Lead_Group Analysis Assistance

[PhilipM]

I’ve begun to look at my large STN-DBS dataset using Lead_Group and I am excited about the possibilities offered by your platform, although I would appreciate some guidance with analysis steps.

I am interested in finding the distance between the active contact and each subdivision of the Accolla STN, and I seem to be able to do this.

I am also interested in knowing for a given estimated volume of activation, whether that is within a particular subdivision (yes / no) and if possible what proportion of that VAT is within the subdivision. There is mention of this on a previous thread but when I try to use the “Correlation Between Regressors and Volume Intersections” button unfortunately I receive errors “Reference to non-existent field ‘ft’.” I have tried entering in separate scalar regressors to make sure it wasn’t that which was missing.

Is there a way to pull this data out directly from a .mat file – either ea_stats.mat or LEAD_groupanalysis.mat?

I’m also having some issues with visualisation. In 3D visualisation (also in Lead_DBS) not all subdivisions of the Accolla STN atlas are displayed – the motor portion on the right and the limbic / associative regions on the left. The stats all seem to calculate properly, indicating that all regions are taken into account bilaterally when measuring distance etc.

Unfortunately I’m also having issues with 2D visualisation “Reference to non-existent field ‘backdrop’.”

The visualisation issues are not key at present, however, and the visualisation on Lead_DBS works on an individual level as a ‘sanity check’

Thanks again, much appreciated.

[andreashorn]

Dear Philip,

yes, lead group is widely undocumented unfortunately, sorry about that.
Before you click “Correlation Between Regressors and Volume Intersections”, you need to calculate the DBS stats, i.e. click on the “Prepare stats” button. I hope this works, haven’t tried for some while – let me know if it doesnt.

The info should also be inside the ea_stats.mat file of each patient after that.

Regarding the Accolla atlas, it seems that you have an outdated data folder from an older version than 1.5 – you can simply delete the atlas_index.mat file and it will be regenerated based on the niftis on the next visualization call. Still, I’d suggest to update the datafiles, i.e. make a clean new download from the website.

Re the 2D viz – this seems to be a bug – I guess you could circumvent it by once clicking on the settings dialog of the 2D visualizations and saving these settings. Seems there are no default settings stored in your installation somehow.

Hope this helps, let me know if you have further questions!

[PhilipM]

Thanks Andreas,

Regenerating the Accolla atlas file worked for me.

Unfortunately I am still getting the same error with the “Correlation Between Regressors and Volume Intersections / Fibrecounts” even after adding the patients, adding stimulation intensities, setting group colours and adding regressors, then pressing “Prepare Stats”. The full error is:

Reference to non-existent field ‘ft’.

Error in lead_group>preparedataanalysis (line 1143)
for side=1:size(M.stats(pt).ea_stats.stimulation(usewhichstim).ft,2)

Error in lead_group>corrbutton_Callback (line 365)
stats=preparedataanalysis(handles);

Error in gui_mainfcn (line 95)
feval(varargin{:});

Error in lead_group (line 42)
gui_mainfcn(gui_State, varargin{:});

Error in
matlab.graphics.internal.figfile.FigFile/read>@(hObject,eventdata)lead_group(‘corrbutton_Callback’,hObject,eventdata,guidata(hObject))
Error using ea_showatlas (line 284)
Error while evaluating UIControl Callback

Processing Lead_DBS_003.

When I then try to extract the information from the .mat file (Lead_groupanalysis.mat – stats – ea_stats – stimulation – vat) there is a 2×4 structure available, but the fields AtlasIntersection and nAtlasIntersection are all [0,0,0] even though the VATs are certainly imposed upon the STN subdivisions.

Am I following the right steps?

Much appreciated.

[andreashorn]

Hi Philip,

I haven’t used the feature in a while but just tried on a small test dataset and can confirm it worked for me. I guess the problem is that you actually need to add a group connectome to the process even if you are not at all interested in connectivity. I never found the time to support the case of preparing DBS stats only for the VTA.

So please make sure that you have some valid .mat files in your lead/fibers folder. You can e.g. download the 2013 group connectome from our download page here on the site. In theory you can even install it by clicking on the group connectome button from within lead.
Then – since you’re not interested in fibers, choose the thinned out x 500 option as in my screenshot – this wont take a lot of processing time.
Make sure you have entered all stim params for each patient correctly and chosen the Mädler or Kuncel VTA model.
Then click on prep. DBS stats. It will prepare all the stats for all patients. When this is done, you should be able to access the data in the mat files as well. Or you can add a regressor into lead group and e.g. click “Correlation between Regressors and Volume Intersections / Fibercounts” in lead group directly.

I think it’s great you’re trying this, it should be done more. One day I may find some time to make lead_group a bit more robust and intuitive.. I guess it’s quite powerful with many features but that also makes it quite hard to use. Also, not all features may always be up to date for the currentmost single-subject processing features. LG used to be a personal tool for quite some time and grew with features I needed at the moment :)

Best, Andy

[andreashorn]

Hi Philip,

I think I informed you wrong and stats are not properly built. Will need to investigate a bit more on this.

Best, Andy

[andreashorn]

Hi Philip,

so at least I wrote a hotfix which should work now.
The easiest option would be to install lead dbs via github as described here:
https://leaddbs.gitbooks.io/leaddbs-manual/content/Installation.html

– then it should work. You could delete all ea_stats.mat files within your patient folders just to make sure (although I think this is not necessary).

Alternatively, we’ll probably upload a new minor release soon, too.

Best, Andy

[PhilipM]

Thanks Andy, much appreciated. I can also load the .nii volumes manually into Matlab and calculate overlapping voxels myself, which gets around the problem too.

Can I point out one other issue that got me confused for a short while?

The 2D display flips the L-R orientation such that contact 0-3 are actually on the RIGHT electrode and contacts 8-11 are on the LEFT electrode. This threw me a bit, as the Medtronic and Boston leads are generally coded the other way around. At least here in Australasia.

Cheers, Phil

[andreashorn]

Dear Philip,

thanks for letting me know. Yes, the nifti option is another one.
I think 0-3 right/left is a center-wise convention, I fear.
I heard that some centers do it reverse.
I checked whether there is an official guideline but didnt find one. If you have any official document (e.g. Medtronic/Boston specification) stating that 0-3 should be left, I’d be happy to change it (although it would be a rather major hack).

Best, Andy

[Author]

Posts