Nimg-27. Radiogenomic Analysis of Glioblastoma Reveals Textural Features From Mri That Correlate With Genomic Immune Score and Are Also Predictive of Chemo-radiation Treatment Response
Year: 2018
Type: conference-abstract
Abstract: Elevated immune response in Glioblastoma (GBM) is known to improve chemo-radiation therapy (CRT) response. Understanding this manifestation at a radiologic scale may establish non-invasive features as potential predictive biomarkers. We seek to discover pre-treatment radiomic MRI features associated with immune response, and further evaluate their capability to predict CRT response. 74 GBM patients from TCIA and TCGA with pre-treatment MRI and corresponding RNA sequencing prior to CRT were retrospectively analyzed. Immune score (IS) was derived from a 140-gene immune signature via the ESTIMATE algorithm. 1400 3D-radiomic features (Gabor, Haralick, Laws) were extracted from T1w, T2w, FLAIR within expert annotated regions (enhancing tumor, edema, necrosis) per patient. Least absolute shrinkage and selection operator (LASSO) was performed in 100 iterations of 3-fold cross-validation on the training set (n=40) with respect to IS. The 4 most selected radiomic features were used to derive a multilinear regression model of IS. This model was applied to predict IS within an independent set (n=34). Within test set, responders (R) were defined as patients with progression-free survival (PFS) of >= 4-months, while non-responders (NR) had PFS of <4-months. Unsupervised clustering of radiomic features was then implemented to categorize every study as R (n=18) or NR (n=16) to CRT. The radiomic model was significantly associated with IS in training (R2=.61, p <.0001) and testing (R2=.16, p =.018) sets. Top IS-associated features captured wave-like enhancement patterns and intensity value heterogeneity within the tumor or peritumoral edema on T1w and T2w. Within the testing set, unsupervised clustering yielded accuracy=.71 (Sensitivity=0.78, Specificity=.63) in distinguishing R from NR. Elevated IS, indicative of a robust immune response preceding treatment, may manifest as subtle changes in peritumoral edema detectable on pre-treatment MRI. Radiomic correlates of IS may provide means of quantifying immune activity and help predict response to CRT.
Source: Neuro-Oncology
Institution: Case Western Reserve University
Language: English
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2899727660