DeepFCD

Detect focal cortical dysplasias using a generative modelling approach as presented on OHBM 2019 in Rome (see our homepage for e-Poster, OHBM oral session and more information: https://www.translationalneuroimaging.de/ohbm-gan-talk ).

Repository is WIP, nothing interesting to see here yet. If you need a state-of-the-art program to automatically detect the FCDs in your datset, please check out MAP18 (https://www.translationalneuroimaging.de/map18).

Software used

Python 3.7.4
Tensorflow 2.1.0 (& 2.2.0-dev for saving and loading models using a CPU only)
Nibabel 2.5.1
Numpy 1.17.2
PIL 6.2.0

For multi-channel 2D-GAN and 3D-CNN training:

CUDA release 10.2

GAN inspired by pix2pix (Isola, Zhu et al., 2016)
(github repo: https://github.com/phillipi/pix2pix)

GAN example output

real FLAIR - real T1 - synthetic FLAIR - synthetic T1