Deep Learning and Machine Learning for Alzheimer’s Disease Biomarker Identification

Background and Aims: Alzheimer's Disease (AD), the most prevalent cause of dementia globally, is characterized by progressive cognitive decline that gradually impairs independent daily functioning. Current approaches to classify AD stages incorporate medical history, neuropsychological testing, genetic factors, and neuroimaging, including Magnetic Resonance Imaging (MRI). The application of Machine Learning (ML) and Deep Learning (DL) to early Alzheimer's Disease diagnosis and automated stage classification has attracted substantial research interest in recent years. This growing attention stems from both advances in neuroimaging technology and genetic analysis, coupled with these algorithms' unique capability to identify hidden patterns in complex biomedical data that may elude conventional analytical methods. Therefore, this study aims to (۱) analyze hidden patterns in AD concerning gene expression through unsupervised learning and (۲) develop a deep learning framework for accurate, automated AD classification using neuroimaging data. Methods: We first applied an unsupervised clustering method to the GSE۵۲۸۱ AD microarray dataset (NCBI) to identify hidden patterns. Next, we trained a Convolutional Neural Network (CNN) on the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset to classify AD stages, comparing its performance against traditional machine learning models. Results: The clustering validity indices (such as Rand Index and Silhouette Score) reveal distinct gene expression patterns that may be associated with AD pathology. Subsequent evaluation of our CNN model on the ADNI dataset demonstrates superior classification performance (accuracy=۰.۹۳%), significantly outperforming traditional machine learning approaches in AD stage differentiation Conclusion: This study demonstrates the effectiveness of unsupervised clustering and CNN-based approaches for identifying AD biomarkers, while also highlighting current methodological limitations and promising future directions for improving Alzheimer's Disease diagnosis. Keywords: Machine learning, Deep learning, Alzheimer's Disease

کلیدواژه ها:

Machine learning ، Deep learning ، Alzheimer's Disease