Results

When teams submit their segmentation results, the evaluation results will be sent to the team contact person by e-mail and will be listed below.

RankTeam nameSubmission nameSubmission dateScoreSequences usedDuration
1XMU_SmartDSP23D Spatial Weighted U-Net for Multi-modality Brain MRI Segmentation29-08-1839T1; T1_IR; FLAIR10 min
2Smartdsp37133D weighted U-shape fully convolutional network26-07-1953T1; T1_IR; FLAIR2 min
3TailHotMulti-modality aggregation network313-04-1868T1; T1_IR; FLAIR13 sec
4WTA23D Cascade convolutional architecture - Method 2223-05-1886T1; T1_IR; FLAIR2 min
5XMU SmartDSP3D CNN with a Cross-modality Channel Attention Scheme317-08-1892T1; T1_IR; FLAIR10 min
6XLab3D Fully CNN with Multi-Modality Feature Fusion310-08-18104T1; T1_IR; FLAIR2 min
7LIVIA_ETSHyperDenseNet206-02-18106T1; T1_IR; FLAIR4 min
8CU_DL2"3D Deep Learning; voxnet2"28-06-16122T1; T1_IR; FLAIR2 min
9CU_DL"3D Deep Learning; voxnet13"16-06-16130T1; T1_IR; FLAIR2 min
10muyaFusion Weighted Network304-07-19134T1; T1_IR; FLAIR10 min
11LRDEFully Convolutional Network20-12-16141T12 sec
12CVSSCNN with Attention Mechanism312-03-19144T1; T1_IR; FLAIR1 min
13MSL-SKKUDeep Convolutional Neural Network19-06-17148T1; T1-IR; FLAIR1.5 min
14QL111111Multi-modality Aggregation Network with Self-attention and Deep Feature Reconstruction218-03-19171T1; T1_IR; FLAIR10 sec
15MDGRUMulti-Dimensional Gated Recurrent Units327-07-16180T1; T1_IR; FLAIR2 min
16FBI/LMB FreiburgU-Net (3D)01-05-16185T1-1mm; T1-IR; FLAIR2 min
17ASTRI_DLU-Net inspired CNN model with 3D context information225-07-18187T1; T1_IR; FLAIR7 min
18PyraMiD-LSTM2NOCC with rounds323-05-16192T1; T1-IR; FLAIR2 min
19THUityModified U-Net315-08-18193T140 sec
20NISTFully CNN for Multi-Modality Brain Tissue Segmentation14-08-18201T1; T1_IR; FLAIR2 min
21AOCAtlas of Classifiers24-12-17220T16 sec
22IDSIAPyraMiD-LSTM05-06-15225T1; T1_IR; FLAIR2 min
23UDSC_DLMulti-UDense-Net214-09-19229T1; T1_IR; FLAIR4 min
24STHHybrid ANN-based Auto-context method203-06-16240T1; T1-IR; FLAIR5 min
25nanand2DeepLab-v3+ with pre-training on FreeSurfer automated labels316-08-18241T115 sec
26holloww2D Rotation-Driven Convolutions30-07-20249T1; T1_IR; FLAIR13 sec
27ISI-NeonatologyMulti-stage voxel classification31-05-14251T11.5 hours
28UNC-IDEALINKS:Learning-based multi-source integration09-02-15254T1; T1_IR; FLAIR3 min
29ASTRI_MSARevised VoxResNet27-08-18275T1; T1_IR; FLAIR1 min
30BCH_CRL_IMAGINE3D patch-wise DenseNet and Patch Fusion224-05-18280T1; T1-IR; FLAIR2 min
31LfBMulti-modality Mix Network17-01-19284T1; T1_IR; FLAIR40 sec
32medical_imaging_nitdgpU-Net Inception and Morphological Gradient Channel24-01-20288T1; FLAIR30 sec
33KSOM GHMFASeTs: MAP-Based with Manifold learning13-05-14291T1; T1_IR; FLAIR23 min
34MNAB2Random Forests21-02-14293T1; T1_IR; FLAIR25 min
35WTA3D Cascade convolutional architecture - Method 1315-05-18295T1; T1_IR; FLAIR5 min
36vicorob UdG T1_FMSSEG using T1 + FLAIR (T1-IR skull)14-01-16302T1; IR; FLAIR2 min
37VBM12VBM12_r738 with WMHC=207-10-15305T16 min
38BIGR2Multi-Feature SVM Classification26-09-13317T1; T1_IR; FLAIR35 min
39vicorob UdG T1MSSEG using only T1 (T1-IR skull)21-01-16323T1; IR2 min
40UofL BioImagingMAP-Based Framework26-09-13358T16 sec
41mehran007Tsallis-Entropy Segmentation through MRF04-04-19368T16.5 min
42NarsilSegmentation using Ensemble Trees26-09-13375T1; FLAIR< 2 min
43CMIVModel-guided Level Sets and Skeletons26-09-13379T1; FLAIR3 min
44UB VPML MedMulti-Atlas with Multiway Cut26-09-13379T1; T1_IR; FLAIR30 min
45draiVGG based Fully Convolutional Neural Network20-01-17391T12 sec
46bigr_neuroAutomatically Trained kNN Classifier26-09-13399T1; FLAIR2 hours
47VIBOT10 M-T1+FConvolutional Neural Network27-02-17401T1; FLAIR3 min
48SPM12_T1_F*SPM12 with T1 and FLAIR sequence31-08-15410T1; FLAIR3 min
49S2_QM2Atlas-based segmentation and AdaBoost28-07-14410T1; T1-IR; FLAIR45 min
50KM-MRF-MASMAP on Intensity, and Local and Multi-atlas Priors28-10-16417T110 min
51RobartsMulti-Atlas with Hierarchical Max-Flow26-09-13418T1-1mm; T1-IR16 min
52SJCEFuzzy Clustering12-04-17421T1<1 min
53CSIM-lab_Mqe-MMRFHybrid Modified Entropy-Based Segmentation with Modified MRF25-06-18425T16 min
54SPM12_T1_IR*SPM12 with T1 and IR sequence31-08-15432T1; IR3 min
55SPM12_T1*SPM12 with T1 sequence31-08-15442T13 min
56MNABRandom Decision Forests26-09-13456T1; T1-IR; FLAIR15 min
57SPM12_T1_IR_F*"SPM12 with T1; IR and FLAIR sequence"31-08-15492T1; IR; FLAIR4 min
58FSL-Seg*FSL (v5.0) seg file31-08-15496T110 min
59LUH-TNTLevel-sets with dictionary learning27-10-14504T1< 1 min
60FreeSurfer*FreeSurfer (v5.3.0)31-08-15505T1-1mm1 hour
61FSL-PVSeg*FSL (v5.0) PVSeg file31-08-15508T110 min
62Jedi Mind MeldAutomated Walks using Machine Learning26-09-13523T1; T1-IR; FLAIR27 sec
63mlmi2014ssAdaBoostM228-07-14526T1; T1-IR; FLAIR5 min
64S2_QMBayesian-based Adaptive Mean-Shift26-09-13540T1; T1-IR; FLAIR1.5 hours
65CSIM-LabMultiple Logistic Classification24-10-17554T1
66ISTB UniBeDecision Forests with spatial regularization04-06-14568T1; T1-IR; FLAIR3 min
67LNMBrainsGaussian Intensity Model26-09-13568T1; T1-IR; FLAIR5 min
68ECOESHModified Expectation Maximization06-01-14611T15 min
* These methods (SPM, FSL and FreeSurfer) were applied by the MRBrainS organizers as described in the MRBrainS journal paper.
2, 3 At the end of the submission name indicates that this is the 2nd or 3rd submission (max 3 resubmissions allowed).

Note: As described below, when teams have equal scores, the standard deviations of the three tissue types and the three evaluation measures over all patients are taken into account.

How is the ranking determined?

Ranking is based on the evaluation results of all 15 test datasets and determined as follows: For each evaluation measure (Dice Coefficient (DC), Modified Hausdorff Distance (HD) and Absolute Volume Difference(AVD)), the mean value over all 15 datasets is determined for white matter (WM), gray matter (GM) and cerebrospinal fluid (CSF). Each team receives a rank (1=best) for each tissue type (GM, WM, CSF) and each evaluation measure (DC, HD-95, AVD) based on the mean value of the evaluation measures over all 15 datasets. The final score is determined by adding the ranks of all tissue types and evaluation measures for each team. The team with the lowest score is ranked number 1. In case two teams would have an equal score, the team with the lowest standard deviation over the tissue types will be ranked first.