Adaptive Frequency-Optimized Wavelet Networks for Early Detection of Subjective Cognitive Decline via Resting-State fMRI.

BACKGROUND: Early detection of subjective cognitive decline (SCD), a preclinical stage of Alzheimer's disease (AD), remains a clinical challenge due to its subtle manifestations. This study aims to address these challenges by introducing a novel approach to enhance the detection and analysis of SCD. METHODS: A Frequency Self-Adaptive Wavelet Transform (FSAWT) model was developed and optimized for functional brain network (FBN) construction using resting-state functional MRI (rs-fMRI) data. The model dynamically selected "golden frequencies" to improve the accuracy and interpretability of brain connectivity patterns. FBNs from 240 participants (106 SCD, 134 controls) were analyzed and compared using traditional methods, pearson correlation (PC) and sparse representation (SR). Receiver operating characteristic-area under the curve (ROC-AUC) analysis validated the classification results. RESULTS: Our findings demonstrate that individuals with SCD exhibit distinct functional connectivity alterations, including reversed parahippocampal gyrus-superior parietal gyrus connectivity-suggesting early DMN disintegration, weakened temporoparietal pathways linked to memory deficits, and enhanced fusiform gyrus-orbitofrontal connectivity. The frequency-optimized SRWT method achieved superior diagnostic performance (83.71% accuracy, AUC = 0.84) with 82.11% sensitivity and 85.71% specificity, significantly outperforming traditional approaches (61.93% accuracy for PC), highlighting its potential for early SCD detection through these network-based biomarkers. CONCLUSIONS: The FSAWT model offers a robust framework for early SCD detection by integrating frequency-specific and cross-frequency dynamics. While these findings highlight potential contributions to precision diagnostics and personalized interventions for neurodegenerative disorders, such applications remain to be established in future studies. Future applications may also explore multimodal neuroimaging and broader cognitive impairments.