From Healthy Scans to Annotated Tumors: A Tumor Fabrication Framework for 3D Brain MRI Synthesis

TL;DR

This paper introduces Tumor Fabrication, a two-stage automated framework that synthesizes paired 3D brain MRI tumor images using limited data, significantly enhancing tumor segmentation performance in low-data clinical settings.

Contribution

The paper presents a novel unpaired 3D brain tumor synthesis framework that requires only healthy scans and limited annotations, improving data augmentation for medical image segmentation.

Findings

Synthetic data improves segmentation accuracy in low-data regimes.

The framework reduces manual modeling and expert knowledge requirements.

Generated images effectively augment training datasets.

Abstract

The scarcity of annotated Magnetic Resonance Imaging (MRI) tumor data presents a major obstacle to accurate and automated tumor segmentation. While existing data synthesis methods offer promising solutions, they often suffer from key limitations: manual modeling is labor intensive and requires expert knowledge. Deep generative models may be used to augment data and annotation, but they typically demand large amounts of training pairs in the first place, which is impractical in data limited clinical settings. In this work, we propose Tumor Fabrication (TF), a novel two-stage framework for unpaired 3D brain tumor synthesis. The framework comprises a coarse tumor synthesis process followed by a refinement process powered by a generative model. TF is fully automated and leverages only healthy image scans along with a limited amount of real annotated data to synthesize large volumes of paired synthetic data for enriching downstream supervised segmentation training. We demonstrate that our synthetic image-label pairs used as data enrichment can significantly improve performance on downstream tumor segmentation tasks in low-data regimes, offering a scalable and reliable solution for medical image enrichment and addressing critical challenges in data scarcity for clinical AI applications.