Integrative Diffusion-Weighted Imaging and Radiogenomic Network Analysis of Glioblastoma multiforme

Abstract

In the past, changes of the Apparent Diffusion Coefficient in glioblastoma multiforme have been shown to be related to specific genes and described as being associated with survival. The purpose of this study was to investigate diffusion imaging parameters in combination with genome-wide expression data in order to obtain a comprehensive characterisation of the transcriptomic changes indicated by diffusion imaging parameters. Diffusion-weighted imaging, molecular and clinical data were collected prospectively in 21 patients. Before surgery, MRI diffusion metrics such as axial (AD), radial (RD), mean diffusivity (MD) and fractional anisotropy (FA) were assessed from the contrast enhancing tumour regions. Intraoperatively, tissue was sampled from the same areas using neuronavigation. Transcriptional data of the tissue samples was analysed by Weighted Gene Co-Expression Network Analysis (WGCNA) thus classifying genes into modules based on their network-based affiliations. Subsequent Gene Set Enrichment Analysis (GSEA) identified biological functions or pathways of the expression modules. Network analysis showed a strong association between FA and epithelial-to-mesenchymal-transition (EMT) pathway activation. Also, patients with high FA had a worse clinical outcome. MD correlated with neural function related genes and patients with high MD values had longer overall survival. In conclusion, FA and MD are associated with distinct molecular patterns and opposed clinical outcomes.

Predicting Survival in Glioblastoma Patients Using Diffusion MR Imaging Metrics—A Systematic Review

“…Jamjoom et al [ 59 ] found that lower min MD within enhancing tumour regions predicted shorted OS. Similar behaviour was observed in a study by Heiland et al [ 73 ] for mean MD. Conversely, the mean MD and values were found not predictive for OS in studies by Huber et al and Mohan et al [ 67 , 81 ] and, normalized for normal contralateral brain tissue, predictive for OS only in the univariate analysis performed by Zikou et al [ 47 ], but not in the multivariate one.…”

“…Controversial results were also observed concerning FA. In a study by Heiland et al [ 73 ], a lower FA in the contrast-enhancing region was significantly associated with better OS. Moreover, Flores-Alvares et al [ 79 ] found significant association between the peritumoral oedema measurement of FA with OS.…”

Diffusion kurtosis imaging evaluating epithelial–mesenchymal transition in colorectal carcinoma xenografts model: a preliminary study

“…Recently, machine learning-based approaches including radiomics, radiogenomics, and so on, have drawn more and more attention and demonstrated the potential feasibility in lesion detection and classification, including but are not limited to prostate cancer, bladder cancer, glioblastoma, and non-small cell lung cancer (NSCLC) 18 , 33 – 35 . Radiogenomics referring to the combination of imaging and gene expression is reported to have the potential to give insight into tumor biology.…”

“…Heiland et al . 35 reported that DWI-derived parameter fractional anisotropy (FA) was strongly associated with EMT pathway activation by network analysis with diffusion data. Yamamoto et al .…”

Radiogenomics for Precision Medicine With a Big Data Analytics Perspective

“…The objective here is not to perform an exhaustive review, but rather highlight selected non-TCIA based studies by different groups that use different datasets and also investigate different cancer subtypes. Table V summarizes such approaches, namely for non-small cell lung cancer (NSCLC) [129], [130], [136], LGG [139], and GBM [126], [132], [133], [140], using different radiomic features stemming from multimodal MRI, PET, and CT image modalities.…”