I think not, unless you normalized it.

I hope these resources could help.

http://www.seniam.org

http://www.ncbi.nlm.nih.gov/pubmed/11018443

Of course you can use it, depending for what? If isometric not isotonic, to measure MVC (maximum voluntary cintraction)

I woud like to check the effect of the core training treatmants as far as the muscular endurance is concerned. Do we use peak, average values or should we use integrals for the 3 seconds surface?

I agree with Emmanuelle Cugy, you need to normalized the signal first

Hi Tijana, you will need to use an average of the signal over a selected time interval during your protocol - EMG is too variable a signal to base your interpretation off of a single peak value.  Consider using a root mean square (RMS) approach over this interval to calculate the average peak. 

Isometric contraction is more easy to evaluate with averaged signal like Emmanuelle and Zulkifli say, RMS as Michale says is better to use for intermittent contraction as by example from KeyPoint device with FFI protocol.

I thought you mean peaks not peak.

When needle EMG we assess muscle firing on the neuromuscular joint visually! and this is what a Neurophysiologist do.

Surface EMG is different and yes more NMJ are registered so you need to average them. Look methodology in both papers I am sending here.

But be aware amplitude is not always more power.. Look to frequency , because if you registered only few fibers NMJ this means that you have a damaged neuropathic root..

Maybe RMS is the best or surface EMG averaged if healthy subjects!

My best,

Ovi

Couple of factors to be considered for isometric contraction & surface EMG 

Factor 1: Maximum voluntary contraction levels - Low levels have a different behavior than at higher levels (70% -90%)

Factor 2: Duration of isometric contraction - There are distinct patterns during the first 3 seconds vs say after 60 seconds for the same muscle (refer work done by Farina/ Merletti/ Dinesh Kumar Kant)

Factor 3: Muscle under isometric - The Tibialis anterior characteristics are different from Biceps Brachii or Triceps brachii. The firing rate and recruitment pattern of motor units vary by muscle, levels of contraction, task, time-into-contraction

Factor 4: Normal contraction vs Fatigue - The characteristics of signal vary based on onset of fatigue or into fatigue stages

Factor 5: Subject is healthy or has neuromuscular condition vs Age group can influence the signal characteristics of sEMG signal

sEMG signals are nonstationary and it is also an electrical summation of motor unit action potential. There may be some instances of peaks due to its nonlinearity and non-gaussian nature.

Normalize or Not-Normalize - This may not change the peak. Instead of having -5.4 units to +7.8 units (for example), on normalization, we may end up between -1 to +1. Normalization against MVC makes more sense.

Lastly, muscle is a multidimensional complex structure. sEMG signals are 1 dimensional time series. So, we have already reduced the complex structure series to a simple one dimensional series. 

Is Peak value 'alone' is a good indication. Definite NO. We (NIID Lab at IIT Madras) have recorded sEMG signals from over 100 subjects / different muscles/ various contraction. Based on my experience this alone may not be suitable even when you normalize 

Try considering nonlinear/nonstationary features which has been proved to be more effective. Check the below references.

References:

M. González-Izal, A. Malanda, E. Gorostiaga, and M. Izquierdo, “Electromyographic models to assess muscle fatigue.,” J. Electromyogr. Kinesiol., vol. 22, no. 4, pp. 501–12, Aug. 2012.

R. Merletti and D. Farina, “Surface EMG processing: Introduction to the special issue,” Biomed. Signal Process. Control, vol. 3, no. 2, pp. 115–117, Apr. 2008

Marri, K., & Swaminathan, R. (2016). Analysis of Biceps Brachii Muscles in Dynamic Contraction Using sEMG Signals and Multifractal DMA Algorithm. International Journal of Signal Processing Systems, 4(1), 79–85. doi:10.12720/ijsps.4.1.79-85