<![CDATA[Hypoglossal nerve stimulation (HNS) has emerged as a promising intervention for patients with obstructive sleep apnea (OSA) who are non-adherent or intolerant to continuous positive airway pressure (CPAP). This study presents a novel, integrative meta-analysis combining swarm intelligence, deep reinforcement learning, and graph neural network (GNN) architectures to investigate the cognitive and mental health outcomes associated with HNS across diverse adult populations. A total of 1,418 patients from nine high-quality studies— including RCTs, prospective registries, and observational cohorts—were included. Cognitive outcomes were measured using validated scales such as MoCA, PVT, and DSST, while mental health changes were assessed via PHQ-9, GAD-7, and PHQ-ADS. Our findings show that patients undergoing HNS therapy demonstrated a mean reduction in PHQ-9 scores of –2.66, indicating meaningful improvements in depressive symptoms. In particular, non-adherent patients had significantly higher anxiety scores (GAD-7: 8.27 vs. 3.90, p = 0.049) and depression scores (PHQ-ADS: 19.20 vs. 10.05, p = 0.035), reinforcing the importance of long- term adherence. Swarm-optimized feature selection identified Follow-up Duration, HNS Type, and Risk of Bias as dominant predictors of mental health outcomes. GNN-based modeling further captured inter-study relational structures with high predictive reliability (training loss < 0.02 at epoch 200). These results establish a robust data-driven framework to support the use of HNS in improving mental health and cognitive domains among OSA patients, while also demonstrating the analytic power of bio-AI hybrid modeling for personalized therapeutic insights.]]>
Copyrights © 2025
