Hello! I’m working on a project analyzing local field potential data from chronically-implanted DBS electrodes. So far I’ve been rolling my own analysis but MNE seems like a great way to streamline my analysis pipeline. I’m just starting to jump in but I have a few questions that aren’t obvious from the documentation.
The data from the DBS recording system comes to me as JSON files. My first step seems to be getting this data into MNE’s Raw object, which seems straightforward to do with Numpy arrays. However, each JSON file may contain several recording sessions from an individual subject. MNE’s data model seems to assume one contiguous block of data per Raw object, is this correct? If I’m understanding correctly, I should just create several Raw objects for each recording block. This may be the easiest solution anyway, but if a single Raw object is capable of handling discontinuous recordings it’d be good to know about that option.
It looks like there is a “dbs” type to identify the channel types in the Info data structure. Am I correct in assuming this is for LFP (aka iEEG) recordings from a DBS electrode? Just confirming, since I think of “DBS” as a stimulation modality, and “LFP” as the recording modality. (As a side note, my background is in basic as opposed to clinical research, and “LFP” might be a more obvious search term to some basic scientists who are doing the same recordings but without any stimulation associated.)
For the future: is this approach of building Raw objects via intermediate Numpy arrays pretty much what would be expected for a contribution to the codebase? If I’m able to get things working well, I expect I should be able to contribute it to let others read these Medtronic JSON files as well, assuming there’s no rights/agreements issues with my employer.
The data from the DBS recording system comes to me as JSON files. My first step seems to be getting this data into MNE’s Raw object, which seems straightforward to do with Numpy arrays. However, each JSON file may contain several recording sessions from an individual subject. MNE’s data model seems to assume one contiguous block of data per Raw object, is this correct? If I’m understanding correctly, I should just create several Raw objects for each recording block. This may be the easiest solution anyway, but if a single Raw object is capable of handling discontinuous recordings it’d be good to know about that option.
you can use mne.concatenate_raws to have all in one file and MNE will insert “edge” annotations to avoid filtering for example
over the edges
It looks like there is a “dbs” type to identify the channel types in the Info data structure. Am I correct in assuming this is for LFP (aka iEEG) recordings from a DBS electrode? Just confirming, since I think of “DBS” as a stimulation modality, and “LFP” as the recording modality. (As a side note, my background is in basic as opposed to clinical research, and “LFP” might be a more obvious search term to some basic scientists who are doing the same recordings but without any stimulation associated.)
see the channel type as a way to process blocks of channels differently.
For the future: is this approach of building Raw objects via intermediate Numpy arrays pretty much what would be expected for a contribution to the codebase? If I’m able to get things working well, I expect I should be able to contribute it to let others read these Medtronic JSON files as well, assuming there’s no rights/agreements issues with my employer.
if you can share the code you have to write to solve your input problem we can see if something is worth contributing directly in MNE.