[Mristudio-users] HARDI data analyze using dtistudio

[susumu mori]

> Hi , professor, I have three questions:
> (1) can dtistudio analyze high angular resolution diffusion
> imaging(hardi) data?
>

Sorry, NO.


> (2) can we process hardi data as diffsuion weighted data? Will the data
> quality be weaken if we do so?
>

Yes, you can process HARDI data just as a plane DTI data. Diffusion
"Tensor" Imaging tries to fit the 6-parameter tensor to the observation
(DWIs); effectively reducing all the information into the 6 numbers. You
theoretically need only 6 DWIs for the calculation but to enhance SNR, 10 -
50 DWIs are routinely acquired. If you acquire a HARDI data with, for
example, 128 DWIs, you can use all of them for simple DTI calculation. Then
you may think HARDI inherently include DTI data and would provide more by
using advanced non-tensor analysis.

This is, in theory, true, but in practice, there is an issue because HARDI
requires two things; more directions AND high b-values.
The former (more direction) is a matter of scanning time. 128 direction
should take about 15 min of scanning time.
The latter (higher b-value) is more problematic. Typical DTI uses b =
700-1,000. HARDI usually requires 3,000 - 30,000.
The high-b-value leads to large signal loss, leading to poorer SNR. It also
requires longer TE, which further reduces SNR. It causes more motion
artifacts and eddy current distortion, which are difficult to correct
because each raw DWI has poor SNR. As a result, HARDI tends to use lower
spatial resolution. Modern 3T scanners should provide DTI data with 2-2.2mm
resolution with decent SNR with 5 min scan time. I am not sure how long the
scan you would need if you want similar resolution and SNR using the HARDI
protocol.

Because of this issue, HARDI data is not routinely used as a substitute of
regular DTI, which can provide higher resolution and higher SNR with
shorter scan time.


> (3) What's the measurement of hardi data, FA, ADC or some others?
>
> Well , this is the most important question. It has been shown that HARDI
can retrieve more anatomical information than DTI. However, I think the
community is struggling to provide quantitative numbers out of it. These
numbers need to carry anatomical meanings which are different from what DTI
can provide. We can think of indices like the number of fiber populations,
crossing angles, FA of each population, etc, but I'm not sure how many of
them have been adopted in the community. Such indices need to be reasonably
robust against noise and hopefully not much dependent on the algorithm or
imaging parameters. At this moment, HARDI seems to be used mostly as a
prior for advanced tractography, but not for quantitative image analysis.

We have been closely monitoring the development of HARDI-type data
acquisition and analysis but haven't decided whether we should be committed
to any of the available approaches. At present, spherical deconvolution
approach by Tournier et al seems the closest for quantitative analysis and
I encourage you to read their papers as well as trying MRtrix software.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mristudio.org/pipermail/mristudio-users/attachments/20130325/afe7f9b2/attachment.html