TCIA TCGA-LGG MRI Collection
Overview
The TCIA TCGA-LGG collection provides pre-operative multi-parametric MRI scans for 108 lower-grade diffuse glioma (DIFG) patients drawn from the TCGA-LGG cohort (originating set n=199 from 5 institutions). Bakas et al. 2017 released expert-revised tumour segmentation labels (whole tumour, tumour core, enhancing/non-enhancing tumour) and >700 extracted radiomic features for these cases. The molecular and genomic counterpart is the cBioPortal study lgg_tcga. PMID:28872634
Composition
- Cancer type: lower-grade diffuse glioma (DIFG), n=108 (from originating set of 199).
- Modality: multi-parametric MRI — T1, T1-Gd (post-contrast), T2, T2-FLAIR; retrospective standard-of-care acquisitions from GE, Siemens, Philips, and Hitachi scanners at field strengths 0.5–3 T.
- Pre-processing: identical pipeline to TCGA-GBM — re-orientation to LPS, affine co-registration via FSL FLIRT, 1 mm³ resampling, skull-stripping with BET.
- Annotations: expert-revised segmentation labels; LGG cases without an apparent enhancing tumour region were labelled as non-enhancing tumour (NET) only or NET+oedema, reflecting the lower blood-brain-barrier disruption typical of low-grade glioma biology. PMID:28872634
Assays / panels (linked)
- Volumetric MRI segmentation — expert-revised sub-region labels.
- GLISTRboost — automated segmentation baseline.
- CaPTk — radiomic feature extraction.
Papers using this cohort
- PMID:28872634 — Bakas et al. 2017, Scientific Data: primary data resource paper; releases expert-revised MRI segmentation labels and >700 radiomic features for 108 TCGA-LGG cases.
Notable findings derived from this cohort
- N=44 LGG cases overlap with BraTS’15 training set; N=15 with BraTS’15 testing set. The revised labels became the BraTS’17 reference standard. PMID:28872634
- LGG cases without apparent enhancing tumour are annotated as NET-only or NET+oedema, creating a biologically grounded labelling scheme distinct from the GBM annotation protocol. PMID:28872634
Sources
- TCIA collection: TCGA-LGG — https://www.cancerimagingarchive.net/collection/tcga-lgg/
- cBioPortal genomic counterpart: lgg_tcga
- PMID:28872634 — Bakas et al. 2017, Scientific Data, DOI 10.1038/sdata.2017.117.
This page was processed by crosslinker on 2026-05-04.