Prediction Models for Early Detection of Alzheimer: Recent Trends and Future Prospects

Alzheimer's Disease (AD) is a neurodegenerative condition characterized by irreversible cognitive decline. Detecting AD early is challenging as symptoms typically manifest years after the disease onset, necessitating the identification of subtle biomarker changes, often detectable through various neuroimaging modalities. Computer-aided diagnostic models leveraging machine learning and deep learning offer promising avenues for analyzing diverse input modalities to aid in early AD detection. The present study aims to analyze recent trends in the methods utilized by researchers for early prediction of Alzheimer along with identifying key challenges in existing research. The study follows PRISMA methodology to provide a comprehensive analysis of studies published in the last five years, resulting in sixty-four studies. The studies are sourced from significant data repositories after careful inclusion and exclusion criteria. The analysis of studies reveals the utilization of various machine learning and deep learning architectures, emphasizing practitioner-oriented perspectives such as data sources, input modalities, feature extraction strategies, and validation techniques. Performance comparison of the methods elucidates the effectiveness of deep learning frameworks, particularly in handling multimodal data and facilitating multiclass classification. Notably, structural MRI emerges as the most utilized input modality, with potential improvements observed when combined with Diffusion Tensor Imaging (DTI). Furthermore, current challenges within the existing literature are addressed and provides recommendations for future research directions. This review serves as a valuable resource for both novice and experienced researchers, offering insights into the state of the art and guiding efforts towards improved Alzheimer's disease prediction methodologies.