Automated decomposition of electromyographic signals recorded with surface electrode arrays

• Main Author: Chen, Yunquan
• Language: English
• Published: 2009
• Online Access: http://hdl.handle.net/2429/6993

This thesis presents results of various investigations which led to the development and validation of a new system for multi-channel surface electromyographic (EMG) recording for clinical examinations (McSERCE) and of a new method for multi-channel surface EMG decomposition (McSED). To obtain individual motor unit action potentials (MUAP5) non-invasively, we invented 1) a brush-tip pin (BTP) electrode made of a bundle of fine wires for contacting the skin without electrolytic jelly; 2) a miniature active surface electrode array consisting of the BTP electrodes and signal buffers built into the device for sampling the MUAPs distributed on the skin; and 3) a surface EMG recording method which makes use of the bipolar electrode configuration for spatial filtering and the Hilbert transform for correcting for the phase-frequency non-linearity of the bipolar electrode transfer function. Based on these novel electrodes and recording method, the McSERCE system can be used to obtain MUAPs with high selectivity, resolution and signal-to-noise ratios. To extract individual MUAPs from EMG signals recorded on the skin with the McSERCE system, McSED, an effective, efficient, and robust multi-channel surface EMG decomposition method, was developed. The McSED method incorporates the following new techniques: 1) the detection of MUAPs using a multi-pass scheme which locates and matches action potentials detected simultaneously at slightly different locations on the skin; 2) the representation of MUAPs with their spatial distribution and temporal propagation patterns and times of occurrences; 3) the measurement of dissimilarity between MUAPs in terms of these patterns and times of occurrences; 4) the classification of MUAPs using both the “chaining” and the “dissection” characteristics of the single and complete linkage clustering methods, respectively; and 5) the estimation of the firing parameters of an incomplete MUAP train using a maximum-likelihood estimator. Using McSERCE and McSED, a clinical study was conducted on the MU conduction velocities of the abductor pollicis brevis muscle. Results showed that 2% of the 894 MUs detected in the muscles of 32 control subjects were “slow MUs” (v < 2.5 m/sec), whereas 27.7% of the 141 MUs detected in the muscles of four patients with myopathic disorders were slow MUs.