Search

In this work, we propose FlowMRI-Net, a novel deep learning-based framework for fast reconstruction of accelerated 4D flow magnetic resonance imaging (MRI) using physics-driven unrolled optimization and a complexvalued convolutional recurrent neural network trained in a self-supervised manner. The generalizability of the framework is evaluated using aortic and cerebrovascular 4D flow MRI acquisitions acquired on systems from two different vendors for various undersampling factors (R=8,16,24) and compared to state-of-the-art compressed sensing (CS-LLR) and deep learning-based (FlowVN) reconstructions. Evaluation includes quantitative analysis of image magnitudes, velocity magnitudes, and peak velocity curves. FlowMRINet outperforms CS-LLR and FlowVN for aortic 4D flow MRI reconstruction, resulting in vectorial normalized root mean square errors of $0.239\pm0.055$, $0.308\pm0.066$, and $0.302\pm0.085$ and mean directional errors of $0.023\pm0.015$, $0.036\pm0.018$, and $0.039\pm0.025$ for velocities in the thoracic aorta for R=16, respectively. Furthermore, FlowMRI-Net outperforms CS-LLR for cerebrovascular 4D flow MRI reconstruction, where no FlowVN can be trained due to the lack of a highquality reference, resulting in a consistent increase in SNR of around 6 dB and more accurate peak velocity curves for R=8,16,24. Reconstruction times ranged from 1 to 7 minutes on commodity CPU/GPU hardware. FlowMRI-Net enables fast and accurate quantification of aortic and cerebrovascular flow dynamics, with possible applications to other vascular territories. This will improve clinical adaptation of 4D flow MRI and hence may aid in the diagnosis and therapeutic management of cardiovascular diseases.

Event-based vision sensors produce an asynchronous stream of events which are triggered when the pixel intensity variation exceeds a predefined threshold. Such sensors offer significant advantages, including reduced data redundancy, micro-second temporal resolution, and low power consumption, making them valuable for applications in robotics and computer vision. In this work, we consider the problem of video reconstruction from events, and propose an approach for dynamic feedback control of activation thresholds, in which a controller network analyzes the past emitted events and predicts the optimal distribution of activation thresholds for the following time segment. Additionally, we allow a user-defined target peak-event-rate for which the control network is conditioned and optimized to predict per-column activation thresholds that would eventually produce the best possible video reconstruction. The proposed OnTheFly control scheme is data-driven and trained in an end-to-end fashion using probabilistic relaxation of the discrete event representation. We demonstrate that our approach outperforms both fixed and randomly-varying threshold schemes by 6-12% in terms of LPIPS perceptual image dissimilarity metric, and by 49% in terms of event rate, achieving superior reconstruction quality while enabling a fine-tuned balance between performance accuracy and the event rate. Additionally, we show that sampling strategies provided by our OnTheFly control are interpretable and reflect the characteristics of the scene. Our results, derived from a physically-accurate simulator, underline the promise of the proposed methodology in enhancing the utility of event cameras for image reconstruction and other downstream tasks, paving the way for hardware implementation of dynamic feedback EVS control in silicon., Comment: 17 pages, 5 figures, ECCV 2024, NEVI workshop

Dynamic nuclear polarization (DNP) involves a polarization transfer from unpaired electrons to hyperfine coupled nuclei and can increase the sensitivity of nuclear magnetic resonance (NMR) signals by several orders of magnitude. The hyperfine coupling is considered to suppress nuclear dipolar flip-flop transitions, hindering the transport of nuclear hyperpolarization into the bulk (''spin-diffusion barrier''). Possible polarization-transfer pathways leading to DNP and subsequent spin diffusion between hypershifted nuclei in a two-electron two-nucleus four-spin system are investigated. The Schrieffer-Wolff transformation is applied to characterize transitions that are only possible as second-order effects. An energy-conserving electron-nuclear four-spin flip-flop is identified, which combines an electron dipolar with a nuclear dipolar flip-flop process, describing spin diffusion close to electrons. The relevance of this process is supported by two-compartment model fits of HypRes-on experimental data. This suggests that all nuclear spins can contribute to the hyperpolarization of the bulk and the concept of a spin-diffusion barrier has to be reconsidered for samples with significant electron and nuclear dipolar couplings.

Si nanoparticles (NPs) have been actively developed as a hyperpolarized magnetic resonance imaging (MRI) contrast agent with an imaging window close to one hour. However, the progress in the development of NPs has been hampered by the incomplete understanding of their structural properties that correspond to efficient hyperpolarization buildup and long polarization decays. In this work we study dynamic nuclear polarization (DNP) of single crystal porous Si (PSi) NPs with defined doping densities ranging from nominally undoped to highly doped with boron or phosphorus. To develop such PSi NPs we perform low-load metal-assisted catalytic etching for electronic grade Si powder followed by thermal oxidation to form the dangling bonds in the Si/SiO$_2$ interface, the $P_b$ centers. $P_b$ centers are the endogenous source of the unpaired electron spins necessary for DNP. The controlled fabrication and oxidation procedures allow us to thoroughly investigate the impact of the magnetic field, temperature and doping on the DNP process. We argue that the buildup and decay rate constants are independent of size of Si crystals between approximately 10 and 60 nm. Instead, the rates are limited by the polarization transfer across the nuclear spin diffusion barrier determined by the large hyperfine shift of the central $^{29}$Si nuclei of the $P_b$ centers. The size-independent rates are then weakly affected by the doping degree for low and moderately doped Si although slight doping is required to achieve the highest polarization. Thus, we find the room temperature relaxation of low boron doped PSi NPs reaching $75 \pm 3$ minutes and nuclear polarization levels exceeding $\sim 6$ % when polarized at 6.7 T and 1.4 K. Our study thus establishes solid grounds for further development of Si NPs as hyperpolarized contrast agents., Comment: Main text: 47 pages, 6 figures. Supporting information: 26 pages, 25 figures. Main text and supporting information are combined into one PDF file. Nanoscale, Advance Article (2024)

Dynamic nuclear polarization enables the hyperpolarization of nuclear spins beyond the thermal-equilibrium Boltzmann distribution. However, it is often unclear why the experimentally measured hyperpolarization is below the theoretical achievable maximum polarization. We report a (near-) resonant relaxation enhancement by microwave (MW) irradiation, leading to a significant increase in the nuclear polarization decay compared to measurements without MW irradiation. For example, the increased nuclear relaxation limits the achievable polarization levels to around 35% instead of hypothetical 60%, measured in the DNP material TEMPO in 1H glassy matrices at 3.3 K and 7 T. Applying rate-equation models to published build-up and decay data indicates that such relaxation enhancement is a common issue in many samples when using different radicals at low sample temperatures and high Boltzmann polarizations of the electrons. Accordingly, quantification and a better understanding of the relaxation processes under MW irradiation might help to design samples and processes towards achieving higher nuclear hyperpolarization levels.better understanding of the relaxation processes under MW irradiation might help to design samples and processes towards achieving higher nuclear hyperpolarization levels.

Magnetic Resonance cardiac diffusion tensor imaging (cDTI) and cardiac intravoxel incoherent motion imaging enables probing of in vivo myofiber architecture and myocardial perfusion surrogates. To study the impact of experimental parameters such as resolution, off-resonances and heart-rate variations, we propose a numerical open-source framework called MRXCAT-CDTI. It allows simulating diffusion and perfusion contrast for spin-echo (SE) and stimulated echo acquisition mode (STEAM) cDTI sequences. The Fourier encoder supports in-plane and/or through-slice off-resonance effects, as well as T2* effects during single-shot image encoding. Optional lesions are included to mimic ischemic and infarcted myocardial regions. MRXCAT-CDTI allows assessing realistic influences on data acquisition, and how these affect the data encoding process and subsequent data processing. As an example, heart-rate variations lead to differences in partial saturation and relaxation of magnetization that end up in errors of 9 to 30% for cDTI angle metrics if not accounted for. For SE echo-planar cDTI, in-plane off-resonance effects more adversely affect cDTI metrics compared to through-slice off-resonances. With this work we propose an open-source MRXCAT-CDTI numerical simulation framework that offers realistic image encoding effects found in cardiac diffusion and perfusion data to systematically study influences of data encoding, reconstruction, and post-processing to promote reproducible research., Comment: 43 pages, 8 figures, 4 supplementary figures, 7 supplementary tables. MRXCAT-CDTI and example scripts to simulate the examples shown in this publication are available online via: https://gitlab.ethz.ch/ibt-cmr-public/mrxcat-cdti-public

Purpose: To compare the segmentation and detection performance of a deep learning model trained on a database of human-labelled clinical diffusion-weighted (DW) stroke lesions to a model trained on the same database enhanced with synthetic DW stroke lesions. Methods: In this institutional review board approved study, a stroke database of 962 cases (mean age 65+/-17 years, 255 males, 449 scans with DW positive stroke lesions) and a normal database of 2,027 patients (mean age 38+/-24 years,1088 females) were obtained. Brain volumes with synthetic DW stroke lesions were produced by warping the relative signal increase of real strokes to normal brain volumes. A generic 3D U-Net was trained on four different databases to generate four different models: (a) 375 neuroradiologist-labeled clinical DW positive stroke cases(CDB);(b) 2,000 synthetic cases(S2DB);(c) CDB+2,000 synthetic cases(CS2DB); or (d) CDB+40,000 synthetic cases(CS40DB). The models were tested on 20%(n=192) of the cases of the stroke database, which were excluded from the training set. Segmentation accuracy was characterized using Dice score and lesion volume of the stroke segmentation, and statistical significance was tested using a paired, two-tailed, Student's t-test. Detection sensitivity and specificity was compared to three neuroradiologists. Results: The performance of the 3D U-Net model trained on the CS40DB(mean Dice 0.72) was better than models trained on the CS2DB (0.70,P <0.001) or the CDB(0.65,P<0.001). The deep learning model was also more sensitive (91%[89%-93%]) than each of the three human readers(84%[81%-87%],78%[75%-81%],and 79%[76%-82%]), but less specific(75%[72%-78%] vs for the three human readers (96%[94%-97%],92%[90%-94%] and 89%[86%-91%]). Conclusion: Deep learning training for segmentation and detection of DW stroke lesions was significantly improved by enhancing the training set with synthetic lesions., Comment: This manuscript has been accepted for publication in Radiology: Artificial Intelligence

Purpose: To simulate the intravoxel incoherent perfusion magnetic resonance magnitude signal from the motion of blood particles in three realistic vascular network graphs from a mouse brain. Methods: In three networks generated from the cortex of a mouse scanned by two-photon laser microscopy, blood flow in each vessel was simulated using Poiseuille law. The trajectories, flow speeds and phases acquired by a fixed number of simulated blood particles during a Stejskal-Tanner monopolar pulse gradient scheme were computed. The resulting magnitude signal as a function of b-value was obtained by integrating all phases and the pseudo-diffusion coefficient D* was estimated by fitting an exponential signal decay. To better understand the anatomical source of the IVIM perfusion signal, the above was repeated by restricting the simulation to various types of vessels. Results: The characteristics of the three microvascular networks were respectively: vessel lengths [mean +/- std. dev.]: 67.2 +/- 53.6 um, 59.8 +/- 46.2 um, and 64.5 +/- 50.9 um; diameters: 6.0 +/- 3.5 um, network 2: 5.7 +/- 3.6 um, and network 3: 6.1 +/- 3.7 um; simulated blood velocity: 0.9 +/- 1.7 um/ms, 1.4 +/- 2.5 um/ms and 0.7 +/- 2.1 um/ms. Exponential fitting of the simulated signal decay as a function of b-value resulted in the following D* [10-3 mm2/s]: 31.7, 40.4 and 33.4. The signal decay for low b-values was the largest in the larger vessels, but the smaller vessels and the capillaries accounted more to the total volume of the networks. Conclusion:This simulation improves the theoretical understanding of the IVIM perfusion estimation method by directly linking the MR IVIM perfusion signal to an ultra-high resolution measurement of the microvascular network and a realistic blood flow simulation., Comment: 22 pages, 5 Figures

Phase-contrast magnetic resonance imaging (MRI) provides time-resolved quantification of blood flow dynamics that can aid clinical diagnosis. Long in vivo scan times due to repeated three-dimensional (3D) volume sampling over cardiac phases and breathing cycles necessitate accelerated imaging techniques that leverage data correlations. Standard compressed sensing reconstruction methods require tuning of hyperparameters and are computationally expensive, which diminishes the potential reduction of examination times. We propose an efficient model-based deep neural reconstruction network and evaluate its performance on clinical aortic flow data. The network is shown to reconstruct undersampled 4D flow MRI data in under a minute on standard consumer hardware. Remarkably, the relatively low amounts of tunable parameters allowed the network to be trained on images from 11 reference scans while generalizing well to retrospective and prospective undersampled data for various acceleration factors and anatomies., Comment: 15 pages, 6 figures