Sliding window averaging for the extraction of representative waveforms from motor unit action potential trains. (May 2016)

• Record Type: Journal Article
• Authors: Malanda, Armando; Rodriguez-Carreño, Ignacio; Navallas, Javier; Rodriguez-Falces, Javier; Porta, Sonia; Gila, Luis
• Is Part Of: Biomedical signal processing and control. Volume 27(2016)
• Page Start: 32
• Page End: 43
• Publication Date: 2016-05
• Subjects: EMG -- Averaging -- MUAP -- Waveform -- Sliding window; DEP Derivative error power -- EA Ensemble averaging -- EMG Electromyography -- FCA Five-closest averaging -- GSMW Gold standard MUAP waveforms -- MA Median averaging -- MUAP Motor unit action potential -- MWP MUAP waveform parameters -- NEP Normalized error power -- NPM Number of potentials per MUAP -- REP Residual error power -- SLER Significantly large errors range -- SPMF Signal processing merit figures -- SWSA Sliding window selective averaging; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2016.01.003
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: A new averaging algorithm for the extraction of representative waveforms from MUAP trains is presented. The algorithm is based on selection and averaging segments of potentials within the local scope of a sliding window. On normal muscles, it performs better than other relevant algorithms, regarding signal processing and quantitative MUAP waveform figures of merit. The algorithm requires a smaller number of potentials in the MUAP train, to estimate quantitative MUAP waveform parameters. Abstract: In quantitative electromyography (EMG), the set of potentials that constitute a motor unit action potential (MUAP) train are represented by a single waveform from which various parameters are determined in order to characterize the MUAP for diagnostic analysis. Several methods that extract such a waveform are currently available, and they are, in essence, based on two operations: averaging and selection, which are performed either sample-by-sample or on the whole-potential. We present a new approach that carries out selection and averaging on a local interval basis. We tested our algorithm with a dataset of MUAP records extracted from the tibialis anterioris muscle of healthy subjects and compared it with some of the most relevant state-of-the-art methods considered in a previous work (Malanda et al., J. Electromyogr. Kinesiol., 2015). The comparison covered general purpose signal processing figures of merit and clinically used MUAP waveform parameters. Significantly better results in both sets of figures of merit were obtained with the new approach. In addition, relative to the other algorithms tested, the new approach required fewer potentials from the MUAP set to obtain an accurate representative waveform.