Predictive Natural Language Metrics of Alzheimer’s Disease and Cognitive Decline Trend Analysis

Early detection of Alzheimer’s disease (AD) and cognitive decline remains a critical frontier in biomedical informatics. This thesis develops resource-efficient and explainable large language model (LLM) architectures to predict future cognitive decline and idea density metrics from longitudinal life narrative interviews. By analyzing the representation geometry of fine-tuned transformers, sliding-window token dynamics, and linguistic complexity indices, we establish robust, quantifiable language markers capable of tracking subtle cognitive trajectories prior to acute clinical manifestation.

Computer Science and Engineering Theses and Dissertations, Southern Methodist University