Binary Classification of Alzheimer's Disease Using Siamese Neural Network for Early Stage Diagnosis

Alzheimer's Disease (AD) is a cognitive disease. In individuals with disease, increased brain cell loss is observed over time. This situation leads to deficiencies in memory and thinking ability over time. As a result, significant impairments occur in individuals’ ability to perform primary function. According to research results, the rate of thos disease doubles every five years among people aged between 65 and 85. The causes of AD are unknown and nowadays not definite cure. Early diagnosis of the disease in clinical cure as it has the potential to slow or stop progression. This study aimed to make a prediction based on Magnetic Resonance (MR) images. Images in the standard Alzheimer dataset obtained from the open access database Kaagle were enhanced by applying Gaussian and Median filters. Siamese Neural Network (SNN) categorizes disease stages by learning the similarity between these images. Two categories of images were used from the dataset: Very Mild Dementia (VMD) and Non-Dementia (ND). According to this proposed study, the training accuracy was %99.62 and the validation accuracy %97.67.