Jon Newman
2011-05-03 10:46:55 

Outline for: Effective hardware and parameters for optogenetic control of
population activity within dissociated cortical cultures grown on MEAs.

Motivation:
  General:
    1. Activation of full neuronal sub-population, genetically chosen, instead of
    random subset of cells with axons that pass near electrodes.
    2. Location of neuronal activation is localized at soma instead of in the middle
    of axons, which causes back propagation [ in the process of confirming this ].
    3. When combined with MEA technology, Higher bandwidths of communication  between
    experimenter and neural network compared to e-stim are possible because of reduced 
    electrical artifact and the ability to have more refined spatial excitation of cells.
    
  
  For cortical cultures:
    1. Allows researcher to decouple population activation of a specific cell type
    from other electrophysiological processes via combination with synaptic or
    extra-synaptic receptor competitive antagonists. This will allow a means for
    breaking down the mechanisms of plasticity with greater precision than before.
    2. Allows researcher increased bandwidth for the induction of long lasting
    network changes and to relate these changes to affects on info processing.
    3. Analog, continuous stimulation patterns are possible.
    
A. Simple system for optogentic control of neuronal populations in-vitro:
  1. Light stimulation system
    - simple circuit that can be reproduced by other labs that allows control 
    of up to 4 light modules and current measurements at LEDs.
    - circuit should be modular in design so it is easy to remove and use in a 
    system that is not NR.
    - electro-mechanical LED interface that is modular from the control circuit.
    Allows user to 'plug in' different LED colors easily.
    - optical interface to culture. Again, modular. Allows user to easily remove'
    from mechanical interface if they want
    - cooling system. Modular. Modified Peltier as designed by Wagenaar. Allows temp control
    of cultures for long protocols in the incubator.
    - low cost is a feature compared to laser based systems (800 dollars each just
    for lasers, excluding optics and delivery system).
  2. Control software
    - Heavily modified NeuroRighter. (most of the way there already)
    - Modular closed loop control interface is built in (just started).
      - Allows user to select from any recording stream
          - Raw data
          - SALPA filtered
          - Band-filtered
          - Spikes
          - LFP
          - EEG
          - MUA
          - Aux digital
          - Aux analog in
      - To control three base outputs
          - e-Stimulation (packaged AO and DO through NR's board)
          - Aux AO, general purpose analog out, up to 12 channels
          - Aux DO, general purpose digital out, up to 32 bits
      - Loop time determined by user, down to 100 ms.
      
B. Effective parameters for controlling first and second order statistics of population firing:
  1. What is the right light power? i.e. at what power does the effect of ChR2 or
  Halo Rhodopsin saturate.
    - Apply total synaptic blockade
    - Apply 1 ms light pulses of random current amplitude, while measuring optical power
    - Find integrals of post-stimulus firing rate histograms. Plot vs corresponding light power.
    - Find the saturation point in these curves. This is nominal light power.
    - This process can be automated in closed loop control.
    - How much cross talk is there?
      - use culture only expressing only ChR2 or Halo to quantify how much cross activation is present
      - Also, what is the effect of FR on non-infected cultures?
    
  2. Open-loop stimulation parameter testing
    - Relies on first result
    - Performed in the absence of synaptic blockers and then blocking the big three.
    - Need to answer: which parameter/stimulation scheme is best for
      1. Control of population FR?
      2. Control of second order statistics (like isi histogram)?
    - Three schemes to test
      1. Periodic (using nominal power from 1.)
        - freq
        - pulse width
        - duty cycle
      2. Random pulse (using nominal power from 1.)
        - avg. freq
        - pulse width
        - duty cycle
      3. Continuous random (using nominal power from 1. as maximal power)
        - avg. power
        - update freq
        - duty cycle
        
  3. Estimating non-stationarity effects over 
        
C. Effect of synaptic blockade
    - Use the most effective open loop protocols from B. to look at how blockade of different
    synaptic types influences the population recruitment (temporally).
    - Note that because of these changes, the only way to reliably control firing rate
    in their presence is the use of a closed loop control system
    
D. Non-stationary effects over long time periods
  - Cortical networks are recurrent, non-stationary and highly nonlinear (i.e. they have 
  wild dynamics).
  - This is especially true over time periods wherein the cultures can undergo heavy
  morphological change.
  - Use the most effective open loop protocols from B. to examine degradation in performance
  over long time periods (i.e. days).
    
D. Closed loop controller
  - Estimator for something
    - avg. firing rate over 100 ms
    - avg firing rate of 1 sec
    - median firing rate
    - variance of ISI distribution 
    - probably Kalman filter to deal with non-stationarity
  - Feedback scheme (probably PID, or just P).
  - Show performance enchancement in controlling firing rate over long time periods and 
  in the face of pham. perturbations (like the application of GABAzine or AP5).
  - Compare to identical open-loop control signal.