Articles tagged with Data Sets
Researchers developed an AI model that can diagnose COVID-19 with high accuracy, using federated learning to preserve patient data privacy. The model was trained on over 9,000 CT scans from 23 hospitals in the UK and China, and validated against a panel of radiologists.
Researchers have published an extensive 7T fMRI dataset to study how humans perceive and interpret naturalistic photographs. The Natural Scenes Dataset provides a massive scale of brain data for training complex deep-learning models that predict brain activity.
The MDI Biological Laboratory has been awarded a grant to promote cloud computing among researchers in Maine, aiming to level the playing field by providing access to sophisticated computing resources. The program will provide training on Google Cloud Platform and assist institutions in implementing cloud computing services.
Researchers developed an algorithm to differentiate life-threatening gunshot events from non-life-threatening plastic bag explosion events. The study found that 75% of plastic bag pop sounds were misclassified as gunshot sounds, highlighting the need for a diverse dataset of similar sounds.
A new database has been launched to systematically record findings on perovskite semiconductors, featuring over 42,000 individual data sets and analysis tools for interactive exploration. The FAIR principles guide the preparation of the data, enabling easy searching with modern algorithms and artificial intelligence.
Researchers from Kaunas University of Technology have developed an AI-based approach for contactless machine failure detection, using sound data from existing equipment. The solution is sustainable and relatively cheap, with no need for new sensors or equipment installation.
A new dataset from Canterbury earthquakes provides over 15,000 case histories for liquefaction, significantly augmenting model training and testing. The dataset enhances hazard assessments and improves engineering solutions in earthquake recovery, benefiting society as a whole.
A team of University of Pennsylvania biologists used citizen science data to create a comprehensive abundance map of the black-legged tick, responsible for transmitting Lyme disease. By correcting biases in the data, they were able to increase its value and provide insights into tick distribution across the Northeast US.
A new AI application called DeepMReye uses MRI signals to read eye movements and infer thoughts, memories, and goals. It can also diagnose brain diseases by analyzing characteristic eye movement patterns.
Convolutional neural networks trained to identify abnormalities on upper extremity radiographs are susceptible to a ubiquitous confounding image feature: radiograph labels. Covering these labels increases accuracy, while using them alone leads to decreased performance.
Research by Binghamton University economists found that COVID-19 lockdowns led to a significant improvement in air quality for minority neighborhoods in rural New York, narrowing the existing gap with majority white neighborhoods. The study suggests stronger regulation can improve air quality in polluted areas.
A new study explores the problem of shortcuts in a popular machine learning method and proposes a solution that can prevent shortcuts by forcing the model to use more data. By removing simpler characteristics and asking the model to solve the task two ways, researchers reduce the tendency for shortcut solutions and boost performance.
Researchers at Children's Hospital of Philadelphia have developed a novel therapy that targets proteins essential for tumor growth and survival. Using a multi-omics approach, they identified peptides unique to neuroblastoma tumors, which are then targeted by peptide-centric chimeric antigen receptors (PC-CARs).
Researchers identified a gene family called MEF2 that controls a genetic program promoting resistance to cognitive decline. Enriching activities appear to activate MEF2, which may help prevent age-related dementia.
A new study improves AI diagnoses by penalizing algorithms for false negatives, which can be more urgent than accuracy. Researchers achieved significant improvements in precision and recall for chronic kidney disease and other conditions using cost sensitivity techniques.
A study has created a massive database of academic papers documenting global adaptation actions to climate change. The research found that people are taking action, but these efforts tend to be fragmented and may not be enough to deal with the expected effects of climate change.
The project aims to address rising healthcare costs, health disparities, and expands digital health through advances in AI. It also trains students from different disciplines to develop and apply data science techniques.
A new MIT study suggests that pedestrians choose routes that point most directly toward their destination, even if those routes are longer. This strategy, known as vector-based navigation, may have evolved to allow the brain to devote more power to other tasks.
The StEER Network's post-event reconnaissance helped assess building damage from Hurricane Michael, revealing widespread wind- and surge-induced damage. The dataset has been used to develop data-driven fragilities, train machine learning applications, and inform policy and practice improvements for coastal communities.
Researchers have developed a federated analytics system, FAMHE, that enables healthcare providers to collaborate on statistical analyses and machine learning models without exchanging underlying datasets. The system has been proven mathematically secure and accurately reproduced published studies in multi-centric settings.
A Texas A&M team led by Dr. Byung-Jun Yoon has received $2.4 million to develop new computational techniques for reducing the size of large scientific data sets. The goal is to preserve quantities of interest while minimizing data storage and processing needs.
Researchers at GlaxoSmithKline and CCDC combined proprietary and published datasets to train machine learning models for predicting stable polymorphs in new drug candidates. The approach leverages the large volume and variety of data in the Cambridge Structural Database, resulting in more confident predictions and improved model accuracy.
Researchers analyzed over 120 million English-language tweets to find a decrease in negative posts about COVID-19, particularly in countries with high vaccination rates. The study suggests that increased vaccination may have contributed to the drop in negativity, but further analysis is needed to fully understand this phenomenon.
A team of Harvard researchers created an integrated pipeline, STAMPScreen, to help genetic engineers identify target genes and perform screening studies. The protocol combines computational tools with lab experiments to quickly and efficiently test gene function in living cells.
Researchers used extensive 2D and 3D broadband seismic reflection data to visualize and understand the subsurface structures of the Orphan Basin. The study provides context for future assessments of source rock, reservoir, and seal strata in oil and gas exploration.
The COVID-19 pandemic has triggered unprecedented life expectancy losses across 29 countries, with a decline of over half a year observed in most nations. The largest declines were seen among males, particularly in the US, where a 2.2-year loss was reported.
Researchers at PSU and UT Arlington will assess how crowdsourced data can help establish pedestrian activity on streets. The study aims to provide a more accurate alternative to traditional counting methods, enabling safer studies and better urban planning.
Researchers at Lawrence Livermore National Laboratory calculated moment tensors for 130 underground nuclear and 10 chemical tests to aid explosion monitoring. The study's database of carefully documented explosions will be useful for researchers, allowing them to distinguish explosions from earthquakes and estimate yield.
Researchers analyzed 1,785 ancient human genomes to determine parental relatedness, revealing that cousin marriages occurred only 3% of the time. The new method allowed for more efficient screening of ancient DNA, also providing insights into population dynamics and demographic impact of agriculture.
Scientists compiled a 71-year potential risk index dataset to estimate TC impact severity on the Chinese mainland, showing increased damage over time. Recent upward trends in precipitation and wind values contribute to larger TC impacts along coastal China.
Researchers analyzed 1.4 million field observations and 73,000 museum records to find a strong correlation between species abundance in nature and their presence in museum collections. This method enables scientists to study species decline and estimate past abundances, providing insights into conservation efforts.
A European project developed standard solutions for energy refurbishment, considering various climate zones. These packages include prefabricated façades, decentralized ventilation systems, and smart ceiling fans to reduce energy consumption and improve comfort. Pilot sites showed significant savings and improvements in tenant well-being.
The world's largest repository of raw genomic sequences is missing critical data necessary for monitoring and preserving biodiversity. Only 14% of archived datasets contain information about when and where organisms were sampled, highlighting the need for community standards to preserve metadata.
A new study reveals that climate change will affect Antarctic seals, such as crabeater and Weddell seals, in distinct ways. Crabeater seals are more vulnerable due to their specialized diet and breeding on unstable pack-ice, while Weddell seals are less affected.
A new analytics platform, RDCA-DAP, will be launched to accelerate rare disease treatment innovation by hosting and standardizing rare diseases data. The platform is expected to empower patients and families to drive innovation in the field.
A new dataset combines demographic and financial information to identify communities at risk of climate-related gentrification. The Socio-Economic Physical Housing Eviction Risk (SEPHER) dataset provides insights into the impact of climate hazards on vulnerable populations, including increased evictions and rising housing costs.
A new study showcases the COVIDome Explorer, a public online portal for real-time COVID-19 data analysis, visualization and sharing. The platform enables rapid hypotheses testing, hypothesis generation and discoveries by experts and non-experts.
A new study found that larger cities produce more income inequality, with the poor experiencing a decline in quality of life. The researchers analyzed data from municipal areas across the US and found that the top 10% of earners gain an increasingly large portion of wealth as cities grow.
Earthquakes generated by controlled fluid injection at the Rangely oil field were caused by destabilizing fault pore pressure changes, according to a new mechanistic model. The study revisits data from the decades-old project in light of increased seismicity due to fluid injection.
A database of 3,000 lead isotope analyses has been compiled for the Iberian Peninsula, providing a comprehensive resource for geological and archaeological research. The IBERLID database includes standardized data on minerals, rocks, and metallic objects, facilitating comparison and analysis.
Researchers developed a novel evidence-based material recommender system that predicts high entropy alloy formation without data descriptors, overcomes data bias and poor availability. The method recommends an FeMnCoNi alloy as the most probable HEA and successfully synthesizes it, confirming its validity.
A recent study published in eLife journal analyzed mortality data from 103 countries to determine the true rate of COVID-19 deaths. The researchers found significant underreporting of pandemic deaths, with some nations reporting 2-50 times higher excess deaths than official numbers.
Researchers analyzed mortality data from 103 countries to track excess deaths during the COVID-19 pandemic. They found significant increases in deaths in some countries, while others saw a decrease due to social distancing measures.
Researchers applied supervised machine learning to lidar data from fishery surveys, automating the identification process in regions with a strong possibility of harboring fish. The technique decreased manual inspection in datasets from Yellowstone Lake and Gulf of Mexico studies by 61.14% and 26.8%, respectively.
A recent study found that meeting the Paris Agreement's 2°C target will increase net energy sector jobs globally by about 8 million by 2050. The majority of these new jobs will be in the solar and wind industries, while fossil fuel extraction jobs are expected to decline rapidly.
A recent study found that controversial posts on Reddit spread faster and reached more people than noncontroversial posts, with over 60,000 comments on the former. The researchers analyzed over 47,000 posts to understand why controversy can make online content more attention-grabbing.
A new analysis reveals that global toxic pollution and climate change disproportionately affect low-income countries, with the top-one-third of at-risk nations accounting for over two-thirds of the world's population. The study provides a list of top-ten countries most equipped to immediately begin risk reduction efforts.
A new neural model can detect 'inappropriateness' in chatbot messages, which are defined as content that may harm the reputation of the speaker's company without being toxic. The study provides a large collection of labeled datasets for further research and has been made publicly available.
A new portal, gEAR, enables researchers to easily utilize complex genetic data from sophisticated studies, advancing medical discoveries in hearing loss. The tool provides visualizations of datasets and allows inputting of own data for comparison, saving time and facilitating analysis.
Researchers found that subjective decisions play a critical role in shaping empirical results, emphasizing the need for open data, systematic robustness checks, and transparency in academic research. This study highlights the benefits of diverse analytical approaches and the importance of humility when communicating findings.
A decade's worth of data shows a significant decrease in mechanical ventilation use and an increase in non-invasive respiratory support, such as CPAP, for preterm infants. This shift is associated with improved health outcomes, including reduced lung injury and neurodevelopmental risks.
A new study maps spirituality and religiosity to a specific brain circuit centered in the periaqueductal gray (PAG), a brainstem region implicated in fear conditioning, pain modulation, altruistic behaviors, and unconditional love. The findings suggest that spiritual acceptance can be localized to this circuit.
Researchers develop AI system that uses cooperative strategies to improve human-AI collaboration in negotiations. The system, called CaSiNo, is trained on a dataset of negotiation dialogues and improves upon traditional menu-driven interfaces. It has real-world applications in business, education, and entrepreneurship.
A new study reveals similarities and differences in online conversation topics between Angelenos and New Yorkers. Online, Angelenos tend to discuss healthcare, jobs, and entertainment, while New Yorkers focus on art, politics, and nightlife.
Researchers at UMSOM aim to expand genetic databases to include under-represented minority populations, improving polygenic risk score accuracy for clinical use. The goal is to identify best practices for translating scores into clinical practice across diverse populations.
Researchers analyzed visual complexity of written language in 47,000+ characters from 133 scripts, finding large scripts have more complex characters and complexity driven by linguistic units, not evolutionary change.
A comprehensive study published in Nature Communications found that more than one million deaths worldwide were attributed to fossil fuel combustion in 2017. The research used a global dataset of air pollution emissions and modeling to identify specific sources of pollution, including coal, energy production, and cookstoves.
Researchers aim to identify when individuals are falling out of the flow state by detecting physiological cues with off-the-shelf sensors. AI will be used to introduce interactive stimuli that nudge subjects back into a higher cognitive state.
Researchers have developed a decision support system to identify the environmental impact of nanomaterials faster and more cost effectively. The library contains full physicochemical characterization of 69 nanomaterials, including calculated molecular descriptors.
Aalto University researchers have developed a machine learning-based method to produce synthetic data that preserves privacy and can be used for statistical analyses. The new method solves a major problem with existing methods, enabling researchers to share sensitive information without compromising individual identities.