Short-term progression risk stratification in glioblastoma using post-resection structural connectivity biomarkers

Abstract ‌ Background ‌: While structural connectome analysis enables preoperative mapping of glioblastoma (GBM) infiltration, mass effect-induced distortion compromises the accuracy of peritumoral tract assessment. We aimed to investigate fiber disruption characteristics and predict short-term progression based on structural connectivity features after eliminating mass effect. Methods : We retrospectively analyzed 113 GBM patients with ≥ 90% resection and 65 healthy controls. Diffusion tensor imaging (DTI) data were processed to construct structural connectomes, which were segmented into three compartments relative to the resection cavity: Tumor disrupted cerebral regions, anatomically confined to FLAIR hyperintense areas and direct fiber disruption; Distant disrupted cerebral regions, outside FLAIR hyperintense areas but exhibiting direct fiber disruption; Indirect disrupted cerebral regions, remote from FLAIR lesions with indirect fiber disruption. The patterns of differential disruption across compartments and progression timelines were quantified, along with their correlations to the Karnofsky performance status (KPS). The Area Under the Curve (AUC) evaluated how well disrupted fibers predict progression time. Patients with fiber disruption counts exceeding the Youden index were classified as high-risk versus low-risk for progression, validated by Kaplan-Meier analysis and Chi-square test. Structural connectivity disruption were used to predict short-term progression via Cox regression. Results : After eliminating mass effects, widespread structural connectome disruption was observed. Among 49 within 1-year progressers, tumor-disrupted regions showed more severe fiber disruption than later-progressing patients (F = 32.5, P  < 0.001). Fiber disruption in tumor-disrupted compartment negatively correlated with pre-radiotherapy KPS score (r=-0.349, P  < 0.001), and best predicted progression time (AUC = 0.803, P  < 0.001). High-risk patients progressed faster (10 months) than low-risk patients (15 months) ( P  < 0.001). 81% of low-risk and 71% of high-risk patients were correctly identified (χ²=30.29, P  < 0.001). Incorporating structural connectivity disruption significantly improved multivariable Cox regression performance over clinical/imaging variables alone ( P  < 0.001). Conclusions : Structural connectivity quantitatively maps postoperative regional cerebral disruption in GBM. Fiber disruption within the tumor-disrupted compartment may identify patients for short-term progression.