Max Planck Research Group MR Physics

Head: Benedikt Poser (since 2026)
 

MPRG MR Physics Evolution
Building and Maintaining the MRI Lab at the MPI for Human Development
Scientific MR Study Support
MR Physics Research

MPRG MR Physics Evolution

Research Group Genesis and Strategic Transition (2023–2024)

The MPRG MR Physics commenced its independent operations in 2024, transitioning from a former support function within the Center for Lifespan Psychology (LIP). This deliberate institutional shift into an independent research and support unit represents a strategic commitment by the Institute to initiate the ambitious, multi-million Euro project involving the procurement and installation of multi-field strength magnets (mobile 1.5 T, stationary 3 T, and 7 T systems) and to ensure long-lasting scientific support across the different platforms.

The group's goal is defined by a dual mission: first, to deliver state-of-the-art multi-field magnetic resonance (MR) technology as a shared, intramural resource for human imaging studies; and second, to pioneer novel MR physics methodologies.

Leadership Transition and Focused Strategy

The first group leader, Siawoosh Mohammadi (2023–2024), established the group's foundational scientific identity. In addition to leading the group's efforts in the planning and procurement of the mobile 1.5 T, stationary 3 T, and 7 T systems, he set a strong focus on quantitative MRI (qMRI). This included developing novel methods for the fusion of high-resolution qMRI data to enable MRI-based in vivo histology (hMRI), specifically targeting key white matter properties such as myelin density, fiber density, and the g-ratio.

After Siawoosh Mohammadi accepted the appointment to become a Professor of Advanced MRI from Universität zu Lübeck in 2024, Sebastian Schröder (2024–2025) became the interim group leader. As the head of the Scientific Technology & Computing (STC) unit, he placed a strong focus on improving the MR Physics group's organizational structure. He continued and further strengthened the group's efforts in the planning and procurement of the mobile 1.5 T, stationary 3 T, and 7 T systems. This work included the acceptance measurements of the 7 T system at the MPI for Empirical Aesthetics (MPIEA), conducted in collaboration with researchers from the MPI for Human Cognitive and Brain Sciences (MPI-CBS), as well as a study to investigate potential differences between stationary and mobile 1.5 T systems.

The current leader, Benedikt Poser (2026), has strategically expanded the group's focus, aligning this scientific direction with the commissioning of the 3 T and 7 T Ultra-High Field (UHF) systems. The current scientific priorities leverage this new infrastructure to address major technical challenges inherent in advanced neuroimaging, specifically by working on the compensation of MRI system imperfections, which involves developing advanced methods to mitigate artifacts arising from hardware non-idealities such as gradient non-linearity and B₀/B₁ field inhomogeneity, thereby ensuring the quantitative reliability of UHF data, and also by focusing on advanced structural and functional pulse sequences through the innovative application of parallel imaging and sophisticated non-Cartesian sampling methods, a step crucial for translating UHF-MRI's improved sensitivity and resolution into feasible acquisition times.

Summary 

This structure showcases a logical and strategic evolution: Siwawoosh Mohammadi established the fundamental requirement for robust, quantitative MRI (in vivo qMRI), which is now strategically empowered by Benedikt Poser's focused development and optimization of UHF technology and physics methods. This organizational trajectory enables two key outcomes: first, providing support for high-quality neuroscience MRI studies performed by collaborative research units (such as the Center for Environmental Neuroscience [CEN], LIP, and the Emmy Noether Group Real-world Applications of Variability in Episodic memory and Neurodevelopment [RAVEN]) through the deployment of state-of-the-art methods; and second, advancing the Institute's overall neuroscience research program through boundary-pushing technical innovations conducted by the MPRG MR Physics. 

Building and Maintaining the MRI Lab at the MPI for Human Development

In close collaboration with the STC unit, the MPRG MR Physics spearheaded the simultaneous procurement and phased installation of a state-of-the-art multi-field neuroimaging facility. This critical, multi-year facility build-out—planned and requiring detailed coordination with vendors and contractors for site planning and design—is pivotal for solidifying the site's credibility as a world-class neuroimaging center. This strategic multi-field approach is implemented to provide application-tailored solutions for diverse neuroscience objectives: a mobile 1.5 T system for field research, a 3 T system as the primary workhorse for general neuroscientific research, and a 7 T system for ultra-high-resolution applications. To maintain long-term reliability for our research, the MPRG MR Physics established robust, continuous operational protocols. Weekly quality assurance (QA) and longitudinal performance monitoring, including geometric accuracy and image uniformity, ensure reproducible and reliable neuroscience data by mitigating hardware imperfections.

The final phase of system commissioning and operational go-live involves rigorous calibration and vendor acceptance, starting with the mobile 1.5 T in January 2026, followed by the 3 T toward the end of 2026, and culminating with the capstone achievement of the 7 T system some time later.

Scientific MR Study Support

Technical Consultation and Protocol Optimization

The MPRG MR Physics offers comprehensive scientific and technical support to neuroscience researchers at the Institute, spanning from initial concept development through to study completion. Consultation services include optimizing imaging parameters to meet specific neuroimaging goals and MR safety. The following table provides a brief overview of the neuroscientific studies supported across different centers and research groups at the Institute during the reporting period.

Advanced Data Management and Quality Control

The MPRG MR Physics further ensures data integrity and accessibility through rigorous management services. This includes comprehensive image handling via an online system, weekly quality assurance (QA), secure data backup, and anonymization. Furthermore, staff provide critical workflow integration, connecting the acquisition process to an internal server platform for data download, and supporting the automation of post-processing and analysis pipelines with complete traceability. Figure 1 shows the results of this QA and the temporal stability of the 3 T scanner used for the neuroscience studies at the Institute.

MR Physics Research

Beyond supporting MRI site planning, procurement, and external research, the MPRG MR Physics maintains an active MR Physics research portfolio. This research work is organized into six interconnected themes: (i) Assessment of MRI system imperfections, (ii) Compensation of MRI system imperfections, (iii) Mobile 1.5 T MRI, (iv) High-resolution diffusion MRI, (v) Quantitative T₁ mapping and segmentation using MP2RAGE, and (vi) MR microstructure and in vivo histology.

Assessment of MRI System Imperfections 
The group has actively investigated the crucial trade-offs between increasing field strength and gradient system performance, providing critical technical benchmarks for the acquisition strategy across 1.5 T, 3 T, and 7 T systems, together with the MPIEA and MPI-CBS (Aigner et al., in press). A comparative study of four different gradient systems, including a 3 T Connectom scanner with powerful gradients (300 mT/m) and two 7 T Terra systems, assessed the temporal stability and image quality for blood-oxygen-level-dependent (BOLD) fMRI and diffusion-weighted imaging (DWI). This work confirmed the expected qualitative improvement in anatomical detail at 7 T, where optimized protocols achieved a significant resolution gain (voxels 9 times smaller than at 1.5 T). However, the study also revealed a critical performance trade-off: 7 T systems face challenges due to increased B₀/B₁+ variations and compromised DWI stability, despite generating a 5.2-fold higher phantom temporal signal-to-noise ratio (tSNR) compared to 1.5 T. In contrast, the 3 T Connectom, leveraging its high-performance gradient system, achieved the best temporal stability and excellent tSNR, as depicted in Figure 2, notably yielding the smallest mean Apparent Diffusion Coefficient (ADC) variation, highlighting the essential need for future 7 T scanners to incorporate Connectom-class gradient technology to minimize system imperfections. The study established that the 1.5 T platform is primarily resolution-limited, emphasizing that the optimal choice of scanner depends critically on the study's scientific objective (resolution vs. stability).

Over the past years, the work of the MR Physics group has focused on the development, validation, and application of a novel approach for transmit field (B₁⁺) mapping, to enable reliable correction of structural brain images (MPRAGE and MP2RAGE) and accurate T₁ quantification. This activity builds on a dedicated, non-commercial pulse sequence implemented in the Siemens IDEA environment capable to achieve specific magnetization trajectories by means of a series of composite pulses. Such a sequence is not available on standard clinical scanners and has been further refined, optimized, and systematically validated during the reporting period. In parallel, the group has independently developed the full simulation framework and a comprehensive MATLAB-based analysis pipeline to model signal behavior, identify optimal acquisition strategies, and quantitatively evaluate performance in simulations, phantoms, and in vivo studies.

The resulting method achieves accurate B₁+ correction with reduced acquisition time and radiofrequency (RF) power requirements, making it suitable both for high-field research and for applications where protocol duration and the specific absorption rate (SAR) are critical, such as CEN's newly established neonatal studies at 3 T (cf. Figure 3). This work is also strategically aligned with the Institute's recent purchase of a 7 T MRI system, for which the method provides a robust foundation for ultra-high-field and future parallel-transmit applications (Santoro et al., in press).

Building on the experience gained with proton imaging, the MR Physics group has also developed a new and distinct B₁+ mapping method for ²³Na MRI at 7 T in view of future multinuclear studies exploring the potential of ²³Na MRI in the brain. This new sequence is specifically designed to exploit the fast repetition times and signal properties of sodium imaging and based on a new trajectory, as the ¹H MRI would not be applicable to ²³Na MRI. As shown in Figure 4, this sodium-specific approach enables accurate correction and absolute quantification of tissue sodium concentration and has led to an international collaboration with Guillaume Madelin at New York University (Santoro et al., in press).

Key References

Compensation of MRI System Imperfections

The push toward UHF and high-speed imaging magnifies inherent hardware limitations, making dedicated physics correction work essential for image fidelity. Without these compensation methods, highly detailed images would be geometrically inaccurate, severely limiting their utility for registration and longitudinal monitoring. Parallel Transmit (pTx) is critical for mitigating the inherent challenges of UHF systems, particularly the severe B₁+ inhomogeneity. The group's research portfolio in this area addresses coil and pulse optimization across complex, extended anatomical regions.

A simulation study together with the University of Duisburg-Essen and Physikalisch-Technische Bundesanstalt (Aigner, May, et al., 2025) addressed the need for combined brain and full cervical spine imaging at 7 T using only the standard 8-channel pTx system. This research utilized dynamic kT-point pTx pulses, demonstrating that tailored pulses substantially improved flip-angle (FA) homogeneity across this large region by roughly a factor of five compared to static RF shimming. The results confirmed that dynamic universal-kT-point pTx pulses achieved high FA homogeneity with a coefficient of variation (CV) below 10% as depicted in Figure 5, validating a rapid strategy for simultaneous acquisition of both regions.

The group extended this work to image the entire spinal cord (SC) at 7 T, a challenge due to spatial B₁+ variation and motion interference, in cooperation with the German Cancer Research Center (DKFZ) and Physikalisch-Technische Bundesanstalt (Aigner, Grimm et al., 2025). This was achieved by coupling a remote 32-channel pTx body array with a local 16-channel receiver coil to achieve a large field of excitation (FOX). Tailored static pTx (RF shimming) successfully optimized B₁+ homogeneity, reducing the CV approximately two-fold (e.g., from 41% to 21.4% in the cervicothoracic spine; cf. Figure 6). A key technical finding was that while B₁+ was robust to respiration, the B₀ field exhibited substantial oscillations (up to 300 Hz along the vertebrae). Despite these B₀ challenges, the optimized pTx techniques enabled the successful acquisition of high-quality 2D and 3D multi-echo gradient echo (GRE) datasets, confirming the potential of high-resolution whole SC imaging at 7 T for diagnosing spinal cord pathologies.

Key References

Mobile 1.5 T MRI

The group has extended its focus on system integrity validation to the emerging domain of mobile neuroimaging, mainly addressing CEN's requirements for opportunities to collect reliable MRI data beyond the lab. A recent in vivo study, in cooperation with Physikalisch-Technische Bundesanstalt and Universität zu Lübeck, directly compared the reproducibility and reliability of a mobile 1.5 T scanner against a stationary 1.5 T unit, specifically assessing the stability after relocation (Aigner, Forlim, et al., 2025). The investigation included evaluating B₀/B₁⁺ homogeneity, structural MP-RAGE quality, and functional EPI-BOLD tSNR and default mode network (DMN) consistency. This work demonstrated that the mobile system achieved robust field stability and image quality comparable to the stationary unit, showing that variability from relocation was similar to the inherent variability of the stationary system. This finding confirms the potential of mobile 1.5 T MRI as a transformative technology for enhancing accessibility and enabling large-scale, diverse neuroimaging studies. The study leveraged structural (MP-RAGE) and functional (EPI-BOLD) sequences across seven subjects in four sessions, relocating the mobile unit between sessions to rigorously test field stability. Quantitative analysis demonstrated high system integrity, with the B₁+ FA exhibiting minimal, non-significant variability (below 2%) across all comparisons. Crucially, the functional image quality showed exceptional performance: the whole-brain tSNR (wb-tSNR) had a low 4% between-scanner variation and a minimal 0.6% within-subject variation (see Figure 7). Furthermore, the complexity introduced by moving and repositioning the mobile scanner resulted in variability comparable only to the inherent, day-to-day variability of the stationary system. This robustness supports the integration of mobile, high-field MRI into large-scale environmental neuroscience studies requiring reliable structural segmentation and consistent resting-state functional metrics, such as the DMN, which was successfully and consistently extracted across all scanning sessions.

Key References

High-Resolution Diffusion MRI

This methodological work originated from a cooperation within a project funded by Deutsche Forschungsgemeinschaft (DFG, No. 412917403, PIs: Hauke Heekeren & Radoslaw Cichy) hosted at Freie Universität (FU) Berlin. The project investigates the hippocampal-striatal-prefrontal network supporting memory-guided, value-based decision making. Within this network, the hippocampus provides associative and episodic memory information to reward-related regions such as the ventral striatum and ventromedial prefrontal cortex. These interactions are partly mediated via hippocampal output pathways including the fornix, a small white matter tract connecting the hippocampal formation with subcortical and medial prefrontal structures. It is difficult to resolve with diffusion MRI because of its limited cross-sectional size and proximity to cerebrospinal fluid.

In this context, MR physicist Nils Bodammer, in close cooperation with biologist and project leader Claudia Wolf (FU Berlin), focused on DWI strategies that provide reliable orientation information for segmenting such fine tracts and, at the same time, enable quantitative assessment of tissue microstructure for use in cross-sectional and longitudinal studies. The choice of diffusion models and protocols was informed by their potential sensitivity to microstructural features such as myelination and neurite architecture, as suggested by recent work relating advanced diffusion metrics to myelin-sensitive measures. On the Prisma-fit system of the FU Berlin, the collaboration systematically explored advanced diffusion protocols, including high-resolution acquisitions down to 1.2 mm isotropic, using both the multiband (MB) EPI sequence developed by the Center for Magnetic Resonance Research (CMRR), University of Minnesota, and the readout-segmented EPI (RESOLVE) sequence. While RESOLVE proved technically suitable, its substantially longer scan times were not compatible with the planned applications. The final protocol choice is based on the CMRR multiband DWI sequence with optimized multi-shell sampling.

Within this framework, different gradient schemes, b-value combinations, and diffusion encodings were evaluated (see Figure 8). Monopolar diffusion gradients were selected for their shorter echo times and higher signal, forcing residual eddy-current effects to be handled in the preprocessing pipeline. An isotropic resolution of 1.5 mm was identified as a robust compromise between anatomical detail and signal-to-noise ratio. In practice, the total acceleration (MB factor × in-plane parallel imaging) is kept at or below 6 to avoid visible reconstruction artefacts. The optimized acquisitions are designed to support several diffusion models within a single measurement, in particular diffusion kurtosis imaging (DKI) and neurite orientation dispersion and density imaging (NODDI). In this work, NODDI is implemented using the Accelerated Microstructure Imaging via Convex Optimization (AMICO) framework for efficient microstructure imaging.

Mean kurtosis (MK) from DKI serves as a robust descriptive measure of deviations from Gaussian diffusion and is sensitive to microstructural complexity, while relying on fewer specific model assumptions than multi-compartment approaches. NODDI delivers model-based maps of neurite density (NDI), orientation dispersion (ODI), and free-water fraction (FWF).

Analyses of the high-quality data also reveal important limitations: NDI estimates in very small, highly coherent tracts such as the fornix are strongly influenced by assumptions on noise and Rician bias correction. The resulting NDI maps were found to depend substantially on the assumed SNR used for bias correction. Parameter settings that yielded anatomically plausible tissue contrasts and reduced artifactual NDI saturation in low-SNR regions required assumptions that were not fully consistent with independent SNR estimates from the data. These observations are in line with published discussions of parameter degeneracy and model non-uniqueness in multi-compartment diffusion models and suggest that absolute NDI values should be interpreted with caution. In contrast, MK and, within the NODDI framework, ODI were comparatively stable across a range of plausible noise and bias-correction settings and are therefore preferred as complementary microstructural contrasts, while NDI must be interpreted more cautiously. Further work will be required to better understand the origin of these fitting instabilities and their dependence on image SNR and model assumptions. The established methodology on the Prisma-fit system forms a technical basis for future studies at the Institute, including time-efficient protocols for children receiving training (e.g., joining a choir), where diffusion MRI—combined with other quantitative methods—will be used to monitor microstructural development and experience-related plasticity in vivo.

Quantitative T₁ Mapping and Segmentation Using MP2RAGE

Over the past years, the MR Physics group established and implemented a protocol involving three consecutive MP2RAGE acquisitions, each with distinct inversion times, in a large neuroimaging study including more than 100 children (FLEX study, PI: Yana Fandakova, formerly LIP). The inversion times for the three acquisitions were: A = [488, 1750] ms, B = [800, 2500] ms, and C = [1200, 3450] ms, while all other imaging parameters were identical. This dedicated approach was developed to replace conventional MPRAGE imaging for tissue segmentation, improving robustness and contrast in the developing brain. The MR Physics group initiated the analysis of combined T₁ mapping strategies based on these three MP2RAGE acquisitions, which provide complementary sensitivity to short and long T₁ values (cf. Figure 9). A signal-derivative-based weighting approach was introduced to integrate acquisition-specific information and enhance T₁ estimation across the full dynamic range, including cerebrospinal fluid. Ongoing analyses indicate systematic differences compared to standard single-protocol T₁ maps, highlighting the potential of this consecutive MP2RAGE approach for robust quantitative mapping in children.

MR Microstructure and in Vivo Histology

A core challenge in quantitative MRI is to translate macroscopic, in vivo measurements into biologically interpretable microstructural metrics that traditionally require ex vivo histology. Building on a track record in MR physics and biophysical modelling, Siawoosh Mohammadi's work is centered on developing and validating tissue microstructure models (diffusion and quantitative MRI) that enable "in vivo histology" by linking MRI observables to histological counterparts. This approach is underpinned by strong interdisciplinary collaborations spanning high-resolution qMRI and modelling (e.g., with Nikolaus Weiskopf at MPI-CBS in Leipzig), neuroanatomy and histological methods (e.g., CLARITY and ultra-high-resolution microscopy with Markus Morawski, Leipzig University), and clinical translation of qMRI biomarkers (e.g., his European Research Council project with Thomas Sauvigny at the Medical Center Hamburg-Eppendorf). A notable contribution is the sustained development of widely used open-source toolboxes for quantitative MRI and diffusion MRI artefact correction/modelling (including hMRI and ACID within the SPM ecosystem), supporting robust estimation of microstructure-sensitive metrics across brain and spinal cord applications, including clinical research contexts.

Recent publications illustrate how this methodological foundation is being converted into concrete microstructure–histology bridges. For example, Mordhorst et al. (2025) directly target the central validation problem by using MRI-scale histology to test whether axon-radius-sensitive MRI measures capture spatially meaningful microstructural variation. Complementing this, Papazoglou et al. (2024) focus on strengthening agreement between a key diffusion-derived compartment metric (axonal volume fraction) and an ultrastructural reference standard, addressing a major barrier to using biophysical parameters as credible in vivo histological surrogates.

Key References