Meeting Banner
Abstract #0935

Variational Diffusion Models for Motion Correction: Comprehensive Evaluation

Julio A Oscanoa1, Cagan Alkan2, Aizada Nurdinova3, Daniel Abraham2, Kawin Setsompop3, Morteza Mardani4, Daniel Ennis3, John Pauly2, and Shreyas Vasanawala3
1Department of Bioengineering, Stanford University, Stanford, CA, United States, 2Department of Electrical Engineering, Stanford University, Stanford, CA, United States, 3Department of Radiology, Stanford University, Stanford, CA, United States, 4NVIDIA Inc., Santa Clara, CA, United States

Synopsis

Keywords: AI Diffusion Models, Machine Learning/Artificial Intelligence, Generative Models

Motivation: Diffusion models have shown state-of-the-art performance in solving inverse problems, including MRI reconstruction. However, applicability to blind inverse problems, e.g. motion correction, is still limited.

Goal(s): To compare our method based on diffusion models to state-of-the-art reconstruction methods for reduction of retrospectively simulated and prospective motion artifacts in brain imaging.

Approach: We evaluated the mitigation of retrospective motion artifacts on fastMRI brain data (N=100, 1100 slices). Additionally, we evaluated the correction of prospective motion in healthy subjects (N=3). We compare our method to conventional and machine learning-based methods.

Results: Our method outperforms competing methods in both retrospective and prospective cases.

Impact: We demonstrate the value of our blind inverse problem framework based on diffusion models. Our method outperforms state-of-the-art methods for reconstruction with motion correction in both retrospectively and prospectively corrupted data.

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