Dear Pavel,
I'm not sure that I fully understand your suggestion. In particular, the relative time-shift version of the 4D image sounds, to me, like the slice-timing-corrected output.
It is possible to create a method that tries to apply a 4D interpolation scheme to data that is shifted in both space and time, and is therefore irregularly spaced in the 4D space. Maybe this is what you are suggesting. We did do initial investigations with this when we looked at incorporating slice-by-slice motion correction, but we found that the results were too heavily influenced by artefacts to be reliable in general (though they did work on very good data). Alexis Roche has recently proposed a slice-by-slice motion correction method too, although I do not know how this performs, but you might be interested to investigate this.
Another reason that we have not investigated this more thoroughly is that we do not recommend the use of slice-timing-correction, and instead recommend use of the temporal derivatives as the latter is able to adjust for both slice timing differences in the acquisition as well as timing differences due to variations in the HRF delay. Given that it is known that the HRF varies between subjects, between regions, and even in the same subject and same region under different physiological conditions (e.g. caffeine levels), then we feel that it is better to be able to correct for both timing differences at the same time, rather than leaving the HRF-induced timing differences locked into the data.
All the best,
Mark
On 4 Apr 2014, at 00:31, Pavel Hok <[log in to unmask]> wrote:
> Dear FSL experts,
>
> forgive me if the following idea is a mathematical or logical
> non-sense, but I would like to hear some expert's opinion on that and
> at least know where my thinking went wrong:
>
> Since there is some controversy about slice-timing correction, the way
> it is implemented in FEAT, I tried to figure out how can it be
> improved without introducing any complicated maths. I thought that
> instead of doing the same slice-timing correction on all volumes it
> might be more accurate to perform motion-corrected interpolation for
> each voxel independently. In the first step, a 4D volume would be
> created where relative time-shift (difference between the reference
> and actual slice acquisition time) would be stored for each voxel.
> Next, MCFLIRT will correct the head motion and resample the functional
> series, the same transform and trilinear resampling would be applied
> to the time-shift 4D volume. This will create new time-shift values
> including the partial volume effect. In the final step, interpolation
> would be done between the spatially corresponding voxels using the
> weighting information from the time-shift matrix.
>
> I think this way it is possible to interpolate values at the
> corresponding brain coordinates and at the same time account for that,
> that they actually might be acquired in different slices due to the
> head motion.
>
> Thank you for taking your time to consider this idea and/or putting my
> erroneous thinking straight.
>
> Cheers
>
> Pavel
>
> --
> -------------------------------------------------------------------------------
> -- MUDr. Pavel Hok
> -------------------------------------------------------------------------------
> -- Laboratoř funkční magnetické rezonance
> -- Neurologická klinika
> -- Lékařská fakulta
> -- Univerzita Palackého v Olomouci
> -- Fakultní nemocnice Olomouc
> -------------------------------------------------------------------------------
> -- Laboratory of functional magnetic resonance imaging
> -- Department of Neurology
> -- Faculty of Medicine and Dentistry
> -- Palacky University Olomouc
> -- University Hospital Olomouc
> -- Czech Republic
> -------------------------------------------------------------------------------
> -- I.P. Pavlova 6, 775 20 Olomouc
> -- web: fmri.upol.cz
> -- tel.: +420 588 443 418
> -- e-mail: [log in to unmask]
> -------------------------------------------------------------------------------
|