Shape-consistent Generative Adversarial Networks for multi-modal Medical segmentation maps

TL;DR

This paper introduces a 3D cross-modality GAN that synthesizes cardiac volumes from unpaired CT and MRI data, improving segmentation accuracy on small datasets by leveraging shape consistency and spatial augmentations.

Contribution

The paper presents a novel shape-consistent GAN architecture for multi-modal medical image synthesis and segmentation, specifically addressing limited data scenarios.

Findings

Improved segmentation results on small datasets.

Enhanced generator performance with spatial augmentations.

Outperforms existing methods with only 16 CT and 16 MRI volumes.

Abstract

Image translation across domains for unpaired datasets has gained interest and great improvement lately. In medical imaging, there are multiple imaging modalities, with very different characteristics. Our goal is to use cross-modality adaptation between CT and MRI whole cardiac scans for semantic segmentation. We present a segmentation network using synthesised cardiac volumes for extremely limited datasets. Our solution is based on a 3D cross-modality generative adversarial network to share information between modalities and generate synthesized data using unpaired datasets. Our network utilizes semantic segmentation to improve generator shape consistency, thus creating more realistic synthesised volumes to be used when re-training the segmentation network. We show that improved segmentation can be achieved on small datasets when using spatial augmentations to improve a generative adversarial network. These augmentations improve the generator capabilities, thus enhancing the performance of the Segmentor. Using only 16 CT and 16 MRI cardiovascular volumes, improved results are shown over other segmentation methods while using the suggested architecture.