Tweet
Dear Biomch-L members,
Following is a summary of replies to my question regarding amplitude
normalization during walking. The initial question was:
Dear Biomch-list members,
I am interested in quantifying co-activation of antagonistic muscles
during walking in 2 different adult population groups. Surface EMG
will be used to measure the EMG signals from selected muscles during
the activity, and the raw EMG will be processed to derive a
time-series RMS signal for each muscle. The time-series RMS signals
will be temporally normalized with regard to the cycle time, i.e. to a
percentage of the gait cycle.
My question is, should the amplitude of the EMG measured from a muscle
during walking be normalized to a percentage of either: a) the peak
EMG amplitude measured during a maximal voluntary isometric
contraction, b) the peak EMG amplitude measured during a maximal
voluntary isotonic contraction, or, c) the peak EMG amplitude measured
during the activity?
>From my review of the literature, it appears that methods a) & c) are
the most common and both appear to have their strengths and
weaknesses. Any comments on which of the above methods are:
- the most valid,
- show the greatest intra-subject reliability and
reproducibility,
and/or
- show the least inter-subject variability,
would be greatly appreciated. Also if anyone has used or knows of
other methods to normalize the amplitude of dynamic EMG during
submaximal activities, please let me know.
I will post a summary of responses in a fortnight.
Thank-you
Peter Mills
Following are the replies that were received. Thanks to all.
---------------------------------------------------------------------
Peter:
I have found the following references useful.
Regarding reliability of normalization schemes:
Yang, J.F., & Winter, D.A. (1984). Electromyographic amplitude
normalization methods: Improving their sensitivity as diagnostic tools
in gait analysis. Archives of Physical Medicine and Rehabilitation, 65
(9), 517-521.
(This reference was also supplied by Birgitte van Don)
Birgitta J.C. van Don, Ph.D.
Boston University
NeuroMuscular Research Center
Motor Control Laboratory
44 Cummington Street>Boston, MA 02215
Phone (617) 353-8984
Fax: (617) 353-5737
e-mail: bvandon@bu.edu
Regarding a useful index of coactivation:
Falconer, K. & Winter, D.A. (1985). Quantitative assessment of
co-contraction at the ankle joint in walking. Electromyography and
Clinical Neurophysiology, 25, 135-149.
It sounds like you are doing something similar to something I've just
done. I worked in collaboration with a colleague to estimate indices
of coactivation during normal and pathologic gait at a variety of
walking speeds. Here was my general procedure:
1. Normalize all EMG data with respect to time (step time or stride
time). This is straightforward. 2. Normalize the amplitude EMG data
from each muscle with respect to the mean ensemble average (6 stride
average) for condition 1 (lowest walking speed). Therefore, at
subsequent walking speeds all EMG data are expressed as a percent of
the baseline speed. I have found coefficients of variation to be
lowest for this normalization scheme. It is also much less time
consuming than any normalization scheme involving MVCs. 3. Compute
coactivation indices using time and amplitude normalized EMG profiles
using the method described in Falconer & Winter (1985). I favor this
method because it is not affected by errors associated with the
identification of muscle activity "onsets" and "offsets". However, you
should be cognizant of the fact that many muscles of the lower
extremity are biarticular in nature and therefore may confound your
ability to make a straightforward interpretation of any coactivation
index that uses EMG data from a biarticular muscle.
Good luck,
Peter
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Peter F. Vint, Ph.D.
Department of Exercise and Sport Science
256 HHP Building, PO Box 26169
University of North Carolina at Greensboro
Greensboro, NC 27402-6169
Phone: (336) 334-3031
Fax: (336) 334-3031
E-mail: pfvint@uncg.edu
URL: http://www.uncg.edu/~pfvint
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
---------------------------------------------------------------------
--------------------------------------------------------------------
To my opinion, the c option is the most suitable for a main reason:
Specificity of pattern.
If the aim is to express EMG normalized i think it can be misleading
to normalize it to a measure which does not have anything to do with a
dynamic performance such as gait. However it presents some problems.
In literature most of the research has been conducted normalizing EMG
to MVC or % of MVC, however it has been shown that MVC can be affected
by a lot of parameters (motivation most of all and familiarity with
the testing procedure). Second, MVC is performed at a specific limb
angle....then how it can represent EMG baseline activity ? I think it
is better to normalize to peak amplitude measured during activity.
Marco Cardinale, M.Sc.
---------------------------------------------------------------------
---------------------------------------------------------------------
> M a r c o
>Thank-you for your reply. I tend to agree with you for
the reason >that you stated, and also because by normalizing using
the peak EMG >measured during the activity, any intersubject
differences with >regard to the maximal amount of force developed by
the individual >muscle (during an MVC) will not alter the
intra-subject comparisons.
This is another good point.
> Using MVC to normalize muscle activity
>during a dynamic task, 2 subjects with equal absolute dynamic force
>patterns during an activity but differences in absolute maximal force
>production (during MVC), of a muscle, would be identified as having
>differences in muscle activation patterns, because the muscle with
>the greater potential force production will have a lower activation
>level at any point during the activity.
I agree, this is why, to me, it is important to normalize EMG to some
dynamic values when dynamic tasks have to be measured. However in
literature most of the people used MVC. I was forced to use MVC
normalization by my tutor in the M.S. Thesis....but I disagree
completely with the results. I was measuring effects of dynamic
exercise...and they obliged me to perform isometric MVC measurement...
Anyway, I think from now it is better to use peak value to normalize
activities such as gait.
>Thank-you and I would welcome any further comments
If something comes from my mind or come up from literature I will keep
you updated. All the best for your research, Marco Cardinale
---------------------------------------------------------------------
Peter,
Did you check the biomechan-list archives? There are a few entries in
the last couple of years summarizing viewpoints on normalizing EMG.
Also, I suggest:
Electroencephalogr Clin Neurophysiol 1987 Nov;67(5):402-11
EMG profiles during normal human walking: stride-to-stride and
inter-subject variability.
Winter DA, Yack HJ
************************************************** ***********
Gordon Chalmers, Ph.D.
Dept. of Physical Education, Health and Recreation
Western Washington University, MS-9067
Bellingham WA
U.S.A.
98225-9067
chalmers@cc.wwu.edu
http://www.wwu.edu/~chalmers
Phone: (360) 650-3113
FAX: (360) 650-7447
---------------------------------------------------------------------
--------------------------------------------------------------------
Gordon,
Thank-you for your reply. I have located summaries regarding
amplitude normalization of dynamic EMG in the Biomch-L archives by
Roger James (1996) and Kevin Sims (1997) that have been helpful.
My concerns regarding the 'best' normalization method stems partly
from an article: Knutson, L. M., Soderberg, G. L., Ballantyne, B. T.,
& Clarke, W. R. (1994). A study of various normalization procedures
for within day electromyographic data. Journal of Electromyography
and Kinesiology, 4(1), 47-59.
This article normalized EMG activity measured from the medial
gastrocnemius during a balance board activity using 3 amplitude
normalization procedures: 1) as a % of the peak ensemble averaged
value measured measured during a MVIC, 2) as a % of the peak ensemble
averaged value measured during the activity, and 3)as a % of the mean
ensemble averaged value measured during the activity. They reported
the lowest intra-subject CV's for the 2nd method, but also reported
higher VR's and lower ICC's for the peak activity normalized data than
for the MVIC normalized data. However, as only 2 trials for each
subject were used, these results may not fully represent the
reliability and reproducibility of these methods.
I personally think it is more appropriate to represent muscle
activation during a dynamic task as a percentage of the maximal muscle
activation during that task, because using a percentage of the EMG
measured during a MVIC can create differences in normalized activity
between individuals that are due solely to differences in the maximal
potential activation level of the muscle measured during an MVIC.
I greatly appreciate your reply and would welcome any further thoughts
ideas.
Peter
---------------------------------------------------------------------
---------------------------------------------------------------------
I am not an expert in this area, but, I notice that Winter DA, Yack HJ
(reference given before) (and Winter is an expert) agree with you.
See>page 405 - normalizing to MVC increased variability observed, so
they normalized to the average value seen during the step cycle. I
know (but don't have references at hand) that others have nomalized
to max seen in step cycle as you suggest. So I suggest you pick one
such method that is established by a prominent publication(s) and/or
researcher and go with it. In fact, normalizing to mean or max level
seen in step cycle is something you can compare in your preliminary
analysis, because both can be done from the same data after the data
is collected.
Good luck GC
Gordon Chalmers, Ph.D.
Dept. of Physical Education, Health and Recreation
Western Washington University, MS-9067
Bellingham WA >U.S.A.
>98225-9067 > >chalmers@cc.wwu.edu >http://www.wwu.edu/~chalmers
>Phone: (360) 650-3113 >FAX: (360)
---------------------------------------------------------------------
---------------------------------------------------------------------
From: "a.l.hof"
Organization: faculty of medical sciences (RuG)
To: p.mills@MAILBOX.GU.EDU.AU
Date sent: Wed, 16 Jun 1999 09:31:07 GMT+0100
Subject: Re: EMG normalization during walking
Priority: normal
Dear peter, I am late in answering, but this is because your question
requires quite a long answer. > > I am interested in quantifying
co-activation of antagonistic muscles during > walking in 2 different
adult population groups. Surface EMG will be used to > measure the
EMG signals from selected muscles during the activity, and the > raw
EMG will be processed to derive a time-series RMS signal for each >
muscle. The time-series RMS signals will be temporally normalized
with > regard to the cycle time, i.e. to a percentage of the gait
cycle. > > My question is, should the amplitude of the EMG measured
from a muscle > during walking be normalized to a percentage of
either: a) the peak EMG > amplitude measured during a maximal
voluntary isometric contraction, b) the > peak EMG amplitude measured
during a maximal voluntary isotonic contraction, > or, c) the peak EMG
amplitude measured during the activity? > > From my review of the
literature, it appears that methods a) & c) are the > most common and
both appear to have their strengths and weaknesses. Any > comments on
which of the above methods are: > > - the most valid, >
- show the greatest intra-subject reliability and reproducibility, >
and/or > - show the least inter-subject variability, > > would
be greatly appreciated. Also if anyone has used or knows of other >
methods to normalize the amplitude of dynamic EMG during submaximal >
activities, please let me know.
What method is the best, depends very much on the subject of your
research. 1) In general, the best EMG calibration is by using a
submaximal isometric contraction with slowly increasing force. This
can give an rmsEMG/force or - moment ratio. (J Biomech 10:529-539,
1977). 2) Scaling with EMG at MVC is much more convenient, no
dynamometer needed, but not accurate. Maximum EMGs are very variable.
3) For walking I use in many cases a regression of the individual
averaged profile Ei with the 'normal' profile E , obtained from a
group of 'normal' subjects.
gain = sum( Ei*E / E^2)
Then (Ei/gain) gives a profile of the same magnitude as the normal
one. Disadvantage: cocontracting muscles are not quantified in moment.
4) For your work, if you cannot do 1), you may record a situation in
which two antagonist muscles are working against each other " keep the
foot / knee as stiff as possible". The muscles are working against
each other, but the net moment is zero. What is the ratio of the EMGs.
In the past several studies have come with a measure of cocontraction
in which just raw µVs of EMG were used. This is of course incorrect.
As you will see, there is not one best answer.
Good luck anyway,
Best wishes,
At Hof
Department of Medical Physiology &
Laboratory of Human Movement Analysis AZG
University of Groningen
Bloemsingel 10
NL-9712 KZ GRONINGEN
THE NETHERLANDS
Tel: (31) 50 3632645
Fax: (31) 50 3632751
e-mail: a.l.hof@med.rug.nl
---------------------------------------------------------------------
---------------------------------------------------------------------
The end. Thanks again!
Peter Mills PES GUGC
p.mills@mailbox.gu.edu.au
---------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
---------------------------------------------------------------
Following is a summary of replies to my question regarding amplitude
normalization during walking. The initial question was:
Dear Biomch-list members,
I am interested in quantifying co-activation of antagonistic muscles
during walking in 2 different adult population groups. Surface EMG
will be used to measure the EMG signals from selected muscles during
the activity, and the raw EMG will be processed to derive a
time-series RMS signal for each muscle. The time-series RMS signals
will be temporally normalized with regard to the cycle time, i.e. to a
percentage of the gait cycle.
My question is, should the amplitude of the EMG measured from a muscle
during walking be normalized to a percentage of either: a) the peak
EMG amplitude measured during a maximal voluntary isometric
contraction, b) the peak EMG amplitude measured during a maximal
voluntary isotonic contraction, or, c) the peak EMG amplitude measured
during the activity?
>From my review of the literature, it appears that methods a) & c) are
the most common and both appear to have their strengths and
weaknesses. Any comments on which of the above methods are:
- the most valid,
- show the greatest intra-subject reliability and
reproducibility,
and/or
- show the least inter-subject variability,
would be greatly appreciated. Also if anyone has used or knows of
other methods to normalize the amplitude of dynamic EMG during
submaximal activities, please let me know.
I will post a summary of responses in a fortnight.
Thank-you
Peter Mills
Following are the replies that were received. Thanks to all.
---------------------------------------------------------------------
Peter:
I have found the following references useful.
Regarding reliability of normalization schemes:
Yang, J.F., & Winter, D.A. (1984). Electromyographic amplitude
normalization methods: Improving their sensitivity as diagnostic tools
in gait analysis. Archives of Physical Medicine and Rehabilitation, 65
(9), 517-521.
(This reference was also supplied by Birgitte van Don)
Birgitta J.C. van Don, Ph.D.
Boston University
NeuroMuscular Research Center
Motor Control Laboratory
44 Cummington Street>Boston, MA 02215
Phone (617) 353-8984
Fax: (617) 353-5737
e-mail: bvandon@bu.edu
Regarding a useful index of coactivation:
Falconer, K. & Winter, D.A. (1985). Quantitative assessment of
co-contraction at the ankle joint in walking. Electromyography and
Clinical Neurophysiology, 25, 135-149.
It sounds like you are doing something similar to something I've just
done. I worked in collaboration with a colleague to estimate indices
of coactivation during normal and pathologic gait at a variety of
walking speeds. Here was my general procedure:
1. Normalize all EMG data with respect to time (step time or stride
time). This is straightforward. 2. Normalize the amplitude EMG data
from each muscle with respect to the mean ensemble average (6 stride
average) for condition 1 (lowest walking speed). Therefore, at
subsequent walking speeds all EMG data are expressed as a percent of
the baseline speed. I have found coefficients of variation to be
lowest for this normalization scheme. It is also much less time
consuming than any normalization scheme involving MVCs. 3. Compute
coactivation indices using time and amplitude normalized EMG profiles
using the method described in Falconer & Winter (1985). I favor this
method because it is not affected by errors associated with the
identification of muscle activity "onsets" and "offsets". However, you
should be cognizant of the fact that many muscles of the lower
extremity are biarticular in nature and therefore may confound your
ability to make a straightforward interpretation of any coactivation
index that uses EMG data from a biarticular muscle.
Good luck,
Peter
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Peter F. Vint, Ph.D.
Department of Exercise and Sport Science
256 HHP Building, PO Box 26169
University of North Carolina at Greensboro
Greensboro, NC 27402-6169
Phone: (336) 334-3031
Fax: (336) 334-3031
E-mail: pfvint@uncg.edu
URL: http://www.uncg.edu/~pfvint
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
---------------------------------------------------------------------
--------------------------------------------------------------------
To my opinion, the c option is the most suitable for a main reason:
Specificity of pattern.
If the aim is to express EMG normalized i think it can be misleading
to normalize it to a measure which does not have anything to do with a
dynamic performance such as gait. However it presents some problems.
In literature most of the research has been conducted normalizing EMG
to MVC or % of MVC, however it has been shown that MVC can be affected
by a lot of parameters (motivation most of all and familiarity with
the testing procedure). Second, MVC is performed at a specific limb
angle....then how it can represent EMG baseline activity ? I think it
is better to normalize to peak amplitude measured during activity.
Marco Cardinale, M.Sc.
---------------------------------------------------------------------
---------------------------------------------------------------------
> M a r c o
>Thank-you for your reply. I tend to agree with you for
the reason >that you stated, and also because by normalizing using
the peak EMG >measured during the activity, any intersubject
differences with >regard to the maximal amount of force developed by
the individual >muscle (during an MVC) will not alter the
intra-subject comparisons.
This is another good point.
> Using MVC to normalize muscle activity
>during a dynamic task, 2 subjects with equal absolute dynamic force
>patterns during an activity but differences in absolute maximal force
>production (during MVC), of a muscle, would be identified as having
>differences in muscle activation patterns, because the muscle with
>the greater potential force production will have a lower activation
>level at any point during the activity.
I agree, this is why, to me, it is important to normalize EMG to some
dynamic values when dynamic tasks have to be measured. However in
literature most of the people used MVC. I was forced to use MVC
normalization by my tutor in the M.S. Thesis....but I disagree
completely with the results. I was measuring effects of dynamic
exercise...and they obliged me to perform isometric MVC measurement...
Anyway, I think from now it is better to use peak value to normalize
activities such as gait.
>Thank-you and I would welcome any further comments
If something comes from my mind or come up from literature I will keep
you updated. All the best for your research, Marco Cardinale
---------------------------------------------------------------------
Peter,
Did you check the biomechan-list archives? There are a few entries in
the last couple of years summarizing viewpoints on normalizing EMG.
Also, I suggest:
Electroencephalogr Clin Neurophysiol 1987 Nov;67(5):402-11
EMG profiles during normal human walking: stride-to-stride and
inter-subject variability.
Winter DA, Yack HJ
************************************************** ***********
Gordon Chalmers, Ph.D.
Dept. of Physical Education, Health and Recreation
Western Washington University, MS-9067
Bellingham WA
U.S.A.
98225-9067
chalmers@cc.wwu.edu
http://www.wwu.edu/~chalmers
Phone: (360) 650-3113
FAX: (360) 650-7447
---------------------------------------------------------------------
--------------------------------------------------------------------
Gordon,
Thank-you for your reply. I have located summaries regarding
amplitude normalization of dynamic EMG in the Biomch-L archives by
Roger James (1996) and Kevin Sims (1997) that have been helpful.
My concerns regarding the 'best' normalization method stems partly
from an article: Knutson, L. M., Soderberg, G. L., Ballantyne, B. T.,
& Clarke, W. R. (1994). A study of various normalization procedures
for within day electromyographic data. Journal of Electromyography
and Kinesiology, 4(1), 47-59.
This article normalized EMG activity measured from the medial
gastrocnemius during a balance board activity using 3 amplitude
normalization procedures: 1) as a % of the peak ensemble averaged
value measured measured during a MVIC, 2) as a % of the peak ensemble
averaged value measured during the activity, and 3)as a % of the mean
ensemble averaged value measured during the activity. They reported
the lowest intra-subject CV's for the 2nd method, but also reported
higher VR's and lower ICC's for the peak activity normalized data than
for the MVIC normalized data. However, as only 2 trials for each
subject were used, these results may not fully represent the
reliability and reproducibility of these methods.
I personally think it is more appropriate to represent muscle
activation during a dynamic task as a percentage of the maximal muscle
activation during that task, because using a percentage of the EMG
measured during a MVIC can create differences in normalized activity
between individuals that are due solely to differences in the maximal
potential activation level of the muscle measured during an MVIC.
I greatly appreciate your reply and would welcome any further thoughts
ideas.
Peter
---------------------------------------------------------------------
---------------------------------------------------------------------
I am not an expert in this area, but, I notice that Winter DA, Yack HJ
(reference given before) (and Winter is an expert) agree with you.
See>page 405 - normalizing to MVC increased variability observed, so
they normalized to the average value seen during the step cycle. I
know (but don't have references at hand) that others have nomalized
to max seen in step cycle as you suggest. So I suggest you pick one
such method that is established by a prominent publication(s) and/or
researcher and go with it. In fact, normalizing to mean or max level
seen in step cycle is something you can compare in your preliminary
analysis, because both can be done from the same data after the data
is collected.
Good luck GC
Gordon Chalmers, Ph.D.
Dept. of Physical Education, Health and Recreation
Western Washington University, MS-9067
Bellingham WA >U.S.A.
>98225-9067 > >chalmers@cc.wwu.edu >http://www.wwu.edu/~chalmers
>Phone: (360) 650-3113 >FAX: (360)
---------------------------------------------------------------------
---------------------------------------------------------------------
From: "a.l.hof"
Organization: faculty of medical sciences (RuG)
To: p.mills@MAILBOX.GU.EDU.AU
Date sent: Wed, 16 Jun 1999 09:31:07 GMT+0100
Subject: Re: EMG normalization during walking
Priority: normal
Dear peter, I am late in answering, but this is because your question
requires quite a long answer. > > I am interested in quantifying
co-activation of antagonistic muscles during > walking in 2 different
adult population groups. Surface EMG will be used to > measure the
EMG signals from selected muscles during the activity, and the > raw
EMG will be processed to derive a time-series RMS signal for each >
muscle. The time-series RMS signals will be temporally normalized
with > regard to the cycle time, i.e. to a percentage of the gait
cycle. > > My question is, should the amplitude of the EMG measured
from a muscle > during walking be normalized to a percentage of
either: a) the peak EMG > amplitude measured during a maximal
voluntary isometric contraction, b) the > peak EMG amplitude measured
during a maximal voluntary isotonic contraction, > or, c) the peak EMG
amplitude measured during the activity? > > From my review of the
literature, it appears that methods a) & c) are the > most common and
both appear to have their strengths and weaknesses. Any > comments on
which of the above methods are: > > - the most valid, >
- show the greatest intra-subject reliability and reproducibility, >
and/or > - show the least inter-subject variability, > > would
be greatly appreciated. Also if anyone has used or knows of other >
methods to normalize the amplitude of dynamic EMG during submaximal >
activities, please let me know.
What method is the best, depends very much on the subject of your
research. 1) In general, the best EMG calibration is by using a
submaximal isometric contraction with slowly increasing force. This
can give an rmsEMG/force or - moment ratio. (J Biomech 10:529-539,
1977). 2) Scaling with EMG at MVC is much more convenient, no
dynamometer needed, but not accurate. Maximum EMGs are very variable.
3) For walking I use in many cases a regression of the individual
averaged profile Ei with the 'normal' profile E , obtained from a
group of 'normal' subjects.
gain = sum( Ei*E / E^2)
Then (Ei/gain) gives a profile of the same magnitude as the normal
one. Disadvantage: cocontracting muscles are not quantified in moment.
4) For your work, if you cannot do 1), you may record a situation in
which two antagonist muscles are working against each other " keep the
foot / knee as stiff as possible". The muscles are working against
each other, but the net moment is zero. What is the ratio of the EMGs.
In the past several studies have come with a measure of cocontraction
in which just raw µVs of EMG were used. This is of course incorrect.
As you will see, there is not one best answer.
Good luck anyway,
Best wishes,
At Hof
Department of Medical Physiology &
Laboratory of Human Movement Analysis AZG
University of Groningen
Bloemsingel 10
NL-9712 KZ GRONINGEN
THE NETHERLANDS
Tel: (31) 50 3632645
Fax: (31) 50 3632751
e-mail: a.l.hof@med.rug.nl
---------------------------------------------------------------------
---------------------------------------------------------------------
The end. Thanks again!
Peter Mills PES GUGC
p.mills@mailbox.gu.edu.au
---------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
---------------------------------------------------------------