Classification of alzheimer's disease on mr using deep neural network methods

Alzheimer's disease is a degenerative disease that progresses over time and leads to loss of mental function. Alzheimer's disease develops and progresses long before it is diagnosed. Early diagnosis of the disease is very difficult. Since it cannot be diagnosed early, it continues to progress for years and continues to progress to the point that it prevents the basic vital functions of the brain from being performed. Alzheimer's disease is known to consist of four main stages. These stages are defined as the ADHD, ADHD, OSHD and ISHD stages. Symptoms that occur in the early stages are usually too mild to be recognized clinically. Alzheimer's disease is clinically diagnosed by listening to the patient's medical and trauma history from the patient and the patient's relatives. In addition, the patient's brain image is taken through computer-aided devices such as MRI and Computed Tomography (CT). Changes in the brain are observed through these images and are expected to support the diagnosis of the disease. In this study, the early stages of Alzheimer's disease were divided into three main workgroups and model training was performed on these workgroups with ResNet-18, ResNet-50, ResNet101, DenseNet161, EfficientNet-B7, ConvNext architectures. The ResNet-50 architecture, which is the most successful model in the study group dealing with Normal-Mild Cognitive Impairment stages; reached an accuracy rate of 95.40%. The most successful model for the study group consisting of Normal-Mild Cognitive Impairment stages was trained with ConvNext architecture. In this study group, the ConvNext model achieved a success rate of 97.53%. In the third study group, which consists of two main stages, Mild Cognitive Impairment and Very Mild Cognitive Impairment, the most successful model was trained with ConvNext architecture and achieved a success rate of 92.17%. The most successful models obtained from the study groups were integrated with the web interface and put into the test environment. It is thought that the study will shed light on clinical tests and real-time applicability in the early diagnosis of Alzheimer's disease.