Login

Forgot password? Request a new one here!

Page navigation

Challenges »
BRATS2014

How to join BRATS 2014: Brain Tumor Image Segmentation Challenge

Register below, select BRATS2014 as the research unit

How to join BRATS 2014 if you are already registered (e.g. in BRATS2012, BRATS2013 or other Research Unit): Navigate to MySMIR, scroll to "Group Membership" apply for a new Membership by selecting BRATS2014
More information can be found at MICCAI-BRATS 2014

Acknowledgements

Brain tumor image data used in this article were obtained from the MICCAI 2013 Challenge on Multimodal Brain Tumor Segmentation. The challenge database contain fully anonymized images from the Cancer Imaging Archive.

Citation

The outcome of the BRATS2012 and BRATS2013 challenges has been summarized in the following publication. Authors using the BRATS dataset are kindly requested to cite this work:

  • Menze et al., The Multimodal Brain TumorImage Segmentation Benchmark (BRATS), IEEE Trans. Med. Imaging, 2015.Get the citation as BibTex
  • Kistler et. al, The virtual skeleton database: an open access repository for biomedical research and collaboration. JMIR, 2013. (BibTex)

Registration

You have to use your institutional email address for the registration.
Please note that while you can use non-institutional emails (e.g. Gmail, Hotmail, Yahoo, etc.) your request might be refused depending on the license model of the dataset (e.g. non-commercial use), hence, we recommend using an educational or non-profit account.
Register

Evaluation

Ground Truth images

All images are stored as signed 16-bit integers, but only positive values are used. The manual segmentations (Truth) of the patient images have the following four different labels:

    1 for necrosis
    2 for edema
    3 for non-enhancing tumor
    4 for enhancing tumor
    0 for everything else

IMPORTANT

here are 3 requirements for the successfull upload and validation of your segmentation:

  1. Use the MHA filetype to store your segmentations (not mhd) [use short or ushort if you experience any upload problems]
  2. Keep the same labels as the provided truth.mha (see above)
  3. Name your segmentations according to this template: VSD.your_description.###.mha

    replace the ### with the ID of the corresponding Flair MR images. This allows the system to relate your segmentation to the correct training truth. Download an example list for the training data and testing data.

Example 1
    Filename: VSD.Brain.XX.O.MR_Flair.40819.mha
     The VSD-Id would be 40819
    Filename of your segmentation: VSD.Seg_HG_001.40819.mha
Example 2
    Filename: VSD.Brain.XX.O.MR_Flair.35529.mha
     The VSD-Id would be 35529
    Filename of your segmentation: VSD.Seg_LG_011.35529.mha

Evaluation details

The evaluation is done for 3 different tumor sub-compartements:

  • Region 1: complete tumor (labels 1+2+3+4 for patient data, labesl 1+2 for synthetic data)
  • Region 2: Tumor core (labels 1+3+4 for patient data, label 2 for synthetic data)
  • Region 3: Enhancing tumor (label 4 for patient data, n.a. for synthetic data)

Evaluation Results: Testing

Testing results are a summary of single-case evaluations that can be used to benchmark approaches. Per-case results are not available to users as to minimize efforts where methods are fine-tuned to the testing dataset.

Position Dice Positive Predictive Value Sensitivity Kappa Complete tumor Rank Tumor core Rank Enhancing tumor Rank
  complete core enhancing complete core enhancing complete core enhancing  
1 0.70 (2) 0.56 (1) 0.55 (1) 0.75 (2) 0.60 (2) 0.51 (2) 0.68 (2) 0.59 (2) 0.66 (2) 0.87 (1) 1.75 1.50 1.50
2 0.70 (1) 0.55 (3) 0.54 (3) 0.67 (4) 0.52 (3) 0.49 (3) 0.77 (1) 0.67 (1) 0.68 (1) 0.87 (2) 2.00 2.25 2.50
3 0.67 (3) 0.56 (2) 0.54 (2) 0.78 (1) 0.64 (1) 0.48 (4) 0.63 (3) 0.53 (3) 0.66 (3) 0.87 (3) 2.50 2.25 2.75
4 0.60 (4) 0.40 (4) 0.45 (4) 0.72 (3) 0.51 (4) 0.53 (1) 0.54 (4) 0.40 (4) 0.50 (4) 0.86 (4) 3.75 4.00 3.25

Evaluation Results: Training

Position Dice Positive Predictive Value Sensitivity Kappa Complete tumor Rank Tumor core Rank Enhancing tumor Rank
  complete core enhancing complete core enhancing complete core enhancing  
1 0.76 (1) 0.63 (1) 0.54 (1) 0.81 (2) 0.61 (2) 0.52 (1) 0.73 (1) 0.69 (1) 0.62 (1) 0.90 (1) 1.25 1.25 1.00
2 0.73 (2) 0.61 (2) 0.51 (2) 0.82 (1) 0.64 (1) 0.47 (2) 0.67 (2) 0.64 (2) 0.62 (2) 0.89 (2) 1.75 1.75 2.00
3 0.20 (3) 0.16 (3) 0.16 (3) 0.20 (3) 0.17 (3) 0.15 (3) 0.21 (3) 0.17 (3) 0.19 (3) 0.25 (3) 3.00 3.00 3.00
4 0.02 (4) 0.01 (4) 0.00 (4) 0.02 (4) 0.02 (4) 0.01 (4) 0.01 (4) 0.01 (4) 0.00 (4) 0.02 (4) 4.00 4.00 4.00

Evaluation Results: Testing (frozen 10.11.2014)

Testing results are a summary of single-case evaluations that can be used to benchmark approaches. Per-case results are not available to users as to minimize efforts where methods are fine-tuned to the testing dataset.

Position Dice Positive Predictive Value Sensitivity Kappa Complete tumor Rank Tumor core Rank Enhancing tumor Rank
  complete core enhancing complete core enhancing complete core enhancing  
1 0.87 (1) 0.76 (1) 0.72 (2) 0.91 (1) 0.80 (1) 0.69 (1) 0.85 (2) 0.76 (2) 0.81 (2) 1.00 (1) 1.25 1.25 1.50
2 0.87 (2) 0.76 (2) 0.73 (1) 0.90 (2) 0.73 (2) 0.66 (2) 0.85 (1) 0.83 (1) 0.87 (1) 1.00 (2) 1.75 1.75 1.50
3 0.49 (5) 0.41 (3) 0.38 (3) 0.50 (8) 0.47 (5) 0.40 (4) 0.49 (6) 0.39 (5) 0.39 (8) 0.57 (4) 5.75 4.25 4.75
4 0.47 (7) 0.41 (4) 0.33 (10) 0.52 (6) 0.46 (6) 0.34 (5) 0.44 (9) 0.40 (3) 0.36 (10) 0.57 (5) 6.75 4.50 7.75
5 0.60 (3) 0.30 (13) 0.29 (13) 0.86 (3) 0.58 (3) 0.56 (3) 0.53 (3) 0.27 (14) 0.28 (14) 0.96 (3) 3.00 8.25 8.25
6 0.49 (4) 0.40 (5) 0.36 (5) 0.47 (10) 0.46 (7) 0.31 (12) 0.52 (4) 0.39 (4) 0.49 (3) 0.57 (9) 6.75 5.75 7.00
7 0.47 (6) 0.38 (9) 0.36 (4) 0.46 (12) 0.44 (9) 0.34 (6) 0.49 (5) 0.35 (9) 0.45 (4) 0.57 (7) 7.50 9.00 4.75
8 0.46 (8) 0.38 (7) 0.34 (7) 0.48 (9) 0.42 (11) 0.32 (11) 0.46 (8) 0.37 (8) 0.41 (5) 0.57 (6) 7.75 8.50 7.50
9 0.44 (11) 0.36 (10) 0.34 (9) 0.54 (5) 0.48 (4) 0.34 (7) 0.39 (12) 0.31 (12) 0.39 (7) 0.57 (8) 9.00 8.00 8.25
10 0.43 (12) 0.38 (6) 0.34 (6) 0.54 (4) 0.43 (10) 0.33 (9) 0.38 (13) 0.38 (6) 0.41 (6) 0.56 (12) 10.25 8.50 8.25
11 0.46 (9) 0.38 (8) 0.34 (8) 0.47 (11) 0.41 (12) 0.34 (8) 0.47 (7) 0.38 (7) 0.39 (9) 0.57 (11) 9.50 9.50 8.50
12 0.44 (10) 0.34 (11) 0.30 (12) 0.50 (7) 0.44 (8) 0.33 (10) 0.41 (10) 0.33 (10) 0.32 (13) 0.57 (10) 9.25 9.75 11.50
13 0.40 (13) 0.31 (12) 0.31 (11) 0.46 (13) 0.41 (13) 0.30 (13) 0.37 (14) 0.27 (13) 0.36 (11) 0.52 (13) 13.25 12.75 12.00
14 0.37 (14) 0.29 (14) 0.27 (14) 0.37 (14) 0.31 (14) 0.26 (14) 0.39 (11) 0.33 (11) 0.34 (12) 0.46 (14) 13.25 13.25 13.50
15 0.01 (15) 0.01 (15) 0.01 (15) 0.01 (15) 0.01 (15) 0.01 (15) 0.01 (15) 0.01 (15) 0.01 (15) 0.02 (15) 15.00 15.00 15.00

Evaluation Results: Training (frozen 10.11.2014)

Position Dice Positive Predictive Value Sensitivity Kappa Complete tumor Rank Tumor core Rank Enhancing tumor Rank
  complete core enhancing complete core enhancing complete core enhancing  
1 0.05 (1) 0.04 (1) 0.03 (1) 0.04 (2) 0.04 (1) 0.03 (1) 0.06 (1) 0.04 (1) 0.04 (1) 0.07 (2) 1.50 1.00 1.00
2 0.05 (2) 0.02 (2) 0.02 (2) 0.05 (1) 0.03 (3) 0.02 (3) 0.05 (2) 0.02 (2) 0.03 (2) 0.07 (1) 1.50 2.25 2.25
3 0.03 (6) 0.02 (4) 0.01 (3) 0.03 (3) 0.03 (2) 0.03 (2) 0.03 (5) 0.02 (5) 0.01 (3) 0.07 (3) 4.25 3.50 2.75
4 0.03 (3) 0.02 (3) 0.01 (4) 0.03 (5) 0.02 (5) 0.01 (5) 0.03 (4) 0.02 (4) 0.01 (4) 0.04 (5) 4.25 4.25 4.75
5 0.03 (5) 0.01 (6) 0.01 (5) 0.03 (4) 0.02 (4) 0.01 (4) 0.03 (6) 0.01 (6) 0.01 (5) 0.04 (4) 4.75 5.00 4.50
6 0.03 (4) 0.02 (5) 0.01 (6) 0.03 (6) 0.02 (6) 0.01 (6) 0.03 (3) 0.02 (3) 0.01 (6) 0.04 (6) 4.75 5.00 5.75