Abstract: Objective
Working memory decline is a core feature of cognitive aging, with its essence lying in the systematic disruption of large-scale brain network coordination. Severe working memory decline significantly impacts older adults' quality of life and imposes substantial burdens on families and society. However, existing neuromodulation techniques face critical limitations: exogenous approaches (e.g., TMS, tDCS) are constrained to single-site interventions with limited precision, while endogenous neurofeedback based on EEG or fMRI, despite its promise, is hindered by high costs, poor portability, or insufficient spatial resolution for long-term application in older populations. Moreover, emerging network-based neurofeedback studies remain in early exploratory stages, with feedback signals typically simplified to single connectivity strengths or summations of network edges, failing to capture the holistic topology of multi-node dynamic coordination. Notably, no study has yet targeted the severely declined subgroup within the older population from a network connectivity perspective.
To address these gaps, this study proposes a connectome-targeted fNIRS neurofeedback intervention framework for severe working memory decline, systematically investigating its intervention effects and cognitive transfer mechanisms.
Study 1: Identification of Specific Network Connectivity Patterns and Target Setting
This study first adopts an extreme groups approach to classify older adults into severely-declined and function-maintained groups based on working memory performance. Using multivariate pattern analysis (MVPA) based on representational similarity analysis (RSA), we identify network connectivity templates that optimally discriminate between the two groups during a working memory task. These templates serve as the core intervention targets, with the average connectivity pattern of the function-maintained group established as the regulatory goal. This data-driven target identification strategy provides a foundation for subsequent individualized intervention, shifting from experience-based target selection to evidence-based targeting.
Study 2: Closed-Loop Connectome-Targeted fNIRS Neurofeedback Intervention
Building on the identified targets, we develop a closed-loop fNIRS neurofeedback training paradigm. Unlike previous studies that simplified feedback signals to single connectivity strengths or summations of network edges, this study employs the spatial similarity between an individual's real-time network connectivity pattern and the target pattern as the feedback index. This index preserves the complete topological structure of network nodes, capturing richer information about multi-node dynamic coordination. A three-group randomized controlled design (connectome-targeted feedback, traditional single-region feedback, and sham feedback) is implemented to systematically evaluate the intervention’s effectiveness, durability, and superiority over traditional single-region feedback approaches.
Study 3: Cognitive Transfer Effects and Dual-Pathway Mechanisms
Given that the targeted network serves not only working memory but also multiple other cognitive functions, regulation of this network may induce broad cognitive transfer effects. This study systematically examines transfer effects across four dimensions—transfer type (near vs. far), direction (positive vs. negative), strength (strong vs. weak), and durability (immediate vs. long-term). Furthermore, we propose and test a "dual-pathway transfer" theoretical framework: the "network direct-driven pathway"—directly improving cognitive functions closely associated with the target network, and the "working memory-mediated pathway"—indirectly facilitating broader cognitive functions through enhancement of working memory as a foundational cognitive process. Mediation analysis is employed to differentiate the relative contributions and interplay of these two pathways.
Theoretical Framework and Innovation
This study advances a three-level theoretical framework: the essence of working memory aging as large-scale network coordination imbalance, the pathway of endogenous self-regulation for network reshaping, and the mechanism of dual-pathway cognitive transfer. The core innovation lies in shifting intervention targets from single brain regions to large-scale network connectivity patterns, developing a spatially similarity-based feedback index that preserves topological information, and proposing a dual-pathway transfer framework that explains how connectome-targeted regulation produces cognitive gains. This paradigm holds promise for providing a safe, effective, and sustainable intervention for severe working memory decline, while offering a generalizable framework for broader network-targeted neuromodulation strategies.

Key words: working memory aging, network connectivity modulation, fNIRS neurofeedback, cognitive transfer, dual-pathway mechanism