MRI Sequence Selection for Brain Tumor Segmentation

GetDATA Team · · 1 min read

The sequences a tumour model needs

Brain-tumour segmentation is inherently multi-parametric: each MRI sequence highlights a different aspect of the lesion. Models in the BraTS lineage expect a standard four: T1, contrast-enhanced T1, T2 and FLAIR.

What each contributes

  • T1c (post-gadolinium): enhancing tumour and active margins.
  • FLAIR: peritumoural edema and infiltration.
  • T2: overall lesion extent and fluid.
  • T1: anatomical reference and necrosis.

Harmonisation across scanners

MRI intensities are not standardised, so multi-site data needs intensity normalisation (and often ComBat) before pooling. Document field strength (1.5T vs 3T) and voxel spacing, and prefer BIDS-organised studies.

Sourcing labelled MRI

Post an MRI request on GetDATA specifying the required sequences, field strength, annotation type and case counts; verified providers contribute expert voxel-level segmentations on compliant, de-identified studies.

Need a specific medical dataset?

Post a request describing exactly what you need — modality, labels, format and volume — and verified hospitals and labs fulfill it with compliant, de-identified data.