Optimal transport estimation improves medical image reconstruction accuracy
Convex Potential Learning for Optimal Transport Between Medical Images
Computer Vision and Pattern Recognition
Summary
Comparing and analyzing shapes in medical images is important but hard when working with large 3D scans. The authors present a new technique to more accurately match and transport information between images by using a special mathematical function that guarantees better consistency. Their method works well on both 2D and 3D medical images, leading to clearer and more precise image comparisons than previous approaches. This can help doctors and medical researchers better understand changes in anatomy.
What this means in practice
- •For medical imaging engineers: Improve the quality of 3D image registration and comparison by integrating more accurate transport maps into analysis pipelines.
- •For computer vision developers: Enhance image morphing and matching software by using curl-free transport maps derived from convex potentials.
Authors
Hongyu Kan, Ivan Medri, Natasha Ironside, Kristofor Pas, Richard Corp, Shinjini Kundu, Gustavo Kunde Rohde
Abstract
Computing optimal transport maps between medical images plays a fundamental role in morphological analysis. However, existing methods are either difficult to scale to large 3D medical images or rely on regularized formulations that provide only approximate transport maps. To improve the accuracy of downstream medical image analysis, we propose a novel approach for estimating optimal transport maps by parameterizing the convex potential function. By construction, the resulting transport map is guaranteed to be curl-free, consistent with Brenier's theorem. We evaluate the proposed method on both 2D and 3D medical imaging datasets. Experimental results demonstrate that our approach achieves lower transport estimation errors and produces higher-quality image reconstructions than existing methods.