2011-12: UCSD MMIL Real Brains

From 2011-2012 I got the chance to work on EEG signal processing and thalamocortical simulations for the UCSD MultiModal Imaging Laboratory.  I had not done a lot of programming for many years, and very little ever on a Linux system.   I had classes in EEG, MEG and FMRI neuroimaging, but little experience.  I gave it my best.  Before starting the job, I worked for weeks reading papers and books to come up to speed.  But it turned out that the reading suggested to me beforehand was not very relevant.  So I had to work really hard from day 1.  There was little time to socialize, less so with constant interruptions from where I sat.

I learned Matlab and CENTOS in a big hurry.  Then I learned how EEG/MEG data sets were processed start to finish.  I built matlab code that mapped out the process from taking in the data, down sampling, marking epochs, etc. etc. right through to printing plots and doing statistical correlation analyses.  My code was all parameterized like an assembly line.  I was also responsible for maintaining older code.  My code relied heavily on many library routines built at MMIL over the years.

Our team was also examining sleep signals, and the occasional K-complex that appears among the spindles.  We needed to adapt a well known thalamocortical simulation of up and down states to have it match our theory of where K-complexes come from.  This was C++ code that challenged my code reading skills.  I made all the necessary changes, and it worked.

So we had data collected from electrodes right from the cortex (ECOG) of sleeping patients who were being prepared for surgery for intractable epilepsy.  We would analyze it with my software package, and much additional analysis was being done by Dr. Mak-McCully then working with Dr. Halgren and Dr. Bazhenov.  I also developed some CENTOS scripts for batch job submission because we had a huge parameter space to explore to find these K-complexes in our simulations.  We often had as many as a dozen simulations running in parallel on our servers with the  code runs all automated.  The end results were published in PLOS and also at an SfN (Society for Neuroscience) poster session.

After getting things pretty well in shape for this project, our P.I. lost his R01 funding and support for several jobs in our group came to an end, including mine.