Dear Carsten,
All the voxels exhibiting pathological behaviour seem to be at the
boundary of the brain mask so, as you say, simply discarding them is
probably the easiest thing to do. To understand what is going on, it
would be useful to extract and display their time series.
I copy this email to Will so that he can add any further comments or
advice on the spatial noise prior.
Best regards,
Guillaume.
On 27/11/2018 19:20, Carsten Allefeld wrote:
> Dear Guillaume,
>
> thanks for replying!
>
>> These results were obtained with which spatial noise prior option?
>> The
>> interface lists five options: UGL, GMRF, LORETA, Tissue-type and
>> Robust.
>
> I used the default, "UGL". Which one would you recommend?
>
> Options in detail:
> fmri_est.method.Bayesian.space.volume.block_type = 'Slices';
> fmri_est.method.Bayesian.signal = 'UGL';
> fmri_est.method.Bayesian.ARP = 6;
> fmri_est.method.Bayesian.noise.UGL = 1;
> fmri_est.method.Bayesian.LogEv = 'No';
> fmri_est.method.Bayesian.anova.first = 'No';
> fmri_est.method.Bayesian.anova.second = 'No';
> fmri_est.method.Bayesian.gcon = struct('name', {}, 'convec', {});
>
>> Could you show a map of where the voxels you are concerned about are?
>
> Attached are plots of the absolute value of the autocorrelation at lags 0 to 127 in a middle slice (#13).
> Comparison with the coregistered T1 indicates that they are located mainly at the outer edge of gray matter (maybe meninges), but also frontally and posteriorly slightly into the longitudinal fissure.
>
> That suggests I should simply exclude these voxels from further analysis.
>
> Do you have a suggestion which criterion to use?
> The data-based threshold (2.02e-9) discards more than half of the brain, and any other threshold seems arbitrary.
>
> Best,
> Carsten
>
>
>>
>> Best regards,
>> Guillaume.
>>
>>
>> On 27/11/2018 17:11, Carsten Allefeld wrote:
>>> Hello all,
>>>
>>> I'm interested in getting local estimates of temporal
>>> autocorrelation in SPM, and for that purpose used Bayesian
>>> 1st-level estimation.
>>> The fMRI data I used to test that have 3 sessions of 128 scans at a
>>> TR of 2 s and 64x64x23 voxels of size 4x4x4 mm, unsmoothed, of
>>> which approximately 13,000 are within brain.
>>>
>>> I then extracted the AR coefficients (order 6) for the first
>>> session (Sess1_AR_0001.nii to Sess1_AR_0006.nii) and used the
>>> Yule–Walker equations iteratively to obtain the corresponding
>>> autocorrelation function across lags 0 to 128.
>>>
>>> The results are strange (see attached plot):
>>> – In 16 voxels the AR coefficients describe a non-stationary
>>> process. After excluding them:
>>> – At lag 127, 7906 voxels have an autocorrelation > 1e-6, 1651
>>> voxels > 1e-3, and 113 voxels > 0.1.
>>> – The largest negative autocorrelation at lag 127 is -2.02e-9. If I
>>> take that as an indicator of numerical/estimation precision, there
>>> are 8185 voxels where the autocorrelation at lag 127 is different
>>> from 0 (> +2.02e-9).
>>>
>>> This makes me suspect that the AR estimation in "Bayesian
>>> 1st-level" is not very reliable. Is there something I might have
>>> done wrong?
>>>
>>> Is there a recommended postprocessing for the AR coefficients or
>>> autocorrelation functions?
>>> I thought about tapering à la FSL, or clustering as a crude form of
>>> spatial regularization.
>>> Or should I simply exclude voxels with unbelievably long-range
>>> autocorrelation?
>>>
>>> Thank!
>>>
>>> Best,
>>> Carsten
>>>
>>
>> --
>> Guillaume Flandin, PhD
>> Wellcome Centre for Human Neuroimaging
>> UCL Queen Square Institute of Neurology
>> London WC1N 3BG
>>
--
Guillaume Flandin, PhD
Wellcome Centre for Human Neuroimaging
UCL Queen Square Institute of Neurology
London WC1N 3BG
|
