Artificial Intelligence for Alzheimer's Disease Diagnosis: From Traditional Machine Learning to Large Language Models

Alzheimer's disease (AD) is the most prevalent neurodegenerative disorder and a leading cause of dementia worldwide, characterized by progressive cognitive decline, memory impairment, and functional deterioration. With the rapid growth of the aging population, AD has become a major global health challenge, imposing substantial burdens on patients, families, and healthcare systems. Despite extensive research, early and accurate diagnosis of AD remains challenging due to disease heterogeneity, overlapping clinical manifestations, and the lack of easily accessible, highly sensitive, and specific diagnostic markers. Recent advances in biomedical technologies, including neuroimaging, multi-omics profiling, electronic health records, and digital health tools, have generated large-scale and heterogeneous datasets, providing new opportunities for improving AD diagnosis. However, extracting clinically meaningful information from these complex data sources remains difficult using conventional statistical approaches. Artificial intelligence (AI) has progressively transformed AD diagnosis by evolving from traditional machine learning (ML) approaches based on handcrafted feature engineering to deep learning (DL) models capable of automated representation learning and multimodal information integration. More recently, large language models (LLMs) have further expanded the scope of AI-driven AD diagnosis by enabling contextual understanding of unstructured clinical information, knowledge-guided reasoning, and integration of multimodal biomedical evidence. This transition reflects a shift from feature-based prediction toward more flexible and intelligent diagnostic frameworks. This review synthesizes recent advances in AI-based AD diagnosis, tracing the evolution from traditional ML to DL and LLMs. Particular emphasis is placed on the emerging role of LLMs in extracting disease-related information from speech and clinical narratives, integrating heterogeneous biomedical data sources, and enabling multimodal frameworks for AD assessment.