Overview
This work introduces an integrated open-source framework designed to support the reproducibility and continuous maintenance of advanced computational Magnetic Resonance Imaging (MRI) methods. It addresses the challenge of jointly designing acquisition and reconstruction techniques by providing an open implementation of both within a unified system.
Research Context
Advanced computational MRI techniques necessitate the joint design of data acquisition protocols and image reconstruction algorithms. For these methods to be reproducible, open implementations encompassing both aspects are crucial. Furthermore, the application of such techniques in clinical settings often requires close integration with MRI scanners, which introduces additional complexities related to long-term reproducibility and system maintenance.
Approach
The core of this research involves augmenting the existing BART toolbox with a new software framework specifically for developing pulse sequences. A key component of this framework is a vendor-specific driver sequence, developed to facilitate the execution of these sequences on clinical MRI scanners. This driver enables online adjustment of all relevant sequence parameters during clinical use. To ensure offline reproducibility, the framework integrates with the Pulseq format, allowing the exact same sequence to be regenerated and processed independently of the scanner environment.
As a proof-of-concept, the researchers implemented several quantitative MRI methods within this proposed framework. These methods included:
- T1 mapping
- Joint water/fat R2* mapping
- B0 mapping
These implementations utilized radial FLASH acquisition and model-based reconstruction techniques. The consistency between online acquisitions (performed on a clinical scanner with the driver) and offline reproductions (using the Pulseq format) was validated through experiments conducted on both phantoms and in vivo subjects.
Findings
The project successfully implemented quantitative MRI methods, incorporating both acquisition and reconstruction components, within the BART framework. The system demonstrated the ability to adapt acquisition parameters and the field of view (FOV) online directly on a clinical MRI system. Consistency was observed between the quantitative parameter maps generated from model-based reconstruction for both online and offline Pulseq-regenerated acquisitions. This agreement indicates that the framework can reliably reproduce MRI data and derived parameters across different operational modes.
Why This Matters
This work establishes a comprehensive, end-to-end open-source framework that facilitates the reproducibility of advanced computational MRI methods. This is particularly relevant given the need for transparent and verifiable research in medical imaging, as well as the transition of research innovations into clinical practice.