1. Background introduction

The brain tumor segmentation challenge (BraTS Challenge) is the oldest among all competitions of the International Medical Image Computing and Computer Assisted Intervention Society (MICCAI), and has been held for 10 consecutive years , is one of the hottest competitions in medical image processing.

Each challenge in 2017 and after contains three data sets, namely training set (Training data), validation set (Validation data) and test set (Test data)

The images and labels of the training set and the images of the verification set can be downloaded through official channels and Kaggle. The address is at the end of the article. The verification set label is not public, but the inference results can be uploaded on the official platform, and the platform will automatically score. The test set is only available for players to download during the competition time.

Table 1 Data scale of the BraTS Challenge over the years

years	total number of cases	Number of cases in the training set	The number of cases in the validation set	Number of cases in the test set
2012	50	35	none	15
2013	60	35	none	25
2014	238	200	none	38
2015	253	200	none	53
2016	391	200	none	191
2017	477	285	46	146
2018	542	285	66	191
2019	626	335	125	166
2020	660	369	125	166
2021	2040	1251	219	570

2. Data description

Each case of BraTS contains four modalities of MRI (Magnetic Resonance Imaging, MRI), and the dimension of each modality is 240×240×155 (L×W×H)

Four modes:

T1 imaging is good for observing the anatomical structure, but the lesion is not clear enough

T1ce

Inject a contrast agent into the blood before the subject undergoes MRI to make the area with active blood flow more obvious in the imaging, which is an important criterion for tumor enhancement

T2 imaging, the lesion is clearly displayed, and the entire tumor can be judged

FLAIR

T2 pressure water image (suppresses the high signal of cerebrospinal fluid), the more water content is more eye-catching, it can judge the edema area around the tumor