Deep Learning Approach based Classification of Alzheimer's Disease Using Brain MRI

Alzheimer's disease is a well-known brain disorder which currently still does not have a cure. However, early diagnosis of the disease can help the patient to get proper treatment and slow down the progressiveness of the disease. In medical imaging, deep learning methods have been widely used to assist the medical professionals in Alzheimer's disease diagnosis to classify the normal brain and the stages of Alzheimer's disease. This paper compares the classification performance of deep learning architecture such as MobileNetV2, ResNet-101, DenseNet-121, and the proposed modified convolutional neural network (CNN) model inspired by the VGG16 network using a brain magnetic resonance imaging (MRI) dataset from Kaggle. We evaluate the model performance based on their accuracy, precision, recall, and F1-score and the obtained results show that the classification of Alzheimer's disease using our proposed model is more accurate than the other model with 97.625% accuracy, 98% recall, 98% precision and 98% F1-score. Our study successfully proved that implementation of dropouts in the CNN model can improve the model accuracy and reduce the model training duration, and made valuable insights for medical image diagnosis and future research.