ResDCE-Diff : Dynamic contrast enhanced MRI translation in prostate cancer using residual denoising diffusion models

Kishore Kumar, Sriprabha Ramanarayanan, Keerthi Ram, Harsh Kumar Agarwal, Ramesh Venkatesan, Mohanasankar Sivaprakasam

Keywords: Prostate cancer, Dynamic contrast enhanced MRI (DCE-MRI), Diffusion model, Medical image-to-image translation

TL;DR: Dynamic contrast enhanced MRI synthesis using diffusion models from multi-modal inputs

Abstract: Dynamic contrast enhanced MRI (DCE-MRI) identifies early perfusion patterns of aggressive prostate tumors, but its reliance on gadolinium contrast agents limits wider clinical adoption due to safety concerns. Recently, diffusion models offers a potential solution to synthesize contrast-enhanced images directly from non-contrast MRI. Previous diffusion models for prostate DCE-MRI require long inference times as they need hundreds or thousands of sampling steps limiting practical use. Moreover, the reverse generation process for DCE-MRI synthesis starts from pure noise without explicitly utilizing the prior information present in the non-contrast inputs in the diffusion process. We propose ResDCE-diff, a residual denoising diffusion model to synthesize early and late phase DCE-MRI images from non-contrast multimodal inputs (T2, Apparent diffusion coefficient and pre-contrast MRI). The diffusion process shifts anatomical, micro-structurally relevant and physics-informed residual features between the non-contrast inputs and DCE-MRI targets. Extensive experiments using PROSTATEx dataset shows that ResDCE-diff, (i) consistently outperforms previous methods across early and late DCE-MRI phases with improvement margins of +1.29 db and +1.17 dB in PSNR, +0.04 and +0.03 in SSIM respectively, (ii) requires significantly lesser diffusion steps (~15) compared to the baseline diffusion model, and (iii) exhibits relatively higher diagnostically relevant synthesis quality.

Primary Subject Area: Image Synthesis

Secondary Subject Area: Generative Models

Registration Requirement: Yes

Reproducibility: https://github.com/kishore-kk-25/ResDCE-diff

Visa & Travel: Yes

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: 207