prediction of epileptic seizure: MATLAB or Raspberry Pi ?

Comments

• wjcroft Mount Shasta, CA

  January 2019

  Hi Walfahad,

  
  

  Here is a recent paper on EEG prediction,

  
  

  https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0099334

  
  

  My quick scan of this paper would say that indeed Matlab / EEGLAB is used in initial phases of the analysis, to build Machine Learning models. So this is sophisticated signal processing and would not work on Raspberry.

  
  

  After the ML models are created and if it is desired to run real-time analysis on a patient -- I am unclear if that signal processing is appropriate to run on a Raspberry. Raspberry may still be underpowered for this. Your best bet is to run signal processing on your laptop until you can quantify how much CPU is consumed in the real-time analysis.

  
  

  Remember the radio link from the Cyton to the dongle, means that the person being monitored can have decent mobility within a room or rooms.

  
  

  Finally, do take note of the Disclaimer listed at the bottom of the Cyton page, "The Cyton Board is not a medical device nor is it intended for medical diagnosis..."

  
  

  Regards,

  
  

  William
• AlainD

  March 2019

  Hi,

  
  

  Interesting paper. We should note however that the Freiburg EEG database collected intracranial measurements of patients admitted for pre-surgical evaluation, that the different stages (pre-ictal, ictal, post-ictal and inter-ictal) were labelled by specialised technicians (because it could be challenging to visually differentiate them) and that artefacts were manually removed with some software. EEG scalp measurements with a cython board will likely be more noisy and harder to label, but it would be nonetheless really interesting to try to see whether we can differentiate the epileptic stages with this technology. 

  
  

  Is there any forum, blog or papers where attempts to measure epileptic activity using openBCI boards are described?

  
  

  If the epileptogenic zone (EZ, to avoid stumbling on the first word:) of a focal epilepsy has been localised accurately enough, using some combinations of modalities (EEG, MRI, MEG, PET, ...) available at a hospital or a research facility, then maybe the 4 channels of the Ganglion board are sufficient to record the action; 3 for the focal area and 1 for the extra-focal recordings ... or 1 for the focal area, 1 for the extra-focal recordings and one for a pulse sensor (depending where the focus is, tachycardia is observed in the pre-ictal and ictal stages, and sometime bradycardia in the post-ictal).

  
  

  Alain D.

  
  

  
  

  
  
• Billh

  March 2019

  1. If the person is resting, in many cases a pulse sensor for detection of EKG looking for a 40% to 70% rise in baseline heart rate can be enough. This can be done easily with OpenBCI equipment.

  
  

  2. Individual difference make it hard to generalize about the ability of OpenBCI equipment to detect the pre-ictal state on a given individual's EEG. It depends on where and how the activity changes. In many cases, just looking at the overall power band in the beta range in each hemisphere, perhaps  20 to 30 Hz, is enough, and in such cases 4 leads, two on each side, is good enough.

  
  

  3. Exact localization of a seizure focus may require depth electrodes or >40 lead equipment.
• AlainD

  December 2019 edited December 2019

  I am wondering if anyone has tried openBCI to detect epileptogenic spikes? Assessing the density of these spikes may inform, for instances, about the efficacy of a treatment or the probability of a seizure coming soon.
• Billh

  December 2019 edited December 2019

  The OpenBCI devices are not in general approved for clinical use in many countries that regulate EEG devices. That said, the device can be used to detect spikes. Spike density (do you mean density in time or space or both?) using scalp EEG may not in general predict a seizure, except of course by detecting that has already started, but may be in some individual subjects with specific types of focal onset seizures. In general, successful predictive EEG spike detection predictive algorithms have used implanted electrodes, not the scalp EEG that OpenBCI has. Here is a reference:
  https://doi.org/10.1016/j.yebeh.2014.06.023
  see also eg.
  https://doi.org/10.1016/S0896-6273(01)00262-8