extracting information from the electrical activity of …ted/sigma_xi_055.pdfextracting information...

Extracting Information from the Electrical Activity of Human Skeletal Muscle

Edward (Ted) A. Clancy

Department of Electrical and Computer EngineeringDepartment of Biomedical Engineering

Worcester Polytechnic Institute, Worcester, MA, USA

EAC06–055

Sigma Xi Lecture – 4 October 2006

Worcester Polytechnic Institute

Presentation Overview• Brief background:

– Muscle electrical activity (EMG)

• Engineering models, methods for extracting “EMG amplitude”

• Applications of EMGamp:– Prosthesis control, gait biofeedback, EMG-

torque, EMG-impedance

• Surface EMG arrays ( newest stuff !!)

Introduction

EAC06–056


The Motor UnitBackground

EAC06–057

(http://academic.wsc.edu/faculty/jatodd1/351/ch6outline.html)

• Muscle fibers contract “all or nothing”

• One mechanical twitch per contraction

• Regulate force via:– Number of active

motor units(recruitment),

– Firing rates

• Firing intervals independent (mostly)

• Electrical “action potential” recorded


Electromyogram (EMG) Signal Model

“Interference pattern” EMG:Superimposition of multiple MUAPs plus measurement

additive noise.

EMG Signal Origin [Basmajian and De Luca, 1985]

Background

EAC06–058

MUAP


EMG Amplitude (EMGamp) Estimation

EAC06–059

EMGamp

EMGamp

EMG amplitude 0–10 mV (~1.5 mV rms)Surface EMG frequency bandlimited at 2000 Hz

• EMG Amplitude (EMGamp): “Intensity” of recorded EMG– “Time-varying standard deviation of EMG signal”

• Original estimator: Inman et al. [1956]– Analog full-wave rectify and RC low pass filter

Electrode-Amplifier(From Liberating Technologies)

– Varies with force


Improving EMGamp Estimation

• EMGamp improved by:– Removal of measurement noise– EMG signal whitening

• Adaptive whiteningTo reject measurement noise

– Multiple EMG channels (for large muscles)– Optimal detectors– Optimal smoothing (bias vs. variance error)

EAC06–060

EMGamp

Increases statistical bandwidth of EMG Reduces variance of amplitude estimate.{


Single Channel, Including Noise

Σ( )H etimejω

wi ni ir

si

mi

vi

Zero Mean, WSS,CE, White Process of Unit Intensity

Shaping Filter

Zero Mean,WSS, CE,Additive NoiseProcess

Measured SurfaceEMG

EMG Amplitude

Σ

iaCable andmotion artifact

ii vr +

Prakash et al., IEEE Trans Biomed Eng 52: 331–334, 2005.

WPI MQP !!!

EAC06–061

EMGamp


Multiple-Channel, Including Noise

EAC06–062

EMGamp

Clancy Ph.D., 1991.


Experimental Investigations• Separate EMGamp evaluation from

EMGamp-torque, EMGamp-impedance

EMGampEstimator

Target Tracking

ˆ s t( )m t( )

EMGamp-Torque

ˆ T

EMGamp-Impedance Jbk ,,

EAC06–063

EMGamp


Experimental Apparatus

EAC06–064

EMGamp

Clancy, IEEE Trans Biomed Eng 46:711–729, 1999.


EMG Electrode Sites

EMG Flexion 3

EMG Flexion 2

EMG Flexion 1

EMG Flexion 0

EMG Extension 3

EMG Extension 2

EMG Extension 0

EAC06–065

EMGamp

EMG Extension 1


Multiple-Channel Whitening

EAC06–066

EMGamp

• Constant-posture, constant-force contraction

Clancy and Hogan, IEEE Trans Biomed Eng 42: 203–211, 1995.


EMGamp: Myoelectric Control of Prosthesis

EAC06–067

Prosthetics

• Use remnant muscle EMG to command electric hand, wrist, elbow– Some lower limb prosthetics research

Major current effort by U.S. military to improve prosthetic limbs

•Improved myoelectric controls

•Embedded neural sensors•To give more control•To provide feedback

•Directly connect limb to boneBoston Elbow


Nerve Re-Innervation for Myocontrol

RIC Arm; www.popsci.com

• Kuiken nerve re-innervation surgery

• Nerves from severed limb re-connected to pectoralis muscle regions

• Several myocontrolsites created

• Substantial improvement

EAC06–068

Prosthetics


Myoelectric Control Example (Movie)

Jesse Sullivan video

EAC06–069

Prosthetics


Opportunities in Myocontrol

• Newer prosthetic limbs — circa 2002– Microprocessor– Digital control– Simultaneous control of multiple motors

• Advanced EMG processing now feasible– Not feasible previously in production arms

EAC06–070

Prosthetics


EMGamp: Gait Biofeedback

EAC06–xxx

Gait

EAC06–071

Re-training after stroke, traumatic

brain injury.


Biofeedback Gait Re-Training• Motivation: Some patients may not properly

relearn gait motor patterns– Incorrect timing of plantar flexor firing

• Long-Term Objective: Re-train proper gait via biofeedback (“normative” gait vs. patients)

• Short-Term Objectives:– Determine normative gastrocnemius firing pattern

over wide speed range (including slow)– Acquire pilot clinical data

EAC06–072

Gait


Healthy Raw Data: Two Walking Speeds

Footswitch

GastrocEMG

EAC06–073

Gait

Clancy et al., Am J Phys Med Rehabil 83: 507–514, 2004.


Healthy Gastroc EMGamp Templates

Heel Strike

EAC06–074

Gait

Clancy et al., Am J Phys Med Rehabil 83: 507–514, 2004.


Pilot Clinical Study• Subjects:

– 8 adults, > 2 years post-stroke, ambulatory• Methods:

– Treadmill walking, comfortable speed– EMG monitored, target EMG pattern shown– 12 training sessions (weekly, ~20–40 min)

• Result of before vs. after gait analysis–– ↑↑ Walking speed (5 subjects)–– ↑↑ Affected side single support (5 subjects)–– ↑↑ Stride length (4 subjects)–– ↑↑ Ankle power (7 subjects)

Aiello et al., 2006.Spaulding Hospital

EAC06–075

Gait


EMGamp: EMG-Torque Uses

EAC06–076

EMG-torque

• Non-invasive torque measurement for scientific studies

• Study/evaluation of worker safety– Repetitive, high-force tasks can lead to

cumulative trauma injuries


Simple Elbow Mechanics Model

EAC06–077

EMG-torque

Elbow joint flexedFlexor muscles contractedExtensor muscles relaxed

Elbow joint extendedExtensor muscles contracted

Flexor muscles relaxed

biceps

triceps

Clancy Ph.D., 1991

TNET = TF+TE

(www.mrothery.co.uk)


EMG-Torque Block Diagram

ExtensorEMG

AmplitudeEstimator

ExtensorEMG

Amplitude toTorque

Estimator

mE1,i

mE2,i

mELE,i

FlexorEMG

AmplitudeEstimator

FlexorEMG

Amplitude toTorque

Estimator

mF1,i

mF2,i

mFLF,i

Σ

ŝE

ŝF

ΤE

ΤExt

ΤF

_

+

……

• Relate biceps, triceps EMGamp to elbow torque• Compare conventional vs. advanced EMGamp processors

EAC06–078

EMG-torque


• Slowly force-varying No dynamics

• Constant Posture

EMGamp-Torque: Static Contraction

Clancy and Hogan, IEEE Trans BiomedEng 44: 1024–1028, 1997.

Tor

que

in A

/D U

nits

Tor

que

in A

/D U

nits

Time in Seconds

Time in Seconds

Unwhitened, Single-Channel Processor

Whitened, Four-Channel Processor

EAC06–079

EMG-torque


EMG-Torque: Constant-Posture, Force-Varying

5 1 0 1 5 2 0 2 5-4 0 0

-3 0 0

-2 0 0

-1 0 0

0

1 0 0

2 0 0

3 0 0

4 0 0E x te n s io n (T r ic e p s M /C )

T im e (s e c )

Raw

EM

G in

%M

VEE

5 1 0 1 5 2 0 2 5-4 0 0

-3 0 0

-2 0 0

-1 0 0

0

1 0 0

2 0 0

3 0 0

4 0 0F le x io n (B ic e p s L /C )

T im e (s e c )

Raw

EM

G in

%M

VEF

5 1 0 1 5 2 0 2 5-5 0

-4 0

-3 0

-2 0

-1 0

0

1 0

2 0

3 0

4 0

5 0R e a l T o rq u e a n d P re d ic te d T o rq u e N o rm a liz e d in % M V C

T im e (s e c )

Torq

ue A

mpl

itude

%M

VC

P o si tiv e T o rq u e - F le x io n

N e g a tiv e T o rq u e - Ex te n sio n

R e a l T o rq u e : M e a su re dP re d ic te d T o rq u e , U n w h i te n e d EM GP re d ic te d T o rq u e , W h i te n e d EM G

Clancy et al., J Biomechanics, 2006.

EAC06–080

EMG-torque


EMG-Torque: Force-Varying, Results

0 5 10 15 20 25 3020

30

40

50

60

70

80

90 M e d ian o f % V A F , F as t T rack in g S p e e d

S ys te m Id en tific a tio n O rd e r, n b

%VA

F

0 5 10 15 20 25 305

10

15M e d ian o f M A E , F ast T rack in g S p e e d

S y ste m Id e n tific a tio n O rd e r, n b

MA

E (P

erce

nt)

0 5 10 15 20 25 3020

30

40

50

60

70

80

90 M e an o f % V A F , F ast T rack in g S p e e d

S ys te m Id en tific a tio n O rd e r, n b

%VA

F

0 5 10 15 20 25 305

10

15M e an o f M A E , F as t T rack in g S p e e d

S y ste m Id e n tific a tio n O rd e r, n b

MA

E (P

erce

nt)

S -CH-UNW HITS -CH-W HITM -CH-UNW HITM -CH-W HIT




55%

80%

10%

7.3%

Clancy et al., J Biomechanics, in press.EAC06–081

EMG-torque


EMG-Torque Summary

• Future directions– Applications in ergonomics

• Constant-posture tasks– Relax postural constraints– Multi-joint systems

Better EMGamp Better EMGamp Better TorqueBetter Torque

EAC06–082

EMG-torque


EMG-Impedance Block Diagram

EMGAmplitude toImpedanceEstimator

k (stiffness)

ExtensorEMG

AmplitudeEstimator

mE1,i

mE2,i

mELE,i

ŝE…

FlexorEMG

AmplitudeEstimator

mF1,imF2,i

mFLF,i

ŝF…

b (viscosity)

J (inertia)

Higher-quality EMG amplitude estimates should give better impedance estimation.

Second-order, linear model; constant operating point.

EAC06–083

EMG-impedance


Elbow Mechanical Impedance Measurement

Subject seated, shoulder 90o

abducted, elbow 90o flexed.

Right hand immobilized in cuff (2), connected to actuated joystick (1)

Medio-lateral pseudo-random forceperturbations (3)

Measure medio-lateral movementthrough joystick encoders

Assume second order linear system: k, b, J

Top ViewSlide courtesy Denis Rancourt.

EAC06–084

EMG-impedance


System Validation for Mechanical Impedance

Known inertia, stiffness

0–15 Hz torque input (20 sec)

For 0, 20 and 40 degrees

2% variation in stiffness

% VAF = 92 %

( )( )

−

−=

∑

∑

=

=n

kk

n

kkk

f

ffVAF

1

2

1

2

ˆ

ˆ

1*100%

10 10.5 11 11.5 12 12.5 13 13.5 14 14.5 15

-10

-5

0

5

10

Temps (s)

Forc

e (N

)

Force x brute

Slide courtesy Denis Rancourt.

EAC06–085

EMG-impedance


Impedance Project Status• Much characterization of mechanical

apparatus performance• Data collected from ~16 subjects

EAC06–086

EMG-impedance

• Analysis inprogress


Motor Unit Decomposition

EAC06–087

EMG Arrays

Sweep speed = 10 ms/div; Sensitivity = 1.0mV/div

Motor unit potential with reinnervationDecomposition Shows:

• Firing times, firing shapes

• Scientific, physiologic, clinical information

(http://www.neuro.wustl.edu/neuromuscular/pathol/diagrams/linkedpot.htm)

Partial Denervation: EMG

• High amplitude, long duration, polyphasic, late potentials


Electrodes for EMG Decomposition

• Needles (traditional)+ High spatial resolution

(↓ superimpositions)– Invasive patient discomfort; cover small region

• Surface arrays (experimental)+ Noninvasive+ Larger spatial range

fiber conduction velocity– Less spatial resolution– Mostly record surface fibers

EAC06–088

5 mm

www.neurosupply.com

EMG Arrays


• Spatially selective; Resolve individual MUs– Locate neuromuscular junction– Measure conduction velocity– Non-invasively decompose EMG

• Output = Weighted sum of mono. electrodes

High-Resolution Surface EMG

( ) ( ) ( )tetetVFilt 21 −=

( ) ( ) ( ) ( )( ) ( )te4 5⋅−+

++=te

tetetetVFilt

4

321

–2

+1

+1

–1

+1

BipolarLinear Double

Difference (LDD)Normal Double

Difference (NDD)

–4+1

+1

+1

+1

EAC06–089

EMG Arrays

( ) ( ) ( ) ( )tetetetVFilt 321 2 +⋅−=


Array EMG Hardware• Common mode rejection ratio (CMRR):

– Requires precise weights– Rejects power-line; permits spatial combination

• For high CMRR, weight channels in hardware– Separate analog circuit for each channel– Too many combinations possible!

• So, either:– Pre-wire a few montages– Accept lower CMRR from software combination

NOTE: Problem is an instrumentation issue.

EAC06–090

EMG Arrays


Software-Equalized Array

Analog Signal Conditioning

Circuit #1e1

Linear Equalization

Filter #1A/D EMG1


Circuit #2e2

Linear Equalization

Filter #2A/D EMG2


Circuit #NeN

Linear Equalization

Filter #NA/D EMGN

... ...

Hardware Software

MonopolarElectrodeInputs

EqualizedMonopolarChannels

EAC06–091

EMG Arrays


Equalization CMRR Results

• Hardware: 1% resistors, 5% capacitors

• Testing at 60 Hz

• Sampled at 4096 Hz

5 mm

Bipolar Montage

EAC06–092

EMG Arrays

Desire CMRR ≥ 90 dB


=

)()()()()()()()()(

332313

322212

312111

tetetetetetetetete

Data

• Simulate each monopolarmonopolar surface electrode– Many xyz tripole (AP) locations within muscle

• Select desired spatial filterdesired spatial filter output voltage– Ones for regions of interest– Zeroes for regions of non-interest

• Solve for optimal weights using least squares– Monopolar potentials & output voltages known– Filter weights unknown

• Example:

Pilot Spatial Filter Technique

EAC06–093

EMG Arrays

)()(

)()()(

11113131

21211111

tewtew

tewtewtVSpatial

++

++=

L

{ – Similar to DSPlinear filter design


Example Results

• Filtered Surface Potentials

– Top left: Monopolar Filter3 dB Area = 35.22 mm2

– Top right: NDD Filter3 dB Area = 22.24 mm2

– Bottom Left: LS Filter3 dB Area = 8.95 mm2

• Second selectivity metric

– Bottom right: Case of two APs, LS and NDD Filters

• Y-offset = 2 mm• LS Ratio = 0.121• NDD Ratio = 0.779

EAC06–094

EMG Arrays


Conclusions of Pilot Work• Preliminary simulation results:

– Least squares filters have superior spatial selectivity than NDD, monopolar configurations

– Only simple simulation results so far!!

• Selectivity of LS filters dependent on:– Inter-electrode spacing, action potential depth, number of

electrodes

• Next steps:– More realistic electrophysiologic models– Incorporate measurement noise– More mature optimization– Test with physiologic data

EAC06–095

EMG Arrays


“EMGlab” on the Web

– Open-source software for EMG decomposition/ viewing

– EMG signal data base

– Data and annotation standards development

EAC06–096

EMGlab

A forum for sharing software, data, and information related to EMG decomposition

http://emglab.stanford.edu/ Kevin McGill, VA Palo Alto & Stanford


Research Collaborators

EAC06–097

Collaborators

… with several photos missing.… and, hopefully, many more to add in the years to come!!

extracting information from the electrical activity of …ted/sigma_xi_055.pdfextracting information...

Documents