Monitoring deep-sea biodiversity is challenging because conventional methods, such as net-sampling and visual surveys, require substantial logistical effort. Environmental DNA (eDNA) offers an alternative for detecting marine organisms; however, its application in deep-sea ecosystems is limited by the sampling accessibility. Because the influence of long‑pipeline transportation on eDNA is unknown, we compared two deep-sea water sampling methods: pumped deep-sea water (pumped method) and water collected near the intake using a Niskin bottle (Niskin method). Samples were collected from pumping stations in Sagami and Suruga bays, Japan, and baited camera observations were conducted to verify eDNA-based detections. In Sagami Bay, both sampling methods detected similar numbers of fish species. In Suruga Bay, the Niskin method exhibited higher species diversity and steeper accumulation curves than the pumped method. The habitat depth ranges of identified species corresponded to the water intake depths in both bays, supporting the effectiveness of the pumped method. Fish community composition showed substantial overlap between methods, and baited cameras recorded 11 and 17 fish taxa in Sagami and Suruga bays, respectively, with 7 and 10 taxa matching eDNA detections. These findings indicate that eDNA metabarcoding of pumped deep-sea water is a feasible method for monitoring deep-sea fish biodiversity.
To develop and validate DeepAdapter, a novel deep learning algorithm that integrates self-supervised learning (SSL) and unsupervised domain adaptation (UDA) to enhance model generalisability for retinopathy of prematurity (ROP) screening. DeepAdapter was developed in two stages. First, SSL was applied to 500 000 unlabelled infantile colour fundus photographs (CFPs) retrospectively collected from four Chinese clinical centres to learn general fundus representations. Second, a much smaller labelled dataset of Normal and ROP CFPs was constructed, and a UDA module was incorporated to mitigate domain shift, defined as distribution differences between training and deployment datasets that may impair model performance. Correlation Alignment (CORAL) distance was employed to quantify these domain shifts. External validation was performed using data from an independent clinical centre and two public multiethnic datasets. A total of 500 000 non-overlapping unlabelled CFPs and 39 456 labelled CFPs were included. Compared with the supervised method, DeepAdapter decreased CORAL distance from 0.297 to 3.550×10-6. Both algorithms achieved high internal accuracy (0.989) in ROP prediction, but DeepAdapter significantly outperformed the supervised method on external testing (accuracy: 0.828 (95% CI 0.823 to 0.833) vs 0.739 (95% CI 0.733 to 0.745), p <0.001) and on the public cross-ethnicity dataset (accuracy: 0.839 (95% CI 0.818 to 0.859) vs 0.813 (95% CI 0.791 to 0.835), p=0.014). A web-based system integrating quality control and ROP diagnosis was developed for large-scale, multicentre clinical screening. DeepAdapter effectively mitigated domain shift and significantly improved model generalisability for ROP screening, providing a valuable reference for developing generalisable models in other medical specialties.
Time efficient and reliable pipelines for quantitative evaluation of structural brain MRI are essential to utilize the potential of morphometry tools for large scale research projects as well as to pave the path towards future clinical applications. In our work, we have explored this idea by evaluating three deep learning models for brain segmentation and cortex parcellation (DeepSCAN, FastSurferCNN and QuickNAT) as input for an 11-min surface reconstruction pipeline adapted from the well studied open source software package FreeSurfer. Performance was assessed using both, large publicly available human MRI datasets and a synthetic dataset with known metrics and reference surfaces. Evaluation criteria included closeness to the surface reconstruction by FreeSurfer's full recon-all pipeline, reproducibility within same-session rescans, performance stability across a wide age range, sensitivity to variations of the grey-white contrast in the MRI and accuracy regarding metrics of synthetic surfaces. Metrics derived from the DeepSCAN-based pipeline demonstrated the highest agreement with FreeSurfer in the human data and the greatest fidelity to the expected metrics in the synthetic dataset. Our findings identify the DeepSCAN-based surface reconstruction pipeline as a rapid, yet reliable alternative to established research-grade structural MRI processing. Time expenditure and reliability suggest it is suitable for research applications with high-throughput requirements. This is an essential first step towards necessary subsequent studies aimed at evaluating robustness, pathological variability, and utility in the context of clinical diagnostics.
Accurate photovoltaic power forecasting is critical for grid stability, but remains challenged by weather uncertainties and difficulty integrating historical observations with forward-looking forecasts. We re-evaluate existing architectural choices in deep-learning models for time-series forecasting and demonstrate the importance of full encoder-decoder architectures and channel dependence modeling when both weather forecasts and historical data are available. Based on this insight, we propose Cross-Unet, a Transformer-based architecture featuring multi-scale temporal encoding, correlation-aware channel attention, and hierarchical cross-attention decoding to fuse historical generation data with weather forecasts. Evaluated on open-source datasets from four utility-scale plants in northern China and one aggregated plant in central Australia, using three types of forward-looking inputs: numerical weather prediction, satellite-derived irradiance, and AI-based weather model forecasts. Across the majority of evaluated configurations spanning five photovoltaic power stations, five forecasting horizons (4 hours to 7 days), and three forecast sources, Cross-Unet outperforms ten deep learning baselines and traditional operational benchmarks. By integrating advanced forecasting systems, such as modern AI weather models, into an end-to-end forecasting pipeline, Cross-Unet enables operational 15-minute-resolution predictions over 4-hour to 7-day horizons, supporting grid scheduling and energy trading.
Biodegradable Zn alloys have attracted considerable attention as candidates for load-bearing bone-fixation implants, yet simultaneously optimizing mechanical strength, corrosion-wear resistance, and multifunctional biofunctionalities remains challenging. Herein, a Zn-3Cu-0.8Sr (ZCS) alloy was successfully fabricated by a synergistic processing route that integrated hot rolling (HR) with deep-cryogenic rolling (DCR). The HR+DCR processing effectively refined coarse and brittle SrZn13 and primary ε-CuZn5 phases into uniformly dispersed, well-bonded fine reinforcements, while simultaneously promoting grain coarsening and precipitate growth by suppressing dynamic recovery and restricting atomic diffusion at cryogenic temperatures. This microstructural engineering strategy produced an optimal combination of mechanical properties, including an ultimate tensile strength (σuts) of ∼301.7 MPa, a yield strength of ∼245.0 MPa, an elongation at break (ε) of ∼33.5%, the lowest σuts loss of 12.5% and ε loss of 6.6% after 30 d of immersion in Hanks' Balanced Salt Solution, and the highest biotribological resistance among all thermomechanically processed specimens. The HR+DCR processed specimen exhibited the lowest electrochemical corrosion rate of ∼162 µm/y and degradation rate of ∼20.1 µm/y in Dulbecco's Modified Eagle Medium with fetal bovine serum among all thermomechanically processed specimens. Notably, the alloy displayed enhanced osteoblast viability, osteogenic differentiation and mineralization, and near-complete antibacterial activity against Staphylococcus aureus in both in vitro and in vivo settings. Moreover, the alloy effectively modulated the immune response, driving macrophage polarization toward a pro-healing M2 phenotype. Overall, the alloy combines high mechanical, biotribological, degradation, osteogenic, antibacterial, and immunomodulatory biofunctions, underscoring its potential for next-generation biodegradable orthopedic-fixation devices. STATEMENT OF SIGNIFICANCE: This work reports a multifunctional Zn-3Cu-0.8Sr alloy fabricated using a synergistic hot rolling and deep-cryogenic rolling process for next-generation orthopedic applications. The alloy exhibits exceptional mechanical properties: σUTS of ∼301.7 MPa, σYS of ∼245.0 MPa, and ε of ∼33.5%, with minimal strength/ductility loss after 30-day immersion in Hanks' Balanced Salt Solution (HBSS). It demonstrates a favorable electrochemical corrosion rate (∼162 µm/y), degradation rate (∼20.1 µm/y), and superior biotribological resistance in Dulbecco's Modified Eagle Medium supplemented with fetal bovine serum (DMEM+FBS). Biologically, it enhances osteoblast viability and mineralization while providing near-complete S. aureus antibacterial efficacy in vitro and in vivo. This synergistic combination of strength, corrosion-wear resistance, and bioactivity highlights the alloy's significant potential for advanced biodegradable orthopedic applications.
This study addresses critical gaps in automated lymphoma segmentation from PET/CT imaging, often overlooked in prior work. While deep learning has been applied to this task, few studies evaluate generalizability on external or out-of-distribution data. Similarly, intra- and inter-observer variability analyses remain rare, limiting understanding of task difficulty. Moreover, most methods emphasize global segmentation metrics, neglecting lesion-level characteristics that are crucial for clinical decision-making. We propose a clinically-relevant evaluation framework to assess four commonly used deep segmentation networks (ResUNet, SegResNet, DynUNet, SwinUNETR) on 611 PET/CT cases from multi-institutional datasets spanning varied lymphoma subtypes and lesion characteristics. In addition to the Dice similarity coefficient (DSC), we compute prediction errors on clinical lesion measures and analyze DSC performance as a function of these measures. Additionally, we use traditional lesion-specific detection criteria (1 and 2), providing insights into network's performance in identifying and localizing lesions respectively, and propose an additional Criterion 3 for segmenting lesions based on metabolic characteristics. Finally, we contextualize network performance by comparing it to expert human observers through intra- and inter-observer variability analyses. Networks perform best on large, metabolically active lesions. Their error patterns closely resemble those of expert annotators, while small and faint lesions remain challenging for both networks and physicians. Our clinically-relevant benchmarking framework enables more consistent and meaningful evaluation of lymphoma segmentation models, supporting robust decision-making in patient care. The approach is extensible to other architectures and disease types. Code is available at: https://github.com/microsoft/lymphoma-segmentation-dnn.
Brain age prediction has gained significant attention due to its strong correlation with neurological and cognitive disorders. The discrepancy between an individual's chronological age and their predicted brain age-known as the Brain Age Gap-has been linked to conditions such as schizophrenia, Alzheimer's disease, cognitive decline, and lifestyle factors like stress and poor health. A positive Brain Age Gap is often associated with accelerated aging and neurodegeneration, highlighting the need for precise and reliable estimation methods. In this study, we propose a novel deep learning model that incorporates time-distributed, convolutional and bidirectional LSTM layers for brain age estimation. Using MRI data from the OpenBHB dataset, processed through Voxel-Based Morphometry (VBM), our model undergoes rigorous preprocessing, including outlier detection, data augmentation, and MRI slice selection, to enhance learning efficiency. The model is optimized with the Adam optimizer with a scheduled learning-rate decay and evaluated using Mean Absolute Error (MAE) and [Formula: see text] Score. Experimental results demonstrate that our model achieves an MAE of 3.1573 years, outperforming previous methods and improving brain age prediction accuracy. These findings underscore the importance of advances in deep learning and data preprocessing in enhancing brain age estimation.
Wireless capsule endoscopy (WCE) enables painless, minimally invasive visualization of the gastrointestinal tract. Still, its diagnostic potential is limited by incomplete mucosal coverage and poor transferability of existing navigation methods across patient anatomies. We propose a transferable, anatomical landmark-guided deep reinforcement learning framework for robust autonomous gastric navigation. Leveraging a lightweight edge-contour-depth fusion module, our policy operates on stable, low-dimensional landmark coordinates rather than high-dimensional video streams. This design effectively bridges the sim-to-real visual gap and ensures robustness across diverse anatomies, enabling low-cost deployment by reducing computational overhead. In simulations across eight patient-derived models, the method achieves >97% coverage within 50 s, significantly outperforming vanilla Proximal Policy Optimization, Soft Actor-Critic, and Deep Q-Network agents by enhancing coverage and minimizing variance. To ensure deployment reliability, a two-stage sim-to-real pipeline supported by an adaptive dynamic programming controller actively mitigates physical disturbances, including actuator latency and peristalsis. Ex vivo experiments across five independent scans demonstrate high coverage stability, achieving a mean coverage of 87% and a 53% reduction in procedure time compared with expert manual control. This study establishes a scalable paradigm for autonomous, high‑coverage endoscopic navigation, advancing the clinical deployment of intelligent WCE systems for GI diagnostics.
Two-dimensional covalent organic frameworks (2D COFs) are widely used as stationary phases in liquid chromatography, yet their simple pore structures often mandate surface modifications to attain the necessary separation selectivity. In contrast, three-dimensional COFs (3D COFs) feature customizable hierarchical channel systems with complex pore structures, high specific surface areas, and spatially diversified active sites, enabling superior separation selectivity and enhanced column efficiency. Furthermore, their rigid frameworks prevent structural collapse of stationary phase, offering unique advantages for advanced separations. Therefore, the development of innovative 3D COF@silica stationary phase represents a promising strategy that leverages the established benefits of silica supports while incorporating the tunable porosity and functionality of 3D COFs. This work presents an eco-friendly deep eutectic solvent-mediated synthesis of an imine-linked 3D-Azo-TFPM-COF@silica stationary phase, which eliminates hazardous solvents and reduces reaction time. The synthesized material demonstrated excellent chromatographic performance in separating polar compounds and positional isomers. Meanwhile, the 3D-Azo-TFPM-COF@silica column was applied to cosmetic analysis for quantifying nicotinamide content in commercial whitening essences with high accuracy (recovery: 95.82-102.92%) and good precision (RSD < 0.4%). This work presents a straightforward and green synthetic approach for 3D COF based stationary phases, broadening their application in chromatographic separations and demonstrating their potential for advanced analytical techniques.
Fine-grained aircraft type recognition in remote sensing imagery is of great value for military intelligence analysis and battlefield situation awareness. However, the limited transparency of deep learning decision-making processes restricts its application in high-stakes decision-making scenarios. To address this issue, this paper proposes an explainable aircraft type classification framework, termed EC-AT, which integrates transfer learning, explanation consistency analysis, and trustworthiness evaluation. Specifically, eight representative aircraft categories are selected from the MAR20 dataset to evaluate multiple deep neural network models under two settings: transfer-learning fine-tuning and random initialization. By leveraging pretrained model parameters and fine-tuning the classification layers, the proposed framework effectively improves feature representation capability and convergence efficiency under limited and class-imbalanced data conditions. To further interpret model decisions, three representative explainable artificial intelligence methods, namely Grad-CAM, LIME, and RISE, are introduced to analyze discriminative image regions from complementary perspectives, including class-discriminative activation, local perturbation-based approximation, and randomized input sampling. In addition, an IoMin-based spatial overlap metric is designed to quantitatively assess the consistency among regions generated by different explanation methods. Beyond explanation consistency analysis, the proposed framework is further evaluated using the Multisource AI Scorecard Table (MAST), which provides a structured assessment of model trustworthiness from multiple dimensions, including accuracy, consistency, logical argumentation, and visualization. Experimental results show that the transfer-learning-based models achieve excellent classification performance, with classification accuracy reaching up to 99.9%. The explainability analysis demonstrates that the proposed framework can identify key aircraft-related regions while revealing potential background interference and decision biases. Furthermore, the explanation consistency analysis and MAST-based evaluation jointly provide objective evidence for explanation results and the overall trustworthiness of the framework. Overall, the proposed EC-AT framework significantly improves the classification performance and explainability of fine-grained aircraft recognition in remote sensing imagery, offering a useful reference for trustworthy intelligent remote sensing analysis.
Deep brain stimulation (DBS) is a neuromodulatory intervention primarily utilized in treatment-refractory motor symptoms of Parkinson disease (PD), dystonia and essential tremor (ET). Primary DBS target sites for treatment of movement disorders include subcortical regions with important connections to the limbic, associative and motor systems, potentially leading to neuropsychiatric symptoms. We sought to highlight the role that consultation-liaison (CL) psychiatrists can play in multidisciplinary care for patients with movement disorders by reviewing the literature on psychiatric disorders and clinically significant psychiatric symptoms before and after DBS. This narrative review was conducted utilizing a literature search using PubMed and Google Scholar. The review focused on: 1) the prevalence of psychiatric disorders and/or clinically significant psychiatric symptoms in treatment-refractory PD, dystonia and ET, and the impact of these symptoms on quality of life (QOL) and functioning; and 2) the incidence and/or prevalence of, and risk factors for, psychiatric disorders and/or clinically significant psychiatric symptoms after DBS for these movement disorders. Over two-thirds of patients with PD will experience clinically significant psychiatric symptoms over the course of the disease. Over 90% of patients with dystonia may develop a psychiatric disorder over their lifetime. Nearly three-quarters of patients with ET will experience clinically significant psychiatric symptoms. Psychiatric symptoms have been associated with negative impacts on QOL and functioning across all three movement disorders. Up to three-fifths of patients undergoing DBS for treatment refractory movement disorders may experience clinically significant psychiatric symptoms within weeks of surgery. Suicides have been reported following DBS for PD and dystonia. Potential risk factors for post-DBS clinically significant neuropsychiatric symptoms in patients with PD include advanced age, pre-DBS cognitive impairment and psychiatric disorders, and stimulation of ventral non-motor subcortical brain regions. Psychiatric disorders and clinically significant psychiatric symptoms are common in patients with PD, dystonia or ET prior to DBS therapy. Clinically significant neuropsychiatric symptoms, including completed suicides, are prevalent following DBS for PD or dystonia. CL psychiatrists can play a vital role in the multidisciplinary care of patients with treatment-resistant movement disorders undergoing DBS through routine psychiatric screening, evaluation and follow-up.
Protein-protein interactions (PPIs) are fundamental to cellular processes, and essential for understanding biological function and disease mechanisms. In this review, we emphasize recent deep learning-based methods for protein interaction study. Focusing on three closely related tasks of proteome-wide PPI prediction, PPI interface prediction, and PPI co-complex structure prediction, we discuss how emerging concepts and computation approaches have evolved to shape these fields We categorize recent approaches according to their methodological paradigms, summarize their strengths and limitations, and further explore diverse biological and biomedical applications, highlighting how computational methods in PPI prediction, PPI interface prediction, and PPI structure prediction jointly contribute to understanding of complex biological systems.
Grade 4 glioma is inherently lethal due to inevitable recurrence. Current radiotherapy guidelines recommend uniform target volume margins, disregarding molecular and clinical variables. We hypothesized that spatiotemporal recurrence may be predicted by molecular and phenotypic features present already at diagnosis. We analyzed 390 paired MRIs of grade 4 gliomas from Norwegian multi-centers and TCIA (2015-2025). Deep learning segmentation identified contrast-enhancing (CEcore) and non-enhancing (NE) tumor compartments and afflicted anatomical regions. We quantified primary CEcore/NE volume ratio, anatomical tumor trajectories, and spatiotemporal progression using Hausdorff-95-analyzed in relation to survival, MGMT, IDH, age, extent of resection, sex, and anatomical location, using machine learning and Cox regression. Primary tumor composition was independently prognostic: CEcore/NE volume ratio ≤0.324 predicted improved overall survival (adjusted HR = 0.56, 95% CI 0.37-0.84, p = 0.006), independent of age, MGMT-status, and individual compartments. Within IDH-wildtype, MGMT-unmethylated patients, low volume ratio ≤0.324 demonstrated a 4.3 months survival benefit (median 17.6 vs 13.3 months, p = 0.0209). Longer time to progression correlated with increased HD95 distances (p < 0.03). Tumors originating in occipital lobe had highest propensity to migrate to new sites (57.1%) and the shortest time to progression (adjusted HR = 1.90, p = 0.026). These findings support molecular-demographic-anatomical risk stratification that may inform personalized margin-determination in radiotherapy planning.
Sugarcane tops and leaves (STL) represent an abundant yet underutilised lignocellulosic residue with significant biorefinery potential for the valorisation of its three principal components: cellulose, hemicellulose, and lignin. However, existing fractionation studies often prioritise lignin removal, while hemicellulose recovery is compromised by acid-catalysed degradation and sugar dehydration. Here, we report a one-pot, two-step deep eutectic solvent (DES) fractionation strategy that decouples hemicellulose extraction from lignin solubilisation using a low-cost oxalic acid:choline chloride (1:1 molar ratio) system. In the first step, a hydrated DES (75 wt%) at 80 °C for 60 min selectively recovered hemicellulose, achieving 95.6% combined conversion to xylose and xylobiose while limiting glucose (5.3%), furfural (3.3%), and 5-hydroxymethylfurfural (1.3%) formation. The second step employed neat DES at 120 °C for controlled delignification, achieving 95.4% lignin recovery within 10 min at 87.8% purity, with reduced condensation compared to a single-step lignin extraction approach. The resulting cellulose-rich pulp contained 80.1% cellulose with enhanced crystallinity (CrI = 34.8%). Importantly, DES regeneration via recrystallisation reduced downstream energy demand. This integrated strategy enables sequential and selective recovery of all three biomass fractions within a single solvent system, advancing circular lignocellulosic biorefinery design.
Midfoot osteoarthritis affects approximately 12% of adults over 50 years and is a common cause of chronic foot pain and disability. When conservative management fails, midfoot arthrodesis remains the standard surgical treatment but carries significant morbidity, including nonunion rates of 3-10% and prolonged non-weightbearing. Deep peroneal nerve (DPN) neurectomy has emerged as a motion-preserving alternative; however, clinical evidence has not been systematically evaluated. A systematic review following PRISMA 2020 guidelines was conducted (PROSPERO: CRD420251266344). MEDLINE, Cochrane CENTRAL, and Web of Science were searched from inception to October 2025. Quality was assessed using the ROBINS-I tool. Data were synthesised narratively due to heterogeneity in outcome reporting. Four retrospective case series (Level IV evidence) comprising 88 patients (106 feet) met inclusion criteria. Mean age was 66.3 years; 80% were female. Follow-up ranged from 4 weeks to 51 months. Patient satisfaction was approximately 75% across studies, though variably defined and best regarded as an illustrative estimate.The pooled surgical complication rate was 5.7% (6/106 feet), comprising predominantly minor wound complications. Reoperations, including revision neurectomy, exostosis excision, and conversion to arthrodesis, were required in 7.5% of feet (8/106). Conversion to midfoot arthrodesis occurred in 4.7% of feet (5/106). All four studies were assessed as having serious risk of bias using ROBINS-I. DPN neurectomy may provide clinically meaningful pain relief with low complication rates in selected patients with midfoot osteoarthritis, offering faster recovery than arthrodesis. Current evidence is limited to small retrospective case series; prospective comparative studies are needed.
Accurate prediction of drug-target interactions (DTIs) is a fundamental challenge in early-stage drug discovery, particularly in the absence of reliable three-dimensional structural information. In this study, we propose a fully sequence-based DTI prediction framework that eliminates dependence on structural data while achieving docking-comparable predictive performance. The proposed framework introduces a unified representation that systematically integrates physicochemical protein descriptors, protein 3-gram sequence motifs, and sequence-like drug encodings into a single feature space, enabling effective learning across heterogeneous models. A diverse set of machine learning, deep learning, and ensemble classifiers is evaluated under stratified five-fold cross-validation with class imbalance correction using Synthetic Minority Over-sampling Technique (SMOTE). Beyond individual models, the framework incorporates advanced ensemble strategies, including a stacking classifier that combines Random Forest, Support Vector Machine, and Logistic Regression, resulting in robust performance with ROC-AUC values exceeding 0.90 and a maximum AUC of 0.914. Importantly, the framework explicitly addresses model interpretability through feature importance analysis, revealing biologically meaningful protein sequence motifs associated with binding interactions. To further substantiate the reliability of the proposed approach, molecular docking experiments are conducted on a subset of predicted drug-target pairs, and the observed agreement between docking scores and predicted binding probabilities provides independent validation. Collectively, this study demonstrates that carefully engineered sequence-derived representations, coupled with optimized ensemble learning, constitute a scalable, interpretable, and computationally efficient alternative to structure-dependent DTI prediction methods.
To evaluate whether pre-conception surgery for deep endometriosis (DE) is associated with obstetric outcomes compared with conservative management. Retrospective cohort study. Single-center tertiary hospital. Nulliparous women with singleton pregnancies and diagnosed DE who delivered between 2017 and 2024. Pre-conception surgery for DE. Obstetric outcomes were compared between women undergoing pre-conception DE surgery and those managed non-surgically. The primary outcome was a composite adverse obstetric outcome, including major maternal, fetal, and neonatal complications, analyzed using multivariable logistic regression, adjusted for maternal age, body mass index, adenomyosis, anatomical disease extent (#Enzian classification), and mode of conception, with additional sensitivity analyses to assess robustness. A total of 298 nulliparous women were included, of whom 65.8% (196/298) were managed without surgery, and 34.2% (102/298) underwent pre-conception DE surgery. The composite adverse obstetric outcome occurred in 37.3% (38/102) of the surgery group and 18.9% (37/196) of the no-surgery group (p<0.001). After multivariable adjustment, pre-conception surgery remained associated with higher odds of the composite outcome (adjusted OR 2.3, 95% CI 1.2-4.3). The surgery group had higher rates of placenta previa, gestational hypertensive disorders, postpartum hemorrhage, blood transfusion, and cesarean delivery, whereas fetal and neonatal outcomes were comparable between groups. Pre-conception surgical treatment of DE was associated with increased maternal obstetric morbidity; however, given the observational design and baseline differences in disease severity, this finding should be interpreted as an association rather than evidence of causation.
Traditional separation of inventory management and Prognostics and Health Management (PHM) often leads to resource misallocation. While Deep Reinforcement Learning (DRL) offers a promising solution for joint decision-making, standard agents typically treat Prognostics and Health Management predictions as deterministic ground truths. However, in real-world scenarios, remaining useful life (RUL) predictions inherently contain stochastic errors. Ignoring this uncertainty leads to risk-blind policies that fail to buffer against sudden failures when prediction confidence is low. To address this, this paper proposes an Uncertainty-Aware collaborative adaptive inventory strategy. First, we introduce a Bayesian uncertainty quantification mechanism using Monte Carlo Dropout to estimate not only the RUL value but also its prediction variance. Second, to overcome the agent's myopic behavior, a novel Asymmetric Cost-Aware Reward Shaping mechanism is designed. By strategically decoupling the training and evaluation reward functions-specifically by introducing safety stock penalties and attenuating holding costs during training-the agent is guided to establish robust inventory buffers against supply chain uncertainties. Simulation results demonstrate that the proposed Risk-Sensitive PPO strategy significantly outperforms deterministic baselines, reducing total costs by 40.3% under high-noise environments.
The efficient recovery of carbohydrates from fruit by-products is limited by their restricted accessibility, as they are embedded within the complex cell wall matrix. Overcoming this barrier requires particle-size reduction, matrix disruption and selection of appropriate solvent. Recently, natural deep eutectic solvents (NADES) have emerged as promising sustainable alternative for polysaccharide recovery. However, their high viscosity can hinder mass transfer and consequently extraction yield. This study investigates the combination of NADES with grinding to enhance cell wall carbohydrates accessibility in lemon peel and apple pomace. Vibratory ball grinding was performed in presence of three chloride-based NADES: lactic acid (CCLA), malic acid (CCMlic), and malonic acid (CCMnic). Combining grinding with CCLA and CCMlic significantly increased mass yields, up to four-fold, reaching a maximum of 9.0%. This increase was associated to increased surface area and mechanical energy input by impact, as supported by modelling. In contrast, CCMnic showed limited efficiency with yields up to 4.3%, likely due to weaker solvent-matrix interactions. Compositional analyses revealed the co-extraction of polysaccharides with other components, including lignin (3.0 to 15%). Additionally, the initial disruption degree of the fruit by-product was identified as a critical factor for polysaccharides recovery.
To evaluate the impact of cone position on prognostic outcomes of deep anterior lamellar keratoplasty (DALK) in advanced keratoconus. This retrospective study analyzed 49 eyes (48 patients) with advanced keratoconus undergoing DALK between 2015 and 2023. Cases were categorized as central (n = 26) or paracentral (n = 23) based on the position of the cone. Primary outcomes included uncorrected visual acuity (UCVA), best-corrected visual acuity (BCVA), manifest refraction, keratometric characteristics, corneal thickness, and quality of life assessments. The results showed that the postoperative UCVA and BCVA in the central group were significantly better than those in the paracentral group (UCVA: 0.492 ± 0.192 vs. 0.617 ± 0.241, P = 0.049; BCVA: 0.192 ± 0.102 vs. 0.396 ± 0.199, P < 0.001). Although the paracentral group exhibited a more significant recovery in central corneal thickness (CCT) (CCT: 528.654 ± 30.900 vs. 549.696 ± 36.394, P = 0.034; postop CCT-preop CCT: 127.615 ± 55.697 vs. 155.565 ± 37.490, P = 0.048), no significant difference was found in postoperative keratometric characteristics between the 2 groups (P > 0.05). Quality-of-life assessments revealed that the central group achieved significantly higher postoperative scores in the far vision and driving subscales compared with the paracentral group (both P < 0.05). In contrast, the paracentral group demonstrated significantly better postoperative scores in the color vision and peripheral vision subscales (P = 0.017 and P = 0.038, respectively). Cone position exhibited a positive correlation with UCVA, BCVA, and CCT (P < 0.001). The corneal cone position significantly affects DALK prognosis, influencing postoperative quality of life. Central group shows better visual acuity, whereas paracentral group demonstrates significant CCT recovery. Tailoring surgical strategies based on cone position may mitigate prognostic disparities and optimize visual rehabilitation.