Available here.
All BraTS multimodal scans are available as NIfTI files (.nii.gz) and describe a) native (T1) and b) post-contrast T1-weighted (T1Gd), c) T2-weighted (T2), and d) T2 Fluid Attenuated Inversion Recovery (T2-FLAIR) volumes.
Annotations comprise the GD-enhancing tumor (ET — label 4), the peritumoral edema (ED — label 2), and the necrotic and non-enhancing tumor core (NCR/NET — label 1).
multimodal slices with segmented mask:
3d projections of multimodal scans and segmented mask:
You can also see 3D data projection here
-
- Each pixel must be labeled “1” if it is part of one of the classes (NCR/NET — label 1, ED — label 2, ET — label 4), and “0” if not.
-
- Make a prediction of age and survival days for each unique identifier in the data.
-
- For automatical segmentation was used Unet3d with group normal layers. - unet
-
- To predict age and number of days of survival - the autoencoder was trained to scale the space from 4 * 240 * 240 * 150 to 512, then statistical values, and hidden representations were extracted for each identifier in the data, encoded by the pretrained autoencoder. after wich SVR was trained on this data. - autoencoder
Unet Result:
AutoEncoder Result:
