View Full Version : Summary of replies: EMG Data Presentation

10-20-2009, 08:20 PM
Dear all,
Thanks to all those who took time to reply to the original thread posted below. As can be seen from the summary of replies, there still remains discussion surrounding best methods for presentation and interpretation of EMG data, which may indeed continue for some time.
With assistance here we will endeavour to bring some of the methods described into our clinical practice.
Thanks again,

++++++++++++++++++++++++++++++++++++++++++++++++++ ++++++++++++

Dear Subscribers,

We are currently carrying out an audit of our service and
are in discussion with referrers and multi-disciplinary professionals with
regard to the presentation of Electromyographic data. Currently, we present EMG data as activity
only (i.e. burst (or not) of muscle on or off).
As most will be aware, this is useful information in a largely useless
from. I am writing to request your input
with regard to how clinical gait laboratories present EMG data, and how this is
perceived and received by the interpretation teams.

I think the question of how best to display EMG data may be
a long one unanswered, yet would appreciate any information you are willing to

With best regards

Mark Tucker

Pre-registered Clinical Scientist


South Birmingham Community Health

Clinical Measurements

West Midlands Rehabilitation

91 Oak Tree Lane

Selly Oak

Birmingham B29 6JA

++++++++++++++++++++++++++++++++++++++++++++++++++ +


Stuart Miller Brunel University


Whilst analysing muscle use during gait, I have found it helpful to present it
as a simple line graph with % cycle against normalised EMG.

This way, you can average the data from a series of consecutive strides, and it
gives you the story of what the muscle is doing through the stride. If you can
combine this with data of the muscle length change on the same diagram, you can
get a very good idea of how the muscle is working, when it is, and at what
intensity (much like the recent work of Komi and colleagues looking at the
Gastrocnemius - by attaching an ultrasound imaging device to the lower limb
over the muscle).


++++++++++++++++++++++++++++++++++++++++++++++++++ ++++++++++++++++++++++

Julie Stebbins Nuffield Orthopaedic Centre, Oxford

Hi Mark

We currently also only present raw (high pass filtered at 60hz) data. We
present 3 gait cycles from each trial, and 4 trials for consistency
check. We then only interpret timing of muscle activation
(subjectively). We are also currently reviewing what we do with EMG
presentation of data.

Best wishes


Dr Julie Stebbins

Clinical Scientist and Operational Lead

Oxford Gait Laboratory

++++++++++++++++++++++++++++++++++++++++++++++++++ ++++++++++++++++++++++

Edmund Cramp Motion Lab Systems, LA

Hi Mark,

We (Motion Lab Systems) are a company that make multichannel EMG systems, and
sell EMG software, so this is a question that I too have been interested in
over the years. I've worked closely with many people in the clinical gait
world and I've maintained an interest for a long time in trying to make sure we
do the best job possible. My conclusions to date on the presentation of
clinical data are:

Most clinicians in the USA
look at ON/OFF timing of the muscle with respect to the gait cycle and compare
this to "normal" On/OFF times. This sounds easy but raises a
couple of issues:

1. What are the "normal" ON/OFF
activity times?

2. How do we determine muscle firing

Our EMG software supports a variety of activity timings and allows users to
enter their own "normal" activity times if they don't like the ones
that we provide. In the USA most people seem to trust David Sutherlands
data for young children and either the published data from Jackie Perry or the
Hagy-Mann-Keller data from the San Francisco Shrine Hospital - there's a good
summery of the history here in Gait and Posture 14 (2001) 61-70. It's
worth noting that several Motion Capture vendors have released software
packages with "normal" EMG timing data that differs substantially
from these sources and that the published Rancho data itself is an ensemble

Determining when a muscle is ON or OFF is relatively easy - if you have good
EMG data to start with. But getting good EMG data means that you have to
be confident that you have placed the EMG sensors accurately on the muscle to detect
the correct muscle EMG - *AND* - that you are recording a good clean EMG
signal. Performing a couple of MVC tests can help verify that you're
recording a signal from the correct muscle and reviewing the raw EMG signal is
the only way to guarantee that you're making ON/OFF decisions on real EMG data.

On the assumption that you have good EMG raw data and a set "normal"
EMG activity times that are comfortable with then all that you have to do is
process the raw data to obtain an averaged envelope and pick an envelope level
that we'll call "ON" and "OFF" - but remember that
decisions about filtering of the EMG data and the envelope method used will
affect the ON/OFF times.

As a result, in our EMG Graphing and Analysis software, we prefer two display methods
- our software defaults to displaying the raw EMG together with
"normal" bars from one of the published trusted data sources or the
users own trusted activity timing; and secondly, raw EMG from individual
subject gait cycles overlaid with "normal" activity bars and the
ON/OFF timing detected in the subject EMG. We feel that these are the
most honest display methods in that they enable the clinician to assess both
the timing and the probability that the subject ON/OFF timing data is accurate.

As far as other display methods go, we offer a multi-level ON/OFF activity
report that displays EMG activity as several levels - I've always thought that
this is a nice idea but clinically it's not been much appreciated. I
suspect that this is simply because it presents too much information.
I've always thought that the methods used by Gage et al at
Newington/Connecticut Children's Hospitals where the EMG activity is overlaid
on the kinematic data is very useful but again, clinically there does not seem to
have been much acceptable of this - possibly information overload again.
Other display methods seem to come and go but to my mind most of the fancier
ones all fall short on verifiability. Pretty reports are not, by their
nature, more trustworthy than the traditional methods and often less so.

The comments above apply strictly to the Clinical Presentation of EMG
data. Permission granted to repost this reply to BIOMCH-L as part of a
summary of replies.


Edmund Cramp - eac@motion-labs.com

++++++++++++++++++++++++++++++++++++++++++++++++++ ++++++++++++++++++++++

Gasper Morey Klasping

Hello Mark,

perhaps your question is not precise enough. Presentation of EMG data should be
dependent on the question behind your EMG recordings.

In gait Analysis it is quite common to time normalize the EMG to the gait cycle
or even split into ground contact and swing time. At the same time the signal
is amplitude normalized to some sort of reference measurement or baseline
(which is valid in many cases but not always).

On the other hand when studying trips, or some other kind of responses to
perturbations, time normalization could flaw the results, so, and this is my
personal opinion, I would NOT time normalize the results, since the temporal
information is relevant. However an amplitude normalisation is necessary if we
aim to do any kind of comparison. So there is need to define a well described
reference contraction of the involved muscles.

Personally I like to smooth the signal strong enough to allow a point by point
(i.e. millisecond per millisecond) comparison of the means. It works pretty
good and the plots are very intuitive and easily understood. The interpretation
behind will obviously depend on the experiment itself and the questions to be

Good luck,