Figure 2
From: Radiomics prognostication model in glioblastoma using diffusion- and perfusion-weighted MRI
Analysis pipeline for this study. The imaging analysis includes acquisition, co-registration, signal intensity normalization for conventional magnetic resonance imaging data, and segmentation. A Cox regression with least absolute shrinkage and selection operator method (LASSO) was applied to select significant radiomic features. The individualized radiomic score is calculated as the sum of each radiomic variable multiplied by a non-zero coefficient from LASSO. Subsequently, a composite prognostic model was built using the features showing a significant association, including the individual radiomic score and clinical predictors.
