Importance:Mania or hypomania, the pathognomonic feature of bipolar disorder (BD), is associated with reward hypersensitivity and impulsive decision-making, but the neural mechanisms underlying these distinct behavioral facets remain unclear. Reward expectancy (RE)-associated activation in the left ventrolateral prefrontal cortex (L-vlPFC) and pre-supplementary motor area (pre-SMA) is elevated in individuals at risk of mania or hypomania, but it is unknown whether elevated activation in these regions and associations with risk-related behavioral facets reflect convergent or dissociable pathways to mania or hypomania risk or whether depression severity impacts these associations. Objective:To test whether RE-associated L-vlPFC and pre-SMA activation exhibit distinct associations with delay discounting behavior-indexing reward valuation (lower discounting) vs impulsive decision-making (higher discounting)-and whether these associations are moderated by depression severity. Design, Setting, and Participants:In this cross-sectional study, 2 independent samples of adults aged approximately 18 to 30 years at varying risk of mania or hypomania (142 in the discovery set and 86 in the replication set) completed a functional magnetic resonance imaging reward task and a delay discounting assessment. Major exclusion criteria included current or lifetime BD, primary psychotic disorders, neurological disorders, recent substance use disorders, systemic medical illnesses, and magnetic resonance imaging contraindications. The study took place from 2019 to 2026 at the University of Pittsburgh Medical Center in Pittsburgh, Pennsylvania. Exposure:RE-associated L-vlPFC and pre-SMA activation. Main Outcomes and Measures:Main outcomes were mania or hypomania risk per Mood Spectrum Self-Report-Lifetime mania score and delay discounting behavior per 27-Item Monetary Choice Questionnaire rate. Associations were tested using regression models, including moderation by depression severity. Results:Across 228 individuals (discovery: mean [SD] age, 23.79 [3.32] years; 96 [67.6%] female; replication: mean [SD] age, 26.09 [3.15] years; 64 [74.4%] female), L-vlPFC and pre-SMA activation were both positively associated with mania or hypomania risk in the discovery sample (l-vlPFC: β, 0.544; 95% CI, 0.440 to 0.649; z, 10.24; P < .001 and pre-SMA: β, 0.496; 95% CI, 0.341 to 0.650; z, 6.30; P < .001) and replication sample (l-vlPFC: β, 0.503; 95% CI, 0.241 to 0.765; z, 3.77; P < .001 and pre-SMA: β, 0.684; 95% CI, 0.410 to 0.957; z, 4.90; P < .001). In the discovery sample, these regions showed dissociable delay discounting associations: greater L-vlPFC activation was associated with lower discounting (β, -0.305; 95% CI, -0.561 to -0.050; P = .02), whereas greater pre-SMA activation was associated with higher discounting (β, 0.521; 95% CI, 0.146 to 0.895; P = .007). The L-vlPFC-lower discounting association replicated (β, -0.704; 95% CI, -1.400 to -0.008; P = .048). A pre-SMA activation × depression severity interaction was observed in the replication sample (β, -0.427; 95% CI, -0.792 to -0.063; P = .02), such that higher depression severity attenuated the association between elevated pre-SMA activation and higher discounting. This interaction was detectable in the discovery sample in individuals with elevated pre-SMA activation (β, -0.285; 95% CI, -0.565 to -0.005; t63 = -2.03; P = .046). Conclusions and Relevance:The findings in this cross-sectional study suggest that RE-associated L-vlPFC and pre-SMA activation may represent dissociable neural pathways to mania or hypomania risk. Elevated L-vlPFC activation was associated with sensitivity to reward value, whereas elevated pre-SMA activation was associated with impulsive decision-making, and this latter association was moderated by depression severity.

Enterprise data analysis is emerging as a distinct frontier for autonomous agents. Compared with general-purpose interaction and software engineering, it operates in an open, ambiguous, and continuously evolving environment. These characteristics call for a data-agent architecture that treats semantics, methodology, execution, and evolution as first-class system concerns. To this end, we introduce QwenPaw-Data, an agentic data system designed for enterprise intelligent data analysis. QwenPaw-Data consolidates heterogeneous assets from warehouses, dashboards, documents, interaction logs, and historical tasks into reusable, governable, and evolvable analysis assets, then turns natural-language requests into end-to-end analytical workflows spanning data understanding, retrieval, analysis, report generation, and decision support. Its architecture decomposes the problem into three collaborative subsystems: DataBridge provides trustworthy semantic grounding through interconnected metadata, knowledge, and trace graphs; Skill-Hub codifies expert analytical methodology into reusable and verifiable skills; and Host materializes these evidence and method assets into controllable, artifact-centric runtime execution. Across these subsystems, semantics, methods, traces, and feedback are continuously deposited back into the system, forming a self-evolving asset flywheel. Experiments on public benchmarks and real-world industrial BI workloads show that QwenPaw-Data improves both verifiable data access capability and higher-level analytical quality, offering a practical foundation for reliable, traceable, and continuously improving enterprise data agents.

Neuromelanin-sensitive magnetic resonance imaging (NM-MRI) is used as a proxy biomarker of midbrain catecholaminergic metabolism and synthesis capacity, particularly within dopaminergic nuclei. However, despite strong translational evidence for sexual dimorphism in dopaminergic systems, inconsistent sex differences in NM-MRI signal limit its interpretability as a non-invasive measure of catecholaminergic synthesis capacity. A source of variability may be the reliance on contrast-to-noise ratio (CNR) normalization, which presupposes NM-MRI signal stability within the reference region, canonically the crus cerebri. To address this limitation, the present study introduced a novel normalization framework combining a data-driven reference region with residual-based, rather than ratio-based, normalization. NM-MRI data were acquired in a transdiagnostic sample of 101 adults, aged 18-40 years, including individuals with elevated familial risk for bipolar disorder or major depressive disorder. Sex effects on NM-MRI signal were compared across the substantia nigra pars compacta (SNc), ventral tegmental area (VTA), and locus coeruleus (LC) using conventional crus cerebri-based CNR versus residual-based normalization. Residual-based normalization revealed higher NM-MRI signal in males compared with females across all catecholaminergic nuclei (SNc, VTA, LC; all adjusted p < 0.001). In contrast, conventional CNR normalization detected no sex differences in any region (all adjusted p > 0.20). Moreover, sex differences in NM-MRI signal were observed within the crus cerebri itself (p < 0.001), indicating that the canonical reference region is not biologically invariant. Sexual dimorphism in midbrain NM-MRI signal is revealed in vivo for the first time using a novel residual-based normalization framework. By reducing reference-region bias, this approach advances NM-MRI toward candidate biomarker deployment for assessing catecholaminergic dysfunction in psychiatric and neurological disorders.

Mania/hypomania is pathognomonic of bipolar disorder (BD), yet early identification remains challenging. Impulsivity is a key feature of mania/hypomania and of externalizing disorders that may predispose to BD, but neural markers of impulsivity-related risk remain unknown. This study aimed to identify reward expectancy (RE)-related neural correlates of impulsivity facets, test moderation by current affective/anxiety symptoms, and determine whether such markers differentiate BD and/or externalizing disorders from low impulsivity individuals. Two independent BD-risk samples aged 18-30 years, including individuals with prior externalizing disorder diagnoses but not BD, were recruited; a euthymic BD group was also recruited. Impulsivity facets were assessed via Behavioral Activation System (BAS) and UPPS-P scales. Whole-brain regressions identified neural correlates of impulsivity facets during RE. Linear models tested replication and current affective/anxiety symptom moderation. ANCOVA compared neural activity among BD, externalizing, and non-BD/externalizing impulsivity tertile groups. Whole-brain regressions revealed a positive association between BAS Fun Seeking and pre-supplementary motor area (pre-SMA) activity (pFWE = 0.003, k = 167), which replicated when depressive symptoms were covaried (discovery: β = 2.73, p < 0.001; replication: β = 0.88, p = 0.036; combined: β = 1.49, p < 0.001). A significant pre-SMA × depression interaction (β = -0.08, p = 0.037) indicated depressive symptoms attenuated the pre-SMA-Fun Seeking association. Group comparisons revealed greater pre-SMA activity in high-Fun Seeking (p < 0.001) and externalizing disorder groups (p = 0.039) versus low-Fun Seeking, with similar trends observed in BD once individuals taking medications, particularly benzodiazepines (p = 0.012), were excluded. Pre-SMA hyperactivity during RE is a robust neural correlate of BAS Fun Seeking, moderated by depression severity. This pattern represents a trait-linked neural marker of impulsivity associated with vulnerability to BD and externalizing disorders, informing early risk identification and intervention.

Objective neural markers that reflect the underlying pathophysiological mechanisms of affective disorders are needed to facilitate early identification of individuals most at risk of future affective disorders and ultimately provide neural targets to guide therapeutic interventions. Using an emotional n-back paradigm designed to examine working memory (WM) and emotional regulation (ER) capacity, we previously showed that WM-related elevated left dlPFC activity (a key node of the central executive network (CEN)) and elevated right precuneus activity (a key node of the default mode network (DMN)), as well as ER-related elevated left dlPFC activity were positively associated with future depression severity in young adults at risk for affective disorders. We now aimed to replicate and extend these previous longitudinal findings by examining relationships among right precuneus activity and left dlPFC activity during WM and ER tasks and future depression severity in a new independent young adult sample (n = 77: 50 female, age = 24.68), and a larger combined sample (n = 121: 83 female, age = 23.81) comprising the original and new samples. The Hamilton Rating Scale for Depression (HAM-D) and Young Mania Rating Scale (YMRS) were measured at 12 months post scan to assess future depression and mania/hypomania severity respectively. In the new sample, we showed patterns of left dlPFC activity and right precuneus activity during WM, and left dlPFC activity during ER that were consistent with the original sample. In both the new and combined samples, future depression severity was robustly predicted by WM-related left dlPFC activity and right precuneus activity, and ER-related left dlPFC activity (all ps < 0.05 qFDR). These findings were specific to future depression severity. The effect sizes (pseudo R-squared values) for the full models including all IVs in the new and combined samples ranged from approximately 25-43%, with left dlPFC and right precuneus activity during WM explaining 15.59% of variance in future depression severity in the new sample; and left dlPFC activity during ER explaining 14.63% of variance in future depression severity in the combined sample. These replicated, longitudinal findings provide candidate neural markers to guide risk identification and targeting of new interventions for individuals with and those at risk for future affective disorders.

BACKGROUND:Identifying reproducible neural markers of bipolar disorder(BD) risk is critical for early detection and differentiation from unipolar/major depression. We previously demonstrated that inter-amygdala functional connectivity(FC) and bilateral-ventrolateral-right-dorsolateral-prefrontal-cortex(vlPFC-dlPFC)-FC were positively associated with both mania/hypomania and depression risk and mania/hypomania risk, respectively, as measured by the Mood Spectrum Self-Report(MOODS-SR) 'mood' subdomains, replicated in three independent samples. We tested whether these neural markers also generalized to the MOODS-SR 'cognition' and 'energy' subdomains and whether they are elevated in individuals with BD versus those at-risk. METHODS:Three independent young adult samples without BD(n=299/ages 18-30) completed an fMRI approach emotion-processing task(Discovery n=114/age=21.60±1.91; Test sample-1 n=103/21.57±2.09; Test sample-2 n=82/23.43±2.86), and a fourth sample with BD(n=32/25.11±3.73) completed the same protocol. Poisson loglinear models tested whether previously identified neural markers of MOODS-SR mood subdomains were also associated with manic and depressive cognition and energy subdomains. One-way ANOVAs compared neural variables showing significant relationships with subdomain scores between low-risk, high-risk, and BD groups. RESULTS:Across all risk samples, inter-amygdala-FC was positively associated with manic and depressive mood and cognition subdomains(qFDRs<0.001-0.01); vlPFC-dlPFC-FC was positively associated with manic mood and cognition subdomains(qFDRs<0.001-0.048). Inter-amygdala-FC differed significantly across groups(F2,328=3.56, P=0.03) and was higher in BD versus low-risk and in high-risk versus low-risk groups(Ps=0.033). CONCLUSIONS:Inter-amygdala-FC emerged as a robust, cross-dimensional correlate of BD risk, linking subsyndromal risk to syndromal BD, whereas vlPFC-dlPFC-FC was specific to mania risk. Findings support a multidimensional, circuit-based model of BD risk supporting early identification and prevention.

Objectives: Bipolar Disorder (BD)-characterized by mania, affective lability, and elevated psychosis risk-is associated with sustained attention deficits. However, evidence regarding performance on the AX-Continuous Performance test task (AX-CPT), a task reliably impaired in schizophrenia-spectrum psychosis, is mixed. This study examined whether individual differences in mania/affective lability risk and psychosis risk, across cohorts of BD and BD-at-risk individuals, were associated with altered AX-CPT performance. Methods: The standard measures d'context and A-cue bias quantified AX-CPT performance. Factors from the Mood Spectrum Self Report (MOODS-SR-L) indexed lifetime mania/affective lability risk (psychomotor activation, suicidality, and mixed instability) and lifetime psychosis risk. Linear regressions were performed for dimensional continuous aims (Aims 1-2) for both d'context and A-cue bias. Tests between BD with low and high-risk groups (Aims 3-4) were completed with ANCOVAs for both AX-CPT output measures. Results: In the final sample of euthymic BD (n = 27) and at-risk individuals (n = 121), higher levels of psychosis and mania/affective lability risk were associated with reduced target discrimination (d'context; absolute t's > 2.015, p's < 0.047), but not with altered A-cue bias. Group comparisons showed no significant differences for either AX-CPT measure. Associations with d'context from Aims 1-2 remained after covarying for current depressive symptoms but were removed when covarying for current mania severity (absolute t's < 1.26, all p's > 0.203). Conclusions: Target discrimination deficits were associated dimensionally with psychosis and mania/affective lability risk but not categorically with BD diagnosis. This suggests scope for dimensional risk models for understanding sustained attention deficits in BD and at-risk individuals and highlights contributions of mania/affective lability and psychosis risk.

Psychotic disorders, characterized by perceptual abnormalities and cognitive decline, affect millions of people. Auditory hallucinations (AH), or the perception of non-existent sounds, are particularly taxing. While AH is often viewed as a perceptual disorder, patient distress is more closely related to cognitive factors. Research suggests that dysfunctional attentional control of auditory systems may contribute to AH development, and auditory cognitive control may be a key factor in the link between AH and functional decline. This study used magnetoencephalography (MEG) to investigate attentional control of ASSR in first-episode psychosis (FEP) within left and right primary auditory cortex (A1) at initial clinical contact and 4-12 months later. We investigated the relationships among ASSR deficits and attention-mediated sensory gain in 40 FEP and 40 matched healthy comparison subjects (HC). We measured ASSR to click trains with attention directed toward or away from the auditory stimulus at both timepoints. Results indicated that FEP patients showed reduced attentional modulation of ASSR, particularly in the right A1, and increased ASSR during passive listening in left A1, correlating with AH severity. ASSR measurements were reliable, with persistent group differences despite symptom reduction. These findings highlight the role of selective attention deficits in psychosis and suggest ASSR as a potential biomarker for temporal lobe dysfunction. Early identification of ASSR deficits could enable targeted treatments for individuals at highest risk of developing a chronic disorder.

With the increasing growth of video data, limited bandwidth and hardware resource constraints demand more efficient video compression. Current learned video compression methods have shown promising performance. However, these methods mainly rely on the optical flow networks to perform temporal prediction, which may suffer from inaccurate motion estimation and introduce extra artifacts to reconstructed frames. In this paper, we propose a spatio-temporal feature enhancement method for learned video compression to better model the inter-frame motion patterns and reduce compression artifacts. Specifically, we introduce a spatio-temporal motion enhancement module that further extracts the feature representation of original motion vector to enhance corresponding spatial and temporal components. Then, we introduce an in-loop filtering enhancement module that employs cascaded residual blocks to progressively enhance feature textures and provide higher- quality temporal domain reference signals for subsequent reconstruction. More importantly, our proposed method can be integrated into the widely-used residual coding and contextual coding schemes. Comprehensive experiments demonstrate that our integrated methods are superior to the previous learned methods on JCTVC, UVG and MCL-JCV benchmark datasets. In addition, our integrated methods also outperform the latest generalized video coding standard (H.266/VVC) by a larger margin in terms of MS-SSIM metric.

This paper explores the application of enhancement filtering techniques in neural video compression. Specifically, we categorize these techniques into in-loop contextual filtering and out-of-loop reconstruction enhancement based on whether the enhanced representation affects the subsequent coding loop. In-loop contextual filtering refines the temporal context by mitigating error propagation during frame-by-frame encoding. However, its influence on both the current and subsequent frames poses challenges in adaptively applying filtering throughout the sequence. To address this, we introduce an adaptive coding decision strategy that dynamically determines filtering application during encoding. Additionally, out-of-loop reconstruction enhancement is employed to refine the quality of reconstructed frames, providing a simple yet effective improvement in coding efficiency. To the best of our knowledge, this work presents the first systematic study of enhancement filtering in the context of conditional-based neural video compression. Extensive experiments demonstrate a 7.71% reduction in bit rate compared to state-of-the-art neural video codecs, validating the effectiveness of the proposed approach.

Video stabilization is a critical technology for enhancing visual content quality in dynamic shooting scenarios, especially with the widespread adoption of mobile photography devices and Unmanned Aerial Vehicle (UAV) platforms. While traditional digital stabilization algorithms can improve frame stability by modeling global motion trajectories, they often suffer from excessive cropping or boundary distortion, leading to a significant loss of valid image regions. To address this persistent challenge, we propose the View Out-boundary Synthesis Algorithm (VOSA), a symmetry-aware spatio-temporal consistency framework. By leveraging rotational and translational symmetry principles in motion dynamics, VOSA realizes optical flow field extrapolation through an encoder–decoder architecture and an iterative boundary extension strategy. Experimental results demonstrate that VOSA enhances conventional stabilization by increasing content retention by 6.3% while maintaining a 0.943 distortion score, outperforming mainstream methods in dynamic environments. The symmetry-informed design resolves stability–content conflicts and outperforms mainstream methods in dynamic environments, establishing a new paradigm for full-frame stabilization.

This report presents Wan, a comprehensive and open suite of video foundation models designed to push the boundaries of video generation. Built upon the mainstream diffusion transformer paradigm, Wan achieves significant advancements in generative capabilities through a series of innovations, including our novel VAE, scalable pre-training strategies, large-scale data curation, and automated evaluation metrics. These contributions collectively enhance the model's performance and versatility. Specifically, Wan is characterized by four key features: Leading Performance: The 14B model of Wan, trained on a vast dataset comprising billions of images and videos, demonstrates the scaling laws of video generation with respect to both data and model size. It consistently outperforms the existing open-source models as well as state-of-the-art commercial solutions across multiple internal and external benchmarks, demonstrating a clear and significant performance superiority. Comprehensiveness: Wan offers two capable models, i.e., 1.3B and 14B parameters, for efficiency and effectiveness respectively. It also covers multiple downstream applications, including image-to-video, instruction-guided video editing, and personal video generation, encompassing up to eight tasks. Consumer-Grade Efficiency: The 1.3B model demonstrates exceptional resource efficiency, requiring only 8.19 GB VRAM, making it compatible with a wide range of consumer-grade GPUs. Openness: We open-source the entire series of Wan, including source code and all models, with the goal of fostering the growth of the video generation community. This openness seeks to significantly expand the creative possibilities of video production in the industry and provide academia with high-quality video foundation models. All the code and models are available at https://github.com/Wan-Video/Wan2.1.

Objectives:Accurate segmentation of craniomaxillofacial (CMF) structures and individual teeth is essential for advancing computer-assisted CMF surgery. This study developed CMF-ELSeg, a novel fully automatic multi-structure segmentation model based on deep ensemble learning. Methods:A total of 143 CMF computed tomography (CT) scans were retrospectively collected and manually annotated by experts for model training and validation. Three 3D U-Net-based deep learning models (V-Net, nnU-Net, and 3D UX-Net) were benchmarked. CMF-ELSeg employed a coarse-to-fine cascaded architecture and an ensemble approach to integrate the strengths of these models. Segmentation performance was evaluated using Dice score and Intersection over Union (IoU) by comparing model predictions to ground truth annotations. Clinical feasibility was assessed through qualitative and quantitative analyses. Results:In coarse segmentation of the upper skull, mandible, cervical vertebra, and pharyngeal cavity, 3D UX-Net and nnU-Net achieved Dice scores above 0.96 and IoU above 0.93. For fine segmentation and classification of individual teeth, the cascaded 3D UX-Net performed best. CMF-ELSeg improved Dice scores by 3%-5% over individual models for facial soft tissue, upper skull, mandible, cervical vertebra, and pharyngeal cavity segmentation, and maintained high accuracy Dice > 0.94 for most teeth. Clinical evaluation confirmed that CMF-ELSeg performed reliably in patients with skeletal malocclusion, fractures, and fibrous dysplasia. Conclusion:CMF-ELSeg provides high-precision segmentation of CMF structures and teeth by leveraging multiple models, serving as a practical tool for clinical applications and enhancing patient-specific treatment planning in CMF surgery.

Accurately predicting the probability distribution of quantized latent representations is a critical challenge for entropy models in learned video compression (LVC). Existing mainstream LVC methods typically adopt ready-made entropy models based on image compression, which fail to fully exploit the information of spatial-temporal correlation. To address this issue, we propose a spatial correlation priors and hierarchical temporal attention (SCP-HTA) model, which exploits the spatial correlation information from the current video frames and refine the temporal information from the context. First, we extract the spatial correlation of the current frame to guide the generation of masks, enabling the frame to leverage more information during encoding and decoding process. Additionally, to obtain more accurate temporal information, we introduce a hierarchical temporal attention module at channel level when we generate the context. Experimental results demonstrate that the proposed SCP-HTA model achieve 15.76% bitrate saving in PSNR and 61.82% in MS-SSIM on average across all test datasets when compared with VTM-13.2 (LDP).

Mania/hypomania, the pathognomonic feature of bipolar disorder (BD), is characterized by elevated impulsivity, often assessed via delay discounting-the preference for smaller, immediate versus larger, delayed rewards. It remains unclear whether delay discounting differentiates BD from non-BD individuals or serves as an objective behavioral marker of mania/hypomania versus depression risk. Bipolar disorder (n = 40) and non-BD (n = 187) individuals were recruited, with the latter encompassing a range of mania/hypomania and depression risk and stratified into mania/hypomania and depression risk tertiles. Kruskal-Wallis and Dunn's tests evaluated delay discounting rates (k values), assessed via the 27-Item Monetary Choice Questionnaire, across both risk groups compared to the BD group. Significant group effects were found for overall and geomean k values in both mania/hypomania (overall k: χ2(3) = 8.15, p = 0.043; geomean k: χ2(3) = 8.40, p = 0.038) and depression risk groups (overall k: χ2(3) = 8.30, p = 0.04; geomean k: χ2(3) = 8.75, p = 0.033). Only k values for medium reward magnitudes were significant for both mood risk stratifications (corrected α = 0.05/3 = 0.0167). Bipolar disorder had significantly higher k versus low-risk mania/hypomania individuals (adjusted p = 0.012), as did high-risk versus low-risk mania/hypomania individuals (adjusted p = 0.039). Bipolar disorder had higher k versus high-risk depression individuals (adjusted p = 0.005), as did low-risk versus high-risk depression individuals (adjusted p = 0.029). Bipolar disorder had significantly higher k for medium reward magnitudes versus high-risk depression-only (W = 398, p < 0.001), but not versus high-risk mania/hypomania-only (W = 587.5, p = 0.368) individuals. Delay discounting for medium reward magnitudes differentiates BD from non-BD individuals and distinguishes heightened mania/hypomania risk from depression risk, supporting its potential as an objective behavioral marker for mania/hypomania risk detection.