Meeting Banner
Abstract #2657

Investigating the Impact of Control Information on Fidelity of Detail Recovery in Latent Diffusion Models for Undersampled MRI Reconstruction

Xingjian Tang1,2, Kai Tong1,2, Linge LI3, Yu ZHANG4, Li YAN1, and Jingwei GUAN1
1Shenzhen Technology University, Shenzhen, China, 2Shenzhen University, Shenzhen, China, 3Huawei, Shenzhen, China, 4Qilu University of Technology, Jinan, China

Synopsis

Keywords: AI Diffusion Models, AI/ML Image Reconstruction

Motivation: ControlNet-based latent diffusion models (LDMs) are widely used in image restoration, but the effectiveness of the control information in MRI reconstruction task is unexplored.

Goal(s): This study investigates the impact of varying control information quality impacts on the generation fidelity of LDMs, aiming for facilitating the translation to clinical applications.

Approach: Three levels of conditional inputs quality generated from different data modalities and network structures were involved to guide the ControlNet-based LDM's reconstructing process.

Results: Visual and quantitative evaluations were conducted, revealing the crucial role of control quality in optimizing model performance and ensuring more reliable reconstructions.

Impact: A key prerequisite for translating LDM-based MRI reconstruction methods into clinical practice is resolving the trade-off between detail richness and fidelity. Our research contributes to advancing solutions for the reconstruction fidelity challenge.

How to access this content:

For one year after publication, abstracts and videos are only open to registrants of this annual meeting. Registrants should use their existing login information. Non-registrant access can be purchased via the ISMRM E-Library.

After one year, current ISMRM & ISMRT members get free access to both the abstracts and videos. Non-members and non-registrants must purchase access via the ISMRM E-Library.

After two years, the meeting proceedings (abstracts) are opened to the public and require no login information. Videos remain behind password for access by members, registrants and E-Library customers.

Click here for more information on becoming a member.

Keywords