วีดีโอ

Thank you Sam! Kudos to Renzo for making this project works.

Panagiotis. The reason, why high-resolution fMRI studies usually register the anatomy to EPI and not the other way around is to keep the functional data as unaffected as possible. every spatial alignment comes along with a spatial interpolation step, which ultimately blurs the signal. See this paper for more info (e.g. Fig. 1): dx.doi.org/10.1016/j.neuroimage.2017.04.053 This is usually not a bit issue in conventional fMRI, where the signal is blurred anyway.

Thanks for your kind words :) . Indeed, AFNI has an implementation of Weickert 1998. I think I must have mentioned it at some point or maybe in the previous lecture. One difference is that I sometimes use custom equations to tweak the diffusion behavior to fit special needs (e.g. case mentioned around 40:47, now published work: elifesciences.org/articles/56963 ).

@@ofgulban Awesome, can't wait to read your paper! Yes you mentioned it in this lecture but didn't get the name quite right, and you didn't mention it in your paper or (from what i can see) your github either. It's all good, I just see time and time again AFNI tools getting looked over (partly their fault for not formally publishing everything they do).

Salvatore Torrisi yeah, I would have definitely preferred realizing that AFNI already had an easy to use implementation a few years ago, before implementing myself :). Not enough people knew about AFNI around me unfortunately at that time. Though it is not too bad of an exercise to implement something to fully understand it compared to only being a user. Just to clarify, I did not mention it in the paper since I refer directly to the origin, Weickert’s 1998 book, which is what I have studied to implement it from scratch while tweaking some parts in the way I see fit. Nevertheless I have updated my github page now to direct some well-deserved attention to 3danisosmooth of AFNI: github.com/ofgulban/segmentator/wiki/Nonlinear-anisotropic-smoothing . Thanks for taking the time highlight this.

@@ofgulban Very kind of you to make that change to the page. Yours is much more than an "exercise" and rather a truly valuable contribution to the field! Thank you for your very high quality work.

Thanks Arman, Thanks for this excellent question. Yes, the TR and the slow HRF are something that we can improve in the future. I think one of many reasons, why the data looked best for longer TRs was that the reconstruction of GRAPPA data in two Phase-encoding directions. And how it deals with phase errors. E.g. here we see that it the specifics of the Recon algorithm matter in this protocol. See example artifacts here: twitter.com/layerfMRI/status/1243519834194882560/photo/1 It is without question, that we will miss the neurally driven fluctuations at high temporal frequencies. But after all, the HRF is slow. It cannot capture the interesting neural fluctuations in the range of alpha, betha, gamma bands. And the fMRI sensitivity to measure sub-second changes is orders of magnitude smaller than the magnitude of slower changes (Laura Lewis studies). So it remains to be seen how big of a hit in tSNR I should find acceptable, just so that I don't miss the HRF-attenuated fast fluctuations, that are very hard to see to begin with.

Hi tayebeh, Sure, the Registration.sh is part of ANTS github.com/ANTsX/ANTs. I tried to explain how to use it here: layerfmri.com/2019/02/11/high-quality-registration/ I hope that helps.

Thanks again. I want to register two structural T1-W images. I have used your script and performed exactly like what you have explained in this tutorial video. but I got an error: transform file does not exist: registered1_0GenericAffine.mat Could you please help me to fix this error?

Sorry, I don't. Maybe one of their organisers (Pierre-Francois Van de Moortele, Gregor Adriany, Xiaoping Wu, and Yigitcan Eryaman) knows, if they recorded the sessions?

Yes, Artie moved ;-) He is a son (creation) of Jacco, no? -> Cryptography has Alice and Bob -> The forensics department has Jane/John Doe -> Computer science has Foo and Bar -> CPR has rescue Annie I think MRI should have Artie Fischl ;-) much better than "test" and all the other pseudo names

Your cool

You are awesome!