AI News Archive: July 8, 2026 — Part 19
Sourced from 500+ daily AI sources, scored by relevance.
- Learning social norms enhances compatibility in dynamic human-AI coordination
Humans continuously coordinate with others in dynamic interactions, often through implicit, hard-to-quantify social norms that act as shared tacit expectations among interacting agents. As AI agents, including large language models (LLMs), become embedded in daily life, they increasingly participate...
- Multimodal Smart Glove for Sign Language Recognition Using Deep Learning
Sign language recognition technologies can improve communication between deaf individuals and the broader community, but many existing systems face challenges in real-world deployment. This paper presents a deployable smart glove system for sign language recognition that integrates wearable sensing ...
- Two-player Alternate Uses Test: A Controlled Testbed for Interactive Human-AI and Human-Human Co-Creation
Controlled research on AI ideation typically compares independent agents, while field studies of human-AI collaboration sacrifice experimental control. We introduce a controlled, two-player extension of the Alternate Uses Test (AUT) that enables comparison of human-human and human-AI co-creation und...
- ShapeTalk: Combining Natural Language and Sketch for Time-Series Pattern Querying
Searching for time-series segments that match user-defined patterns is important in domains such as finance, climate science, and healthcare. However, existing visual query tools often struggle to support vague, composite, or fuzzy pattern descriptions, often requiring users to express their intent ...
- Unlearning to Protect: A Distilled Reinforcement Learning Framework with Privacy-Preserving Feature Unlearning and XAI for IoT Security
Botnets pose a significant cybersecurity threat, enabling attacks such as DDoS, data theft, and service disruptions on IoT devices. These devices often lack built-in botnet traffic filtering, leaving them highly exposed. Existing AI-based solutions improve detection capabilities but have limitations...
- Mitigating Taint-Style Vulnerabilities in MCP Servers via Security-Aware Tool Descriptions
Large language models (LLMs) are increasingly deployed as autonomous agents that interact with external tools and services via the Model Context Protocol (MCP), a standardized interface for dynamic tool invocation. While MCP simplifies integration, it also expands the attack surface and enables gene...
- Beware of Agentic Botnets: Scalable Untargeted Promptware Attacks via Universal and Transferable Adversarial HalluSquatting
The growing adoption of agentic LLM applications has introduced a new threat previously named as promptware. While prior work has established that adversaries can exploit direct channels to LLM applications to apply promptware under weak threat models, many applications do not provide any direct cha...
- The AI Resilience Gap: Bringing Artificial Intelligence Inside the Operational Resilience Perimeter
The rapid adoption of artificial intelligence across regulated firms has produced an extensive governance response oriented around trustworthiness: the EU AI Act, ISO IEC 42001, the NIST AI Risk Management Framework, and the United Kingdom's principles-based approach all address safety, fairness, tr...
- Thinking More, Harnessing Better: State Machine Guided Harness Automatic Generation with Project Digestion and Workflow Decomposition
High-quality fuzz harnesses are essential for effective gray-box fuzzing. While Large Language Models (LLMs) offer promise for automating this task, existing one-turn generation methods suffer from hallucinations and inadequate coverage due to coarse-grained function targeting and misaligned generat...
- SA-DRL: Security-Aware Deep Reinforcement Learning for Ransomware Detection with Asymmetric Reward Design
Ransomware detection is a security-critical task in which false negatives and false positives have unequal operational consequences. Conventional machine learning detectors often use symmetric objectives that penalize missed ransomware detections and benign false alarms equally, although a false neg...
- Continual Learning With Participation Privacy: An Auditable Buffering-Aggregation Recipe
Modern federated and streaming learning systems often release intermediate models, so privacy must hold for the full trajectory under adaptive interaction. Motivated by participation privacy, we study single-edit neighboring user streams, where one insertion/deletion shifts all subsequent updates an...
- Certifying Ghosts: How Cybersecurity AI Agents Break the EU Cyber Resilience Act
The EU Cyber Resilience Act (CRA) makes a smart bet. It does not demand that products be free of vulnerabilities, but only that manufacturers run a process: assess risk, handle flaws, ship updates. The bet pays off if four things about the world stay true: (P1) finding vulnerabilities is slow, skill...
- Cyber Dynamics I: Finite Macrostates for Behavioral Anomaly Detection in Network Telemetry
Entropy-based methods have long been used for network anomaly detection, but most existing approaches treat entropy as a scalar statistic on narrow observables rather than as part of a broader behavioral state-space for cyber systems. We propose a finite-dimensional macrostate framework for network ...
- Rethinking Code Performance Benchmarks for LLMs
Many function-level performance benchmarks have been proposed to evaluate whether large language models (LLMs) can generate efficient programs. However, results on these benchmarks often show that LLM-generated implementations have little or no execution-time difference from canonical solutions. In ...
- The Poisoned Chalice of LLM Evaluation Report
Large language models are increasingly used to evaluate and support software engineering tasks, yet the validity of these evaluations is often undermined by uncertainty about whether benchmark instances were seen during pretraining. This can lead to data contamination, which may inflate performance ...
- ORCAID: Oblique Rule-Based Continuous-Action Interpretation for Deep RL Policies
Explainability remains a key issue in reinforcement learning (RL). Distilling an interpretable policy from an agent trained in a complex environment is particularly challenging when the action space is continuous. We introduce ORCAID, a novel method for extracting interpretable rule-based policies f...
- Gimitest: A Comprehensive Tool for Testing Reinforcement Learning Policies
Reinforcement learning (RL) policies can be unsafe and vulnerable to attacks. Ensuring their reliability is often a pain point as existing automated testing methods target only selected environments, testing scenarios, and RL algorithms. To address this, we propose a comprehensive framework for test...
- Mining Workflow Graphs for Black-Box Boundary Testing of Conversational LLM Agents
Conversational LLM agents can cause real-world harm when their internal workflows fail, such as completing a transaction without confirmation. Testing these state-dependent failures is difficult because critical boundaries, such as identity checks and confirmation gates, are hidden behind multi-turn...
- Quantum Software Engineering in Practice: FPGA and AI Integration for Quantum Certification
The emergence of Quantum Software Engineering (QSE) responds to the need for systematic, disciplined, and quantifiable approaches to the development, operation, and maintenance of quantum software. Within this context, quantum computer certification represents a significant challenge: verifying that...
- What Makes a Good Bug Report for an AI Agent?
Automated program repair (APR) agents are transitioning from research benchmarks to developer workflows, yet they still begin with bug reports written for human developers. While decades of research have established what makes a good bug report for humans (e.g., steps to reproduce, stack traces), it...
- Biased or Personalized? The Impact of Personal Information on AI-driven Development
Generative AI is increasingly permeating software engineering, enabling developers to generate functions, files, and even entire applications from natural language specifications. AI systems are also becoming more personalized, adapting outputs based on inferred user characteristics and interaction ...
- Text-Independent Speaker Verification Using Discrete Audio Tokens
Neural audio codecs (NACs) enable efficient audio compression and have achieved success in downstream tasks such as speech synthesis. However, their discrete representations consistently underperform traditional spectral features in automatic speaker verification (ASV). We empirically demonstrate th...
- Decoupling Conversational Dynamics in Full-Duplex Spoken Models through Reinforcement Learning
Recent full-duplex spoken dialogue models have demonstrated compelling progress toward human-like interaction, enabling agents to respond with low latency, produce backchannels, and handle user barge-ins. Yet these improvements in conversational dynamics often come with weaker reasoning and instruct...
- UBG-Net: An Uncertainty-aware Bayesian Gating Network for Robust Audio-Visual Speech Recognition
Audio-Visual speech recognition systems often degrade in real-world scenarios due to signal corruption and distribution shifts. To address this, we propose a unified uncertainty-modeling framework, namely the uncertainty-aware Bayesian gating network (UBG-Net). UBG-Net features a Modality Uncertaint...
- Interpretable Uncertainty for Adaptive Retrieval and Reasoning in Question Answering
Large language models (LLMs) achieve a strong performance in question answering (QA), but remain prone to hallucinations and suffer from limited transparency. Retrieval-augmented generation (RAG) can improve factuality, yet decisions about when and how to retrieve from external resources are typical...
- Seeing and Reflecting: Multimodal Memory-Enhanced Agent Collaboration for Recommendation
Large language model (LLM)-based agentic recommender systems show promise in modeling user preferences through natural-language reasoning, yet they remain limited by text-centric inputs and coarse-grained memory updates, making agents prone to missing visual evidence, semantic noise, and preference ...
- Simultaneous quantification of dynamic bacterial deformation and motility by machine learning
We developed a convolutional neural network-based machine learning technique to simultaneously analyze the morphology and motility of spirochetal bacteria swimming with continuous cellular deformation. Matching probabilities between experimental images and learned models realizes quantification of cell morphology and association with motility. This method can be applied to diverse transformable cells, offering critical biophysical insights into microbial dynamics.
- Residual Multi-Modal Learning for Pan-Breast-Cancer Drug Response Prediction
Predicting drug sensitivity across diverse cancer cell lines remains a fundamental challenge in precision oncology, particularly for data-scarce cell lines where per-cell-line models overfit and lookup-table approaches cannot generalise to unseen biological contexts. We present DL4DR, a Two Tower Residual Late Fusion deep learning model that addresses this challenge through content-based, identity-free genomic conditioning. The Cell Line Tower encodes each cell line as a 3 x 139 x 139 genomic image - encoding gene expression, mutation severity, and copy-number variation as RGB channels - using a convolutional encoder that maps directly from biological content, never from a cell line ID. The Compound Tower combines three complementary molecular representations: D-MPNN graph message passing, ORNN octave convolutional image features, and an ECFP hard-memorization head that preserves activity-cliff resolution. Predictions are composed as a residual sum: f = fhard + {lambda}(zc). fresidual, where the learned gate $lambda$ modulates how much interaction signal supplements the memorization baseline. Evaluated across 51 breast cancer cell lines(136,342 records), Residual Fusion outperforms the ECFP-Only baseline in 48/51 cell lines (94.1%), with {Delta}R2 > 0.02 in 26/51 (51.0%). On the leave-cell-line-out split - the decisive test of genomic generalisation - the mean {Delta} R2 = 0.016 across all 51 lines demonstrates that the genomic encoder learns transferable biological signal beyond cell line identity. External validation on 601 cell lines across 27 cancer tissue types (CellTiter-Glo dataset; 0 cell line overlap with training) achieves median R2 = 0.627, within the range of the internal random-split performance (R2 = 0.61--0.69), confirming pan-cancer generalisation. GradCAM interpretability on the Cell Line Tower recovers TP53 among the top-five cross-cell-line genomic activators (5/51 cell lines) alongside several uncharacterised candidate genes (e.g.FSIP2, 6/51) - without any prior pathway annotation - providing partial biological validation of the learned representation, while also indicating that a substantial share of the encoder's top-ranked signal corresponds to genes with no current annotation as breast cancer drivers. Code and data are available at https://github.com/bayjuan5/DL4DR.
- scSpark: an AI-assisted cloud platform for traceable interpretation of single-cell transcriptomic results
Single-cell RNA sequencing now routinely produces detailed maps of cell types and states, but interpreting a finished project remains harder than it should be. Once the analysis is done, the results are usually handed over as static reports, figure panels and supplementary tables. A biologist who later wants to revisit an annotation, recompute a cell-type proportion or check whether a pathway is specific to one group typically has to return to a bioinformatician rather than explore the data directly. We developed scSpark to close this gap. The platform takes the completed outputs of a single-cell project: cell annotations, embeddings, differential-expression tables, trajectories, cell-cell communication networks and enrichment result and serves them through a web browser as an interactive workspace. Heavy computation stays upstream: scSpark indexes the precomputed objects under a single project structure and exposes them through six modules for cell annotation, differential analysis, trajectory exploration, cell-cell communication, functional interpretation and AI-assisted result interrogation. Every action in these modules, from a query to a label change, an export or an AI-generated summary, is linked to a specific project version, data object, parameter set and output file, so that any conclusion can be traced back to the evidence behind it. We illustrate the platform by reworking a published periodontitis dataset through this interface. scSpark does not replace upstream pipelines or expert judgement; it is a layer that makes their results easier to inspect, revise and reuse, and that turns a single-cell project from a one-off report into an interpretation others can follow and check.
- Evaluation of Large Language Models for Post-Cystectomy Sexual Health Counseling in Women: A Pilot Study
Abstract Objective To evaluate the adherence to guidelines and readability of large language model-generated sexual health information related to female sexual dysfunction following cystectomy, and to determine whether adherence differs across models and prompt formats. A secondary objective was to introduce an analytic strategy using principal component analysis to examine the dimensions of readability metrics. Methods Three large language models (LLMs), ChatGPT, Gemini, and Perplexity were prompted with six clinical questions related to sexual function after cystectomy. Questions were phrased in long-form and short-form language. Responses were independently graded by two reviewers, derived from guideline recommendations. Linear mixed-effects models predicted adherence as functions of LLM, prompt, and reviewer, with clinical questions as a random intercept. Readability was assessed using five metrics, and principal component analysis (PCA) was used to determine latent structure. Results ChatGPT demonstrated the highest (estimated marginal mean [emm] = 0.769), outperforming Gemini (0.499) and Perplexity (0.457). Shorter, less complex prompts elicited higher adherence than more complex, clinical prompts. All models produced content that exceeded recommended reading levels. PCA demonstrated that a single dominant component accounted for 76.7% of variance across readability indices, indicating a shared underlying construct. Conclusion ChatGPT produced the most guideline-concordant information overall. High linguistic complexity was seen across models, highlighting a barrier to patient comprehension. These findings characterize large language models as variable medical information systems whose outputs rely heavily on prompt structure and model type.
- Is simple better? Comparing Computational Cost and Carbon Impact of Machine Learning Models for Traumatic Brain Injury Prediction; A Case Study for Sustainable Digital Health Implementation
Background Machine learning (ML) models for traumatic brain injury (TBI) prediction increasingly demand extensive data, computational resources, and energy consumption, yet simpler models may offer comparable clinical benefit with lower barriers to deployment. This study compares predictive performance, computational efficiency, carbon footprint, and real-world feasibility of resource-efficient ("pauci-parameter") versus data-intensive ("multiparameter") ML models for predicting TBI patient care pathways and outcomes. Methods External validation study in a level 1 trauma center (n=534 adult TBI patients with GCS<9 and/or intracranial injuries). Seven models tested: two pauci-parameter models using only routine prehospital variables (PREHOSP, 15 variables) or CT image analysis (CT-TIQUA), and five multiparameter models integrating clinical and imaging data. Primary outcome: positive likelihood ratio for predicting neurocritical care intensity, mortality (7/30-day, 6-month), and functional outcome (Glasgow Outcome Scale Extended). Secondary outcomes: computation time, carbon footprint, clinical implementability. Results Multiparameter models showed superior performance but did not consistently translate to better clinical utility. PREHOSP (pauci-parameter) showed comparable performance to complex models for most outcomes. The best-performing multiparameter model (MULTI-PRE) required 100-fold longer inference time and 10-fold higher carbon emissions per prediction versus simple models, while net clinical benefit was nearly identical (0.06 vs 0.05). Models using only prehospital data demonstrated greater generalizability and lower deployment barriers. Interpretation Computational complexity and resource intensity should factor equally with predictive performance in clinical AI deployment decisions. For sustainable digital health implementation--especially in resource-limited settings--simpler models with comparable clinical benefit may enable broader access while reducing environmental and financial costs.
- Exploring the Application of the Observational Medical Outcomes Partnership Common Data Model to Multi-site Stroke Rehabilitation Research Data
Background: Emerging artificial intelligence and machine learning (AI/ML) tools can help generate robust knowledge to support precision rehabilitation approaches for varied patient populations. There is a large amount of research-generated and clinical rehabilitation data available for this purpose; however, a pronounced lack of interoperability prevents large-scale data aggregation. Common data models (CDMs) such as Observational Medical Outcomes Partnership (OMOP) have improved data interoperability across healthcare settings, and more recently, for clinical rehabilitation data, specifically. However, the application of these CDMs to research-generated data has not yet been explored. Therefore, as a foundational step, our study evaluated the breadth and depth of OMOP CDM coverage for data in a multi-site repository of harmonized rehabilitation research data: the Enhancing NeuroImaging Genetics through Meta-Analysis Stroke Recovery (ENIGMA-SR) database. Methods: Two raters independently mapped data elements representing 46 demographics and medical history (DMH) ENIGMA-SR variables and 95 distinct ENIGMA-SR rehabilitation assessments to OMOP standard concepts. Initial rater agreement was assessed for data element inclusion in OMOP and for specific OMOP concepts used (primary metric: Gwet's agreement coefficient [AC]). Mapping differences were reconciled, and final mappings were descriptively analyzed to examine (1) overall OMOP inclusion, (2) inclusion of more granular levels (subscales, items) of complex assessments, and (3) mapped OMOP concept characteristics. Results: Initial rater agreement was good/very good for overall OMOP inclusion of DMH and assessment data elements and for OMOP concepts mapped across almost all assessment data elements (Gwet's AC: 0.79-0.89). Initial OMOP concept agreement was more variable for DMH data elements; however, all mapping differences were successfully reconciled to 100%. Overall, DMH data elements had higher OMOP inclusion than rehabilitation assessments: 84.8% (39/46) vs. 58.9% (56/95). OMOP coverage was particularly limited for complex assessment subscale- and item-level data elements (9.4% [3/32]; 19.2% [14/73]) and did not match the granularity level represented in ENIGMA-SR data for 56.2% (41/73) of complex assessments. DMH and top-level assessment data elements were frequently mapped to multiple OMOP concepts (median: 6, 2; range: 1-23, 1-8), and for > 50% of these data elements the concepts spanned 2-3 different OMOP domains. Conclusion: For ENIGMA-SR, the OMOP CDM has good coverage of DMH data, moderate top-level coverage of rehabilitation assessments, and very limited coverage of assessment subscales and items. This uneven coverage, combined with variability in OMOP concepts and domains mapped to equivalent data points, presents challenges for aggregating clinical and research-generated rehabilitation data into AI/ML-ready datasets. Moreover, software tools currently available to facilitate the mapping process do not effectively accommodate content- and structure-related features inherent to research-generated data. Going forward, the utility of the OMOP CDM to aggregate multi-source rehabilitation data may be improved by expanding the catalogue of OMOP rehabilitation-related concepts, building cross-walks to research-oriented data standards, and adapting emerging computational tools to streamline the mapping process.
- A scalable neuroinformatics pipeline for harmonizing routine clinical electroencephalograms across public hospitals
We propose a study protocol for routine clinical electroencephalograms (EEGs) from public hospitals, which represents a vast resource for neuroscience research. These non-invasive measures of brain function, paired with rich clinical annotations from large and diverse patient populations, are critical for developing robust artificial intelligence (AI) models and conducting population-level studies. This protocol presents a scalable methodology for curating and harmonizing extensive clinical EEG datasets, encompassing over 40,000 individual studies, to facilitate research applications. Key steps include: (i) integration of raw EEG recordings with corresponding clinical records, including neurological reports, diagnostic codes, and potentially medication data; and (ii) spatial standardization of EEG signals by mapping them to a common brain space defined by functional and anatomical landmarks. The resulting harmonized datasets enable the development of large-scale EEG foundation models, the discovery of novel EEG waveform representations, and the creation of normative "brain charts" for electrophysiological assessment across the lifespan. By enabling standardised, large-scale analyses of real-world clinical EEG data, this protocol supports data-intensive solutions for EEG applications and addresses the challenge of generalising AI models. Our approach promotes the translation of AI tools from research to diverse patient populations, advancing population neuroscience.
- Effect of initiating an ARB- versus ACEI-based regimen on dementia risk, a target trial emulation of 2.5 million US Veterans
Background: Hypertension is a modifiable risk factor for dementia, yet the comparative effectiveness of angiotensin receptor blockers (ARBs) versus angiotensin converting enzyme inhibitors (ACEIs) on dementia risk remains uncertain. Objective: To compare the risk of dementia and dementia-free death of ARB versus ACEI initiation among US Veterans with incident hypertension. Methods: We conducted a retrospective target trial emulation using a new-user, active-comparator design among Veterans with incident hypertension. We analyzed longitudinal electronic health records from 2,577,000 individuals who initiated ARBs or ACEIs between 1/1/2000-12/31/2017, with up to five years of follow-up. The exposure was initiation of an ARB-based versus ACEI-based antihypertensive regimen. Co-primary outcomes were dementia, identified using natural language processing of clinical notes, and dementia-free death. We used inverse probability of treatment weights based on 66 pretreatment covariates to estimate the cumulative incidence of the outcomes for each treatment group. Weighted risk ratios and absolute risk differences through five years were computed with bootstrapped 95% CIs. Secondary outcomes included all-cause death and a composite of dementia or death, evaluated using a weighted Kaplan-Meier approach. Results: Among 2,577,000 Veterans (mean age, 63 years; 4.5% female; 65% White; 15% Black), 10% initiated ARBs and 90% initiated ACEIs. Over five years of follow up, 6% developed dementia, 12% died without dementia, and 13% died overall. ARB initiation yielded consistently lower risk of dementia (risk ratio, 0.88; 95% CI, 0.83-0.93 at 6 months to 0.92; 95% CI, 0.90-0.94 at 5 years) and dementia-free death (risk ratio, 0.90; 95% CI, 0.86-0.96 at 6 months to 1.00; 95% CI, 0.98-1.01 at 5 years) than ACEI initiation. Effects on secondary outcomes were similar to those for primary outcomes. Greater protective dementia effects were observed in older and male Veterans and non-statin users, with similar effects on dementia-free death. Discussion: Among US Veterans with incident treated hypertension, initiation of ARB versus ACEI antihypertensive regimen conveyed a modestly lower risk of dementia. Given the high prevalence of hypertension, these modest effects may confer meaningful population-level benefits on brain health. Future research estimating per-protocol effects using a more generalizable population is needed to confirm our findings. Key words: antihypertensive medication, dementia, natural language processing, target trial emulation, Veteran
- Artificial Intelligence-Enabled Detection of Vascular Perfusion Defects on Ventilation/Perfusion (V/Q) Scintigraphy for Pulmonary Embolism
Accurate interpretation of planar ventilation-perfusion (V/Q) scintigraphy, used for diagnosing pulmonary embolism (PE) based on PIOPED/EANM guidelines, requires objective assessment of mismatched V/Q defects. Manual delineation of V/Q defects is time-consuming, subject to interobserver variability, and rarely performed in practice, limiting standardized reporting and quantification of disease burden. To address these challenges, we evaluated four modern AI models for automated segmentation of vascular perfusion defects in planar V/Q scans and compared their performance to human annotators. We retrospectively identified 2,118 patients who underwent planar V/Q scans at The Ottawa Hospital (June 2019-February 2023). Six standard projections (ANT, POST, LAO, RAO, LPO, RPO) were included. Four 2D neural networks (U-Net, nnU-Net, Swin UNETR, and a Bottleneck Transformer U-Net [BTU-Net]) were trained on 1,313 patients (7,878 projections) and validated on 329 (1,974 projections) using physician-annotated defects. A hold-out test set of 46 high probability patients was used to evaluate segmentation quality, and defect detection accuracy using free-response receiver operating characteristic (FROC) analysis, where BTU-Net was the only model performing on par with human readers, showing robust sensitivity across the entire range of segmentation probabilities. At 1.5 false positives per projection rate (FPPR), BTU-Net outperformed other models with a sensitivity of 0.529 {+/-} 0.026, On a separate hold-out set of low likelihood of disease patients (n=430), the lowest FPPR was 0.08 {+/-} 0.01 for BTU-Net (P<0.0001). BTU-Net enables rapid, consistent, and accurate interpretation of planar V/Q scans. Such tools may enhance diagnostic efficiency, standardize reporting, and support non-expert readers in evaluating PE.
- Retina-derived Quantitative Biomarkers of Brain Health
Accurate and scalable assessment of quantitative neuroimaging biomarkers, such as white matter hyperintensities (WMH) and hippocampal (HIP) volumes, is essential for understanding and monitoring brain health, preventing neurological diseases and improving healthspan. However, population-level evaluation of these neuroimaging biomarkers relies on inaccessible, costly and time-consuming magnetic resonance imaging (MRI). Here we propose RetiBrain, a cross-modal deep learning framework that predicts these neuroimaging biomarkers from retinal color fundus photography (CFP) images. By distilling latent structural representations from MRI-based models into a CFP-based model, RetiBrain establishes biologically grounded eye-to-brain mapping. In a CFP-MRI paired cohort, RetiBrain accurately estimates six WMH- and HIP-related biomarkers and outperforms the state-of-the-art retinal foundation model RETFound, improving the mean Pearson correlation coefficient by 0.309 (from 0.240 to 0.549) and achieving a coefficient of 0.640 for periventricular WMH prediction. By integrating structural, topological and geometric feature analyses from CFP images, RetiBrain identifies interpretable retinal representations associated with neurodegeneration and cerebrovascular injury, hallmarks of major neurological diseases such as dementia and stroke. In a longitudinal cohort comprising 2,082 participants (4,164 CFP images with up to 15 years of follow-up), RetiBrain-predicted neuroimaging biomarkers robustly estimated neurological disease risk, as illustrated by dementia prediction (AUROC of 0.824, hazard ratio 2.500 per standard deviation increase, 95% CI: 2.201-2.840). RetiBrain provides a robust, scalable, cost-effective and convenient approach for the assessment of neuroimaging biomarkers, and has potential for long-term brain health monitoring in large-scale general population settings.
- A large language model-assisted workflow for generating a living evidence base for climate-sensitive foodborne disease
Abstract Climate change is altering environmental conditions that influence foodborne disease transmission, yet traditional systematic reviews cannot keep pace with expanding evidence. We assessed whether an LLM-assisted workflow could generate a rapid, repeatable, and policy-relevant living evidence base for climate-sensitive foodborne disease. We combined structured PubMed searches (2010-2023), gold-standard human labelling, and iterative refinement of a GPT?4?Turbo?based auto-labeller within the SysRev platform. Pathogens of public-health importance in England were selected a priori. Model performance was evaluated against human reviewers using recall, precision, specificity, accuracy, and balanced accuracy. The refined inclusion model achieved 89{middle dot}2% recall, 59{middle dot}2% precision, 84{middle dot}5% specificity, and 85{middle dot}4% accuracy across 1,044 screened abstracts, identifying 436 studies for inclusion. Post-hoc re-evaluation of discordant abstracts showed that records excluded by the model but included during initial human screening did not meet the refined inclusion criteria. Frequently identified climate exposures included rainfall, temperature, seasonality, and humidity; norovirus, Salmonella, Campylobacter, and Cryptosporidium were the most common pathogens. An LLM-assisted workflow can generate living evidence for climate-sensitive foodborne disease with high recall and improved screening consistency. The approach is scalable, auditable, and suitable for secure institutional environments, supporting horizon scanning and climate-health risk assessment.
- Serum potassium elevation and acute respiratory acidosis during thoracoscopic esophagectomy with intrathoracic carbon dioxide insufflation: a multicenter retrospective observational study
Background Thoracoscopic esophagectomy with intrathoracic carbon dioxide insufflation and lung collapse, usually performed in the prone position, can markedly alter respiratory physiology and acid-base balance. Serum potassium elevation is often observed during acute respiratory acidosis in these procedures, despite the conventional view that respiratory acidosis has little effect on potassium. We quantified the intraoperative potassium change and explored associated factors. Methods This multicenter retrospective study included adults undergoing thoracoscopic esophagectomy with carbon dioxide insufflation in the prone or lateral decubitus position during 2022-2024. Arterial blood gas variables were evaluated after anesthesia induction and at the time of the lowest arterial pH during carbon dioxide insufflation. The primary outcome was the paired difference in serum potassium. Sensitivity, subgroup, and regression analyses were performed. Results All 131 patients were included: 117 in the prone position and 14 in the lateral decubitus position. Serum potassium increased from 3.96 {+/-} 0.38 to 4.59 {+/-} 0.63 mEq/L (mean increase, 0.64 mEq/L; 95% confidence interval, 0.55-0.73; p < 0.001). During the same period, pH decreased from 7.387 to 7.247 and arterial carbon dioxide tension increased from 41.48 to 58.60 mmHg. After excluding marked metabolic acidosis, the increase remained significant and similar in magnitude (0.618 mEq/L). In the centered multivariable model, lactate change was independently associated with potassium change ({beta} = 0.303; p = 0.004), whereas arterial carbon dioxide change and preoperative renal function were not. The intercept remained significantly positive. Conclusions A clinically meaningful potassium increase was observed during thoracoscopic esophagectomy with carbon dioxide insufflation, while acute respiratory acidosis developed during the same period. The increase persisted after excluding marked metabolic acidosis and may not be explained solely by metabolic stress. Arterial blood gas assessment, including potassium measurement, is warranted during significant hypercapnia, and potential electrolyte consequences should be considered when permissive hypercapnia is accepted.
- Cognitive performance, frailty and functional dependence in community-dwelling older adults: Results of the FREEDOM cohort.
BACKGROUND: Aging is associated with a progressive decline in cognitive performance and functional autonomy, both closely related to frailty. Understanding the interrelation between these domains is essential to identify modifiable factors influencing cognitive impairment in older adults. OBJECTIVES: To evaluate the relationship between physical frailty, cognitive performance, and functional dependence, and to identify sociodemographic and clinical variables associated with cognitive impairment in community-dwelling older adults. DESIGN: Cross-sectional study. SETTING: FREEDOM-LNA cohort, a population-based study conducted by the University Hospital of Limoges, France. PARTICIPANTS: A total of 753 community-dwelling older adults aged [≥]75 years, or [≥]65 years with at least two comorbidities, were included. MEASUREMENTS: Cognitive function was assessed using the Mini Mental State Examination (MMSE), 5-word test (5WT), clock drawing test (CDT), and verbal fluency tests. Frailty was defined according to Frieds physical criteria, and functional independence was evaluated using ADL and IADL scales. Sociodemographic, clinical, and lifestyle factors were analyzed using multivariate models to identify predictors of cognitive impairment. RESULTS: Of the participants, 34.4% had a pathologic MMSE, 46.0% failed the CDT, 68.0% the verbal fluency test, and 17.0% the 5WT. Cognitive performance was significantly lower among frail compared to prefrail and robust individuals. Older adults with pathologic cognition were more frequently dependent in activities of daily living. Independent predictors of poor cognitive performance included non-modifiable factors (age, sex, education) and modifiable ones (low BMI, hypertension, alcohol consumption, smoking, and polypharmacy). CONCLUSIONS: Cognitive impairment was highly prevalent among frail older adults and was strongly associated with loss of independence. Interventions targeting modifiable risk factors such as low BMI, hypertension, alcohol consumption, and smoking may help preserve cognitive and functional abilities in aging populations. Interventions to improve BMI and reduce alcohol consumption, smoking, and hypertension may preserve cognition in older adults.
- Population Impact of a hybrid strategy combining maternal RSV vaccination and nirsevimab immunization on lower respiratory tract infections in infants under 1 year in Bogota: a counterfactual analysis
Objective: To assess the impact of a hybrid RSV immunization strategy on hospitalizations, pediatric intensive care unit (ICU) admissions, outpatient visits, and mortality due to LRTI among infants under one year of age in Bogota. Methods: We conducted an ecological interrupted time-series study using weekly surveillance data from Bogota from EW 1 of 2023 to EW 24 of 2026 (181 weeks). Outcomes included weekly rates of all-cause viral LRTI-related general hospitalizations, pediatric ICU admissions, outpatient visits, and deaths among infants younger than one year. Segmented negative binomial regression models adjusted for secular trends, seasonality using Fourier terms, and autocorrelation were used to estimate changes associated with maternal RSVpreF vaccination and nirsevimab implementation. Counterfactual analyses were performed to estimate cases averted and relative risk reductions. Results: Compared with the same period in 2025, the 2026 LRTI hospitalization rate decreased significantly (rate ratio [RR] 0.66; 95% CI 0.64-0.68), as did pediatric ICU admissions (RR 0.78; 95% CI 0.73-0.85) and outpatient visits (RR 0.78; 95% CI 0.77-0.79). Interrupted time-series analyses identified a significant weekly decline in hospitalization trends following maternal RSVpreF introduction (3.9% per week; p=0.023) and a smaller but significant decline in ICU admissions (-2.8% per week; p=0.039). The cumulative relative reduction in hospitalizations was estimated at 47.1% (95% CI 13.9-70.4), corresponding to 7.605 hospitalizations averted over the post-intervention period (EW 47/2025-EW 24/2026). No statistically significant changes were observed for outpatient visits or mortality. Conclusions: Implementation of a hybrid RSV prevention strategy was associated with a substantial reduction in severe LRTI among infants during the first respiratory season following introduction in Bogota. These findings provide the first real-world population-level evidence from Latin America supporting hybrid RSV immunization as a feasible and potentially cost-effective strategy for reducing severe infant respiratory disease in middle-income settings. Keywords: Respiratory syncytial virus (RSV); Maternal RSV vaccination; Nirsevimab; Hybrid immunization strategy; Population impact; Lower respiratory tract infection (LRTI); Interrupted time series ( ITS); Bogota, Colombia.
- Climate Change, Place, and Mental Health in Sub-Saharan Africa: A Multi-Country Analysis of Lived Experiences Following Extreme Weather Events
Background: Climate change is an escalating global health threat, with sub-Saharan Africa disproportionately affected due to entrenched spatial inequalities, high exposure to environmental hazards, and limited adaptive capacity. Increasingly frequent extreme weather events (EWEs), including floods and cyclones, are reshaping the material and social conditions of place, with implications for mental health and wellbeing. However, evidence remains limited, particularly multi-country qualitative research that examines how mental health impacts are produced through lived experiences of place in contexts of recurring environmental disruption and structural vulnerability. This study explored the mental health and wellbeing impacts of EWEs among individuals with lived experience of such events in Mozambique, Burkina Faso, South Africa, and Kenya, using participatory methods that centred community narratives and place-based accounts of everyday life. Methods: This qualitative study employed digital storytelling as a participatory visual method to examine how EWEs are experienced and narrated across diverse socio-spatial contexts. A total of 37 participants (8 to 10 per country) were recruited from rural, peri-urban, and informal urban settlements with recent exposure to flooding or cyclone events. Participants produced digital stories during facilitated five-day workshops. These narratives were analysed using inductive and deductive thematic analysis informed by Braun and Clarke's framework, with attention to the spatial and relational production of distress and coping. Results: Across Mozambique, Burkina Faso, South Africa, and Kenya, findings show that the mental health impacts of EWEs are deeply embedded in place-based conditions and are cyclical, cumulative, and relational rather than confined to discrete disaster events. Participants described how repeated environmental disruptions reconfigured everyday life in place, generating ongoing uncertainty, anticipatory anxiety during rainfall periods, and acute fear during floods and cyclones. Loss of housing, livelihoods, infrastructure, and social anchors of place contributed to enduring psychological distress, which was frequently reactivated by subsequent environmental cues such as heavy rain, wind, and deteriorating physical environments. Persistent anxiety, hypervigilance, sleep disturbance, and emotional distress were reported across all sites. While social and community networks constituted critical infrastructures of care within place, these were often simultaneously overwhelmed as entire communities experienced shared disruption. Limited and delayed institutional responses further compounded spatial and social precarity. Conclusions: This study provides a comparative participatory account of how EWEs shape mental health through their embeddedness in place across diverse sub-Saharan African contexts. The findings demonstrate that psychological distress is produced through the interaction of repeated environmental exposure, structural inequality, and disrupted place-based infrastructures of daily life, rather than emerging solely as a post-disaster outcome. These results underscore the need for climate-responsive mental health and psychosocial support that is integrated into place-based disaster risk governance, alongside strengthened social protection and community infrastructure that can sustain wellbeing in contexts of recurring environmental instability.
- Artificially sweetened beverage intake and risk of liver-related adverse events in individuals with MASLD: A prospective UK Biobank cohort study
Purpose Metabolic dysfunction-associated steatotic liver disease (MASLD) is a major cause of chronic liver disease and liver-related morbidity worldwide. Although dietary factors may influence MASLD progression, the long-term liver-specific implications of artificially sweetened beverage (ASB) intake remain unclear. We aimed to examine the association between ASB intake and the risk of liver-related adverse events and liver-related death among individuals with MASLD. Methods This prospective cohort study included 50,562 participants with MASLD from the UK Biobank. ASB intake was assessed using 24-hour dietary recalls and categorized as 0, >0-1, and >1 serving/day. Multivariable Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for liver-related adverse events and liver-related death. Restricted cubic spline models were used to assess dose-response patterns, and competing-risk analyses were performed by treating liver-related death as a competing event for liver-related adverse events. Additional substitution, subgroup and sensitivity analyses were conducted to evaluate the robustness of the findings. Results During a median follow-up of 12.8 years, 292 liver-related adverse events and 91 liver-related deaths occurred. Compared with participants reporting no ASB intake, those consuming >1 serving/day had a higher risk of liver-related adverse events in the fully adjusted model (HR 1.40, 95% CI 1.02-1.93; P = 0.039), whereas the association for >0-1 serving/day was not statistically significant (HR 1.26, 95% CI 0.92-1.71; P = 0.149). The risk of liver-related adverse events increased across ASB intake categories (P for trend = 0.023). Restricted cubic spline analysis indicated a positive linear association between ASB intake and liver-related adverse events (P-overall <0.001; P-nonlinearity = 0.72). In competing-risk analysis, the association for >1 serving/day remained consistent after accounting for liver-related death as a competing event (sub-HR 1.40, 95% CI 1.02-1.93; P = 0.038; Gray test P = 0.006). The association was robust in sensitivity analyses. ASB intake was not significantly associated with liver-related death, and beverage substitution analyses showed no significant associations. Conclusion Among individuals with MASLD, high ASB intake, particularly >1 serving/day, was associated with an increased risk of liver-related adverse events, but not liver-related death. This association was consistent across dose-response, competing-risk, and sensitivity analyses, suggesting that high ASB intake may represent a potential dietary risk marker for adverse liver outcomes in MASLD.
- Autoantibodies From Connective Tissue Diseases Penetrate Cells and Exert Functional Properties
Antinuclear antibodies (ANAs) are a hallmark of connective tissue diseases (CTDs) and serve as robust diagnostic biomarkers. Because their cognate antigens are intracellular, ANAs have long been considered non-pathogenic in CTDs. Here, using systemic sclerosis (SSc) associated anti-topoisomerase I antibodies (ATAs) as a model, we provide data challenging this view. We show that ANAs enter living cells, accumulate in nuclei, and engage their intracellular antigen. Nuclear ATAs inhibit topoisomerase I enzymatic activity, induces DNA damage, fibrosis, and, through the STING pathway, activates type I interferon production. We further identify neonatal Fc receptor (FcRn)-dependent intracellular trafficking as a key determinant of ANAs nuclear access and demonstrate that pharmacological FcRn blockade impairs ATAs functionality. These findings reveal a previously unrecognized intracellular effector function of ANAs and establish a mechanistic framework by which ANAs may directly contribute to tissue injury in CTDs.
- A district-level model review system to strengthen coverage and quality of Medical Certification of Cause of Death in India: Protocol for a population based feasibility and effectiveness study
Abstract Background: The current coverage of MCCD in India is only 22%. This is due to incomplete coverage of hospitals under MCCD and also lack of a system for non-institutional deaths in the country. The quality of MCCD in the country is also poor. One of the main reasons for this is the lack of review and feedback at the district level. This study would be the first of its kind study in the country to test the effectiveness and feasibility of involving the district level CRS/Health dept officials in review of MCCD Objectives: To assess the feasibility and effectiveness of a district level review system for MCCD in improving the coverage and quality of MCCD Methods: The study would be conducted in Chikkaballapura district for a period of 2 years. Local Registrars would do a first level of review of MCCD forms for completeness, use of abbreviations, legibility. They would also ensure that form 4/4A is written for all registered deaths in their area. A MCCD review committee would assess the quality of MCCD forms on a monthly basis and provide feedback to the certifying doctors. Comparison of the pre-test and post-test coverage and quality of MCCD will be done. Results: Constitution of the audit committee, training of local registrars, doctors and committee members and baseline assessment have been completed. Intervention has been started from Nov 2025. Expected Outcomes: Improved coverage and quality of MCCD and as a result cause of death data of the district
- Integrating multi-ancestry common and rare variant mapping accelerates therapeutic target discovery
Integrating human genetics into therapeutic discovery accelerates drug development. However, ancestral biases in historical cohorts have left critical functional variation largely uncharted. Here, we leverage the diverse NIH All of Us Research Program to conduct comprehensive common- and rare-variant association analyses for 624 quantitative traits across 369,655 ancestrally diverse individuals. We identified 6,181 genome-wide significant locus-trait associations (526 novel) and 416 gene-trait associations (105 novel) via rare-variant burden testing. By integrating fine-mapping with computational variant-effect predictors, we systematically prioritized rare, likely causal variants driving these signals. Jointly modeling common and rare variation with protein-class annotations significantly improved the identification of known drug targets compared to common-variant analysis alone. Notably, we identified NRG4 as a high-confidence candidate therapeutic target for preserving kidney function. Our findings demonstrate that characterization of rare and common variation across diverse populations enhances causal gene discovery and identifies novel, actionable therapeutic targets.
- Genomic Evidence Links Inflammation to Residual Pulmonary Vascular Obstruction and Risk of Pulmonary Embolism Recurrence
Background and Aims: Residual pulmonary vascular obstruction (RPVO) defined as the persistence of thrombotic material within the pulmonary arteries several months after an acute pulmonary embolism (PE) is associated with an increased risk of severe complications, including recurrent events and chronic pulmonary hypertension. However, the genomic architecture underlying RPVO in unprovoked PE remains poorly understood, and this study aims to address this gap. Method: By leveraging genetic and imaging RPVO data from three independent cohorts totaling 586 unprovoked PE patients, we conducted a meta-analysis of genome wide association study (GWAS) of RPVO using a dedicated statistical method to handle the semi-continuous distribution of RPVO. The meta-GWAS was complemented by haplotype association analyses and transcriptome wide association studies as well as Mendelian Randomization (MR) approaches based on plasma metabolites and proteins. Results: Through meta-GWAS, we identified one locus, OSTN, associated with RPVO (lead variant rs59109356 associated with a ~2-fold increase of RPVO, p=3.92x10-8). A second locus, CCN4, previously reported to associate with pulmonary fibrosis, was also identified, with evidence of association approaching genome-wide significance (p=6.7x10-8). We also identified a common haplotype spanning over AHSG/HRG/KNG1 associated with a ~3-fold increase of RPVO (p=2.96x10-8). Using plasma protein-based MR, we demonstrated that one unit increase in genetically determined plasma levels of IL-1 R AcP encoding IL1RAP was associated with a 28% (p=1.32x10-6) reduction in RPVO. We also observed statistical evidence that the CCN4 (p=0.06) and IL1RAP (p=0.02) loci associate with the risk of PE recurrence in a sample of 1,617 unprovoked PE patients. Conclusions: By identifying novel molecular determinants of RPVO that map to loci involved in inflammatory pathways and vascular remodeling, our study provides evidence that inflammation is the predominant, and likely the key mechanism underlying RPVO, whereas impaired fibrinolysis appears to play a more limited role.
- Tune Out: A randomised controlled trial to investigate the impact of an online program on tinnitus severity, handicap, and psychological symptoms in adults with tinnitus.
Objective: To evaluate the efficacy, engagement, and usability of Tune Out, an unguided, self-paced online tinnitus management program, for reducing tinnitus severity in adults with tinnitus. Design: A two-arm, parallel-group randomised controlled trial was conducted with Australian adults reporting diagnosed or self-reported tinnitus. Participants were randomised to immediate access to Tune Out or a waitlist control group. Outcomes were assessed at baseline, 6 weeks, and 12 weeks. The primary outcome was tinnitus severity measured using the Tinnitus Functional Index (TFI). Secondary outcomes included tinnitus handicap, psychological symptoms, program engagement, self-efficacy, and usability. Results: Eighty-eight participants were randomised: 43 to the intervention group and 45 to the waitlist control group. The primary outcome analysis included 63 participants at 12 weeks. A significant Group x Time interaction was observed for TFI total score, indicating greater reductions in tinnitus severity over time in the intervention group compared with waitlist control, F(2, 102.57) = 5.95, p = .004, partial 2= .104. Significant effects were also observed for tinnitus handicap, F(2, 106.76) = 4.12, p = .019, partial 2 = .072. Effects on psychological symptoms were less consistent, although anxiety showed a significant Group x Time interaction, F(2, 116.85) = 3.63, p = .030, partial 2 = .059. At 12 weeks, 23.1% of intervention participants achieved a clinically meaningful reduction in tinnitus severity compared with 5.4% of controls. Program use was highly variable, with a median use of 1.10 hours, and 25.6% of intervention participants recording no use. Usability ratings were favourable among respondents, with a mean System Usability Scale score of 73.13. Conclusions: Tune Out demonstrated preliminary efficacy for reducing tinnitus severity and tinnitus handicap compared with waitlist control. Effects on broader psychological symptoms were less consistent. Although usability was rated positively, low and variable engagement highlights the need for strategies to support uptake and sustained use in unguided digital tinnitus interventions.
- Urinary CD4+ Effector Memory CD38+ HLA-DR+ T Cells for Diagnosis of Acute Interstitial Nephritis
Introduction Acute interstitial nephritis is an important differential diagnosis in patients with deteriorating kidney function. Diagnosis currently requires kidney biopsy, an invasive procedure associated with risks. We hypothesized that urinary T cells may serve as a non-invasive biomarker for acute interstitial nephritis. Methods A total of 320 patients undergoing clinically indicated kidney biopsy were enrolled in a discovery cohort at Charite Berlin (n = 80), an internal validation cohort at Charite (n = 100), and an external validation cohort at The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou (n = 140). Urinary immune cells were assessed by flow cytometry. Renal T cell infiltration was evaluated by immunofluorescence in kidney biopsy specimens from the discovery and internal validation cohorts, including 16 patients with acute interstitial nephritis and 9 patients without acute interstitial nephritis. Additionally, CXCL9 was measured by ELISA in 102 urine samples from these cohorts. Results Across all cohorts, 27 patients (8.4%) were diagnosed with acute interstitial nephritis. In the discovery cohort, multiple urinary T cell subsets were increased in acute interstitial nephritis, with activated CD4+ effector memory T cells expressing CD38 and HLA-DR showing the strongest diagnostic performance. This marker outperformed urinary monocytes, eosinophils, and CXCL9 and was validated in both independent cohorts. Across all cohorts, the area under the receiver operating characteristic curve was 0.84 and increased to 0.91 after exclusion of 8 patients receiving corticosteroids. A cutoff of 211 activated CD4+ effector memory T cells per 100 mL urine yielded a sensitivity of 78% and a specificity of 81%. Urinary activated CD4+ effector memory T cell counts correlated with renal CD4+ and CD4+ CD38+ T cell infiltration in acute interstitial nephritis. Conclusions Urinary activated CD4+ effector memory T cells expressing CD38 and HLA-DR represent a promising non-invasive biomarker for the diagnosis of acute interstitial nephritis.
- Aortic Geometric Atlas: Centile-Based Reference Charts and Pathological Signatures Across the Adult Lifespan
The aorta is a site of major cardiovascular burden, yet its clinical assessment on computed tomography (CT) remains limited to manual diameter measurements. Here we present the Aortic Geometric Atlas, a comprehensive characterization of thoracic aortic geometry, realized through the Aortic Geometry Toolkit (AGT), an automated pipeline we developed to extract 38 aortic geometric phenotypes (AGPs) across anatomically delineated subsegments. Applying AGT to 62,366 participants representing 140,319 CT studies, we constructed sex-specific, continuous, centile-based reference ranges spanning nine decades of the adult lifespan from 35,648 participants without aortic disease. We then performed a phenome-wide time-to-event analysis for incident disease, identifying 861 prognostic associations across 155 phecodes, with non-caliber geometry contributing predictive value beyond diameter, and derived disease-specific AGP signatures for cardiovascular risk stratification. Together, the Aortic Geometric Atlas provides a population-scale reference for individualized aortic assessment, positioning AGPs as early subclinical markers of incident cardiovascular disease.
- Complex harmonic manifolds in mindfulness-based cognitive therapy for major depressive disorder
Major depressive disorder (MDD) is a heterogeneous mental disorder characterised by rumination. Mindfulness-based cognitive therapy (MBCT) is an evidence-based treatment developed to target rumination and recurrence risk. Ongoing studies have begun to identify neural changes associated with treatment effects. However, the low-dimensional organisation underlying whole-brain dynamics remains largely unexplored and may provide a more complete characterisation of the neural processes through which MBCT exerts its therapeutic effects in MDD. Here, we investigated functional magnetic resonance imaging (fMRI) of a randomised controlled trial of MBCT with treatment as usual (TAU), or TAU alone, in a group of MDD patients (N=80). We applied a novel framework, complex harmonics decomposition (CHARM), to uncover low-dimensional manifolds in the spacetime domain, capturing local as well as non-local interactions made possible by brain criticality and amplified by the anatomical long-range connectivity. We successfully identified distinct distributed spatiotemporal manifolds across brain states and outperformed traditional dimensionality reduction techniques. During rumination after MBCT we found consistent recruitment of regions involved in bodily and interoceptive processing integrated within the whole-brain across manifolds, changes in latent configurations associated with clinical and behavioural improvements, and greater flexibility within the reduced space. Integration of bodily and interoceptive processing regions within distributed whole-brain manifolds and greater brain flexibility may be associated with reduced 'stickiness' of ruminative thinking patterns following mindfulness training in depression. Our findings highlight the promise of low-dimensional manifolds and long-range interactions arising from critical brain dynamics in understanding how mindfulness targets depressive ruminative processing.