Conditional Learned Reconstruction for Medical Imaging

Nikita Moriakov, George Yiasemis, Jan-Jakob Sonke, Jonas Teuwen

Keywords: Inverse Problems, Medical Imaging, Conditional Learned Reconstruction

TL;DR: We propose a conditional reconstruction framework that adapts learned iterative networks to acquisition settings and show that this consistently improves MRI and CT reconstruction over non-conditional baselines.

Abstract: Medical imaging utilizes a handful of different imaging modalities such as tomography and magnetic resonance (MRI) imaging that require solving an inverse problem to reconstruct an image from the acquired measurements. Reconstruction methods based on learned iterative schemes have been widely explored recently, however, these modalities involve variability in hardware- and protocol-dependent acquisition parameters such as tube current and projection count in case of tomography and acceleration factor or field strength in case of MRI, which are typically not accounted for in the architecture. In this work we propose the framework of conditional learned iterative schemes, where the network weights are explicitly adapted as learned functions of the acquisition parameters. We compare conditional learned iterative schemes to their counterparts without conditioning for both tomography and MRI and demonstrate their effectiveness.

Primary Subject Area: Image Acquisition and Reconstruction

Secondary Subject Area: Application: Radiology

Registration Requirement: Yes

Reproducibility: https://github.com/NKI-AI/direct/blob/nuconv/direct/nn/conv/modulated_conv.py

Visa & Travel: No

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Midl Latex Submission Checklist: Ensure no LaTeX errors during compilation., Replace NNN with your OpenReview submission ID., Includes \documentclass{midl}, \jmlryear{2026}, \jmlrworkshop, \jmlrvolume, \editors, and correct \bibliography command., Did not override options of the hyperref package., Did not use the times package., Use the correct spelling and format, avoid Unicode characters, and use LaTeX equivalents instead., Any math in the title and abstract must be enclosed within $...$., Did not override the bibliography style defined in midl.cls and did not use \begin{thebibliography} directly to insert references., Avoid using \scalebox; use \resizebox when needed., Included all necessary figures and removed *unused* files in the zip archive., Removed special formatting, visual annotations, and highlights used during rebuttal., All special characters in the paper and .bib file use LaTeX commands (e.g., \'e for é)., No separate supplementary PDF uploads., Acknowledgements, references, and appendix must start after the main content.

Submission Number: 61