Hi all,

I'm looking for MRI datasets with available reference segmentations in order to test some segmentation methods using these ground truth. Up to now i've just found some OASIS segmentations from a MICCAI challenge (MICCAI 2012 Grand Challenge and Workshop on Multi-Atlas Labeling). I also dove into ADNI database (https://ida.loni.usc.edu) but i coudn't find any, although some papers (e.g. Segmentation of MR images via discriminative dictionary learning and sparse coding: Application to hippocampus labeling) point out that there exist.

I was wondering if you can help me find more datasets providing ground truth segmentations, such as MIRIAD, ICBM or the same ADNI.

Thanks,
Oualid

>

Please see the Mindboggle 101 manually labeled brain
images:

http://mindboggle.info/data.html

Cheers,
@rno

Hi Oualid,

In addition to the mindboggle data mentioned by Prof. Klein, you can also find labeled ground truth scans at https://www.nitrc.org/projects/ibsr/.  There are two groups of scans there: "IBSR_V2.0" are the best quality and "20 Normals" are older scans.  IBSR scans have been downloaded and use in hundreds of publications.  

The mindboggle and IBSR scans are available for immediate download at no cost.  There are even more expert-labeled scans available, and while they are not free, they are the highest quality available anywhere.  Please see: https://www.nitrc.org/projects/manuallabels/.  The payment for a subscription to Neuromorphometrics' ever-improving database of labeled scans goes toward increasing the number of labeled scans available in the subscription.  There are 63 scans currently available in the 2012 subscription with subject ages ranging from 5 through 96.  The 2013 subscription (available in 2014) will include additional test-retest scans (same subject scanned and labeled twice) along with 30 scans from the ADNI database!

Andy.

Complete disclosure: I founded Neuromorphometrics 15 years ago (today!)

Hi Oualid,

There is an updated page in the SPM WikiBook on atlases, which now lists several sets of labelled images:

https://en.wikibooks.org/w/index.php?tit...

I hope this is helpful, and that others will feel welcome to add to it and improve it.

Best wishes,
Ged