ABSTRACT
In recent years, deep learning using a convolutional neural network (CNN) has become the cutting-edge technology in diagnosing Alzheimer’s disease (AD) due to its advancement in image classification, segmentation, and detection tasks. The robustness of CNN architecture has led to a surge of research adopting it in AD disease recognition, replacing conventional machine learning methods. However, CNNs require a large-scale dataset to achieve a better classification performance. Also, it is still struggling with overfitting issues and spatial invariance behavior. This chapter reviews the emergence of an advanced deep learning approach involving advanced modules and CNN architectures employed for AD pattern recognition on magnetic resonance imaging (MRI) data. In this review, several studies from well-known databases have been selected and analyzed to answer the identified research questions relating to data and recent advanced approaches utilizing attention mechanisms that have been integrated with CNN architecture to further improve the architecture in handling medical images, thus enhancing the classification performance in diagnosing AD. Considering the insights drawn from this chapter, the need to investigate further the potential of enhancement modules in CNN frameworks for enhancing the performance of AD pattern recognition was identified as an important direction for future study.
