Pediatric Tumor MRI Image Generation with Mass Effect

Introduction

This repo contains the code to generate children's MRI images with tumor from healthy image. We consider the deformation of surrounding tissue induced by the tumor, which is called mass effect.
The whole pipeline is stated as follow:

• Training:
• Inference:

File structure

Mass effect prediction

The mass effect prediction method is proposed by .To use it, you should have segmented

Mask2Image algorithm

The domain adaptation method that we choose is SPADE, which is proposed by Nvlab in 2019.

Evaluation

We synthesize 1260 new images. 3 different segmentation methods to

Acknowledgements

Our code borrows heavily from the following repos:
TumorMassEffect
SPADE
pytorch-fid
Pytorch-Medical-Segmentation