Alzheimer Classification Based on Inception V3 Convolutional Neural Network

Abstract

Alzheimer's disease (AD)is an incurable disease that occurs in old age and early old age, and is characterized by neuron death and brain shrinkage. In this paper, an improved convolutional neural network based on Inception V3 is proposed for the recognition of Alzheimer's disease. We used the SMOTE technique to balance the data and adjusted the study rate using the ReduceLROnPlateau callback function. And at the same time, to improve the robustness of our model, we use the Batch Normalization method. Through 60 times of training, our model can quickly and accurately classify the input images. Finally, the accuracy of our model reached 94.40% on the training set, 90.28% on the test set and 90.48% on the validation set. Besides, the AUC of the model on test set can reach 0.9857, the precision rate is 90.06%, the recall rate is 90.87%, and the Fl-score rate is 90.23%.