Search

Current interactive segmentation approaches, inspired by the success of META's Segment Anything model, have achieved notable advancements, however, they come with substantial limitations that hinder their practical application in real clinical scenarios. These include unrealistic human interaction requirements, such as slice-by-slice operations for 2D models on 3D data, a lack of iterative refinement, and insufficient evaluation experiments. These shortcomings prevent accurate assessment of model performance and lead to inconsistent outcomes across studies. IntRaBench overcomes these challenges by offering a comprehensive and reproducible framework for evaluating interactive segmentation methods in realistic, clinically relevant scenarios. It includes diverse datasets, target structures, and segmentation models, and provides a flexible codebase that allows seamless integration of new models and prompting strategies. Additionally, we introduce advanced techniques to minimize clinician interaction, ensuring fair comparisons between 2D and 3D models. By open-sourcing IntRaBench, we invite the research community to integrate their models and prompting techniques, ensuring continuous and transparent evaluation of interactive segmentation models in 3D medical imaging., Comment: Undergoing Peer-Review

Self-Supervised Learning (SSL) presents an exciting opportunity to unlock the potential of vast, untapped clinical datasets, for various downstream applications that suffer from the scarcity of labeled data. While SSL has revolutionized fields like natural language processing and computer vision, their adoption in 3D medical image computing has been limited by three key pitfalls: Small pre-training dataset sizes, architectures inadequate for 3D medical image analysis, and insufficient evaluation practices. We address these issues by i) leveraging a large-scale dataset of 44k 3D brain MRI volumes and ii) using a Residual Encoder U-Net architecture within the state-of-the-art nnU-Net framework. iii) A robust development framework, incorporating 5 development and 8 testing brain MRI segmentation datasets, allowed performance-driven design decisions to optimize the simple concept of Masked Auto Encoders (MAEs) for 3D CNNs. The resulting model not only surpasses previous SSL methods but also outperforms the strong nnU-Net baseline by an average of approximately 3 Dice points. Furthermore, our model demonstrates exceptional stability, achieving the highest average rank of 2 out of 7 methods, compared to the second-best method's mean rank of 3., Comment: Arxiv Preprint. Currently under Review

Fragmentation and evolution for the molecular shells of the compact HII regions are less explored compared to their evolved counterparts. We map nine compact HII regions with a typical diameter of 0.4 pc that are surrounded by molecular shells traced by CCH. Several to a dozen dense gas fragments probed by H13CO+ are embedded in these molecular shells. These gas fragments, strongly affected by the HII region, have a higher surface density, mass, and turbulence than those outside the shells but within the same pc-scale natal clump. These features suggest that the shells swept up by the early HII regions can enhance the formation of massive dense structures that may host the birth of higher-mass stars. We examine the formation of fragments and find that fragmentation of the swept-up shell is unlikely to occur in these early HII regions, by comparing the expected time scale of shell fragmentation with the age of HII region. We propose that the appearance of gas fragments in these shells is probably the result of sweeping up pre-existing fragments into the molecular shell that has not yet fragmented. Taken together, this work provides a basis for understanding the interplay of star-forming sites with an intricate environment containing ionization feedback such as those observed in starburst regions., Comment: Accepted by MNRAS, 24 pages, 14 figures

In this paper, we present a detailed analysis of the IRS 17 filament within the intermediate-mass protocluster IRAS 08448-4343 (of $\sim\,10^3\,\rm L_{\odot}$), using ALMA data from the ATOMS 3-mm and QUARKS 1.3-mm surveys. The IRS 17 filament, which spans $\sim$54000 au ($0.26\,\rm pc$) in length and $\sim$4000 au ($0.02\,\rm pc$) in width, exhibits a complex, multi-component velocity field, and harbours hierarchical substructures. These substructures include three bundles of seven velocity-coherent fibers, and 29 dense ($n\sim 10^8\,\rm cm^{-3}$) condensations. The fibers have a median length of $\sim 4500\,\rm au$ and a median width of $\sim 1400\,\rm au$. Among these fibers, four are identified as ``fertile", each hosting at least three dense condensations, which are regarded as the ``seeds" of star formation. While the detected cores are randomly spaced within the IRS\,17 filament based on the 3-mm dust continuum image, periodic spacing ($\sim1600\,\rm au$) of condensations is observed in the fertile fibers according to the 1.3-mm dust map, consistent with the predictions of linear isothermal cylinder fragmentation models. These findings underscore the crucial role of fibers in star formation and suggest a hierarchical fragmentation process that extends from the filament to the fibers, and ultimately, to the smallest-scale condensations., Comment: 19 pages, 10 figures, 4 tables, accepted by ApJ

Medical imaging is spearheading the AI transformation of healthcare. Performance reporting is key to determine which methods should be translated into clinical practice. Frequently, broad conclusions are simply derived from mean performance values. In this paper, we argue that this common practice is often a misleading simplification as it ignores performance variability. Our contribution is threefold. (1) Analyzing all MICCAI segmentation papers (n = 221) published in 2023, we first observe that more than 50% of papers do not assess performance variability at all. Moreover, only one (0.5%) paper reported confidence intervals (CIs) for model performance. (2) To address the reporting bottleneck, we show that the unreported standard deviation (SD) in segmentation papers can be approximated by a second-order polynomial function of the mean Dice similarity coefficient (DSC). Based on external validation data from 56 previous MICCAI challenges, we demonstrate that this approximation can accurately reconstruct the CI of a method using information provided in publications. (3) Finally, we reconstructed 95% CIs around the mean DSC of MICCAI 2023 segmentation papers. The median CI width was 0.03 which is three times larger than the median performance gap between the first and second ranked method. For more than 60% of papers, the mean performance of the second-ranked method was within the CI of the first-ranked method. We conclude that current publications typically do not provide sufficient evidence to support which models could potentially be translated into clinical practice., Comment: Paper accepted at MICCAI 2024 conference

Artificial intelligence and machine learning are increasingly used to offload decision making from people. In the past, one of the rationales for this replacement was that machines, unlike people, can be fair and unbiased. Evidence suggests otherwise. We begin by entertaining the ideas that algorithms can replace people and that algorithms cannot be biased. Taken as axioms, these statements quickly lead to absurdity. Spurred on by this result, we investigate the slogans more closely and identify equivocation surrounding the word 'bias.' We diagnose three forms of outrage-intellectual, moral, and political-that are at play when people react emotionally to algorithmic bias. Then we suggest three practical approaches to addressing bias that the AI community could take, which include clarifying the language around bias, developing new auditing methods for intelligent systems, and building certain capabilities into these systems. We conclude by offering a moral regarding the conversations about algorithmic bias that may transfer to other areas of artificial intelligence., Comment: Distribution Statement A. Approved for public release; distribution is unlimited

Using the upGREAT instrument on SOFIA, we have imaged the [C II] 158 {\mu}m fine structure line emission in bright-rimmed pillars located at the southern edge of the IC1848 H II region, and carried out pointed observations of the [O I] 63 and 145 {\mu}m fine structure lines toward selected positions. The observations are used to characterize the morphology, velocity field, and the physical conditions in the G1 - G3 filaments. The velocity-resolved [C II] spectra show evidence of a velocity shift at the head of the brightest G1 filament, possibly caused by radiation pressure from the impinging UV photons or the rocket effect of the evaporating gas. Archival Herschel PACS and SPIRE data imply H2 column densities in the range 10^{21} - 10^{22} cm^{-2}, corresponding to maximum visual extinction AV = 10 mag, and average H2 volume density of about 4500 cm^{-3} in the filaments. The [C II] emission traces ~ 17% of the total H2 column density, as derived from dust SED fits. PDR models are unable to explain the observed line intensities of the two [O I] fine structure lines in IC1848, with the observed [O I] 145 {\mu}m line being too strong compared to the model predictions. The [O I] lines in IC1848 are overall weak and the signal-to-noise ratio is limited. However, our observations suggest that the [O I] 63/145 {\mu}m intensity ratio is a sensitive probe of the physical conditions in photon dominated regions such as IC1848. These lines are thus excellent targets for future high-altitude balloon instruments, less affected by telluric absorption., Comment: Accepted for publication in Astronomy and Astrophysics

Explainable AI (XAI) is a rapidly growing domain with a myriad of proposed methods as well as metrics aiming to evaluate their efficacy. However, current studies are often of limited scope, examining only a handful of XAI methods and ignoring underlying design parameters for performance, such as the model architecture or the nature of input data. Moreover, they often rely on one or a few metrics and neglect thorough validation, increasing the risk of selection bias and ignoring discrepancies among metrics. These shortcomings leave practitioners confused about which method to choose for their problem. In response, we introduce LATEC, a large-scale benchmark that critically evaluates 17 prominent XAI methods using 20 distinct metrics. We systematically incorporate vital design parameters like varied architectures and diverse input modalities, resulting in 7,560 examined combinations. Through LATEC, we showcase the high risk of conflicting metrics leading to unreliable rankings and consequently propose a more robust evaluation scheme. Further, we comprehensively evaluate various XAI methods to assist practitioners in selecting appropriate methods aligning with their needs. Curiously, the emerging top-performing method, Expected Gradients, is not examined in any relevant related study. LATEC reinforces its role in future XAI research by publicly releasing all 326k saliency maps and 378k metric scores as a (meta-)evaluation dataset. The benchmark is hosted at: https://github.com/IML-DKFZ/latec., Comment: Accepted at NeurIPS 2024

An open problem in artificial intelligence is how systems can flexibly learn discrete abstractions that are useful for solving inherently continuous problems. Previous work in computational neuroscience has considered this functional integration of discrete and continuous variables during decision-making under the formalism of active inference (Parr, Friston & de Vries, 2017; Parr & Friston, 2018). However, their focus is on the expressive physical implementation of categorical decisions and the hierarchical mixed generative model is assumed to be known. As a consequence, it is unclear how this framework might be extended to learning. We therefore present a novel hierarchical hybrid active inference agent in which a high-level discrete active inference planner sits above a low-level continuous active inference controller. We make use of recent work in recurrent switching linear dynamical systems (rSLDS) which implement end-to-end learning of meaningful discrete representations via the piecewise linear decomposition of complex continuous dynamics (Linderman et al., 2016). The representations learned by the rSLDS inform the structure of the hybrid decision-making agent and allow us to (1) specify temporally-abstracted sub-goals in a method reminiscent of the options framework, (2) lift the exploration into discrete space allowing us to exploit information-theoretic exploration bonuses and (3) `cache' the approximate solutions to low-level problems in the discrete planner. We apply our model to the sparse Continuous Mountain Car task, demonstrating fast system identification via enhanced exploration and successful planning through the delineation of abstract sub-goals., Comment: 11 pages (+ appendix). Accepted to the International Workshop on Active Inference 2024. arXiv admin note: substantial text overlap with arXiv:2408.10970

Medical image classification in radiology faces significant challenges, particularly in generalizing to unseen pathologies. In contrast, CLIP offers a promising solution by leveraging multimodal learning to improve zero-shot classification performance. However, in the medical domain, lesions can be small and might not be well represented in the embedding space. Therefore, in this paper, we explore the potential of visual prompt engineering to enhance the capabilities of Vision Language Models (VLMs) in radiology. Leveraging BiomedCLIP, trained on extensive biomedical image-text pairs, we investigate the impact of embedding visual markers directly within radiological images to guide the model's attention to critical regions. Our evaluation on the JSRT dataset, focusing on lung nodule malignancy classification, demonstrates that incorporating visual prompts $\unicode{x2013}$ such as arrows, circles, and contours $\unicode{x2013}$ significantly improves classification metrics including AUROC, AUPRC, F1 score, and accuracy. Moreover, the study provides attention maps, showcasing enhanced model interpretability and focus on clinically relevant areas. These findings underscore the efficacy of visual prompt engineering as a straightforward yet powerful approach to advance VLM performance in medical image analysis., Comment: Accepted at ECCV 2024 Workshop on Emergent Visual Abilities and Limits of Foundation Models

Zero-field optically pumped magnetometers (OPMs) have emerged as an important technology for biomagnetism due to their ulta-sensitive performance, contained within a non-cryogenic small-scale sensor-head. The compactness of such OPMs is often achieved through simplified detection schemes, which typically provide only single-axis magnetic field information. However, multi-axis static magnetic fields on non-measurement axes cause a systematic error that manifests as amplitude and phase errors across the measurement axis. Here we present a triaxial operational technique for a compact zero-field OPM which suppresses multi-axis systematic errors through simultaneous measurement and closed-loop active control of the static magnetic fields across all axes. The demonstrated technique requires magnetic modulation across two axes while providing static field information for all three axes. We demonstrate this technique on a rubidium laboratory-based zero-field magnetometer, achieving a bandwidth of 380 Hz with sensitivities of $<25$ fT/$\sqrt{\rm{Hz}}$ across both transverse axes and $65$ fT/$\sqrt{\rm{Hz}}$ along the beam axis. Using the proposed triaxial technique, we demonstrate precise tracking of a 2 Hz triaxial vector test signal and suppression of systematic cross-axis projection errors over an extended period, $\simeq20$~min.

An open problem in artificial intelligence is how systems can flexibly learn discrete abstractions that are useful for solving inherently continuous problems. Previous work has demonstrated that a class of hybrid state-space model known as recurrent switching linear dynamical systems (rSLDS) discover meaningful behavioural units via the piecewise linear decomposition of complex continuous dynamics (Linderman et al., 2016). Furthermore, they model how the underlying continuous states drive these discrete mode switches. We propose that the rich representations formed by an rSLDS can provide useful abstractions for planning and control. We present a novel hierarchical model-based algorithm inspired by Active Inference in which a discrete MDP sits above a low-level linear-quadratic controller. The recurrent transition dynamics learned by the rSLDS allow us to (1) specify temporally-abstracted sub-goals in a method reminiscent of the options framework, (2) lift the exploration into discrete space allowing us to exploit information-theoretic exploration bonuses and (3) `cache' the approximate solutions to low-level problems in the discrete planner. We successfully apply our model to the sparse Continuous Mountain Car task, demonstrating fast system identification via enhanced exploration and non-trivial planning through the delineation of abstract sub-goals., Comment: 4 pages, 3 figures

We combine a new Galactic plane survey of Hydrogen Radio Recombination Lines (RRLs) with far-infrared (FIR) surveys of ionized Nitrogen, N+, to determine Nitrogen abundance across Galactic radius. RRLs were observed with NASA DSS-43 70m antenna and the Green Bank Telescope in 108 lines-of-sight spanning -135 degrees < l < 60 degrees, at b=0 degrees. These positions were also observed in [N II] 122 um and 205 um lines with the Herschel Space Observatory. Combining RRL and [N II] 122 um and 205 um observations in 41 of 108 samples with high signal-to-noise ratio, we studied ionized Nitrogen abundance distribution across Galactocentric distances of 0-8 kpc. Combined with existing Solar neighborhood and Outer galaxy N/H abundance determinations, we studied this quantity's distribution within the Milky Way's inner 17 kpc for the first time. We found a Nitrogen abundance gradient extending from Galactocentric radii of 4-17 kpc in the Galactic plane, while within 0-4 kpc, the N/H distribution remained flat. The gradient observed at large Galactocentric distances supports inside-out galaxy growth with the additional steepening resulting from variable star formation efficiency and/or radial flows in the Galactic disk, while the inner 4 kpc flattening, coinciding with the Galactic bar's onset, may be linked to radial flows induced by the bar potential. Using SOFIA/FIFI-LS and Herschel/PACS, we observed the [N III] 57 um line to trace doubly ionized gas contribution in a sub-sample of sightlines. We found negligible N++ contributions along these sightlines, suggesting mostly singly ionized Nitrogen originating from low ionization H II region outskirts., Comment: Accepted for publication at the Astrophysical Journal. 25 pages, 13 figures

Using new continuum and molecular line data from the ALMA Three-millimeter Observations of Massive Star-forming Regions (ATOMS) survey and archival VLA, 4.86 GHz data, we present direct observational evidence of hierarchical triggering relating three epochs of massive star formation in a ring-like H II region, G24.47+0.49. We find from radio flux analysis that it is excited by a massive star(s) of spectral type O8.5V-O8V from the first epoch of star formation. The swept-up ionized ring structure shows evidence of secondary collapse, and within this ring a burst of massive star formation is observed in different evolutionary phases, which constitutes the second epoch. ATOMS spectral line (e.g., HCO$^+$(1-0)) observations reveal an outer concentric molecular gas ring expanding at a velocity of $\sim$ 9 $\rm km\,s^{-1}$, constituting the direct and unambiguous detection of an expanding molecular ring. It harbors twelve dense molecular cores with surface mass density greater than 0.05 $\rm g\,cm^{-2}$, a threshold typical of massive star formation. Half of them are found to be subvirial, and thus in gravitational collapse, making them third epoch of potential massive star-forming sites., Comment: 18 pages, 7 figures, accepted for publication in The Astrophysical Journal Letters

Selective Classification, wherein models can reject low-confidence predictions, promises reliable translation of machine-learning based classification systems to real-world scenarios such as clinical diagnostics. While current evaluation of these systems typically assumes fixed working points based on pre-defined rejection thresholds, methodological progress requires benchmarking the general performance of systems akin to the $\mathrm{AUROC}$ in standard classification. In this work, we define 5 requirements for multi-threshold metrics in selective classification regarding task alignment, interpretability, and flexibility, and show how current approaches fail to meet them. We propose the Area under the Generalized Risk Coverage curve ($\mathrm{AUGRC}$), which meets all requirements and can be directly interpreted as the average risk of undetected failures. We empirically demonstrate the relevance of $\mathrm{AUGRC}$ on a comprehensive benchmark spanning 6 data sets and 13 confidence scoring functions. We find that the proposed metric substantially changes metric rankings on 5 out of the 6 data sets.

Bose-Einstein condensates (BECs) have been proposed for many applications in atom interferometry, as their coherence over long evolution times promises unprecedented sensitivity. To date, BECs can be efficiently created in devices using atom chips, but these are still complex and place high demands on size, weight and power. To further simplify these setups, we equipped an atom chip with a nano-structured diffraction-grating to derive all beams for the magneto-optical trap (MOT) from a single laser beam. Moreover, using a 2D$^+$-MOT as an atomic source and a beam with uniform intensity for the grating illumination, we capture $1\times10^9$ atoms in one second, cool them to $14\,\mu$K, and demonstrate magnetic trapping using the atom chip. This is a major step towards the simplification of portable BEC devices for quantum sensing on earth and in space., Comment: 12 pages, 3 figures

Semantic segmentation is an essential component of medical image analysis research, with recent deep learning algorithms offering out-of-the-box applicability across diverse datasets. Despite these advancements, segmentation failures remain a significant concern for real-world clinical applications, necessitating reliable detection mechanisms. This paper introduces a comprehensive benchmarking framework aimed at evaluating failure detection methodologies within medical image segmentation. Through our analysis, we identify the strengths and limitations of current failure detection metrics, advocating for the risk-coverage analysis as a holistic evaluation approach. Utilizing a collective dataset comprising five public 3D medical image collections, we assess the efficacy of various failure detection strategies under realistic test-time distribution shifts. Our findings highlight the importance of pixel confidence aggregation and we observe superior performance of the pairwise Dice score (Roy et al., 2019) between ensemble predictions, positioning it as a simple and robust baseline for failure detection in medical image segmentation. To promote ongoing research, we make the benchmarking framework available to the community., Comment: This work has been submitted for possible publication

Tsetlin machines (TMs) have been successful in several application domains, operating with high efficiency on Boolean representations of the input data. However, Booleanizing complex data structures such as sequences, graphs, images, signal spectra, chemical compounds, and natural language is not trivial. In this paper, we propose a hypervector (HV) based method for expressing arbitrarily large sets of concepts associated with any input data. Using a hyperdimensional space to build vectors drastically expands the capacity and flexibility of the TM. We demonstrate how images, chemical compounds, and natural language text are encoded according to the proposed method, and how the resulting HV-powered TM can achieve significantly higher accuracy and faster learning on well-known benchmarks. Our results open up a new research direction for TMs, namely how to expand and exploit the benefits of operating in hyperspace, including new booleanization strategies, optimization of TM inference and learning, as well as new TM applications., Comment: 9 pages, 17 figures

Identifying predictive biomarkers, which forecast individual treatment effectiveness, is crucial for personalized medicine and informs decision-making across diverse disciplines. These biomarkers are extracted from pre-treatment data, often within randomized controlled trials, and have to be distinguished from prognostic biomarkers, which are independent of treatment assignment. Our study focuses on the discovery of predictive imaging biomarkers, aiming to leverage pre-treatment images to unveil new causal relationships. Previous approaches relied on labor-intensive handcrafted or manually derived features, which may introduce biases. In response, we present a new task of discovering predictive imaging biomarkers directly from the pre-treatment images to learn relevant image features. We propose an evaluation protocol for this task to assess a model's ability to identify predictive imaging biomarkers and differentiate them from prognostic ones. It employs statistical testing and a comprehensive analysis of image feature attribution. We explore the suitability of deep learning models originally designed for estimating the conditional average treatment effect (CATE) for this task, which previously have been primarily assessed for the precision of CATE estimation, overlooking the evaluation of imaging biomarker discovery. Our proof-of-concept analysis demonstrates promising results in discovering and validating predictive imaging biomarkers from synthetic outcomes and real-world image datasets., Comment: 19 pages, 12 figures

Mamba state-space models (SSMs) were recently shown to outperform state-of-the-art (SOTA) Transformer large language models (LLMs) across various tasks. Despite subsequent widespread adaptation, little work has focused on Mamba LLMs' amenability for fine-tuning frameworks ubiquitously used for Transformer-based LLMs, e.g., mixed-precision fine-tuning (MPFT) and parameter-efficient fine-tuning (PEFT). For the former, it currently remains an open question whether Mamba's recurrent dynamics are robust to small input changes, such as those encountered during MPFT. Using dynamical systems theory (in particular, Lyapunov exponents), we answer this question in the affirmative. We empirically validate this result through several experiments, showing that Mamba SSMs are significantly more stable to changes introduced by mixed-precision than comparable Transformers, even when both MPFT and PEFT are combined. For PEFT, we show how targeting specific memory buffers in Mamba's customized CUDA kernels for low-rank adaptation regularizes SSM parameters, thus providing both parameter efficient learning and computational savings. Finally, with both MPFT and PEFT enabled, we explore the impact of instruction tuning Mamba SSMs for in-context learning (ICL) on natural language tasks. While pretrained Mamba and Mamba-2 models only achieve 38% and 82% (respectively) of the ICL improvements of comparable Transformer-based LLMs, we show that instruction tuning allows Mamba models to narrow this gap to 81% and Mamba-2 models to skyrocket over this gap to 132%., Comment: 18 pages, 4 figure panels, 5 tables

Ultracold atoms are crucial for unlocking truly precise and accurate quantum metrology, and provide an essential platform for quantum computing, communication and memories. One of the largest ongoing challenges is the miniaturization of these quantum devices. Here, we show that the typically macroscopic optical lattice architecture at the heart of many ultra-precise quantum technologies can be realized with a single input laser beam on the same diffractive chip already used to create the ultracold atoms. Moreover, this inherently ultra-stable platform enables access to a plethora of new lattice dimensionalities and geometries, ideally suited for the design of high-accuracy, portable quantum devices., Comment: 7 pages, 2 figures

The hippocampus has been associated with both spatial cognition and episodic memory formation, but integrating these functions into a unified framework remains challenging. Here, we demonstrate that forming discrete memories of visual events in sparse autoencoder neurons can produce spatial tuning similar to hippocampal place cells. We then show that the resulting very high-dimensional code enables neurons to discretize and tile the underlying image space with minimal overlap. Additionally, we extend our results to the auditory domain, showing that neurons similarly tile the frequency space in an experience-dependent manner. Lastly, we show that reinforcement learning agents can effectively perform various visuo-spatial cognitive tasks using these sparse, very high-dimensional representations.

On 24 September 2023, the NASA OSIRIS-REx mission dropped a capsule to Earth containing approximately 120 g of pristine carbonaceous regolith from Bennu. We describe the delivery and initial allocation of this asteroid sample and introduce its bulk physical, chemical, and mineralogical properties from early analyses. The regolith is very dark overall, with higher-reflectance inclusions and particles interspersed. Particle sizes range from sub-micron dust to a stone about 3.5 cm long. Millimeter-scale and larger stones typically have hummocky or angular morphologies. A subset of the stones appears mottled by brighter material that occurs as veins and crusts. Hummocky stones have the lowest densities and mottled stones have the highest. Remote sensing of the surface of Bennu detected hydrated phyllosilicates, magnetite, organic compounds, carbonates, and scarce anhydrous silicates, all of which the sample confirms. We also find sulfides, presolar grains, and, less expectedly, Na-rich phosphates, as well as other trace phases. The sample composition and mineralogy indicate substantial aqueous alteration and resemble those of Ryugu and the most chemically primitive, low-petrologic-type carbonaceous chondrites. Nevertheless, we find distinct hydrogen, nitrogen, and oxygen isotopic compositions, and some of the material we analyzed is enriched in fluid-mobile elements. Our findings underscore the value of sample return, especially for low-density material that may not readily survive atmospheric entry, and lay the groundwork for more comprehensive analyses., Comment: 73 pages, 22 figures

The release of nnU-Net marked a paradigm shift in 3D medical image segmentation, demonstrating that a properly configured U-Net architecture could still achieve state-of-the-art results. Despite this, the pursuit of novel architectures, and the respective claims of superior performance over the U-Net baseline, continued. In this study, we demonstrate that many of these recent claims fail to hold up when scrutinized for common validation shortcomings, such as the use of inadequate baselines, insufficient datasets, and neglected computational resources. By meticulously avoiding these pitfalls, we conduct a thorough and comprehensive benchmarking of current segmentation methods including CNN-based, Transformer-based, and Mamba-based approaches. In contrast to current beliefs, we find that the recipe for state-of-the-art performance is 1) employing CNN-based U-Net models, including ResNet and ConvNeXt variants, 2) using the nnU-Net framework, and 3) scaling models to modern hardware resources. These results indicate an ongoing innovation bias towards novel architectures in the field and underscore the need for more stringent validation standards in the quest for scientific progress., Comment: Accepted at MICCAI 2024

Penalized transformation models (PTMs) are a novel form of location-scale regression. In PTMs, the shape of the response's conditional distribution is estimated directly from the data, and structured additive predictors are placed on its location and scale. The core of the model is a monotonically increasing transformation function that relates the response distribution to a reference distribution. The transformation function is equipped with a smoothness prior that regularizes how much the estimated distribution diverges from the reference distribution. These models can be seen as a bridge between conditional transformation models and generalized additive models for location, scale and shape. Markov chain Monte Carlo inference for PTMs can be conducted with the No-U-Turn sampler and offers straightforward uncertainty quantification for the conditional distribution as well as for the covariate effects. A simulation study demonstrates the effectiveness of the approach. We apply the model to data from the Fourth Dutch Growth Study and the Framingham Heart Study. A full-featured implementation is available as a Python library.

PURPOSE: Deep learning methods for classifying prostate cancer (PCa) in ultrasound images typically employ convolutional networks (CNNs) to detect cancer in small regions of interest (ROI) along a needle trace region. However, this approach suffers from weak labelling, since the ground-truth histopathology labels do not describe the properties of individual ROIs. Recently, multi-scale approaches have sought to mitigate this issue by combining the context awareness of transformers with a CNN feature extractor to detect cancer from multiple ROIs using multiple-instance learning (MIL). In this work, we present a detailed study of several image transformer architectures for both ROI-scale and multi-scale classification, and a comparison of the performance of CNNs and transformers for ultrasound-based prostate cancer classification. We also design a novel multi-objective learning strategy that combines both ROI and core predictions to further mitigate label noise. METHODS: We evaluate 3 image transformers on ROI-scale cancer classification, then use the strongest model to tune a multi-scale classifier with MIL. We train our MIL models using our novel multi-objective learning strategy and compare our results to existing baselines. RESULTS: We find that for both ROI-scale and multi-scale PCa detection, image transformer backbones lag behind their CNN counterparts. This deficit in performance is even more noticeable for larger models. When using multi-objective learning, we can improve performance of MIL, with a 77.9% AUROC, a sensitivity of 75.9%, and a specificity of 66.3%. CONCLUSION: Convolutional networks are better suited for modelling sparse datasets of prostate ultrasounds, producing more robust features than transformers in PCa detection. Multi-scale methods remain the best architecture for this task, with multi-objective learning presenting an effective way to improve performance., Comment: early draft, 7 pages; Accepted to SPIE Medical Imaging 2024

Traditionally, segmentation algorithms require dense annotations for training, demanding significant annotation efforts, particularly within the 3D medical imaging field. Scribble-supervised learning emerges as a possible solution to this challenge, promising a reduction in annotation efforts when creating large-scale datasets. Recently, a plethora of methods for optimized learning from scribbles have been proposed, but have so far failed to position scribble annotation as a beneficial alternative. We relate this shortcoming to two major issues: 1) the complex nature of many methods which deeply ties them to the underlying segmentation model, thus preventing a migration to more powerful state-of-the-art models as the field progresses and 2) the lack of a systematic evaluation to validate consistent performance across the broader medical domain, resulting in a lack of trust when applying these methods to new segmentation problems. To address these issues, we propose a comprehensive scribble supervision benchmark consisting of seven datasets covering a diverse set of anatomies and pathologies imaged with varying modalities. We furthermore propose the systematic use of partial losses, i.e. losses that are only computed on annotated voxels. Contrary to most existing methods, these losses can be seamlessly integrated into state-of-the-art segmentation methods, enabling them to learn from scribble annotations while preserving their original loss formulations. Our evaluation using nnU-Net reveals that while most existing methods suffer from a lack of generalization, the proposed approach consistently delivers state-of-the-art performance. Thanks to its simplicity, our approach presents an embarrassingly simple yet effective solution to the challenges of scribble supervision. Source code as well as our extensive scribble benchmarking suite will be made publicly available upon publication.

Content-based image retrieval (CBIR) has the potential to significantly improve diagnostic aid and medical research in radiology. Current CBIR systems face limitations due to their specialization to certain pathologies, limiting their utility. In response, we propose using vision foundation models as powerful and versatile off-the-shelf feature extractors for content-based medical image retrieval. By benchmarking these models on a comprehensive dataset of 1.6 million 2D radiological images spanning four modalities and 161 pathologies, we identify weakly-supervised models as superior, achieving a P@1 of up to 0.594. This performance not only competes with a specialized model but does so without the need for fine-tuning. Our analysis further explores the challenges in retrieving pathological versus anatomical structures, indicating that accurate retrieval of pathological features presents greater difficulty. Despite these challenges, our research underscores the vast potential of foundation models for CBIR in radiology, proposing a shift towards versatile, general-purpose medical image retrieval systems that do not require specific tuning.

At first glance, the ruins of the Roman Theatre in the Italian town of Volterra have little in common with cassette tapes containing Atari games. One is certainly considered an important historical landmark, while the consensus on the importance of the other is partial at best. Still, both are remnants of times vastly different from the present and are at risk of oblivion. Unearthed architectural structures are exposed to the elements just as the deteriorating signals stored on magnetic tapes. However, the rate of deterioration is much faster with the magnetic media, as their life expectancy is counted in decades, whereas the Roman Theater, which is already in ruin, measures its lifespan in centuries. Hence, both would benefit from some form of digital preservation and reconstruction. In this panel, we discuss how to sustainably preserve tangible and intangible cultural artifacts for future generations., Comment: 8 pages, including 1 page of bibliography, 9 figures. Panel summary to be published in proceedings from 11th Machine Intelligence and Digital Interaction MIDI Conference

This work presents an algorithm for determining the parameters of a nonlinear dynamic model of the respiratory system in patients undergoing assisted ventilation. Using the pressure and flow signals measured at the mouth, the model's quadratic pressure-volume characteristic is fit to this data in each respiratory cycle by appropriate estimates of the model parameters. Parameter changes during ventilation can thus also be detected. The algorithm is first refined and assessed using data derived from simulated patients represented through a sigmoidal pressure-volume characteristic with hysteresis. As satisfactory results are achieved with the simulated data, the algorithm is evaluated with real data obtained from actual patients undergoing assisted ventilation. The proposed nonlinear dynamic model and associated parameter estimation algorithm yield closer fits than the static linear models computed by respiratory machines, with only a minor increase in computation. They also provide more information to the physician, such as the pressure-volume (P-V) curvature and the condition of the lung (whether normal, under-inflated, or over-inflated). This information can be used to provide safer ventilation for patients, for instance by ventilating them in the linear region of the respiratory system.

Star formation is essential in galaxy evolution and the cycling of matter. The support of interstellar clouds against gravitational collapse by magnetic (B-) fields has been proposed to explain the low observed star formation efficiency in galaxies and the Milky Way. Despite the Planck satellite providing a 5-15' all-sky map of the B-field geometry in the diffuse interstellar medium, higher spatial resolution observations are required to understand the transition from diffuse gas to gravitationally unstable filaments. NGC 2024, the Flame Nebula, in the nearby Orion B molecular cloud, contains a young, expanding HII region and a dense filament that harbors embedded protostellar objects. Therefore, NGC 2024 is an excellent opportunity to study the role of B-fields in the formation, evolution, and collapse of filaments, as well as the dynamics and effects of young HII regions on the surrounding molecular gas. We combine new 154 and 216 micron dust polarization measurements carried out using the HAWC+ instrument aboard SOFIA with molecular line observations of 12CN(1-0) and HCO+(1-0) from the IRAM 30-meter telescope to determine the B-field geometry and to estimate the plane of the sky magnetic field strength across the NGC 2024. The HAWC+ observations show an ordered B-field geometry in NGC 2024 that follows the morphology of the expanding HII region and the direction of the main filament. The derived plane of the sky B-field strength is moderate, ranging from 30 to 80 micro G. The strongest B-field is found at the northern-west edge of the HII region, characterized by the highest gas densities and molecular line widths. In contrast, the weakest field is found toward the filament in NGC 2024. The B-field has a non-negligible influence on the gas stability at the edges of the expanding HII shell (gas impacted by the stellar feedback) and the filament (site of the current star formation)., Comment: 36 pages, 26 figures Accepted for publication in Astronomy & Astrophysics

Uncertainty estimation is an essential and heavily-studied component for the reliable application of semantic segmentation methods. While various studies exist claiming methodological advances on the one hand, and successful application on the other hand, the field is currently hampered by a gap between theory and practice leaving fundamental questions unanswered: Can data-related and model-related uncertainty really be separated in practice? Which components of an uncertainty method are essential for real-world performance? Which uncertainty method works well for which application? In this work, we link this research gap to a lack of systematic and comprehensive evaluation of uncertainty methods. Specifically, we identify three key pitfalls in current literature and present an evaluation framework that bridges the research gap by providing 1) a controlled environment for studying data ambiguities as well as distribution shifts, 2) systematic ablations of relevant method components, and 3) test-beds for the five predominant uncertainty applications: OoD-detection, active learning, failure detection, calibration, and ambiguity modeling. Empirical results on simulated as well as real-world data demonstrate how the proposed framework is able to answer the predominant questions in the field revealing for instance that 1) separation of uncertainty types works on simulated data but does not necessarily translate to real-world data, 2) aggregation of scores is a crucial but currently neglected component of uncertainty methods, 3) While ensembles are performing most robustly across the different downstream tasks and settings, test-time augmentation often constitutes a light-weight alternative. Code is at: https://github.com/IML-DKFZ/values, Comment: ICLR 2024 (oral)

Only a handful of massive starless core candidates have been discovered so far, but none of them have been fully confirmed. Within the MM1 clump in the filamentary infrared dark cloud G34.43+0.24 that was covered by the ALMA-ATOMS survey at Band 3 ($\sim2\arcsec$, 6000\,au) and the ALMA-QUARKS survey at Band 6 ($\sim 0.3\arcsec$, 900\,au), two prestellar core candidates MM1-C and E1 with masses of 71 and 20 \solarmass~and radii of 2100--4400\,au were discovered. The two cores show no obvious sign of star-formation activities. In particular, MM1-C is a very promising massive prestellar core candidate with a total gas mass of 71\,\solarmass. Within MM1-C, we detected two extremely dense substructures, C1 and C2, as characterized by their high densities of $\rm n_{H_2}\sim 10^{8-9} cm^{-3}$. Moreover, evidence of further fragmentation in C2 was also revealed. We have detected the primordial fragmentation in the earliest stage of massive star formation, and we speculate that MM1-C would be the birthplace of a massive multiple system. However, we cannot fully rule out the possibility that the massive prestellar core MM1-C will just form a cluster of low-mass stars if it undergoes further fragmentation., Comment: 12 pages, 6 figures

Autoregressive and Masked Transformers are incredibly effective as generative models and classifiers. While these models are most prevalent in NLP, they also exhibit strong performance in other domains, such as vision. This work contributes to the exploration of transformer-based models in synthetic data generation for diverse application domains. In this paper, we present TabMT, a novel Masked Transformer design for generating synthetic tabular data. TabMT effectively addresses the unique challenges posed by heterogeneous data fields and is natively able to handle missing data. Our design leverages improved masking techniques to allow for generation and demonstrates state-of-the-art performance from extremely small to extremely large tabular datasets. We evaluate TabMT for privacy-focused applications and find that it is able to generate high quality data with superior privacy tradeoffs., Comment: NeurIPS 2023 (https://openreview.net/forum?id=qs4swxtIAQ)

Carbon monoxide (CO) is a poor tracer of H$_{2}$ in the diffuse interstellar medium (ISM), where most of the carbon is not incorporated into CO molecules unlike the situation at higher extinctions. We present a novel, indirect method to constrain H$_{2}$ column densities ($N_{H_{2}}$) without employing CO observations. We show that previously-recognized nonlinearities in the relation between the extinction, $A_{V} ({H}_{2})$, derived from dust emission and the HI column density ($N_{HI}$) are due to the presence of molecular gas. We employ archival $N_{H_{2}}$ data, obtained from the UV spectra of stars, and calculate $A_{V} ({H}_{2})$ towards these sight lines using 3D extinction maps. We derive an empirical relation between $A_{V} ({H}_{2})$ and $N_{H_{2}}$, which we use to constrain $N_{H_{2}}$ in the diffuse ISM. We construct a $N_{H_{2}}$ map of our Galaxy and compare it to the CO integrated intensity ($W_{CO}$) distribution. We find that the average ratio ($X_{CO}$) between $N_{H_{2}}$ and $W_{CO}$ is approximately equal to $2 \times 10^{20}$ cm$^{-2}$ (K km s$^{-1}$)$^{-1}$, consistent with previous estimates. However, we find that the $X_{CO}$ factor varies by orders of magnitude on arcminute scales between the outer and the central portions of molecular clouds. For regions with $N_{H_{2}} \gtrsim 10^{20}$ cm$^{-2}$, we estimate that the average H$_{2}$ fractional abundance, $f_{{H}_{2}}$ = $2N_{H_{2}}$/(2$N_{H_{2}}$ + $N_{{HI}}$), is 0.25. Multiple (distinct) largely atomic clouds are likely found along high-extinction sightlines ($A_{V} \geq 1$ mag), hence limiting $f_{{H}_{2}}$ in these directions. More than $50 \%$ of the lines of sight with $N_{H_{2}} \geq 10^{20}$ cm$^{-2}$ are untraceable by CO with a $J$ = 1-0 sensitivity limit $W_{CO} = 1$ K km s$^{-1}$., Comment: Accepted for publication in A&A (edited author list)