Deep Learning Based Processing of EEG Signals for Detection and Recognition of Alzheimer Disease

The aim of this study is to provide early diagnosis of Alzheimer's disease (AD). Twodimensional color image transformations of EEG signals will be processed for early diagnosis. Alzheimer's disease; It is a common type of dementia neurological disease that occurs in advanced ages and causes a decrease in thought, memory and behavioral functions. There is no definitive cure for Alzheimer's disease. However, it is possible to both slow down the process and reduce the severity of some symptoms. For this, early and accurate diagnosis is of great importance. EEG data were obtained from the dataset of 88 participants (35 healthy people, 31 with mild Alzheimer's disease and 22 with Alzheimer's disease) Alzheimer's disease and healthy people. EEG signals converted to Gramian Summation Angular Field images were firstly processed through various preprocessing steps. The 2D color image data obtained was trained and tested using the AlexNet deep learning model. AlexNet, a Convolutional Neural Network model, consists of 8 layers. In the literature review, 16 channels were selected in various studies and the 3 channels with the highest results were used. The same channels were also selected for this study. Among these channels, F7, T3 and T5 channels have the highest success rate. The three channels with the highest results were used in this study. GASF images of selected F7, T3 and T5 channels were used to train the AlexNet CNN model over 50 epochs. The developed model achieved promising performance with 98.03% accuracy, 98.21% sensitivity and 97.84% specificity. In addition, the AlexNet CNN model was trained and tested over 50 epochs with 4-fold Cross Validation. As a result of this study, the developed model achieved the highest results with 98.02% accuracy, 98.01% sensitivity and 98.04% specificity