Desıgn of an EEG devıce for real-tıme and aı-based analysıs of mental dısorders

This study aims to develop a machine learning model for the automatic identification of six distinct neurological and neuropsychiatric disorders, specifically Alzheimer's, Depression, Epilepsy, Migraine, Parkinson's, and Schizophrenia, using EEG signals. The methodological framework of the study is based on a careful preprocessing procedure, followed by the extraction of a rich set of statistical features from the delta and theta frequency bands of the EEG data. These extracted features are designed to capture subtle biological differences between the diseases. To assess the classification performance of the extracted features, three widely used algorithms—Random Forest (RF), XGBoost, and Multilayer Perceptron (MLP)—were applied. The experimental results revealed that among the algorithms, the RF model demonstrated superior performance, achieving an overall accuracy of 89,47%, followed by XGBoost at 82,38% and MLP at 68,19%. These findings provide empirical evidence that ensemble learning methods can serve as a strong alternative to traditional singular or deep learning models in the development of complex EEG-based multi-diagnosis systems. Class-based performance analysis showed that the model achieved particularly high classification accuracy for healthy controls and Alzheimer's patients, while the performance was relatively limited for the Depression and Schizophrenia groups, likely due to the neurobiological complexity of these disorders. This study, unlike approaches in the literature that typically focus on binary classification (healthy vs. diseased), presents a holistic machine learning framework capable of distinguishing a wide range of neurological and psychiatric disorders. In conclusion, the thesis scientifically demonstrates that it is both feasible and reliable to differentiate these disorders based on EEG signals, offering a potential tool to support clinical diagnostic processes.