Machine learning techniques for diagnosis of alzheimer disease, mild cognitive disorder, and other types of dementia

Abstract

Alzheimer’s disease (AD) is one of the most common form of dementia which mostly affects elderly people. AD identification in early stages is a difficult task in medical practice and there is still no biomarker known to be precise in detection of AD in early stages. Also, AD is not a curable disease at this time and there is a high failure rate in clinical trials for AD drugs. Researchers are making efforts to find ways in early detection of AD to help in slowing down its progression. This paper reviews the state-of-the-art research on machine learning techniques used for detection and classification of AD with a focus on neuroimaging and primarily journal articles published since 2016. These techniques include Support Vector Machine, Random forest, Convolutional Neural Network, K-means, among others. This review suggests that there is no single best approach; however, deep learning techniques such as Convolutional Neural Networks appear to be promising for diagnosis of AD, especially considering that they can leverage transfer learning which overcomes the limitations of availability of a large number of medical images. Research is still on-going to provide an accurate and efficient approach for diagnosis and prediction of AD. In recent years, a number of new and powerful supervised machine learning and classification algorithms have been developed such as the Enhanced Probabilistic Neural Network, Neural Dynamic Classification algorithm, Dynamic Ensemble Learning Algorithm, and Finite Element Machine for fast learning. Applications of these algorithms for diagnosis of AD have yet to be explored.