I send a sample file with this EMG sygnal. Channel nb 1 is VastusLateralis, Ch2= Rectus Femoris, Ch3=Vastus Medialis, Ch4=Biceps Femoris.

It is a text file, you can open it in Matlab, Office...

Strictly speaking, EMG signal as no reliable correlation with muscle force. When the task is dynamic, the correlation is even less likely to exist. In general, there is a disagreement in the literature as to which muscle fatigue inducing protocol is the best (see our recent paper https://www.researchgate.net/publication/230820140_The_effects_of_lower-extremity_muscle_fatigue_on_the_vertical_ground_reaction_force_A_meta-analysis). One option in your case would be the use of a maximum force task in the static condition before and after the application of fatigue protocol. You can then measure the drop in the level of maximum muscle force. Of course, it is not an optimal test, but there is not much else that you can do.

Article The effects of lower-extremity muscle fatigue on the vertica...

The frequency content of the EMG signal bears some indication of the level of fatigue. The mean (or median) power frequency can be determined with Fast Fourier Transformation (FFT) analyses but these are usually only applicable for so-called 'stationary' signals, i.e. from the isometric contraction you wanted to test but could not get from the bikers. I assume that some of the more sophisticated methods might be applicable for your dynamic recordings, maybe you can look into wavelet or principal component analyses.

Do you know JASA (http://www.sciencedirect.com/science/article/pii/S1050641100000225)? An advancement of the method (german language) at http://www.thumedi.de/THUMEDI_KG/Download/Messung_musk_Ermuedung_INET.pdf.

When isolating a small portion of the rotation it's perhaps useful, but for doing this you need angular information of the crank.

Dear Amir,

My knowledge tells me that the study of the EMG signal due to muscle fatigue is not the good solution in dynamic condition. So how do you write an option of measurement drop in the level of force is at least an objective measure of the whole system, but it is a comparison between the start and end point. I'm more interested in what happens during the exercise / test. Besides, I do not have at this point, the ability to measure torque and power itself, thank you for article.

Dietmar, I've never used to wavelet analysis, but has already used factor analysis. I also found a few of work, for example, Mark Latash, who used this type of analysis to examine muscle synergies. I need to delve into this subject.

Most cheered me answer Uli, because I see a proposed new way to solution of my problem. I have information about the angle of the knee and ankle so I guess I can use the JASA.

I would like to ask one more question:

if selecting a small slice of time in which the change in angle is low, could I calculate frequency and amplitude parameters of the EMG signal , and use them to determine muscle fatigue? Is this part of the observation can be regarded as stationary?

Greg

Despite of knowing, what's exactly your goal and focusing to the EMG-Problem in my opinion it's the only way to cut a piece of perhaps 20° around horizontal position of the crank, take the mean-values of 10 revolutions for amplitude and frequency from, the beginning and the end of your 90s trials and additionally leave ot the first and last 5s of the trial. That's all you can do to get an as much as possible reliable signal for your purpose. Is it perfect? No, but it's not too bad I think and if you don't have a better solution try this one ;-)

The load must be high is you want to see relevant changes in only 90s and the mesurement have to be of high quality, so you've no artifacts because of the movement - sorry, I'm not able to import the given data to excel correctly to check this in a rough way

You need to use a filter to avoid signal intereferences of heart and other muscles which you are not interested

Try to look any article of Hug and Dorel. They had a lot of researches using ergometers and EMG.

You can use time/frequency to analyze fatigue too.

von Tscharner has published a couple papers with very thorough analyses of EMG in hand muscles during fatigue protocols. They tend to favour wavelet analysis, but they have used other techniques as well.

http://www.ncbi.nlm.nih.gov/pubmed/21094054

http://www.ncbi.nlm.nih.gov/pubmed/17986395

Hi all,

thanks for next answers, It's helpful for me.

Renato - what you mean time/frequency analysis? It is analysis of mean/median frequency?

Uli - I try to do this in next few days (cut the signal regarding to position of crank) and compare it to present results from whole cycle of motion.

Thanks to all contributors

Greg

Sobota,

First, I think what Renato means by time/frequency is the decomposition of signal in time / frequency / amplitude. You can do that using either short time fourier transform or wavelet transform. My choice goes to wavelet. Like says by sean, you can see von Tscharner papers on that topic.

Dieter proposed you to use MPF or median frequency, but these measurements are discussed in dynamic condition. To my knowledge, there no consensus to use these parameters to quantify muscle fatigue except in some specific case like isometric contraction. Therefore i would prefer wavelet analysis.

Then, in my own research, in order to estimate the muscle fatigue through EMG measurements in dynamic condition, we've simply compared the amount of average EMG for one mouvement to the other (corresponding to one crank revolution in your case). This method could seems simplistic, but in dynamic condition as said by Amir previously, their is no good relationship between force and EMG (especially in stretch-shortening cycle movement like jumping, running and also cycling).

Regards,

Cédric

The relation between EMG properties and fatigue have been studied for many years. At the moment the subject is less en vogue, and with good reason. There are indeed distict changes in the EMG after long lasting activity: the amplitude increases and the power spectrum shifts to lower frequencies. But no consistent relation with fatigue has, to my knowledge, ever been verified.

As it seems the observed phenomena are related to the decrease in blood flow during muscle contraction. This is especially pronounced in continuous isometric contractions. In the more 'natural' dynamic and intermittent contractions such effects are much harder to find. A relation between blood lactate and EMG frequency changes could not be found (Jansen et al. 1997).

Refs: Ament et al. Eur J Appl Physiol 74:180-186 (1996)

Ament et al. J Electromyogr Kinesiol 3: 39-45 (1993)

Jansen et al. Eur J Appl Physiol 75:102-108 (1997)

I would suggest the recent paper in IEEE TNSRE (2011) by Kumar and Naik- regarding Synchronization, IIS index and endurance limit (alternate for muscle fatigue). If you are unable to find it, let me know.

Yes, please check the paper in IEEE TNSRE 2011 by Kumar, Arjunan and Naik on muscle endurance.

Tracking muscle fatigue via EMG is difficult in general. In a dynamic situation you have confounding factors such as change force and joint angle (MacIsaac et al. (EMBS Mag 2001) "Influence of dynamic factors on myoelectric signal estimation" provides a simple and succinct explanation). With isometric, isotonic contractions, conventional methods such as MF and MDF work okay. With dynamic contractions, recent work on multiparameter methods and multi-scale methods show some promise but there is no accepted consensus yet. Some of my lab's recent work includes "A Lempel-Ziv Complexity Measure for Muscle Fatigue Estimation" (JEK 2011)

The short answer is- yes, it is possible to identify muscle fatigue from EMG. However, spectral analysis does not work, and you should have two channels of EMG recorded from the same muscle. Then you measure the independence between these channels- you can do this using ICA or entropy- not using correlation. The results are very good, and it works for isometric or dynamic contraction, and also for rather low levels of muscle contraction. The technique is called IIS index. Check a reference in IEEE TNSRE 2011- Kumar et al. Let me know how it goes. Just make sure that the inter-electrode distance of the two channels is the same.