Region-wise Comparison of DCE-MRI based Pharmacokinetic Model Fitting in Glioblastoma

Motivation: Glioblastoma (GBM) has poor survival rates, and DCE-MRI aids in understanding tumor vascularity. However, selecting an optimal pharmacokinetic (PK) model remains challenging. Goal(s): To identify the best-fitting PK model for GBM regions using an Akaike Information Criterion (AIC)-based parsimonious model selection. Approach: DCE-MRI data from 46 GBM patients were segmented into enhancing, non-enhancing, and edema regions. Five PK models were fitted per region, with AIC used to select the best model per voxel. A parsimonious model was created from these best-fits. Results: The parsimonious model outperformed individual models across regions, with Extended Tofts, 2CXM and Shutter-Speed models excelling in enhancing/non-enhancing regions and edema. Impact: This study improves GBM assessment by optimizing DCE-MRI PK model selection. It tailors models to specific tumor regions, enabling more accurate quantification of tumor characteristics, aiding diagnosis & treatment planning. This approach can be applied to other tumors and imaging protocols.