I was given an EEG data sample (normal vs mild cognitive impaired). I used EEGlab on Octave to load the files. After loading, i find that each one of the samples are measured over different time points.
That is, on doing,
size(eeg.data) returns (19, 11990) for one sample and returns (19,10212) for another sample. SO, the number of channels is the same across the board but the number of time points over which data is gathered is varying.
Is this normal in EEG data. Are'nt all samples supposed to have the same dimensions ? How to go ahead with applying any ML techniques on this data?
The number of total time points in a given EEG file may vary because whoever recorded the data may have started and stopped the acquisition manually rather than coding a specific start/stop trigger into the paradigm. This is pretty normal, but it's good to double check and make sure that files were not stopped early (e.g., check to make sure that all expected stimulus/response triggers are present). As you analyze your data, you will likely be segmenting into epochs based on the task. Your time points for the segments should all line up across participants (i.e., each segment for each participant contains the same number of total data points).
Brittany K. Taylor thanks ma'am! That was a nice explanation.
It is mentioned that the file has already been manually denoised by ICA and epochs rejected. I do not understand what this means. So, all i have are 10 .set files of MCI patients Vs. 10 .set files of normal patients, with eyes closed throughout. So, i suppose no events need to be extracted. Although, on using the Annotations API of MNI, i do find many events associated with a subject, like below:
OrderedDict([('onset', 30.598), ('duration', 7.024), ('description', 'boundary'), ('orig_time', None)]) OrderedDict([('onset', 43.71), ('duration', 8.0), ('description', 'boundary'), ('orig_time', None)]) OrderedDict([('onset', 113.082), ('duration', 1.216), ('description', 'boundary'), ('orig_time', None)]) OrderedDict([('onset', 154.286), ('duration', 3.604), ('description', 'boundary'), ('orig_time', None)]) OrderedDict([('onset', 178.782), ('duration', 3.152), ('description', 'boundary'),('orig_time', None)])
So, i am told i can simply analyse this data (extract features) directly. I'd really like some help with going about this feature extraction (preferably in python as i am using MNE). Should i extract these above events or simply consider this data as a whole?
- Badges
- Science topic
- Similar topics
- Filing