Abstract
BACKGROUND: In recent years, the incidence of cognitive diseases has also risen with the significant increase in population aging. Among these diseases, Alzheimer disease constitutes a substantial proportion, placing a high-cost burden on health care systems. To give early treatment and slow the progression of patient deterioration, it is crucial to diagnose mild cognitive impairment (MCI), a transitional stage. OBJECTIVE: In this study, we use autobiographical memory (AM) test speech data to establish a dual-modal longitudinal cognitive detection system for MCI. The AM test is a psychological assessment method that evaluates the cognitive status of subjects as they freely narrate important life experiences. METHODS: Identifying hidden disease-related information in unstructured, spontaneous speech is more difficult than in structured speech. To improve this process, we use both speech and text data, which provide more clues about a person's cognitive state. In addition, to track how cognition changes over time in spontaneous speech, we introduce an aging trajectory module. This module uses local and global alignment loss functions to better learn time-related features by aligning cognitive changes across different time points. RESULTS: In our experiments on the Chinese dataset, the longitudinal model incorporating the aging trajectory module achieved area under the receiver operating characteristic curve of 0.85 and 0.89 on 2 datasets, respectively, showing significant improvement over cross-sectional, single time point models. We also conducted ablation studies to verify the necessity of the proposed aging trajectory module. To confirm that the model not only applies to AM test data, we used part of the model to evaluate the performance on the ADReSSo dataset, a single time point semistructured data for validation, with results showing an accuracy exceeding 0.88. CONCLUSIONS: This study presents a noninvasive and scalable approach for early MCI detection by leveraging AM speech data across multiple time points. Through dual-modal analysis and the introduction of an aging trajectory module, our system effectively captures cognitive decline trends over time. Experimental results demonstrate the method's robustness and generalizability, highlighting its potential for real-world, long-term cognitive monitoring.