Conversational generative AI has demonstrated remarkable promise for empowering biomedical practitioners, but current investigations focus on unimodal text. Multimodal conversational AI has seen rapid progress by leveraging billions of image-text pairs from the public web, but such general-domain vision-language models still lack sophistication in understanding and conversing about biomedical images. In this paper, we propose a cost-efficient approach for training a vision-language conversational assistant that can answer open-ended research questions of biomedical images. The key idea is to leverage a large-scale, broad-coverage biomedical figure-caption dataset extracted from PubMed Central, use GPT-4 to self-instruct open-ended instruction-following data from the captions, and then fine-tune a large general-domain vision-language model using a novel curriculum learning method. Specifically, the model first learns to align biomedical vocabulary using the figure-caption pairs as is, then learns to master open-ended conversational semantics using GPT-4 generated instruction-following data, broadly mimicking how a layperson gradually acquires biomedical knowledge. This enables us to
Recent biosignal foundation models (FMs) have demonstrated promising performance across diverse clinical prediction tasks, yet systematic evaluation on long-duration multimodal data remains limited. We introduce SignalMC-MED, a benchmark for evaluating biosignal FMs on synchronized single-lead electrocardiogram (ECG) and photoplethysmogram (PPG) data. Derived from the MC-MED dataset, SignalMC-MED comprises 22,256 visits with 10-minute overlapping ECG and PPG signals, and includes 20 clinically relevant tasks spanning prediction of demographics, emergency department disposition, laboratory value regression, and detection of prior ICD-10 diagnoses. Using this benchmark, we perform a systematic evaluation of representative time-series and biosignal FMs across ECG-only, PPG-only, and ECG + PPG settings. We find that domain-specific biosignal FMs consistently outperform general time-series models, and that multimodal ECG + PPG fusion yields robust improvements over unimodal inputs. Moreover, using the full 10-minute signal consistently outperforms shorter segments, and larger model variants do not reliably outperform smaller ones. Hand-crafted ECG domain features provide a strong baseli
The Bio Image and Signal Processing (BISP) Technical Committee (TC) of the IEEE Signal Processing Society (SPS) promotes activities within the broad technical field of biomedical image and signal processing. Areas of interest include medical and biological imaging, digital pathology, molecular imaging, microscopy, and associated computational imaging, image analysis, and image-guided treatment, alongside physiological signal processing, computational biology, and bioinformatics. BISP has 40 members and covers a wide range of EDICS, including CIS-MI: Medical Imaging, BIO-MIA: Medical Image Analysis, BIO-BI: Biological Imaging, BIO: Biomedical Signal Processing, BIO-BCI: Brain/Human-Computer Interfaces, and BIO-INFR: Bioinformatics. BISP plays a central role in the organization of the IEEE International Symposium on Biomedical Imaging (ISBI) and contributes to the technical sessions at the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), and the IEEE International Conference on Image Processing (ICIP). In this paper, we provide a brief history of the TC, review the technological and methodological contributions its community delivered, and highlight
Fine-grained few-shot entity extraction in the chemical domain faces two unique challenges. First, compared with entity extraction tasks in the general domain, sentences from chemical papers usually contain more entities. Moreover, entity extraction models usually have difficulty extracting entities of long-tailed types. In this paper, we propose Chem-FINESE, a novel sequence-to-sequence (seq2seq) based few-shot entity extraction approach, to address these two challenges. Our Chem-FINESE has two components: a seq2seq entity extractor to extract named entities from the input sentence and a seq2seq self-validation module to reconstruct the original input sentence from extracted entities. Inspired by the fact that a good entity extraction system needs to extract entities faithfully, our new self-validation module leverages entity extraction results to reconstruct the original input sentence. Besides, we design a new contrastive loss to reduce excessive copying during the extraction process. Finally, we release ChemNER+, a new fine-grained chemical entity extraction dataset that is annotated by domain experts with the ChemNER schema. Experiments in few-shot settings with both ChemNER+
We present {\it Hubble Space Telescope} Advanced Camera for Surveys multicolor coronagraphic images of the recently discovered edge-on debris disk around the nearby ($\sim10$ pc) M dwarf AU Microscopii. The disk is seen between $r = $0\farcs 75 -- 15'' (7.5 -- 150 AU) from the star. It has a thin midplane with a projected full-width-at-half-maximum (FWHM) thickness of 2.5 -- 3.5 AU within $r < 50$ AU of the star that increases to 6.5 -- 9 AU at $r \sim 75$ AU. The disk's radial brightness profile is generally flat for $r < 15$ AU, then decreases gradually ($I \propto r^{-1.8}$) out to $r \approx 43$ AU, beyond which it falls rapidly ($I \propto r^{-4.7}$). Within 50 AU the midplane is straight and aligned with the star, and beyond that it deviates by $\sim3^{\circ}$, resulting in a bowed appearance that was also seen in ground-based images. Three-dimensional modelling of the disk shows that the inner region ($r < 50$ AU) is inclined to the line-of-sight by $<1^{\circ}$ and the outer disk by $\sim3^{\circ}$. The inclination of the outer disk and moderate forward scattering ($g \approx 0.4$) can explain the apparent bow. The intrinsic, deprojected FWHM thickness is 1.5 --
Recent results in d+Au and p+Pb collisions at RHIC and the LHC provide evidence for collective expansion and flow of the created medium. We propose a control set of experiments to directly compare particle emission patterns from p+Au, d+Au, and He3+Au or t+Au collisions at the same sqrt(sNN). Using Monte Carlo Glauber we find that a He3 or triton projectile, with a realistic wavefunction description, induces a significant intrinsic triangular shape to the initial medium and that, even with viscous damping, this survives into a significant third order flow moment v3. By comparing systems with one, two, and three initial hot spots, one can disentangle the effects from the initial spatial distribution of the deposited energy and viscous damping. These are key tools to answering the question of how small a droplet of matter is necessary to form a quark-gluon plasma described by nearly inviscid hydrodynamics.
The noise in bias frames for all four readout amplifiers in the Advanced Camera for Surveys (ACS) Wide Field Channel (WFC) is dependent on row number. This is because dark current accumulated during readout increases across the detector, influencing and increasing the read noise as a function of row number. In this report, we investigate bias frames taken with the ACS/WFC to explore the column dependence of read noise for each of the amplifiers for different anneal periods. Analyzing the data, we find that there is no column dependence of read noise and that the read noise values for the physical pre-scans are approximately 0.5 e$^-$ lower than in the science arrays because there is no readout dark accumulated in this area. We further investigate 1) the evolution of read noise over an anneal period, 2) a linear decrease in read noise within the initial columns per amplifier, and 3) pixels in elevated read noise columns. We conclude that 1) there is no visual trend of read noise over an anneal period, 2) amplifiers A and C have an initial linear decrease of read noise in the science arrays, and 3) masking unstable hot pixels in a column will decrease its read noise values.
Bio image analysis has recently become one keystone of biological research but biologists tend to get lost in a plethora of available software and the way to adjust available tools to their own image analysis problem. We present BIII, BioImage Informatic Index (www.biii.eu), the result of the first large community effort to bridge the communities of algorithm and software developers, bioimage analysts and biologists, under the form of a web-based knowledge database crowdsourced by these communities. Software tools (> 1300), image databases for benchmarking (>20) and training materials (>70) for bio image analysis are referenced and curated following standards constructed by the community and then reaching a broader audience. Software tools are organized as full protocol of analysis (workflow), specific brick (component) to construct a workflow, or software platform or library (collection). They are described using Edam Bio Imaging, which is iteratively defined using this website. All entries are exposed following FAIR principles and accessible for other usage.
In their article published in Phys. Med. Biol. 60 (2015) 4149-4168, Lin et al studied the radiosensitizing effect of gold nanoparticles (GNPs) using radiation transport simulations and a biological model for the survival of irradiated cells. This comment points out several caveats to the methodlogy used by Lin et al. that may not be evident to readers and may contribute to confusion in the literature about the radiation effects of gold nanoparticles. The two main caveats are the high mass fraction of gold considered and a potential problem with the modified local effect model used to predict cell survival.
We present a new approach to identify satellite trails (or other linear artifacts) in ACS/WFC imaging data using a modified Radon Transform. We demonstrate that this approach is sensitive to features with mean brightness significantly below the background noise level, and it is resistant to the influence of bright astronomical sources (e.g., stars, galaxies) in most cases. Comparing with a set of satellite trails identified by eye, we find a trail recovery rate of 85\% and a false detection rate (after removing diffraction spikes that are easily filtered) of 2.5\%. By performing an analysis using a much larger ACS/WFC data set where false trails are identified by their persistence across multiple images of the same field, we identify the Radon Transform parameter space and image properties where our algorithm is unreliable, and estimate a false detection rate of $\sim10\%$ elsewhere. We apply our method to ACS/WFC data taken between 2002 and 2022 to determine both the frequency of satellite trail contamination in science data and also the typical trail brightness as a function of time. We find the rate of satellite trail contamination has increased by approximately a factor of two
We present a dedicated study of CCD serial ($x$-direction) charge transfer efficiency (CTE) in ACS/WFC. Following past studies of parallel ($y$-direction) CTE, we use the serial CTE trails behind hot pixels in calibration dark frames to characterize charge trapping and release in the serial registers of the WFC detectors. Serial CTE trails are sharper and longer than parallel CTE trails. Many fewer charge traps come into play during serial pixel transfers than parallel transfers, which explains why parallel CTE is much worse than serial CTE. We find that serial CTE can cause losses of $\sim$0.005-0.02~mag in stellar photometry and shift stellar centroids by $\sim$0.01-0.035 pixels. The pixel-based algorithm in CALACS that corrects for parallel CTE losses in WFC data has been modified to include a correction for serial CTE losses. The PCTETAB reference file has also been updated to include serial CTE parameters. The pixel-based correction for serial CTE currently runs only on full-frame WFC images obtained after SM4 (May 2009). Shortly following the publication of this report, science data corrected for both parallel and serial CTE will be available in the MAST archive.
Using repeat imaging of a galaxy cluster taken over a seventeen-year baseline, we examine the impact that degraded Charge Transfer Efficiency (CTE) has on photometric measurements of extended sources using the ACS/WFC on HST. We examine how measured brightnesses depend on time since ACS installation, source location on the WFC detectors, source brightness, and local background level in individual exposures. We find that global brightness measurements using large apertures are generally reliable within $\sim$0.05 magnitudes across the WFC detectors if exposure backgrounds are above $20e^-/{pixel}$ and sources are brighter than $\sim300e^-$ in a single exposure. However, brightness measurements on smaller scales can suffer deficiencies in excess of 0.1 mags (sometimes, significantly more) in recent data unless sources are very close to the CCD serial registers ($\lesssim 512$ pixels), or brighter than $\sim3000\,e^-$ in a single exposure. We also show how degraded CTE can result in artificial asymmetries in galaxy light distributions, which are largely mitigated if backgrounds are $>20e^-/{pixel}$ and targets are not far ($>1536$ pixels) from the serial registers. As expected,
In their article published in Phys. Med. Biol. 59 (2014) 7675-7689, Lin et al studied the dose enhancement of of gold nanoparticles (GNPs) for proton therapy, which they compared with the case of photon irradiation. This comment points out two caveats to the methodlogy used by Lin et al that may not be evident to readers and may contribute to confusion in the literature about the dose enhancement by gold nanoparticles.
Recent analysis of 23 years of Hubble Space Telescope ACS/SBC data has shown that background levels can vary considerably between observations, with most filters showing over an order of magnitude variation. For the shorter-wavelength filters, the background is understood to be dominated by airglow; however, what precisely drives background variations is not well constrained for any filter. Here, we explore the causes of the background variation. Using over 8,000 archival SBC observations, we developed a machine learning model that can accurately predict the background for an observation, based upon a set of 23 observational parameters. This model indicates that, depending on filter, the SBC background is generally dominated by Solar elevation, Solar separation angle, Earth limb angle of observation, SBC temperature, and target Galactic latitude.
The representation of arbitrary data in a biological system is one of the most elusive elements of biological information processing. The often logarithmic nature of information in amplitude and frequency presented to biosystems prevents simple encapsulation of the information contained in the input. Criticality Analysis (CA) is a bio-inspired method of information representation within a controlled self-organised critical system that allows scale-free representation. This is based on the concept of a reservoir of dynamic behaviour in which self-similar data will create dynamic nonlinear representations. This unique projection of data preserves the similarity of data within a multidimensional neighbourhood. The input can be reduced dimensionally to a projection output that retains the features of the overall data, yet has much simpler dynamic response. The method depends only on the rate control of chaos applied to the underlying controlled models, that allows the encoding of arbitrary data, and promises optimal encoding of data given biological relevant networks of oscillators. The CA method allows for a biologically relevant encoding mechanism of arbitrary input to biosystems, cr
In 2024, due to some operational changes, the Hubble Space Telescope began exhibiting undeclared loss of lock events. This loss of lock can result in the smearing out of light from the target field during an exposure, which leads to data degradation, which in turn may require data to be retaken. In this work, we investigate this `roll-drift' effect in ACS/WFC images. We quantify the impact of roll-drift on measurable parameters in data by using simulations and existing data reduction techniques. We identify a threshold of one such measurable parameter beyond which data may likely be affected by roll-drift, so that users can quickly and easily assess whether their data needs further attention.
Maximally embedding dimension (MED) numerical semigroups are a wide and interesting family, with some remarkable algebraic and combinatorial properties. Associated to any numerical semigroup one can construct a MED closure, as it is well known. This paper shows two different explicit methods to construct this closure which also sheds new light on the very nature of this object.
We examined the long-term behavior of the superbias calibration frames for the Advanced Camera for Surveys Wide Field Channel (ACS/WFC) aboard the Hubble Space Telescope (HST). Superbias frames are used to remove detector-level bias structure from science images and are currently generated after an anneal and delivered monthly. The primary goal of this study was to determine whether the frequency of superbias generation could be reduced without compromising calibration quality, potentially aligning with the Wide Field Camera 3 UVIS (WFC3/UVIS) approach of generating only one superbias per year. We analyzed superbias frames produced from 2007 through 2024 to investigate whether these calibration products have changed significantly over time, and whether the frequency of superbias generation and delivery could be safely reduced without loss of calibration accuracy. In addition to visual inspections and pixel-level comparisons, we employed Principal Component Analysis (PCA) to evaluate whether any long-term, global structure exists beneath the apparent noise in these frames. Our findings show that the superbias structure has remained fairly stable post-Servicing Mission 4 (SM4), a 15-
Recently, the ACS team applied an Ubercal framework to assess the photometric repeatability of stars observed across the WFC detector using 15 years of post-SM4 calibration data in the globular cluster 47 Tuc (Ryan et al., 2024). A surprising finding was an apparent 0.05 mag global difference in sensitivity between the WFC1 and WFC2 chips, which had not been seen in prior tests of sensitivity variations around the field-of-view. Given the many degenerate variables within the Ubercal framework such as CTE losses, time-dependent sensitivity, and flat-field corrections, we obtained new calibration data to perform a straightforward test of the reported $\sim$5$\%$ flux offset between detectors. We observed three white dwarf standards with three filters at four positions on the detector (each on a different amplifier), but with the same number of x and y pixel transfers to mitigate differential CTE-related effects. For the F606W and F814W filters, the agreements are good to 0.4$\%$ on average, and always 1$\%$ or better in individual cases. The consistency of these two filters over all three stars and the four dither positions provides very strong evidence against the large global sensi
Here we develop an approach to bio-structural robustness integrated with structure-function relationship in a unified conceptual and methodological framework, and envision its study using adequate computational and experimental methods. To distinguish this structural robustness from the abstract organizational robustness of systems, we call it anapedesis, and define it as the scale-independent property of biological objects, from biomolecules to organisms, to deform and recover while minimizing and/or repairing the damage produced by stretch. We propose to study the consequences of deformation of biological objects closer to their structural and/or functional failure than previously considered relevant. We show that structural robustness is present as a basic principle in many facets of biomedicine: many pathological conditions may derive from the failure of molecules, cells and their higher-order assemblies to maintain robustness against deformation. Furthermore, structural robustness could have been the key selective criterion during pre-biotic evolution and afterwards, and its universality can be demonstrated by modeling using genetic algorithms. Thus, the specific investigation