Cooperative Training and Latent Space Data Augmentation for Robust Medical Image Segmentation

  title={Cooperative Training and Latent Space Data Augmentation for Robust Medical Image Segmentation},
  author={Chen Chen and Kerstin Hammernik and Cheng Ouyang and Chen Qin and Wenjia Bai and Daniel Rueckert},
Deep learning-based segmentation methods are vulnerable to unforeseen data distribution shifts during deployment, e.g. change of image appearances or contrasts caused by different scanners, unexpected imaging artifacts etc. In this paper, we present a cooperative framework for training image segmentation models and a latent space augmentation method for generating hard examples. Both contributions improve model generalization and robustness with limited data. The cooperative training framework… 
1 Citations

Figures and Tables from this paper

Causality-inspired Single-source Domain Generalization for Medical Image Segmentation
This work investigates the single-source domain generalization problem: training a deep network that is robust to unseen domains, under the condition that training data is only available from one source domain, which is common in medical imaging applications.


Generalizing Deep Learning for Medical Image Segmentation to Unseen Domains via Deep Stacked Transformation
A deep stacked transformation approach for domain generalization that can be generalized to the design of highly robust deep segmentation models for clinical deployment and reaches the performance of state-of theart fully supervised models that are trained and tested on their source domains.
Realistic Adversarial Data Augmentation for MR Image Segmentation
This work proposes an adversarial data augmentation method for training neural networks for medical image segmentation, and shows that such an approach can improve the generalization ability and robustness of models as well as provide significant improvements in low-data scenarios.
Improving the Generalizability of Convolutional Neural Network-Based Segmentation on CMR Images
It is demonstrated that a neural network trained on a single-site single-scanner dataset from the UK Biobank can be successfully applied to segmenting cardiac MR images across different sites and different scanners without substantial loss of accuracy.
Anatomical Priors for Image Segmentation via Post-Processing with Denoising Autoencoders
Post-DAE is proposed, a post-processing method based on denoising autoencoders (DAE) trained using only segmentation masks that can improve the quality of noisy and incorrect segmentation mask obtained with a variety of standard methods, by bringing them back to a feasible space, with almost no extra computational time.
A survey on Image Data Augmentation for Deep Learning
This survey will present existing methods for Data Augmentation, promising developments, and meta-level decisions for implementing DataAugmentation, a data-space solution to the problem of limited data.
U-Net: Convolutional Networks for Biomedical Image Segmentation
It is shown that such a network can be trained end-to-end from very few images and outperforms the prior best method (a sliding-window convolutional network) on the ISBI challenge for segmentation of neuronal structures in electron microscopic stacks.
MRI k-Space Motion Artefact Augmentation: Model Robustness and Task-Specific Uncertainty
This work model patient movement as a sequence of randomly-generated, ‘de-meaned’, rigid 3D affine transforms which, by resampling artefact-free volumes, are then combined in k-space to generate realistic motion artefacts.
Models Genesis: Generic Autodidactic Models for 3D Medical Image Analysis
The authors' extensive experiments demonstrate that their Models Genesis significantly outperform learning from scratch in all five target 3D applications covering both segmentation and classification, and are attributed to the unified self-supervised learning framework, built on a simple yet powerful observation.
Deep Learning Techniques for Automatic MRI Cardiac Multi-Structures Segmentation and Diagnosis: Is the Problem Solved?
How far state-of-the-art deep learning methods can go at assessing CMRI, i.e., segmenting the myocardium and the two ventricles as well as classifying pathologies is measured, to open the door to highly accurate and fully automatic analysis of cardiac CMRI.
Self-Challenging Improves Cross-Domain Generalization
A simple training heuristic, Representation Self-Challenging (RSC), is introduced that significantly improves the generalization of CNN to the out-of-domain data and is presented theoretical properties and conditions of RSC for improving cross-domain generalization.