Skip to main content
Brain Informatics
Fig. 1 | Brain Informatics

Fig. 1

From: Deep learning-based algorithm for postoperative glioblastoma MRI segmentation: a promising new tool for tumor burden assessment

Fig. 1

Application of the final BraTS harmonization pipeline for the patients in the Molinette dataset, comprising SRI-24 co-registration performed with CaPTk and skull-stripping by SynthStrip. A Preoperative (first and whole row—FLAIR; second, necrosis, core rows—T1ce). B Postoperative (first and whole rows—FLAIR; second, cavity, enhancing rows—T1ce). The previous figure comprises in detail all the steps required in the pipeline: N4 bias correction for magnetic field inhomogeneities, LPS/RAI voxel re-orientation, SRI-24 co-registration and skull-stripping. The first two rows show the harmonization pipeline for two examples coming from the BraTS dataset (left, preoperative) and the Molinette dataset (right, postoperative). It is worth noticing that the Atlas co-registration modifies the depth dimension (i.e., the number of slices), therefore the most similar interpolated slice is here shown for visual representation

Back to article page