arXiv subjects

Klaus Scheffler

Publications and source records attributed to Klaus Scheffler.

18 recordsLinked to original sources

arXiv2026

Overcoming the limitations of NMR Field Probes: A Novel Integrated Sensor Utilizing Pre-Polarization for (Ultra) Low Field MRI

Access to magnetic resonance imaging (MRI) remains severely limited in low- and middle-income countries, especially in sub-Saharan Africa, despite rising rates of non-communicable diseases. Low-field MRI presents an affordable, locally developable diagnostic solution, but its performance is constrained by magnetic field instability. We present a novel NMR field probe designed to overcome these challenges using a rapid non-adiabatic switch-off of a pre-polarization field resulting in precessing spin magnetization. Achieved by first use of high-voltage silicon carbide transistors operating in controlled avalanche breakdown, it measures the Larmor frequency without prior field knowledge, unlike conventional probes. This capability is crucial during magnet development with often unknown fields, allowing early detection of magnet issues, and offering an urgently needed tool for magnet design and image-quality improvement. Validated from 1 mT to 45 mT (up to 1,000 times stronger than similar systems) its low-cost, modular design supports replication, upgrades, and enhanced field control, helping expand global MRI access.

physics.med-ph↗

arXiv2026

Group-kernel auto-calibration and group-patch k-space reconstruction: Fast MRI with time-variant B0 kernels partitioned into time-invariant subsets

Purpose: Pushing MRI speed further demands more spatially-encoded information captured per unit time, e.g., by superimposing additional field modulations during oversampled readout. However, this can introduce calibration errors and increase reconstruction time. Thus, we propose a continuous field calibration approach and an efficient k-space reconstruction technique. Theory and Methods: Our auto-calibration generalizes GRAPPA kernels to explicitly extract continuous B0 modulation kernels, solving interpolation relationships between two ACS regions differing only in the extra field modulation. The k-space locations sharing the same instantaneous image-space modulation are grouped, so that subsets of time-invariant kernels can be separately estimated, as a generalized solution for Wave-CAIPI/FRONSAC-type scans. This view further inspires a k-space subregion-wise reconstruction technique, as an efficient alternative to conventional hybrid-space reconstruction. At 9.4T, FLASH accelerated by a local B0 coil array and Wave-CAIPI were tested with retrospective undersampling. Results: Artifact-free images were reconstructed, under diverse rapid B0 modulation schemes, reaching maximum acceleration factors of 8-fold in 2D and 14.6-fold in 3D. Some nonlinear gradients modulation schemes reach similar sampling efficiency as linear gradients modulation. The proposed reconstruction shows potentials in reducing reconstruction time. Conclusion: Rapid B0 modulations and widely-adopted parallel imaging can share a common mathematical framework, and consequently, achieve similarly-robust reconstructions. Specifically, for scans using dynamic B0 and static RF kernels, not only signal encoding, but also auto-calibration and reconstruction can be performed in k-space. This paves the way to robustly remove eddy currents, and explore more complex B0 modulation strategies towards ultimate MRI speed.

physics.med-ph↗

arXiv2026

Group-Patch Joint Compression: Compressing Dynamic B0 and Static RF Spatial Modulations Across k-Space Subregion Groups for Highly Accelerated MRI

Purpose: To accelerate MRI further, rapid B0 field modulations can be applied during oversampled readout to capture additional physical information, as in Wave-CAIPI/FRONSAC/local B0 coils modulation techniques. These methods, however, turn the Fourier readout into a non-Fourier-encoded dimension that cannot be reconstructed by FFT, posing significant reconstruction challenges especially in compressed-sensing or neural-network frameworks. Theory and methods: Because the rapid B0 modulations still vary slowly relative to the oversampled ADC dwell time, we exploit this encoding redundancy by compressing k-space patch-by-patch across subregions, each of which is jointly encoded by a distinct subset of B0 and RF (receive) spatial encoding functions. For each subset, a compression matrix is computed once and reused to compress all patches encoded by the same B0-RF spatial modulations. This can be implemented by feeding subsets of B0 and RF spatial encoding maps into an adapted conventional RF array compression algorithm, mimicking an expanded set of virtual receiver channels. This approach was evaluated on human brain scans at 9.4 T/3 T. Results: The proposed group-patch joint compression achieves substantially higher compression factors than conventional RF-only compression, while minimally compromising encoding efficiency. Typically, joint compression factors of 11x-20x led to negligible encoding loss, dramatically reducing reconstruction time and peak memory usage. For example, compressed-sensing reconstruction took 1.4-5.1 s/2D slice, 177 s-10.1 min/3D volume on a high-memory CPU node. Conclusion: Given joint encoding of dynamic B0 and static RF fields, compressing multidimensional k-space patches in separate groups outperforms compressing RF receivers alone. This substantially mitigates a fundamental computational bottleneck when combining rapid B0 and RF-receiver modulations.

physics.med-ph↗

arXiv2025

Coaxial Dipole Array with Switching Transmit Sensitivities for ultrahigh field MRI

Purpose: To investigate dipole antennas with electronically switchable transmit field patterns to improve flip angle homogeneity in ultra-high field MRI Methods: An array of eight coaxial dipoles with electronically switchable $B_{1}^{\!+}$ field profiles was constructed. Alteration of the field profiles was accomplished by modulating the currents along the dipoles using a combination of PIN diodes and lumped inductances. The behavior of these reconfigurable elements was studied in numerical electromagnetic simulations and 9.4T MRI measurements, investigating rapid switching of transmit sensitivities during excitation pulses in both single-channel and pTx mode operation. Results: For the simulated dipole elements, modulating the current densities along the dipole's axis causes a $\sim$30% change of the $B_{1}^{\!+}$ field between superior and inferior regions of the brain. When rapidly switched during excitation pulses, this degree of freedom can improve flip angle homogeneity, e.g. by a factor of $\sim$2.2 for a two kT points pTx pulse. For the constructed prototype array, the switching effect was observable but weaker, causing $\sim$10% superior-inferior $B_{1}^{\!+}$ variation. Conclusion: The proposed coaxial dipole array with switchable transmit sensitivities offers a novel degree of freedom for designing excitation pulses. The approach has the potential to improve flip angle homogeneity without necessitating an expensive increase in the number of independent transmit channels.

physics.med-ph↗

arXiv2025

High-resolution deuterium metabolic imaging of the human brain at 9.4 T using bSSFP spectral-spatial acquisitions

We demonstrated the feasibility of using bSSFP acquisitions for off-resonance insensitive high-resolution [6,6'-2H2]-glucose deuterium metabolic imaging (DMI) studies in the healthy human brain at 9.4T. Balanced SSFP acquisitions have potential to improve the sensitivity of DMI despite the SNR loss of phase-cycling and other human scanner constraints.We investigated two variants of bSSFP acquisitions, namely uniform-weighted multi echo and acquisition-weighted CSI to improve the SNR of deuterium metabolic imaging (DMI) in the brain with oral labelled-glucose intake. Phase-cycling was introduced to make bSSFP acquisitions less sensitive to B0 inhomogeneity. Two SNR optimal methods for obtaining metabolite amplitudes from the phase-cycled data were proposed. The SNR performance of the two bSSFP variants was compared with a standard gradient-spoiled CSI acquisition and subsequent IDEAL processing. In addition, in vivo T1 and T2 of water, glucose and Glx (glutamate+glutamine) were estimated from non-localized inversion recovery and spin-echo measurements.High-resolution whole-brain dynamic quantitative DMI maps were successfully obtained for all three acquisitions. Phase-cycling improved the quality of bSSFP metabolite estimation and provided additional spectral encoding. The SNR improvement was only observed for the CSI variant of bSSFP acquisitions with an average increase of 18% and 27% for glucose and Glx, respectively, compared to the vendor's CSI. ME-bSSFP acquisition achieved higher resolutions than acquisition-weighted CSI and exhibited several qualitative improvements.

physics.med-ph↗

arXiv2024

Indirect Zero Field NMR Spectroscopy

This study pioneers the two-field correlation spectroscopy (COSY) in zero to ultralow field (ZULF) liquid state NMR, employing hyperpolarized [1-$^{13}$C]pyruvate as a model system. We demonstrate the successful integration of signal amplification by reversible exchange (SABRE) for hyperpolarization, enabling the detection of ZULF COSY spectra with increased sensitivity. The use of field cycling allows the acquisition of two-field COSY spectra at varying magnetic field strengths, including zero-field conditions. This enables insight into both J-coupling and Zeeman-dominated regimes benefiting from ULF field observation sensitivity and mitigation of low-frequency noise by conducting readout at higher fields (>5 $\mu$T). Our study explores the effects of polarization transfer, apodization techniques, and the potential for further application of ZULF NMR in chemical analysis exemplified for three X-nuclei and three corresponding molecules: [1-$^{13}$C]pyruvate, [$^{15}$N]acetonitrile and [3-$^{19}$F]pyridine. These findings pave the way for more sensitive and cost-effective NMR spectroscopy in low-field regimes.

physics.med-ph↗

arXiv2024

UltraCortex: Submillimeter Ultra-High Field 9.4 T Brain MR Image Collection and Manual Cortical Segmentations

The UltraCortex repository (https://www.ultracortex.org) houses magnetic resonance imaging data of the human brain obtained at an ultra-high field strength of 9.4 T. It contains 86 structural MR images with spatial resolutions ranging from 0.6 to 0.8 mm. Additionally, the repository includes segmentations of 12 brains into gray and white matter compartments. These segmentations have been independently validated by two expert neuroradiologists, thus establishing them as a reliable gold standard. This resource provides researchers with access to high-quality brain imaging data and validated segmentations, facilitating neuroimaging studies and advancing our understanding of brain structure and function. Existing repositories do not accommodate field strengths beyond 7 T, nor do they offer validated segmentations, underscoring the significance of this new resource.

cs.CV↗

arXiv2023

DISGAN: Wavelet-informed Discriminator Guides GAN to MRI Super-resolution with Noise Cleaning

MRI super-resolution (SR) and denoising tasks are fundamental challenges in the field of deep learning, which have traditionally been treated as distinct tasks with separate paired training data. In this paper, we propose an innovative method that addresses both tasks simultaneously using a single deep learning model, eliminating the need for explicitly paired noisy and clean images during training. Our proposed model is primarily trained for SR, but also exhibits remarkable noise-cleaning capabilities in the super-resolved images. Instead of conventional approaches that introduce frequency-related operations into the generative process, our novel approach involves the use of a GAN model guided by a frequency-informed discriminator. To achieve this, we harness the power of the 3D Discrete Wavelet Transform (DWT) operation as a frequency constraint within the GAN framework for the SR task on magnetic resonance imaging (MRI) data. Specifically, our contributions include: 1) a 3D generator based on residual-in-residual connected blocks; 2) the integration of the 3D DWT with $1\times 1$ convolution into a DWT+conv unit within a 3D Unet for the discriminator; 3) the use of the trained model for high-quality image SR, accompanied by an intrinsic denoising process. We dub the model "Denoising Induced Super-resolution GAN (DISGAN)" due to its dual effects of SR image generation and simultaneous denoising. Departing from the traditional approach of training SR and denoising tasks as separate models, our proposed DISGAN is trained only on the SR task, but also achieves exceptional performance in denoising. The model is trained on 3D MRI data from dozens of subjects from the Human Connectome Project (HCP) and further evaluated on previously unseen MRI data from subjects with brain tumours and epilepsy to assess its denoising and SR performance.

eess.IV↗

arXiv2023

Pretraining is All You Need: A Multi-Atlas Enhanced Transformer Framework for Autism Spectrum Disorder Classification

Autism spectrum disorder (ASD) is a prevalent psychiatric condition characterized by atypical cognitive, emotional, and social patterns. Timely and accurate diagnosis is crucial for effective interventions and improved outcomes in individuals with ASD. In this study, we propose a novel Multi-Atlas Enhanced Transformer framework, METAFormer, ASD classification. Our framework utilizes resting-state functional magnetic resonance imaging data from the ABIDE I dataset, comprising 406 ASD and 476 typical control (TC) subjects. METAFormer employs a multi-atlas approach, where flattened connectivity matrices from the AAL, CC200, and DOS160 atlases serve as input to the transformer encoder. Notably, we demonstrate that self-supervised pretraining, involving the reconstruction of masked values from the input, significantly enhances classification performance without the need for additional or separate training data. Through stratified cross-validation, we evaluate the proposed framework and show that it surpasses state-of-the-art performance on the ABIDE I dataset, with an average accuracy of 83.7% and an AUC-score of 0.832. The code for our framework is available at https://github.com/Lugges991/METAFormer

cs.CV↗

arXiv2023

A Three-Player GAN for Super-Resolution in Magnetic Resonance Imaging

Learning based single image super resolution (SISR) task is well investigated in 2D images. However, SISR for 3D Magnetics Resonance Images (MRI) is more challenging compared to 2D, mainly due to the increased number of neural network parameters, the larger memory requirement and the limited amount of available training data. Current SISR methods for 3D volumetric images are based on Generative Adversarial Networks (GANs), especially Wasserstein GANs due to their training stability. Other common architectures in the 2D domain, e.g. transformer models, require large amounts of training data and are therefore not suitable for the limited 3D data. However, Wasserstein GANs can be problematic because they may not converge to a global optimum and thus produce blurry results. Here, we propose a new method for 3D SR based on the GAN framework. Specifically, we use instance noise to balance the GAN training. Furthermore, we use a relativistic GAN loss function and an updating feature extractor during the training process. We show that our method produces highly accurate results. We also show that we need very few training samples. In particular, we need less than 30 samples instead of thousands of training samples that are typically required in previous studies. Finally, we show improved out-of-sample results produced by our model.

eess.IV↗

arXiv2023

LIGHT-SABRE hyperpolarizes 1-^{13}C-pyruvate continuously, without magnetic field cycling

Nuclear spin hyperpolarization enables real-time observation of metabolism and intermolecular interactions in vivo. 1-13C-Pyruvate is the leading hyperpolarized tracer currently under evaluation in several clinical trials as a promising molecular imaging agent. Still, the quest for a simple, fast, and efficient hyperpolarization technique is ongoing. Here, we describe that continuous, weak irradiation in the audio-frequency range of the 13C spin at 121 {\mu}T magnetic field (\sim twiceEarth\apos s field) enables spin order transfer from parahydrogen to 13C magnetization of 1-13C-pyruvate. These so-called LIGHT-SABRE pulses couple nuclear spin states of parahydrogen and pyruvate via the J-coupling network of reversibly exchanging Ir-complexes. Using \sim 100% parahydrogen at ambient pressure, we polarized 51 mM of 1-13C-pyruvate in the presence of 5.1 mM Ir-complex continuously and repeatedly to a polarization of 1.1% averaged over free and catalyst-bound pyruvate. The experiments were conducted at -8{\deg}C), where almost exclusively bound pyruvate was observed, corresponding to an estimated 11% polarization on bound pyruvate. The obtained hyperpolarization levels closely match those obtained via SABRE-SHEATH under otherwise identical conditions. The creation of three different types of spin orders was observed: transverse 13C magnetization along the applied magnetic field, 13C z-magnetization along the main field B_0, and 13C-1H zz-spin-order. With a superconducting quantum interference device (SQUID) for detection, we found that the generated spin orders result from tiny 1H-13C J-coupling interactions, which are not visible even with our narrow linewidth below 0.3 Hz.

physics.chem-ph↗

arXiv2022

Prediction of motion induced magnetic fields for human brain MRI at 3T

Objective Maps of B0 field inhomogeneities are often used to improve MRI image quality, even in a retrospective fashion. These field inhomogeneities depend on the exact head position within the static field but acquiring field maps (FM) at every position is time consuming. Here we explore different ways to obtain B0 predictions at different head positions. Methods FM were predicted from iterative simulations with four field factors: 1) sample induced B0 field, 2) system's spherical harmonic shim field, 3) perturbing field originating outside the field of view, 4) sequence phase errors. The simulation was improved by including local susceptibility sources estimated from UTE scans and position-specific masks. The estimation performance of the simulated FMs and a transformed FM, obtained from the measured reference FM, were compared with the actual FM at different head positions. Results The transformed FM provided inconsistent results for large head movements (>5 degree rotation), while the simulation strategy had a superior prediction accuracy for all positions. The simulated FM was used to optimize B0 shims with up to 22.2% improvement with respect to the transformed FM approach. Conclusion The proposed simulation strategy is able to predict movement induced B0 field inhomogeneities yielding more precise estimates of the ground truth field homogeneity than the transformed FM.

physics.med-ph↗

arXiv2021

High-resolution neural network-driven mapping of multiple diffusion metrics leveraging asymmetries in the balanced SSFP frequency profile

We suggest to utilize the rich information content about microstructural tissue properties entangled in asymmetric balanced steady-state free precession (bSSFP) profiles to estimate multiple diffusion metrics simultaneously by neural network (NN) parameter quantification. A 12-point bSSFP phase-cycling scheme with high-resolution whole-brain coverage is employed at 3 T and 9.4 T for NN input. Low-resolution target diffusion data are derived based on diffusion-weighted spin-echo echo-planar-imaging (SE-EPI) scans, i.e., mean, axial, and radial diffusivity (MD, AD, RD), fractional anisotropy (FA) as well as the spherical coordinates (azimuth ${\Phi}$ and inclination ${\Theta}$) of the principal diffusion eigenvector. A feedforward NN is trained with incorporated probabilistic uncertainty estimation. The NN predictions yielded highly reliable results in white matter (WM) and gray matter (GM) structures for MD. The quantification of FA, AD, and RD was overall in good agreement with the reference but the dependence of these parameters on WM anisotropy was somewhat biased, e.g., in corpus callosum. The inclination ${\Theta}$ was well predicted for anisotropic WM structures while the azimuth ${\Phi}$ was overall poorly predicted. The findings were highly consistent across both field strengths. Application of the optimized NN to high-resolution input data provided whole-brain maps with rich structural details. In conclusion, the proposed NN-driven approach showed potential to provide distortion-free high-resolution whole-brain maps of multiple diffusion metrics at high to ultra-high field strengths in clinically relevant scan times.

physics.med-ph↗

arXiv2021

Coherent evolution of signal amplification by reversible exchange in two alternating fields (alt-SABRE)

Parahydrogen (pH2) is a convenient and cost efficient source for magnetic resonance signal enhancement. Transient interaction of pH2 with a metal organic complex in a signal amplification by reversible exchange (SABRE) experiment enabled more than 10% polarization for some 15N molecules. Here, we analyzed a variant of SABRE, consisting of an outer magnetic field alternating between a low field of ~1 \muT, where a polarization transfer takes place, and a higher field >50 \muT (alt-SABRE). We found effects of both of these fields on amplitude and the frequency of polarization transfer. Deviation of a lower magnetic field from a "perfect" condition of level anti-crossing increases the frequency of polarization transfer that can be exploited for polarization of short-lived transient SABRE complexes i.e. some substrates. Moreover, the coherences responsible for polarization transfer at a lower field persisted during magnetic field variation and continued their spin evolution at higher field with a frequency of 2.5 kHz at 54 \muT. The latter should be taken into consideration for an efficient alt-SABRE.

physics.chem-ph↗

arXiv2020

Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: Evaluation in Alzheimer's disease

Background: Although convolutional neural networks (CNN) achieve high diagnostic accuracy for detecting Alzheimer's disease (AD) dementia based on magnetic resonance imaging (MRI) scans, they are not yet applied in clinical routine. One important reason for this is a lack of model comprehensibility. Recently developed visualization methods for deriving CNN relevance maps may help to fill this gap. We investigated whether models with higher accuracy also rely more on discriminative brain regions predefined by prior knowledge. Methods: We trained a CNN for the detection of AD in N=663 T1-weighted MRI scans of patients with dementia and amnestic mild cognitive impairment (MCI) and verified the accuracy of the models via cross-validation and in three independent samples including N=1655 cases. We evaluated the association of relevance scores and hippocampus volume to validate the clinical utility of this approach. To improve model comprehensibility, we implemented an interactive visualization of 3D CNN relevance maps. Results: Across three independent datasets, group separation showed high accuracy for AD dementia vs. controls (AUC$\geq$0.92) and moderate accuracy for MCI vs. controls (AUC$\approx$0.75). Relevance maps indicated that hippocampal atrophy was considered as the most informative factor for AD detection, with additional contributions from atrophy in other cortical and subcortical regions. Relevance scores within the hippocampus were highly correlated with hippocampal volumes (Pearson's r$\approx$-0.86, p<0.001). Conclusion: The relevance maps highlighted atrophy in regions that we had hypothesized a priori. This strengthens the comprehensibility of the CNN models, which were trained in a purely data-driven manner based on the scans and diagnosis labels.

eess.IV↗

arXiv2020

MRzero -- Fully automated discovery of MRI sequences using supervised learning

Purpose: A supervised learning framework is proposed to automatically generate MR sequences and corresponding reconstruction based on the target contrast of interest. Combined with a flexible, task-driven cost function this allows for an efficient exploration of novel MR sequence strategies. Methods: The scanning and reconstruction process is simulated end-to-end in terms of RF events, gradient moment events in x and y, and delay times, acting on the input model spin system given in terms of proton density, T1 and T2, and $\Delta$B0. As a proof of concept, we use both conventional MR images and T1 maps as targets and optimize from scratch using the loss defined by data fidelity, SAR penalty, and scan time. Results: In a first attempt, \textit{MRzero} learns gradient and RF events from zero, and is able to generate a target image produced by a conventional gradient echo sequence. Using a neural network within the reconstruction module allows arbitrary targets to be learned successfully. Experiments could be translated to image acquisition at the real system (3T Siemens, PRISMA) and could be verified in the measurements of phantoms and a human brain \textit{in vivo}. Conclusions: Automated MR sequence generation is possible based on differentiable Bloch equation simulations and a supervised learning approach.

physics.med-ph↗

arXiv2018

DeepCEST: 9.4 T Chemical Exchange Saturation Transfer MRI contrast predicted from 3 T data - a proof of concept study

Purpose: Separation of different CEST signals in the Z-spectrum is a challenge especially at low field strengths where amide, amine, and NOE peaks coalesce with each other or with the water peak. The purpose of this work is to investigate if the information in 3T spectra can be extracted by a deep learning approach trained by 9.4T human brain target data. Methods: Highly-spectrally-resolved Z-spectra from the same volunteer were acquired by 3D-snapshot CEST MRI at 3 T and 9.4 T with similar saturation schemes. The volume-registered 3 T Z-spectra-stack was then used as input data for a 3-layer deep neural network with the volume-registered 9.4 T fitted parameter stack as target data. The neural network was optimized and applied to training data, to unseen data from a different volunteer, and as well to a tumor patient data set. Results: A useful neural net architecture could be found and verified in healthy volunteers. The principle gray-/white matter contrast of the different CEST effects was predicted with only small deviations. The 9.4 T prediction was less noisy compared to the directly measured CEST maps, however at the cost of slightly lower tissue contrast. Application to a tumor patient measured at 3 T and 9.4 T revealed that tumorous tissue Z-spectra and corresponding hyper/hypo-intensities of different CEST effects can also be predicted. Conclusion: Deep learning might be a powerful tool for CEST data processing and deepCEST could bring the benefits and insights of the few ultra-high field sites to a broader clinical use. Vice versa deepCEST might help for determining which subjects are good candidates to measure additionally at UHF.

physics.med-ph↗

arXiv2015

Task-related edge density (TED) - a new method for revealing large-scale network formation in fMRI data of the human brain

The formation of transient networks in response to external stimuli or as a reflection of internal cognitive processes is a hallmark of human brain function. However, its identification in fMRI data of the human brain is notoriously difficult. Here we propose a new method of fMRI data analysis that tackles this problem by considering large-scale, task-related synchronisation networks. Networks consist of nodes and edges connecting them, where nodes correspond to voxels in fMRI data, and the weight of an edge is determined via task-related changes in dynamic synchronisation between their respective times series. Based on these definitions, we developed a new data analysis algorithm that identifies edges in a brain network that differentially respond in unison to a task onset and that occur in dense packs with similar characteristics. Hence, we call this approach "Task-related Edge Density" (TED). TED proved to be a very strong marker for dynamic network formation that easily lends itself to statistical analysis using large scale statistical inference. A major advantage of TED compared to other methods is that it does not depend on any specific hemodynamic response model, and it also does not require a presegmentation of the data for dimensionality reduction as it can handle large networks consisting of tens of thousands of voxels. We applied TED to fMRI data of a fingertapping task provided by the Human Connectome Project. TED revealed network-based involvement of a large number of brain areas that evaded detection using traditional GLM-based analysis. We show that our proposed method provides an entirely new window into the immense complexity of human brain function.

q-bio.NC↗