GAN for tumor prediction

03 Sep 2020

This project aims to predict brain glioma growth from a baseline MRI. It builds upon proof-of-concept results using a extended version of Pix2Pix with pairs of 2D FLAIR brain MRI tagged with timelapse information. We intend to investigate the following sophistications:
(1) Explore Pix2Pix on 3D MRI brain tumour images
(2) Explore the recent Edge-aware Ea-GAN architecture to improve prediction qual- ity
(3) Add a Time Predictor Model to improve growth prediction
(4) Try using several input images from the same patient to predict an ulterior MRI