HNAS-reg: hierarchical neural architecture search for deformable medical image registration

TL;DR

This paper introduces HNAS-Reg, a hierarchical neural architecture search framework that automatically designs efficient CNN architectures for deformable medical image registration, outperforming existing methods in accuracy and size.

Contribution

The paper proposes a novel hierarchical NAS framework combining operation and topology search specifically for medical image registration, reducing computational costs with a partial channel strategy.

Findings

Improved registration accuracy over traditional and unsupervised methods

Reduced model size while maintaining performance

Validated on three MRI datasets with 636 images

Abstract

Convolutional neural networks (CNNs) have been widely used to build deep learning models for medical image registration, but manually designed network architectures are not necessarily optimal. This paper presents a hierarchical NAS framework (HNAS-Reg), consisting of both convolutional operation search and network topology search, to identify the optimal network architecture for deformable medical image registration. To mitigate the computational overhead and memory constraints, a partial channel strategy is utilized without losing optimization quality. Experiments on three datasets, consisting of 636 T1-weighted magnetic resonance images (MRIs), have demonstrated that the proposal method can build a deep learning model with improved image registration accuracy and reduced model size, compared with state-of-the-art image registration approaches, including one representative traditional approach and two unsupervised learning-based approaches.