关键词: 疼痛, 脑网络, 疼痛韧性, 认知功能, 功能磁共振成像

Abstract: Chronic pain profoundly undermines patients’ physical, psychological, and social functioning, highlighting the urgent need for effective coping and management strategies. Psychological resilience plays a pivotal role in buffering the adverse effects of pain, and enhancing resilience has become essential for patients to face biopsychosocial challenges. However, the key factors that promote pain resilience and their underlying mechanisms remain insufficiently understood. Previous studies has demonstrated a positive association between cognitive function and resilience, with interventions targeting cognitive interventions shown to enhance resilience. Building on this work, studies employing neuroimaging and machine learning techniques further revealed that gray matter volume and functional activity within cortical regions of the central executive network (CEN) are associated with pain resilience. Building on these findings, the present research program hypothesizes that cognitive function and the CEN are not only pivotal to the enhancement of pain resilience but also predictive of its longitudinal development. To test this hypothesis, a multicenter, longitudinal study integrating self-report measures and functional magnetic resonance imaging (fMRI) is proposed.
This research program comprises three sub-studies. Study 1 adopts a multi-time point longitudinal survey design to examine the relationship between cognitive function and pain resilience among patients with chronic musculoskeletal pain and in healthy controls. Cognitive functioning, pain resilience, and pain-related outcomes are assessed at baseline, 6 months, and 12 months. Cross-lagged panel modeling is used to characterize the directional and temporal associations between cognitive function and pain resilience. Additionally, multi-group structural equation modeling is further applied to determine whether these associations differ between patients and healthy controls.
Study 2 employs a multimodal neuroimaging design across two clinical centers. At baseline (Time Point 1), demographic characteristics, cognitive function, pain outcomes (including pain intensity, pain interference, and depressive symptoms), pain resilience and structural and resting-state fMRI data are collected from patients with chronic musculoskeletal pain. Pain resilience is reassessed at 6 months (Time Point 2) and 12 months (Time Point 3). Mediation analyses are conducted to test whether resting-state functional connectivity within the CEN mediates the association between baseline cognitive function and pain resilience at Time Point 3. Internal CEN connectivity (i.e. functional interactions among nodes within the network) serves as the mediator, with seed regions selected based on prior literature. The mediation model is established using data from Center 1 and independently validated in Center 2, enabling robust identification of neural mechanisms underlying cognitive contributions to pain resilience.
Study 3 extends the first two studies by developing a multimodal deep learning prediction model using a multicenter prospective framework to predict pain resilience. Guided by the multimodal integration framework, the model incorporates complementary feature domains, including demographic variables, pain outcome indicators, cognitive function measures, and neuroimaging features. Demographic variables enhance clinical applicability and mitigate group-level confounds (e.g., age-related brain changes), thereby improving interpretability of neural predictors. Pain outcomes index clinical symptomatology, cognitive function measures capture psychological context, and neuroimaging features provide a biological foundation. By integrating these multimodal data, the model aims to identify robust, clinically meaningful biomarkers of pain resilience. At baseline (Time Point 1), all domains are assessed. Pain resilience, cognitive function and pain outcomes are reassessed at 6 months (Time Point 2) and 12 months (Time Point 3), with pain resilience additionally evaluated at 18 months (Time Point 4). A Gated Recurrent Unit (GRU) model is used to integrate longitudinal (cognitive function and pain outcomes) and static baseline data (demographics and neuroimaging features) to predict pain resilience at Time Point 4. Training is conducted using data from Center 1, with external validation in Center 2 to ensure robust individualized prediction.
Collectively, this research program is expected to clarify the cognitive and neural mechanisms through which cognitive function shapes psychological resilience in individuals with chronic pain. By delineating the cognitive processes and CEN-based neural pathways that support pain resilience, the findings are positioned to advance theoretical models of adaptive coping in chronic pain and inform the development of targeted, personalized interventions for chronic pain management.

Key words: pain, brain networks, pain resilience, cognitive function, functional magnetic resonance imaging