Tumor location, genomic alterations, and radiomic features as predictors of survival in glioblastoma: a Multi-Modal analysis.

PURPOSE

This study aims to identify the impact of tumor location on the survival of glioblastoma (GBM) patients and the associated genetic alterations, using MRI scans from The Cancer Imaging Archive (TCIA) and genomic data from The Cancer Genome Atlas (TCGA). It also seeks to uncover non-invasive radiomic markers related to poor survival outcome for improved prognosis and treatment planning.

METHODS

We analysed pre-operative MRI scans and genomic data from 123 GBM patients (TCIA and TCGA). Tumor locations were determined using our in-house tool, "tumorVQ", followed by Kaplan-Meier survival analysis based on tumor position. Genomic analysis included somatic mutations, copy number variations, fusion genes, and differential gene expression to identify factors linked to poor survival. We extracted radiomic features from T1ce MRI scans using pyRadiomics to analyse their relationship with survival outcomes.

RESULTS

Kaplan-Meier analysis showed worse survival for tumors in the parietal lobe compared to other lobes, especially frontal lobe tumors. Genomic analysis revealed high prevalence of PTEN mutations, and exclusive fusion genes FGFR3-TACC3 and EGFR-SEPT14 in parietal lobe tumors. Differential gene expression showed upregulation of PITX2, HOXB13, and DTHD1, linked to tumor progression, while ALOX15 downregulation increased relapse risk. Copy number alterations, like LINC00290 deletions, were associated with aggressive parietal lobe tumors. Radiomic features, lower GLDM DependanceEntropy (LLL) and higher FirstOrder Mean (HLL), were strongly linked to increase risk.

CONCLUSION

This study highlights poor survival outcomes in GBM patients with parietal lobe tumors. Key genetic alterations, such as PTEN mutations and fusion genes, drive tumor progression and chemoresistance in parietal lobe tumors. The association between radiomic features and survival indicates their potential as non-invasive prognostic biomarkers, which could aid in personalized treatment and improved patient management.