Research using machine learning to solve scientific problems. Covers AI quality control, sea-surface temperature estimation, label-free cell analysis, and cyber defense.
Omnipose, a deep learning software, can identify various types of tiny objects in micrographs with high precision, including bacteria of all shapes and sizes. It overcomes limitations of previous approaches by handling object overlap and detecting cell intoxication, making it a game-changer for biological image analysis.
A University of Groningen team created two machine learning models to predict app removal risks, achieving accuracy rates of up to 79.2%. The models can help developers avoid bans and users protect their data.
Researchers developed a machine-learning model to predict heat capacity of MOFs, enabling more efficient applications in energy and climate change. The model's accuracy was improved by removing solvent from pores during synthesis.
A recent study analyzed the records of over 33,000 dementia patients in Wales and found that those closest to their diagnosis were taking an average of three or more medications. This suggests a possible link between polypharmacy and the development of dementia.
A study published in the Canadian Journal of Cardiology found that smartwatch health apps detecting atrial fibrillation generated a high rate of false positives and inconclusive results, especially in patients with certain cardiac conditions. Better algorithms and machine learning may help improve the accuracy of these devices.
A team of Illinois Tech researchers used machine learning to estimate the age and gender of individual users with high accuracy, raising questions about data security and privacy. The study highlights the need for better regulations and best practices to protect personal information from being misused.
Researchers successfully taught human and mouse neurons to play the video game Pong in real-time, showcasing their ability to exhibit sentience and adapt to a changing environment. The study's findings have potential applications in disease modeling, drug discoveries, and expanding our understanding of brain function.
Researchers from McGill University and MIT developed an AI system that can learn the rules and patterns of human languages on its own. The model automatically generates higher-level language patterns that can be applied to different languages, achieving better results.
Researchers used AI to analyze seismic signals and predict future fault friction and next failure time with high resolution in laboratory earthquakes. The technique goes beyond previous work by predicting the future state of the fault's physical system.
Researchers developed a smart mouthguard that translates complex bite patterns into instructions to control devices such as computers, smartphones and wheelchairs. The device achieves 98% accuracy and has the potential to support individuals with limited dexterity or neurological disorders.
A University at Buffalo-led research team has been awarded a $5 million grant to develop digital tools that can help older adults recognize and protect themselves from online deceptions. The project, DART, aims to reduce online fraud among older adults, who lose billions of dollars each year due to scams.
Researchers developed a generatively designed patient-specific bone fixation device using Generative Design technology. The implants are tailored to each patient's anatomy and biomechanical needs, resulting in lighter, less prominent, and minimally invasive designs that promote faster healing and reduced revision surgery.
Researchers found that machine learning models outperformed traditional risk prediction models in predicting suicide-related outcomes. These models can identify patterns associated with suicide risk and have been shown to correctly predict 66% of people who would experience a suicide outcome.
Researchers aim to create a unified database network for battery data, facilitating AI analysis and predictions. The Battery Data Genome will collect data across the entire battery lifecycle, from discovery to deployment, with uniform standards for metadata.
Researchers at FAU's Schmidt College of Medicine are developing a pilot program to tackle health disparities by leveraging electronic health records, artificial intelligence, and machine learning. The project aims to improve AI/ML delivery and research operations in community health centers and federally qualified health centers.
A new study found an association between college student FoMO and maladaptive behaviors such as academic misconduct, drug and alcohol use, and breaking the law. Brief FoMO assessments may be valuable risk prediction tools for counselors assisting students in transition to college.
Researchers developed algorithms that can predict volleyball players' actions with over 80% accuracy, combining visual data with hidden variables. The algorithms also showed promise in predicting multiple actions across sequences of up to 44 frames.
Researchers used machine learning algorithms to optimize climate models, increasing their accuracy and detail. By applying Generative Adversarial Networks (GANs) to climate simulations, the team was able to improve the models' ability to represent extreme precipitation events.
Researchers developed a new technique that enables on-device training using less than a quarter of a megabyte of memory, reducing the need for powerful computers and central servers. This approach preserves privacy by keeping data on the device, making deep learning more accessible for low-power edge devices.
Researchers developed a non-invasive algorithm to identify patients with compensated cirrhosis at highest risk for severe complications. The online calculator uses widely available laboratory parameters and is simple, non-invasive, and cost-effective.
A machine learning model developed by Carnegie Mellon University can accurately predict how stay-at-home orders affect people with multiple sclerosis. The model uses passively gathered smartphone and fitness tracker data to identify factors that signal changes in health.
The new computer chip uses a transistor-free design that eliminates data transfer time and minimizes energy consumption. It offers up to 100 times faster performance than conventional computing architectures, making it ideal for AI applications.
A Newcastle University study has developed a machine learning tool that can predict the performance properties of land plant Rubisco proteins with high accuracy. This prediction will enable researchers to identify and engineer 'supercharged' Rubisco proteins that can increase atmospheric CO2 uptake and store in crops such as wheat.
Neuronal silencing periods enable efficient temporal sequence identification, allowing the brain to remember phone numbers and PINs. A new AI mechanism utilizing this mechanism also protects against stolen cards by recognizing personal handwriting style and timing.
A study published in Cell Press found that when humans are involved, computer decisions are perceived as fairer. Participants deemed decisions related to positive outcomes fairer than negative ones and had concerns over fairness in systems with higher stakes. The results suggest that automated decision-making systems need careful desig...
Researchers developed a machine learning model to quickly recognize predictive risk factors and their importance for undesirable hospitalization outcomes. The model achieved an accuracy of 95.6% and identified modifiable risk factors that can be mitigated through clinical interventions.
Researchers developed a computational platform to identify metabolic vulnerabilities in ovarian cancer genes, suggesting opportunities for targeted therapies. The study found that certain genetic alterations can create vulnerabilities in cancer cell metabolism, which can be exploited to selectively kill cancer cells.
Researchers at Ohio State University have developed a new machine learning method called next-generation reservoir computing that can learn spatiotemporal chaotic systems in a fraction of the time. The algorithm is more accurate and uses less training data, making it easier to predict complex physical processes like Earth's weather.
A recent grant will fund a project developing new hardware for machine learning, aiming to curb unsustainable energy use in AI systems. The new algorithms being developed are made available to the research community and compatible with an openly shared computing platform.
Current AI models are restricted by a lack of experience in real-world environments, despite achieving significant advancements in virtual settings. Researchers are now exploring ways to bridge this gap with foundation models that can operate in physical spaces.
A recent study published in Aging-US found that feeling lonely, unhappy, or hopeless increases one's biological age more than smoking. The research used digital models of aging to analyze the effects of various factors on aging rates, revealing a significant correlation between mental health and accelerated aging.
A blood test taken at the time of Covid-19 infection could predict who is most likely to develop long Covid. Researchers identified a 'signature' in protein levels that predicted persistent symptoms after one year.
The use of voice control smart devices in children may impede critical thinking, empathy, and compassion. Devices can't teach children how to behave politely and lack non-verbal communication skills.
Researchers discovered a pattern of DNA mutations that links bladder cancer to tobacco smoking using a powerful new machine learning tool. The tool identified four mutational signatures, including one tied to tobacco smoking, which could lead to more customized treatments for patients with specific cancers.
A machine learning multi-city model is developed to predict municipal solid waste generation in China, using a wide range of feature variables. The model shows good performance with an R-squared value of 0.939 and is suitable for predicting MSW generation in China.
Physicists used machine learning to compress a complex quantum problem into four equations, capturing the physics of electrons on a lattice with high accuracy. The approach could revolutionize how scientists investigate systems containing many interacting electrons and potentially aid in designing materials with sought-after properties.
The 2022 EXPLORE Lunar Data Challenge identifies hazards on the Moon's surface using images from the Lunar Reconnaissance Orbiter. Participants train models to recognize craters and boulders, then create a map of optimal rover routes to avoid hazards.
Researchers use classical computers to make predictions about quantum systems, helping to solve physics and chemistry problems. Machine learning tools provide a bridge between the human world and quantum reality.
Researchers reviewed mobile sensing designs, outcomes, and limitations to better understand its capacity for remote detection, longitudinal tracking, and exposure tracing. Despite technical and societal challenges, advances in data analytics and machine learning may improve data quality and scalability.
A team of scientists from China published a perspective paper on the use of AI in skin diseases, highlighting its potential to assist clinicians. They propose several recommendations to improve AI-assisted diagnosis systems, including establishing a robust database and adapting algorithms to existing real-world databases.
Researchers developed an AI tool using natural language processing and machine learning to identify people who inject drugs in electronic health records. The model accurately identified PWIDs in 1,000 records from 2003-2014, significantly improving clinical decision making and resource allocation.
Researchers developed an electronic laboratory notebook that uses knowledge graphs to describe material properties and experimental processes. The platform enables automated analysis, lossless sharing, and discovery of new materials with potential applications in energy-related devices.
The study used a weakly supervised deep learning algorithm to analyze human brain autopsy tissues and predict the presence or absence of cognitive impairment. The model identified a signal associated with decreasing myelin staining, which was linked to cognitive impairment in the white matter.
A multidisciplinary team at UC San Diego received NSF funding to create a system that remotely monitors patient posture, movement, and physical therapy treatments using wearable technology and machine learning. The goal is to enable personalized physical therapy treatments and improve health outcomes.
Researchers have developed an algorithm that uses smartphone camera and flash to detect low blood oxygen levels. The method produced accurate results in 80% of the participants, showing promise for remote monitoring and early detection of conditions like COVID-19.
Researchers developed a new software tool called ProteinMPNN to create protein molecules more accurately and quickly than before. The team used machine learning algorithms, including AlphaFold, to generate new protein shapes and sequences, paving the way for novel vaccines, treatments, and sustainable biomaterials.
Researchers at Emory University used machine learning and fMRI to analyze a dog's brain activity while watching videos. The results show that dogs are more attuned to actions in their environment than to who or what is performing the action. This study offers proof of concept for decoding canine visual perception.
Researchers developed an AI diagnostic tool called CheXzero that can detect diseases on chest X-rays from natural-language descriptions contained in accompanying clinical reports. The model performed on par with human radiologists and was trained without laborious human annotation of data, making it a major advance in clinical AI design.
The Florida Summer Institute in Biostatistics and Data Science recruits students from underrepresented groups, providing a rigorous quantitative education. The program introduces college students to statistical methods and applications in biomedical research, highlighting health disparities and social drivers of health.
University of Minnesota scientists have developed a method to fine-tune traditional compressed sensing for high-quality images using modern data science tools and machine learning ideas. This approach closes the gap between traditional and deep learning methods, providing a new direction for MRI reconstruction research.
The Center for BrainHealth has launched three research projects to develop objective metrics of improved brain systems in response to interventions. These projects aim to determine the changes in the brain's physiology, structure, and function linked to gains in comprehensive psycho-social measurements over time.
Healthcare researchers caution against misusing AI algorithms in clinical research, highlighting concerns about bias, transparency, and data quality. The team advocates for evaluating ML methods against traditional statistical approaches and ensuring clinician decision-making is complemented, not replaced.
The University of California, San Diego is part of the National Institutes of Health's Bridge to Artificial Intelligence program, aiming to create comprehensive AI-ready datasets. The program will support researchers in developing interpretable and trustworthy AI technologies to improve human health.
Researchers developed a unified machine learning model that analyzes data from patients with and without treatment, predicting depression outcomes better than separate models. The approach enables personalized medicine by designing treatment plans specific to each patient's needs.
City digital twin technology is used to create synthetic training data for deep learning models, which are then trained on a combination of real and synthetic data. This approach yields promising results for architectural segmentation tasks, particularly for modern building styles.
Researchers find XAI methods improve AI performance and explain bias in data, enabling accurate applications like insurance decision-making. Implementing XAI helps manage power consumption and optimize AI systems for efficient Industry 4.0 growth.
Researchers developed an AI method to predict how well new COVID-19 variants infect human cells and evade antibodies. The system can analyze a million mutated variants, enabling the development of next-generation antibody therapies and vaccines that provide broader protection against potential future variants.
A team led by York University has developed a new technique to keep drinking water safe in refugee settlements using machine learning and ensemble forecasting systems. The approach can predict the probability of residual chlorine remaining in stored water, providing critical information for aid workers to ensure safe drinking water.
A new method using machine learning corrects damaged DNA and unveils true mutation processes in tumour samples, helping early cancer detection and accurate diagnosis. The tool predicted over 90% of developing cancer processes, offering a significant advancement in cancer patient care.
Researchers from GIST developed an AI model that adjusts videogame difficulty based on player emotions, incorporating aspects such as challenge, competence, flow, and valence. The model has been verified to improve players' overall experience, regardless of their preference, and has potential applications in various fields beyond gaming.