As part of  EuCanImage project, the BCN-AIM team recently released stable version 1.0.0 of the medigan python library. 

Using pretrained generators, the library allows to quickly genrate synthetic data that can be used in experiments. Below is a simple example that generates a dataset of e.g. 1000 synthetic polyp images with segmentation marks:

Step 1: pip install medigan

Step 2: python3

from medigan import Generators

Generators().generate(model_id=10, num_samples=1000 install_dependencies=True)

A paper describing the library in detail was written in collaboration with the University of Arkansas for Medical Sciences which can be found here.