The impact of electrode reduction in the diagnosis of dyslexia

Abstract

Dyslexia is a learning disorder and involves disability in reading. It is a deficit with a brain origin despite the presence of good intelligence. Dyslexic patients may have lower rates of learning compared to healthy individuals of the same age. This is a critical problem in the learning process at school years, which makes it important to determine the origin of dyslexia in the brain for treatment. There are different methods to investigate how the brain works. One of these methods is to record brain signals (Electroencephalography (EEG)). Dyslexic children have shown some anxiety and restlessness due to inability to perform tasks properly. Thus, their additional movements may cause an error in the signal recording. Reducing the number of connections decreases the possibility of measurement errors in EEG recording. We determined the optimal group of electrodes for Identification Dyslexic Patients in this research. The reduction in the number of electrodes makes the test easier and more practical. Classification accuracy can also improve with the removal of irrelevant channels. Bhagavatula (2009) and Modrzejewski (1993) increased the accuracy of the classification by removing inefficient electrodes. For this purpose, we extracted the best features including RSP features, mean, standard deviation, skewness and kurtosis, hjorth and AR parameters. Then, both SVM and Bayes classifiers were used to separate two classes. We used Mutual Information (MI) to electrode reduction. The aim of the proposed method is to apply reduced electrodes on dyslexic children and reach acceptable results for diagnosis. Finally, we succeeded to reduce the number of electrode channels from 19 to 2-6 and attain a classification accuracy of 70%.