This paper studies the application of machine learning in the analysis and diagnosis of electromyography data. Firstly, 2,352 electromyography examination reports have been recorded from Sichuan Provincial Hospital of Traditional Chinese Medicine for ten months. The data cleaning has been conducted based on the specific-designed inclusion criteria. Next, two data sets have been established, containing 575 facial motor nerve conduction study reports and 233 auditory brainstem response reports, respectively. And then, four machine learning algorithms including random forest, linear regression, support vector machine and logistic regression have been employed to the data sets. The performance comparisons of accuracy and recall rate among different algorithms indicate that the random forest algorithm has the optimal performance over the other two in both data sets. Moreover, the comparisons have been carried out in the cases with and without deviation standardization for each algorithm, and the results demonstrate that the deviation standardization has a certain effect on the accuracy improvement. Additionally, it is found that the random forest algorithm can present the ranking of the features in order of importance. Consequently, the random forest is proven to be an optimal algorithm for computer-aided diagnosis systems. Furthermore, it is worth mentioning that the feature ranking in order of importance can facilitate clinical diagnosis and has a certain clinical potential in diagnosis and diagnostic assessment.

No abstract

No abstract