Multi-echo fMRI support in 4.8.0

87 views
Skip to first unread message

Markus Handal Sneve

unread,
Oct 28, 2024, 5:40:24 AM10/28/24
to 'Glasser, Matthew' via HCP-Users
Hi,

The 4.8.0 release announcement mentioned multi-echo support for fMRI preprocessing.

I’ve found a github commit 
but documentation seems to be lacking.

The implementation looks straight-forward, however just to double-check:

1) I need to include the new optional flag `—echoTE` which takes the echo-times used in milliseconds


2) the fMRI input timeseries file (`—fmritcs`) should be a concatenation of the individual echo timeseries

Anything else I need to consider?

Thanks a lot for your fantastic contributions!

Best,
Markus Handal Sneve
Center for Lifespan Changes in Brain and Cognition (LCBC)
Department of Psychology
University of Oslo

Glasser, Matthew

unread,
Oct 28, 2024, 7:47:06 AM10/28/24
to hcp-...@humanconnectome.org, Rosen, Burke

That is correct.  Also SBRef (--fmriscout) is a concatenation of the SBRef echos.  Perhaps Burke after returning from vacation can make these things clearer in the documentation.

 

Note that we were disappointed with our multi-echo experience.  We had hoped to find that multi echo produced better data than single echo, but this was not the case when we tested them head-to-head with a carefully matched protocol and optimized preprocessing.  The multi-echo data were not bad, they just weren’t better than matched single echo data in terms of functional CNR (at least at the 2.4mm isotropic, TR=1000ms spatial and temporal resolution target that we set).  Some of the literature out there was not entirely forthcoming about their MRI protocols and once the actual protocols were obtained it was found that certain unmatched protocol settings favored multi-echo over single echo but were not properly disclosed.  As it stands currently, they are quite a pain to acquire on Siemens XA software, so we have dropped the idea for now.


Matt.

--
You received this message because you are subscribed to the Google Groups "HCP-Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hcp-users+...@humanconnectome.org.
To view this discussion visit https://groups.google.com/a/humanconnectome.org/d/msgid/hcp-users/A35B1ED7-F9E4-45B1-8D05-7D897C656BC6%40psykologi.uio.no.

 


The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.

Andre Zamani

unread,
Oct 28, 2024, 11:34:46 AM10/28/24
to hcp-...@humanconnectome.org
Hi Matt,

This is really interesting. We have been trying to develop a ME sequence on our Philips scanner at UBC, but are finding it just simply performs worse than SE in terms of tSNR. 

Would you expand on what you mean by the unmatched protocol settings that favored ME but were not disclosed?

Andre 

Glasser, Matthew

unread,
Oct 28, 2024, 11:58:40 AM10/28/24
to hcp-...@humanconnectome.org

I would have to look again to find everything, but the big one was the reported resolution was 2.4mm isotropic, but in fact there was a reduced resolution in the phase direction (which increases SNR, but reduces actual resolution).


Matt.

Andre Zamani

unread,
Oct 28, 2024, 2:15:32 PM10/28/24
to hcp-...@humanconnectome.org
Mmm interesting. 

If you have time and energy, it would be great to see a more exhaustive list. 

Sounds like the finding is better described as "specific combination of acquisition parameters improved tSNR and FC reliability but it is not entirely clear how this happens bc we didn't systematically vary them"


Glasser, Matthew

unread,
Oct 28, 2024, 4:45:36 PM10/28/24
to hcp-...@humanconnectome.org

I can tell you what we did.  We kept everything the same except for the TE and the required acceleration (MB=4xIPAT=1 vs MB=6xIPAT=2).  We used the same acceleration scheme as in the paper.

Reply all
Reply to author
Forward
0 new messages