Search

ECG-gated cine imaging in breath-hold enables high-quality diagnostics in most patients, arrhythmia and inability to hold breath, however, can severely corrupt outcomes. Real-time cardiac MRI in free-breathing leverages robust and faster investigations regardless of these confounding factors. With the need for sufficient acceleration, adequate reconstruction methods, which transfer data into high quality images, are required. Undersampled spiral real-time acquisitions in free-breathing were conducted in a study with 16 healthy volunteers and 5 patients. Image reconstructions were performed using a novel score-based diffusion model, as well as a variational network and different compressed sensing approaches. The techniques were compared by means of an expert reader study, by calculating scalar metrics and difference images with respect to a segmented reference, and by a Bland-Altman analysis of cardiac functional parameters. In participants with irregular RR-cycles, spiral real-time acquisitions showed superior image quality with respect to the clinical reference standard. Reconstructions using the diffusion model, the variational network and l1-wavelets offered an overall comparable image quality, however sharpness was slightly increased by the diffusion approach. While slightly larger ejection fractions for the real-time acquisitions were exhibited with a bias of 1.6% for healthy subjects, differences in the data acquisition procedure resulted in uncertainties of 7.4%. The proposed real-time technique enables free-breathing acquisitions of spatio-temporal images with high-quality, covering the entire heart in less than one minute. Evaluation of the ejection fractions using the segmented reference can be significantly corrupted due to arrhythmias and averaging effects. Prolonged inference times of the diffusion model represent the main obstacle to overcome for clinical translation., Comment: 29 pages, 8 figures

Machine learning architectures, including transformers and recurrent neural networks (RNNs) have revolutionized forecasting in applications ranging from text processing to extreme weather. Notably, advanced network architectures, tuned for applications such as natural language processing, are transferable to other tasks such as spatiotemporal forecasting tasks. However, there is a scarcity of ablation studies to illustrate the key components that enable this forecasting accuracy. The absence of such studies, although explainable due to the associated computational cost, intensifies the belief that these models ought to be considered as black boxes. In this work, we decompose the key architectural components of the most powerful neural architectures, namely gating and recurrence in RNNs, and attention mechanisms in transformers. Then, we synthesize and build novel hybrid architectures from the standard blocks, performing ablation studies to identify which mechanisms are effective for each task. The importance of considering these components as hyper-parameters that can augment the standard architectures is exhibited on various forecasting datasets, from the spatiotemporal chaotic dynamics of the multiscale Lorenz 96 system, the Kuramoto-Sivashinsky equation, as well as standard real world time-series benchmarks. A key finding is that neural gating and attention improves the performance of all standard RNNs in most tasks, while the addition of a notion of recurrence in transformers is detrimental. Furthermore, our study reveals that a novel, sparsely used, architecture which integrates Recurrent Highway Networks with neural gating and attention mechanisms, emerges as the best performing architecture in high-dimensional spatiotemporal forecasting of dynamical systems.

A joint image reconstruction and segmentation approach based on disentangled representation learning was trained to enable cardiac cine MR imaging in real-time and under free-breathing. An exploratory feasibility study tested the proposed method in undersampled real-time acquisitions based on an in-house developed spiral bSSFP pulse sequence in eight healthy participants and five patients with intermittent atrial fibrillation. Images and predicted LV segmentations were compared to the reference standard of ECG-gated segmented Cartesian cine in repeated breath-holds and corresponding manual segmentation. On a 5-point Likert scale, image quality of the real-time breath-hold approach and Cartesian cine was comparable in healthy participants (RT-BH: 1.99 $\pm$ .98, Cartesian: 1.94 $\pm$ .86, p=.052), but slightly inferior in free-breathing (RT-FB: 2.40 $\pm$ .98, p<.001). In patients with arrhythmia, image quality from both real-time approaches was favourable (RT-BH: 2.10 $\pm$ 1.28, p<.001, RT-FB: 2.40 $\pm$ 1.13, p<.001, Cartesian: 2.68 $\pm$ 1.13). Intra-observer reliability was good (ICC=.77, 95%-confidence interval [.75, .79], p<.001). In functional analysis, a positive bias was observed for ejection fractions derived from the proposed model compared to the clinical reference standard (RT-BH mean EF: 58.5 $\pm$ 5.6%, bias: +3.47%, 95%-confidence interval [-.86, 7.79%], RT-FB mean: 57.9 $\pm$ 10.6%, bias: +1.45%, [-3.02, 5.91%], Cartesian mean: 54.9 $\pm$ 6.7%). The introduced real-time MR imaging technique is capable of acquiring high-quality cardiac cine data in 1-2 minutes without the need for ECG gating and breath-holds. It thus offers a promising alternative to the current clinical practice of segmented acquisition, with shorter scan times, higher patient comfort and increased robustness to arrhythmia and patient incompliance., Comment: Submitted to the Journal of Cardiovascular Magnetic Resonance

Page Stream Segmentation (PSS) is an essential prerequisite for automated document processing at scale. However, research progress has been limited by the absence of realistic public benchmarks. This paper works towards addressing this gap by introducing TABME++, an enhanced benchmark featuring commercial Optical Character Recognition (OCR) annotations. We evaluate the performance of large language models (LLMs) on PSS, focusing on decoder-based models fine-tuned with parameter-efficient methods. Our results show that decoder-based LLMs outperform smaller multimodal encoders. Through a review of existing PSS research and datasets, we identify key challenges and advancements in the field. Our findings highlight the key importance of robust OCR, providing valuable insights for the development of more effective document processing systems.

We use branes to generalize the Distance Conjecture. We conjecture that in any infinite-distance limit in the moduli space of a $d$-dimensional quantum gravity theory, among the set of particle towers and fundamental branes with at most $p_\text{max}\leq d-2$ spacetime dimensions, at least one has mass/tension decreasing exponentially $T\sim \exp(-\alpha\Delta)$ with the moduli space distance $\Delta$ at a rate of at least $\alpha\geq 1/\sqrt{d-p_\text{max}-1}$. Since $p_\text{max}$ can vary, this represents multiple conditions, where the Sharpened Distance Conjecture is the $p_\text{max}=1$ case. This conjecture is a necessary condition imposed on higher-dimensional theories in order for the Sharpened Distance Conjecture to hold in lower-dimensional theories. We test our conjecture in theories with maximal and half-maximal supersymmetry in diverse dimensions, finding that it is satisfied and often saturated. In some cases where it is saturated -- most notably, heterotic string theory in 10 dimensions -- we argue that novel, low-tension non-supersymmetric branes must exist. We also identify patterns relating the rates at which various brane tensions vary in infinite-distance limits and relate these tensions to the species scale., Comment: 41 pages plus appendices, 14 figures

The Emergent String Conjecture constrains the possible types of light towers in infinite-distance limits in quantum gravity moduli spaces. In this paper, we use these constraints to restrict the geometry of the scalar charge-to-mass vectors $(-\vec{\nabla}\log m)$ of the light towers and the analogous vector $(-\vec{\nabla}\log\Lambda_{\text{QG}})$ of the species scale. We derive taxonomic rules that these vectors must satisfy in each duality frame. Under certain assumptions, this allows us to classify the ways in which different duality frames can fit together globally in the moduli space in terms of a finite list of polytopes. Many of these polytopes arise in known string theory compactifications, while others suggest either undiscovered corners of the landscape or new swampland constraints., Comment: 69 pages plus appendices, 30 figures

Active fluids, such as suspensions of microswimmers, are known to self-organize into complex spatio-temporal flow patterns. An intriguing example is mesoscale turbulence, a state of dynamic vortex structures exhibiting a characteristic length scale. Here, we employ a minimal model for the effective microswimmer velocity field to explore how the turbulent state develops from regular vortex patterns when the strength of activity resp. related parameters such as nonlinear advection or polar alignment strength - is increased. First, we demonstrate analytically that the system, without any spatial constraints, develops a stationary square vortex lattice in the absence of nonlinear advection. Subsequently, we perform an extended stability analysis of this nonuniform "ground state" and uncover a linear instability, which follows from the mutual excitement and simultaneous growth of multiple perturbative modes. This extended analysis is based on linearization around an approximation of the analytical vortex lattice solution and allows us to calculate critical activity parameters. Above these critical values, the vortex lattice develops into mesoscale turbulence in numerical simulations. Utilizing the numerical approach, we uncover an extended region of hysteresis where both patterns are possible depending on the initial condition. Here, we find that turbulence persists below the instability of the vortex lattice. We further determine the stability of square vortex patterns as a function of their wavenumber and represent the results analogous to the well-known Busse balloons known from classical pattern-forming systems. Here, the region of stable periodic patterns shrinks and eventually disappears with increasing activity parameters. Our results show that the strength of activity plays a similar role for active turbulence as the Reynolds number does in driven flow exhibiting inertial turbulence.

Motivated by indirect measurements and applications from nanometrology with a mixed noise model, we develop a novel algorithm for jointly estimating the posterior and the noise parameters in Bayesian inverse problems. We propose to solve the problem by an expectation maximization (EM) algorithm. Based on the current noise parameters, we learn in the E-step a conditional normalizing flow that approximates the posterior. In the M-step, we propose to find the noise parameter updates again by an EM algorithm, which has analytical formulas. We compare the training of the conditional normalizing flow with the forward and reverse KL, and show that our model is able to incorporate information from many measurements, unlike previous approaches., Comment: Published in Machine Learning: Science and Technology

We present a complete proof of the Weak Gravity Conjecture in any perturbative bosonic string theory in spacetime dimension $D\ge6$. Our proof works by relating the black hole extremality bound to long range forces, which are more easily calculated on the worldsheet, closing the gaps in partial arguments in the existing literature. We simultaneously establish a strict, sublattice form of the conjecture in the same class of theories. We close by discussing the scope and limitations of our analysis, along with possible extensions including an upcoming generalization of our work to the superstring., Comment: 33 pages plus appendices, 6 figures

By exploiting complementary sensor information, radar and camera fusion systems have the potential to provide a highly robust and reliable perception system for advanced driver assistance systems and automated driving functions. Recent advances in camera-based object detection offer new radar-camera fusion possibilities with bird's eye view feature maps. In this work, we propose a novel and flexible fusion network and evaluate its performance on two datasets: nuScenes and View-of-Delft. Our experiments reveal that while the camera branch needs large and diverse training data, the radar branch benefits more from a high-performance radar. Using transfer learning, we improve the camera's performance on the smaller dataset. Our results further demonstrate that the radar-camera fusion approach significantly outperforms the camera-only and radar-only baselines., Comment: EUSIPCO 2023

Standard axion electrodynamics has two closely related features. First, the coupling of a massless axion field to photons is quantized, in units proportional to the electric gauge coupling squared. Second, the equations of motion tell us that a time-dependent axion field in a background magnetic field sources an effective electric current, but a time-dependent axion field in a background electric field has no effect. These properties, which manifestly violate electric-magnetic duality, play a crucial role in experimental searches for axions. Recently, electric-magnetic duality has been used to motivate the possible existence of non-standard axion couplings, which can both violate the usual quantization rule and exchange the roles of electric and magnetic fields in axion electrodynamics. We show that these non-standard couplings can be derived from SL(2,Z) duality, but that they come at a substantial cost: in non-standard axion electrodynamics, all electrically charged particles become dyons when the axion traverses its field range, in a dual form of the standard Witten effect monodromy. This implies that there are dyons near the weak scale, leads to a large axion mass induced by Standard Model fermion loops, and dramatically alters Higgs physics. We conclude that non-standard axion electrodynamics, although interesting to consider in abstract quantum field theory, is not phenomenologically viable., Comment: 33 pages, 5 figures; v2: small clarifications, accepted for publication in JHEP

We study towers of light particles that appear in infinite-distance limits of moduli spaces of 9-dimensional $\mathcal{N}=1$ string theories, some of which notably feature decompactification limits with running string coupling. The lightest tower in such decompactification limits consists of the non-BPS Kaluza-Klein modes of Type I$'$ string theory, whose masses depend nontrivially on the moduli of the theory. We work out the moduli-dependence by explicit computation, finding that despite the running decompactification the Distance Conjecture remains satisfied with an exponential decay rate $\alpha \ge \frac{1}{\sqrt{d-2}}$ in accordance with the sharpened Distance Conjecture. The related sharpened Convex Hull Scalar Weak Gravity Conjecture also passes stringent tests. Our results non-trivially test the Emergent String Conjecture, while highlighting the important subtlety that decompactification can lead to a running solution rather than to a higher-dimensional vacuum., Comment: 64 pages, 15 figures