Alzheimer's Disease Classification Model Based on MED-3D Transfer Learning

Abstract

Alzheimer's disease (AD) is one of the most common forms of dementia and a common condition of old age. Its pathogenic mechanism is unknown, and it is also an irreversible neurodegenerative disease. Therefore, early detection and treatment are particularly critical for patients. In this paper, we adopt the idea of transfer learning, use the ADNI dataset to retrain the Med-3D network, and change the segmentation network into the whole connection layer of classification. At the same time, we retrain the Resnet-3D network and compare the Med-3D network. Experimental results show that the convergence rate of this migration model is faster, and the accuracy is higher than that of the ResNet network with corresponding layers. The final accuracy rate reached 83%.