A study has developed a method using dark-field microscopy and deep learning algorithms to identify microplastics in human cells, achieving an accuracy of 93% for 1-micron polystyrene particles. The technique has the potential to screen microplastics in various samples, reducing time-consuming data acquisition and processing steps.
A new study by Ritsumeikan University reveals that whole-body dynamic balance training increases brain connectivity and improves motor learning. The research found significant changes in brain activity associated with offline learning, opening up new avenues for studying neural networks during natural tasks.
Researchers found that re-identifying individuals from genomic data using public face images is harder than previously thought, with success rates well below idealized settings. They developed a method to alter social media photos and reduce the risk of privacy breaches.
Researchers at Japan Advanced Institute of Science and Technology create method to estimate five-room acoustic parameters and speech transmission index using a short conversation, with potential applications in monitoring auditoriums during concerts and saving lives through smart speakers
A new AI model has been developed to classify colorectal polyps, demonstrating accuracy and sensitivity at a level comparable to practicing pathologists. The model was tested in a clinical trial involving 15 pathologists, showing significant improvements in accuracy compared to traditional methods.
The research team developed a technology for remotely assessing the condition of a building during an earthquake based on the readings from the building's seismometer. The new method uses CNN machine learning to quickly assess damage levels and determine if a building can continue to be used.
A study involving Drosophila found that a constant and precisely regulated energy supply is essential for nerve development, particularly during the degradation of nerve connections. Malnutrition was shown to intensify defects in this process.
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 suggests that larger bee colonies with comfortable food stores are less willing to take risks, while smaller colonies with limited resources are more likely to ignore warning signals. This study provides insights into the complex communication system of bees and its implications for understanding biological collectives.
A study at the University of Turku found that brain function regulating satiety and appetite is altered before obesity develops, with family background risk factors contributing to these changes. The findings suggest the brain and central nervous system are key targets for treating obesity.
Researchers at the University of Waterloo have created a deep neural network that detects disease biomarkers with high accuracy, achieving 98 per cent detection of peptide features. This breakthrough could enable earlier and more accurate disease detection through tissue sample analysis.
Scientists at Gladstone Institutes discovered that non-convulsive epileptic activity drives chronic brain inflammation in Alzheimer's models, which can be reversed by eliminating protein tau or using the anti-epileptic drug levetiracetam. This link between brain networks and immune cells may hold promising treatments for Alzheimer's di...
Scientists studied rats in a virtual reality maze to gain a deeper understanding of the hippocampus' circuit-level functions. They discovered that hippocampal neurons encoded multiple aspects of an animal's location, including distance traveled and body direction.
Researchers have trained a neural network to detect anomalies in medical images, adapting it to the nature of medical imaging and achieving better results. The new method uses weakly supervised training and can spot small-scale anomalies, accelerating the work of histopathologists and radiologists.
A pilot study examines graph-theoretical properties of brain networks in traumatic brain injury and controls, showing association with balance impairment and structural damage. The study uses EEG-based functional connectivity measures during a balance perturbation task to explore underlying neural mechanisms.
Researchers at Gladstone Institutes found that reducing tau levels impacts both excitatory and inhibitory cells, leading to a reduction in excitation-inhibition ratios. This effect counteracts diseases that cause abnormal increases in this ratio, potentially improving the brain's ability to perform its functions.
Researchers developed an attention-based deep neural network to detect multiple ship targets, exceeding conventional networks' performance. The model focused on inherent features of the two ships simultaneously, outperforming traditional approaches.
A novel machine learning approach has been developed to understand symmetry and trends in materials, enabling researchers to group similar classes of material together. The technique uses a large, unstructured dataset gleaned from 25,000 images to identify structural similarities and trends.
Researchers have made progress in creating a brain atlas of the mouse brain, which will help develop tools for studying the human brain. The study describes the diversity of neurons in the mouse brain and establishes new methods for characterizing cell types and neural connections.
Researchers at the Sainsbury Wellcome Centre have discovered a new brain circuit that enables mice to override their instincts based on previous experience. The ventral lateral geniculate nucleus (vLGN) inhibits threat reactions when animals feel safe, but activates them when danger is perceived.
Researchers developed a novel approach to 3D image segmentation, segmenting the gaps between parts instead of contours, to automate tedious tasks. The technique demonstrates promising results in diagnosing TMJ-related issues and has potential applications in other fields.
A team of researchers has found a 390-million-year-old hyper-facet eye system in trilobites that is unique to the animal kingdom. The discovery suggests that this ancient eye may have been an adaptation for life in low light conditions, and could provide insights into the evolution of visual systems.
Researchers used fractal analysis to study brain network patterns while listening to a story. The results show that complex thoughts are reflected in high-order dynamic correlations in neural activity patterns.
Scientists discovered that recurrent neural networks (RNNs) play a crucial role in the frontal cortex, responsible for decision-making, expressive language, and voluntary movement. The research also found that RNNs are more complex than previously thought, with a unidirectional structure.
Scientists at Aarhus University are working on a nano-sized brain-inspired computer that can harvest its own energy, making it the smallest and most efficient AI system yet. The project aims to reduce power consumption by 12 orders of magnitude compared to modern supercomputers.
Researchers have developed a new microscopy technique, light beads microscopy, that captures detailed images of activity from one million neurons across the mouse brain at high speed and single-cell resolution. This innovation allows scientists to investigate biological questions in a way that was not possible before.
Researchers at University of Chicago and Argonne National Laboratory found that primate neurons receive two to five times fewer excitatory and inhibitory synaptic connections than similar mouse neurons. The study suggests that metabolic costs may drive larger neural networks to be sparser, as seen in primates versus mouse neurons.
Researchers developed a robust, deep neural network model to analyze automobile traffic impacts of construction zones. The model estimates hourly traffic volumes without adjustment factors, helping transportation agencies plan for efficient work zone operations.
Researchers at Nanyang Technological University and Tan Tock Seng Hospital have developed an AI-powered system to diagnose glaucoma from stereo fundus images, achieving an accuracy of 97% in diagnosing the condition. The automated method could potentially be used in less developed areas where patients lack access to ophthalmologists.
Researchers at The Hebrew University of Jerusalem have developed a new deep learning artificial infrastructure inspired by individual neurons. Their approach uses complex mathematical modeling to replicate the brain's electrical processes and create more intelligent AI systems.
Using complex-valued layers can improve performance against adversarial attacks without sacrificing efficiency. This technique, combined with gradient regularization, allows neural networks to resist small perturbations and maintain accuracy.
Researchers discovered that a brain area traditionally thought to specialize in old habits also plays a role in learning new actions. The study found that the dorsolateral striatum is involved in consolidating action learning immediately after the new action has been learned.
University of South Australia researchers create a computer vision system to detect premature babies' faces and vital signs from digital cameras, outperforming electrocardiogram machines. The technology has the potential to replace contact-based sensors, reducing skin tearing and infections.
Researchers at Tokyo Institute of Technology developed a tunable neural network framework that achieves high accuracy and efficiency for sparse CNNs. The new architecture employs a Cartesian-product MAC array and pipelined activation aligners to enable dense computing of sparse convolution, resulting in better resource utilization.
A team of researchers used remote sensing data and deep machine learning to identify hundreds of new shell ring sites in the southeastern US. The study provides a better understanding of how people lived in the area and offers a way to locate undiscovered shell rings.
A machine learning algorithm can accurately sex Caucasian 2D footprints with up to 90 percent accuracy, surpassing expert analysis. This breakthrough has significant implications for forensic and anthropological applications.
Researchers from Skoltech and KU Leuven used machine learning to reconstruct 3D micro-CT images of fibrous materials, overcoming the difficulties faced by humans in analyzing these complex materials. The team employed GANs to fill a gap in available inpainting tools, enabling precise material analysis and simulation.
A study published in Science Advances reveals a previously unknown mechanism behind compulsive alcohol use, which may be targeted by medication. A small group of nerve cells in the central amygdala promote alcohol use despite negative consequences.
Researchers have developed a neural network model called BiteNetPp to detect protein-peptide binding sites, enabling the design of peptide-based drugs. The model consistently outperforms existing methods and can analyze a single protein structure in under a second, making it suitable for large-scale studies.
A new study from Washington University in St. Louis shows that guided by sparsity, silicon neurons learn to pick the most energy-efficient perturbations and wave patterns, enabling an emergent phenomenon of efficient communication between neurons. This research has significant implications for designing neuromorphic AI systems.
Researchers from Skoltech and their colleagues developed a neural network that can efficiently generate IUPAC names for organic compounds in accordance with the IUPAC nomenclature system. The network, trained using the Transformer architecture, achieved an accuracy of nearly 99%, outperforming traditional rule-based solutions.
Researchers at NYU Tandon School of Engineering developed a framework called DeepReDuce that streamlines neural networks to be more adept at computing on encrypted data. The team found that DeepReDuce improved accuracy and reduced ReLU count by up to 3.5% and 3.5×, respectively.
Researchers found that bats can compress their echoes by up to 90% without losing essential information for sonar-based tasks. This efficient encoding strategy allows bats to navigate complex environments with minimal neural machinery, enabling them to detect location and movement with high accuracy.
Researchers at Université libre de Bruxelles compare popular neuro-evolutionary methods for offline robot swarm design, observing a 'reality gap' where simulated neural networks fail in the real world. To address this, they propose reducing method 'power' to adopt simpler approaches with predefined building blocks.
Researchers introduce RoseTTAFold, a neural network approach that accurately predicts protein structures, outperforming traditional methods and rivalling DeepMind's AlphaFold2. The tool's code and public server are now accessible to the scientific community, enabling rapid solution of challenging structure determination problems.
Researchers from Skoltech have developed a new augmentation technique called MixChannel to help train computer vision algorithms with limited data. This approach outperformed state-of-the-art solutions in testing with three neural networks and can be combined with other methods for even more training data.
Skoltech researchers create a neural network that can guide the controlled deformation of semiconductor crystals, enabling superior properties for next-gen chips and solar cells. The approach combines various data sources and active learning to boost accuracy and convergence.
The brain's globally sparse yet locally compact modular topological characteristics reduce resource consumption for establishing connections. The research model shows that rewiring the network to a more biologically realistic modular structure significantly reduces running consumption and building cost.
Researchers developed machine learning models that can predict daily solar radiation using only thermal data, improving upon existing methods in various geo-climatic conditions. The models have been tested in nine locations across southern Spain and North Carolina, showing significant improvements in accuracy.
A Cornell University-led team developed a machine learning tool called Correlation Convolutional Neural Networks (CCNN) to parse quantum matter and make distinctions in the data. CCNN can identify relationships among microscopic properties that are impossible to determine at the scale of quantum systems.
A team of researchers has developed an AI system that can detect COVID-19 through automatic cough analysis. The system uses spectrogram features and demonstrates improved accuracy when incorporating gender information, which is found to be a significant factor in distinguishing between male and female coughs.
University of Illinois engineers develop physics-informed neural networks to predict outcomes of complex 3D printing processes. The model accurately recreates experiments and predicts temperature and melt pool length with high accuracy.
Researchers used convolutional neural networks to analyze facial photos before and after facelift surgery in 50 patients. The AI algorithms recognized a 4.3-year reduction in age, which correlated with patient satisfaction scores, averaging 75 for facial appearance and over 80 for quality of life.
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 DZNE's Dresden site develop i3D-Markers, a cutting-edge technology platform that uses high-density microelectrode arrays and 3-dimensional neuronal networks to predict the reaction of neurons to compounds. This platform aims to optimize drug candidate selection and accelerate brain disease development.
Researchers found that fluid flow vortices have high information processing capabilities before transitioning to a Karman vortex street. Virtual physical reservoir computing using numerical simulation revealed the relationship between vortices and info processing capacity.
A new study by California Institute of Technology researchers found that a computer program can accurately predict which paintings a person will like, using low-level visual attributes such as contrast, saturation, and hue. The program achieved similar accuracy to deep convolutional neural networks in predicting art preferences.
A recent study by Hebrew University researchers identified molecular factors that allow birds to fly, differing from mammals and reptiles. The ephrin-B3 molecule plays a crucial role in coordinating wing movement, enabling birds to flap and take flight.
The European Virtual Institute will study the neural basis of emotion using a Marie Sklodowska-Curie Innovative Training Network, focusing on the role of the cerebellum in controlling emotions. The network aims to develop new therapeutic strategies for emotional disorders by combining fundamental and clinical research.
Scientists at Osaka University employ machine learning algorithms to assess the remaining useful life of mechanical rolling bearings, which may lead to industrial cost savings and fewer discarded parts. The new method improves prediction accuracy by about 32%.