Home
Welcome to the homepage of project jemris, a high performace computing open source simulator for MRI.
MRI simulations are needed in many cases and deliver many advantages over real systems. Safety considerations for living tissue as well as for the MRI hardware suggest the use of simulations prior to the implementation of new processes on real MRI hardware. Likewise, the work involved in implementing new ideas on MRI scanners oftentimes requires thorough evaluation through simulations. Simulators can also serve as a basis for comparison in quality assurance for data acquisition as well as data processing.
Different approaches to development of an MRI simulations have been published previously. Yet, we argue that jemris differs substantially to those because it:
- utilises a highly optimised library for numerical solutions of differential equations. jemris thus provides exact results where no analytical solution is available. For example, in case of simple slice selection where RF pulses and gradients are applied simultaneously. It is thus well suited and has been successfully applied for designing multi-channel 3D spatially selective excitations.
- out of the box, it can deal with arbitrary RF and gradient shapes, arbitrary multi-channel Tx-Rx coil geometries and configurations, as well as many important physical concepts such as random and deterministic off-resonance effects, concomitant fields, nonlinear gradients, etc.
- utilises a library for symbolic mathematics. Thus many extensions are readily available with no additional programming involved.
- is easy to use. GUIs are provided with which all sequence, hardware and simulation parameters are configured, including nearly arbitrary complex MRI sequences.
- is available on many operating systems, since it has been coded only in ANSI C++, and has been used for computations on Linux, Windows and Mac OS X.
- utilises massive parallel processing and it is available on different hardware architectures. jemris can perform up to reasonable sample sizes and experiment setups on single core hardware. It has been extensively used for simulations on small-scale clusters but has also been deployed on massively parallel setups such as JUGENE, the world's fifth fastest civilian super computer with 292288 computer cores and more than 1 PFLOPS peak performance.
- was developed with extensibility in mind. The incorporated physics model, sequence framework as well as the simulated samples are reconfigurable, replaceable and extensible. For instance a solver is currently implemented, which optionally plugs in replacing the standard Bloch equation solver.
jemris is open source and distributed for a range of operating systems under the GPLv3 license. A scientific publication describing the physical and computer scientific concepts is in press with MRM.