A5003443752- Cell Image Analysis Techniques
- Machine Learning in Materials Science
- Electron and X-Ray Spectroscopy Techniques
- Advanced Electron Microscopy Techniques and Applications
- Industrial Vision Systems and Defect Detection
- Advanced Fluorescence Microscopy Techniques
- Photoacoustic and Ultrasonic Imaging
- Regulation of Appetite and Obesity
- Hydrogen Storage and Materials
- Machine Learning in Bioinformatics
- Hybrid Renewable Energy Systems
- Nuclear Materials and Properties
- Medical Imaging and Analysis
- Experimental Learning in Engineering
- Ferroelectric and Piezoelectric Materials
- Anomaly Detection Techniques and Applications
- Magnetic and transport properties of perovskites and related materials
- Metabolomics and Mass Spectrometry Studies
- Brain Tumor Detection and Classification
- Metalloenzymes and iron-sulfur proteins
- Electronic and Structural Properties of Oxides
- Isotope Analysis in Ecology
- Machine Learning and Data Classification
- Data Analysis with R
- Sleep and Wakefulness Research
Pacific Northwest National Laboratory
2022-2024
Washington State University
2012
Artificial intelligence (AI) promises to reshape scientific inquiry and enable breakthrough discoveries in areas such as energy storage, quantum computing, biomedicine. Scanning transmission electron microscopy (STEM), a cornerstone of the study chemical materials systems, stands benefit greatly from AI-driven automation. However, present barriers low-level instrument control, well generalizable interpretable feature detection, make truly automated impractical. Here, we discuss design...
Understanding the relationship between evolution of microstructures irradiated LiAlO2 pellets and tritium diffusion, retention release could improve predictions tritium-producing burnable absorber rod performance. Given expert-labeled segmented images unirradiated pellets, we trained Deep Convolutional Neural Networks to segment into defect, grain, boundary classes. Qualitative microstructural information was calculated from these facilitate comparison pellets. We tested modifications...
The recent growth in data volumes produced by modern electron microscopes requires rapid, scalable, and flexible approaches to image segmentation analysis. Few-shot machine learning, which can richly classify images from a handful of user-provided examples, is promising route high-throughput However, current command-line implementations such be slow unintuitive use, lacking the real-time feedback necessary perform effective classification. Here we report on development Python-based graphical...
Neural sites that interact with the suprachiasmatic nuclei (SCN) to generate rhythms of unrestricted feeding remain unknown. We used targeted toxin, leptin conjugated saporin (Lep-SAP), examine importance receptor-B (LepR-B)-expressing neurons in arcuate nucleus (Arc) for generation circadian rhythms. Rats given Arc Lep-SAP injections were initially hyperphagic and rapidly became obese (the "dynamic phase" weight gain). During this phase, rats arrhythmic under 12:12-h light-dark (LD)...
Light-sheet microscopy has made possible the 3D imaging of both fixed and live biological tissue, with samples as large entire mouse brain. However, segmentation quantification that data remains a time-consuming manual undertaking. Machine learning methods promise possibility automating this process. This study seeks to advance performance prior models through optimizing transfer learning. We fine-tuned existing TrailMap model using expert-labeled from noradrenergic axonal structures in By...
To advance our ability to predict impacts of the protein scaffold on catalysis, robust classification schemes define features proteins that will influence reactivity are needed. One these is a protein's metal-binding ability, as metals critical catalytic conversion by metalloenzymes. As step toward realizing this goal, we used convolutional neural networks (CNNs) enable metal cofactor binding pocket within scaffold. CNNs images be classified based multiple levels detail in image, from edges...
We report here the creation of a graphical user interface (GUI) for Data Extraction Integrated Multidimensional Spectrometry (DEIMoS) tool. DEIMoS is Python package that processes data from high-dimensional mass spectrometry measurements. It divided into several modules, each representing processing step such as peak detection, alignment, and tandem spectra extraction deconvolution. The inputs outputs can include millions N-dimensional points, which be challenging to visualize in way...
Abstract Light-sheet microscopy has made possible the 3D imaging of both fixed and live biological tissue, with samples as large entire mouse brain. However, segmentation quantification that data remains a time-consuming manual undertaking. Machine learning methods promise possibility automating this process. This study seeks to advance performance prior models through optimizing transfer learning. We fine-tuned existing TrailMap model using expert-labeled from noradrenergic axonal...
The functionality of materials is critically dependent on structures and electronic properties. valence transition metal (TM) cations in complex oxides key each TM exhibits a fixed range stable values. For Ni, the formal oxidation state 2+, but can reach 3+ for certain compositions. By limiting SrNiO${}_{3}$ film thickness to 1 unit cell (SrNiO${}_{3}$)${}_{1}$/(SrTiO${}_{3}$)${}_{n}$ superlattices, authors demonstrate that Ni exceeds whereas Ti remains at 4+, neither changes with $n$. This...
An abstract is not available for this content so a preview has been provided. As you have access to content, full PDF via the ‘Save PDF’ action button.
Measurement of uncertainty predictions from machine learning methods is important across scientific domains and applications. We present, to our knowledge, the first such technique that quantifies a classifier accounts for both classifier's belief performance. prove method provides an accurate estimate probability outputs two neural networks are correct by showing expected calibration error less than 0.2% on binary classifier, 3% semantic segmentation network with extreme class imbalance....