I have been using conn batch scripting to pre-process and denoise data for a few cohorts now, and everything is working perfectly. 
Recently I've been working on a new cohort and for some reason the denoising step creates botched brain images with diagonal lines.
Lately I've realized that the slices of each image are mixed- in each slice I can see two or more slices intertwined with diagonal lines between them. Do you have any idea why this phenomenon is happening?
In addition, this phenomenon only happens in the lower slices of the brain- maybe this is a clue for what is going on?
An example is attached

Thank you!

Inbar

Hi Inbar,

What you describe sounds to me like either the effect of subject-motion (e.g. intra-scan motion during an interleaved acquisition sequence) or the effect of image reslicing when the interslice distance or voxel-size is large (which will occur, depending on your preprocesing steps, during realignment and/or during normalization), but not something that could be expected from the denoising step. Could you perhaps double-check, tracing back your images until their original un-preprocessed step, to see where those effects might first appear?

Best
Alfonso
Originally posted by inbar_meni:

Alfonso thank you for your answer!

All the subjects have these artifact but only for the denoised output. The first preprocessing stage results with an image with a 'u' prefix for realignment and unwarping and it is perfectly fine. I tries using the gui again, to try and control the different denoising stages but nothing helps. I even cancelled all of them- removed detrending, changed the filter to 0-Inf and cancelled all the regression parameters (as attached below), and still the output is botched. Any ideas as to what happened?

Thank you,
Inbar

I can see that the bounding box of my 'du' output is different from my 'u' output beforehand. Is it possible that the reslicing is the source of this artifact?
Originally posted by Alfonso Nieto-Castanon:

Hi Inbar,

Yes, please check in Setup.Options the 'analysis.space' option, that defines the space where the data will be resampled to (typically that space is MNI with 2mm voxels and a standard bounding box, which is the same as the space that the default preprocessing pipeline will have put your functional and anatomical data into, but in your case these spaces may be different and that may be causing this additional/problematic resampling step?)

Best
Alfonso
Originally posted by inbar_meni:

Amazing thank you! It worked!!!

I was looking for the way to incorporate it in my matlab batch script and I've fount it here: BATCH.Setup.voxelresolution in the conn batch manual.
attaching the full directions below in case someone else is interested:

Description: Specifies the space of voxel-level analyses (e.g. volume- or surface- based
analyses). For volume-based analyses this option also specifies the
resolution/registration/bounding-box of voxel-based analyses. Set to 1 for volume-based analyses
using a fixed template resolution (by default 2mm isotropic voxels or same as
BATCH.Setup.voxelmaskfile if specified); 2 for volume-based analyses using the same
resolution/registration as the structural volumes; 3 for volume-based analyses using the same
resolution/registration as the functional volumes; and 4 for surface-based analyses.
Example:
>> clear BATCH;
>> BATCH.Setup.voxelresolution=1;
>> conn_batch(BATCH);