Early Diagnoses of Alzheimer using EEG data and Deep Neural Networks classification

Alzheimer's disease has always been a challenge to be detected at early stages as it has always been mistaken as normal aging. It can be recognized when the patient starts to have Mild Cognitive Impairment (MCI) and by that stage, only little can be done as no treatment can reverse its effect but only to delay its progression. In this work, we present a method for early diagnoses of AD by creating a low-cost EEG device along with using a deep neural network that can classify the suspected patients into three classes: MCI patients, AD patients, and healthy patients. This is done using collecting brain wave signals using Electroencephalography (EEG) device during a 3 level N-Back working memory test. This is called event-related potential (ERP), later the data will be cut into small frames, FFT transformed to extract brainwave subbands (theta, alpha, and beta) and then projected to 2D images where it will be used in training convolution neural network for classification. Early detection of Alzheimer will reduce the progression of AD at early stages. We implement and evaluate our hardware and classification with an accuracy of 90.36% for MCI and 92.52% for Alzheimer's disease detection.