Adaptive Weights Integrated Convolutional Neural Network for Alzheimer's Disease Diagnosis

Magnetic resonance imaging (MRI) is a common medical imaging technique, which is not harmful to the human body and is often used in the diagnosis of Alzheimer's disease (AD). We propose an adaptive weighted integrated convolutional neural network (AWI-CNN) to effectively identify patients with Alzheimer's disease and Mild Cognitive Impairment in this paper. The research data used in this paper are T1-weighted images from ADNI. First, after pre-processing, we obtained the grey matter images of the same individual in the coronal plane, axial plane, and sagittal plane. Second, we use the improved convolution neural network to train and recognize the three planes. We are using the idea of group decision making to adaptive weighted voting of three planes obtained from Softmax layer. Finally, using the technology of integrated learning, the data of different planes are integrated to obtain the final recognition accuracy. The results of our experiment achieves higher classification accuracy than many traditional methods: AD versus NC is 99.38% and MCI versus NC is 98.19%.