Abstract:

This study organized a classification structure consisting of five processing stages for the diagnosis of ALS and Myopathy diseases, the most common variety of neuromuscular diseases. The EMG (Electromyogram) signs have been passed through the pre-treatment, division, accumulation and authentication stages before classification. In the stage of the processing, the method was used. The time then tested a total of 25 vectors, including the vectors of specificity in the frequency space and the vectors of multiple specificity obtained by their different combinations. In the next stage, the data is classified with the Support Vector Machine (DVM), the K-Last Neighbor (K-EYK) algorithm and the Diskiriminant Analysis (DA) algorithm. A cross-validity method has been used as a verification measure. The value of k is selected 10. Experimental findings have shown that among the recommended authenticity vectors, multiple authenticity vectors are more successful than single authenticity vectors. In the case of the use of multi-specific vectors, the DVM classifier has classified the EMG signs with higher accuracy compared to the K-EYK and DA classifiers. The total accuracy is 97.39% for ALS, while 86.74% for myopathy. This study has shown that the success of classification depends on the high degree of inter-class separability of the subjectivity vector.

Keywords: