For each clinical case, the provided data are:
- Pre-operative 3T MRI images, including Gadolinium-enhanced T1w and T2 FLAIR scans
- Intra-operative Ultrasound (iUS) images, as a 3D volume covering the entire tumor region after craniotomy but before dura opening
- Expert-labeled homologous anatomical landmarks, defined on all image modalities. These landmarks will be used to evaluate registration algorithms and rank the challenge participants.
All medical images used for the challenge were acquired for routine clinical care at St Olavs University Hospital (Trondheim, Norway), after patients gave their informed consent. The imaging data are available in both MINC and NIFTI formats. All images, MRI and iUS, are in the same referential.
Training dataset (23 cases)
Training data have been released to the public previously as a dataset article in Medical Physics in 2017, and the 23 patients collection is named REtroSpective Evaluation of Cerebral Tumors (RESECT). Details can be found in the related publication:
- Y. Xiao, M. Fortin, G. Unsgård , H. Rivaz, and I. Reinertsen, “REtroSpective Evaluation of Cerebral Tumors (RESECT): a clinical database of pre-operative MRI and intra-operative ultrasound in low-grade glioma surgeries”. Medical Physics, Vol. 44(7), pp. 3875-3882, 2017.
The dataset can be downloaded at the following address: https://archive.norstore.no/pages/public/datasetDetail.jsf?id=10.11582/2017.00004
The RESECT database contains iUS volumes obtained at different surgical stages: before, during and after tumor resection. For the first edition of the challenge, the participating teams will only use the before resection iUS scans, which consists a subset of 22 patients.
For each of the 23 cases, images to register are named:
Pre-operative MRI: CaseXX-FLAIR.nii.gz and CaseXX-T1.nii.gz
Intra-operative US, before resection: CaseXX-US-before.nii.gz
The landmarks of the data are in MINC tag file format with the extension .tag.
An easy way to visualize the MINC image pair with the homologous landmarks is to use the software 'register' that comes with MINC Toolkit (https://bic-mni.github.io). To view the images, please use the following command in your terminal:
register image1.mnc image2.mnc landmarks.tag
Each row of six entries, in world coordinates (unit = mm), represent a landmark pair. The first 3 columns are the x-, y-, and z-coordinate of landmarks belonging to image1 , and the latter 3 columns are the x-, y-, and z-coordinate of landmarks belonging to image2.
For this challenge, the file CaseXX-MRI-beforeUS.tag contains the landmarks in the MRI and before-resection iUS images.
Methods will be evaluated by comparing the MRI landmarsk after registration with the homologous iUS landmarks (ground truth).
Testing dataset (10 cases)
On Aug 24, 2018, 10 new cases of tumour resection will be distributed only to the participating teams who have submitted their methodology manuscripts before the deadline, upon assigning the agreements. The format of the new cases will be exactly the same as the training data. However, only anatomical landmarks for the MRI images will be given.