PDA

View Full Version : EMG analysis software



uro1
08-29-2011, 10:10 AM
I was wondering if anyone knows of any EMG analysis software that is capable of processing large amounts of data. My goal is to capture and analyze tracings of about an hour or so at a time. Any help would be appreciated.

lroren39
09-05-2011, 08:09 AM
Dear Kyle,

Using a "normal" capture frequency of 1000Hz, an hour's worth of EMG data would mean 3600 x 1000 = 3,600,000 data points, or about 14MB of data per channel if each point is stored as a 4-byte floating number. This isn't a huge amount for modern PCs, so assuming you have enough RAM and harddisk space you should be fine. Some file formats used for storing the data (for example C3D) may have certain limitations to the number of fields that can be stored in the file, so this is worth bearing in mind.

The key question is what information you wish to extract from the data. Are you interested in "bursts" of activity that could happen sometime during the hour-long capture, but you're not sure when? Or are you planning to analyze the entire hour's worth of EMG to look for long-term trends, for example to see how the median or mean frequency of the signal changes?

In the first case, one question is whether the software lets you easily extract the short periods of useful information from the long periods of uninteresting data. If you graphed the hour's worth of data on a screen 1500 pixels wide, each pixel would correspond to more than 2s worth of data, and it may be very difficult to identify the periods of interest without having to manually zoom in on subsections of the data - which could be very tedious indeed. Therefore, some kind of automatic method for identifying muscle onset/off would be essential, which would use some signal processing and threshold detection to identify muscle activity. If the software is able to then calculate whatever information you need (e.g. the length and timing of each burst) and export this data in some convenient format, you'd be fine.

In the second case, the amount of data could become a problem since advanced calculations done on such a big dataset will be time-consuming. However, accepting that you may need to patiently wait for the results, this should also be doable, especially if the software lets you choose the interval between each calculation. In other words, assuming you're looking for a trend in the data over an hour, you're unlikely to need a value for each and every sample, so perhaps one value per second would be sufficient. This would dramatically cut the number of calculations the software would have to perform, but such a facility would need to be provided by the software, of course.

In summary, if you could provide some detailed information on the experiment you're intending to conduct, I may be able to comment further. I personally develop professional EMG software which is offered commercially by my company. I will post more details in the "Products and Services" forum about this if you would like further information.

jcannon41
09-05-2011, 08:09 AM
Hi Kyle,

I think with a file this size Matlab would be for best bet... Vert flexible and you can crunch the data any way you see fit...

Regards,
Jack