ModeTv2: GPU-accelerated Motion Decomposition Transformer for Pairwise Optimization in Medical Image Registration

TL;DR

This paper introduces ModeTv2, a GPU-accelerated transformer-based network for efficient and accurate pairwise optimization in deformable medical image registration, improving speed, realism, and usability over traditional methods.

Contribution

The study presents a novel pyramid network with an enhanced motion decomposition transformer and CUDA-optimized operator, advancing deep learning-based deformable image registration.

Findings

Outperforms traditional methods in accuracy and speed

Demonstrates robustness across multiple medical imaging datasets

Provides improved realism and interpretability of deformation fields

Abstract

Deformable image registration plays a crucial role in medical imaging, aiding in disease diagnosis and image-guided interventions. Traditional iterative methods are slow, while deep learning (DL) accelerates solutions but faces usability and precision challenges. This study introduces a pyramid network with the enhanced motion decomposition Transformer (ModeTv2) operator, showcasing superior pairwise optimization (PO) akin to traditional methods. We re-implement ModeT operator with CUDA extensions to enhance its computational efficiency. We further propose RegHead module which refines deformation fields, improves the realism of deformation and reduces parameters. By adopting the PO, the proposed network balances accuracy, efficiency, and generalizability. Extensive experiments on three public brain MRI datasets and one abdominal CT dataset demonstrate the network's suitability for PO, providing a DL model with enhanced usability and interpretability. The code is publicly available at https://github.com/ZAX130/ModeTv2.