Synthetic 3D brain data

• Full pipeline from data preprocessing to synthetic data generation
• Designed to work with moderately low amount (≈1000 samples) of input data

Input(s):

• Real 3D brain scans
• MRI, CT, …

Main feature(s):

• Data preprocessing, registration, augmentation
• Multi-GPU GAN training

Output(s):

• Synthetic 3D brain data

• Implemented as Docker images, which can be converted to Singularity images
• Multiple GPUs and lots of RAM (>100 GB) recommended

Researchers, data scientists.

• Mostly built on open-source technology
• Component licenses: Apace 2.0 and MIT