On the Degrees of Freedom of Gridded Control Points in Learning-Based Medical Image Registration

2026-03-15

AI Summary

Medical image registration often becomes ill-posed in homogeneous or noisy regions, and conventional voxel-wise dense decoders incur substantial computational and memory costs. To address these challenges, this work proposes GridReg, a framework that replaces dense displacement fields with sparse grid-based control points, integrating multi-scale 3D encoder features via a cross-attention mechanism to achieve efficient and robust registration. The method innovatively introduces an adaptive grid training strategy, enabling flexible support for varying grid densities during inference without retraining, and systematically quantifies the impact of control point degrees of freedom on registration performance. Evaluated on three datasets—prostate, pelvic organs, and neural structures—GridReg significantly outperforms existing dense displacement field (DDF) and keypoint-based displacement prediction approaches at comparable or lower computational cost.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments

Abstract

Many registration problems are ill-posed in homogeneous or noisy regions, and dense voxel-wise decoders can be unnecessarily high-dimensional. A sparse control-point parameterisation provides a compact, smooth deformation representation while reducing memory and improving stability. This work investigates the required control points for learning-based registration network development. We present GridReg, a learning-based registration framework that replaces dense voxel-wise decoding with displacement predictions at a sparse grid of control points. This design substantially cuts the parameter count and memory while retaining registration accuracy. Multiscale 3D encoder feature maps are flattened into a 1D token sequence with positional encoding to retain spatial context. The model then predicts a sparse gridded deformation field using a cross-attention module. We further introduce grid-adaptive training, enabling an adaptive model to operate at multiple grid sizes at inference without retraining. This work quantitatively demonstrates the benefits of using sparse grids. Using three data sets for registering prostate gland, pelvic organs and neurological structures, the results suggested a significant improvement with the usage of grid-controled displacement field. Alternatively, the superior registration performance was obtained using the proposed approach, with a similar or less computational cost, compared with existing algorithms that predict DDFs or displacements sampled on scattered key points.

Problem

• medical image registration
• ill-posed problem
• sparse control points
• degrees of freedom
• deformation representation

Innovation

• sparse control points
• learning-based registration
• grid-adaptive training
• cross-attention deformation prediction
• medical image registration