aestrivex
commented
8 years ago That approach is rather different from what I've done so far. Our algorithm tries to extract all the electrode locations from a CT along with possibly some noise. Then it will try to fit them to a plane with local curvature. It doesn't need to find all the electrodes this way, but it expects to find most of them (at least 75%).
I've never worked with postoperative MR scans and I'm not sure how hard it is or if the same approach would work. Maybe you could provide some for testing.
Certainly it's feasible and useful to build support for this as a separate module than processing CTs. parts of gselu/ielu will be the same, the labeling, visualization and editing would not change. On Feb 1, 2016 10:19 AM, "Jean-Rémi KING" notifications@github.com wrote:
Hi @aestrivex https://github.com/aestrivex, cc @choldgraf https://github.com/choldgraf
Sorry if this is not the best place to discuss this
I am trying to find a robust solution https://githubcom/mne-tools/mne-python/issues/2829#issuecomment-178001549 to detect the location of intracranial electrodes from MRI and not CT As you know, this kind of detection is much more difficult than with CT because it relies on a noisy MRI dropout
From what I see, your current API makes use of an automatic electrode localization
Would it be conceivable to build a GUI with an MRI viewer so that :
- the user indicates he/she'll add a new grid n x m electrodes and indicate the fixed distance between neighboring electrodes
- the user then manually click on an 3-view MRI on the location of some of the electrodes that are clearly identifiable
- an algorithm fits a linear (for rigid depth electrodes) or a polynomial (for surface ECoG) functions, and predicts the location of the remaining electrodes
- the users confirm or adjust all electrodes locations ?
This process could also make use of small tricks: eg plot both the coregistered pre-op and post-op MRI, as well as their differences so as to highlight the electrodes locations, incorporate the location of the cortical surfaces in the fitting to bias the prediction towards particular locations etc
There already are some interactive MRI viewers here http://nipy/nibabel#251 and there https://githubcom/AlexandreAbraham/pynax, so such enhancement would mainly require work for the online fitting and prediction of the electrodes given specific dimensions and partial locations
Of course, I'd be happy to help, but I'll need a bit of direction to interact with your code
— Reply to this email directly or view it on GitHub https://github.com/aestrivex/gselu/issues/8.
kingjr
Hi @aestrivex, cc @choldgraf
Sorry if this is not the best place to discuss this.
I am trying to find a robust solution to detect the location of intracranial electrodes from MRI and not CT. As you know, this kind of detection is much more difficult than with CT because it relies on a noisy MRI dropout.
From what I see, your current API makes use of an automatic electrode localization.
Would it be conceivable to build a GUI with an MRI viewer so that :
This process could also make use of small tricks: e.g plot both the coregistered pre-op and post-op MRI, as well as their differences so as to highlight the electrodes locations, incorporate the location of the cortical surfaces in the fitting to bias the prediction towards particular locations etc.
There already are some interactive MRI viewers here and there, so such enhancement would mainly require work for the online fitting and prediction of the electrodes given specific dimensions and partial locations.
Of course, I'd be happy to help, but I'll need a bit of direction to interact with your code.