The data do look a bit anomalous to me, but it's been a long time since I've looked at raw EMG data. I'm not 100% confident that this caused the spikes in your case, but I have seen very large "spikes" occur due to poor adhesion of the surface electrode to the skin. (I assume you are using surface EMG). Sometimes, when the electrode would break contact with the skin, our system would interpret this as abnormally huge numbers, and the signal was clipped (i.e. 100% of the max voltage). Any chance this might have happened in your case?

Hi Sean,

Thank you for getting back to me.

I would not rule out that the electrode may have broken off the skin, but we did use tubular taping to help fixate the surface electrode onto the skin to try to prevent this from happening.

If you know of any other possible cause please let me know so I can try to prevent the problem occurring again! :)

Maybe a simple contact with cables or electrodes (is normally for the first experience).

Please send me a photo of the experiment.... When is possible to see muscle EMG procedure

In my opinion, the better way to determine whether the signal is abnormal is to view the raw signal. Some details can provide information.

It's probably motion artifact of electrodes relative to skin or contact of wires. What filtering technique are you using? Although it's best practice to prevent collection of these artifacts during collection, a band pass filter (20-500hz) might help.

You can see SENIAM Web site to solving this gup

If you have maximum voluntary contraction recordings for the same muscle, comparing the data you have with those could give you a better idea if these spikes are indeed artifacts. Knowing the activation level of the muscle when the data is recorded and having a value (i.e. max. voluntary contraction) to normalize that activation level may help you better determine the nature of the spikes you see... but these do seem like artifacts, perhaps the electrode cables tugging on the electrodes? Taping the cables to the subject may help with that, if you haven't done that already....

I would suggest that you need to normalise your EMG to some maximum value, whether it be during a maximum contraction or the maximal M wave following stimulation of the peripheral nerve. If the maximum EMG during the movement is above 100% this is likely to indicate that the responses are due to an artefact associated with electrode or cable displacement.

Agree with Sean Osis, the most plausible answer is that both spikes are artificial. A simple and effective way to control EMG signals for artifacts is simply listening to the signal at real-time using the EMG-system's loudspeaker. Make yourself familiar with the sound of EMG and with the sound of typical artifacts be producing them intentionally, for example by shaking the electrode cables or the electrodes.

The big spikes indeed suggest artifacts, caused by moving of the cables or the electrode skin contact. Basically, you cannot 100% determine whether this concerns artifacts, when you use the processed (full wave rectified) EMG signals. You need the original raw signals and when you do a power spectrum anaysis, the most simple thing is to determine the amount of power below 10 Hz, If this is more then 5 % these are certainly artifacts

Hi Charlotte: This looks like an artefact of some sort to me - either movement of wires or electrodes. Strenuous skin prep may help. Muscles don't activate in this pattern even with lunges. Do you have some measure of the movement (e.g. limb position, velocity or acceleration) which would allow you to correlate the "spike" with the phase of movement?

Dear Charlotte,

this looks like an artefact, You shoul fornish the raw data and a phto of the recording to better understyand.

Happy new year,

Ferdi

Dear Charlotte,

the best guess is that those are artifacts. Depending on the procedure the raw data went through those may be eliminated in some of your tries but not in others.

We had the same problem a couple of years ago. Since then we always fix the cables of the electrodes with a small loop behind the electrode so that movements cannot "pull" on the cable itself.

Happy new year.

Chris

I agree with everyone else here, these do seem to represent artifacts. In addition to the suggestions made by others:

1. Normalize against MVC or activation during a standard task

2. Purposely create suspected artifacts in your system so that you can recognize them in the data you have collected

3. Look At the raw data when looking for artifacts; your post-processing is spreading the effect of the artifacts across a time window that is wider than the artifact itself (i.e. 100 ms window).

4. Although band pass filtering may help some, big spikes like that are going to excite your whole frequency spectrum. It would be better IMHO to reject epochs containing artifacts based on your raw signal.

In addition I would provide a bit more information:

5. What was the system you were using? Amplification level, active vs. Passive electrodes? Wireless (e.g. Noraxon) or wired? Filtering applied within the system?

6. I noticed that you had markers In your plot. What do they represent?

7. Is the second signal in your for the same subject or activity In the same muscle in the contra-lateral limb?

Very high peaks, more than 10x the usual signal are always suspect, but in some cases they turn out to be reflexes. You should always be prepared for electrode artifacts. Especially in movements with sudden impacts, like your lunge action, they are hard to avoid and impossible to remove by a linear filter. Make sure that the electrodes are glued on very well and that there are no dangling cables.

I can give two hints for the present case.

- Listen to the signals by means of a loudspeaker, a box with a good bass reproduction. The normal EMG sounds like a distant rumble, electrode artifacts like sharp cracking sounds.

- Inspect the signal before high-pass filtering, if that is possible. An electrode artifact shows itself usually as a sudden shift of the DC level. After HP filtering a peak results.

If you have a synchronized video recording with your measurements you can maybee see if there is something special that happenes that causes these spikes.

I think there is always a good idea to have a synchronized video together with the data for quality control of data. There are easy and feasible ways of synchronize video with data offline without expensive measurement systems.

/Peter

Hello,

Firstly, I discerned two distinct diagrams in the file you attached; in essence, they are more prominent under spikes. Secondly, following the notion of motion artifacts in previous posts, and your comments that lunge movement was performed repeatedly, it is expected to observe spiky patterns, repeatedly as well.

Taken together, I think that definite conclusions should be avoided before frequency analysis of your data (raw data).

Regards,

Dear Charlotte,

You ask if there is a way of determining if the EMG is of poor quality. I simply need to see the data as is, with no pre-processing (apart from applying standard low and high pass filters).

Eleftherios Papathanasiou

Clinical Neurophysiologist

Nicosia, Cyprus

This is an interesting problem that many face. The issue with biological signals is that it can be hard to discern the signal and noise (or artifact). Some knowledge of the signal can help of course. As noted, the bandwidth of EMG is pretty much known, with the majority of the power between 20-300 Hz (roughly, some will say as low as 1 Hz or 10 Hz and as high as 500 Hz; since you seem to be interested in relative EMG signal power, I would suggest filtering between 30-300 Hz since that is where the majority of the power would be and outside that range should be little signal power). If the large spikes are due to things outside of this range, you can be pretty sure that they are artifacts (e.g., motion artifact due to movement of the electrodes, cables, etc.). You can test this quite easily by bandpass filtering the signal prior to processing it (you should be doing that anyways). As noted the raw data and power spectrum would help discern what might be doing on, as would knowledge of the experimental setup (e.g., electrode type).

If anyone is interested, I am pursuing some research on signal quality assessment, which is trying to find ways to detect errors in EMG data!

Without having raw data it is difficult to tell about EMG signal. However I think it is a artifact. You must see raw signal frequency spectrum in the first step to detect any artifact and noises and their harmonies. Then you can use appropriate filters to reduce their effect on EMG signal

Dear Charlotte,

At least, in the masticatory muscles, these irregularities can be related to a somatic situation, for example facial paralysis, already resolved but with sequels on the electric microvoltage count. Have you explored this option?

Kind regards,

A. Gilbert

Eem to be a noise that can be caused by mateiral shoe or the sole of which does not necessarily isolating the earth.

After removing the DC offset, the EMG signal should be properly filtered to remove low-frequency motion artifacts. Did you apply such filtering? If not, your EMG-recording may be contaminated by such artifacts (see attachments, see also the guidelines for appropriate conditioning of EMG signals which can be found in the rear of each issue of the Journal of Electromyography and Kinesiology).

Anton van Boxtel

Here is another attachment.

EMG muscle shows the pathological involvement. Muscle strength does not give objective information about.

Hi Everyone,

To clarify a few points on the EMG system and set-up:

Surface EMG activity was recorded using a wireless telemetric system (Telemyo2400 G2, Noraxon) . Each electrode had a differential amplifier attached (input range +/- 3.5 mV, input impedance > 100MOhm, CMR > 100 dB, base gain of 500). The analogue signal was hardware bandpass filtered (10Hz-500Hz) converted to a digital signal by the transmitter data acquisition system (16 bit). Additionally the signal was processed using a digital bandpass filter (10-400Hz).

Is it still possible to have noise from cables with a wireless system?

No MVC was performed this time because firstly it would have been too time consuming with the experimental set-up and secondly because I prefer to normal to a peak or average peak of a normal condition (depending on the set up). It can be hard for participants who are not trained to produce their maximum contraction, especially in the smaller muscles. However, for a signal quality check - it would be useful.

The raw data is attached, it is still fairly hard to determine whether there is an abnormality or not due to the variability of EMG data. It probably is...

Inspecting the power frequency spectrum is a good tip for the frequency cut-off values is a good tip ti use.

As is listening to the EMG's loudspeaker.

Thank you for all the advice.

Charlotte

you have to take care when using MVC to normalise the EMG. Basically you do this in order to decrease intersubject variability, but it might also decrease the contrast between a patient group and healthy controls. This happens when both the EMG during MVC and the experimental task are equally influenced by the condition of the patient. The ratio (%EMG at MVC) is then going towards a normal ratio. We encountered this several times, especially in patients with neurological conditions or pain. In these conditions MVC is also difficult to get in a reliable way.

Hi Charlotte,

yes it is still possible to get artifacts from cables even in a wireless system, since there has to be a cable connection between the recording site and the radio transmitter.

But knowing that you use a wireless system: this can also be a source of artifacts if there are other electrical sources around and the signal itself is not coded.

Kind regards. Chris

It is better if you have more set of data by repeating the experiment several times or having more number of subject. With that you can know if there is data that is unusual. So this would go back to the experiment protocol. Perhaps, the subject experience something unusual during the experiment, for example the muscle might be twitched unexpectedly.