Artificial Intelligence for the Classification of Neuromuscular Diseases Using Dominant MUAP

Motor-unit-action-potentials in an electromyographic isolated signals shape & sounds are valuable diagnosis information for neuro-muscular disorders-treatment& management. An expert can analyse these parameters qualitatively or quantitative by pattern recognition techniques. Because of benefits of the quantitatively method of EMG, producing robust automated MUAP types is investigated, & many systems is produced for this purpose, but correctness of the previous methods isn't peak enough to use in clinical settings. The developed system uses an EMG signal decomposition mechanism to extract both time &frequency domain for M-U-A-P's retrieved from E-M-G. Un-similar type of algorithms was studied, including 1 & many types with many sub-sets characteristics. The multi-classifier methods proposed here performed well in experiments by real set EMG. The multi-classifier, which aggregates base classifiers using multiple feature sets &both trainable &non-trainable fusion techniques, performed the best of the methods studied, with an average precision is 98%.