MorphSeek: Fine-grained Latent Representation-Level Policy Optimization for Deformable Image Registration

TL;DR

MorphSeek introduces a novel, fine-grained policy optimization framework for deformable image registration that models spatially continuous deformations in latent space, improving accuracy and efficiency.

Contribution

It proposes a representation-level policy learning paradigm with a stochastic Gaussian policy head, enabling spatially coherent, data-efficient registration across multiple benchmarks.

Findings

Achieves consistent Dice improvements over baselines.

Maintains high label efficiency with minimal parameter cost.

Operates with low latency overhead.

Abstract

Deformable image registration (DIR) remains a fundamental yet challenging problem in medical image analysis, largely due to the prohibitively high-dimensional deformation space of dense displacement fields and the scarcity of voxel-level supervision. Existing reinforcement learning frameworks often project this space into coarse, low-dimensional representations, limiting their ability to capture spatially variant deformations. We propose MorphSeek, a fine-grained representation-level policy optimization paradigm that reformulates DIR as a spatially continuous optimization process in the latent feature space. MorphSeek introduces a stochastic Gaussian policy head atop the encoder to model a distribution over latent features, facilitating efficient exploration and coarse-to-fine refinement. The framework integrates unsupervised warm-up with weakly supervised fine-tuning through Group Relative Policy Optimization, where multi-trajectory sampling stabilizes training and improves label efficiency. Across three 3D registration benchmarks (OASIS brain MRI, LiTS liver CT, and Abdomen MR-CT), MorphSeek achieves consistent Dice improvements over competitive baselines while maintaining high label efficiency with minimal parameter cost and low step-level latency overhead. Beyond optimizer specifics, MorphSeek advances a representation-level policy learning paradigm that achieves spatially coherent and data-efficient deformation optimization, offering a principled, backbone-agnostic, and optimizer-agnostic solution for scalable visual alignment in high-dimensional settings.