Differential Diagnostic Methods for Cognitive Disorders Using Neural Networks Based on Electroencephalogram Signals: A Systematic Review

1. Alzheimer Disease 2. Cognitive Dysfunction 3. Electroencephalography Introduction: Alzheimer’s disease is a brain disorder that gradually des troys cognitive function and eventually the ability to carry out daily routine tasks. Early diagnosis of this disease has attracted the attention of many physicians and scholars, and several methods have been used to detect it in early phases. Evaluation of artificial neural networks is low-cos t with no side effect method that is used for diagnosing and predicting Alzheimer’s disease in subjects with mild cognitive impairment based on electroencephalogram signals. Materials and Methods: for this sys tematic review, keywords Alzheimer’s, Artificial Neural network and EEG were searched in IEEE, PubMed central, ScienceDirect, and Google Scholar databases between 2000 to 2019. Then, they were selected for critical evaluation based on the mos t relevance to the subject under s tudy. Results: The search result in these databases was 100 articles. Excluding unrelated articles, only 30 articles were s tudied. In the present study, different types of artificial neural networks were described, Next, the accuracy of the classification obtained by these methods was inves tigated. The results have shown that some methods, despite being less used in research or have simple architecture, have the highes t accuracy for classification. In many s tudies, artificial neural networks have also been considered in comparison with other classification methods and the results show the superiority of these methods. Conclusion: Artificial neural networks can be used as a tool for early detection of Alzheimer’s disease. This approach can be evaluated for its classification accuracy, speed, architecture, and common use. Some networks are accurate at classifying and unders tanding data, but are slow or require specific hardware/software environments. Some other networks work better with simple architectures than complex networks. Furthermore, changing the architecture of conventional networks or combining them with other methods resulted in significantly different results. Increasing accuracy of classification of these networks in the diagnosis of mild cognitive impairment could help to predict Alzheimer’s disease appropriately. ABSTRACT Article Info: