A review of deep learning techniques in Alzheimer's disease with emphasis on data tools and transfer learning.
Authors
Affiliations (4)
Affiliations (4)
- Department of Computer Science and the Environment, Liverpool Hope University, L16 9JD UK. Electronic address: [email protected].
- Department of Computer Science and Technology, Zhejiang Normal University, Jinhua, Zhejiang 321004, China.
- Department of Computing Science and Mathematics, Dundalk Institute of Technology, A91 K584 Ireland.
- Department of Computer Science and the Environment, Liverpool Hope University, L16 9JD UK.
Abstract
Alzheimer's disease (AD) is a significant neurological condition that is marked by the gradual decline of memory and cognitive function, with a higher incidence observed in older individuals. The mental deterioration associated with this condition is irreversible, resulting in substantial consequences for both affected individuals and society as a whole. Despite relentless research efforts, a definitive cure for AD remains elusive. However, interventions targeting the early stages of the disease have shown promise in slowing its progression. Deep learning-based approaches introduced better results for the early identification of AD stages, which can be curable. Due to less annotated data, those models have many problems regarding model over-fitting and class imbalance issues, directly impacting the model's performance. Researchers developed transfer learning-based approaches to overcome those issues, which can produce improved results on fewer annotated data samples. The primary motivation behind this article is to provide a review of the article, which is directly based on the transfer learning techniques for classifying AD stages using MRI and PET modalities. This article also provides a complete review of pre-processing tools for data extraction. It discusses the challenges that affect the performance of the models, as well as generalization challenges and biases in transfer learning.