Dealing with Class Imbalance and Multi-Class Classification in Alzheimer's Disease and Dementia: An Innovative Methodology

Alzheimer's Disease (AD) and various forms of Dementia present significant challenges in the field of healthcare, affecting millions of people worldwide. It is crucial to accurately and promptly classify these neurodegenerative disorders to provide effective treatment and care for patients. However, the complex nature of dealing with class imbalances and multi-class classification within the AD and Dementia domain makes it difficult to develop precise diagnostic models. In this paper, we introduce a novel approach BD2EMNET (DEMentia NETwork including Borderline-SMOTE and DenseNet-121 architecture) that simultaneously addresses the issue of class imbalance and handles the intricacies of multi-class classification in the context of AD and Dementia. Through extensive experimentation on diverse datasets, we demonstrated that our approach surpasses the traditional methods. The Traditional methods that are similar to the specified problem were listed in the literature survey and one best experiment among them is first executed and later compared with the proposed approach to indicate the improvement in the performance. Our approach achieves an impressive accuracy of 99.58% for Dementia classification of four class and 99.88% for AD classification of five class. Notably, the approach outperforms existing solutions, particularly in scenarios involving class imbalance and multi-class situation. The implications of the research was profound as it bridges the gap between class imbalance and multi-class complexity, thereby enhancing the accuracy of AD and Dementia classification. This advancement contributes to early detection, personalized medicine, and improved patient outcomes. Furthermore, the adaptability of the approach suggests potential applications in other medical diagnostic contexts facing similar challenges of this. The findings highlight the potential of innovative methodologies to transform disease classification ultimately leading to a more effective healthcare strategies and interventions. DOI: https://doi.org/10.17762/ijisae.v12i4.7028