PDA

View Full Version : Summary: techniques to reduce artefact content of EMG data(posted 12 sep 2002)



Andrew Chapman
09-18-2002, 03:44 PM
Thank you to all those who responded. A special thanks to those who offered
to view sample data.

Original message posted Thu, 12 Sep 2002 16:05:06 +1000 with subject
"techniques to reduce artefact content of EMG data" has been successfully
distributed to the BIOMCH-L list:

"I am attempting to collect telemetered EMG data during running, but the
artefact content of the data is overwhelming.

I have searched the Biomech-L archives for previous postings relating to
EMG noise content, but all suggestions mentioned in these postings have
failed to resolve my current problems.

The data is being captured with a Noraxon TeleMyo 900 8 channel
telemetered EMG system. This system has a fixed output voltage range (+/-
2.5 volts when in bipolar mode) and a fixed gain of 2000.

To date, all measures taken to reduce the noise content of the data have
failed. Any running, even at the slowest of speeds, causes clipping and
thus a high frequency artefact which cannot be filtered - rending the data
useless. Even a comparatively gentle foot 'stomp' causes an artefact of
amplitude beyond the range of the voltage band and thus clipping.

When the signal does not contain clipped data (for example: from isolated
limb movements), filtering is effective, and the quality of the data,
while not extraordinary, is reasonable. With a fixed output range and
fixed gain, I can not see any adjustments that can be made to expand the
voltage band and thus prevent the clipping without actually reducing the
noise. However, if such adjustments are possible, I would welcome advice
as to how these adjustments can be made.

Interestingly, the artefact content of the data is enhanced when fine wire
electrodes are used, but regardless, the artefact is at levels which
render the data useless even when surface electrodes are used as per the
intended use the Noraxon system (thus the fine wire electrodes themselves
do not appear to be the main/only source of the problem).

The fine wire electrodes used consist of 75 micrometer teflon coated
stainless steel with ~1mm bared tips for recording. Custom made connectors
were used to connect the wire to the Noraxon EMG cables, via an alligator
clip to allow connection to the electrode wires, and a 'male' press stud
to fit the 'female' press stud connection of the Noraxon cable.

All cables were secured to the subject as best as possible to minimize
movement, and all cables were wound around one another. Channels not in
use were shorted out with back to back surface electrodes. The cable was
secured to prevent tension on the input point to the transmitter (which
was strapped as tightly as possible around the subject's waist) and the
input point to the electrodes.

Varying the ground electrode position does not alter the artefact content
of the recorded signals.

I would appreciate any suggestions as to how I can reduce the noise
content of my EMG recordings? I can provide sample data if this will
assist others in determining the cause of this noise.

Regards,
Andrew Chapman

BPhty(Hons.)(UQ)
PhD Candidate
Department of Physiotherapy
The University of Queensland
St. Lucia, 4072, QLD, Australia.

and

Department of Physical Therapies
Australian Institute of Sport
P.O.Box 176
Belconnen, 2616, ACT, Australia.

email: andrewchapman@ozemail.com.au
phone: + 61 2 6214 7943
fax: + 61 2 6214 7953
mobile: + 61 438 115 605"
-----

Summary of responses:

-----
We currently use the 8 channel MIE telemetry system and don't have artefact
problems.
In the past with a wire based system or for smaller experiments we use Pals
electrodes which are a steel mesh woven into a cloth pad with a gel surface
that sticks to the skin very well. The electrodes are available from
http://www.axelgaard.com/ . I would also try with a non-telemetry system to
locate further the source of the artefact. In fact trying each part
individually should help here.

Hope this is useful,
Jonathan Norton.
-----
It sounds like you are getting motion artifact from movement between the
electrodes and the skin and/or movement of the electrode wires between the
electrodes and the preamp (you are using "passive" electrodes). With fine
wire electrodes the problem would be even greater since there is no
preamplification of the emg signal and movement of the wires would cause
artifact which would be amplified by the preamps. I havent worked with
telemetry systems so I dont know if the telemetry may add additional noise.

I don't know of any sure fire solution. If it is possible you may try to
increase the cutoff frequency of the highpass filter to take out more
motion artifact. Active electrodes where the preamp is built into the back
of the electrode may also help. I think the best you can do with what
you've got is to try to constrain the movement of any wires and electrodes
as much as possible, but I guess you've already done that.

Kelvin J. Chen [kelvin@ieee.org]
-----
Andrew,
I encountered similar difficulties when collecting data for my
dissertation with the same system. Interestingly, I had clean data for
surface EMG channels, but I was encountering the same impact spikes for
my single indwelling channel. I had a home-built interface unit for the
fine-wires, which I ended up replacing with a unit professionally built
from Run Technologies and that took care of my problem. I came to the
conclusion that the problem was at the interface between the wires and
the fine-wire adapter (where they snap together). I don't know how old
your cables are, but perhaps new cabling will help for all of the
channels. In general, the Noraxon unit provided clean data during
running after I fixed that one problem. I don't know if this is
helpful, but feel free to contact me if you have any questions.

Regards,
Kris
Kristian M. O'Connor, Ph.D.
Assistant Professor of Biomechanics
Department of Human Movement Science
University of Wisconsin - Milwaukee
P.O. Box 413
Milwuakee, WI 53201
(414) 229-2680
-----
Andrew,
The main issue in your case is that the gain is too high. 2.5V /
2000 = 1.25 mV
For surface EMG (not fine-wire) this might do, but in movement
there are artifacts of much higher amplitude.
Can Noraxon provide you with 200x amplifiers?
The artifacts, which are of high amplitude but low frequency, can
then be filtered out with a high-pass filter.
If the output with 200x preamp. is too low for your A/D, apply first
an analog high-pass filter and the amplify some 10x after the filter.
An other (but $$) option is to abandon the Noraxon system and
buy the PORTI system as we use. This system has some very
clever interference suppression techniques and 22-bits A/D.
Good luck,
************************************************** *****
At Hof
Institute of Human Movement Sciences &
Laboratory of Human Movement Analysis AZG
University of Groningen
A. Deusinglaan 1, room 321
postal address:
PO Box 196
NL-9700 AD GRONINGEN
THE NETHERLANDS
Tel: (31) 50 363 2645
Fax: (31) 50 363 3150
e-mail: a.l.hof@med.rug.nl
************************************************** *******
-----
Hi Andrew,
I am sorry you are having such difficulties. Has the unit made the
trip back to Noraxon? I suspect something major with the electronics, but
who knows. Can you collect walking data with acceptable results? Is it the
impacts that cause the clipping? Try leaving the unused channels/wires
unplugged--perhaps movement of these electrodes is the cause. Use coban or
some other stretchy tape to bind the wires to the runners legs. Do not wind
the wires to avoid an induction coil.
I have seen clipping, but only on a Noraxon system with two analog
channels designed for A to D (accelerometers, etc.). Never have I seen this
with EMG. If you don't get it figured out I'd ask Noraxon for a loaner and
send back the suspect unit.

Michael
Seattle VA Gait Lab

Michael Orendurff [morendurff@hotmail.com]
-----
Hi Andrew

Are your subjects running on a treadmill? Static electricity from the belt
can
cause huge artefacts.

Regards, Brian
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Brian L. Davis, Ph.D.
Department of Biomedical Engineering (ND20)
Cleveland Clinic Foundation
davis@bme.ri.ccf.org
http://www.lerner.ccf.org/bme/biomechanics/index.html

-----
Dear Andrew

Can you send me any noisy signal to analyze.
I hope that I can help you!

Sincerely

Rodrigo

Rodrigo Lício Ortolan [ortolan@sel.eesc.sc.usp.br]
-----
Dear Andrew,

Have you tried using an antialiasing lowpass filter before recording?, you
can
stablish the cut off frequency of the filter to the maximum emg frequency
that
you want as long as the sampling frequency you are using is at least 5 times
the desired frequency. This will avoid any higher frequencies to affect the
data. I am not familiar with that EMG system and dont' know if this filter
can
be added. My guess is to place it, if possible between the receiver and
the
computer (I am assuming you are using a computer to store your data)

Try avoiding the alligator clips, the movement between the alligator and the
electrode is a source of noice

I hope this helps

Kind regards

Jose

PS, Do you have more explicit specifications of this emg system? or a
website
where they can be found?

Jose Salazar [J.D.J.Salazar-Torres@newcastle.ac.uk]
-----
Hi Andrew.

I'm interpreting your note to mean that the raw EMG voltages from running
are so strong that they overload the signal conditioner and lead to
clipping.

If this is the case, why not reduce the magnitude of the signals before
they get to the EMG processing unit?

The signal path would look like:

wire electrode--> amplifier with gain 0.5 --> noraxon signal processing unit

The amplifier gain could be as small as necessary to avoid clipping.
This could introduce additional noise, and would reduce the signal content
but if it eliminated the clipping it might be an overall benefit. It might
also require a custom designed amplifier to make it small and portable, but
hopefully your university has some good electronics people. A first test
might be to use a research quality but full size amplifier in the lab and
see if it works for a foot stomp.

Paul Ostic

Paul Ostic [ostic@me.queensu.ca]
-----
Andrew,

My first thought is that the fixed gain of 2000 may be excessive for
recording EMG during high effort such as running. This would account for
the clipping. If you plot peak (rectified and smoothed) EMG amplitude vs.
gait speed, what does this curve look like? Can you verify or predict that
EMG amplitude will saturate for a certain level of effort, and clip above
that level? If so, you will need to reduce the gain for high-effort tasks.

Best wishes,

Andy Hoffer
-----
This is a very real problem especially with surface electrodes. If you
bump a set of electrodes will always get such an artifact. I suggest that
you try to prevent electrodes from rubbing against clothing etc. and space
pairs of electrodes far enough apart that they cannot hit each
other. Another solution might be to glue pairs together so that they move
as-one or obtain electrodes from Delsys which come paired.

I assume you are not using a treadmill, which may collect a static
charge. If so, try to ground the belt, like some trucks or high-speed
printers, with charge dissipators.

D. Gordon E. Robertson, Ph.D. [dger@uottawa.ca]
-----
Hi Andrew,

Are you sure the gain is fixed? I'm not sure what model our Noraxon is, but
the gains are adjustable. There are dip switches inside the transmitter box
on the waist band that allow us to alter gain.

There shouldn't be any movement artifact with the Noraxon. I don't know
what kind of filtering goes on, I imagine it is substantial, because we can
have quite gross cable movements and still find no artifact present.

Another thing to look for is a loose electrical connection somewhere. We
have had occasional problems with the battery pack causing the signal to
drop in and out. When it comes in, there is a big spike in the signal.

If you want to send me a text file containing some data you've collected, I
would be happy to have a look and try to interpret your problem. It's
really hard to guess the cause without seeing what the trace looks like.

Are you going to LaTrobe in November? I believe abstracts are due on Monday.

Cheers,

Peter Sinclair
Lecturer in Biomechanics
School of Exercise and Sport Science
The University of Sydney
East St Lidcombe NSW 1825
Phone (02) 9351 9137
p.sinclair@fhs.usyd.edu.au
-----
Andrew,

When we used alligator clips to connect to male press
studs of EMG electrodes we found that considerable
noise was introduced by the metal-on-metal
movement between the alligator clip and press-stud.
This was still present in walking with leads and pre-
amps securely fastened. This noise was removed by
simply securing the respective alligator clip to the
EMG electrode to prevent any movement between the
two.

Hope this helps

Allan Carman
Research Fellow
School of Physiotherapy
University of Otago
-----
Andrew

Perhaps what you are getting are the result of auto-gain control. And, this
could be compounded by motion artefacts. My suggestion is that try to repeat
the process with the muscle steady (isometric) and the load varies suddenly.
The result of this experiment may shed more light. Let me know.

I have never used this equipment though.

Cheers

Dinesh
Dinesh Kant Kumar [dinesh@rmit.edu.au]
-----
Hi Andrew

We recently upgraded our old Noraxon system (which we could alter the output
range) and discovered, to our horror, that the new "updated" Noraxon system
has several features that now make it not nearly as user friendly in
research (ie fixed output range; an inability to display the output on the
receiver of more than one channel at a time etc). We have also experienced
similar types of problems as you have with either extremely small signals
and clipping. If you would
like to give me a ring I can chat through some of the option we tried (with
varying levels of success) to obtain cleaner signals. Let me know via email
when a good time is to call.

Cheers

Julie Steele
-----
Dear Andrew,

My name is Peter Konrad and besides my own scientific work here in
Cologne/Dortmund - Germany, I am associated with Noraxon since 91, now being
responsible for the software design.

The Noraxon EMG telemetry system is a very well established standard in
numerous labs all over the world, so I must conclude that there is a local
problem in your lab. It may relate to hardware or handling problems.

I propose to first try the following tests, based on surface EMG:

1) Check system noise vs external noise

What is the average baseline noise if you short connect the two electrodes
of one lead? Too high noise could mean amplifier problems or heavy
interfering external noise (typically 50 or 60Hz ground noise which can
easily been check in the EMG power spectrum). Next step: is it constant over
all the channels you have? If so, it indicates external noise problems.

2) Noise Constancy

Assuming that you stay in battery mode(recommended because more stable
against ground noise interferences), move around (with the complete system),
check other rooms to verify that you do not struggle against local
electrical problems (=strong electrical motors, bad electricity conditions
etc...)
3) Skin impedance
Unfortunately you didn't describe your technique of skin preparation
(surface EMG), which should be checked prior to fine wire use. Did you clean
and slightly rubber the skin? Did you check the impedance between two
electrodes?
This could be an important point...
We do not use fine wire electrodes, but plan to do, so any exchange is
appreciated! But what we already learned is that it needs a high knowledge
and training. Obvisiously it is much more difficult to sample clean wire
EMGs than surface (where we never had a problem).

4) Records

You report about high amplitudes "clipped" away by the system signal range.
Did you properly setup the software A/D settings/Hardware/Device? Did you
run an FFT to check your power spectrum?

Anyways, yes I would like check some of your sample data
Best regards - Peter
Dr. Peter Konrad
KinLab - University of Dortmund FB 16 Sports Science Germany.
Konrad@sport.uni-dortmund.de
-----
End of message

---------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
---------------------------------------------------------------