Abstract

Methods for medical image registration infer geometric transformations that align pairs, or groups, of images by maximising an image similarity metric. This problem is ill-posed as several solutions may have equivalent likelihoods, also optimising purely for image similarity can yield implausible deformable transformations. For these reasons regularization terms are essential to obtain meaningful registration results. However, this requires the introduction of at least one hyperparameter, often termed λ, which serves as a trade-off between loss terms. In some approaches and situations, the quality of the estimated transformation greatly depends on hyperparameter choice, and different choices may be required depending on the characteristics of the data. Analyzing the effect of these hyperparameters requires labelled data, which is not commonly available at test-time. In this paper, we propose a novel method for evaluating the influence of hyperparameters and subsequently selecting an optimal value for given pair of images. Our approach, which we call HyperPredict, implements a Multi-Layer Perceptron that learns the effect of selecting particular hyperparameters for registering an image pair by predicting the resulting segmentation overlap and measures of deformation smoothness. This approach enables us to select optimal hyperparameters at test time without requiring labelled data, removing the need for a one-size-fits-all cross-validation approach. Furthermore, the criteria used to define optimal hyperparameter is flexible post-training, allowing us to efficiently choose specific properties (e.g. overlap of specific anatomical regions of interest, smoothness/plausibility of the final displacement field). We evaluate our proposed method on the OASIS brain MR standard benchmark dataset using a recent deep learning approach (cLapIRN) and an algorithmic method (Niftyreg). Our results demonstrate good performance in predicting the effects of regularization hyperparameters and highlight the benefits of our image-pair specific approach to hyperparameter selection.