Alzheimer’s Disease classification Using Bilateral Residual Adaptive Intelligent Multiclass Network

Abstract

Early Alzheimer's disease (EAD) diagnosis enables individuals to take preventative actions before irreversible brain damage occurs. Cross-sectional imaging studies of AD demonstrate that the characteristics of the abrasion sites in AD patients, as revealed by magnetic resonance imaging (MRI), are highly diverse and distributed across the image space. In AD memory and cognitive abilities deteriorate, affecting the capacity to do basic activities. In and around brain cells, aberrant amyloid and tau protein accumulation is believed to cause it. Amyloid deposits create plaques surrounding brain cells, whereas tau deposits form tangles inside brain cells. The plagues and tangles harm healthy brain cells, causing shrinkage. This damage seems to be occurring in the hippocampus, a brain region involved in memory formation. There are presently no methods that provide the most accurate outcomes and suggestions. The current techniques do not identify AD early. So, we proposed Bilateral Residual Adaptive Intelligent Multiclass Network (BRAIM-Net) method for identifying the earlier prediction of AD. In BRAIM-Net, two datasets are used, namely an MRI image dataset and a text dataset. The MRI image dataset has been trained with CNN-deep residual network (ResNet) layers. Deep ResNet enables this ResNet model to extract more information from network levels. The Modified Adam Optimization has selected the best feature information from MRI scans of Alzheimer's patients and transferred it to another area while keeping the most important data. Using the BRAIM-Net approach, a multiclass classification has been carried out. Finally, users can enter their queries and the system will retrieve medical advice. The experimental results indicate that the classification accuracy of the approach proposed in this research can reach 97.86%..