There exists growing interest in integrating social and ecological sciences to elucidate human-nature interactions from the perspective of social-ecological systems (SES). The compelling logic in the past is that people inevitably harm nature as they use it; however, people have learned to use it while protecting or even improving it. Nowadays though there is a growing emphasis on transformability into a more desirable SES with deliberate human actions rather than adapting to the existing conditions, this theory has not been tested, due to the lack of large-scale and long-term records. The millennia-long records of humans to alter SES in China's Loess Plateau provide a unique means to test ecological theories and heuristic models. Here, we demonstrated human activities can erode ecological resilience or improve it to provide better ecoservices than the pristine. Relative resilience, calculated as a ratio of annual sediment discharge anomaly to the pristine rate, decreased from 0 at the pristine state to -1 in the 1950s due to deforestation and agricultural expansion, whereas increased to about 0.8 at the present by human intervention of environmental slow variables. Building resilience by attending slow variables rather than controlling disturbance should be the goal of adaptive ecosystem management in a SES.
Medical data science education often separates theoretical instruction from practical application, resulting in fragmented learning experiences that fall short of preparing students for real-world data analysis challenges. To better familiarize students with such challenges, we have developed and evaluated an integrated 8 European Credit Transfer and Accumulation System (ECTS) course for master's students in computer science that combines lectures, seminars, and exercises focusing on practical key challenges of health data processing. The course employs a matrix structure (15 intensive care diseases × 4 informatics foci) based on disease-specific projects using the freely available MIMIC-IV Demo dataset. Particular emphasis was placed on teaching technical skills that are essential for processing raw data, including across different industries: groups of students developed complete (E)xtract, (T)ransform, (L)oad (ETL) pipelines in accordance with the medallion architecture (bronze-silver-gold levels) and, in parallel, conducted structured reviews in accordance with PRISMA guidelines. 87 % of enrolled students qualified for examinations. The examination pass rate was 91 %. Students particularly value authentic data challenges and transferable data processing frameworks, such as the medallion architecture. This integrated design successfully bridges theory and practice in medical data science education, providing transferable skills through real-world data and systematic methodology. The freely available dataset enables reproducibility by other institutions.
To determine if a pharmacist-led digital enablement for medication optimization (DEMO) program was non-inferior to standard physician consultation in achieving a blood pressure (BP) target of <140/90mmHg within six months. Secondary outcomes included assessing BP changes and time to exit in DEMO group, system usability scale (SUS), and willingness to pay (WTP) for the DEMO program. A prospective cohort study was conducted in a hospital outpatient setting (Mar 2023 -Oct 2024). The intervention group used a mobile application to transmit home BP readings to a clinical dashboard, with pharmacist-led teleconsultations every 2 - 4 weeks, during which antihypertensives were adjusted. Control group received standard physician consultations. Using 1:4 nearest neighbour propensity score matching, 145 control patients were matched to 56 DEMO participants. The probability of achieving BP <140/90mmHg within six months was 53% (95% CI 44-62) in DEMO vs 39% (95% CI 31-47) in controls. DEMO participants had a 36% higher relative probability of achieving the BP target (RD 0.14, 95% CI 0.01-0.27; P=0.041). Mean change in systolic and diastolic BP in DEMO group was -16.4mmHg (95% CI -21.9 to -10.9; P<0.001) and of -7.7mmHg (95% CI -10.5 to -4.9; P<0.001) respectively. Median time to exit DEMO was 46 days (IQR 20.5-74). Mean SUS was 82.2±13.7, indicating excellent usability. Mean WTP was SGD$13.2 per month (95% CI 0-32.0). The pharmacist-led DEMO program was non-inferior to standard physician consultation in achieving a BP target of 140/90mmHg within six months, with high usability and modest WTP.
The goal of this study is to empirically evaluate the Decoupled Momentum Optimizer (DeMo) in medical image segmentation while demonstrating its extensibility to applications outside LLMs. We aim to characterize the behavior of each parameter group and their adherence to conjectures underlying DeMo's function. DeMo leverages spatial redundancy in gradients through a spatially partitioned frequency decomposition compression algorithm, reducing network traffic and smoothing gradient noise. DeMo provides up to a 150x traffic reduction and 1.6x wall-time speedup on lung segmentation of COPDGene CTs. Analysis of gradients support the conjectures that the primary components of the gradient exhibited higher spatial autocorrelation and lower temporal variance. We find that these conjectures are not uniformly true across all parameters, but rather are predominantly observed in a small subset of them. We also introduce DeMoDropout, a modification to the algorithm that selectively compresses only the largest gradients to significantly reduce computational overhead while maintaining effective overall compression. Using the Beyond the Cranial Vault dataset, we demonstrate potential speed-ups at bandwidths of 1 Gb/s and 100 Mb/s (1.6x vs 1.5x and 6.151 vs 6.31x for DeMoDropout and DeMo, respectively).
Gene set analysis (GSA) is a foundational approach for interpreting genomic data of diseases by linking genes to biological processes. However, conventional GSA methods overlook clinical context of the analyses, often generating long lists of enriched pathways with redundant, nonspecific, or irrelevant results. Interpreting these requires extensive, ad-hoc manual effort, reducing both reliability and reproducibility. We introduce cGSA, a novel AI-driven framework that enhances GSA by incorporating context-aware pathway prioritization. cGSA integrates gene cluster detection, enrichment analysis, and large language models to identify pathways that are not only statistically significant but also biologically meaningful. Benchmarking on 102 curated gene sets across 19 diseases and ten disease-related biological mechanisms shows that cGSA outperforms baseline methods by over 30%, with expert validation confirming its increased precision and interpretability. Two independent case studies in melanoma and breast cancer further demonstrate its potential to uncover context-specific insights and support targeted hypothesis. The demo website is publicly available at https://www.ncbi.nlm.nih.gov/CBBresearch/Lu/Demo/cGSA/, while the data and code can be accessed at https://github.com/ncbi-nlp/cGSA.
To create and evaluate OpenRad ( https://konstvr.github.io/OpenRad/index.html ), a curated, standardized repository that aggregates open-access radiology artificial intelligence (AI) models enriched with metadata from the corresponding code repositories regarding availability of pretrained weights and interactive applications. Retrospective analysis of literature from PubMed, arXiv, and Scopus until 12/2025 (5239 works). After duplicate removal and relevance screening, 1694 articles describing open-access AI models were processed. Model records were generated using a locally hosted large language model (LLM) (gpt-oss:120b), based on the RSNA AI Roadmap JSON schema, and then manually verified by ten expert reviewers. The stability of LLM outputs was assessed on 225 randomly selected papers using text similarity metrics. A statistical analysis of the collected works was also performed. The included 1694 models span all imaging modalities (computed tomography (CT), magnetic resonance imaging (MRI), X-ray, ultrasound (US)) and radiology subspecialties. Automated extraction demonstrated high stability for structured fields (Levenshtein ratio > 90%), with 78.5% of edits, during expert review, being minor corrections. Statistical analysis of the repository revealed convolutional neural network (CNN) and transformer architectures as dominant, while MRI was the most commonly used modality (in 621 neuroradiology AI models). Research output was mostly concentrated in China and the United States. The proposed web interface enables model discovery via keyword search and filters for modality, subspecialty, intended use and demo availability, alongside live statistical dashboards. The community can also contribute new models through a dedicated portal. OpenRad contains ~1700 open-access, curated radiology AI models with standardized metadata, supplemented with analysis of code repositories, thereby creating a comprehensive, searchable resource for the radiology community. Question Current repositories of AI models in radiology are limited, fragmented and include models that are not readily available for use. Findings OpenRad, a curated repository, includes ~1700 open-access, standardized radiology AI models with verified code repositories. Clinical relevance OpenRad enables the radiology community to reliably access AI models with readily available code, weights and demos.
Realistic simulated data is critical for advancing methodological development and optimizing study design in genetics research. However, many genetic simulation tools are unable to replicate the distribution of rare variants or incorporate key genetic information, such as functional annotations and linkage disequilibrium. RAREsim, an accurate rare variant simulation algorithm that uses real genetic haplotypes, was developed to address these limitations. Here, we introduce RAREsim2, an update that provides both streamlined software and new functionalities for simulating individual-level differences (e.g., case-control status, technological or batch effects) and variant-level differences to represent a variety of causal models. We demonstrate RAREsim2's utility with three rare variant association methods (Burden, SKAT, and SKAT-O) across several simulation scenarios, including various genetic ancestries, gene sizes, strengths of association, and proportions of risk variants. Type I Error was maintained and the test with the highest power matched previously known patterns. Importantly, real genetic regions can be simulated to include known variant functions and disease associations. Ultimately, RAREsim2 offers additional flexibility and ease in simulating a multitude of realistic genetic scenarios. The RAREsim2 Python package is publicly available on Github (https://github.com/Hendricks-Research-Team/RAREsim2), PyPI (https://pypi.org/project/raresim/), and Zenodo (https://doi.org/10.5281/zenodo.19442523). Code for the example demonstration can be found at https://github.com/JessMurphy/RAREsim2-demo.
Neural cellular automata (NCA) are a class of cellular automata where the update rule is parameterized by a neural network that can be trained using gradient descent. In this article, we focus on NCA models used for texture synthesis, where the update rule is inspired by partial differential equations (PDEs) describing reaction-diffusion systems. To train the NCA model, the spatiotemporal domain is discretized, and Euler integration is used to numerically simulate the dynamics. Crucially, it is unclear whether a ground-truth PDE even exists for the task, and NCA training only supervises the final steady state without any trajectory supervision, leaving it an open question whether a trained NCA truly learns continuous dynamics or merely overfits the discretization used during training. We study NCA models at the limit where space-time discretization approaches continuity. We find that existing NCA models tend to overfit the training discretization, especially in the proximity of the initial condition, also called a "seed." To address this, we propose a solution that utilizes uniform noise as the initial condition. We demonstrate the effectiveness of our approach in preserving the consistency of NCA dynamics across a wide range of spatiotemporal granularities. We further show that the resulting model is robust to a stochastic updating scheme and modest additive Gaussian noise. Our improved NCA model enables two new test-time interactions by allowing continuous control over the speed of pattern formation and the scale of the synthesized patterns. We demonstrate this new NCA feature in our interactive online demo. Our work reveals that NCA models can learn continuous dynamics and opens new avenues to studying NCA as a class of PDEs and from a dynamical system's perspective.
Fecal microbiota transplantation (FMT) has emerged as a highly effective treatment for recurrent Clostridioides difficile infection and is being actively investigated for numerous other conditions. While multi-omics studies have revealed dynamic changes in microbial communities and host metabolism following FMT, existing approaches are primarily descriptive and lack the ability to model individual patient trajectories or identify early biomarkers of treatment response. Small-sample, multi-omics, longitudinal prediction presents unique computational challenges: high dimensionality ($p \gg n$), multi-omics integration, temporal dynamics, and interpretability. Here, we present Hierarchical Multi-Omics Trajectory Prediction (HMOTP), a purpose-built machine learning framework that addresses these challenges through hierarchical feature construction, multilevel attention mechanisms, and patient-specific trajectory prediction. We evaluated HMOTP on 15 patients with recurrent Clostridioides difficile infection who underwent FMT, with lipidomics and metagenomics profiling at four timepoints spanning 6 months. Notably, naively concatenating multi-omics features degraded Random Forest performance ($93.33\%$ to $87.18\%$ accuracy), whereas HMOTP's hierarchical integration benefited from the additional omics layer, demonstrating that its advantage stems from structure, not from access to more data. Through hierarchical interpretability, HMOTP identified key biomarkers and revealed cross-omics associations between host lipid metabolism and microbial energy pathways, demonstrating utility for longitudinal modeling and biological discovery in FMT response. HMOTP provides a generalizable, principled framework for personalized medicine applications across small-sample multi-omics problems. Source code and a demo dataset are publicly available.
Mendelian randomization (MR), polygenic risk score (PRS), Geno Ontology (GO), and the Kyoto Encyclopedia of Genes and Genomes (KEGG) are powerful bioinformatic analysis tools. However, the analysis of MR, PRS, GO, and KEGG may pose a challenge for novices. This article intends to introduce a program that adeptly guides beginners in implementing these analysis functions, ensuring that even those new to the field can confidently use them. The MPGK program was developed to run on the command line. It conveniently implements the MR, PRS, GO, and KEGG analysis functions by calling well-written R programs. The results of our analyses were validated using genome-wide association study (GWAS) summary data for diabetes and psoriasis, as well as gene sequencing data for diabetes. Three demo analyses using the MPGK program demonstrate the comprehensive capabilities of the MPGK program in conducting advanced bioinformatics analysis. First, the MPGK program revealed a causal relationship between diabetes and psoriasis. Additionally, the PRS analysis generated polygenic risk scores for diabetes, demonstrating the implementation of PRS analysis within the MPGK framework. Furthermore, the GO and KEGG analyses indicated that psoriasis is associated with infection and T helper 17 cells. These findings are consistent with the previous literature. MPGK can be easily used to perform comprehensive analysis, including MR, PRS, GO, and KEGG analyses, by both beginners and researchers.
Large language models (LLMs) demonstrate strong general reasoning and language understanding, yet their performance degrades in domains governed by strict formal rules, precise terminology, and legally binding structure. Tax law exemplifies these challenges, as correct answers require exact statutory citation, structured legal argumentation, and numerical accuracy under rigid grading schemes. We curate SteuerEx, the first open benchmark derived from authentic German university tax-law examinations, processed through a structured statement-level decomposition pipeline. SteuerEx comprises 115 expert-validated examination questions spanning six core tax law domains and multiple academic levels, and employs a statement-level, partial-credit evaluation framework that closely mirrors real examination practice. We further present SteuerLLM, a domain-adapted LLM for German tax law trained on a large-scale synthetic dataset generated from authentic examination material using a controlled retrieval-augmented pipeline. SteuerLLM (28B parameters) consistently outperforms general-purpose instruction-tuned models of comparable size and, in several cases, substantially larger systems, demonstrating that domain-specific data and architectural adaptation are more decisive than parameter count for performance on realistic legal reasoning tasks. All benchmark data, training datasets, model weights, and evaluation code are released openly to support reproducible research in domain-specific legal artificial intelligence. A web-based demo of SteuerLLM is available at https://steuerllm.i5.ai.fau.de .
Phosphodiesterase-4 (PDE4) inhibitors, including apremilast, crisaborole and roflumilast, are widely prescribed for chronic inflammatory diseases. However, sex-specific safety profiles of these agents in routine clinical practice remain poorly characterized. We conducted a retrospective pharmacovigilance study using the US Food and Drug Administration Adverse Event Reporting System (FAERS) from Q1-2004 to Q1-2025. Data cleaning and deduplication were performed following the FDA Guidance for Industry: Pharmacovigilance Practices and Pharmacoepidemiologic Assessment (FDA, 2005). Specifically, deduplication was executed by retaining only the most recent FDA_DT for each unique CASEID, and demographic (DEMO), drug (DRUG), and reaction (REAC) files were merged. Reports listing apremilast, crisaborole, or roflumilast as primary suspect drugs formed three target cohorts. Disproportionality analyses were performed using reporting odds ratio (ROR), proportional reporting ratio (PRR), Bayesian confidence propagation neural network (BCPNN, information component), and multi-item gamma Poisson shrinker (MGPS, empirical Bayes geometric mean). Positive signals required meeting all four algorithmic thresholds. Signals were summarised at Preferred Term (PT) and System Organ Class (SOC) levels, stratified by sex, and visualised with volcano plots, forest plots, and SOC-level heatmaps. Time-to-onset (TTO) was calculated as the interval between treatment start and event onset. A total of 127,516 Apremilast, 7,562 Crisaborole, and 2,142 Roflumilast reports were included. Clear sex-specific disproportionality patterns emerged across the three agents. For Apremilast, females exhibited higher reporting of dizziness, palpitations, and infection-related adverse events, whereas Crisaborole showed a modest male predominance for "product use issue," and Roflumilast demonstrated male-skewed signals, particularly for malignancy- and metabolism-related events. System Organ Class (SOC) analyses further revealed distinct organ-system involvement for each agent. Time-to-onset profiles showed substantially delayed onset for Apremilast compared with the more immediate onset observed for Crisaborole and Roflumilast. The median TTO (IQR) was 24 (7-86) days for Apremilast, 4 (1-15) days for Crisaborole, and 42 (14-128) days for Roflumilast. This large-scale real-world analysis reveals pronounced sex-specific and drug-specific heterogeneity in the safety profiles of PDE4 inhibitors. These findings highlight the need for sex-stratified risk communication, individualized monitoring strategies, and further mechanistic investigations.
Soybean (Glycine max (L.) Merr.) is one of the important legume crops, rich in protein, vegetable oil, maintain soil fertility and used as human food and livestock feed. Despite its importance, adoption and productivity of soybean in South Ethiopia is limited by poor agricultural practices and poor access to inputs like biofertilizers. This work aimed at trapping and screening elite rhizobia strains to use as inoculants in farmers' fields for improved soil health and crop productivity. Forty-four rhizobia strains isolated from different locations in South Ethiopia region were screened along with a commercial inoculant Bradyrhizobium japonicum strain (obtained from Menagesha Biotech Industry, Addis Ababa) in greenhouse using modified Leonard jars (MLJ). Three top performing rhizobia strains SB19, SB22, and SB24 based on the MLJ experiment and the commercial inoculant were further evaluated in farmers' fields at different locations with varying eco-physiological conditions because environmental factors differ from place to place, affecting how living things grow, survive and function. MLJ experiment revealed that the new isolates SB19, SB22, and SB24 had significantly higher relative symbiotic effectiveness (SE%) (p < 0.05) than un-inoculated and N-fertilized control plants. Except SB19, the top performing strains did not differ from the commercial strains regarding SE%. Strain SB19 produced 34 to 61 number of nodules, while the commercial strain induced 22 to 49 nodules. In the first field experiment (2023) inoculation of soybean with SB19 resulted in average grain yields of 3.1 and 2.92 tons ha⁻1 at Arba Minch University (AMU) demo farm and Abaya campus experimental sites to be consistent with, respectively. In experiment 2 (2024), SB19 strain resulted in the average grain yields of 2.39 and 2.45 tons ha-1 at Abaya and AMU demo sites, respectively. Across all locations, the commercial strain produced an average yield of 2.25 to 2.40 tons ha⁻1, which was significantly lower (p < 0.05) than that of the native strains, but higher than the control plants, which yielded 1.87 to 2.02 tons ha⁻1. Among the evaluated strains, SB19 consistently exhibited the most promising performance across all fields and locations compared to the commercial one and others. This finding highlights the presence of highly effective, locally adapted rhizobial strains capable of nodulating soybean in South Ethiopian soils, and demonstrate their promise for selection and improvement into superior inoculant strains to enhance soybean productivity.
Spatial transcriptomics provides a multi-modal perspective by simultaneously capturing gene expression profiles, spatial coordinates, and histological images. While existing methods focus on maintaining view consistency to handle distribution shifts, they frequently neglect semantic conflicts introduced by distorted views-a common limitation arising from technical data acquisition and processing constraints. These conflicts lead to distorted consensus representations. To address this challenge, we propose Holistic Invariant RetrAcing for mitigating representation distortion (HiraST). Our framework explicitly corrects distorted multi-view representations through two complementary mechanisms: 1) Cross-view invariant retracing, which jointly aligns instance-level features and pseudo-label distributions to retrace invariant information. This dual alignment ensures that semantically similar cells or tissue regions remain consistent across heterogeneous modalities, even in the presence of acquisition-induced distortions; and 2) holistic prototype learning, which leverages low-frequency structural components to recalibrate corrupted views and enhance robustness against noise. Extensive experiments on spatial transcriptomics datasets and incomplete multi-view clustering benchmarks demonstrate our framework's state-of-the-art performance. Meanwhile, HiraST demonstrates strong capability across various downstream tasks. The demo code of this work is publicly available at https://github.com/hexiao0275/HiraST.
Relapsed small cell lung cancer (SCLC) is widely considered as a difficult-to-treat disease with an adverse prognosis and scarce therapeutic options, especially in the case of platinum-resistance. Tarlatamab (IMDELLTRA™), a first-in-class, Delta-like ligand-3 (DLL3)-targeted bispecific T-cell engager (BiTE), works by creating a molecular bridge between DLL3 on tumor cells and CD3 on T-cells, leading to T-cell activation and Τ-cell-mediated tumor cell lysis. Tarlatamab demonstrated promising efficacy in early-phase trials at the cost of immune-mediated toxicities like cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS). CRS and ICANS emerge primarily during the first two cycles of treatment, have low to moderate severity and are generally manageable with general supportive measures and specialized immunosuppressive treatment including corticosteroids and monoclonal antibodies like tocilizumab. Tarlatamab appears to be a promising choice for relapsed SCLC, based on the results of the Phase III DeLLphi-304 trial, which demonstrated a clinically and statistically meaningful improvement in overall survival (OS) with its use compared to approved second-line chemotherapy (ChT) options. Having been recently granted FDA approval for use in patients with SCLC who progressed on or after platinum-based ChT, tarlatamab is currently being evaluated in multiple settings of SCLC, including first-line and maintenance treatment. Small cell lung cancer is a very aggressive type of lung cancer. The management of patients whose cancer comes back after the initial treatment is considered as a difficult-to-solve riddle with no significant breakthrough for many years and a few therapeutic options based on chemotherapy. Tarlatamab (IMDELLTRA™) is a new, targeted drug that establishes a connection between the cancer cells and a subgroup of the patient’s immune cells, enabling the latter to localize and destroy the tumor. This drug has been shown to be more effective than standard chemotherapy for patients whose cancer has come back after platinum-based treatment. Specifically, a major study (DeLLphi-304) found that tarlatamab helped patients live longer and kept the disease under control for a longer period compared to chemotherapy. While tarlatamab can cause side effects related to the immune system or to the nervous system due to its mechanism of action, these are usually mild and happen early in the treatment. Based on this study, tarlatamab has recently been approved by the FDA for the treatment of relapsed small cell lung cancer and is under ongoing assessment for other indications like newly diagnosed small cell lung cancer.
Primary open-angle glaucoma (POAG) is a leading cause of irreversible blindness globally. Emerging evidence suggests glaucoma is a vasculopathy with associated systemic microvascular dysfunction. A known feature of endothelial dysfunction is transudation, typified by albuminuria. This aimed to determine the relationship between POAG and albuminuria in patients attending a Nigerian tertiary hospital with a view to enhancing the evaluation and management of POAG patients for blindness prevention, particularly in resource-limited settings. This hospital-based, comparative cross-sectional study included 210 adults: 105 POAG patients and 105 age-and sex-matched non-glaucomatous participants. Data on demographics and medical history were collected using an interviewer-administered questionnaire. Blood samples were analyzed for fasting blood sugar, lipid profile, and estimated glomerular filtration rate (GFR). Spot urine samples were collected to determine albuminuria using the urinary albumin-to-creatinine ratio (UACR, mg/g), based on measurements of urinary albumin and creatinine. A total of 210 patients were enrolled. These included 45 males and 60 females in each study and control group (male-to-female ratio of 1:1.3). The mean age of POAG participants was 43.91 ± 14.8 years. Among POAG participants, 69 (65.7%) had mild, 26 (24.8%) had moderate, and 10 (9.5%) had severe POAG. A significantly higher proportion of albuminuria was observed in the POAG group (8 patients, 7.6%) compared to the non-glaucoma group (3 patients, 2.9%) (p = 0.049). The odds of albuminuria were nearly three times higher in participants with severe POAG compared to those with mild POAG (OR = 2.993; 95% CI, 1.003-7.574; p = 0.048), and participants with POAG were approximately 2.5 times more likely to have albuminuria compared to the non-glaucoma group (OR = 2.483; 95% CI, 1.071-4.552; p = 0.042). Median UACR was significantly higher in the POAG group (7.69 mg/g; interquartile range [6.2-10.3]) compared to the non-glaucoma group (3.83 mg/g; range [1.7-9.3]) (p < 0.001). This study demonstrated a significant association between POAG and albuminuria, with increasing severity of POAG correlating with higher odds of albuminuria. These findings suggest that albuminuria could serve as a systemic marker of disease severity in POAG. Early identification and monitoring of albuminuria in glaucoma patients may improve disease management and outcomes, potentially reducing the incidence of blindness.
Although a large body of literature describes Takotsubo syndrome (TTS) and spontaneous coronary artery dissection (SCAD) as having overlapping clinical features and benign outcome measures, they differ significantly in their pathophysiological mechanisms, demo-graphic profile, and natural history. Herein, we sought to investigate differences in clinical profile and outcomes between patients of these two conditions. Following PRISMA guidelines, we compared TTS and SCAD in adult patients regarding epidemiological, clinical, and prognostic features. A systematic search of PubMed, Embase, and Cochrane Library identified eligible studies, with data extracted and quality assessed using the Newcastle-Ottawa scale. Random effects models were applied for statistical analysis, with heterogeneity evaluated by I2 and sensitivity analysis conducted to ensure robustness. Takotsubo syndrome patients presented more often with dyspnea (46.8% vs. 0.9%; p < 0.001), while SCAD patients displayed typical angina (p < 0.001). In-hospital outcomes were worse for TTS patients, with higher mortality (4.4% vs. 0.8%; RR = 7.41, p = 0.001) and major adverse cardiac events (43.3% vs. 5.2%; RR = 8.35, p < 0.001). At one year, TTS patients had higher all-cause mortality (12.5% vs. 0.8%; p < 0.001) and stroke (2.1% vs. 0.6%; RR = 5.08, p = 0.02). Poorer outcomes are associated with TTS compared to SCAD. SCAD patients demonstrate better prognoses but remain at risk for recurrent ischemic events.
Rendering complex dynamic scenes offers the advantage of observing and understanding the real world. However, existing Dynamic Scene Rendering (DSR) methods remain challenged by suboptimal reconstruction fidelity. These limitations stem from relying on a single, unified deformation model, which struggles to capture complex motions involving multiple sub-motions and abrupt geometric transitions. While temporal decomposition methods could alleviate such shortcomings, they introduce the additional challenge of ignoring motion correlations and increasing storage requirements. To address these issues, we introduce Keyframe-driven Predictable Gaussian Splatting (KPGS)-an efficient framework for high-fidelity complex dynamic scene rendering. First, we present a patch-wise HSV clustering for extracting keyframes. Second, a prediction network based on the Transformer is utilized to calculate the deformable Gaussians at discrete keyframe times via voxelization. Third, we propose an inter-frame deformation network and a mutual supervision between adjacent segments to maintain the temporal continuity. Extensive experiments on our newly built dataset (MotionGS), as well as public benchmarks HyperNeRF and Neu3D, demonstrate that KPGS could achieve a higher average view synthesis performance than SOTA approaches, while maintaining a balance between storage cost and performance. More details of the demo and dataset are available at KPGS Supplementary.
Electroencephalography (EEG) is a non-invasive method that records brain electrical activity from scalp electrodes, offering millisecond temporal resolution but limited spatial detail due to sparse sensor layouts. We present DiBiMa-EEGSR, a bidirectional Mamba-2 diffusion framework for spatio-temporal EEG super-resolution that reconstructs high-resolution signals from standard low-density recordings without additional hardware. The method formulates super-resolution as conditional generative inference and integrates a diffusion process with a bidirectional state-space backbone to model long-range temporal dependencies with linear complexity. Conditioning on low-resolution inputs, electrode positions and task labels enables anatomically coherent and context-aware reconstruction. A one-step sampling strategy substantially reduces inference time while preserving fidelity. Across two public benchmarks, the approach improves reconstruction accuracy, spatial coherence and spectral preservation over convolutional, transformer-based and prior diffusion models in both spatial and temporal upsampling tasks, providing a scalable pathway toward high-resolution electrophysiological imaging. Code to reproduce ablation experiments, training and evaluation of the proposed BiMa and DiBiMa EEGSR models are available at https://github.com/UgoLomoio/DiBiMa-EEGSR.git. Model weights are available at https://huggingface.co/Ugo96/DiBiMa-EEGSR while an interactive demo for EEG spatial super-resolution using our models can be found at https://huggingface.co/spaces/Ugo96/DiBiMa-EEGSR-Demo.
Deep Mutational Scanning (DMS) experiments generate large volumes of sequencing data that must be processed through multi-step computational pipelines to yield interpretable variant scores. At least twelve dedicated tools have been published for this purpose, yet the diversity of experimental designs, scoring strategies, and software implementations has produced a fragmented landscape in which no single tool accommodates the full range of workflows encountered in practice. Here we present CountESS (Count-based Experiment Scoring and Statistics), an open-source pipeline tool that provides a modular, graphical interface for constructing flexible DMS analysis workflows. CountESS supports a wide range of input formats, barcode translation, HGVS variant calling, and user-defined scoring functions, enabling it to accommodate diverse experimental designs including selection assays, time-series experiments, and bin-based assays such as VAMP-seq. Implemented in Python with DuckDB as a computational backend, the software provides high-performance, memory-efficient processing suitable for large datasets. CountESS is freely available at https://github.com/CountESS-Project/CountESS under the 3-Clause BSD Licence. Supplementary data, including demonstration pipelines and example datasets, are available at https://github.com/CountESS-Project/countess-demo.