Integrated Wireless Neural Recording System for
Neuroprosthetics and Advanced Neuroscience Research
Moo Sung Chae
Integrated Bioelectronics Group Department of Electrical Engineering University of California, Santa Cruz
Overview
Neural Recording Systems
Motivations
Design Issues
Noise
Optimization
Wireless
Prototype Systems and Test Results
Summary
A recording system to monitor neuron’s activities
Neurons communicate by “action potentials”
Action potential is an all-or-nothing signal
Neural Recording Systems
Near field signal 1) Intra-cellular action potential : rapid change in trans membrane
potential cause by the voltage dependent trans membrane conductance (H-H model)
2) Extra-cellular action potential : solenoidal current around neurons, few hundred µm
3) Local field potential (LFP) : coherent low frequency change, a few mm
Far field signal 1) Electrocorticograms (EcoG) : sub-dural
2) Electroencephalograph (EEG) : on the scalp
Neural Signals
Signal Type Bandwidth (Hz) Range (mVpp)
EEG 0.05 ~ 128 0.02 ~ 0.4
ECoG 0.1 ~ 64 0.02 ~ 1
EMG 1 ~ 128 0.02 ~ 1
LFPs 0.1 ~ 100 0.1 ~ 1
EAPs 100 ~ 10K 0.04 ~ 0.2
IAPs 100 ~ 10K ~ 100
Neural Signals – Properties
Adjustability of the amplifier is necessary
Study of complex neural networks of the animals in their natural environment
A large number of recording channels
To remove tethering wire
Miniaturized and low power consumption
Motivations (1) – Advanced Neuroscience
Artificial devices to replace or improve the function of an impaired nervous system
Upper and lower limb prosthesis [3],[4]
Bladder and bowel movement [5],[6]
Respiration control for SCI patients [7]
Hand grasping [8]
Motivations (2) – Neural Prostheses
Challenges
Simultaneous recording from a large number of channels (more than 100 channels) [9]
Untethered wireless transmission of recorded data in free running animals
On-the-fly spike sorting
Programmability and Versatility
Integration and Miniaturization
Low-power consumption
Proposed System
Design methodology to optimize the power and area
Programmable gain and bandwidth of the amplifiers
On-the-fly spike sorting engine
High data rate & Low power transmitter for simultaneous recording of 128 channels (90 Mbps)
Front-end Blocks
Decide the resolution of ADC
Optimize preamplifier
Decide the number of channel per one ADC
System resolution by electrode noise
Proposed Architecture
Component modeling
1. Fixed design constants
2. Design variables to optimize
Design Methodology
Noise Sources
Neuronal Noise
~ 20µVrms
Tissue/Electrode Noise
~ 10µVrms
Electronics Noise
~ 5µVrms
Z. Yang, Q. Zhao, E. Keefer and W. Liu, "Noise Characterization, Modeling, and Reduction for In Vivo Neural Recording," to appear in Neural Information Processing Systems (NIPS), 2010.
Biological and interface noise is larger than circuit noise.
Need to optimize the performance of the amplifiers
Different techniques for different noise sources
Neural Signal Processing Engine
UWB
1. FCC assigned 3.1 ~ 10.6 GHz spectrum for unlicensed medical and communication systems.
2. Short range applications with large bandwidth 3. Small antenna size due to high frequency nature
IR UWB vs Multi-carrier UWB
Impulse radio (IR) UWB is more suitable than
multi-carrier UWB such as OFDM.
Use UWB pulse for data symbol (telegraph)
Simple CMOS pulse generator for Tx
Low power at Tx, but high power at Rx
UWB Wireless Tx & Rx
H L H L L H H H L
10 01 10 01 01 10 10 10 01 00 00 00 00 00 00 00
serialized 9-bit sampled data Redundant data UWB pulse(PPM)
Manchester coded data
sampled data
UWB pulse(OOK)
UWB Pulse generator
Pulse shaping
Filter
UWB antenna
Encoder
OOK modulator
PPM modulator
dCLK pCLK
From ADC
MU
X
Pad & package parastics
9
Enc
oded
da
ta
OOKin
PPMin
Seriallizer
data
dCLK Mode sel
BPF @ 4 GHz LPF @ 100 MHz
LNA stage Diode Lowpass Filter
FPGA
Analog . Amplifier BPF UWB
Antennas
UWB Prototype
1. Custom designed TX chip in 0.35µm CMOS process 2. Manchester coding for clock recovery and channel
separation@RX 3. PPM and OOK 4. Adjustable pulse width 5. Receiver is implemented using commercial components.
UWB Prototype Test Results
[Output of CMOS TX] [PSD of the UWB pulse@TX]
[Recovered data@RX]
Maximum data rate of 90Mbps
1mW power consumption with 90Mbps OOK
Digital clock recovery by the FPGA in the RX
USB interface with PC to handle high bandwidth
LPF output
FPGA output
FCC emission mask
Architecture of Prototype System
Bench Top Test Results
1. Ex-vivo extracellular action potential recording from a dissected snail brain (A circumoesophageal ring dissected from a H. Aspersa)
2. Concentric bipolar electrode with 25µm platinum tip with suction
3. Measured impedance of electrode is 190KΩ at 1KHz.
Ex-vivo Recording
In-vitro Recording
1. Snail neurons were cultured on Ayanda MEAs 2. Electrode impedance is 1MΩ. 3. Recording from a cluster of neurons 4. Field potentials were observed.
1. 4-channel recording IC (cut-down version of 128ch recording IC)
2. EEG, ECG, EMG
3. USB2.0 PC interface
FPGA
USB2.0
FSK
FSK
PIC Microcontroller
Another Prototype System
[A mouse wearing the recording system. Photo was taken at Arizona State University]
1. Recording electrode implanted in a live mice brain
2. The impedance of the electrode is 100KΩ.
3. Extracellular action potentials
In-vivo Recording
Human EEG and ECG Recording
Summary & Acknowledgement
Need for versatile wireless recording systems for advanced neuroscience research and neuroprothetics
Challenging requirements in terms of power and area
Need various technologies for the implementation of the system
An example of system implementation
Chip fabrication by National Semiconductor