Transcript MAVOORI

Neurochips and Neural Telemedicine
Jaideep Mavoori
University of Washington
(currently at Neurovista)
Collaborators:
Andy Jackson+, Eb Fetz (Biophysics and Neurophysiology)
Tom Daniel (Biology)
Chris Diorio (Computer Science)
+Currently
at Newcastle University
Neural Telemedicine
Acquire
Monitor
Detect
abnormalities &
their evolution
Diagnose
Single neurons
Local field potentials
ECoG
EEG
EMG
JM
Track pathological
waveforms
Issue alerts
Initiate curative
actions
Trigger repair
mechanisms
Drug delivery
Electrical stimulation
2
Biological Motor Control
Photo courtesy of UW PWB program
JM
3
Conventional neurophysiology of restrained primates
Filters + Amplifiers
Recording
Analog to Digital
Converter
Spike
discriminator
Stimulator
JM
Analysis
4
Option 1 - Telemetry systems:
• high power consumption
• limited range
• transmission delays
JM
5
Option 2 - Implantable microelectronics:
• autonomous operation
• low power
• limited processing capability
JM
6
Primate Brain Computer Interface
Connector
50μm diameter
tungsten wire
Polyamide
guide-tubes
JM
Skull
7
Neurochip BCI User interfaces:
PDA (Lyme)
PC (MatLab)
JM
8
Architecture of the neurochip
Two Cypress Programmable
System-on-Chips (PSoCs)
 Front-end signal processing
(filtering, DC offset + amplification)
 Neural signal sampled at 12ksp/s
 2 EMG signals sampled at 2.7ksp/s
 Real-time spike discrimination
 Spike rate and mean rectified EMG
compiled for user-defined timebins
 2 x 8Mb non-volatile FLASH
memory
 Biphasic, constant-current stimulator
(±15V, ~100μA)
 JM
Infra-red link to PC or PDA
9
M1 and muscle activity during natural behaviour:
IEEE TNSRE, 2006
JM
10
M1 and muscle activity during natural behaviour:
IEEE TNSRE, 2006
JM
11
M1 and muscle activity during natural behaviour:
IEEE TNSRE, 2006
JM
12
Long-term recording of cell activity:
Continuous recording of a single M1 neuron for 2 weeks.
J. Neurophysiol. 2007
JM
13
MI-Motor Model
Cortical
activity
System
properties?
Muscle
output
How time-invariant is this system ?
How does the model compare in task and free behaviours ?
Can we alter the system properties ?
JM
14
Motor Pathway Modeling
JM
15
Motor Pathway Modeling
JM
16
MI-Motor Model
Findings:
 Over several neurons and muscles, aspects
of the system are linear and time-invariant.
 The relationships translate from task to freebehaviour as well as from day to day.
 Advantageous for neural prosthetics:
Parameters for limb mechanics can possibly
be learnt during a training segment and
applied during a wide range of daily activities.
JM
17
Altering system properties:
Cortical remapping with the Neurochip
Neurons that fire together, wire together.
Induce correlated firing between neighboring sites
Long-lasting changes in biological wiring
Nature 2006
JM
18
Altering system properties:
Cortical remapping with the Neurochip
Neurons that fire together, wire together.
Induce correlated firing between neighboring sites
Long-lasting changes in biological wiring
Nature 2006
JM
19
Neurochip conditioning
JM
20
Neurochip conditioning
JM
21
Motor cortex plasticity
induced by Neurochip
conditioning
Movements evoked from the
recording site changed to resemble
those evoked from the stimulation
site.
JM
22
Motor cortex plasticity
induced by Neurochip
conditioning
Additional findings:
 Timing from spike to stimulation
is critical. Delay of 20 ms
produced strongest conditioning
effect.
 Conditioning effects last for
several days.
Useful for repairing damage caused
by spinal chord injury or neural
disorder.
JM
23
Monitor
JM
Detect early onset
Diagnose
Repair
24
Monitor
Detect early onset
Diagnose
Repair
Early stages of neural telemedicine
Early stages of neural disorders
JM
25
Miniature chips for insect flight studies
1st Generation
2nd Generation
(top)
(top)
1cm
(bottom)
1cm x 3cm x 0.5cm
1.47g (without battery)
JM
1cm
(bottom)
1cm x 1.9cm x 0.4cm
0.85g (without battery)
3rd Generation
5th Generation
4th
Generation
(top)
(top)
(bottom)
0.9cm x 1cm
0.6g (no
battery)
1cm x 1.25cm x 0.25cm
0.25g (without battery)
(bottom)
1cm x 1.27cm
0.42g (no battery)
26
Acknowledgements:
Chris Diorio
Eb Fetz
JM
Tom Daniel
Andy Jackson
Supported by NIH, ONR,
UW Royalty Research Fund,
Packard Foundation.
27
JM
28