Abstract BACKGROUND Integrating connectomics and functional brain activity into the study of brain tumors is among neuro-oncology’s challenges. Functional MRI (fMRI) studies showed that gliomas are integrated with brain networks’ activity, exerting effects beyond their margins. We investigated longitudinal modifications of fMRI gradients and graph-theory parameters in glioma patients, assessing their role in cognitive prognostication. MATERIAL AND METHODS Diffuse non-enhancing glioma patients underwent fMRI and cognitive assessments (OCS-Bridge) before, within 72h after, 3 and 12 months after surgery. Cognitive data were normalized and averaged into a global score. Differences between the last and preoperative scores were computed (“cognitive difference”). fMRI data underwent preprocessing, cortical parcellation, computation of timeseries, connectivity matrices, gradients and graph-theory measures. To quantify gradient dispersion, we computed a circularity metric (“CM”) of the hull perimeter of the first two gradients’ distribution. T-tests, partial correlations (correcting for volume and surgery radicality) and linear regression (regressors: age, volume, preoperative cognition, surgery radicality) were used to investigate associations with cognitive difference. RESULTS 17 patients (22-56 years, 8 females) were included, 4 had long-term cognitive worsening. Postoperative circularity metric, especially of the healthy intra- and inter-hemispheric connectivity (“HIIHC”), was higher than the preoperative (0.43 vs 0.32, t=2.4, p=0.03). Month3-postoperative and month3-preoperative differences in HIIHC CM were higher in patients with compared to those without cognitive worsening (0.33 vs -0.06, t=6.9, p<0.0001 and 0.27 vs 0.009, t=4.3, p=0.001). Both were negative predictors of the cognitive difference (the first: R=-0.94 with p=0.0002; β=-1.1 with p=0.001 and regression model’s p=0.003; the second: R=-0.76 with p=0.006; β=-1.1 with p=0.02 and model’s p=0.05). Month3-preoperative network modularity difference correlated with the second metric (R=-0.72, p=0.01) and the cognitive difference (R=0.67, p=0.02), while the preoperative modularity with the first metric (R=0.77, p=0.014). Similar results were obtained using a predetermined number of modules (Yeo’s 7 networks). Month12-preoperative difference in healthy intra-hemispheric connectivity CM also correlated with cognitive difference (R=-0.91, p=0.002). CONCLUSION Functional gradients seem dispersed after glioma surgery. Their recovery or worsening at month 3 could predict opposite long-term cognitive outcomes, putatively due to connectivity changes mainly involving the healthy hemisphere. Low preoperative network modularity and its increase at month 3 are associated with such gradient dispersion. Cognitive rehabilitation might be beneficial in the early post-surgical phase.
Read full abstract