Probabilistic muscle characterization using QEMG: application to neuropathic mus

[Guest guest]

Muscle Nerve. 2010 Jan;41(1):18-31.

Probabilistic muscle characterization using QEMG: application to neuropathic

muscle.

Pino LJ, Stashuk DW, Boe SG, Doherty TJ.

Systems Design Engineering, University of Waterloo, 200 University Avenue West,

Waterloo, Ontario, Canada N2L 3G1.

Clinicians who use electromyographic (EMG) signals to help determine the

presence or absence of abnormality in a muscle often, with varying degrees of

success, evaluate sets of motor unit potentials (MUPs) qualitatively and/or

quantitatively to characterize the muscle in a clinically meaningful way. The

resulting muscle characterization can be improved using automated analysis.

As such, the intent of this study was to evaluate the performance of automated,

conventional Means/Outlier and Probabilistic methods in converting MUP

statistics into a concise, and clinically relevant, muscle characterization.

Probabilistic methods combine the set of MUP characterizations, derived using

Pattern Discovery (PD), of all MUPs detected from a muscle into a

characterization measure that indicates normality or abnormality.

Using MUP data from healthy control subjects and patients with known neuropathic

disorders, a Probabilistic method that used Bayes' rule to combine MUP

characterizations into a Bayesian muscle characterization (BMC) achieved a

categorization accuracy of 79.7% compared to 76.4% using the Mean method (P >

0.1) for biceps muscles and 94.6% accuracy for the BMC method compared to 85.8%

using the Mean method (P < 0.01) for first dorsal interosseous muscles.

The BMC method can facilitate the determination of " possible, " " probable, " or

" definite " levels for a given muscle categorization (e.g., neuropathic) whereas

the conventional Means and Outlier methods support only a dichotomous " normal "

or " abnormal " decision. This work demonstrates that the BMC method can provide

information that may be more useful in supporting clinical decisions than that

provided by the conventional Means or Outlier methods.