have used Flash 5/30 sequences in the past with fewer than 8
echoes without any problems.
Actually, in Heidelberg, we didn't have a multi-echo FLASH sequence
available to us when we first started using the MNE software, but we've
managed to do just fine using only a single echo (TE: 1.86 ms) from each of
two flip angles (5 and 30 degrees) corresponding to the left-most images of
Fig. 9 of this paper
<http://nmr.mgh.harvard.edu/~fischl/reprints/sequence_independent_segmentation_reprint.pdf>
.
Thanks all! I appreciate the feedback. A memory upgrade would be nice but
is not likely in the near future. The 5-echo FLASH sequence is working
fine for me on the Trio. The surfaces look good and are nearly identical
with either 8 or 5 echoes.