medigan: a Python library of pretrained generative models for medical image synthesis

TL;DR

medigan is an open-source Python library that provides easy access to a variety of pretrained generative models for medical image synthesis, facilitating data augmentation and research in medical imaging.

Contribution

medigan introduces a modular, scalable framework offering pretrained generative models, simplifying their use and promoting wider adoption in medical imaging research and clinical applications.

Findings

21 pretrained models across 9 GAN architectures

Demonstrated applications in data sharing, model evaluation, and clinical tasks

Analyzed Fréchet Inception Distance variability in medical images

Abstract

Synthetic data generated by generative models can enhance the performance and capabilities of data-hungry deep learning models in medical imaging. However, there is (1) limited availability of (synthetic) datasets and (2) generative models are complex to train, which hinders their adoption in research and clinical applications. To reduce this entry barrier, we propose medigan, a one-stop shop for pretrained generative models implemented as an open-source framework-agnostic Python library. medigan allows researchers and developers to create, increase, and domain-adapt their training data in just a few lines of code. Guided by design decisions based on gathered end-user requirements, we implement medigan based on modular components for generative model (i) execution, (ii) visualisation, (iii) search & ranking, and (iv) contribution. The library's scalability and design is demonstrated by its growing number of integrated and readily-usable pretrained generative models consisting of 21 models utilising 9 different Generative Adversarial Network architectures trained on 11 datasets from 4 domains, namely, mammography, endoscopy, x-ray, and MRI. Furthermore, 3 applications of medigan are analysed in this work, which include (a) enabling community-wide sharing of restricted data, (b) investigating generative model evaluation metrics, and (c) improving clinical downstream tasks. In (b), extending on common medical image synthesis assessment and reporting standards, we show Fr\'echet Inception Distance variability based on image normalisation and radiology-specific feature extraction.