Work packages

WP1: Reference MR scanner - hardware development, characterisation, and optimisation

The aim of this work package is to develop, evaluate and test three reference MR low-field scanners. It includes the definition of the technical specifications of the system, functionality, and imaging tests. In total three scanners will be constructed and operated at three national metrology institutes (PTB, INRIM, TUBITAK). The low-field MRI will be based on the open-source low-field scanner OSI² ONE utilising various open-source software and hardware modules from www.opensourceimaging.org/projects. First prototypes of OSI² ONE at LUMC and PTB demonstrated excellent imaging performance and the feasibility to be used for a complete open-source reference MR system.

WP2: Reference MR scanner - software development

The aim of this work package is to develop an open source and portable software platform which allows controlling the reference low-field scanner developed in WP1. The software will provide options for several tasks, including a) starting and shutting down hardware components of the scanner, b) performing adjustment and tune-up tasks, c) registering the subject or object of investigation, d) storing MR protocols and loading previously stored protocols to the measurement control interface, e) executing the pulse sequences, f) retrieving and storing the raw MR data, g) triggering and performing the image reconstruction and h) viewing and exporting the image data. Such a modular platform will enable controlling other low-field MRI systems as well.

WP3: Evaluation and optimisation of MR imaging performance for low-field MR

The aim of this work package is to develop qualitative and quantitative imaging approaches optimised for low-field applications. These methods will be evaluated in phantoms on the low-field reference MR systems built in WP1 and in phantoms and in-vivo on existing low-field MR scanners. The imaging performance of the low-field reference MR scanners is evaluated by comparing the three systems built in WP1 to each other and to at least three other existing low-field MR scanners with a field strength between 0.05 T and 0.6 T.

Open-source image reconstruction techniques will be implemented for the T1-weighted and T2-weighted sequences developed in WP2. In a second step, they will be extended to quantitative T2 and T1,rho mapping. In order to compensate for low SNR we will adopting a MR Fingerprinting approach, i.e. varying the scan parameters during data acquisition and utilising an advanced MR signal model (e.g. extended phase graphs) to obtain quantitative (bio-)physical parameters such as T2 and T1,rho, which have been shown to be promising parameters for characterisation of inflammation and fibrosis in different diseases. This approach will then also be extended to obtain B0 maps without the need of additional scans. The obtained B0 maps will be used to correct for any distortions of the quantitative scan but also any other scans obtained in the same scan session. In order to further improve image quality and accuracy of the obtained quantitative parameters, physics-guided deep-learning based image reconstruction techniques will be developed. Data driven deep learning approaches have been proposed for image reconstruction, however commonly they do not incorporate information about the physics of the measurement system and encoding process. Here we will develop physics-guided approaches that combine both the model of the MR acquisition (including system imperfections) and training data. Training of a deep-learning-based image reconstruction approach will be carried out using simulated data such that ground truth information is available. Based on the in-vivo data obtained in Task 3.4 the performance of these models applied to in-vivo data will be evaluated and improved using transfer learning strategies. The low-field reference systems will be compared to each other and to a range of existing scanners with different designs and field strength using phantom experiments. Parameters to be assessed will include B0 homogeneity and SNR, as well as reproducibility across the low-field reference systems. All the imaging methods developed in this Task will be made open access and implemented into the reconstruction platform developed in WP2.

WP4: Documentation and uncertainties

The aim of this work package is to determine the uncertainties of the low-field MRI and to provide technical documentation for the design and production process of the reference systems in accordance with the requirements of the MDR (EU)2017/745. An important project goal is a fully documented, fully transparent and metrologically characterised low-field MRI reference system. For this purpose, open-source hardware and open-source software documentation will be prepared to allow reproducibility and efficient communication amongst the participating sites and participating network to further improve its performance.