This new resource provides 1,251 unified eight-class anatomy-lesion label sets derived from the BraTS 2023-GLI cohort. It addresses label noise in joint segmentation tasks by accounting for coexisting white matter hyperintensities (WMH) that previous datasets treated as normal tissue.
HOW THIS AFFECTS YOU
●
researcherYou can now train models on more granular, anatomically accurate labels that include pathological noise.
●
healthThis resource enables more accurate automated MRI segmentation for clinical glioma diagnostics.