Automated Detection and Classification of Alzheimer's Disease from Brain Images using Machine Learning Techniques

Abstract

    Electroencephalography (EEG), a non-invasive technique, captures subtle voltage variations caused by ionic current flows within the neurons of the cerebral cortex. These recordings are invaluable for diagnosing brain disorders such as tumors and epileptic seizures. However, EEG signals are often distorted by ocular artifacts (OAs) caused by eye movements and blinking, which overlap with EEG signals of similar frequencies. The presence of these artifacts can significantly affect the accuracy of signal analysis and classification. This study proposes a two-step approach to enhance epileptic seizure classification. The first step involves the detection and removal of ocular artifacts from the UCI Epileptic EEG dataset using a combination of Independent Component Analysis (ICA) and Discrete Wavelet Transform (DWT), optimized with a tailored wavelet function. The second step employs a deep learning-based modified Gated Recurrent Unit (GRU) model to classify epileptic seizures. The results demonstrate that removing ocular artifacts improves signal clarity, yielding superior classification performance. The clean EEG dataset achieved a classification accuracy of 99.50%, with enhanced precision, recall, and F1-score metrics compared to the contaminated dataset. The modified GRU model proved effective in improving EEG-based epileptic seizure classification, highlighting its potential for reliable applications in Brain-Computer Interface (BCI) systems and advancing the field of medical signal processing.