Search

In this paper, we tackle the challenge of instance segmentation for foreign objects in chest radiographs, commonly seen in postoperative follow-ups with stents, pacemakers, or ingested objects in children. The diversity of foreign objects complicates dense annotation, as shown in insufficient existing datasets. To address this, we propose the simple generation of synthetic data through (1) insertion of arbitrary shapes (lines, polygons, ellipses) with varying contrasts and opacities, and (2) cut-paste augmentations from a small set of semi-automatically extracted labels. These insertions are guided by anatomy labels to ensure realistic placements, such as stents appearing only in relevant vessels. Our approach enables networks to segment complex structures with minimal manually labeled data. Notably, it achieves performance comparable to fully supervised models while using 93\% fewer manual annotations.

Two-dimensional material (2DM)-based field-effect transistors (FETs), such as molybdenum disulfide (MoS${_2}$)-FETs, have gained significant attention for their potential for ultra-short channels, thereby extending Moore's law. However, MoS${_2}$-FETs are prone to the formation of Schottky barriers at the metal-MoS${_2}$ interface, resulting in high contact resistance (R${_c}$) and, consequently, reduced transistor currents in the ON-state. Our study explores the modification of MoS${_2}$ to induce the formation of conductive 1T-MoS${_2}$ at the metal-MoS${_2}$ interface via reverse sputtering. MoS${_2}$-FETs exposed to optimized reverse sputtering conditions in the contact area show R${_c}$ values reduced to less than 50% of their untreated counterparts. This reduction translates into improvements in other electrical characteristics, such as higher ON-state currents. Since reverse sputtering is a standard semiconductor process that enhances the electrical performance of MoS${_2}$-FETs, it has great potential for broader application scenarios in 2DM-based microelectronic devices and circuits., Comment: 33 pages

A reliable and scalable transfer of 2D-TMDCs (two-dimensional transition metal dichalcogenides) from the growth substrate to a target substrate with high reproducibility and yield is a crucial step for device integration. In this work, we have introduced a scalable dry-transfer approach for 2D-TMDCs grown by MOCVD (metal-organic chemical vapor deposition) on sapphire. Transfer to a silicon/silicon dioxide (Si/SiO$_2$) substrate is performed using PMMA (poly(methyl methacrylate)) and TRT (thermal release tape) as sacrificial layer and carrier, respectively. Our proposed method ensures a reproducible peel-off from the growth substrate and better preservation of the 2D-TMDC during PMMA removal in solvent, without compromising its adhesion to the target substrate. A comprehensive comparison between the dry method introduced in this work and a standard wet transfer based on potassium hydroxide (KOH) solution shows improvement in terms of cleanliness and structural integrity for dry-transferred layer, as evidenced by X-ray photoemission and Raman spectroscopy, respectively. Moreover, fabricated field-effect transistors (FETs) demonstrate improvements in subthreshold slope, maximum drain current and device-to-device variability. The dry-transfer method developed in this work enables large-area integration of 2D-TMDC layers into (opto)electronic components with high reproducibility, while better preserving the as-grown properties of the layers., Comment: 29 pages

Lesion segmentation in PET/CT imaging is essential for precise tumor characterization, which supports personalized treatment planning and enhances diagnostic precision in oncology. However, accurate manual segmentation of lesions is time-consuming and prone to inter-observer variability. Given the rising demand and clinical use of PET/CT, automated segmentation methods, particularly deep-learning-based approaches, have become increasingly more relevant. The autoPET III Challenge focuses on advancing automated segmentation of tumor lesions in PET/CT images in a multitracer multicenter setting, addressing the clinical need for quantitative, robust, and generalizable solutions. Building on previous challenges, the third iteration of the autoPET challenge introduces a more diverse dataset featuring two different tracers (FDG and PSMA) from two clinical centers. To this extent, we developed a classifier that identifies the tracer of the given PET/CT based on the Maximum Intensity Projection of the PET scan. We trained two individual nnUNet-ensembles for each tracer where anatomical labels are included as a multi-label task to enhance the model's performance. Our final submission achieves cross-validation Dice scores of 76.90% and 61.33% for the publicly available FDG and PSMA datasets, respectively. The code is available at https://github.com/hakal104/autoPETIII/ ., Comment: AutoPET III challenge submission

Layered two-dimensional (2D) semiconductors have shown enhanced ion migration capabilities along their van der Waals (vdW) gaps and on their surfaces. This effect can be employed for resistive switching (RS) in devices for emerging memories, selectors, and neuromorphic computing. To date, all lateral molybdenum disulfide (MoS${_2}$)-based volatile RS devices with silver (Ag) ion migration have been demonstrated using exfoliated, single-crystal MoS${_2}$ flakes requiring a forming step to enable RS. Here, we present volatile RS with multilayer MoS${_2}$ grown by metal-organic chemical vapor deposition (MOCVD) with repeatable forming-free operation. The devices show highly reproducible volatile RS with low operating voltages of approximately 2 V and fast switching times down to 130 ns considering their micrometer scale dimensions. We investigate the switching mechanism based on Ag ion surface migration through transmission electron microscopy, electronic transport modeling, and density functional theory. Finally, we develop a physics-based compact model and explore the implementation of our volatile memristors as artificial neurons in neuromorphic systems., Comment: 43 pages

Two-dimensional materials (2DMs) have been widely investigated because of their potential for heterogeneous integration with modern electronics. However, several major challenges remain, such as the deposition of high-quality dielectrics on 2DMs and the tuning of the 2DM doping levels. Here, we report a scalable plasma-enhanced atomic layer deposition (PEALD) process for direct deposition of a nonstoichiometric aluminum oxide (AlOX) dielectric, overcoming the damage issues associated with conventional methods. Furthermore, we control the thickness of the dielectric layer to systematically tune the doping level of 2DMs. The experimental results demonstrate successful deposition without detectable damage, as confirmed by Raman spectroscopy and electrical measurements. Our method enables tuning of the Dirac and threshold voltages of back-gated graphene and MoS${_2}$ field-effect transistors (FETs), respectively, while also increasing the charge carrier mobility in both device types. We further demonstrate the method in top-gated MoS${_2}$ FETs with double-stack dielectric layers (AlOX+Al${_2}$O${_3}$), achieving critical breakdown field strengths of 7 MV/cm and improved mobility compared with the back gate configuration. In summary, we present a PEALD process that offers a scalable and low-damage solution for dielectric deposition on 2DMs, opening new possibilities for precise tuning of device characteristics in heterogeneous electronic circuits., Comment: 28 pages

Artificial neural networks (ANNs) have evolved from the 1940s primitive models of brain function to become tools for artificial intelligence. They comprise many units, artificial neurons, interlinked through weighted connections. ANNs are trained to perform tasks through learning rules that modify the connection weights. With these rules being in the focus of research, ANNs have become a branch of machine learning developing independently from neuroscience. Although likely required for the development of truly intelligent machines, the integration of neuroscience into ANNs has remained a neglected proposition. Here, we demonstrate that designing an ANN along biological principles results in drastically improved task performance. As a challenging real-world problem, we choose real-time ocean-wave prediction which is essential for various maritime operations. Motivated by the similarity of ocean waves measured at a single location to sound waves arriving at the eardrum, we redesign an echo state network to resemble the brain's auditory system. This yields a powerful predictive tool which is computationally lean, robust with respect to network parameters, and works efficiently across a wide range of sea states. Our results demonstrate the advantages of integrating neuroscience with machine learning and offer a tool for use in the production of green energy from ocean waves., Comment: 23 pages, 6 figures

Infra-gravity waves are generally known as small-amplitude waves of periods between 25 seconds and 5 minutes. They originate from the presence of wave groups in the open ocean waves and can move freely after being released near the surf zone where they can be further fueled with energy from the spatially varying break point of swell waves . As these waves approach the shore, the relative importance of the infra-gravity wave signal increases, and its impact on the shorter waves gets stronger. In addition, infra-gravity waves drive strong cross-shore currents, which lead to significant back-and-forth motion of the underlying sea water. This strong cross-shore motion has been made visible by recent field studies, where significant cross-shore movement was detected and found to be correlated with the infra-gravity wave signal. In the present work, the connection between infra-gravity waves is explored further using linear wave theory and an established numerical nearshore wave model (BOSZ). It is shown that in all cases, the presence of infra-gravity waves leads to strong cross-shore motion. This behavior can be understood by considering the infra-gravity waves as separate free waves, and then following the fluid particle trajectories excited by these waves. As it is shown, these trajectories have a very large horizontal extent which -- if not separated from the main gravity wave field -- appears as a large, but often not directly visible back-and-forth motion, underlying the more readily observable gravity ocean waves., Comment: 13 pages, 5 figures

Using data from a recent field campaign, we evaluate several breaking criteria with the goal of assessing the accuracy of these criteria in wave breaking detection. Two new criteria are also evaluated. An integral parameter is defined in terms of temporal wave trough area, and a differential parameter is defined in terms of maximum steepness of the crest front period. The criteria tested here are based solely on sea surface elevation derived from standard pressure gauge records. They identify breaking and non-breaking waves with an accuracy between 84% and 89% based on the examined field data., Comment: Eight figures, two tables, submitted to Physics of Fluids

In recent decades, a growing number of discoveries in fields of mathematics have been assisted by computer algorithms, primarily for exploring large parameter spaces that humans would take too long to investigate. As computers and algorithms become more powerful, an intriguing possibility arises - the interplay between human intuition and computer algorithms can lead to discoveries of novel mathematical concepts that would otherwise remain elusive. To realize this perspective, we have developed a massively parallel computer algorithm that discovers an unprecedented number of continued fraction formulas for fundamental mathematical constants. The sheer number of formulas discovered by the algorithm unveils a novel mathematical structure that we call the conservative matrix field. Such matrix fields (1) unify thousands of existing formulas, (2) generate infinitely many new formulas, and most importantly, (3) lead to unexpected relations between different mathematical constants, including multiple integer values of the Riemann zeta function. Conservative matrix fields also enable new mathematical proofs of irrationality. In particular, we can use them to generalize the celebrated proof by Ap\'ery for the irrationality of $\zeta(3)$. Utilizing thousands of personal computers worldwide, our computer-supported research strategy demonstrates the power of experimental mathematics, highlighting the prospects of large-scale computational approaches to tackle longstanding open problems and discover unexpected connections across diverse fields of science., Comment: 21 pages, 6 figures, and 1 table; with 9 appendix sections totaling 12 pages, 1 figure, and 4 tables

The Whitham equation is a model for the evolution of surface waves on shallow water that combines the unidirectional linear dispersion relation of the Euler equations with a weakly nonlinear approximation based on the KdV equation. We show that large-amplitude, periodic, traveling-wave solutions to the Whitham equation and its higher-order generalization, the cubic Whitham equation, are unstable with respect to the superharmonic instability (i.e. a perturbation with the same period as the solution). The threshold between superharmonic stability and instability occurs at the maxima of the Hamiltonian and $\mathcal{L}_2$-norm. We examine the onset of wave breaking in traveling-wave solutions subject to the modulational and superharmonic instabilities. We present new instability results for the Euler equations in finite depth and compare them with the Whitham results. We show that the Whitham equation more accurately approximates the wave steepness threshold for the superharmonic instability of the Euler equations than does the cubic Whitham equation. However, the cubic Whitham equation more accurately approximates the wave steepness threshold for the modulational instability of the Euler equations than does the Whitham equation.

We present a general methodology that learns to classify images without labels by leveraging pretrained feature extractors. Our approach involves self-distillation training of clustering heads based on the fact that nearest neighbours in the pretrained feature space are likely to share the same label. We propose a novel objective that learns associations between image features by introducing a variant of pointwise mutual information together with instance weighting. We demonstrate that the proposed objective is able to attenuate the effect of false positive pairs while efficiently exploiting the structure in the pretrained feature space. As a result, we improve the clustering accuracy over $k$-means on $17$ different pretrained models by $6.1$\% and $12.2$\% on ImageNet and CIFAR100, respectively. Finally, using self-supervised vision transformers, we achieve a clustering accuracy of $61.6$\% on ImageNet. The code is available at https://github.com/HHU-MMBS/TEMI-official-BMVC2023., Comment: Accepted at BMVC2023. Code at https://github.com/HHU-MMBS/TEMI-official-BMVC2023

In this work, a novel Boussinesq system is put forward. The system is naturally nonlinearly entropy/energy-stable, and is designed for problems with sharply varying bathymetric features. The system is flexible and allows tuning of the dispersive parameters to the relevant wavenumber range of the problem at hand. We present a few such parameter sets, including one that tracks the dispersive relation of the underlying Euler equations up to a nondimensional wavenumber of about $30$. In the one-dimensional case, we design a stable finite-volume scheme and demonstrate its robustness and accuracy in a suite of test problems including Dingemans's wave experiment. We generalise the system to the two-dimensional case and sketch how the numerical scheme can be straightforwardly generalised.

When modeling longitudinal biomedical data, often dimensionality reduction as well as dynamic modeling in the resulting latent representation is needed. This can be achieved by artificial neural networks for dimension reduction, and differential equations for dynamic modeling of individual-level trajectories. However, such approaches so far assume that parameters of individual-level dynamics are constant throughout the observation period. Motivated by an application from psychological resilience research, we propose an extension where different sets of differential equation parameters are allowed for observation sub-periods. Still, estimation for intra-individual sub-periods is coupled for being able to fit the model also with a relatively small dataset. We subsequently derive prediction targets from individual dynamic models of resilience in the application. These serve as interpretable resilience-related outcomes, to be predicted from characteristics of individuals, measured at baseline and a follow-up time point, and selecting a small set of important predictors. Our approach is seen to successfully identify individual-level parameters of dynamic models that allows us to stably select predictors, i.e., resilience factors. Furthermore, we can identify those characteristics of individuals that are the most promising for updates at follow-up, which might inform future study design. This underlines the usefulness of our proposed deep dynamic modeling approach with changes in parameters between observation sub-periods.

We demonstrate the design, fabrication, and characterization of wafer-scale, zero-bias power detectors based on two-dimensional MoS$_2$ field effect transistors (FETs). The MoS$_2$ FETs are fabricated using a wafer-scale process on 8 $\mu$m thick polyimide film, which in principle serves as flexible substrate. The performances of two CVD-MoS$_2$ sheets, grown with different processes and showing different thicknesses, are analyzed and compared from the single device fabrication and characterization steps to the circuit level. The power detector prototypes exploit the nonlinearity of the transistors above the cut-off frequency of the devices. The proposed detectors are designed employing a transistor model based on measurement results. The fabricated circuits operate in Ku-band between 12 and 18 GHz, with a demonstrated voltage responsivity of 45 V/W at 18 GHz in the case of monolayer MoS2 and 104 V/W at 16 GHz in the case of multilayer MoS$_2$, both achieved without applied DC bias. They are the best performing power detectors fabricated on flexible substrate reported to date. The measured dynamic range exceeds 30 dB outperforming other semiconductor technologies like silicon complementary metal oxide semiconductor (CMOS) circuits and GaAs Schottky diodes., Comment: 28 pages

We are concerned with numerical approximations of breather solutions for the cubic Whitham equation which arises as a water-wave model for interfacial waves. The model combines strong nonlinearity with the non-local character of the water-wave problem. The equation is non-integrable as suggested by the inelastic interaction of solitary waves. As a non local model, it generalizes, in the low frequency limit, the well known modified KdV (mKdV) equation which is a completely-integrable model. The mKdV equation has breather solutions, i.e. periodic in time and localized in space biparametric solutions. It was recently shown that these breather solutions appear naturally as ground states of invariant integrals, suggesting that such structures may also exist in non-integrable models, at least in an approximate sense. In this work, we present numerical evidence that in the non-integrable case of the cubic Whitham equation, breather solutions may also exist., Comment: 7 pages, 6 figures, updated abstract and references