Summary
When medical images are taken over time but only at a few points, guessing what happens in between can be tricky because different ways of moving can fit the start and end images equally well. The authors create a way to learn typical motion patterns from many examples, which helps limit these confusing guesses while still allowing for differences in individual cases. Their method breaks down movement into basic directions that can be combined and adjusted smoothly, helping to reconstruct intermediate images more reliably. Tests on heart and lung scans show their method works better than others that don't use intermediate image data, and competitively compared to those that do.
What this means in practice
- •For medical imaging engineers: Improve time-sequence image reconstruction in cardiac and lung scans when intermediate images are unavailable by constraining motion estimates with learned motion patterns.
- •For computational radiology teams: Enhance 4D image analysis workflows by reducing ambiguous motion estimates, supporting more accurate models of organ movement over time.
Authors
Haojin Li, Huazhu Fu, Heng Li, Jiang Liu
Abstract
Unsupervised 4D medical image interpolation reconstructs intermediate anatomical states from temporally sparse observations without intermediate supervision. Distinct motion fields may attain comparable endpoint agreement while producing different intermediate anatomy, giving rise to motion ambiguity. This ambiguity is difficult to constrain when many independently adjustable motion variables must be estimated from the observed endpoints. We develop a framework that learns a motion manifold to restrict candidate motion fields while retaining sample-specific variation. Cross-Sample Spectral Basis Learning extracts recurring motion structure from endpoint-derived dense estimates and encodes it in a shared spectral basis. Sample-Conditioned Cayley Basis Reorientation uses predicted motion coordinates to rotate the effective motion directions for each endpoint pair without increasing the number of independent coordinates. Their composition defines a nonlinear family of complete velocity fields across endpoint pairs, accommodating anatomical and motion-state differences within a restricted representation. Integrating each predicted field supports bidirectional reconstruction at arbitrary intermediate times. Theoretical analysis establishes sufficient conditions under which endpoint observations distinguish nearby motions within the learned family. Experiments on ACDC and 4D-Lung demonstrate state-of-the-art performance among the evaluated methods trained without intermediate volumes and competitive results against methods using intermediate supervision.