Application of deep learning for the classification of the stage of Alzheimer on the basis of MRI of the brain

Abstract

Diagnosis of Alzheimer’s disease at an early stage of development plays a significant role in the treatment of this disease, since determining the severity of the disease and the risk of its progression allows for preventive measures to be taken in a timely manner, before irreversible brain damage occurs. Alzheimer’s disease is a chronic degenerative disease associated with damage to neurons in the brain. To diagnose this disease, along with other methods, MRI of the brain is used. Of interest are formalized automated MRI analysis tools that can serve as a decision support tool for making a diagnosis. An effective mechanism for developing such tools in the presence of a large training sample can be deep learning methods, in particular methods based on the construction of convolutional neural networks. A review of research in this area reflects a number of successful computational experiments in the application of convolutional neural networks to medical image analysis. This work attempts to use a convolutional neural network (CNN) to classify the stage of Alzheimer’s disease based on brain MRI. The following main classes (levels of disease) are distinguished: NonDementia (no dementia), VeryMildDementia (early dementia), MildDementia (moderate dementia), ModerateDementia (severe dementia). The model proposed in the work demonstrates good quality in terms of the main classification metrics and allows one to determine all stages of the disease with great accuracy, and the VeryMildDemented class is best determined. Recognizing this particular stage of the disease is very important from the point of view of selecting treatment that prevents the development of the disease.