TITLE:
Macroscale Cortical Hierarchy and Spontaneous Speech for Preclinical Alzheimer’s Disease: A Narrative Review Focused on Remote Screening and Longitudinal Prognosis
AUTHORS:
Zhengwei Chen, Mingtian Lu, Zhejing Ding, Tianyang Guan, Zheng Lai, Zhongliang Li, Jing Huang, Guomin Huang, Yueling Lyu, Yu Liu
KEYWORDS:
Alzheimer’s Disease, Default Mode Network (DMN), Digital Biomarkers, Spontaneous Speech, Resting-State fMRI
JOURNAL NAME:
Health,
Vol.18 No.3,
March
10,
2026
ABSTRACT: Amyloid-β (Aβ) pathology can be detected years before clinical Alzheimer’s disease (AD), yet forecasting who will decline—and how rapidly—remains difficult in cognitively unimpaired (CU) and subjective cognitive decline (SCD) populations. Resting-state fMRI connectome gradients provide a low-dimensional description of macroscale cortical hierarchy, typically spanning unimodal systems to transmodal association cortex anchored in the default mode network (DMN). Alterations in gradient range, dispersion, and template similarity have been reported across the AD continuum and are increasingly described in preclinical Aβ+ cohorts. In parallel, spontaneous speech has emerged as a scalable digital phenotype: acoustic timing and pausing can be captured remotely and repeatedly, extracted automatically without manual transcription, and tracked longitudinally with minimal practice effects. This narrative review synthesizes: 1) why gradient-based hierarchy metrics are biologically plausible early functional readouts of preclinical AD, 2) the emerging evidence for atypical hierarchy in CU/SCD Aβ+ individuals, 3) the speech feature families most suitable for scalable early detection and prognosis—emphasizing timing/pause measures—and 4) practical standards for longitudinal models predicting cognitive decline slope and conversion to mild cognitive impairment (MCI). We conclude that gradients and speech—especially transcription-free timing markers—are complementary and potentially synergistic, but translation requires rigorous confound control (motion, hearing, affect, device variability), transparent reporting, calibration, and external validation.