Meeting Banner
Abstract #0387

Self-Consistency-Driven Test-Time Prompt Tuning for All-in-One MR Reconstruction Model

Zhuo-Xu Cui1, Taofeng Xie2, Xuemei Wang3, Wenwu He4, Congcong Liu1, Qingyong Zhu1, Yuanyuan Liu1, Jing Cheng1, Yihang Zhou1, and Dong Liang1
1Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China, 2Inner Mongolia Medical University, Hohhot, China, 3Inner Mongolia Medical University Affiliated Hospital, Hohhot, China, 4Fujian University of Technology, Fuzhou, China

Synopsis

Keywords: AI/ML Image Reconstruction, AI/ML Image Reconstruction

Motivation: Recently, all-in-one image restoration models based on prompt learning have been proposed. However, MRI is influenced by numerous factors, introducing additional complexity that challenges the development of an all-in-one model specifically MRI.

Goal(s): Develop an all-in-one model tailored for MRI.

Approach: It remains challenging to build an MRI model generalized across scan location, sequence parameters, sampling methods through prompt learning. Inspired by scan-specific models, we incorporate test-time tuning of prompts (TTTP) using MRI physics-informed priors, enabling scan-specific adjustments to achieve robust generalization across various scenarios.

Results: Through TTTP, an all-in-one MRI model is constructed, effectively generalizing across scan regions, sequence parameters, sampling methods.

Impact: This model can adapt to all MRI scenarios, facilitating simplified installation and maintenance.

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