An interpretable deep learning approach to detect Alzheimer using MRI images

Abstract

Alzheimer’s disease (AD) is a serious neurological condition that causes loss of long term memory, cognitive difficulties, disorientation, inconsistent behavior, and even tually death. Also, AD is caused by the destruction of brain cells that are responsible for proper brain function. The main focus of our research is to provide an efficient model for the rapid diagnosis of Alzheimer’s disease. In this research, we design and demonstrate an interpretable deep-learning approach to detect Alzheimer’s us ing MRI images. For the experiment, brain MRIs are utilized, and by using this data, the model is able to determine the disease. Additionally, this model is de signed based on multiclass classification (MildDemented, ModerateDemented, Non Demented, VeryMildDemented) for helping patients belonging to different phases of Alzheimer’s disease. For this research, we experimented with four different ar chitectures of Convolutional Neural Networks. From the models, we obtained an accuracy of 92.65% for VGG-19, 89.18% for DenseNet-169, 87.84% for ResNet-50 V2, and 80.10% for Inception V3. By comparing and contrasting the performance of the models, the result can be improved by up to 92.65%, and it is decided to im plement the best-performing architecture (VGG19) into the system. Although there was a lack of data and it was difficult to tell the difference between a brain suffering from Alzheimer’s disease and a normal brain, the findings obtained revealed accu rate identification and categorization of Alzheimer’s disease and its phases. Lastly, GradCam (Gradient-Weighted Class Activation Mapping) is implemented to make the application of Explainable AI(XAI) apparent. Therefore, the proposed system would enable the detection and interpretation of Alzheimer’s disease effectively.