Abstract
Dynamic MRI supports a range of clinical applications, including cardiac function assessment, organ motion tracking, and radiotherapy guidance. However, fully sampling dynamic k-space is often infeasible due to acquisition time constraints and physiological motion, such as respiration and heartbeat. This necessitates undersampling, which degrades reconstructed image quality and can impair deformation field estimation for registering dynamic images to a static reference. Accurate registration is critical for motion correction, treatment planning, and quantitative analysis, particularly in cardiac imaging and MR-guided radiotherapy. Extending E2E-ADS-Recon, we add deformable registration and jointly optimize its adaptive sampling–reconstruction pipeline with registration. The framework first uses a DL-based adaptive sampling strategy to optimize dynamic k-space acquisition for each case. A reconstruction module then produces images from undersampled moving data that are optimized for deformation field estimation, followed by a registration module that aligns the reconstructed dynamic images with a static reference. The framework is independent of specific reconstruction and registration architectures, enabling plug-and-play integration of different modules. It is jointly trained using supervised and unsupervised losses, allowing end-to-end optimization across all components. Controlled experiments and ablations evaluate warped-image similarity on undrsampled cardiac cine MRI, a suitable dynamic setting, and externally on aortic MRI. -8pt Dynamic MRI Cardiac MRI Adaptive Sampling Image Reconstruction and Registration Deep Learning End-to-end Learning
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/RIME_006.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=lrT4gpSDZl
BibTex
@InProceedings{YiaGeo_Endtoend_MICCAISAT2026,
author = { Yiasemis, George AND Sonke, Jan-Jakob AND Teuwen, Jonas},
title = { { End-to-end Adaptive k-space Sampling, Reconstruction and Registration for Dynamic MRI } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17277},
month = {pending},
page = {pending}
}