Brain Tumor Sequence Registration (BraTS-Reg) Challenge: Establishing Correspondence between Pre-Operative and Follow-up MRI
The Brain Tumor Sequence Registration (BraTS-Reg) challenge is now accepted to run in conjunction with the MICCAI 2022 scientific conference. We will keep updating this page with all related information..
Registration of Magnetic Resonance Imaging (MRI) scans containing pathologies is challenging due to tissue appearance changes, and still an unsolved problem. We organize the first Brain Tumor Sequence Registration (BraTS-Reg) challenge, focusing on estimating correspondences between baseline pre-operative and follow up scans of the same patient diagnosed with a brain glioma. The BraTS-Reg challenge intends to establish a benchmark environment for deformable registration algorithms. The dataset associated with this challenge comprises de-identified multi-institutional multi-parametric MRI (mpMRI) data, curated for each scan’s size and resolution, according to a common anatomical template. The clinical experts of our team have generated extensive annotations of landmarks points within the scans. The “training data” along with these ground truth annotations will be released to participants to design their registration methods, whereas annotations of the “validation” and “test” data will be withheld by the organizers and used to evaluate the containerized algorithms of the participants. We will conduct the quantitative evaluation of the submitted algorithms using several metrics, such as Median Absolute Error and Robustness.
(All deadlines are for 23:59 Eastern Time)
|8 Apr 2022||Registration opens for BraTS-Reg @ MICCAI.|
|8 Apr 2022||Training & Validation data release.|
|8 Apr 2022||Evaluation Platform goes live!|
|2 May 2022||Leaderboard goes live! (https://www.cbica.upenn.edu/BraTSReg22/)|
|1 Jul 2022||Submission of short paper reporting method & results.|
|10 Jul 2022||Submission of containerized algorithm.|
|18 Aug 2022||Contacting top-ranked teams to prepare oral presentations.|
|18-22 Sep 2022||Announcement of top-ranked teams at MICCAI.|
|30 Oct 2022||Camera-ready submission of extended papers for inclusion in the associated workshop proceedings.|
|(All deadlines are for 23:59 Eastern Time)|
Registration of baseline pre-operative (treatment-naïve) and follow-up brain tumor MRI scans is challenging, yet a clinically important task for a multitude of reasons. Brain tissue shows heavy deformations induced by the apparent tumor (also known as mass effect) that following its resection are relaxed due to the relieving pressure from the resected tissue. Such deformations affect the whole brain (including the lateral ventricles) and are not limited to the vicinity of the tumor. This is particularly important as the relationship of the tumor to the lateral ventricles and the deformations to the rest of the brain tissue are important factors in prognosis and treatment planning. Further changes in the peritumoral edematous/infiltrated tissue, potential tumor recurrence, as well as treatment related changes, also affect the brain tissue elasticity. The resected tissue/tumor also relates to missing correspondences, and inconsistent intensity profiles between the follow up and the baseline pre-operative scans.
Taking all the above into consideration, finding spatial correspondences between two longitudinal scans of brain tumor patients, i.e., the registration between the baseline pre-operative and follow-up MRI scans, can advance our mechanistic understanding for these tumors. Specifically, for tumor infiltration and potential recurrence, further contributing in the generation of predictive modelling for related pathophysiological processes, but also in understanding biophysical dynamic and plasticity characteristic of brain tissues, as well as for neurosurgical planning.
The registration between pre-operative and follow-up MRI scans of brain glioma patients, is important yet challenging task. In this challenge, participants are invited to develop deformable image registration algorithm by using the provided clinically acquired training data and the annotations done by our expert clinical neuroradiologists.
The evaluation of the registration between the two scans will be based on manually seeded landmarks (ground truth) in both the pre-operative and the follow-up scans. The performance will be quantitatively evaluated in terms of Median Absolute Error (MAE) and Robustness.
We have identified, curated, and pre-processed retrospective multi-institutional data. The data comprises of pairs of pre-operative baseline and follow-up MRI brain scans (each pair being of the same patient) diagnosed and treated for glioma. The exact multi-parametric MRI (mpMRI) sequences of each timepoint are i) native (T1) and ii) contrastenhancedT1-weighted (T1-CE), iii) T2-weighted and iv) T2Fluid Attenuated Inversion Recovery (FLAIR).
In training phase, participants will be provided with the ground truth annotations along with the MRI data. The ground truth consists of location of some unique landmark points found in baseline scan and their corresponding locations in follow-up scan to develop the deformable registration algorithms.
These landmarks are defined on anatomical markers such as blood vessel bifurcations, the anatomical shape of the cortex, and anatomical landmarks of the midline of the brain. The total number of landmarks vary from case to case and across all cases in the range of 6-50 per scan.
The validation data will be provided to the participants as scan pairs of baseline and follow-up with landmarks provided only for the follow-up scan. Participants will submit coordinates of warped landmark locations in the baseline scan. Also, they will be called to upload their method in a containerized way for evaluation on testing data.
Training & Validation Data availability (April 2022)
Register for the BraTS-Reg challenge, to get access to the NIFTI, skull-stripped, and annotated training data. An independent set of validation scans will be made available to the participants, with the intention to allow them assess the generalizability of their methods on unseen data, via the official evaluation platform. Note that the training data will include its corresponding ground truth, whereas the validation data will have its ground truth data held by the organizers at all times.
Short Paper submission deadline (July 2022)
Participants will have to evaluate their methods on the training and validation datasets, and submit short paper describing their method and results. The organizers will review the paper for sufficient details required to understand and reproduce the algorithm. The challenge participants will be given a chance with the option to extend their individual papers, and hence publish their methods in Springer LNCS proceedings.
Testing phase (July 2022)
Participants will need to submit their method in a containerized form. Note that only participants that have submitted a short paper will be considered in the testing phase.
The BraTS-Reg test data will not be made available to the participating teams at any point during the lifecycle of this challenge.
We will communicate more details on how to create the expected containers soon.
Oral Presentations at MICCAI (August 2022)
The top-ranked participants of the testing phase will be contacted by August to prepare slides for orally presenting their method in MICCAI 2022. The final results of the challenge will also be reported at MICCAI.
Joint post-conference journal paper
Finally, we intend to coordinate a journal meta-analysis manuscript in one of the reputed journals in the domain by extending the preprint appropriately to describe the challenge design, data, clinical relevance, and summarizing the results and insights of the challenge.
Challenge data may be used for all purposes, provided that the challenge is appropriately referenced using the citation given at the bottom of this page.
To request the training and the validation data of the BraTS-Reg 2022 challenge, please follow the registration steps below. Please note that the i) training data includes ground truth annotations, ii) validation data does not include annotations, and iii) testing data are not available to the public or to the challenge participants.
To register for the challenge and request the data of the BraTS-Reg challenge, please follow the steps below:
- Create an account in CBICA's Image Processing Portal (ipp.cbica.upenn.edu) and wait for its approval. Note that a confirmation email will be sent so make sure that you also check your Spam folder.
- Once your IPP account is approved, login to ipp.cbica.upenn.edu and then click on the application "BraTS-Reg 2022: Registration and Data Request", under the "Brain Tumor Sequence Registration (BraTS-Reg) Challenge" group.
- Fill in the requested details and press "Submit Job".
- Once your request is recorded, you will receive an email pointing to the "results" of your submitted job. You need to login to IPP, access the "Results.zip" file, in which you will find the file “REGISTRATION_STATUS.txt” that will provide the links to download the BraTS-Reg data. The training data will include for each subject the 4 structural modalities and ground truth landmarks, whereas the validation data will include only the 4 modalities.
Please note that you are expected to use CBICA's IPP to evaluate your method against the ground truth labels of the validation datasets.
You are free to use and/or refer to the BraTS-Reg datasets in your own research, provided that you always cite the following manuscript:
 B.Baheti, D.Waldmannstetter, S.Chakrabarty, et al., "The Brain Tumor Sequence Registration Challenge: Establishing Correspondence between Pre-Operative and Follow-up MRI scans of diffuse glioma patients", arXiv preprint 2112.06979 (2021)
Feel free to send any communication related to the BraTS-Reg challenge to firstname.lastname@example.org
The leaderboard of Training and Validation phase of BraTSReg challenge is now live at: https://www.cbica.upenn.edu/BraTSReg22/
This is the unranked leaderboard including the numbers of cases each team has uploaded and the latest evaluation metrics for these cases.
The test data will not be released publicly and the participants need to submit their method in a containerized form (Singularity) to be evaluated by the organizers on the hidden test data.
The specific instructions for creating singularity containers will be provided soon. The challenge organizers will run all the methods on the test data using own computational infrastructure to confirm reproducibility of results.
- Bhakti Baheti (University of Pennsylvania)
- Diana Waldmannstetter (University of Zurich and Technical University of Munich)
- Satrajit Chakrabarty (Washington University in Saint Louis)
- Hamed Akbari (University of Pennsylvania)
- Aristeidis Sotiras (Washington University in Saint Louis)
- Bjoern Menze (University of Zurich and Technical University of Munich)
- Spyridon Bakas (University of Pennsylvania)
Clinical Experts & Data Contributors
- Michel Bilello (University of Pennsylvania)
- Benedikt Wiestler (Technical University of Munich)
- Julian Schwarting (Technical University of Munich)
- Syed Abidi (Washington University in Saint Louis)
- Mina Mousa (Washington University in Saint Louis)
- Evan Calabrese (University of California San Francisco)
- Jeffrey Rudie (University of California San Francisco)
- Javier Villanueva-Meyer (University of California San Francisco)
- Daniel S. Marcus (Washington University in Saint Louis)
- Christos Davatzikos (University of Pennsylvania)