Unsupervised 3D registration through optimization-guided cyclical self-training

TL;DR

This paper introduces a novel self-supervised learning approach for 3D registration that leverages cyclical self-training and feature-based differentiable optimization, outperforming existing methods without requiring manual annotations or synthetic data.

Contribution

It proposes a cyclical self-training framework for unsupervised 3D registration that improves feature learning stability and accuracy without manual labels or synthetic data.

Findings

Outperforms state-of-the-art registration methods on abdomen and lung datasets.

Demonstrates the effectiveness of cyclical self-training in stabilizing feature extraction.

Achieves superior registration accuracy compared to metric-based supervision.

Abstract

State-of-the-art deep learning-based registration methods employ three different learning strategies: supervised learning, which requires costly manual annotations, unsupervised learning, which heavily relies on hand-crafted similarity metrics designed by domain experts, or learning from synthetic data, which introduces a domain shift. To overcome the limitations of these strategies, we propose a novel self-supervised learning paradigm for unsupervised registration, relying on self-training. Our idea is based on two key insights. Feature-based differentiable optimizers 1) perform reasonable registration even from random features and 2) stabilize the training of the preceding feature extraction network on noisy labels. Consequently, we propose cyclical self-training, where pseudo labels are initialized as the displacement fields inferred from random features and cyclically updated based on more and more expressive features from the learning feature extractor, yielding a self-reinforcement effect. We evaluate the method for abdomen and lung registration, consistently surpassing metric-based supervision and outperforming diverse state-of-the-art competitors. Source code is available at https://github.com/multimodallearning/reg-cyclical-self-train.