DeepAtlas: Joint Semi-Supervised Learning of Image Registration and Segmentation

TL;DR

DeepAtlas introduces a joint semi-supervised learning framework for image registration and segmentation, leveraging existing labels and data augmentation to improve accuracy with limited training data in medical imaging.

Contribution

It proposes a novel deep learning approach that jointly learns registration and segmentation, utilizing weak supervision and data augmentation for enhanced performance.

Findings

Achieves large improvements in registration and segmentation accuracy.

Effective with very limited training data, including one-shot scenarios.

Outperforms independent training methods on knee and brain MRI datasets.

Abstract

Deep convolutional neural networks (CNNs) are state-of-the-art for semantic image segmentation, but typically require many labeled training samples. Obtaining 3D segmentations of medical images for supervised training is difficult and labor intensive. Motivated by classical approaches for joint segmentation and registration we therefore propose a deep learning framework that jointly learns networks for image registration and image segmentation. In contrast to previous work on deep unsupervised image registration, which showed the benefit of weak supervision via image segmentations, our approach can use existing segmentations when available and computes them via the segmentation network otherwise, thereby providing the same registration benefit. Conversely, segmentation network training benefits from the registration, which essentially provides a realistic form of data augmentation. Experiments on knee and brain 3D magnetic resonance (MR) images show that our approach achieves large simultaneous improvements of segmentation and registration accuracy (over independently trained networks) and allows training high-quality models with very limited training data. Specifically, in a one-shot-scenario (with only one manually labeled image) our approach increases Dice scores (%) over an unsupervised registration network by 2.7 and 1.8 on the knee and brain images respectively.