Computer Aided Assessment of Certain Neuromuscular Disorders Based on Surface Electromyography

The objective of this study was to investigate the usefulness of Surface Electromyography (SEMG), in the clinical laboratory. Towards this goal, recordings from 94 normal subjects (NOR) and 20 neuromuscular subjects (11 myopathy (MYO) and 9 neuropathy (NEURO)) were recorded at 10, 30, 50, 70 and 100% of maximum voluntary contraction. These signals were analysed in the time, frequency and bi-frequency domains. The time-domain parameters, turns and zero crossings per second, increased with force level (FL), for all three groups. Frequency-domain, median and mean frequencies, decreased with FL for the NOR group. Frequency at maximum power was unaffected with FL, but differences were found between all groups of subjects. Maximum and total power increased with FL for all groups, with MYO subjects having significantly higher values for the latter parameter. Bi-frequency parameter, bispectrum peak amplitude increased with FL only for the NOR group and significant differences were found between all three groups. Significant differences between the Gaussianity of the NEURO group and the other two groups were found. The location and slope of the bispectrum peak amplitude was unaffected with FL, however significant differences between the NEURO group and the other two groups were found. Repeatability and reliability measurements on a selected group of subjects gave very good results. Correlation analysis on all possible parameter pairs was performed for normal values. Excellent and good results are presented and compared with the corresponding pairs for patient data, where differences for certain parameters were found. It was shown that all parameters investigated were not affected by age or gender. Cluster analysis using the KNN algorithm was used to perform two (a. NORMAL/ABNORMAL (MYO and NEURO) and b. MYO/NEURO), and three class (NOR/MYO/NEURO) classification models. The percentage of correct classifications score for NORMAL/ABNORMAL and MYO/NEURO models were 3 94.4% and 70% respectively. The percentage of correct classifications score for the three class models varied from 59.3% when an equal number of subjects (9 NOR/9 MYO/9 NEURO) was used to 82.9% when 91 NOR/11 MYO/9 NEURO subjects were used. Self Organising Feature Maps give visually the mapping of two (a. NORMAL/ABNORMAL and b. MYO/NEURO) and three (NOR/MYO/NEURO) class models, with the former group presenting a better discrimination. Based on the findings of this study, it is concluded that the SEMG system investigated can be used successfully and effectively as an additional tool by the neurophysiologist for the assessment and monitoring of muscle function. It is proposed, that the system can be used “blindly” without sharing any prior information or knowledge about the possible classification of the subject, by the referring neurophysiologist