Automatic multi-class alzheimer's disease detection from brain MRI images using deep learning methods

Diagnosis of Alzheimer's disease by doctors by analyzing brain MRI images is a difficult, human error-prone and time-consuming process. In recent years, many researchers have been working on early and accurate diagnosis of Alzheimer's disease using deep learning architectures. However, the success rates observed in the studies examined in the literature are not sufficient for clinical use in real hospitals. In this study, it is aimed to classify the images of Alzheimer's disease with fast and high accuracy using deep learning architectures. In addition, the proposed study can be used in hospitals to support specialist doctors in disease detection and thus play an important role in reducing incorrect treatments. In the training, validation and testing phases of the study, the open source licensed Augmented Alzheimer MRI dataset obtained through kaggle was used. The proposed architecture was developed by adding SE block attention mechanism to EfficientNet deep learning model. The results of the experiments showed that the proposed deep learning architecture achieved the highest success rate compared to other studies in the literature. In the accuracy evaluation metric, the proposed architecture achieved a success rate of 0,9903 indicating that it can successfully classify images of Alzheimer's disease. The proposed method will increase the efficiency in the detection of Alzheimer's disease, reduce the workload of specialist doctors, and improve the quality of diagnosis and treatment processes.