Cloud MR: an Open-Source Software Framework to Democratize MRI Training and Research

Project Summary/Abstract Cloud MR: an open-source software framework to democratize MRI training and research This project is a competing continuation of our project entitled Novel Software Tools for Rational Design and Assessment of MR Coils, which yielded seminal advances in understanding radiofrequency coil performance at high and ultra-high field. It also delivered novel computational tools for rapid coil simulation using Integral Equation techniques, and introduced the ultimate intrinsic transmit efficiency as an absolute metric for transmit coil performance. Our continuing project will integrate these advances into the development of Cloud MR, a comprehensive framework to simulate all aspects of the MRI experiment. By means of an intuitive web-based user interface, Cloud MR will allow the development of RF coils, pulse sequences and image reconstruction methods within an interconnected simulation environment that will enable users to optimize them jointly or individually. We will introduce the first web-based tool for modeling and simulation of flexible RF coils, as well as an innovative tool for pulse sequence development based on a Sequence Description Language. We will train a convolutional neural network for the removal of Gibbs artifacts using synthetic brain images generated with Cloud MR. The overall goal is to provide to anyone with an internet connection a powerful, innovative, comprehensive open-source tool for MRI research and training. By providing a virtual simulation environment to test new technology and optimize clinical protocols without operating an actual MR scanner, Cloud MR will reduce the carbon footprint of MRI. Cloud MR will also allow to generate realistic synthetic MR datasets to train neural networks, without the need to access actual patients’ data. We will distribute all software freely and fully documented, including tutorials and examples that could be used to demonstrate physics and engineering concepts in undergraduate and graduate courses.