In article <93012.181409BOYDJ at QUCDN.QueensU.CA> Jeff Boyd <BOYDJ at QUCDN.QueensU.CA> writes:
>>I didn't receive any replies to this question; in the hope that any
>interested parties may still have been away on holidays, I'm posting
>>> I find myself needing a mathematical rule for determining the 'end'
>> of a motor unit action potential. For example, I could arbitrarily
>> decide that a pulse is effectively over when, on its final phase, the
>> amplitude has fallen to 1% of the peak-to-peak amplitude.
>>>> Is there a standard definition which has received general acceptance
>> among EMG researchers?
We record EMGs in our lab, but we've never had the need to define
the end of a single motor unit action potential; nevertheless, talking
to someone here who has worked with such things I gathered this much:
First of all, the motor unit action potential shape can be complex.
It can have several zero crossings or "turns". If you do have a clean
signal, the best way to do what you want to is:
1. define a baseline (mean +/- sd) when there are no motorunits.
2. then use some criterion such as "amplitude has fallen to within
one standard deviation of baseline". The tricky part will be to
see how many consistent "turns" you have in the action potential
waveform and to see you include all such turns.
If you want to get more detailed, you can fit a regression line to the
last consistent "turn" in your action potential and determine where the
slope meets the baseline. That would define your "end".
The bottom line seems to be that there is NO standard way of defining
this. I'm assuming you need this so that you can isolate different
action potential shapes from the same electrode. Let me know if you
need a better explanation of what I've just said and I'll try to
provide one! Also, be warned once again that I'm no authority on this!
(venk at u.washington.edu)
PS: You may also wish to consult the literature on template matching
for action potential recognition - usually applied to neural recordings.
These days people use "fancy" methods such as wavelet analysis.