Landmarks Augmentation with Manifold-Barycentric Oversampling

TL;DR

This paper introduces a manifold-preserving data augmentation method using optimal transport theory, specifically designed for landmark detection and medical segmentation, improving over existing augmentation techniques.

Contribution

It presents a novel augmentation approach that maintains data within the original manifold using Wasserstein barycenters, enhancing GAN training and medical data analysis.

Findings

Reduces overfitting in landmark detection tasks

Improves segmentation quality metrics

Outperforms popular augmentation methods

Abstract

The training of Generative Adversarial Networks (GANs) requires a large amount of data, stimulating the development of new augmentation methods to alleviate the challenge. Oftentimes, these methods either fail to produce enough new data or expand the dataset beyond the original manifold. In this paper, we propose a new augmentation method that guarantees to keep the new data within the original data manifold thanks to the optimal transport theory. The proposed algorithm finds cliques in the nearest-neighbors graph and, at each sampling iteration, randomly draws one clique to compute the Wasserstein barycenter with random uniform weights. These barycenters then become the new natural-looking elements that one could add to the dataset. We apply this approach to the problem of landmarks detection and augment the available annotation in both unpaired and in semi-supervised scenarios. Additionally, the idea is validated on cardiac data for the task of medical segmentation. Our approach reduces the overfitting and improves the quality metrics beyond the original data outcome and beyond the result obtained with popular modern augmentation methods.