Artikel
Integrative analysis of MRI diffusivity parameters and genome-wide expression in glioblastoma multiforme
Suche in Medline nach
Autoren
Veröffentlicht: | 8. Juni 2016 |
---|
Gliederung
Text
Objective: MRI is the standard diagnostic method in high-grade gliomas (HGG). Diffusion tensor imaging (DTI) is commonly used for functional imaging and fiber tracking. This method is a promising way to characterize microstructural changes. Our aim was to study the correlation of diffusion-based MRI data and genome-wide expression in HGG and the potential impact on progression-free (PFS) and overall survival (OS).
Method: We retrospectively analyzed 25 patients with a primary glioblastoma between 2010 and 2014. All patients underwent a presurgical MRI. Diffusion weighted images with DTI scalars (fractional Anisotropy FA, radial diffusivity RD, axial diffusivity AD and mean diffusivity MD) were used to complement conventional MR imaging. Expression analyses were done by gene-expression array (Affymetrix HuGene 2.0) and integrative analysis was performed using individual R-software pipelines. Differentially gene expression analysis was done by paired t-test (limma r-tool) and clustered in Cluster 3.0. PFS and OS of these subgroups were compared by Kaplan-Meier statistics and Cox regression model.
Results: By analyzing all diffusivity parameters and correlated gene expression patterns together, we identified two main groups representing tumors with high and low diffusivity parameters in an unsupervised clustering. We detected a highly significant association of radial, mean and axonal diffusivity with ion transport and synaptic transmission (p<0.05) but a negative correlation with cell-cycle processes and mitotic activation (p<0.05), the mTOR pathway and G2M-checkpoint associated pathways (p <0.05). Both groups with high diffusivity showed a significant better PFS (p<0.05, mean PFS: Cluster I 6.78 months; Cluster II 8.01 months; Cluster III 12.89 months). Cox regression and Kaplan-Meier statistics showed a non-significant trend for a longer overall survival in the subgroup with higher diffusivity values.
Conclusions: We show that diffusivity-based data can improve basic neuroradiology diagnostics and is associated with activation or inhibition of specific pathways in glioblastoma with an impact on the clinical course of the disease. Further analysis with an increased number of patients might allow identification of MRI surrogate parameters with the potential to predict the association with a specific genetic pattern and a typical clinical course.