Detection of Dementia: Using Electroencephalography and Machine Learning

Abstract

Introduction This article serves as a background to an emerging field and aims to investigate the use of Electroencephalography signals in detecting dementia. It offers a promising approach for individuals with dementia, as electroencephalography provides a non-invasive measure of brain activity during language tasks. Method: The methodological core of this study involves implementing various electroencephalography feature extraction and selection techniques, along with the use of machine learning algorithms for analyzing the signals to identify patterns indicative of dementia. In terms of results, our analysis showed that most individuals likely to have dementia are in the 60-69 age bracket, with a higher incidence in females. Result: Notably, the K-means algorithm achieved the highest Silhouette Score at approximately 0.295. Additionally, Decision Tree and Random Forest models achieved the best accuracy at 95.83%, slightly outperforming the support vector machines and Logistic Regression models, which also showed good accuracy at 91.67%. Conclusion: The conclusion drawn from this article is that electroencephalography signals, analyzed with machine learning algorithms, can be effectively used to detect dementia, with Decision Tree and Random Forest models showing promise for future non-invasive diagnostic tools.