ESTRO 2024 - Abstract Book

S826

Clinical - CNS

ESTRO 2024

Material/Methods:

50 WHO grade 2-3 adult glioma survivors (>1 year post-therapy) and 50 matched healthy controls (gender, age and education), underwent a neurocognitive assessment, covering six neurocognitive domains. Raw test scores were transformed to w-scores, corrected for age and education. Furthermore, all subjects were subjected to advanced MR imaging. Weighted whole-brain and nodal graph metrics were calculated based on diffusion-weighted imaging, using multi-shell multi-tissue constrained spherical deconvolution and anatomically constrained tractography. We defined hubs based on 4 nodal metrics (i.e. nodal strength, betweenness centrality, clustering coefficient and characteristic path length) (Figure 1). Nodal graph metrics were compared between patients and controls for the hub nodes and non-hub nodes separately in a mixed-design ANOVA with repeated measures. Group differences in whole-brain graph measures were explored using the Mann-Whitney U test. Graph metrics that significantly differed between the two groups were correlated with w-scores per cognitive domain. Bonferroni correction was applied to correct for multiple testing.

Results:

Among the glioma survivors, a significant portion (60%) exhibited cognitive impairment in at least one cognitive domain following their treatment. In healthy controls 12/78 nodes were identified as hubs (Figure 2). However, in glioma survivors, the left putamen (66%) and right putamen (81%) were significantly less frequently designated as hubs compared to controls (96%, p<.001). This node’s assortativity was positively correlated with attention, and proxy IQ (r(90)>.57,p<.001). Nodal metrics of clustering coefficient, characteristic path length, local efficiency and assortativity were significantly different between groups, independently of hubness (all p.001). All these nodal metrics correlated strongly with cognitive outcomes (r(98)>.4, p<.002). Moreover, attention and proxy IQ were significantly more often correlated with assortativity within hubs compared to non-hub regions (p<.001). Whole brain graph measures of clustering coefficient (r=.69), global (r=.57) and local efficiency (r=.50) correlated positively (r>.50) with proxy IQ (p<.001).