Test-Time Training for Deformable Multi-Scale Image Registration

Abstract

Registration is a fundamental task in medical robotics and is often a crucial step for many downstream tasks such as motion analysis, intra-operative tracking and image segmentation. Popular registration methods such as ANTs and NiftyReg optimize objective functions for each pair of images from scratch, which are time-consuming for 3D and sequential images with complex deformations. Recently, deep learningbased registration approaches such as VoxelMorph have been emerging and achieve competitive performance. In this work, we construct a test-time training for deep deformable image registration to improve the generalization ability of conventional learning-based registration model. We design multi-scale deep networks to consecutively model the residual deformations, which is effective for high variational deformations. Extensive experiments validate the effectiveness of multi-scale deep registration with test-time training based on Dice coefficient for image segmentation and mean square error (MSE), normalized local cross-correlation (NLCC) for tissue dense tracking tasks.

Meta-Learning Initializations for Interactive Medical Image Registration

“…Learning-based methods have used different architectures, such as convolutional neural networks [1][2] and vision transformers [3], different training strategies, such as generative adversarial networks [4,5], supervised [1,4], unsupervised [2,[6][7][8] or reinforcement learning [9][10][11], or different transformation constraints, based on parametric splines [6], diffeomorphism [12] and biomechanics [13]. Semi-supervised learning [14], few-shotand meta-learning [15][16], unsupervised contrastive learning [17], inference-time augmentation [16,18], and amortized hyperparameter learning [19] methodologies have also been used to improve data efficiency and generalizability. For further discussion on these learning-based registration methods, readers are referred to recent systematic surveys [20][21][22].…”