Transcript Zoran_jwb

Analysis, Design, and Control of
Movable Neuro-Probes
Z. Nenadic, E. Branchaud,
W. Collins, and J. Burdick
Engineering and Applied Science
R. Andersen, J. Pezaris,
B. Greger, B. Pesaran
Biology
California Institute of Technology
(Auxiliary Program Started April 1, 2001)
Limitations of Current Neuro-Probe
Technology
Make the probes movable to track/find neurons!
Key Challenge: record high quality signals
from many neurons (for months/years)
Fixed positioning of implant
• Non-optimal (or wrong!) receptive fields.
• Electrode not near cell body.
• Low impedance (poor SNR) design required.
Gliosis and encapsulation
Movable Probe Concept: computer controlled movable probes can
track moving neurons, find new neurons, break through encapsulation.
Project Goals
• Short term: validate concept, enable more complex acute experiments
• Intermediate term: Develop design specs for MEMS devices
• Long term: develop MEMS technology for implantable devices (see talk
by Y.C. Tai)
Current Research Program Outline
Theory – develop algorithms for probe
control using modeled (computational)
environment
• Model extra-cellular neuron potentials
• Characterize local field potentials (LFP)
• Control algorithm development guided by
computational model
Implementation – meso-scale hardware
test-beds
• Validate concept, evaluate algorithms
• Enable testing of Biomechanical issues of
chronic movable probe operation
Single Cell Extracellular Potential Simulation
3720 compartment NEURON pyramidal
cell model (adapted from Mainen & Sejnowski ‘96)
Synaptic inputs scattered uniformly
throughout dendrites.
Laplace equation:
soma
Boundary condition:
Since solution nearly impossible, use
line source approximation (Holt & Koch ‘99)
Spatial Variation of Extracellular Potential
Peak-to-peak amplitude
Tuning curve
Keep electrode in this region!
Added noise -- independent,
Gaussian, zero-mean
Noise variance -- determined by signal-tonoise ratio (SNR)
Movable Probe Feedback
Q: How to find the maximum point of the tuning curve?
A: Stochastic optimization
gradient method
basis function method
splines – under development
(stochastic gradient descent)
future
position
step size
current
position
stochastic gradient
Under the conditions:
the sequence
converges to
almost surely i.e.
Movie 1
Movie 2
Movable Probe Test-bed Development
Multiple development phases to maximize scientific
gain and engineering development along the way.
 Acute: probes inserted in brain tissue for a few hours
 Initial validation of movable probe concept and algorithms
 Enable better short-term prosthetic feedback experiments
 Semi-chronic: electrodes remain, motors removed
 Understand biophysical issues of chronic probe operation
 Track neural populations over days for plasticity studies
 Will set spec.s for future MEMS devices
 Chronic: movable system permanently implanted
 Ultimate goal: needs MEMS development
 Key technology for future neural-prosthetics
Acute Test-Beds
Last time: motorizing the CCMD, a
pre-existing manual 4 probe device
Completed, with lessons learned
– (need to list some lessons here)
Put diagram of Thomas system here + a few words about status
Semi-Chronic Test-bed
Phase I(a): two motor drive that fits inside head cap
• Motors and electrodes stay inserted for a few days
• Power, control, data wires attached at start of each session
• Useful for studies in learning and plasticity
Phase I(b): two motor drive that fits inside head cap
• motors detached at end of every session—electrodes stay
implanted for long periods of time.
• enables testing of long term biophysical impact of chronic
electrode operation (inflammation, gliosis, etc.)
Phase II: 12-16 motor micro-drive with removable motor
assembly. Will enable consistent recordings of many cells
Phase 1 Design Progress
Pretty Picture
Pretty Picture
Conclusions
Developed theory for control of movable probe based on peak-topeak amplitude. Future investigations will include:
• other wave form features, such as phase, shape, frequency, etc.
• event detection algorithms to handle irregularity of spike trains
• the effects of multiple units (inclusion of spike sorting).
• effects of tissue dimpling and relaxation (easily incorporated)
Movable Probe Test-Beds development program started
• CCMD motorization completed, with lessons learned.
• 5-probe acute system developed
• Phase I of semi-chronic system largely designed