445. DISTINCTIVE ASSOCIATION PATTERNS BETWEEN TOPOLOGICAL PROPERTIES OF BRAIN STRUCTURAL NETWORKS AND SOCIAL COGNITION IN EUTHYMIC BIPOLAR DISORDER

Abstract Background Social cognition impairments are a well-recognized feature of bipolar disorder (BD), persisting even during euthymic states. Previous studies using graph theoretical analysis (GTA) in structural imaging have revealed characteristic patterns of white matter (WM) networks disruption, including reduced global integration and increased local segregation in BD, which implicated in various domains of cognitive dysfunction. However, the relationship between alterations in structural brain connectivity and deficits in social cognition remains under-explored. Aims & Objectives This study aims to investigate the relationship between alterations in WM networks and social cognition in euthymic BD (euBD). We employed generalized q-sampling imaging (GQI) and GTA to extract topological properties of brain networks. The primary objective is to identify specific connectivity parameters, such as network integration and segregation, and its association to social cognition in euBD. Method This study included 44 euBD patients and 59 HC. All participants underwent diffusion-weighted imaging using a 3T MRI scanner to reconstruct white matter networks via GQI. GTA was applied to quantify topological properties of the brain structural networks. Social cognitive performance was assessed using three validated tools: Diagnostic Analysis of Nonverbal Accuracy 2, Taiwan version (DANVA-2-TW), Reading the Mind in the Eyes Test (RMEIT), and the Chinese Theory of Mind task (cToM). Group differences in brain network metrics and social cognitive measures were examined, and associations between structural connectivity parameters and social cognition were explored in both groups. Results The HC group significantly outperformed the euBD group in all social cognitive tasks (p=0.004 in ToM, p=0.001 in RMEIT, p<0.001 in DANVA-2-TW). Regarding GTA, the HC group demonstrated higher levels of network integration (assortativity, p=0.037), while the euBD group exhibited increased network segregation (normalized clustering coefficient, p=0.026). Furthermore, the BD group showed enhanced small-world properties (sigma, p=0.016) compared to the HC group. Correlation analysis revealed negative correlation between DANVA-2-TW total scores and network segregation (gamma and modularity) in both groups, as well as the small-world index (sigma). Social cognition was positively associated with network integration properties in both groups, including that assortativity positively correlated with DANVA-2-TW total scores in the HC group, and global efficiency was positively correlated with cToM performance in the euBD group. Discussion & Conclusions The BD group showed lower network integration and higher local segregation, aligning with previous studies and suggesting that over-segregation of WM networks may underlie social cognition impairments in BD. Integration properties, such as global efficiency, were linked to better theory of mind (cToM) performance in BD, while segregation metrics correlated negatively with emotion recognition across groups. These findings indicate that different social cognition domains may rely on distinct patterns of brain network property alterations, as well as the importance of balancing global and local WM connectivity for social cognition. In conclusion, this study advances our understanding of the neurobiological mechanisms underlying social cognition impairments in BD by identifying distinctive patterns of structural network alterations.