Public Metrics Research Articles

Causal Discovery from Temporal Data: An Overview and New Perspectives

Temporal data, representing chronological observations of complex systems, has always been a typical data structure that can be widely generated by many domains, such as industry, finance, healthcare and climatology. Analyzing the underlying structures, i.e. , the causal relations, could be extremely valuable for various applications. Recently, causal discovery from temporal data has been considered as an interesting yet critical task and attracted much research attention. According to the nature and structure of temporal data, existing causal discovery works can be divided into two highly correlated categories i.e. , multivariate time series causal discovery, and event sequence causal discovery. However, most previous surveys are only focused on the multivariate time series causal discovery but ignore the second category. In this paper, we specify the similarity between the two categories and provide an overview of existing solutions. Furthermore, we provide public datasets, evaluation metrics and new perspectives for temporal data causal discovery.

Twitter data as a linked digital phenotype in the Avon Longitudinal Study of Parents and Children

This data note describes the collection and linkage of participants' Twitter data as a digital phenotype in the Avon Longitudinal Study of Parents and Children (ALSPAC) multi-generational birth cohort study. Twitter (renamed X in 2023) is a social media platform based around a micro-blog format. Digital phenotyping represents a novel opportunity for cohort studies to collect data with a low participant burden, and outside of discrete measurement periods. The ALSPAC governance framework supports the ethical consenting, storage and sharing of social media data, and linking Twitter data with wider cohort data provides opportunities to assess Twitter data quality concerns in a research context. All adults currently participating in ALSPAC (N=26,205) were invited to take part, which included the index cohort and their parents. N=3,247 indicated that they were Twitter users, 26% of these (N=835) consented and 19% (N=623) had their data successfully linked. Data were collected using our open-source software, Epicosm in February 2023. Approximately two thirds of the linked Twitter cohort are from the index cohort generation, and the remainder from the parent generation. In general, linked participants are representative of the general ALSPAC cohort, with the exception of having slightly higher educational attainment. This is consistent with previous research into the demographics of Twitter users. Overall the linked dataset contains 1,488,517 posts (tweets) from between 2008 and 2023, with 27% of these being 'retweets'. The available data includes information derived from a range of commonly used sentiment scoring algorithms, type of tweet, public metrics such as likes and retweets, and the time and date of the tweet. Controls are in place to maintain the anonymity of cohort participants, and data linkage is managed by ALSPAC’s data linkage team to reduce disclosure risk. This ensures high standards of data security and ethical use of social media data.

Morphology of mergers and their impact on performance and accounting metrics in HEIs: evidence from a small European country

ABSTRACT The purpose of this study is to analyze the structural elements from mergers in higher education institutions (HEIs) of a small European country, Greece, and evaluate how merger morphology is related to performance criteria and ratios imposed by the Greek state. More specifically, this study provides a depiction of ten mergers of HEIs in Greece that were introduced by the state in order to absorb the previously existent thirteen technological educational institutes (TEIs) from Greek universities. The morphology of mergers is provided by examining several factors, such as: type of merger (merger by absorption or consolidation); type and number of entity involvement (two-partner or multi-partner); geographical area of newly formed HEIs (single-campus or multi-campus), with provision of number of campuses in different cities, etc. These merger characteristics are associated with the HEI performance in the post-merger era, and their performance is evaluated with public accounting metrics and funding on specific ratios imposed by the Greek state. The results provide evidence that public HEIs in Greece throughout the process of joining entities have demonstrated different levels of synergies after mergers, in terms of state funding, with better results for the case where there are campuses in three different cities. Also, this study supports that a state initiative for legislative intervention in HEI mergers could lead to different levels of synergies. Last, this study tries to provide new knowledge for the gap of economic impact of mergers’ characteristics of HEIs and offers an open framework for further research on a fruitful area.

Regional variation of age-related blood pressure trajectories and their association with blood pressure control rates, disability-adjusted life years and death

Abstract Background/Introduction While the relationship between blood pressure (BP) changes and increasing age is well described, age-related BP trajectories may vary substantially across different populations and associated with differential CV risk and outcomes. Purpose The dataset of the May Measurement Month (MMM) international campaigns enables a unique investigation of regional, socio-economic and sex differences in age-BP trajectories and their association with major public-health outcomes. Method Global BP data from more than 4,5 million adult (≥18 years) screenees obtained during the cross-sectional MMM campaigns in 2017, 2018, 2019, and 2021 were used for analysis. Average BP and slopes of BP changes across age groups were calculated within regional, socio-economic and sex strata. BP slopes of individuals between 30 and 70 years of age were derived from adjusted linear regression models, along with baseline BP estimates at young age. These BP slopes were correlated in regression models adjusted for baseline BP at young age with country-specific all-cause rates of DALYs (Disability-Adjusted Life Years), death (both from Global Burden of Disease Study 2019), life expectancy (World Bank data SP.DYN.LE00.IN 2019) and BP control rates (NCD Risk Factor Collaboration). Results Age-related BP changes differed distinctly across global geographic regions, income levels, and between sexes. The steepest increase in BP with age was observed in populations from Africa and Europe, whereas a more gradual rise in BP with increasing age was seen in South-East Asian populations. Women had lower BP levels than men at younger ages (twenties and thirties), but subsequently experienced a more pronounced age-related BP increase resulting in diminishing sex differences across older age groups. Geographically divergent age-related BP trajectories were significantly associated with major national public health indicators. Globally, the steeper the age-related BP slope, the stronger the association with poor BP control (β = -0.46 [95% CI: -0.74; -0.19], p=0.001), increased DALY’s (β = 66155.68 [95% CI: 44847.0; 87464.36], p<0.001) and death rates. (β = 1017.3 [95% CI: 351.72; 1682.89], p=0.003). Conclusion Age-related BP trajectories vary considerably across global populations and contribute considerably to the variability in BP-related risks and adverse outcomes across regions. Effective public health strategies may require region-specific targeting of lifestyle and other factors contributing to adverse BP trajectories to improve health outcomes.Average blood pressure by ageAssiciations with public health metrics

#Storytime: An analysis of COVID-19 origin narratives and conspiracies on TikTok

ABSTRACT This study delves into the construction, communication, and reception of crisis narratives on TikTok, specifically focusing on videos discussing the purported origins of COVID-19 via the hashtag #covidorigin. Employing a content analysis approach, this research examines patterns in content types and their relationship with public engagement metrics such as views, likes, comments, and shares. This study observed that popular TikTok videos discussing the origin of COVID-19 majorly incorporated at least one of the following: a first-hand narrative (a story in the first-person), a secondhand narrative (an account someone heard from other people), and/or an informational report (e.g., legislative documents, news reports, and book summaries). Additionally, while U.S. intelligence agencies have yet to reach a consensus on COVID-19’s origin, this research detected a surge in #covidorigin videos following a 2023 interview in which FBI Director Wray stated that COVID-19 likely originated from a laboratory incident in China. Overall, this study suggests the influential role public leaders and media sources have in disseminating COVID-19 information and the importance of coherent crisis narratives in shaping public understanding. It also considers the effects of content types and communication formats in terms of how the public engages with speculative claims online.

Controllable image synthesis methods, applications and challenges: a comprehensive survey

Controllable Image Synthesis (CIS) is a methodology that allows users to generate desired images or manipulate specific attributes of images by providing precise input conditions or modifying latent representations. In recent years, CIS has attracted considerable attention in the field of image processing, with significant advances in consistency, controllability and harmony. However, several challenges still remain, particularly regarding the fine-grained controllability and interpretability of synthesized images. In this paper, we comprehensively and systematically review the CIS from problem definition, taxonomy and evaluation systems to existing challenges and future research directions. First, the definition of CIS is given, and several representative deep generative models are introduced in detail. Second, the existing CIS methods are divided into three categories according to the different control manners used and discuss the typical work in each category critically. Furthermore, we introduce the public datasets and evaluation metrics commonly used in image synthesis and analyze the representative CIS methods. Finally, we present several open issues and discuss the future research direction of CIS.

Deep learning for surgical workflow analysis: a survey of progresses, limitations, and trends

Automatic surgical workflow analysis, which aims to recognize the ongoing surgical events in videos, is fundamental for developing context-aware computer-assisted systems. This paper reviews representative surgical workflow recognition algorithms based on deep learning, outlining their merits, limitations, and future research directions. The literature survey was performed on three large bibliographic databases, covering 67 lary sources, which were comparatively analyzed in terms of spatial feature modeling, spatio-temporal feature modeling, input pre-processing, regularization and post-processing algorithms, as well as learning strategies. Then, common public datasets and evaluation metrics for surgical workflow recognition are also described in detail. Finally, we discuss all literature from different perspectives, and point out the challenges, possible solutions and future trends. The need for more diverse and larger datasets, the potential of unsupervised and semi-supervised learning approaches, comprehensive and equitable metrics, establishing complete regulatory and data standards, and interoperability will be key challenges in translating models to clinical operating rooms. And we propose that surgical activity anticipation and employing large language model as training assistant are interesting research directions in surgical workflow analysis.

Project SIGNAL: A Dashboard for Supporting Community Confidence in Making Data-Driven Decisions.

Data dashboards have emerged as critical tools for surveillance and informing resource allocation. Despite their utility and popularity during COVID-19, there is a growing need to understand what tools and training are tailored to nonprofit community-based organizations that may partner with public health officials. In June 2021, the Rhode Island Department of Health and Brown University partnered to create Project SIGNAL (Spatiotemporal Insights to Guide Nuanced Actions Locally), which utilizes spatiotemporal analytics to identify Rhode Island's largest disparities in COVID-19-related outcomes (eg, testing, diagnosis, vaccinations) at the neighborhood level. Results were hosted in an interactive online dashboard (signal-ri.org) designed using principles of the CDC Clear Communication Index. The target audience included a network of 15 geographic areas called Health Equity Zones, funded by the health department to provide critical grassroots public health programs to address social, health, and economic outcomes in their communities. To disseminate the dashboard, a 6-hour virtual workshop series was created to train leaders to use the dashboard and increase their confidence in understanding common public health data terminology and concepts and better prepare attendees for rapid decision making during future public health emergencies. The Project SIGNAL dashboard was launched in August 2022 and has been accessed over 7500 times. A total of 84 community leaders were trained to use this dashboard, increasing their confidence in applying common public health metrics to make decisions about their COVID-19-related activities. While several studies have outlined best practices for data dashboards, this is among the first to examine incorporating these practices into a spatiotemporal decision tool designed specifically for community organizations. Project SIGNAL demonstrates that by incorporating design best practices and pairing data dashboards with hands-on training, we can empower community leaders to utilize advanced spatiotemporal methods to identify health disparities and take localized action.

Epidemiological study on Covid-19 frequency, severity, and territorial spreading

This study analyzes the descriptive epidemiology and testing performance in India has been thoroughly analyzed, revealing a robust and effective approach to reporting and managing COVID-19 cases. COVID-19 scenario reports from January 22, 2020 and April 30, 2020 in India. This research examines COVID-19 incidence, severity and spread. The examination frequency was explained spatially and temporally. Along with public health metrics and events, we showed trends and characterized instances by period location and individuals. During January 22, 2020 and April 30, 2020, people were tested for SARS-CoV-2; 40 184 (3,9 %) of the tests produced positive findings in India. Age-specific assault rates were lowest among children under 10 and highest among those aged 50-69. Positive instances were more common in encounters with complaints and in non-responsive people than in patients with SARI, overseas travel experiences or healthcare professionals. Males had a greater assault rate (41,64 %) over females (24,32 %). Six percent of attacks were secondary. In all, 71,1 % of 736 districts reported COVID-19 instances and 99 % percent of districts confirmed treatment.

Integrating Sexual and Reproductive Health Equity Into Public Health Goals and Metrics: Comparative Analysis of Healthy People 2030's Approach and a Person-Centered Approach to Contraceptive Access Using Population-Based Data.

The Healthy People initiative is a national effort to lay out public health goals in the United States every decade. In its latest iteration, Healthy People 2030, key goals related to contraception focus on increasing the use of effective birth control (contraceptive methods classified as most or moderately effective for pregnancy prevention) among women at risk of unintended pregnancy. This narrow focus is misaligned with sexual and reproductive health equity, which recognizes that individuals' self-defined contraceptive needs are critical for monitoring contraceptive access and designing policy and programmatic strategies to increase access. We aimed to compare 2 population-level metrics of contraceptive access: a conventional metric, use of contraceptive methods considered most or moderately effective for pregnancy prevention among those considered at risk of unintended pregnancy (approximating the Healthy People 2030 approach), and a person-centered metric, use of preferred contraceptive method among current and prospective contraceptive users. We used nationally representative data collected in 2022 to construct the 2 metrics of contraceptive access; the overall sample included individuals assigned female at birth not using female sterilization or otherwise infecund and who were not pregnant or trying to become pregnant (unweighted N=2760; population estimate: 43.9 million). We conducted a comparative analysis to examine the convergence and divergence of the metrics by examining whether individuals met the inclusion criteria for the denominators of both metrics, neither metric, only the conventional metric, or only the person-centered metric. Comparing the 2 approaches to measuring contraceptive access, we found that 79% of respondents were either included in or excluded from both metrics (reflecting that the metrics converged when individuals were treated the same by both). The remaining 21% represented divergence in the metrics, with an estimated 5.7 million individuals who did not want to use contraception included only in the conventional metric denominator and an estimated 3.5 million individuals who were using or wanted to use contraception but had never had penile-vaginal sex included only in the person-centered metric denominator. Among those included only in the conventional metric, 100% were content nonusers-individuals who were not using contraception, nor did they want to. Among those included only in the person-centered metric, 68% were currently using contraception. Despite their current or desired contraceptive use, these individuals were excluded from the conventional metric because they had never had penile-vaginal sex. Our analysis highlights that a frequently used metric of contraceptive access misses the needs of millions of people by simultaneously including content nonusers and excluding those who are using or want to use contraception who have never had sex. Documenting and quantifying the gap between current approaches to assessing contraceptive access and more person-centered ones helps clearly identify where programmatic and policy efforts should focus going forward.

Disparities in public transport accessibility in London from 2011 to 2021

Addressing urban inequalities has become a pressing concern on both the global sustainable development agenda and for local policy. Improving public transport services is seen as an important area where local governments can exert influence and potentially help reduce inequalities. Existing measures of accessibility used to inform decision-making for public transport infrastructure in London show spatial disparities, yet there is a gap in understanding how these disparities vary across demographic groups and how they evolve over time—whether they are improving or worsening. In this study, we investigate the distribution of public transport accessibility based on ethnicity and income deprivation in London over the past decade. We used data from the Census 2011 and 2021 for area-level ethnicity characteristics, English Indices of Deprivation for income deprivation in 2011 and 2019, and public transport accessibility metrics from Transport for London for 2010 and 2023, all at the small area level using lower super output areas (LSOAs) in Greater London. We found that, on average, public transport accessibility in London has increased over the past decade, with 78% of LSOAs experiencing improvements. Public transport accessibility in London showed an unequal distribution in cross-sectional analyses. Lower income neighbourhoods had poorer accessibility to public transportation in 2011 and 2023 after controlling for car-ownership and population density. These disparities were particularly pronounced for underground accessibility. Temporal analyses revealed that existing inequalities with respect to income deprivation and ethnicity are generally not improving. While wealthier groups benefited most from London Underground service improvements; lower income groups benefited more from bus service improvements. We also found that car ownership levels declined in areas with substantial increases to public transport accessibility and major housing developments, but not in those with moderate improvements.

A Survey on Surface Defect Inspection Based on Generative Models in Manufacturing

Surface defect inspection based on deep learning has demonstrated outstanding performance in improving detection accuracy and model generalization. However, the small scale of defect datasets always limits the application of deep models in industry. Generative models can obtain realistic samples in a very cheap way, which can effectively solve this problem and thus has received widespread attention in recent years. This paper provides a comprehensive analysis and summary of the current studies of surface defect inspection methods proposed between 2022 and 2024. First, according to the use of generative models, these methods are classified into four categories: Variational Auto-Encoders (VAEs), Generative Adversarial Networks (GANs), Diffusion Models (DMs), and multi-models. Second, the research status of surface defect inspection based on generative models in recent years is discussed from four aspects: sample generation, detection objective, inspection task, and learning model. Then, the public datasets and evaluation metrics that are commonly used for surface defect inspection are discussed, and a comparative evaluation of defect inspection methods based on generative models is provided. Finally, this study discusses the existing challenges for the defect inspection methods based on generative models, providing insights for future research.

Efficient artificial intelligence approaches for medical image processing in healthcare: comprehensive review, taxonomy, and analysis

In healthcare, medical practitioners employ various imaging techniques such as CT, X-ray, PET, and MRI to diagnose patients, emphasizing the crucial need for early disease detection to enhance survival rates. Medical Image Analysis (MIA) has undergone a transformative shift with the integration of Artificial Intelligence (AI) techniques such as Machine Learning (ML) and Deep Learning (DL), promising advanced diagnostics and improved healthcare outcomes. Despite these advancements, a comprehensive understanding of the efficiency metrics, computational complexities, interpretability, and scalability of AI based approaches in MIA is essential for practical feasibility in real-world healthcare environments. Existing studies exploring AI applications in MIA lack a consolidated review covering the major MIA stages and specifically focused on evaluating the efficiency of AI based approaches. The absence of a structured framework limits decision-making for researchers, practitioners, and policymakers in selecting and implementing optimal AI approaches in healthcare. Furthermore, the lack of standardized evaluation metrics complicates methodology comparison, hindering the development of efficient approaches. This article addresses these challenges through a comprehensive review, taxonomy, and analysis of existing AI-based MIA approaches in healthcare. The taxonomy covers major image processing stages, classifying AI approaches for each stage based on method and further analyzing them based on image origin, objective, method, dataset, and evaluation metrics to reveal their strengths and weaknesses. Additionally, comparative analysis conducted to evaluate the efficiency of AI based MIA approaches over five publically available datasets: ISIC 2018, CVC-Clinic, 2018 DSB, DRIVE, and EM in terms of accuracy, precision, Recall, F-measure, mIoU, and specificity. The popular public datasets and evaluation metrics are briefly described and analyzed. The resulting taxonomy provides a structured framework for understanding the AI landscape in healthcare, facilitating evidence-based decision-making and guiding future research efforts toward the development of efficient and scalable AI approaches to meet current healthcare needs.

Evaluating exposure-response relationship in 1,3-butadiene and leukemia studies.

1,3-Butadiene (BD) exposure's link to leukemia is under regulatory scrutiny. The assessment methods for BD exposure risks have evolved from early animal and limited human studies to advanced exposure-response modelling with comprehensive quantitative data. The objec- tive of this study is to explore the nuances of exposure-response modelling, investigating how various statistical methods have influenced the quan- tification of exposure-response relationships. Although this study was not conducted as a formal systematic review, a search was performed in Medline/Pubmed to identify all human studies on leukemia risk assessment for BD exposure. This search included articles written in English. The electronic search spanned from inception of records until July 23, 2023, using the search term: "butadiene AND (leukaemia OR leukemia OR myeloid OR lymphoid)" and was restricted to human species. Focusing on the synthetic styrene-butadiene rubber (SBR) industry cohort study conducted by the University of Alabama at Birmingham, USA, this review evaluates various statistical models and factors influenc- ing exposure-response modelling. Peak exposures to BD may be more influential in the dose-response relationship than cumulative or long-term exposure. The authors recommend utilizing β-coefficients derived from the latest SBR study update, employing Cox proportional hazard modelling, non-lagged and non-transformed cumulative BD exposure, and adjusting for age and peak BD exposure. The study reveals that statistical model selection has a limited impact on the calculated dose-response effects. The significant variation in estimated cancer mortality values arises from additional assumptions needed for metrics like the excess leukemia risk or the occupational BD effective concentration. In con- clusion, this study provides insights into exposure-response modelling for BD exposure and leukemia mortality, highlighting the importance of peak exposures. The recommended statistical approach offers a reliable basis for regulatory risk assessment and public health population metrics. Int J Occup Med Environ Health. 2024;37(3):300-10.

Integrating community-based partnerships for enhanced health outcomes: A collaborative model with healthcare providers, clinics, and pharmacies across the USA

The Community Health Enhancement Initiative represents a significant advancement in the field of public health within the United States, targeting the improvement of health outcomes through a comprehensive, collaborative approach. This initiative, in partnership with an array of healthcare entities including renowned hospitals (e.g., Cleveland Clinic, Mayo Clinic, Johns Hopkins Hospital), federally qualified community health centers (e.g., Henry J. Austin Health Center, Chase Brexton Health Care), primary care practices (e.g., Kaiser Permanente, Geisinger Health System, Providence Medical Group), specialized clinics (e.g., Dana-Farber Cancer Institute, Cleveland Clinic Heart and Vascular Institute), and pharmacies with healthcare services (e.g., CVS Health, Walgreens, Rite Aid), aims to revolutionize community health services across the nation. This paper delves into the strategic framework of the initiative, which is designed to foster health literacy, disease prevention, and wellness in diverse American communities. It examines how the collaborative model not only leverages the strengths of various healthcare providers to offer comprehensive care but also actively involves communities in their health management through educational workshops and accessible services. By bridging the gap between healthcare providers and the public, the initiative seeks to build a more health-literate society that is better equipped to manage and prevent diseases. Furthermore, the paper evaluates the impact of this nationwide initiative on public health metrics, showcasing evidence-based outcomes and the potential for scalability in different regions. The initiative's alignment with national health objectives, its innovative use of partnerships to extend healthcare access, and its emphasis on health literacy as a pivotal component of community well-being underscore its relevance and importance to the broader health policy landscape in the U.S. In addition, the research addresses the challenges and opportunities presented by this model, including the integration of technology in healthcare delivery, the sustainability of partnerships, and the continuous engagement of communities. It concludes with policy recommendations for enhancing the initiative's effectiveness and potential for replication, contributing to the ongoing discourse on improving public health infrastructure and outcomes across the United States.

Enhanced SARS-CoV-2 case prediction using public health data and machine learning models.

The goal of this study is to propose and test a scalable framework for machine learning (ML) algorithms to predict near-term severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) cases by incorporating and evaluating the impact of real-time dynamic public health data. Data used in this study include patient-level results, procurement, and location information of all SARS-CoV-2 tests reported in West Virginia as part of their mandatory reporting system from January 2021 to March 2022. We propose a method for incorporating and comparing widely available public health metrics inside of a ML framework, specifically a long-short-term memory network, to forecast SARS-CoV-2 cases across various feature sets. Our approach provides better prediction of localized case counts and indicates the impact of the dynamic elements of the pandemic on predictions, such as the influence of the mixture of viral variants in the population and variable testing and vaccination rates during various eras of the pandemic. Utilizing real-time public health metrics, including estimated Rt from multiple SARS-CoV-2 variants, vaccination rates, and testing information, provided a significant increase in the accuracy of the model during the Omicron and Delta period, thus providing more precise forecasting of daily case counts at the county level. This work provides insights on the influence of various features on predictive performance in rural and non-rural areas. Our proposed framework incorporates available public health metrics with operational data on the impact of testing, vaccination, and current viral variant mixtures in the population to provide a foundation for combining dynamic public health metrics and ML models to deliver forecasting and insights in healthcare domains. It also shows the importance of developing and deploying ML frameworks in rural settings.

Potential Integration of Metaverse, Non-Fungible Tokens and Sentiment Analysis in Quantitative Tourism Economic Analysis

With the emergence of the metaverse, countries’ digital efforts to create tourism opportunities have given rise to the possibility of capitalising on digital content which, along with physical tourism experiences, can generate further income and enhance a country’s reputation. Non-fungible tokens (NFTs), a unique application of blockchain technology, offer an enabling technology in several sectors, including tourism. Therefore, this study aims to explore the official tourism websites of Croatia and Slovenia and analyse current NFT applications in tourism economics. The methodology focuses explicitly on sentiment analysis, blockchain and machine learning. The paper introduces various applications currently in place, including Slovenia’s “I Feel Nft” project. The research shows that the main benefits of using NFT and sentiment analysis in the tourism economy are the promotion and presentation of major tourist destinations, exhibitions, works of art, and companies’ products in tokens, digital content and souvenirs. The adoption of sentiment analysis and NFTs in the tourism economy is still open to proposals for implementing public quantitative data metrics. Therefore, the scientific contribution of this research is essential in terms of operational recommendations and defining metrics for measuring the effectiveness of those methodologies and their applications in the tourism economy. On top of that, the practical contribution lies in monitoring the influx of tourists, and highlighting their increase over time and the significance of new technology in time series tourism research.

A pilot recruitment strategy to enhance ethical and equitable access to Covid-19 pediatric vaccine trials.

The SARS-CoV-2 pandemic disproportionately impacted communities with lower access to health care in the United States, particularly before vaccines were widely available. These same communities are often underrepresented in clinical trials. Efforts to ensure equitable enrollment of participants in trials related to treatment and prevention of Covid-19 can raise concerns about exploitation if communities with lower access to health care are targeted for recruitment. To enhance equity while avoiding exploitation, our site developed and implemented a three-part recruitment strategy for pediatric Covid-19 vaccine studies. First, we publicized a registry for potentially interested participants. Next, we applied public health community and social vulnerability indices to categorize the residence of families who had signed up for the registry into three levels to reflect the relative impact of the pandemic on their community: high, medium, and low. Finally, we preferentially offered study participation to interested families living in areas categorized by these indices as having high impact of the Covid-19 pandemic on their community. This approach allowed us to meet goals for study recruitment based on public health metrics related to disease burden, which contributed to a racially diverse study population that mirrored the surrounding community demographics. While this three-part recruitment strategy improved representation of minoritized groups from areas heavily impacted by the Covid-19 pandemic, important limitations were identified that would benefit from further study. Future use of this approach to enhance equitable access to research while avoiding exploitation should test different methods to build trust and communicate with underserved communities more effectively.

Learning From Noisy Labels With Deep Neural Networks: A Survey.

Deep learning has achieved remarkable success in numerous domains with help from large amounts of big data. However, the quality of data labels is a concern because of the lack of high-quality labels in many real-world scenarios. As noisy labels severely degrade the generalization performance of deep neural networks, learning from noisy labels (robust training) is becoming an important task in modern deep learning applications. In this survey, we first describe the problem of learning with label noise from a supervised learning perspective. Next, we provide a comprehensive review of 62 state-of-the-art robust training methods, all of which are categorized into five groups according to their methodological difference, followed by a systematic comparison of six properties used to evaluate their superiority. Subsequently, we perform an in-depth analysis of noise rate estimation and summarize the typically used evaluation methodology, including public noisy datasets and evaluation metrics. Finally, we present several promising research directions that can serve as a guideline for future studies.

A Review of Image Inpainting Methods Based on Deep Learning

Image Inpainting is an age-old image processing problem, with people from different eras attempting to solve it using various methods. Traditional image inpainting algorithms have the ability to repair minor damage such as scratches and wear. However, with the rapid development of deep learning in the field of computer vision in recent years, coupled with abundant computing resources, methods based on deep learning have increasingly highlighted their advantages in semantic feature extraction, image transformation, and image generation. As such, image inpainting algorithms based on deep learning have become the mainstream in this domain.In this article, we first provide a comprehensive review of some classic deep-learning-based methods in the image inpainting field. Then, we categorize these methods based on component optimization, network structure design optimization, and training method optimization, discussing the advantages and disadvantages of each approach. A comparison is also made based on public datasets and evaluation metrics in image inpainting. Furthermore, the article delves into the applications of current image inpainting technologies, categorizing them into three major scenarios: object removal, general image repair, and facial inpainting. Finally, current challenges and prospective developments in the field of image inpainting are discussed.