Open‐source magnetic resonance imaging: Improving access, science, and education through global collaboration

Open‐source practices and resources in magnetic resonance imaging (MRI) have increased substantially in recent years. This trend started with software and data being published open‐source and, more recently, open‐source hardware designs have become increasingly available. These developments towards a culture of sharing and establishing nonexclusive global collaborations have already improved the reproducibility and reusability of code and designs, while providing a more inclusive approach, especially for low‐income settings. Community‐driven standardization and documentation efforts are further strengthening and expanding these milestones. The future of open‐source MRI is bright and we have just started to discover its full collaborative potential. In this review we will give an overview of open‐source software and open‐source hardware projects in human MRI research.

form for open-source software, while the Open Source Imaging Initiative (OSI 2 , https://www.opensourceimaging.org)focuses on both software and hardware. 24,25Furthermore, thematically specialized initiatives have emerged to address specific research interests and improve open science practices.Examples include the Open Science Initiative for Perfusion Imaging, 26,27 MRSHub with open-source resources for MR spectroscopy, 28 and the Renal MRI Initiative. 29e of the most ambitious goals was set by OSI 2 upon its foundation in 2016: to develop a complete open-source MRI system.Since then, remarkable progress has been made with the first open-source hardware prototypes being presented in 2022, such as the low-field MRI system OSI 2 ONE. 30,31This rapid advancement highlights the potential of collaborative efforts within the research community.Currently, these systems are diffusing to a wider user range, including industry collaboration.An exciting milestone was achieved when the first OSI 2 ONE system was successfully reproduced in Uganda, a remote place on the MR map. 33To ensure that these technologies eventually reach patients, it is crucial to include sharing of documentation that can be used to facilitate regulatory approval as a medical device.Currently, such documentation blueprints are not publicly available yet, but work is underway to address this challenge. 32[39] The future of open-source MRI is bright, and we have only begun to discover its full collaborative potential.In this review, we aim to provide an overview of current open-source software and open-source hardware in human MRI research.We will delve into the concept of open-source, including its corresponding licensing and documentation models, and discuss the challenges involved in implementing open-source practices.
While there are numerous exciting open-source projects in existence, it is not possible to cover them all comprehensively within this review.For those seeking further information, websites like https://ismrm.github.io/mrhub/and https://www.opensourceimaging.orgserve as excellent starting points.[42][43] Compressed lists of all open-source software and hardware projects covered in this review are given in Table 1 and Table 2.These lists are also available online (https://gitlab.com/osii-one/osii-orga) to provide a dynamic resource that can be updated with new project information or new projects by everyone.

| Open-source licenses and documentation
The term "open-source" was defined by the OSI and is based on the Free Software Definition by the FSF. 129It means that the user is granted full control to freely use, study, modify, redistribute, and also commercially exploit the shared material.These definitions were primarily written for software source code, and software licenses must meet those requirements to qualify as free or open-source software.Both the OSI and FSF maintain a list of officially endorsed licenses. 130However, the "freedoms" or "rights of open source" granted by free/open-source licenses can be applied to any material, including hardware.The OSHWA has published a definition clarifying the requirements for respective licensing terms for open-source hardware. 131Today, there is a large variety of open-source copyright licenses available, optimized for different types of materials (software, hardware, images, fonts, etc.) and specific use cases (e.g., domain-specific liability or associated patents).However, especially for hardware, other domains of intellectual property (IP) rights may apply, most notably patent law or industrial design rights.To make hardware open source, it must be "freed" from all applicable domains of IP restrictions.Most commonly this is done by publishing the hardware design (1) under an open-source license, and (2) in a way that it is easily found on the internet (also called "defensive publishing").The open-source license in turn ensures that the design files can be freely used, studied, modified, and redistributed.Open-source licensing schemes are often classified by their reciprocality, which determines the preservation of rights in derivative works (Table 1).Version 2.0 of the popular hardware-specific CERN Open Hardware License (CERN OHL) is available in three variants, reflecting these licensing schemes. 21,132Licenses that prohibit commercial use or modification of the material (such as CC BY-NC or CC BY-ND) do not meet the licensing requirements of any of the abovementioned definitions and hence are not open-source licenses.
For effective reuse and widespread decentralized development of open-source hardware designs, complete and comprehensive technical documentation is crucial. 19,133This includes the bill of materials, design files in their native format (e.g., 3D computer-aided design [CAD] models) and manufacturing-oriented export formats (e.g., STL files for 3D printing).It is important to note that the open-source hardware projects listed in this review may not adhere to a unified documentation standard, and often files that are needed to fully reproduce the designs are missing.Nevertheless, important pioneering steps of sharing designs have been performed and current community work is focusing on establishing documentation standards to improve documentation quality.Export file formats (e.g., STEP) can be processed using open-source software, making the underlying information extractable.They can also support replication if they are delivered in a productionready format (e.g., STL for 3D printing).A full regularly updated list can be found at https://gitlab.com/osii/cab.In this review we are going to focus on open-source software projects that do not just reproduce the results of a single scientific paper, but provide a wider range of functionality.In addition, we only include software projects that provide reasonable long-time support, that is, which are compatible with up-to-date operating systems and software versions.
We will cover three main areas of open-source software: (i) simulation of MR data acquisition and design of vendor-independent MR sequences; (ii) image reconstruction; and (iii) postprocessing of reconstructed images.It is worth noting that several of the software packages discussed cover multiple areas.

| MR simulations and sequences
Several open-source tools have been developed to simulate different aspects of the MR acquisition process.These tools enable the evaluation of the impact of pulse sequences on the measured data, resulting from changing a simple acquisition parameter such as the echo time to making more profound changes to specific pulse sequence elements, such as the characteristics of a radio frequency (RF) pulse.

| JEMRIS
JEMRIS, written in C++ but taking advantage of a MATLAB GUI, enables fast and efficient MR simulations. 63,138Pulse sequences can be generated by concatenating multiple sequence objects, such as RF and gradient objects, each having its own set of attributes.It enables the simulation of B1 + transmit field inhomogeneity and of multichannel RF transmit systems; the latter is done by providing multiple RF waveforms, each assigned to a given coil channel.Bio-Savart transmit loops can be simulated, and the option to read external coil files is also provided.Besides predefined waveforms, this software can simulate arbitrary RF and gradient waveforms that can be defined using symbolic analytical formulas.
Potential applications include simulating nonlinear gradient encoding or the impact of concomitant fields.This tool has been extended to simulate realistic flow motion. 139More recently, JEMRIS has been integrated within an end-to-end open-source MR imaging workflow. 140This workflow enables running a simulation with JEMRIS or to acquire data on a real scanner using the exact same pulse sequence, making it easier to carry out sequence prototyping and debugging.To execute the pulse sequence on different scanners and reconstruct images, Pulseq interpreters, 77 the ISMRMRD open data format, 61 and open-source reconstruction tools such as BART 48 are utilized (Figure 1).

| MRILab
MRILab also enables carrying out numerical simulations of MRI experiments. 72,141The user interfaces run on Matlab but interface with C++ code and take advantage of GPU capabilities to enable running on a standard PC.When it was first made available, the main differentiating feature of this software was to model tissues composed of multiple compartments.This makes it possible to simulate quantitative magnetization transfer (MT), multicomponent T 1 and T 2 relaxometry, as well as chemical-exchange saturation-transfer (CEST) experiments.This feature is currently also available in JEMRIS.

| KomaMRI.jl
An alternative MRI simulation platform has been implemented recently using the Julia programming language. 65,142It has the advantage of providing cross-platform compatibility and high CPU and GPU parallelization, enabling fast computations.It can read pulse sequence files in Pulseq format (see below) and outputs raw data using the ISMRMD standard.Furthermore, it provides integration with the MRIReco.jlreconstruction software package and other reconstruction programs compatible with the ISMRMD raw data file format.

| ODIN
ODIN is a GUI-supported C++ software framework to develop and simulate MR sequences. 76Contemporary MRI techniques are available: for example, sequence modules for echo planar imaging and spiral-imaging, parallel imaging with GRAPPA reconstruction, 2D-pulses and field mapbased distortion correction.

| Pulseq, PyPulseq, and TOPPE
Pulse sequence design using vendor-specific code requires a steep learning curve because it typically utilizes low-level programming languages.
This can make the testing of new ideas and the translation of a given pulse sequence to systems produced by other manufacturers very complicated.The goal of these tools was to simplify these procedures by relying on high-level programming languages such as MATLAB (Pulseq 77,143 and TOPPE 93,144 ) or the freely available Python language (PyPulseq 78,145 ).
Pulseq proposed a new compact file format (".seq") for describing pulse sequences, based on a hierarchical representation consisting of a timing table that defines when different sequence blocks occur (Figure 2).The seq files can be generated by running high-level scripts or programmed using the JEMRIS GUI and can currently be interpreted and executed on Siemens, GE (via TOPPE), and Bruker commercial scanners.
The development of interpreters specific to other manufacturers' platforms or open-source scanners can be expected in the future.

| Virtual MRI scanner
Virtual MRI scanner, 95,146 a result from the ISMRM 2019 Junior Fellow Challenge, 147 is a web-based platform that enables running virtual MRI experiments, simulating the interaction with a console, patient registration, and selection of the imaging protocol and sequence parameters.Once the experiment is set up, a ".seq" file is written with PyPulseq and Bloch simulations are carried out using a discrete event approach.An important feature of this software is to outsource computationally heavy tasks to a different computer other than that running the GUI.

| Spinach
Spinach is a spin dynamics library that carries out calculations using the quantum description of magnetic resonance phenomena. 90,148Computations of complex systems are made possible in reasonable times by keeping only the main contributions from all possible spin-states in the simulated system.This software has the particularity of being able to support simulations of all forms of MR spectroscopy, including not only nuclear F I G U R E 1 MRI simulation using JEMRIS. 63Left: a pulse sequence is built by adding the RF and gradient objects to the tree and defining all the parameters required for the calculations (e.g., field of view, echo time).Middle: the sequence diagram and the k-space trajectory can be visualized for debugging.Right: having chosen the sample properties and selected the sequence file, the simulation can be run, and the obtained signal (bottom right of the middle column) processed to reconstruct an image.Here, an echo planar imaging sequence is simulated in a brain phantom, including susceptibility effects.The produced image appears distorted as expected.
magnetic resonance (NMR), but also electron paramagnetic resonance (ESR) and dynamic nuclear polarization (DNP), among others.This is possible by splitting the code so that all core operations can be performed using a general algebraic abstraction applicable to any finite-state quantum system, whereas a separate part of the code is designed to deal with the specific assumptions required to model each spectroscopy modality.

| VeSPA
The Versatile Simulation, Pulses and Analysis (VeSPA) software enables users to create RF pulses, simulate their behavior and compare their performance for use in MR applications, including MR spectroscopy. 94,149The software contains a database of pre-existing RF pulse designs that can be refined by the user, and the designed RF pulses can be exported using multiple file formats including ASCII and vendor-specific formats (Siemens).It is also possible to simulate and analyze spectroscopic data using VeSPA.

| SigPy
SigPy is a Python tool that includes toolboxes for MRI reconstruction and RF pulse design. 87,150The latter (SigPy.RF) enables designing, for example, adiabatic, multiband, and B 1 + -selective RF pulses, including k-space trajectory design and parallel transmit capabilities.It also performs Bloch simulations to predict the performance of the designed RF pulses.

| Image reconstruction
Image reconstruction commonly describes the process of solving the inverse problem of obtaining an image from the acquired MR raw data.
Often this problem is ill-posed and prior knowledge needs to be included in the form of additional regularization terms, which can require iterative methods.In addition, image reconstruction also involves other tasks such as prewhitening of the MR raw data, k-space trajectory estimation, calculation of coil sensitivity maps, and more.Here, we focus on image reconstruction software that covers the entire pipeline from the raw data to the final image.

| Gadgetron
In parallel to the development of ISMRMRD, the image reconstruction pipeline Gadgetron was created. 56,152,153The main aims of Gadgetron are to be flexible and modular by utilizing a chain of individual gadgets.These modules carry out tasks such as noise prewhitening, sorting of the data, iterative image reconstruction, and estimation of quantitative parameters.The input and output of each gadget are standardized so that gadgets can be easily combined to generate tailored image reconstruction pipelines.MR raw data (e.g., ISMRMRD) can be sent to Gadgetron via a Transmission Control Protocol/Internet Protocol connection.This allows for the streaming of data directly from the MR scanner to a custom image reconstruction pipeline.The reconstruction can be carried out on a separate reconstruction server with multiple cores of modern CPUs and GPUs that ensure short reconstruction times.This enables, for example, the reconstruction of highly undersampled non-Cartesian MR raw data with iterative image reconstruction approaches during image-guided interventions.Gadgetron is implemented in C++, but high-level interfaces using, for example, Python, are also available, which merge computational speed with rapid prototyping.

| BART
BART focuses on image reconstruction providing a range of solvers and regularization functionals (e.g., total variation, sparsity in the wavelet domain, or low-rank approaches). 48,154It is mainly implemented in C++ and image reconstruction can be carried out via the command line.Highlevel interfaces using, for example, Python, are also available.BART uses its own format for raw and image data but some support for ISMRMRD is already available.

| SIRF
SIRF is a framework that allows for the reconstruction of multimodality medical imaging data such as PET-MR. 89,155Conveniently, the framework provides a unified interface to already existing image reconstruction packages (e.g., Gadgetron for MR and Software for Tomographic Image Reconstruction [STIR] for PET 156 ).In addition to image reconstruction, SIRF also provides image registration capabilities utilizing NiftyReg and advanced iterative optimization techniques using Core Imaging Library (CIL). 157,158These capabilities also allow for motion-corrected image reconstruction. 159

| MRIReco.jl
A more recent software package is MRIReco.jl,which enables MR image reconstruction using Julia. 73,160It provides algorithms for iterative regularized image reconstruction, is ISMRMRD compatible, and also contains an MR simulation module.
Figure 3 shows the reconstruction of the same MR raw data file with Gadgetron, BART, SIRF, and MRIReco.jl.This should not be seen as a comparison of different reconstruction packages but an illustration that all these reconstruction packages can process real MR raw data (a docker file providing all the necessary reconstruction packages and a script to reproduce this figure can be found at https://github.com/ckolbPTB/OpenSourceMrRecon).It can also be used as a starting point to try out and compare different reconstruction packages, avoiding a lengthy installation process.
Apart from Gadgetron, all the other image reconstruction frameworks are aimed at offline image reconstruction, that is, the MR raw data need to be exported from the scanner first.This can be challenging to apply in clinical practice.Yarra is a project that provides a general framework to integrate custom image reconstruction methods in the standard clinical workflow for easy translation of novel developments. 96

| Image postprocessing
Image postprocessing is a method to extract information from an MR image.For general image analysis and postprocessing with a wide range of options, ImageJ is one of the most widely used open-source programs. 60ImageJ can display, edit, save DICOM images, and perform deconvolution, registration, segmentation, and many more features that can be embedded via Plugins.More specialized MR analysis techniques such as more complex image registrations, automatic segmentations or quantitative analysis require more specialized software, details of which are presented below.

| Image registration
Image registration aligns two images so that common features overlap and can be used to compensate for misalignments due to subject motion and geometric distortions, align images from different modalities, timepoints (follow-up scans), or subjects (population studies).Among others, registration makes it possible to combine structural images with functional information and improves tumor delineation for surgical planning.

ITK
ITK emerged in 2002 and is one of the most used open-source tools for image registration in medical imaging. 62,1613][164][165][166] Several Reconstruction results of a turbo spin echo black-blood acquisition on a 3-T Siemens scanner using Gadgetron, BART, MRIReco.jl, and SIRF in three different orientations of the heart.The dataset was undersampled by a factor of $2.The direct reconstruction corresponds to a simple FFT followed by coil-combination and was carried out with Gadgetron.For parallel imaging, the information from the different receiver coils was utilized to remove undersampling artifacts.The type of algorithm used by each method varies but the final results are comparable.See https://github.com/ckolbPTB/OpenSourceMrecoReconfor more details.
scientific groups expanded ITK's core functions.Prominent examples are Elastix and its Python, Java, R, Octave, Lua implementation Sim-pleElastix 88,167 ; ANTs 46,168 and its Python implementation ANTsPy. 470][171][172] It enables visualizing neural activation maps within higher resolution anatomical scans.Slices in each cardinal plane can be viewed simultaneously.Manual placement of markers on anatomical landmarks allows transformation of anatomical and functional scans into stereotaxic (Talairach-Tournoux) coordinates.

FreeSurfer
FreeSurfer is a set of software tools for the study of cortical and subcortical anatomy. 55,173In the cortical surface stream, the tools construct models of the boundaries between white and cortical gray matter, and the pial surface.Once these surfaces are known, several anatomical measures become possible, including cortical thickness, surface area, curvature, and surface normal at each cortical location.Surfaces from individuals can be aligned via an atlas with a high-dimensional nonlinear registration algorithm.

MIRTK
MIRTK is a research-focused image processing toolkit. 70It provides a collection of libraries and command-line tools to assist in processing and analyzing imaging data.The main application of MIRTK is in adult and neonatal brain MR image registration, as well as the reconstruction of cortical surface meshes.

NiftyReg
NiftyReg is a suite of image registration and alignment tools primarily intended for use with MRI and other brain scan images. 74NiftyReg is able to perform rigid, affine, and nonlinear registration of 2D and 3D images stored as Nifti or Analyze (nii or hdr/img).

3D Slicer
3D Slicer supports versatile visualizations and provides advanced functionality such as automated segmentation and registration for a variety of application domains. 44,175Unlike a typical radiology workstation, 3D Slicer is not tied to specific hardware.An example of the 3D Slicer workflow using different open-source postprocessing tools is illustrated in Figure 4.

Dafne
Dafne is a program for MR image segmentation that includes advanced deep learning models (based on Keras 64 ) for automatic segmentation. 52e user can refine the automated results, and the software will learn from the improvements and modify its internal models accordingly.

MONAI
MONAI is based on PyTorch and Numpy libraries especially created for deep learning in healthcare. 71,80,178It aims to capture the best AI development practices for healthcare researchers, focusing on medical imaging.MONAI has a repository (Model Zoo) that hosts medical imaging segmentation models developed by the scientific community.

| Quantitative parameter estimation
Commonly, MRI is used as a qualitative imaging modality, that is, the pixel values of reconstructed images are unitless and individually do not provide diagnostic information.Diagnosis is only possible due to the contrast between pathological and healthy tissue.In recent years this has started to change, and more and more quantitative MR techniques have been proposed, which directly provide quantitative (bio-)physical parameters (e.g., T1 relaxation time or blood flow).These parameters can be estimated by fitting a model to a sequence of qualitative images with different contrasts.Examples of the numerous quantitative tools are electric properties tomography, 53,179 blood perfusion quantification, 67,180 and susceptibility mapping. 85,181Here, we focus on software packages for quantitative analysis that are actively maintained and documented.
qMRLab qMRLab provides a framework implemented in MATLAB/Octave to estimate quantitative parameters such as T1 and T2 relaxation times, B 1 and B 0 field maps, and magnetic transfer ratios for a wide range of MR sequences. 82,182Parameter estimation often requires additional information about data acquisition and image reconstruction, which is frequently not part of the data exported from the MR scanner.qMRLab also advocates the development of qMRI-BIDS, an open-source data format for qMRI. 183This format is derived from the Brain Image Data Structure (BIDS), 184 aiming to provide all the necessary metadata for accurate parameter estimation.

hMRI-toolbox
Another software package for quantitative parameter estimation implemented in MATLAB is hMRI-toolbox. 59,185It relies on the Statistical Parametric Mapping (SPM) software package also available for MATLAB 91 and is targeted for qMRI of the brain.

QMRITools
QMRITools is a software package written in Mathematica. 81,186Similar to qMRLab, its functionality is very broad and ranges from T1 and T2 mapping to fat-water quantification.

PyQMRI
A Python implementation for quantitative parameter estimation is provided by PyQMRI. 79,187This software package is specifically aimed at both image reconstruction and parameter estimation.
F I G U R E 4 3D Slicer and workflow examples for MR image registration and segmentation.(A) 3D Slicer GUI. 44(B) MRI brain registration using the Elastix extension for 3D Slicer. 88Moving images are registered to fixed images.As an output the software provides the registered images and deformation fields.Image source from SlicerElastix. 176(C) Visualization of FreeSurfer 55 segmentation for the assessment of stereotactic surgery in epilepsy.Figure adapted from previous works. 177UIT QUIT is a C++-based software package that allows for quantitative susceptibility mapping, B 1 mapping, perfusion estimation, and many more. 83,188win A similar range of parameter estimates is provided by erwin, a Python-based software package for quantitative parameter estimation.54 erwin can be imported as a Python package or used directly from the command line.
In addition to these software packages mainly aimed at processing reconstructed images, quantitative parameter estimation can also be carried out with the software packages discussed in "Image Reconstruction" (section 2.2).Gadgetron and SIRF allow for parameter estimation after image reconstruction.BART provides reconstruction techniques that directly reconstruct (bio-)physical parameters from the MR raw data without an intermediate step of reconstructing qualitative images. 189

| OPEN-SOURCE MR SCANNERS
Designing the hardware required for a complete MR system involves a great deal of expertise in a wide range of engineering areas.The combination of multidisciplinary expertise, costly infrastructure, and materials costs have long restricted the handling of this technology to a handful of privileged manufacturers.Therefore, designing a complete open-source MR system was, for a long time, elusive.1][192][193][194] These low-field scanners are the most suitable candidates to exercise an open-source strategy, for several reasons: 1. Lower material and instrumentation costs, facilitating access to a larger community on limited budgets and in different places around the globe.
2. Construction feasible with easily accessible manufacturing methods.
4. Safer to produce and use: the lower magnetic field strengths and power usage largely reduce hazards presented to builders and practitioners.
5. Reduced risk on operating a scanner safely on human subjects (overall greatly reduced specific absorption rate [SAR] risks, simplified SAR modeling and monitoring).
6. Low-field MR companies would benefit from an open ecosystem helping with scanner development.
Within this strategy, the OSI 2 ONE low-field MR scanner, which consists largely of open-source software and hardware components, has been presented (Figure 5). 30,31,195The MR magnet optimization is performed by an open-source Python code that finds the optimal location of small cubic permanent magnets in a ringlike arrangement. 58,190,196A 47-mT magnet for imaging of the head and extremities was designed with an inner bore diameter of 30 cm using 12 Â 12 Â 12 mm 3 N48 NdFeB magnets. 122The low-field magnet can be constructed by hand using relatively affordable production machines such as CNCs, laser cutters, or 3D printers.The inevitable deviations through this low-cost manufacturing process led to a substantial difference between the simulated and realized B 0 field distribution of the magnet, which, however, can be corrected for by B 0shimming routines.A three-axis robotic stage COSI Measure 101,197 equipped with a magnetic field sensor can map the initial magnetic field distribution.This information is then applied to run various shimming algorithms to calculate the position and orientation of smaller cubic magnets to homogenize B 0 . 198,199After a few iterations, B 0 homogeneities of $450 ppm over a 20 cm diameter spherical volume can be achieved.
For the gradient coils, several open-source software tools exist to calculate the wiring paths based on predefined requirements such as linearity over a target volume or coil efficiency. 49,50,66,92,200These digital models can then be exported to CAD software to design a holder for the gradient coil wire.For low-field head-only systems, the gradient coils are much more efficient, experience substantially lower Lorentz forces, and can be driven by much lower currents than for whole body high-field systems.This enables relatively easy construction, using 3D printed holders to glue the wires into place.The gradient coils for the OSI 2 One have efficiencies around 0.5 mT/m/A, meaning that for typical imaging experiments no more than 10 A per gradient coil is needed.OSI 2 ONE currently uses a three-channel linear gradient amplifier able to deliver +/-15 A at approximately 12 V, with the voltage chosen so that the system can run off two car batteries for improved portability. 127,201Other designs of linear amplifiers exist with a maximum output current of around 8 A per channel that can be used to drive gradients or B 0 -shimming coils, 103,105,[202][203][204][205] and experimental designs with 20 A/100 V 110,206 are also available.
For RF transmission, solenoidal coils wired on 3D printed forms can be used.Solenoids are more sensitive for low fields than other RF coil designs such as birdcage or saddle coils and are very simple to construct.The design files for a 20.5 cm diameter solenoid for OSI 2 ONE are available here. 126The RF coil versus body noise dominance at low frequencies also implies a simpler receiver chain because coil arrays (> 2 receive channels) are less beneficial for increasing the signal-to-noise ratio.To amplify the RF pulses, several RF power amplifier designs with up to 1 kW of peak power with good linearity are available. 102,104,123,207This power output is more than enough to drive the highly efficient low-frequency RF coils that generate much higher B 1 + /√kW than their high-field counterparts, because component losses increase and tissue losses dominate at higher frequencies.
Regarding the console hardware, there are two open-source hardware-based NMR spectrometers with the capability to transmit RF and receive and digitize MR signals, the Opencore NMR 120,208 and LimeSDR. 109,209However, to the best of our knowledge, no open-source console designs exist yet with the capabilities to control gradient pulses with the fidelity needed for imaging.Nonetheless, relatively affordable proprietary electronic boards have been used for MR imaging by dedicated open-source software packages such as gr-MRI, 57,210 OCRA, 75,211 and MAR-COS. 68,212gr-MRI can be used with GNU Radio 213 compatible software-defined radios (SDRs), thus not limiting the software development to a single piece of hardware, being able to upgrade or change hardware designs without major modifications of the software.OCRA and MARCOS operate on the Red Pitaya hardware, which offers two independent transmit and receive channels and relies on external add-on boards to generate the gradient waveforms. 106,118A similar hardware-agnostic approach to gr-MRI is followed by the compatibility of MARCOS to Pulseq, 77,78 allowing independent and parallel pulse sequence and hardware developments, an important feature to successful open-source developments of a larger decentralized community.
In addition to these specific software solutions to drive the console hardware, more general software tools for scanner control, acquisitions, and postprocessing, such as Scanhub, 84,236 are being developed.Scanhub is a cloud-based generic and modular solution enabling integration of many of the aforementioned software packages (e.g., image reconstruction or postprocessing) to be used with actual MR scanner hardware such as the OSI 2 ONE system.These cloud-based platforms will gain further importance, beyond their core functionality to operate the scanner, by connecting users and experts from various backgrounds, exchanging imaging protocols and postprocessing routines while reducing the local computational requirements, or performing remote quality checks and maintenance of the scanners.An interesting aspect of the OSI 2 ONE project is the goal to also produce and provide documentation that follows the requirements for regulatory approval as a medical device according to EU regulation 2017/745.Such documentation, if publicly available, will greatly facilitate the translation from research prototypes to products, while simultaneously providing valuable documentation blueprints for other medical devices that are currently unavailable.
While the OSI 2 ONE project eventually targets in vivo imaging, other lower cost tabletop systems for NMR relaxometry 116,214 or fully capable MRI scanners 111,117,215 are successfully applied in the educational field (Figure 6).These tabletop scanners come with a full set of components to  For RF coils, an eight-channel transmit/receive design is available 99,216 to, for example, perform 7-T parallel transmission (pTx) experiments.
A similar sized eight-channel RF coil was also constructed for 3 T, allowing an easy measurement-based comparison of pTx techniques between these field strengths. 98,217Another scalable 3-T pTx RF coil design has been applied for 19 F imaging and could in principle also be tuned to 1 H frequency. 218,219 To characterize and investigate the electromagnetic fields of these or similar RF coils, the open-source electromagnetic simulation package MARIE is available. 69,220While MARIE might be an alternative to costly proprietary EM simulation software, other open-source software tools expand the functionality of proprietary software, such as CoSimPy, 51,221 which allows speedy circuit co-simulations (e.g., for fast RF coil optimizations).To homogenize B 1 for multichannel transmit RF coils, the shimming toolbox that implements static B 1 shimming routines can be used. 86,222e shimming toolbox is not restricted to B 1 shimming; it can also be applied for real-time B 0 shimming and B 0 dynamic shimming using AC/DC coils, effectively operated by open-source hardware current drivers. 103,105,202cause parallel transmission hardware is not readily available, pTx systems have been assembled (e.g., for implant safety testing). 125,223This Figures adapted from previous works. 223ource positioning system, 101,197 automated RF safety experiments can be performed outside or inside the MR scanner (Figure 7).This setup has been used to validate electromagnetic and temperature simulations, for RF safety measurements and sensor calibrations of implants.These configurations are extended by providing wireless reference implant hardware that can communicate with the pTx system over Bluetooth and exchange measured implant sensor information. 128,224Other MR-compatible low-cost sensors have been presented, for example, wireless sensors for motion tracking, 114,225 or sensors to measure force during muscular exercises inside the MR scanner. 121,226t all MR systems are equipped with highest end technology and specialized open-source hardware configurations can help to close this gap.A good example is a modular RF preamplifier/attenuator system, which was used to expand the initial four receive channels of a 4.7-T MRI to 32, potentially adjustable to other frequencies (Figure 8). 35,36A similar approach that extends the standard performance of an available MR hardware configuration, enabling novel applications is, for example, a transmit/receive switch design capable of handling much higher average RF powers for thermal MR RF heating experiments. 108,228Simple but useful open-source hardware tools for the laboratory or MR environment are, for example, RF field probes, which can be used to measure the frequency response of an RF coil, electromagnetic interferences, or RF shield attenuations. 107,113r phantoms, several designs are available.A very simple phantom, which can be assembled by mostly 3D printed parts, is the open-source NIST phantom. 119,229Phantoms for very specific applications are also available to assess image distortions, for example, in image-guided interventions such as MR-guided radiation therapy, 115,230 3D printed phantoms to evaluate contrast agent solutions, 112,231 for perfusion imaging, 100 or pneumatic phantoms to mimic respiration-induced artifacts. 124,2324][235] While sucrose gives strong chemical shift artifacts, PVP is often preferred for imaging phantoms.

| Challenges and suggestions for open-source MRI
This section presents a summary of the observations we have made over the years regarding current open-source practices and areas that we believe require further improvement to enhance the generation of useful results.This knowledge helps to manage expectations of newcomers to the field who seek rapid results, and leave under the impression of slow, chaotic, and inefficient processes.Open source is not a sprint, it is a marathon, and so is the development of the underlying infrastructure.

| Documentation and licensing
Clear open-source licensing and complete, comprehensive documentation is crucial for the successful reuse of open-source resources.Numerous projects lack clear licensing terms, thus the use of these resources might lead to copyright infringement.In addition, the quality of documentation across open-source projects varies greatly.Particularly for open-source hardware, the stakes and the risks are much higher because monetary investment and time to build and test a device need to be invested.It is illusive that documentation standards will be unified across all technological disciplines.Therefore, technology-specific documentation criteria (TsDC, DIN SPEC 3105-1 19 ) are defined and evaluated in standardization efforts to improve open-source hardware documentation.As experience shows, well documented projects can be reproduced without any or with minimal input from the original creator.In addition, no infrastructural support exists for developers after the release of a software or hardware.Even though extra efforts must be taken to create an open-source tool, the investment in the development and documentation of the tool is still reimbursed to some extent for the developer (e.g., through journal publications and increased visibility).However, after the release, there are no incentives and no support for ongoing development, bug fixes, maintenance, and user requests.This is particularly frustrating if the tool developed is successful and needed, because the developer is flooded with user requests.Institutional resources that exist, for example, for technology transfer and the costs associated with expensive patent applications or legal offices dealing with nondisclosure agreements, could be extended or partly transformed to strengthen open-source outputs.

| Open-source products
Currently, no, or very few products exist around the open-source projects presented in this review.To accelerate the distribution of technology and results produced with these, more participation is needed by the private sector.Not everyone wants and not everyone is able to reproduce open-source technology and there is strong interest in open-source products.
The challenge is that the principles of open source are difficult to conciliate with standard business models because protection of IP is a common practice to attract investment and to secure sufficient funds for product development.This is especially important for the commercialization of medical devices, which usually take several years to reach the market because of the lengthy certification processes.For such long development times, IP brings some security to investors, at the expense of slower development time, limited adoption, and dramatically reduced collaboration.In open-source-based business models the expertise on product development, support, and customization could be outsourced to the open-source community.Companies are curious but still hesitant to make the effort or take the risk to adapt or even connect with these new ways of value creation.To break the ice and accelerate the process, cooperation with innovation-driven businesses, which traditionally rely on IP restrictions, may need dedicated funding schemes.

| Medical devices
Open-source hardware cannot be used as a medical device on humans unless regulatory approval has been given.Regulatory approval and its documentation are lengthy processes that require substantial resources for companies.Currently Amazon, Google, SAP, Intel, OpenAI, and many others, actively utilize and contribute to open-source software.Likewise, open-source hardware designs are increasingly being commercialized by companies.Prominent examples include Arduino, Sparkfun, Adafruit, and Prusa.These trends are also reflected in the MRI community.The International Society for Magnetic Resonance in Medicine (ISMRM) has increasingly been incorporating open-source content into its annual meetings.Multiple ISMRM study groups actively promote open-source collaboration and various initiatives have emerged to establish open-source standards within MRI.MR-Hub (https://ismrm.github.io/mrhub/)serves as a plat-

T A B L E 1
List of open-source software resources, corresponding license, available files, published platform, and evaluated reference.A full regularly updated list can be found at https://gitlab.com/osii/cab.client v0.62, server v0.9796  T A B L E 2 List of open-source hardware resources, corresponding license, available files, published platform, and evaluated reference.Native file formats may be using proprietary software.

2 |
OPEN-SOURCE SOFTWARE PROJECTS FOR MRIMaking software available as open source has become increasingly popular in MRI.Partly because of community efforts to make research methods used in publications transparent and reproducible, and partly because of online platforms, such as GitHub,134 Bitbucket,135 and GitLab,136 which facilitate the sharing and collaborative development of software.The increased use of open-source high-level programming languages such as Python137 has also contributed to the ease of code sharing.The quality and reusability of open-source software can vary greatly, ranging from simple scripts with limited documentation to large software projects with comprehensive documentation and community-based support.In recent years, the application of machine learning techniques in MRI has led to a proliferation of small open-source software projects.The majority of these consist of the code necessary to reproduce the results of a scientific publication and they provide a very valuable starting point for further developments.

F I G U R E 2
Pulseq 77 and Pypulseq78 workflow, illustrating the design and testing of a diffusion-weighted echo planar imaging (EPI) sequence.Left: the pulse sequence is designed using high-level Matlab (Pulseq) or Python (Pypulseq) code as in this example, and its diagram is displayed below.Middle: a seq file is written, containing all the instructions concerning the RF, gradient, and sampling objects, and is imported into the MR scanner for running the sequence.Right: a set of images with different levels of diffusion-weighting (b-values) is acquired and used to estimate the apparent diffusion coefficient (ADC) map through monoexponential fitting of the signal decay in each voxel.2.2.1 | ISMRMRDMost MR vendors have their own MR raw data format that can only be interpreted using proprietary tools, which makes it cumbersome to create vendor-agnostic image reconstruction and postprocessing software.To facilitate software compatibility regardless of the scanner, at an ISMRM workshop in 2013, a group of MR scientists defined an open-source MR raw data format named ISMRM raw data (ISMRMRD).61,151The ISMRMRD project also provides a standard for reconstructed image data and open-source converters to translate proprietary raw data formats to ISMRMRD.This community-driven initiative was a huge step towards reproducible science and enabled the harmonization of image reconstruction software.
Image segmentation techniques extract image features by subdividing the image into different regions.Manual MRI segmentation is still a common practice; however, it suffers from poor reproducibility, user variability, and is a very time-consuming task.Automatic segmentation tools have emerged in recent years using different segmentation techniques (e.g., threshold, clustering, edge, region or atlas-based image segmentation and, more recently, deep learning-based segmentation).174

3. 1 |
Open-source resources for MR safety, quality assurance, and novel applications Within proprietary MR systems, customized hardware components are often needed to evaluate the feasibility of a novel application, investigate, and extend currently applied safety procedures, and advance science through novel experimental procedures.On the RF transmission side, F I G U R E 6 Two different educational tabletop systems.(A) Schematic, and (B) Photograph of the hardware setup of the instructional tabletop system from MGH/MIT. 111 (C) Photograph of the hardware, and (D) GUI and MR images from the OCRA Tabletop MRI System from Magdeburg University. 117particularly at ultrahigh-field (UHF) or for X-nuclei MRI, dedicated RF coils are often missing, leading to many customized research designs.Similarly, many research sites have already engaged in building their own phantoms.Recently, more open-source resources are becoming available, allowing easier reproduction, evaluation, and improvement by other groups.
allows a relatively flexible configuration of transmission and reception channels (up to 32 channels) over a wide frequency range, covering practically all relevant transmission frequencies from low-field to UHF MR at 7 T. Together with broadband RF power amplifiers and an open-F I G U R E 7 Open-source hardware for RF safety testing in and outside the MR scanner.(A and B) Parallel transmission medical implant safety testbed 125 used to measure E-fields inside the MRI system.(C and D) Parallel transmission (pTx) testbed setup outside the scanner to validate EM simulations by measuring B 1+ with time-domain H-field probes at various locations using COSI Measure101 in an open-source 7-T RF coil.99 Community is at the heart of open-source and its main workforce.Open-source projects are often inclusive, interdisciplinary, and intercultural.Communities are typically decentralized, and nonhierarchical, and community members often work voluntarily without monetary compensation and without strict deadlines.Because IP is organized as a common good, open-source technologies live independently from their originators or primarily distribution entities (e.g., companies or research institutes).Consequently, conventional communication and management structures with mostly clear responsibilities, goals, and decision-making hierarchies, are often missing.There is yet no standard solution for community governance.For large complex pioneer projects, such as open-source MRI technology, this needs to be piloted and iterated as the project grows.Today, F I G U R E 8 Open-source hardware for customized applications.(A) Modular 32-channel preamplifier/attenuator system.36 (B) Low-cost eightchannel digitally programmable current driver with 8 A per channel to drive B 0 -shimming coils.103(C) Simple RF probe for the laboratory in its manufactured form107 and its PCB layout designed with the open-source software KiCAD.227multiple software platforms support decentralized project development and community management (e.g., GitLab and OpenProject), but they are not yet fully streamlined for efficient use in a corporate scenario and also primarily cater to software development.

4 |
, no blueprints are available to facilitate the roadmap towards regulatory approval.It has to be evaluated to what extent and in which documentation format open-source hardware and software documentation can be aligned along the regulatory pathway.This would greatly ease the transition from research prototypes to actual products and would require less investment by the private sector.It is already difficult to establish open-source hardware documentation, let alone introduce documentation standards that go beyond current levels.This, together with the missing infrastructural support from institutions, seems to be a challenging task.However, based on observations from open-source software development over the last decades, professional documentation practices are being increasingly diffused into state-of-the art research.This trend was largely supported by dedicated software platforms (e.g., Github or Gitlab) allowing widespread collaboration, while enforcing certain documentation standards through built-in Wikis, project pages, git versioning, project management tools, etc. Software tools and blueprints to support documentation structures for regulatory approval are needed and would close the current gap between research prototypes and medical technologies used on patients.DISCUSSION AND CONCLUSION Open-source practices and resources in MRI have substantially increased in recent years, indicating a clear trend towards a culture of sharing and establishing nonexclusive global collaborations to advance the field.These highly favorable advancements continue to enhance the reproducibility and reusability of code and designs, while providing a more inclusive approach to low-income settings.Despite these important steps, the documentation quality of current open-source projects is still strongly variable.Developers and scientists are mostly left alone in their efforts to make open-source releases, with little guidance on the publishing process, the documentation requirements, postrelease maintenance, and community management.Community-driven standards and initiatives are forming to address these infrastructural needs, but more dedicated support is needed (e.g., at the institutional level), by funding agencies, and by widely visible academic publishers.Here, in particular, open-source research hardware has the potential to substantially lower the cost and access to MR technology and, in close collaboration with the industry, enable an easier translation from research prototypes to actual products benefiting patients.Open source can be more than an accelerator for science, it can be the foundation for sustainable global collaboration in the development of medical devices, making MRI more accessible and transparent.
learn, understand, and perform MRI experiments of small volumes: magnet, console, RF and gradient coils, amplifiers, transmit/receive switches, GUI, pulse sequences, etc.Such educational systems offer stimulating practical learning experiences for a wide audience.Physics students can design pulse sequences to gain insight into the physical and technical principles that generate an MR image.Medical students can obtain handson experience of different contrast mechanisms of MRI and their interpretation.Chemistry students can exercise chemical analysis through relaxometry and spectroscopy techniques.Electrical engineering students can learn small signal, power, and digital electronics.Mathematics and Information Technology students can explore the information encoding/decoding power of linear algebra, information compression and machine learning with data generated by themselves.
3.2.3| Funding and infrastructure Currently, there are few funding opportunities specifically available for open-source software projects, and even fewer (if any) for open-source hardware projects.Typically, researchers must find alternative paths by incorporating open-source technologies (e.g., as part of novel research proposals).Overall, more scientific results could be produced if essential proprietary technologies are available open source.A prerequisite for that is dedicated funding schemes.Other mechanisms to free designs would be to extend current open science guidelines and the requirements of public funding to open-source software and hardware.