An 14 channel EEG system is used to collect the data for clinical investigation. The spatial resolution of this EEG data must be improved for better clinical diagnosis. In the mean time, is there any possibility to improve the spatial resolution of this data?
Hi, simply perform Laplacian transforms: it acts as a very efficient spatial filter.
Hi, The longer the recording is (the more event of the same type) the better source localization would work. The problem is that you have not just the signal but a huge amount of 'noise', if you can get rid of that than you will have better estimations, and possibly interpolation would give you also better results. Hope this helps a little,
Hi, simply perform Laplacian transforms: it acts as a very efficient spatial filter.
Another way is to exactly localize the position of each electrode on the head (instead of assuming them to be where they are supposed to be) and use this in your analyses. This can be further improved by combining it with a superb material model of conductivity of each individuals head (from PET or MRI scans) that lets you know where bone and skin etc. are.
The general way to get more info from poor data - to use prior knowledge ... In you case - the knowledge about the heart and its e. activity. If you take an advantage of use one of the many known models, particularly any parametric model, you will reduce the problem to the few parameters estimation task, in simplest case to the task of source localization. I have heard that a model of a number of coupled nonlinear dipoles (representting heart departments i.e. EEG fragments) with such parameters as their localization, axis orientation. frequency, coupling coefficients etc gives a good results...
A subtle but important point: The spatial *resolution* of the data is determined entirely by the number of electrodes you have. More electrodes = higher resolution. You mean ways to increase the spatial *precision* of the data. For >50 electrodes, the Laplacian is a good method to increase spatial precision. For 14 electrodes, there is very sparse spatial sampling to begin with. Technically a Laplacian is possible, but it's unlikely that you'd get any additional information out of the results. The Laplacian is a spatial band-pass filter, but with 14 electrodes the spatial Nyquist is already so low that you won't really effectively filter out the low frequencies.
Thus, I'd say if possible start using more electrodes. Otherwise, it's best to work with what you have and try not to focus too much on questions for which spatial precision is important.
We already performed different types of spatial filtering (including Surface Laplacian) to smooth the preprocessed output to enhance the spatial activity of EEG signals. But, this method does not works well or give higher classification rate when we have lesser number of EEG channels (14 channels). However, the same method gives good accuracy when we used with 64 channels data. Therefore, SL filters are mostly useful for higher number of EEG channels compared to lesser number of EEG channels.
Dear Michael Cohen, I agree with your comments. Increasing the number of channels is the best way to improve the spatial resolution.
- Badges
- Science topic
- Similar topics
- Biological Science
- Physiology
- Biosignals
- Biomedical Signal Processing