Evaluation of Radiogenomics for Risk Stratification of Intracranial Aneurysms: A Pilot Study

TL;DR

This pilot study explores combining imaging and blood-based biomarkers to better assess the risk of intracranial aneurysms.

Contribution

The study introduces a novel radiogenomics approach integrating radiomics features and gene expression data for aneurysm risk stratification.

Findings

22 radiomics features and 97 genes were significantly different between symptomatic and asymptomatic aneurysms.

Radiogenomics features provided better separation of high- and low-risk aneurysms than radiomics alone.

Correlations were found between 15 radiomics features and 49 differentially expressed genes.

Abstract

Aneurysm wall enhancement (AWE) has emerged as an imaging biomarker, which could help in risk stratification of intracranial aneurysms (IAs) and also shed light on local pathobiology of the IA wall. In this pilot study, we explored the potential of a radiogenomics approach by combining blood-based biomarkers and AWE for better risk stratification of IAs. Patient specific vessel wall imaging scans and whole blood samples were obtained, and IAs were classified as high-risk or low-risk using two different metrics: symptomatic status (3 symptomatic vs. 13 asymptomatic) and PHASES score (4 with a high score vs. 12 with a low score). Radiomics features (RFs) were extracted from the pre- and post-contrast MRI for all IA sac walls, and significantly different RFs were identified through univariate analysis. RNA sequencing from whole blood samples for these patients was also performed to identify differentially expressed genes (DEGs) between high and low-risk IA groups. Using both risk metrics, we performed principal component analysis (PCA), clustering analysis, and correlation analysis between RFs and genes. Lastly, ontological analyses were carried out to investigate biological mechanisms associated with the DEGs. Our analysis of 16 IAs identified 22 RFs that were significantly different between symptomatic and asymptomatic IAs and 97 genes with at least a 2-fold change and a p-value < 0.01. Examining risk with respect to PHASES score, we identified 10 significant radiomics features and 38 differentially expressed genes. Furthermore, we found a significant correlation between 15 unique RFs and 49 DEGs. Through principal component analysis, we found that DEGs only and radiogenomics features produced a better separation between high- and low-risk, for both risk metrics, than RFs alone. We demonstrated that a radiogenomics approach can help in better risk stratification of IAs.

Figures