LiftReg: Limited Angle 2D/3D Deformable Registration

TL;DR

LiftReg is a deep learning-based 2D/3D deformable registration method trained on simulated data, utilizing backprojected features and a statistical deformation model to improve registration quality, especially in limited angle scenarios.

Contribution

The paper introduces LiftReg, a novel deep registration framework that leverages simulated training data and backprojected features to enhance registration accuracy in limited angle 2D/3D imaging.

Findings

Outperforms existing learning-based registration methods on DirLab dataset.

Utilizes high-quality CT-CT similarity measure for training.

Effectively addresses depth ambiguities in limited angle acquisitions.

Abstract

We propose LiftReg, a 2D/3D deformable registration approach. LiftReg is a deep registration framework which is trained using sets of digitally reconstructed radiographs (DRR) and computed tomography (CT) image pairs. By using simulated training data, LiftReg can use a high-quality CT-CT image similarity measure, which helps the network to learn a high-quality deformation space. To further improve registration quality and to address the inherent depth ambiguities of very limited angle acquisitions, we propose to use features extracted from the backprojected 2D images and a statistical deformation model. We test our approach on the DirLab lung registration dataset and show that it outperforms an existing learning-based pairwise registration approach.