Diffuse Low-grade Gliomas Database

Diffuse Low-grade Glioma Database

 

 

Diffuse low-grade glioma (WHO grade II, DLGG) are infiltrative brain tumors that are generally revealed by seizures, in young patients with a normal social and professional life.

 

In this database, we provide 210 different FLAIR MR Images of DLGG at different levels of evolution. The data is provided in DICOM format, and accompanied with expert manual segmentations in XML format.

 

A MATLAB code for conversion of the XML segmentations to binary maps is available here.

 

The goal of this work is to provide data for the evaluation of tumor segmentation methods, as well as probabilistic studies on the tumors' preferential localisations.

 

 

Data

You are free to download a portion of the dataset for non-commercial research and educational purposes. Each zip file contains a subset of the dataset. Each folder corresponds to a subject's DICOM and XML segmentation file. A summary of the list of images available is provided here.

 

8 subjects - 23 MB

31 subjects - 114 MB

33 subjects - 113 MB

43 subjects - 167 MB

38 subjects - 153 MB

25 subjects - 86 MB

32 subjects - 122 MB

 

Probabilistic Atlas

We provide alongside the database, the probablistic atlas built following the methodology introduced in Parisot et al., MICCAI 2011 and Parisot et al., CVPR 2012, aiming at identifying DLGG's preferential locations. All subjects are affinely registered to the same reference template. Each brain tumor is then represented as a node in a graphical model where the edges between nodes is the distance between them. This graph is then clustered, where each cluster represent a preferential location and is associated to a probability map describing the local spatial extension of tumors.

The archive below contains the average registered brain volume, the assignments of all subjects to a location specific cluster, the coordinates of the centers of the clusters in MNI space and the probability maps associated to each cluster.

 

Probabilistic atlas - 6 MB

 

 

References

Work based on the dataset should cite our CVPR 2012 paper:

 

@INPROCEEDINGS{Parisot2012,

author = {Parisot, Sarah and Duffau, Hugues and Chemouny, St{\'e}phane and

Paragios, Nikos},

title = {Graph-based detection, segmentation \& characterization of brain

tumors},

booktitle = {IEEE Conference on Computer Vision and Pattern Recognition - CVPR},

year = {2012},

pages = {988--995},

organization = {IEEE}

}

© Sarah Parisot - 2015

Contact

 

Sarah Parisot

s.parisot[at]imperial.ac.uk

 

Amélie Darlix

Amelie.Darlix[at]icm.unicancer.fr