Embracing limited and imperfect training datasets: opportunities and challenges in plant disease recognition using deep learning
Ontology highlight
ABSTRACT:
SUBMITTER: Xu M
PROVIDER: S-EPMC10557492 | biostudies-literature | 2023 Jan
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Since the successful application of messenger RNA (mRNA) vaccines in preventing COVID-19, researchers have been striving to develop mRNA vaccines for clinical use, including those exploited for anti-tumor therapy. mRNA cancer vaccines have emerged as a promising novel approach to cancer immunotherapy, offering high specificity, better efficacy, and fewer side effects compared to traditional treatments. Multiple therapeutic mRNA cancer vaccines are being evaluated in preclinical and clinical trials, with promising early-phase results. However, the development of these vaccines faces various challenges, such as tumor heterogeneity, an immunosuppressive tumor microenvironment, and practical obstacles like vaccine administration methods and evaluation systems for clinical application. To address these challenges, we highlight recent advances from preclinical studies and clinical trials that provide insight into identifying obstacles associated with mRNA cancer vaccines and discuss potential strategies to overcome them. In the future, it is crucial to approach the development of mRNA cancer vaccines with caution and diligence while promoting innovation to overcome existing barriers. A delicate balance between opportunities and challenges will help guide the progress of this promising field towards its full potential.
| S-EPMC10511650 | biostudies-literature
Project description:The field of immunology is rapidly progressing toward a systems-level understanding of immunity to tackle complex infectious diseases, autoimmune conditions, cancer, and beyond. In the last couple of decades, advancements in data acquisition techniques have presented opportunities to explore untapped areas of immunological research. Broad initiatives are launched to disseminate the datasets siloed in the global, federated, or private repositories, facilitating interoperability across various research domains. Concurrently, the application of computational methods, such as network analysis, meta-analysis, and machine learning have propelled the field forward by providing insight into salient features that influence the immunological response, which was otherwise left unexplored. Here, we review the opportunities and challenges in democratizing datasets, repositories, and community-wide knowledge sharing tools. We present use cases for repurposing open-access immunology datasets with advanced machine learning applications and more.
| S-EPMC8086961 | biostudies-literature
Project description:Fluorescence microscopy is an important technique in many areas of biological research. Two factors that limit the usefulness and performance of fluorescence microscopy are photobleaching of fluorescent probes during imaging and, when imaging live cells, phototoxicity caused by light exposure. Recently developed methods in machine learning are able to greatly improve the signal-to-noise ratio of acquired images. This allows researchers to record images with much shorter exposure times, which in turn minimizes photobleaching and phototoxicity by reducing the dose of light reaching the sample. To use deep learning methods, a large amount of data is needed to train the underlying convolutional neural network. One way to do this involves use of pairs of fluorescence microscopy images acquired with long and short exposure times. We provide high-quality datasets that can be used to train and evaluate deep learning methods under development. The availability of high-quality data is vital for training convolutional neural networks that are used in current machine learning approaches.
| S-EPMC8099770 | biostudies-literature
Project description:Dengue vaccine development efforts have focused on the development of tetravalent vaccines. However, a recent Phase IIb trial of a tetravalent vaccine indicates a protective effect against only 3 of the 4 serotypes. While vaccines effective against a subset of serotypes may reduce morbidity and mortality, particular profiles could result in an increased number of cases due to immune enhancement and other peculiarities of dengue epidemiology. Here, we use a compartmental transmission model to assess the impact of partially effective vaccines in a hyperendemic Thai population. Crucially, we evaluate the effects that certain serotype heterogeneities may have in the presence of mass-vaccination campaigns. In the majority of scenarios explored, partially effective vaccines lead to 50% or greater reductions in the number of cases. This is true even of vaccines that we would not expect to proceed to licensure due to poor or incomplete immune responses. Our results show that a partially effective vaccine can have significant impacts on serotype distribution and mean age of cases.
| S-EPMC4142437 | biostudies-literature
Project description:Corona Virus Disease-2019 (COVID-19), caused by Severe Acute Respiratory Syndrome-Corona Virus-2 (SARS-CoV-2), is a highly contagious disease that has affected the lives of millions around the world. Chest X-Ray (CXR) and Computed Tomography (CT) imaging modalities are widely used to obtain a fast and accurate diagnosis of COVID-19. However, manual identification of the infection through radio images is extremely challenging because it is time-consuming and highly prone to human errors. Artificial Intelligence (AI)-techniques have shown potential and are being exploited further in the development of automated and accurate solutions for COVID-19 detection. Among AI methodologies, Deep Learning (DL) algorithms, particularly Convolutional Neural Networks (CNN), have gained significant popularity for the classification of COVID-19. This paper summarizes and reviews a number of significant research publications on the DL-based classification of COVID-19 through CXR and CT images. We also present an outline of the current state-of-the-art advances and a critical discussion of open challenges. We conclude our study by enumerating some future directions of research in COVID-19 imaging classification.
| S-EPMC8890789 | biostudies-literature
Project description:Deep learning has witnessed a significant improvement in recent years to recognize plant diseases by observing their corresponding images. To have a decent performance, current deep learning models tend to require a large-scale dataset. However, collecting a dataset is expensive and time-consuming. Hence, the limited data is one of the main challenges to getting the desired recognition accuracy. Although transfer learning is heavily discussed and verified as an effective and efficient method to mitigate the challenge, most proposed methods focus on one or two specific datasets. In this paper, we propose a novel transfer learning strategy to have a high performance for versatile plant disease recognition, on multiple plant disease datasets. Our transfer learning strategy differs from the current popular one due to the following factors. First, PlantCLEF2022, a large-scale dataset related to plants with 2,885,052 images and 80,000 classes, is utilized to pre-train a model. Second, we adopt a vision transformer (ViT) model, instead of a convolution neural network. Third, the ViT model undergoes transfer learning twice to save computations. Fourth, the model is first pre-trained in ImageNet with a self-supervised loss function and with a supervised loss function in PlantCLEF2022. We apply our method to 12 plant disease datasets and the experimental results suggest that our method surpasses the popular one by a clear margin for different dataset settings. Specifically, our proposed method achieves a mean testing accuracy of 86.29over the 12 datasets in a 20-shot case, 12.76 higher than the current state-of-the-art method's accuracy of 73.53. Furthermore, our method outperforms other methods in one plant growth stage prediction and the one weed recognition dataset. To encourage the community and related applications, we have made public our codes and pre-trained model.
| S-EPMC9726777 | biostudies-literature
Project description:Background: In 2018, Stellenbosch University's Ukwanda Centre for Rural Health led a faculty initiative to expand undergraduate health professions training to a new site, 9 hours drive from the health sciences campus in the sparsely populated Northern Cape Province of South Africa in the town of Upington. This is part of a faculty strategy to extend undergraduate health sciences training into an under-resourced part of the country, where there is no medical school. During 2019, the first year of implementation, four final year medical students undertook a longitudinal integrated clerkship at this site, while final year students from other programmes undertook short 5-week rotations, with plans for extending rotations and including more disciplines in 2020. The aim of this study was to understand stakeholder perceptions regarding the development of Upington as a rural clinical training site and how this influenced existing services, workforce sustainability and health professions education. Methods: An iterative thematic analysis of qualitative data collected from 55 participants between January and November 2019 was conducted as part of the case study. A constructivist approach to data collection was utilized to explore participants' perceptions, experiences and understanding of the new training site. Triangulation of data collection and reflexive thematic analysis contributed to the trustworthiness of the data and credibility of the findings. Findings: The perceptions of three key groups of stakeholders are reported: (1) Dr. Harry Surtie Hospital and Academic Programme Managers; (2) Supervising and non-supervising clinical staff and (3) Students from three undergraduate programs of the Faculty. Five themes emerged regarding the development of the site. The themes include the process of development; the influence on the health service; workforce sustainability; a change in perspective and equipping a future workforce. Discussion: This case study provides data to support the value of establishing a rural clinical training platform in a resource constrained environment. The influence of the expansion initiative on the current workforce speaks to the potential for improved capacity and competence in patient management with an impact on encouraging a rural oriented workforce. Using this case study to explore how the establishment of a new rural clinical training site is perceived to influence rural workforce sustainability and pathways, may have relevance to other institutions in similar settings. The degree of sustainability of the clinical training initiative is explored.
| S-EPMC8093558 | biostudies-literature
Project description:Consumer interest in meat and dairy alternatives drives demand for plant-based protein ingredients. While soy and gluten dominate the market, there is a trend to explore alternative crops for functional ingredient production. The multitude of ingredients poses challenges for food manufacturers in selecting the right protein. We investigated 61 commercially available protein ingredients from various sources, categorizing them based on their protein content into protein-rich flours (protein content less than 50%), protein concentrates (protein content between 50% and 80%), and protein isolates (protein content higher than 80%). Methionine, cysteine, and lysine were decisive for the amino acid score, which even varied between ingredients produced from the same raw material. Such differences were also observed in the protein solubility profiles, characterized by their raw material-specific protein pattern. By focusing on soy, pea, and fava bean ingredients, the broad spectra of emulsifying and foaming properties were illustrated. These ranged from non-emulsifying and non-foaming to high emulsifying capacities of 737 mL/g ingredient and foaming activities of 2,278%, accompanied by a foam stability of 100%. Additionally, we demonstrated that the functionality of ingredients obtained from different batches could vary by up to 24% relative SD. Protein solubility, powder wettability, color, and particle size were determined as key properties for the differentiation of soy, pea, and fava bean protein ingredients by principal component analysis. In our study conclusion, we propose essential measures for overcoming challenges in protein ingredient production and utilization to realize their full potential in fostering sustainable and innovative plant-based food production.
| S-EPMC11648661 | biostudies-literature
Project description:Tackling global health challenges demands the appropriate use of available technologies. Although digital health could significantly improve health care access, use, quality, and outcomes, realizing this possibility requires personnel trained in digital health. There is growing evidence of the benefits of digital health for improving the performance of health systems and outcomes in developed countries. However, significant gaps remain in resource-constrained settings. Technological and socio-cultural disparities between different regions or between provinces within the same country are prevalent. Rural areas, where the promise and need are highest, are particularly deprived. In Latin America, there is an unmet need for training and building the capacity of professionals in digital health. This viewpoint paper aims to present a selection of experiences in building digital health capacity in Latin America to illustrate a series of challenges and opportunities for strengthening digital health training programs in resource-constrained environments. These describe how a successful digital health ecosystem for Latin America requires culturally relevant and collaborative research and training programs in digital health. These programs should be responsive to the needs of all relevant regional stakeholders, including government agencies, non-governmental organizations, industry, academic or research entities, professional societies, and communities. This paper highlights the role that collaborative partnerships can play in sharing resources, experiences, and lessons learned between countries to optimize training and research opportunities in Latin America.
| S-EPMC6939247 | biostudies-literature
Project description:IntroductionIn an attempt to address health inequities, many U.S. states have considered or enacted legislation requiring antibias or implicit bias training (IBT) for health care providers. California's "Dignity in Pregnancy and Childbirth Act" requires that hospitals and alternative birthing centers provide IBT to perinatal clinicians with the goal of improving clinical outcomes for Black women and birthing people. However, there is as yet insufficient evidence to identify what IBT approaches, if any, achieve this goal. Engaging the experiences and insights of IBT stakeholders is a foundational step in informing nascent IBT policy, curricula, and implementation.MethodsWe conducted a multimethod community-based participatory research study with key stakeholders of California's IBT policy to identify key challenges and recommendations for effective clinician IBT. We used focus groups, in-depth interviews, combined inductive/deductive thematic analysis, and multiple techniques to promote rigor and validity. Participants were San Francisco Bay Area-based individuals who identified as Black or African American women with a recent hospital birth (n=20), and hospital-based perinatal clinicians (n=20).ResultsWe identified numerous actionable challenges and recommendations regarding aspects of (1) state law; (2) IBT content and format; (3) health care facility IBT implementation; (4) health care facility environment; and (5) provider commitment and behaviors. Patient and clinician insights overlapped substantially. Many respondents felt IBT would improve outcomes only in combination with other antiracism interventions.Health equity implicationsThese stakeholder insights offer policy-makers, health system leaders, and curriculum developers crucial guidance for the future development and implementation of clinician antibias interventions.
| S-EPMC10507933 | biostudies-literature