Learning Diffeomorphic and Modality-invariant Registration using B-splines

Huaqi Qiu, Chen Qin, Andreas Schuh, Kerstin Hammernik, Daniel Rueckert

Keywords: modality-invariant, multi-modal image registration, mutual information, Parzen window, B-splines, diffeomorphic, deep learning

TL;DR: Diffeomorphic modality-invariant deep learning image registration using a differentiable mutual information and a B-spline free form deformation (FFD) parameterisation of Stationary Velocity Field (SVF)

Abstract: We present a deep learning (DL) registration framework for fast mono-modal and multi-modal image registration using differentiable mutual information and diffeomorphic B-spline free-form deformation (FFD). Deep learning registration has been shown to achieve competitive accuracy and significant speedups from traditional iterative registration methods. In this paper, we propose to use a B-spline FFD parameterisation of Stationary Velocity Field (SVF) to in DL registration in order to achieve smooth diffeomorphic deformation while being computationally-efficient. In contrast to most DL registration methods which use intensity similarity metrics that assume linear intensity relationship, we apply a differentiable variant of a classic similarity metric, mutual information, to achieve robust mono-modal and multi-modal registration. We carefully evaluated our proposed framework on mono- and multi-modal registration using 3D brain MR images and 2D cardiac MR images.

Registration: I acknowledge that publication of this at MIDL and in the proceedings requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: both

Primary Subject Area: Image Registration

Secondary Subject Area: Image Registration

Source Code Url: https://github.com/qiuhuaqi/midir

Data Set Url:

https://camcan-archive.mrc-cbu.cam.ac.uk/dataaccess/,
https://www.ukbiobank.ac.uk/