Contributions to Advanced Magnetic Resonance Imaging Simulation: A Model for Dynamic Simulation and a Web Interface for Pulse Sequence Development and Visualization

Abstract

Magnetic Resonance Imaging (MRI) simulation is of particular interest due to its ability to recreate a technique that, despite being extremely suitable for many clinical situations, is expensive and not easily accessible to researchers and trainees. Over the last few years, numerous MRI simulators have emerged, both open-source and proprietary. Among them, KomaMRI stands out as the only open-source tool compatible with all Operating Systems, which includes a graphical user interface (GUI) and vendor-agnostic GPU support. Additionally, it is compatible with widely-used MRI community standards and is written in Julia, which enables efficient and extensible code.

In this context, two key areas have been identified for improving the usability and versatility of the simulator. First, the functionality for defining and simulating dynamic phantoms presents room for improvement, as it only allows motion to be described using analytical expressions. Second, in a previous Bachelor's Thesis, a desktop application for designing MRI sequences was developed. This application also shows potential for improvement, and its conversion to a web application was proposed as a future line of work.

This Master's Thesis thus addresses the dual objective of defining a novel dynamic phantom model within the KomaMRI simulator and creating an enhanced web-based application for the editing and simulation of pulse sequences.

For the first objective, the data structure of the simulator has been extended to include information about the motion of the phantom. To achieve this, the global motion of the model has been defined as a list of independent motions, which allows specifying the type, temporal behaviour, and affected spins for each of them. Additionally, simulation functions have been modified to incorporate the calculation of model displacements, and the KomaMRI visualization tool has been improved to allow for the temporal representation of dynamic phantoms. Finally, a new file format has been defined to facilitate the storage and sharing of these digital phantoms. All of this has been developed using the Julia programming language.

For the second objective, a full-stack development has been carried out, addressing both the front-end and the back-end, as well as the communication mechanisms between them. Specifically, the front-end includes an improved version of the previously developed sequence editor, a 3D visualization tool for the selected slice, and two additional panels: one for visualizing the temporal sequence diagram and the other for displaying simulation results. This implementation combines the Qt framework with web technologies such as HTML, JavaScript, and WebAssembly. The back-end, developed in Julia, includes an HTTP server with a REST API, the MRI simulator, and additional modules including the database and front-end files.

Experiments conducted with the dynamic phantom demonstrate the ease of defining and simulating dynamic anatomical models, while also offering reduced simulation times. Furthermore, the obtained results show a high degree of realism, both in demonstrative experiments and those which compare the enhanced version of KomaMRI with other contributions in the field of dynamic MRI simulation. As for the web-based sequence editor evaluation, it highlights its usefulness, interactivity and smoothness, also demonstrating the ability to design and simulate arbitrarily complex pulse sequences without the need for local installations.

The contributions of this work can be summarized as the enhancement of an MRI simulator with improved dynamic phantom simulation capabilities, the definition of a new file format for digital phantoms, and the development of a freely accessible web application for designing and simulating pulse sequences, which benefits both researchers and technical users.