Transcript Document

Laboratory 13 : Neurophysiology
Dr. Craelius
• Autonomic (Involuntary) and
Somatic (Voluntary) Nervous Systems)
• Methodologies for measuring and controlling
both.
• Relevant to Neuroprostheses. I.e. how can we
“decode” what the brain is telling our organs and
muscles, so that when the information channels
are blocked, we can replace them?
The output is a pulse train: Its frequency contains the information.
In general the higher the frequency, the greater the information
content. Neurons in the MI cortex are specialized for such operations
As kinematics or dynamics.
Neurophysiology Basics
• Muscles and neurons are excitable, and
carry information through their pulse rate.
• I = log2(fm*t +1)
• Action Potential - a reversal in relative
polarity or change in electrical potential of a
cell Neurotransmitters- chemical
messengers
• Central neurons are specialized for function.
ANS Background
• Spectral analysis of HRV reveals 2 limbs of the
ANS.
• ULF (diurnal) HRV is predictive of cardiac
performance.
• HRV signatures using Cepstral Analysis (US
patent 6,390,986).
• HRV manipulation can improve asthma (Vaschillo,
Lehrer et al, 2004).
• Portable digital recorders of ECG facilitate
analysis.
HRV Analysis Pyramid

Complexity
Frequency
Time Domain
Inverse
Problem
Heart Rate (HR)
(HP)
Heart Period (HP)
(HP)
Beat-to-beat (RR)
Pacemaker: Sino-atrial node
Influences: Autonomic + Mechanical+ Metabolic + Environmental
Autonomic Nervous System &
Heart Rate Variability
• Exerts effects on every organ, but the heart
is the most “visible” organ we can examine.
• Para slows it, Sympa speeds it.
Pacemaker
Para
Sympa
Beatmaker
Autonomic (Para and Symp) modulations of HR: A model
What modulates HR?
• SNS has a periodicity of ~ 20 seconds, and
possibly others.
• ANS has a periodicity = breathing rate, and
possibly others.
• Thoracic motions alone can modulate, and vagal
nerve drive of respiration can contribute.
• Thermoregulation, daily activities can modulate
long periodicities.
Low and High Frequency
variations in blood pressure
Respiration controls
HF
What limb of the ANS controls
the LF?
RR
RESP
RR
RESP
High RSA during sleep @ 11:30
PM
Tachogram
Low RSA, High LF During sleep
12:50 AM
RSA disappeared during sleep
11:30 PM
13 breaths/min
1
3
b
1:00 AM
Can HRV identify disease or
specific individuals?
• Age-related normal ranges of overall HRV
= SDNN, are known and are predictive of
survival after MI.
• A brief record of HR can be a signature of
an individual using HR vector cepstral
methods *.
*Curcie, D, Craelius W: Recognition of Individual Heart Rate Patterns with Cepstral Vectors,
Biological Cybernetics, 77/2:103-109, 1997
Analyzing HRV
• Collect sufficiently long , ‘clean’, epoch,
I.e. need at least a few cycles of the rhythmso for LF , get > 3 X 20 seconds.
• First examine tachogram, edit artifacts.
• Do time domain stats, ie, S.D.
• Do spectral analysis if you have sufficient
data, ie. Need several cycles to detect.
Cardiovascular Resonance
•
•
•
•
Vaschillo & Lehrer et. Al.
Get ANS into resonance by biofeedback.
Deep breathing at resonant rate is key.
Resonance can influence performance.
Time Domain Indices
Task Force: Circulation 93:1043-1065, 1996.
Geometric Indices
Task Force: Circulation 93:1043-1065, 1996
TINN analysis
RR Interval histogram
Frequency Bands
Power
(ms2)
ULF
VLF
LF
HF
0.00010.003
0.0030.04
0.040.15
0.150.4
Frequency (Hz)
Ratios
• LF/HF : estimates sympathetic to
parasympathetic activities
• LF-tilt/HF-supine: a more specific estimate
Normal Range
Variable
SDNN
SDANN
RMSSD
Triangular index
Total Power
LF
HF
LF
HF
LF/HF
Units Normal Values
(mean +- S.D.)
ms
141 
39
ms
127 
35
ms
27 
12
ms
37 
15
ms2
3466  1018
ms2
1170  416
ms2
975  203
nu
54
 4
nu
29
 3
1.5- 2.0
HRV Oscillations
Frequency Range
Likely Origin
Component (Hz)
HF
0.15 - 0.40 Parasympathetic outflow
LF
VLF
ULF
0.05 - 0.15 Mostly Sympathetic in
standing position
0.003 Possibly thermoregulatory
0.05
or plasma rennin activity
<0.003
Wide range of
determinants like posture,
Collect
ECG- Lead
II
Detect
fiducial R
points
(Pulse record in our lab)
Time
Domain
Measures
Overall
HRV
Frequency
Domain
Power in
Bands
Estimate
autonomic
activities
Modelling
HR Vector
Classify
individuals
Instantaneous
Heart Rate
Stationary?
Free of Artifact?
Processing Pulse Record
Unfiltered Pulses
High Pass Filtered @ 0.2 Hz
Baseline correction: If you filter too much, you differentiate.
Somatic Nervous System:
Signals
• Originate in a Motor Neuron
– Activated by conscious thought or afferent
input (i.e. reflex)
• Travel through the nervous system to the
target muscle(s) via, depolarization (action
potential) and neurotransmitters : Signal
Degradation
Motor Homunculus: Map of
functions
Motion Control
Volition + Load -(sensation) = error
Volition
+
External Load
Controller
+
_
Proprioception
Vision

Motor Regions for UL
Index Finger
Forearm
Areas for placing
electrodes
Biceps
Bionic Approaches to Restoring Mobility
Brain
• Mobility can be
restored by
several
neuroprosthetic
approaches *.
Computer
Action
BCI
BMI
Robot
Action
Muscles
Action
Computer
Action
Hybrid
BMI
Muscles
PMI
1.* Craelius,W.: "The Bionic Man: Restoring Mobility", Science, Vol 295, 1018-1021,
2002
Training primates to move arm
by decoding neuron signals
Brain-Machine/Computer
Interfaces
• Monkeys in Brooklyn moving arms in
North Carolina, fast learning (Wessberg et
al.)
• Completely paralyzed persons moving
cursors and robotic arms (Kennedy, PR, et
el.)
• Paraplegic with implanted SC chip using
switches on walker (Rabischong)
Record Inside the brain ?
• Need > 1000 Implanted electrodes
• Hence need wireless control from external
controller
• Electrode biocompatibility and migration
• But decoding volition from motoneurons is
surprisingly easy: simple cumulative
summation of firing rates (linear)
Volitional
Degradation/Restoration
•
•
•
•
•
G = H · V ( G and V are column
vectors)
G is the measured response
H is the degradation through the system
V is the volition
To Retrieve volition:

V  H 1  G
Task
Brain
Volition
V
Motors
Degradation
H
Context
Environment
V
Muscular
Output
Controller
Filter
1/H
State Vector
State
G
Register
Linear filter is simplest
And best decoder
How to measure performance of
decoding?
• How accurate is positioning of arm?
• Euclidean distance:
• Speed versus accuracy
Speed/Accuracy Tradeoff
How to quantify?
Attention
Time
As your need for attention
Accuracy
Accuracy
Measuring performance with
Speed/Accuracy tradeoff : Specific
targeting task
90°
Fitts
SAT
SAT test
•
•
•
•
•
•
MT = a + b log2(2A/W)
where
MT = average movement time = Time/# of hits
A = amplitude (distance) of movement between
targets
W = width of the target
a = intercept
b = slope
log2(2A/W) == difficulty level
SAT graph
Difficulty index
Protocol
1. Pulse recording 5 min --- file
2. SAT test
5 min ---- file
3. Deep Breathing w/pulse recording 5 min-- file.
4. SAT test
5 min ---- file
5. Pulse recording 5 min --- file
6. SAT test
5 min ---- file
Analysis
1. Prepare pulse files, with HP filtering if
necessary.
2. Use RR interval program to get intervalsoptimize for minimal artifacts.
3. Put RR & SAT data in Excel- analyze &
graph.
4. If time, further analyze RR data with Loga-Rhythm.
Bionic Interface Types
BMI
For “partial” paralysis
BCI
For complete paralysis Screen Cursor
PMI
For amputees or with
weak muscles
HBMI SC injuries
Robotic Limbs
Parkinson’s
Activa Tremor
Control
CBI
Robotic arms
FreeHand