Normative connectomes and cortical lesion maps (stroke)

[ljh]

Hi Andreas and co.,
I’d like to use the ‘normative connectomes’ feature of LEAD DBS in a stroke patient cohort. I’m hoping to i) draw tracts from the lesion site (a 3d nifti file) and ii) create a “disconnectivity“ matrix from existing parcellations.
I have successfully downloaded lead-dbs and the normative connectomes I plan to use (Horn, 2013).
Using Lead Connectome Mapper, I do the following:
 No patient directory chosen
 1. Seed: a 3d lesion map (where 1 denotes the seed region)
 2. Include structural connectivity: select Horn 2013 connectome – export in 2mm.
 3. Connectivity map from seed
 Run!

I get an error that suggests I haven’t placed an anatomy file in the subject folder (error pasted below). However, I haven’t actually specified a patient folder, and I don’t have individual level anatomical data – just lesion maps that are already normalized etc., Any suggestions on the best way forward here?

Best,
Luke

*************
Running structural connectivity…
Command: seed
Iterating voxels (1/1): 100%
Iterating fibers (1/1): 100%
Done.
Warning: No anatomy information found! Please put either
anat_t1.nii, anat_t2.nii or anat_pd.nii into subject folder.

[andreashorn]

Hi Luke,

as a general rule of thumb, always ignore warnings in lead-dbs (this is just a warning, not an error).
The result files should already be there.

Best, Andy

[ljh]

Hi Andy,
Thank you for the reply – you are indeed correct! The output looks fine.

As a follow up question – I want to investigate the structural connectivity between a seed roi (a lesion) and a common parcellation (rather than at the voxel level). To do this I have entered a the lesion as a seed roi amongst 246 other parcellation seeds and selected the ‘connectivity matrix’ option. This produces a 247 x 247 matrix which I assume denotes the number of streamlines between each of the regions I have entered into the seeds (246 parcellation rois and one lesion) – is that correct? (and if so – why does the diagonal contain values?)

Thanks again for the help – lead dbs is a fantastic tool!

Best,
Luke

[andreashorn]

Hi, correct – the diagonals are just streamlines going through each region alone. Can ignore it or use it to normalize connections or similar.

[ljh]

Thanks again Andy – I had another question (sorry!). Is the above approach what you would suggest for creating a normative connectome in common parcellations (i.e., loading each ROI of a parcellation separately as a seed and then using the ‘connectivity matrix’ option)?

While this works well in the smaller datasets, it seems to use a lot of memory in the NKI 169 atlas. For example, in a 512 parcellation of the cortex is uses over 128 GB of memory which is quite problematic. I thought perhaps there might be a more efficient option, or some extra pre-processing I could do to the dataset to reduce the memory load (although I could not find any documentation).

Thanks again for the help, I promise to stop bothering you soon!

[andreashorn]

Hi, what you could try is to make sure that all your seeds are really binary (i.e. only 0 and 1 in the images). Then it will be much faster and should also not consume as much memory. The lead mapper tool can use weighted seeds and these will be processed more elaborately given tracks need to be weighted for each voxel.
So if you break down a parcellation into single files, make sure to use nearest neighbor interpolation or do it manually making sure your nifti file only contains booleans (e.g. no values like 0.9999 or 0.000001 at bordering regions).

Other than that no clue how to fix it easily.

[area51]

Hi. Totally new to this but interested in normative connectomics. Is there an online tutorial on how to do this in Lead? I have some limbic DBS subjects that do not have pre-operative DTI. I would like to run connectomics on the VAT from the lead-dbs electrode contact reconstructions.

Thanks,
A51

[andreashorn]

Hi A51,

good name there!
Unfortunately I fear the documentation is scarce and we currently lack the resources to improve them.
I guess you already checked the manual? Then the walkthrough-video on the website is probably the best asset.
After you’ve been through that, I’d advise to join our slack channel and ask specific questions there.
Best to clarify what’s the state of your analysis (e.g. electrodes reconstructed, VTAs calculated?) and more importantly what exactly you wish to do. Running connectomics is a bit unspecific, I assume you first want to create connectivity maps from your VTAs? But if so, what’s the next step? I.e. is there a behavioral or clinical parameter that you want to address by means of connectivity to the VTA?
Would definitely need to have a pretty precise plan of what you want to answer to really help. Reason is that there are basically 4 regions in lead dbs where one can do one or the other (running seeds or connectivity matrices -> lead mapper; visualizing and analyzing single patient connectivity -> “convis” module directly in the lead dbs 3D viewer; running parcellation based connectivity stats from your vtas -> lead group; analyzing patient specific connectivity data -> lead connectome; group-wise analysis of such -> lead group connectome). I assume the latter two do not count here since you lack patient DTI. But just wanted to make the point that there are various tools for running connectomics in the lead suite..

Hope this helps!

Best, Andy

[andreashorn]

…also note that most of the tools are heavily under development still and should be seen as “beta” state. IMHO the field of dbs connectomics is still developing and we predominantly put in code while we need it ourselves for specific projects. So bear with us if not everything works as intended..

[area51]

Thanks Andy. To answer you’re questions. I would like to look at the individual patient connectivity from the VTA (electrodes and VTA already constructed with LeadDBS). We have behavioral stimulation mapping data where a behavior occurs only at certain voltages and we would like to map behavior to tracts stimulated.

-A51

[andreashorn]

Hi there,

then easiest is to:
– process these subjects in lead connectome. No new normalization needed but just run “fibertracking” and normalize fibers
– then load the patients up in lead mapper -> select their stimulations as seed and patient specific connectivity in the dMRI source. Then run seed and you’ll get seedmaps from your VTA to other brain regions.
– After that, can use the scripts ea_Amap, ea_Rmap or ea_Cmap (see Horn 2017 AoN for details) to create maps that denote your behavioral changes as a function of connectivity across patients.

Best, Andy

[Author]

Posts