Analyzing sleep EEG

I'm about to start playing with some sleep EEG data from my girlfriend's research group. She's studying neurobiology, and spends this summer working for the Sleep Research Group at Biomedicum Helsinki. They are looking for a way to automatically detect (score) different sleep stages from the raw EEG data. They currently have some semi-automated tools that use a Fourier transformation and some heuristics to detect the different stages, but the tools are not nearly accurate enough. Thus they need to spend countless dull hours manually scoring the data from their rat experiments.

I promised my girlfriend that I'd take a look at the data and see if I could come up with something useful. It's quite likely that I'll just end up admitting defeat (smarter people have tried before...), but it's still a very interesting problem which also gives me a chance to learn something new. She promised to provide me with some real data and any required background information. Here's a list of things I plan to do with the data:

• Find or write a simple tool for displaying the time and frequence graphs of a selected window of the EEG data. This will allow me to get familiar with the data and the possible common patterns.


• Apply tools like genetic algorithms or neural networks (for example self-organizing maps) on the data and let the computer come up with the heuristics for detecting the sleep stages.


• Validate the effectiveness of these tools against the manual scoring and the existing semi-automated tools.

Let's see how this turns out...