On 1 Dec 2010, at 04:08, Rajendra Morey wrote:

Hi Saad, Karla et al :

We have been acquiring DTI for over two years with 55 directions, a b-value of 1000 and TE of 76. We have data on a lot of subjects ( > 100). It takes about 11 minutes on our GE 3T Excite. We have been able analyze this data with TBSS-X to get primary and partial volume fractions for primary and crossing fibers.

However we are starting a new phase of the study that includes a major expansion of the clinical and behavioral measures. So this is a good opportunity to introduce a new DTI acquistion sequence. We will also upgrade to the GE MR750 scanner. The goals of the new sequence are to

1. Be able to analyze crossing fibers.

2. Have a sequence that is not outdated in 3-4 years hence.

3. Acquisition time under 20 minutes

3. If possible be compatible with our earlier sequence but not absolutely essential.

These goals could be met by simply adding another b-value of ~2500 in addition to the current value of 1000. The new b-value would also be with 55 directions. The main problem however here is this higher b-value would require a different TE, around 100, so that it would work for both b-values. It will not be possible to use the old TE of 76 for both b-values (1000 and 2500). Apparently this will cause SNR incompatibility between the two image sets. So the only option with two values and 55 directions is with a TE of 100 and b-values of 1000 and 2500 but sacrifices backward compatibility with the old data. This is still an option to consider.

The final idea is to keep the old sDTI equence, and add another DTI with 30 directions and two b- values with a carefully tuned TE that works for both b-values. This would be around 20 minutes and would be backward compatible. My question is would this allow nice resolution of crossing fibers with modelling of higher-order tensors. In other words with 2 b-values is the high number of directions (~60) still required or can fewer directions (~30) give comparable information about crossing fibers?

Please excuse the long post and if there is a paper or technical report that answers these questions, please refer me to it.

Cheers raj