cIDIR: Conditioned Implicit Neural Representation for Regularized Deformable Image Registration

TL;DR

cIDIR introduces a novel implicit neural representation-based framework for deformable image registration that efficiently adapts to different regularization hyperparameters without retraining, ensuring smooth and anatomically consistent deformation fields.

Contribution

It presents a conditioned implicit neural representation approach that models a continuous deformation field and allows hyperparameter optimization via automatic differentiation, reducing computational costs.

Findings

Achieves high registration accuracy on DIR-LAB dataset.

Demonstrates robustness across various regularization hyperparameters.

Enables seamless integration of advanced regularization techniques.

Abstract

Regularization is essential in deformable image registration (DIR) to ensure that the estimated Deformation Vector Field (DVF) remains smooth, physically plausible, and anatomically consistent. However, fine-tuning regularization parameters in learning-based DIR frameworks is computationally expensive, often requiring multiple training iterations. To address this, we propose cIDI, a novel DIR framework based on Implicit Neural Representations (INRs) that conditions the registration process on regularization hyperparameters. Unlike conventional methods that require retraining for each regularization hyperparameter setting, cIDIR is trained over a prior distribution of these hyperparameters, then optimized over the regularization hyperparameters by using the segmentations masks as an observation. Additionally, cIDIR models a continuous and differentiable DVF, enabling seamless integration of advanced regularization techniques via automatic differentiation. Evaluated on the DIR-LAB dataset, $\operatorname{cIDIR}$ achieves high accuracy and robustness across the dataset.