To explore Alzheimer's disease, there were many classical machine learning techniques have been used, growing from image decomposition techniques such as principal component analysis through high-complexity algorithms. With the appearance of the deep learning methods, the extraction of high-level abstract features directly from MRI images has become promising to describe the distribution of data in low-dimensional space. Our approach includes training sparse autoencoder (SAE) approach to perform unsupervised feature leaning and prediction of AD. We spot the light on developing a SAE model to learn effective features from the AD dataset and then perform classification using the learned features.
IntroductionDementia is a medical disorder categorized by progressive degeneration in two or more cognitive domains, including memory, language, executive and visuospatial function, personality, and behavior, which causes loss of abilities to perform instrumental and/or basic activities of daily living. By far, Alzheimer's disease (AD) is the most common cause of dementia for up to 80% of all dementia diagnoses [1, 2]. In the United States, even though the general death rate from stroke and cardiovascular disease is decreasing, the percentage of deaths related to AD is growing by 89% between 2000 and 2014 [3-6].Medical experts are responsible for evaluating the interpretation of AD medical data, which is very difficult and restricted for a medical expert to interpret images