共找到 20 条结果
暂无摘要（点击查看详情）
Natural Language Processing requires data to be pre-processed to guarantee quality models in different machine learning tasks. However, Swahili language have been disadvantaged and is classified as low resource language because of inadequate data for NLP especially basic textual datasets that are useful during pre-processing stage. In this article we develop and contribute common Swahili Stop-words, common Swahili Slangs and common Swahili Typos datasets. The main source for these datasets were short Swahili messages collected from Tanzanian platform that is used by young people to convey their opinions on things that matters to them. Therefore, we derive list of common Swahili stop-words by reviewing most frequent words that are generated with Python script from our corpus, review common slang with help of Swahili experts with their corresponding proper words, and generate common Swahili typos by analysing least frequent words generated by a Python script from corpus. The datasets were exported into files for easy access and reuse. These datasets can be reused in natural language processing as resources in pre-processing phase for Swahili textual data.
This study was undertaken to report rates of payer blocking of prescribed branded migraine medications in total and by race/ethnicity, income, and education, and to examine the relationship between payer blocking and emergency department and inpatient encounters for migraine-specific and all-cause reasons in a large US claims database. Access to prescribed branded migraine medications can be challenging due to restrictive utilization management protocols and may be related to increased healthcare resource utilization. These barriers may disproportionately affect patients based on sociodemographic factors. This observational cohort study analyzed data from an integrated database containing electronic medical records, specialty pharmacy, and in-office dispensing datasets. Individuals with a migraine diagnosis who experienced payer blocking-defined as denial of prescribed branded migraine medications ≥2 times due to step therapy, prior authorization, or related restrictions-between January 1, 2019, and December 31, 2021, were included. Rates of payer blocking were reported for the total sample and by sociodemographic groups. Emergency department (ED) and inpatient encounter rates for migraine-specific and all-cause reasons were compared by payer blocking status. Differences were assessed using χ2 tests with Bonferroni corrected α (p < 0.0018). Effect sizes were estimated using Cramer's V. Among 7.7 million patients with ≥1 migraine-related prescription claim(s), 370,560 actively treated individuals met inclusion criteria for analysis. One in five (20.6%) experienced payer blocking. There were significant differences in the distribution of race/ethnicity, annual household income, and educational attainment categories across payer blocking status groups (p < 0.001 for all), with small effect sizes (Cramer's V = 0.014-0.034). Patients with a history of payer blocking had significantly higher rates of ED and inpatient encounters for both all-cause and migraine-specific reasons (p < 0.001 for all), although effect sizes were small. This pattern was consistent within racial/and ethnic groups, with significant differences observed for White patients both for migraine-related and all-cause ED and inpatient encounters, and Black/African American and Hispanic/Latino patients for migraine-related encounters (p < 0.001 for all). Payer blocking of branded migraine medications was fairly common, affecting one in five patients and was broadly associated with increased healthcare utilization. These findings suggest that payer blocking may disproportionately impact patients according to race/ethnicity, household income, and educational attainment. Although effect sizes were small, the outcomes may provide useful hypotheses for understanding and addressing healthcare disparities in migraine [Correction added on January 31, 2026 after first online publication: typos errors are removed to ensure accuracy]. This study examined data from 370,560 people with migraine who were denied access to prescribed branded migraine medications two or more times due to insurance restrictions such as step therapy or prior authorization (this is also called “payer blocking”). Payer blocking occurred for about one out of every five patients and was more common among patients with lower income and/or education and varied by race/ethnicity, although the magnitude of effects was small. Those who experienced these forms of payer blocking were more likely to visit the emergency department and/or be hospitalized for migraine and/or other reasons within the 3‐year study window.
Prediction is a core feature of language, which is widely studied across research domains. The Munich Sentence (MuSe) database enhances reproducibility by providing sentence completion norms for 619 German sentences, including cloze probabilities and entropy estimates from up to 232 participants. Sentence completions were collected in two online studies in which participants completed sentence beginnings with a single-word response after either hearing (auditory sample, N = 133) or reading (visual sample, N = 98) the sentence beginning. All responses were manually preprocessed to correct typos and spelling mistakes and to label grammatical errors, proper nouns, and singular and plural variants of the same response. In addition to the sentence norms, we provide trial-level data with participant-level demographic information and subclinical autistic and schizotypal trait measures. Together with open-access R scripts or our web tool, this allows tailoring the cleaning and norming steps to integrate individual-difference measures. For a subset of 479 sentence beginnings, the database also includes professional audio recordings of sentence beginnings, which can be flexibly combined with 531 recordings of unique sentence-final words and implemented in auditory language paradigms. All material is freely accessible via the Open Science Framework ( https://osf.io/ktnze/overview ) and the MuSe webtool ( https://munichsentencedatabase.franziskaknolle.com/ ).
Public omics repositories contain vast amounts of valuable data, but their metadata suffers from extreme heterogeneity, unstandardized terminologies, and quality issues that severely limit data reusability and cross-study integration. While prospective metadata standards exist, the majority of published omics data remain in non-standardized formats requiring retrospective harmonization. We performed comprehensive manual curation and harmonization of metadata, such as participant characteristics and study conditions, from 212 027 omics samples across 468 studies in two repositories: curatedMetagenomicData (93 studies, 22 588 samples) and cBioPortal (375 studies, 189 438 samples). Through systematic ontology mapping, we consolidated redundant, dispersed information into far fewer harmonized columns, reduced unique values, and increased the completeness of major attributes. This curation process revealed common metadata quality issues, including typos, inconsistent terminologies, misplaced values, conflicting annotations, and inappropriately merged information across attributes. We document the challenges, decisions, and solutions during this large-scale metadata harmonization. The harmonized metadata, accessible through the OmicsMLRepoR Bioconductor package, enables repository-wide queries and cross-study analyses previously challenging with heterogeneous metadata. Our experience provides practical guidance for similar curation efforts and demonstrates the value of investing in retrospective metadata improvement for existing public omics resources.
Trace element imbalances are implicated in cancer pathophysiology, yet data from Iraq, a country with distinct environmental and industrial exposures, remain scarce. The stomach plays a central role in trace element absorption through its acidic environment and digestive processes, and gastric cancer disrupts this function, potentially altering systemic trace element homeostasis. This study measured serum zinc (Zn), copper (Cu), and lead (Pb) concentrations in gastric cancer patients and evaluated their diagnostic potential with gender-stratified analysis. A case-control study was conducted in Najaf, Iraq (2022-2024), enrolling 100 gastric cancer patients and 100 age-matched healthy controls (50 males and 50 females per group). Serum Zn and Cu were quantified by flame atomic absorption spectrometry (FAAS); Pb was measured by graphite furnace AAS (GFAAS). Independent samples t-tests compared group differences, and receiver operating characteristic (ROC) analysis assessed diagnostic accuracy. Female patients showed significantly lower serum Zn (100.43 ± 64.25 vs. 818.31 ± 130.26 ppb, p < 0.001) and higher Cu (1466.87 ± 568.69 vs. 977.79 ± 249.30 ppb, p < 0.001) compared to healthy controls. Male patients showed significantly lower Cu (89.50 ± 15.50 vs. 173.14 ± 98.94 ppb, p < 0.001), while Zn did not differ significantly (p = 0.394). Serum Pb was elevated in both sexes (females: 40.31 ± 9.43 vs. 11.98 ± 2.68 ppb; males: 44.79 ± 10.33 vs. 12.38 ± 2.68 ppb; both p < 0.001). ROC analysis showed Pb to be the strongest discriminator in both sexes (AUC = 0.989), while Zn was a strong discriminator in females (AUC = 0.994) but poor in males (AUC = 0.396). Serum Pb showed consistent elevation across both sexes and strong diagnostic accuracy, suggesting its potential as a cancer-associated biomarker in this population. Gender-specific differences in Zn and Cu underscore the role of sex as a biological modifier in trace element metabolism during gastric cancer. Larger multicenter studies with comprehensive environmental and dietary data are warranted.
Automatic segmentation of computed tomography (CT) images is fundamental for quantitative anatomical analysis in a wide range of clinical applications. Despite remarkable advances in artificial intelligence (AI), CT acquisition parameters critically affect image quality, structural fidelity, and model performance. This narrative review investigates how variations in CT acquisition influence image quality and accuracy of AI-based aortic segmentation models. Following PRISMA guidelines, a narrative review was conducted to identify studies assessing how CT technical parameters influence image quality and AI-based aortic segmentation accuracy. From 376 initial records, 13 studies met the inclusion criteria. Thinner slices improved segmentation accuracy, increasing the Dice Similarity Coefficient (DSC) from 0.75 to 0.87 and reducing the 95th Percentile Hausdorff Distance from 2.1 to 1.2 mm. Larger slice spacing (3 mm) worsened accuracy by about 12%, while 1 mm slices reduced aortic diameter errors to below 2%. Faster acquisitions decreased image noise from 32.3 ± 6.3 to 22.1 ± 5.9 HU. Among segmentation models, 2D-3D U-Net ensembles achieved the highest accuracy (DSC = 0.928 ± 0.026), with thrombus segmentation (DSC = 0.782 ± 0.170) outperforming vessel segmentation (DSC = 0.481 ± 0.155). Optimal pixel spacing (0.665 × 0.665 mm) improved image fidelity, yielding a PSNR of 16.28 dB and lower MSE (2729.67). CT acquisition parameters critically impact the reliability of AI-driven aortic segmentation. Optimizing these parameters improves model performance, but the opacity of AI systems still limits clinical interpretability, highlighting the need for standardized acquisition protocols and explainable AI approaches.
Irisin is an exercise-induced myokine that has been proposed to exert beneficial effects on metabolic health. However, its response to different exercise training modalities in individuals with overweight or obesity remains inconsistent. This systematic review and meta-analysis primarily aimed to evaluate the effects of various exercise interventions on circulating irisin levels as the primary outcome in overweight and obese adults. Additionally, we assessed changes in other selected myokines and metabolic markers as secondary outcomes to provide a better understanding of exercise induced physiological adaptations. A systematic search was conducted in Web of Science, EMBASE, Cochrane Library, PubMed, SCOPUS, and Google Scholar (up to 22 April 2025) to identify randomized controlled trials (RCTs) evaluating the effects of different exercise training protocols (aerobic, resistance, concurrent, and high-intensity interval training) on circulating irisin levels in adults with overweight or obesity. The primary eligibility criterion to include studies was the measurement of circulating irisin. Within the RCTs meeting this criterion, we additionally extracted data on selected myokines (follistatin, myostatin, and FGF21) and metabolic markers (glycemic control and lipid profiles) when reported. These secondary outcomes were analyzed to contextualize irisin responses within broader metabolic adaptations, but no separate systematic search was performed for these variables. Pooled effect sizes were calculated using random-effects models and expressed as standardized mean differences (SMDs) with 95% confidence intervals (CIs). Using the median split technique, subgroup analyses were computed according to exercise training modality. A total of 50 studies comprising 1780 participants (1104 in exercise groups and 676 in passive control groups) were included. For the primary outcome, exercise training was associated with a significant increase in circulating irisin (SMD 0.62, 95% CI 0.39-0.85, p < 0.001, n = 76 arms, 1721 subjects) compared with passive controls. Among the secondary outcomes, training was also associated with increases in high-density lipoprotein cholesterol (SMD 0.25, 95% CI 0.03-0.47, p = 0.030], n = 18 arms, 420 subjects), follistatin (SMD 0.89, 95% CI 0.37-1.41, p = 0.008, n = 7 arms, 141 subjects), and fibroblast growth factor 21 (FGF-21; SMD 1.00, 95% CI 0.10-1.91, p = 0.003, n = 13 arms, 280 subjects). In contrast, exercise training did not significantly affect myostatin levels (SMD - 0.45, 95% CI - 1.07 to 0.18, p = 0.160, n = 11 arms, 283 subjects). Additionally, exercise training significantly reduced fasting blood glucose (SMD - 0.61, 95% CI - 0.89 to - 0.33, p < 0.001, n = 28 arms, 696 subjects), insulin (SMD - 0.80, 95% CI - 1.11 to - 0.50, p < 0.001, n = 24 arms, 566 subjects), homeostatic model assessment for insulin resistance (SMD - 0.75, 95% CI - 1.08 to - 0.43, p < 0.001, n = 28 arms, 609 subjects), hemoglobin A1C (HbA1C) (SMD - 0.96, 95% CI - 1.23 to - 0.70, p < 0.001, n = 12 arms, 275 subjects), and low-density lipoprotein cholesterol (SMD - 0.38, 95% CI - 0.74 to - 0.02, p = 0.040, n = 18 arms, 443 subjects). Exploratory subgroup analysis showed significant increases in irisin following resistance training (SMD 0.88 [95% CI, 0.44 to 1.33], [p = 0.001], n = 21 arms, 537 subjects, I2 = 79% [p = 0.001]), high-intensity interval training (SMD 0.61 [95% CI, 0.13 to 1.09], [p = 0.001], n = 17 arms, 346 subjects, I2 = 75% [p = 0.001]), and concurrent training (SMD 0.42 [95% CI, 0.16 to 0.68], [p = 0.002], n = 18 arms, 356 subjects, I2 = 21% [p = 0.20]). Although resistance training demonstrated numerically larger effects, differences between exercise modalities were not statistically significant (p > 0.05). Regular exercise is an effective intervention for increasing circulating irisin levels and favorably modulating other myokines such as follistatin and FGF-21 in adults with overweight or obesity. Resistance training showed larger numerical effects, but the differences between exercise types were not statistically significant. These adaptations, alongside improvements in metabolic markers, support the role of structured exercise as part of a comprehensive strategy for improving metabolic health in this population. PROSPERO CRD42025637476.
Machine learning (ML) and artificial intelligence (AI) offer opportunity and risk in mass trauma response, disasters and crisis. This narrative review synthesizes material from our "AI to the Rescue" panel at the inaugural PreAct Mass Trauma conference in June 2025, integrating relevant literature and the authors' expertise. We examine AI approaches beyond large language models (LLMs), including traditional ML and multimodal systems, while grounding the concept of "AI-made disasters" as a necessary third disaster type alongside Human-made and Natural, supported by emerging evidence of AI-caused psychiatric harm. We present the AI Safety Levels for Mental Health (ASL-MH) framework with six levels - from supportive applications, to autonomous packages, to experimental, high-risk systems - positioned as a practical heuristic for graduated risk governance given the nascent regulatory landscape and the demonstrated fragility of voluntary industry safety commitments. Using the Model for Adaptive Response to Complex Cyclical Disasters (MARCCD) framework, we organize AI applications across four phases: Anticipation, Impact, Adaptation, and Growth & Recovery, with attention to core disaster mental health sequelae and the challenge of differentiating normative distress from psychopathology. Recommendations address research/evidence, governance/regulation, training/literacy, and equity/access. Given our presentation involved live demos of AI applications, we have distilled key elements into this review which cannot be directly shown.
Shp1 is a cytosolic tyrosine phosphatase generally associated with antitumor effects through the inhibition of tyrosine kinase signaling. Herein, we shown that genetic and pharmacological inhibition of Shp1 in breast cancer cells induces accelerated cell migration and promotes a more invasive phenotype. Furthermore, we found that interleukin-8 (IL8), a chemokine with multiple pro-tumorigenic roles within the tumor microenvironment, directly modulates Shp1 activity. In breast cancer, IL8 elicits its functions through the binding to the CXCR2 receptor with the subsequent modulation of several intracellular signaling pathways. We show that in breast MCF7 cells, IL8 induces the PKC-mediated phosphorylation of Shp1 at Ser591, diminishing its enzymatic activity and impairing the dephosphorylation of PP2A; this enhances CXCR2 phosphorylation and alters receptor trafficking by promoting ubiquitination and degradation of CXCR2. This feedback mechanism limits IL8 signaling revealing a previously unrecognized mechanism of receptor turnover and signal attenuation. In addition, we found that Shp1-mediated regulation of CXCR2 directly influences IL8-driven invasiveness in a subtype-specific manner, affecting luminal and triple-negative breast cancer (TNBC) cells but not HER2-positive ones. Transcriptomic and pathway analyses further support Shp1 involvement in cytokine and GPCR signaling, particularly in TNBC, where its downregulation correlates with reduced survival and higher IL8 levels. Taken together, our findings elucidate a novel mechanism of IL8 signaling and identify Shp1 as a promising therapeutic target, highlighting the potential of modulating the CXCR2-Shp1 axis to limit invasiveness and metastasis in aggressive breast cancer subtypes, particularly TNBC.
Large Language Models (LLMs) have demonstrated impressive performance across various natural language processing (NLP) tasks, including text summarization, classification, and generation. Despite their success, LLMs are primarily trained on curated datasets that lack human-induced errors, such as typos or variations in word choice. As a result, LLMs may produce unexpected outputs when processing text containing such perturbations. In this paper, we investigate the resilience of LLMs to two types of text perturbations: typos and word substitutions. Using two public datasets, we evaluate the impact of these perturbations on text generation using six state-of-the-art models, including GPT-4o and LLaMA3.3-70B. Although previous studies have primarily examined the effects of perturbations in classification tasks, our research focuses on their impact on text generation. The results indicate that LLMs are sensitive to text perturbations, leading to variations in generated outputs, which have implications for their robustness and reliability in real-world applications.
Canine vector-borne pathogens (CVBPs) are increasingly relevant, particularly those of zoonotic concern, due to climate change and increased animal mobility. However, there is no information on the actual burden of these infections in canine populations in several Western and South-Central Asian countries, including Afghanistan. Therefore, this study aimed to investigate the molecular prevalence of CVBP infections in stray and owned dogs across regions of Afghanistan. From July 2020 to August 2024, a total of 100 dogs with outdoor lifestyles in four provinces of Afghanistan (i.e., Kabul, Kunduz, Mazar-i-Sharif, and Takhar) were blood-sampled and molecularly screened for Hepatozoon spp., Babesia spp., Leishmania spp., filarioid helminths, and Bartonella spp. Overall, 81% of dogs tested positive for at least one VBP. Hepatozoon canis was the most common (68%; present in all provinces), followed by Bartonella vinsonii subsp. berkhoffii (18%; present in all provinces), Babesia vogeli (4%; present in 3 provinces), Acanthocheilonema sp. (4%; present in 3 provinces), Leishmania infantum (2%; in Kunduz), and Babesia negevi (1%; in Mazar-i-Sharif). Co-infections were detected in 12% of dogs. To the best of our knowledge, this is the first epidemiological study of CVBPs in Afghanistan, demonstrating the sympatric circulation of H. canis, Bartonella vinsonii berkhoffii, L. infantum, B. vogeli, B. negevi, and Acanthocheilonema sp. Overall, the findings underscore the importance of endoparasite and ectoparasite control in owned dogs, as well as the need to control feral animal populations to reduce VBP transmission.
The present erratum is intended to correct some typos in our paper [J. Opt. Soc. Am. A39, 1774 (2022)JOAOD60740-323210.1364/JOSAA.466318].
Integrating biomedical datasets is hindered by inconsistent metadata, where the same concept may be represented in many ways (e.g. "Ca.," "Carcinoma," "tumor" for "Neoplasm"). Metadata harmonization automatically converts these researcher-specific terms into standard vocabulary terms to enable downstream integration. Current solutions, such as Common Data Elements and laboratory information management systems, either require navigating thousands of subtly different terms or disrupt researcher workflows, resulting in siloed datasets. This fragmentation forces researchers to spend over 40% of curation time on manual standardization. We present a language model-based harmonization solution that automatically maps researcher-specific metadata to standard terms across domains including cancer, alcohol research, and infectious disease. Our method fine-tunes GPT-2 models with realistic data augmentation, generating term variations that mimic researchers' documentations, such as typos, abbreviations, and word reordering. This enables harmonization even in domains without curated synonym sets. Fine-tuned models achieve 96% in-dictionary accuracy, reducing manual effort by over 90% when the term exists in the vocabulary, and 17% out-of-dictionary accuracy for previously unseen standards, outperforming traditional heuristics and zero-shot GPT-4o. Larger general models provide modest gains for unseen terms, while domain-specific small models achieve superior performance on specialized terminology, delivering a scalable, low-burden solution for harmonizing biomedical metadata and accelerating downstream data integration. All datasets used in this study, including training, validation, and test splits, are available via the Netrias Hugging Face organization. This includes datasets for the cancer and alcohol-bacteria mix domains used to develop and evaluate harmonization models. Experiment results are provided in the Supplementary Materials. We also share one representative GPT-2 Large cancer model and five GPT-2 Large models trained on different alcohol-bacteria domain mixtures: (100/0, 75/25, 50/50, 25/75, 0/100). All resources are released under the Apache 2.0 license to support reproducibility and reuse.
In countries with access to the electronic health record (EHR), both patients and healthcare professionals have reported finding errors in the EHR, so-called EHRrors. These can range from simple typos to more serious cases of missing or incorrect health information. Despite their potential detrimental effect, the evidence on EHRrors has not been systematically analysed. It is unknown how common EHRrors are or how they impact patients and healthcare professionals. A mixed systematic review will be carried out to address the research gap. We will search PubMed, Web of Science and CINAHL for studies published since 2000, which report original research data on patient-identified and healthcare professional-identified EHRrors. We will analyse (1) the prevalence of EHRrors, (2) the types of EHRrors and (3) their impact on care. Quantitative and qualitative findings will be synthesised following the Joanna Briggs Institute Framework for Mixed Systematic Reviews. Identified studies will be critically appraised for meta-biases and risk of bias in individual studies. The confidence in the emerging evidence will be further assessed through the Grading of Recommendations Assessment, Development and Evaluation approach. Findings will be contextualised and interpreted involving an international team of patient representatives and practising healthcare professionals. The study will not involve collection or analysis of individual patient data; thus, ethical approval is not required. Results will be published in a peer-reviewed publication and further disseminated through scientific events and educational materials. CRD42024622849.
The AOAC Expert Review Panel (ERP) approved a method for the quantification of folic acid in various dietary supplement dosage forms containing tablets, 2-piece capsules, powder drinks, softgels, and gummies with First Action Official MethodSM status. The previously published method summarized a single-laboratory validation with parameters of linearity, LOD, LOQ, repeatability, recovery, specificity, and system suitability. Based on the request from the ERP, the recovery test for the gummies has been reperformed with a revised procedure. Determination of Folic Acid in Various Dietary Supplement Dosage Forms UPLC/PDA. The recovery range of 94.6-106.5% was achieved by spiking 20, 50, and 80% in the gummy samples. Other adjustments or clarification of the method and minor typos were also addressed. After the revised method, report, and results were analyzed and discussed, the ERP adopted the method and provided recommendations for achieving Final Action status. The revised method meets the requirements of Standard Method Performance Requirement (SMPR®) 2022.002.
As the field of glycobiology has developed, so too have different glycan nomenclature systems. While each system serves specific purposes, this multiplicity creates challenges for usability, data integration, and knowledge sharing across different databases and computational tools. We present a practical framework for automated nomenclature conversion that takes any glycan nomenclature as input without requiring declaration of the specific language and outputs a canonicalized IUPAC-condensed format as a standardized representation. Our implementation handles all common nomenclatures including WURCS, GlycoCT, IUPAC-condensed/extended, GLYCAM, CSDB-linear, LinearCode, GlycoWorkbench, GlySeeker, Oxford, and KCF, along with common typos, and manages complex cases including structural ambiguities, modifications, uncertainty in linkage information, and different compositional representations. This Universal Input framework can translate more than 10 nomenclatures in <1 ms per glycan, tested on over 150 000 sequences with 98%-100% coverage, enabling seamless integration of existing glycan databases and tools while maintaining the specific advantages of each representation system. Universal Input is implemented within the glycowork Python package, available at https://github.com/BojarLab/glycowork and our web app https://canonicalize.streamlit.app/.
This study investigates the impact of internet access on creativity and identifies potential hidden costs of internet use for groups. Using the alternative uses task, we randomized participants (N = 244) into separate conditions to generate ideas for nonstandard uses for one of two common objects-a shield or an umbrella-either with or without internet access. Nominal group analysis reveals that while individual creativity may be enhanced by internet access, groups articulate fewer novel solutions when provided internet access, suggesting that internet access may constrain collective creative fluency. We also ran a reanalysis of previous data sets on creativity and internet use and found robust converging evidence across different paradigms, coders, and contexts. We further explore robustness by examining alternative operationalizations of fluency: quality of responses, as measured by coders' evaluations of effectiveness, novelty, and subjective evaluations of creativity. While overall trends suggest an advantage for subjects who do not have internet access, this patterning depends to some degree on variation among coders. Implications for the way digital tools influence creative processes are discussed.
Passive resonators have been widely used in MRI to manipulate RF field distributions. However, optimizing these structures using full-wave electromagnetic (EM) simulations is computationally prohibitive, particularly for massive-element passive resonator arrays with many degrees of freedom. While the EM and RF circuit co-simulation method has previously been applied to RF coil design, this work presents, for the first time, a co-simulation framework tailored specifically for the analysis and optimization of passive resonators. The framework performs a single full-wave EM simulation in which the resonator's lumped components are replaced by ports, followed by circuit-level computations to evaluate arbitrary capacitor/inductor configurations. This allows integration with a genetic algorithm to rapidly optimize the resonator parameters to enhance B 1 fields in a targeted region of interest (ROI). The proposed method was validated across three scenarios of increasing complexity: (1) a single-loop passive resonator on a spherical phantom, (2) a two-loop array on a cylindrical phantom, and (3) a two-loop array on a human head model. In all cases, the co-simulation results showed excellent agreement with full-wave EM simulations, with relative errors below 1%. The genetic-algorithm-driven optimization, involving tens of thousands of capacitor combinations, completed in under 5 minutes-whereas equivalent full-wave EM sweeps would require an impractically long computation time. This work extends co-simulation methodology to passive resonator design for the first time, enabling fast, accurate, and scalable optimization. The approach significantly reduces computational burden while preserving full-wave accuracy, making it a powerful tool for passive RF structure development in MRI.
The administration of surfactant aerosol therapy to preterm infants receiving continuous positive airway pressure (CPAP) respiratory support is highly challenging due to small flow passages, relatively high ventilation flow rates, rapid breathing and small inhalation volumes. To overcome these challenges, the objective of this study was to implement a validated computational fluid dynamics (CFD) model and develop an overlay nasal prong interface design for use with CPAP respiratory support that enables high efficiency powder aerosol delivery to the lungs of preterm infants when needed (i.e., on-demand) and can remain in place without increasing the work of breathing compared with a baseline CPAP interface. Realistic in vitro experiments were first conducted to generate baseline validation data, and then the CFD model, once validated, was used to explore key design parameters across a range of preterm infant nose-throat geometries and aerosol delivery conditions. The most important factors for efficient aerosol delivery were shown to be (i) maintaining the aerosol delivery flow rate below the tracheal flow rate (to minimize CPAP line loss) and (ii) concentrating the aerosol within the first portion of the inhalation waveform. An optimized design was shown to deliver approximately 37-60% of the nominal dose through the system and to the lungs with low intersubject variability (1050-2200 g infants) across two modes of device actuation (automated and manual) with room for further improvement. Ergonomic curvatures and streamlining of the prong geometries were also found to reduce work of breathing and flow resistance compared with a commercial alternative.