Multiparametric MR Brain Tumor Imaging through Radiomic Features as a Metric for Guided Radiation Treatment Planning
Rana Gordji1, Edward Florez1, Juliana Sitta1, Charlene Claudio1, Benjamin Rushing1, Khalid Manzoul1, Niki Patel1, Amy Krecker1, Gerri Wilson1, Elliot Varney1, Stella Powell1, Seth Lirette2, Ali Fatemi1, Candace Howard1
1Department of Radiology, University of Mississippi Medical Center, Jackson, MS
2Department of Data Science, University of Mississippi Medical Center, Jackson, MS
We propose the use of multiparametric MRI combined with radiomic features to improve the differentiation of tumor from edema for GTV definition and to differentiate vasogenic from tumor cell infiltration edema. Twenty-five patients with brain tumor and peritumoral edema were assessed: 17 were diagnosed with glioblastoma multiforme (GBM) and 8 with meningioma. After the acquisition process using a 3T-MRI scanner, two neuroradiologists independently used an in-house algorithm to segment two regions of interest (ROI; edema and tumor) in all patients using functional and anatomical MRI sequences. Radiomic features were extracted from all ROIs through different approaches with and without normalization, leading to the calculation of around 300 different parameters for each ROI. Next, a least absolute shrinkage and selection operator (LASSO) analysis was used to isolate the parameters that best differentiated edema from tumors while irrelevant parameters were discarded. Finally, statistical assessment was performed. Receiver operating characteristic results showcase both the best single discriminator to differentiate tumor from edema and the discriminant capacity of the model using all variables selected by LASSO. T1-weighted sequence postcontrast with normalization offered the best tumor classification (AUC>0.97) for patients with GBM with all MRI sequences. For patients with meningioma, a good model of tumor classification was obtained through the T1-weighted sequence without normalization (AUC>0.71). A small subset of radiomic features showed an excellent ability to distinguish edema from tumor tissue through its most discriminating features.