Manual annotation of spike-wave discharges (SWDs), the electrographic hallmark of absence seizures, is labor-intensive for long-term electroencephalography (EEG) monitoring studies. While machine learning approaches show promise for automated detection, they often struggle with cross-subject generalization due to high inter-individual variability in seizure morphology and signal characteristics. In this study we compare the performance of 16 machine learning classifiers on our own manually annotated dataset of 961 hours of EEG recordings from C3H/HeJ mice, including 22,637 labeled SWDs, and find that a 1D U-Net performs best. We then improve its performance by employing residual connections and data augmentation strategies combining amplitude scaling, Gaussian noise injection, and signal inversion to enhance cross-subject generalization. Our proposed model, AugUNet1D, achieves an average F1-score of 0.90 with balanced precision (0.91) and recall (0.90), representing a 29% relative improvement over the “Twin Peaks” algorithmic baseline and exceptional cross-subject performance. AugUNet1D, pretrained on our manually annotated data, along with the dataset itself, is made public for other users.

To analyze the impact of telemedicine on emergency department (ED) utilization among University of Virginia (UVA) Health System patients, examining which patient characteristics predict reduced ED usage and whether telemedicine reduces ED utilization. We used UVA Electronic Health Records and public datasets to establish clinical and contextual features including demographics, comorbidities, insurance status, and community characteristics. UVA patient data were linked to Virginia Health Information (VHI) data at the individual level, ensuring our utilization measure included ED encounters across all Virginia health systems. We evaluated: (1) patient characteristics associated with reduced ED usage following the first telemedicine encounter using an XGBoost model and (2) associations between telemedicine and ED usage using fixed effects modeling. Younger, healthier patients with high prior ED usage experienced the greatest reduction in ED visits following their first telemedicine visit. Telemedicine was significantly associated with reduced ED utilization across all observation windows (3, 6, and 12 months), with effects attenuating over longer windows. Data pipelines and models were designed to support rapid iteration on varying feature sets and sub-populations and to enable longitudinal model retraining and evaluation. These findings suggest telemedicine reduces ED utilization with significant reductions observed in specific sub-populations in our cohort. Telemedicine engagement is associated with meaningful reductions in ED utilization, particularly among younger, healthier high-utilizer patients. Data science tools can help providers and policymakers optimize telemedicine delivery to benefit patients while reducing health system burden.

Recent research on Vision Language Models (VLMs) suggests that they rely on inherent biases learned during training to respond to questions about visual properties of an image. These biases are exacerbated when VLMs are asked highly specific questions that require focusing on specific areas of the image. For example, a VLM tasked with counting stars on a modified American flag (e.g., with more than 50 stars) will often disregard the visual evidence and fail to answer accurately. We build upon this research and develop a multi-dimensional examination framework to systematically determine which characteristics of the input data, including both the image and the accompanying prompt, lead to such differences in performance. Using open-source VLMs, we further examine how attention values fluctuate with varying input parameters (e.g., image size, number of objects in the image, background color, prompt specificity). This research aims to learn how the behavior of vision language models changes and to explore methods for characterizing such changes. Our results suggest, among other things, that even minor modifications in image characteristics and prompt specificity can lead to large changes in how a VLM formulates its answer and, subsequently, its overall performance.

Fissurella latimarginata and Fissurella cumingi are two sympatric species found along the southwestern coast of South America. We conducted a comparative analysis of oocyte size and their coats between these species and investigated early development using chemically activated gametes. Additionally, heterologous fertilizations were conducted to assess the presence of a reproductive isolation barrier between the species. The mean diameter of the oocyte and gelatinous coat did not show significant differences between species, but there were differences regarding the vitelline coat. However, the practical utility of this trait as taxonomic character for species discrimination is hindered by variance overlap. Chemical activation of gametes enabled homologous fertilizations, resulting in the production of healthy and viable trochophore larvae. Early development, from fertilisation to hatch of these larvae, takes 28.5-45.5 h in both species. Heterologous fertilizations were also successful, resulting in the hatching of trochophore hybrid larvae, indicating effective interactions between the gametes of both species and viable postzygotic larval development. However, further research should focus on determining if there is asymmetry in gametic compatibility, obtaining veliger larvae, or even more advanced ontogenetic stages.

Recent research suggests that Vision Language Models (VLMs) often rely on inherent biases learned during training when responding to queries about visual properties of images. These biases are exacerbated when VLMs are asked highly specific questions that require them to focus on particular areas of the image in tasks such as counting. We build upon this research by developing a synthetic benchmark dataset and evaluation framework to systematically determine how counting performance varies as image and prompt properties change. Using open-source VLMs, we then analyze how attention allocation fluctuates with varying input parameters (e.g. number of objects in the image, objects color, background color, objects texture, background texture, and prompt specificity). We further implement attention-based interventions to modulate focus on visual tokens at different layers and evaluate their impact on counting performance across a range of visual conditions. Our experiments reveal that while VLM counting performance remains challenging, especially under high visual or linguistic complexity, certain attention interventions can lead to modest gains in counting performance.

Introduction Surgical mortality is the third leading cause of death globally, with mortality rates in Africa double those of high-income countries despite patients being younger and undergoing lower-risk procedures. One of the contributors to poor outcomes in low- and middle-income countries (LMICs) is the lack of digital data, which is essential for quality improvement, provider audit and feedback, and early warning track-and-trigger systems. Due to limited financial resources, paper health records remain the standard in LMICs, making readily accessible digital data an urgent priority. This study builds on our previous work in computer vision-based digitization of smartphone-captured anesthesia records by developing a standardized, computer vision-ready anesthesia paper record. Designed for optimal digitization, this record will align with the Minimum Dataset for Surgical Patients in Africa guidelines. Methods The standardized, computer vision-ready anesthesia paper chart was developed with input from anesthesia experts in LMICs and data scientists. Key adaptations designed to facilitate accurate computer vision digitization included replacing traditional free-text entries with predefined categorical checkboxes, pre-printing the name of commonly used medications, supplementing handwritten medication names with numeric codes, and structuring input fields to improve computer vision digitization accuracy.Prior computer vision software was further iterated to improve digitization accuracy for the new standardized chart. Performance of the updated software running on the new computer vision-ready paper chart was then evaluated by comparing the software output to the human-annotated ground-truth data measuring both detection and interpretation accuracy. Results The training dataset consisted of thirty-three standardized, computer vision-ready anesthesia paper charts completed using synthetic data by a group of ten anesthesia providers. Five charts were reserved for validation, while the test dataset consisted of nine charts that were not used for any training or validation purposes. Updated computer vision software demonstrated high detection accuracy for vital signs: systolic blood pressure (93%), diastolic blood pressure (94%), and heart rate (93%), all physiological indicators (100%), and checkboxes (99%). The mean average error for inferring values from model detections were low: systolic (1.98mmHg), diastolic (1.13mmHg), heart rate (3.8/bpm), oxygen saturation (0.19%), end tidal carbon dioxide (0.65 mmHg), inspired oxygen concentration (2.48%). The accuracy for determining which checkboxes were marked vs. unmarked was 99%. Conclusion This study confirms the feasibility and accuracy of a standardized, computer vision-ready anesthesia chart that can be deployed in LMICs to facilitate digital data access. ### Competing Interest Statement The authors have declared no competing interest. ### Clinical Protocols ### Funding Statement This work is funded in part from a grant from the Center for Global Inquiry and Innovation, University of Virginia. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes All software developed for this study is publicly available on our GitHub repository for reproducibility and further validation

This research analyzed the sample entropy (SampEn) of breath-by-breath cardiopulmonary exercise testing (CPET) data from 170 healthy pediatric participants (85 males) 8 to 18-years-old, using a Bayesian statistics approach. SampEn measures the complexity of time series data, providing quantitative insight into the predictability of breathing patterns in pediatric participants. To address non-stationarity, signals were differenced prior to SampEn calculation. In addition to sex and age group comparisons, we examined SampEn before and after the midpoint of each participant’s CPET to assess how SampEn changes as exercise intensity increases. We corroborated previous findings that SampEn decreases in the later half of CPET for healthy pediatric participants for oxygen uptake (V̇O2), carbon dioxide output (V̇CO2), ventilation (V̇E), and heart rate (HR). Females tended to have higher SampEn than their male counterparts, with a statistically significant difference between the sexes in older participants for V̇O2, V̇CO2, V̇E, HR, and respiratory rate (RR). Age-related findings included: significantly higher SampEn in younger males compared to older males for V̇O2 and V̇E and older female participants had a higher SampEn in older females compared to younger females for HR. These findings support SampEn as a sensitive, non-invasive marker of physiological transition during pediatric CPET, with potential applications in exercise physiology research and clinical assessment.

Quantifying the complexity and irregularity of time series data is a primary pursuit across various data-scientific disciplines. Sample entropy (SampEn) is a widely adopted metric for this purpose, but its reliability is sensitive to the choice of its hyperparameters, the embedding dimension (m) and the similarity radius (r), especially for short-duration signals. This paper presents a novel methodology that addresses this challenge. We introduce a Bayesian optimization framework, integrated with a bootstrap-based variance estimator tailored for short signals, to simultaneously and optimally select the values of m and r for reliable SampEn estimation. Through validation on synthetic signal experiments, our approach outperformed existing benchmarks. It achieved a 60 to 90 and a 22 to 45 itself (p ≤ 0.043). Applying our method to publicly available short-signal benchmarks yielded promising results. Unlike existing competitors, our approach was the only one to successfully identify known entropy differences across all signal sets (p ≤ 0.042). Additionally, we introduce "EristroPy," an open-source Python package that implements our proposed optimization framework for SampEn hyperparameter selection. This work holds potential for applications where accurate estimation of entropy from short-duration signals is paramount.

Domain adaptive semantic segmentation is the task of generating precise and dense predictions for an unlabeled target domain using a model trained on a labeled source domain. While significant efforts have been devoted to improving unsupervised domain adaptation for this task, it is crucial to note that many models rely on a strong assumption that the source data is entirely and accurately labeled, while the target data is unlabeled. In real-world scenarios, however, we often encounter partially or noisy labeled data in source and target domains, referred to as Generalized Domain Adaptation (GDA). In such cases, we suggest leveraging weak or unlabeled data from both domains to narrow the gap between them, resulting in effective adaptation. We introduce the Generalized Gaussian-mixture-based (GenGMM) domain adaptation model, which harnesses the underlying data distribution in both domains to refine noisy weak and pseudo labels. The experiments demonstrate the effectiveness of our approach.

We introduce a novel hierarchical Bayesian permutation entropy (PermEn) estimator designed to improve biomedical time series entropy assessments, especially for short signals. Unlike existing methods requiring a substantial number of observations or which impose restrictive priors, our non-centered, Wasserstein optimized hierarchical approach enables efficient MCMC inference and a broader range of PermEn priors. Evaluations on synthetic and secondary benchmark data demonstrate superior performance over the current state-of-the-art, including 13.33-63.67% lower estimation error, 8.16-47.77% lower posterior variance, and 47-60.83% lower prior construction error (p <= 2.42 x 10(-10)). Applied to cardiopulmonary exercise test oxygen uptake signals, we reveal a previously unreported 1.55% (95% credible interval: [0.62%, 2.52%]) entropy difference between obese and lean subjects that diminishes as exercise capacity increases. For individuals capable of completing at least 7.5 minutes of testing, the 95% credible interval contained zero, suggesting potential insights into physiological complexity, exercise tolerance, and obesity. Our estimator refines biomedical signal PermEn estimation and underscores entropy's potential value as a health biomarker, opening avenues for further physiological and biomedical exploration.

Medical disease diagnosis relies on effective labeling and segmentation of cells in biopsy slides. Current deep learning approaches have shown strong results in addressing many of the traditional issues with medical image segmentation. However, none of these approaches address a medical problem by using observation-specific training data to drive further research. A new method named Extremity-Ranked Domain Selection is created by creating an extremity metric, ranking patients based on their extremity, and evaluating a multisource domain adversarial network (MDAN) approach using a full factorial design of experiments. ERDS yields the optimal values in the full factorial experiment of a domain size of 15, domain choice of Extremity Ranked, and a classification threshold of 0.7. Additionally, ERDS provides comparative performance in relation to a standard Monte Carlo Dropout UNet model with multiple factor combinations outperforming (Patients E-139: 0.65, E-247: 0.627, E-147: 0.624, E-92: 0.623) this baseline model (0.571). This work illustrates the impact of extremity metric in future works and drives further research into effectively modifying training data to optimize model performance.

Deep learning for histopathology has been successfully used for disease classification, image segmentation and more. However, combining image and text modalities using current state-of-the-art (SOTA) methods has been a challenge due to the high resolution of histopathology images. Automatic report generation for histopathology images is one such challenge. In this work, we show that using an existing pre-trained Vision Transformer (ViT) to encode 4096x4096 sized patches of the Whole Slide Image (WSI) and a pre-trained Bidirectional Encoder Representations from Transformers (BERT) model for language modeling-based decoder for report generation, we can build a performant and portable report generation mechanism that takes into account the whole high resolution image. Our method allows us to not only generate and evaluate captions that describe the image, but also helps us classify the image into tissue types and the gender of the patient as well. Our best performing model achieves a 89.52% accuracy in Tissue Type classification with a BLEU-4 score of 0.12 in our caption generation task.

Understanding space-use dynamics of wildlife populations is important for informing habitat management and restoration initiatives. In West Virginia, Clemmys guttata (Spotted Turtle) is restricted to the eastern panhandle region and is considered a species of greatest conservation need. The West Virginia Division of Natural Resources is interested in managing and restoring Spotted Turtle habitat, but information on space use is limited. To address this information gap, we used radiotelemetry to track space-use patterns of 9 Spotted Turtles at 2 sites that represent different wetland types in the state from spring 2018 to fall 2020. One site was a wetland complex containing ∼13.19 ha of potential habitat, and the other site was a single isolated wetland containing ∼2.09 ha of potential habitat. Spotted Turtle home-range size and use of potential habitat varied between the sites, with larger home-range sizes and a larger proportion of habitat used at the wetland complex site. Our results represent the first home-range size estimates reported for Spotted Turtle populations in West Virginia and suggest that space-use dynamics may be variable and dependent on site-level wetland characteristics.

Wisconsin encompasses a substantial portion of the Upper Midwest distribution for the globally endangered Glyptemys insculpta (Wood Turtle). However, the paucity of prior research and population monitoring statewide has limited our understanding of their status and population trends in the state. To address this information gap, we conducted standardized Wood Turtle population surveys at 50 sites across 8 HUC-8 watersheds in Wisconsin from 2018 to 2023 to estimate adult abundance and population demographic parameters. We captured turtles at 29 of 50 sites, and detected 250 unique individuals, consisting of 107 adult females, 77 adult males, and 66 juveniles. Site-level estimated adult abundances varied from 0 to 23 (mean = 5 among all sites surveyed). Our results provide a foundation for assessing long-term population trends and responses to conservation and habitat management efforts for Wood Turtles in Wisconsin.

AbstractStandard cardiopulmonary exercise testing (CPET) produces a rich dataset but its current analysis is often limited to a few derived variables such as maximal or peak oxygen uptake (V̇O2). We tested whether breath‐by‐breath CPET data could be used to determine sample entropy (SampEn) in 81 healthy children and adolescents (age 7–18 years old, equal sex distribution). To overcome challenges of the relatively small time‐series CPET data size and its nonstationarity, we developed a Python algorithm for short‐duration physiological signals. Comparing pre‐ and post‐ventilatory threshold (VT1) CPET phases, we found: (1) SampEn decreased by 9.46% for V̇O2 and 5.01% for V̇CO2 (p < 0.05), in the younger, early‐pubertal participants; and (2) HR SampEn fell substantially by 70.8% in the younger and 77.5% in the older participants (p < 0.001). Across all ages, females exhibited greater HR SampEn than males during both pre‐ and post VT1 CPET phases by 14.10% and 23.79%, respectively, p < 0.01. In females, late‐pubertal had 17.6% lower HR SampEn compared to early‐pubertal participants (p < 0.05). Breath‐by‐breath gas exchange and HR data from CPET are amenable to SampEn analysis that leads to novel insight into physiological responses to work intensity, and sex and maturational effects.

Data from the single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq) are now widely available. One major computational challenge is dealing with high dimensionality and inherent sparsity, which is typically addressed by producing lower dimensional representations of single cells for downstream clustering tasks. Current approaches produce such individual cell embeddings directly through a one-step learning process. Here, we propose an alternative approach by building embedding models pre-trained on reference data. We argue that this provides a more flexible analysis workflow that also has computational performance advantages through transfer learning. We implemented our approach in scEmbed, an unsupervised machine-learning framework that learns low-dimensional embeddings of genomic regulatory regions to represent and analyze scATAC-seq data. scEmbed performs well in terms of clustering ability and has the key advantage of learning patterns of region co-occurrence that can be transferred to other, unseen datasets. Moreover, models pre-trained on reference data can be exploited to build fast and accurate cell-type annotation systems without the need for other data modalities. scEmbed is implemented in Python and it is available to download from GitHub. We also make our pre-trained models available on huggingface for public use. scEmbed is open source and available at https://github.com/databio/geniml. Pre-trained models from this work can be obtained on huggingface: https://huggingface.co/databio.