Page History

Description

This dataset contains DICOM-SEG (DSO) conversions of the  Segmentation Labels and Radiomic Features for the Pre-operative Scans of the TCGA-LGG collection  and  Segmentation Labels and Radiomic Features for the Pre-operative Scans of the TCGA-GBM collection analysis datasets, registered to the original spaces of the DICOM volumes from which they were derived. .
 
The MR volumes and segmentations provided in the original segmentation datasets (T1 pre-contrast, T1 post-contrast, T2, FLAIR) are in NIfTI format, co-registered to an atlas space, and re-sampled to 1mm isotropic resolution. The original automatic and semi-automatic segmentations are also in the space and resolution of this atlas space. This dataset contains DICOM-SEG versions of the same segmentations, transformed back into the space of the DICOM patient datasets released in the original original patient resolutions and orientations found in the TCIA’s TCGA-GBM and  and TCGA-LGG datasets. For each patient in the segmentation dataset, each of the original NIfTI MR volumes has been registered and resampled back to their original This allows users to extract features from MR sequences without introducing interpolation artifacts from isotropic resampling.
 
The process for creating these DSO objects is as follows. Patient data from the original NIfTI datasets were registered and resampled from isotropic space to patient space and resolution using using  3DSlicer’s BRAINSFit module . The affine transformation files from these registrations are saved, and used to register and resample both the semi-automatic and automatic NIfTI segmentations into the spaces of each original MR DICOM dataset. The These transformed NIfTI segmentations are then converted into DICOM-SEG datasets using the software package  dcmqi . Because each MR sequence has a unique patient space and resolution, the resulting dataset contains four sets of registered DSO segmentations for each original NIfTI segmentation, as each segmentation has been registered to the unique spacing and resolution of the untransformed pre-contrast T1, post-contrast T1, T2, and FLAIR DICOM datasets. These resulting NIfTI segmentations are then converted into DICOM-SEG datasets using the software package dcmqi. DICOM-SEG metadata values specifying tissue type, algorithm properties, and study qualities are encoded in JSON objects, which are provided in this dataset. Affine transformations from the original NIfTI datasets to the TCGA DICOM datasets are also made available for download in ITK transform format.Please also cite the following original datasets and manuscript when citing this dataset.

Included in this dataset are the converted DSO volumes, DSO metadata values used in the DSO conversion program dcmqi, and affine transformation files from isotropic space to the original patient space saved in ITK format. Original patient DICOM volumes are also available for download below. A key is provided that maps individual DSO objects to their corresponding DICOM Series UID, to facilitate easier data analysis.
 

Download

...

DICOM Segmentations, Registration Transformations, and DCMQI metadata

...