SEARCH · Search NASA

Results for “Comparative visualization”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 217 records · Page 12

All sources NASA NTRS DOE OSTI

DOE OSTI2022

The VTK-m Users' Guide (V.1.9)

High-performance computing relies on ever finer threading. Advances in processor technology include ever greater numbers of cores, hyperthreading, accelerators with integrated blocks of cores, and special vectorized instructions, all of which require more software parallelism to achieve peak performance. Traditional visualization solutions cannot support this extreme level of concurrency. Extreme scale systems require a new programming model and a fundamental change in how we design algorithms. To address these issues we created VTK-m: the visualization toolkit for multi-/many-core architectures. VTK-m supports a number of algorithms and the ability to design further algorithms through a top-down design with an emphasis on extreme parallelism. VTK-m also provides support for finding and building links across topologies, making it possible to perform operations that determine manifold surfaces, interpolate generated values, and find adjacencies. Although VTK-m provides a simplified high-level interface for programming, its template-based code removes the overhead of abstraction. VTK-m simplifies the development of parallel scientific visualization algorithms by providing a framework of supporting functionality that allows developers to focus on visualization operations. Consider the listings in Figure 1.1 that compares the size of the implementation for the Marching Cubes algorithm in VTK-m with the equivalent reference implementation in the CUDA software development kit. Because VTK-m internally manages the parallel distribution of work and data, the VTK-m implementation is shorter and easier to maintain. Additionally, VTK-m provides data abstractions not provided by other libraries that make code written in VTK-m more versatile.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

The VTK-m Users' Guide (V.2.0)

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Development of a Digital Twin for Hydrogen Dispersion and Safety Assessment in an Electrolyzer Based Hydrogen Production Facility

Digital twin models are virtual representations of physical systems that use real-time data to simulate and optimize performance. This study presents the development and initial implementation of a digital twin (DT) for the electrolyzer-based hydrogen production facility at NREL's Advanced Research on Integrated Energy Systems (ARIES), focused on enhancing safety and optimizing sensor placement through physics-based simulations and metadata integration. The DT incorporates detailed facility-specific information, including component layout, leak locations, and controlled release parameters, to model hydrogen dispersion under varying environmental conditions. Using steady-state computational fluid dynamics (CFD) simulations informed by real meteorological data, such as wind speed, direction, and vertical wind profiles, the DT enables visualization of hydrogen plume behavior and spatial concentration distributions. Comparative analysis between high and low wind speed scenarios illustrates the significant influence of wind dynamics on plume shape and extent, with horizontal momentum dominating dispersion at higher speeds, while buoyancy effects become more prominent under low wind conditions. These simulations generate a rich dataset embedded within the DT, allowing users to assess potential leak outcomes and identify optimal sensor locations based on concentration thresholds. The model supports scenario-based analysis to guide safety strategies and equipment deployment for open-area hydrogen infrastructure. The digital twin thus serves as a dynamic platform for virtual prototyping, providing predictive insight into hydrogen behavior and enhancing risk-informed decision-making. This initial phase establishes a validated foundation for future integration of transient, uncontrolled leak scenarios and real-time sensor feedback, positioning the DT as a critical tool for safety design, operational planning, and adaptive monitoring in hydrogen systems. Overall, the approach demonstrates the value of combining environmental data with digital simulations to inform safer and more efficient deployment of hydrogen technologies.

08 HYDROGEN↗

DOE OSTI2026

Development of a Digital Twin for Hydrogen Dispersion and Safety Assessment in an Electrolyzer-Based Hydrogen Production Facility: Preprint

Digital twin models are virtual representations of physical systems that use real-time data to simulate and optimize performance. This study presents the development and initial implementation of a digital twin (DT) for the electrolyzer-based hydrogen production facility at the National Renewable Energy Laboratory (NREL)'s Advanced Research on Integrated Energy Systems (ARIES), focused on enhancing safety and optimizing sensor placement through physics-based simulations and metadata integration. The DT incorporates detailed facility-specific information, including component layout, leak locations, and controlled release parameters, to model hydrogen dispersion under varying environmental conditions. Using steady-state computational fluid dynamics (CFD) simulations informed by real meteorological data, such as wind speed, direction, and vertical wind profiles, the DT enables visualization of hydrogen plume behavior and spatial concentration distributions. Comparative analysis between high and low wind speed scenarios illustrates the significant influence of wind dynamics on plume shape and extent, with horizontal momentum dominating dispersion at higher speeds, while buoyancy effects become more prominent under low wind conditions. These simulations generate a rich dataset embedded within the DT, allowing users to assess potential leak outcomes and identify optimal sensor locations based on concentration thresholds. The model supports scenario-based analysis to guide safety strategies and equipment deployment for open-area hydrogen infrastructure. The digital twin thus serves as a dynamic platform for virtual prototyping, providing predictive insight into hydrogen behavior and enhancing risk-informed decision-making. This initial phase establishes a validated foundation for future integration of transient, uncontrolled leak scenarios and real-time sensor feedback, positioning the DT as a critical tool for safety design, operational planning, and adaptive monitoring in hydrogen systems. Overall, the approach demonstrates the value of combining environmental data with digital simulations to inform safer and more efficient deployment of hydrogen technologies.

08 HYDROGEN↗

DOE OSTI2015

Distributed computing strategies for processing of FT-ICR MS imaging datasets for continuous mode data visualization

High-resolution Fourier transform ion cyclotron resonance (FT-ICR) mass spectrometry imaging enables the spatial mapping and identification of biomolecules from complex surfaces. The need for long time-domain transients, and thus large raw file sizes, results in a large amount of raw data (“big data”) that must be processed efficiently and rapidly. This can be compounded by largearea imaging and/or high spatial resolution imaging. For FT-ICR, data processing and data reduction must not compromise the high mass resolution afforded by the mass spectrometer. The continuous mode “Mosaic Datacube” approach allows high mass resolution visualization (0.001 Da) of mass spectrometry imaging data, but requires additional processing as compared to featurebased processing. We describe the use of distributed computing for processing of FT-ICR MS imaging datasets with generation of continuous mode Mosaic Datacubes for high mass resolution visualization. An eight-fold improvement in processing time is demonstrated using a Dutch nationally available cloud service.

Environmental Molecular Sciences Laboratory↗

DOE OSTI2020

How “hot” are hotspots: Statistically localizing the high-activity areas on soil and rhizosphere images

The topic of microbial hotspots in soil requires not only visualizing their spatial distribution and biochemical analyses, but also statistical approaches to identify these hotspots and separate them from the surrounding activities (background). We hypothesized that each hotspot type (e.g. enzyme activities in the rhizosphere, root exudation, localization of herbicide accumulation) is a result of local process driven by biotic and/or abiotic factors, and the process rates in the hotspots are much faster than those in the soil background. We further hypothesized that the background and hotspot activities in soil belong to different statistical distributions. Consequently, hotspot determination should be based on statistical separation of activities significantly higher than the background. We analyzed for the statistical distributions of grey values on three groups of published images: 1) 14 C images of carbon input by roots into the rhizosphere, 2) 14 C glyphosate accumulation in the plant, and 3) zymogram of leucine aminopeptidase activity in rooted soil. The two Gaussian distributions were fit (the first representing the background, the second the hotspots) to the distribution of grey values in the images, the parameters (means and standard deviations, SD) of the fitted distributions were calculated, and the background was removed. Thus, we identified hotspots as areas outside of the Mean+2SD image intensity (corresponding to the upper ~ 2.5% of activity, being over 97.5% of background values) and finally, visualized images of solely hotspot locations. Finally, these results were compared with previously used decisions on hotspot intensity thresholding (i.e. Top-25% and 17 standard thresholding approaches in ImageJ) and discussed the advantages of the Mean+2SD as well as Mean+3SD approaches. These advantages include: i) simple unification of the thresholding approach for several imaging methods with various principles of activity distribution, ii) identification of hotspots with various activity levels, iii) analysis of “time-specific” hotspots in temporal sequences of images. Compared with 17 standard thresholding methods, we concluded that objectively elucidating and separating the hotspots should be based on statistical distribution analysis, e.g. using the Mean+2SD or Mean+3SD approaches. Furthermore, this simple Mean+2SD approach delivered suitable results for three groups of images and so, helps to understand the processes responsible for the highest activities and elucidate hotspots.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2020

An effective survey method for studying volant species activity and behavior at tall structures

The effects of anthropogenic modification of air space on wildlife, particularly volant species, is not fully understood. Thus, it is essential to understand wildlife-interactions with tall structures to implement effective mitigation strategies. Yet, we are currently lacking standard protocols for visual surveys of wildlife behavior at such heights. Our study sought to determine an effective, repeatable method using readily available night vision and thermal technology to survey wildlife at tall structures. Using bats as the taxonomic group of interest, we (1) created a key to identify bats and their behavior, (2) compared the effectiveness of 2 different technologies, and (3) assessed optimal equipment placement to visually capture bat activity and behavior in proximity to wind turbine towers. For the latter, we tested thermal cameras at four distances from the base of the tower. The results of our study revealed that thermal cameras captured ∼34% more flying animals than night vision at a 2 m distance. However, due to the heat signature of the turbine towers themselves, it was challenging to identify behaviors and interactions that occurred in close proximity to the towers. In contrast, it was difficult to identify bats approaching the towers using night vision, yet we were able to clearly observe interactions with the towers themselves. With regards to equipment placement, we visually captured more bats with the thermal cameras placed 2 m from the tower base compared to farther distances. From our findings, we recommend that when using either thermal or night vision technology at tall structures, they be placed 2 m from the base to effectively observe interactions along the length of these structures. In addition, we further recommend that consideration be given to the use of these two technology types together to effectively conduct such surveys. If these survey techniques are incorporated into standard protocols, future surveys at a variety of tall structures are likely to become comparable and repeatable, thereby more effectively informing any mitigation strategies that may be required.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2015

SLoWPoKES-II: 100,000 WIDE BINARIES IDENTIFIED IN SDSS WITHOUT PROPER MOTIONS

We present the Sloan Low-mass Wide Pairs of Kinematically Equivalent Stars (SLoWPoKES)-II catalog of low-mass visual binaries identified from the Sloan Digital Sky Survey (SDSS) by matching photometric distances. The candidate pairs are vetted by comparing the stellar information. The candidate pairs are vetted by comparing the stellar density at their respective Galactic positions to Monte Carlo realizations of a simulated Milky Way. In this way, we are able to identify large numbers of bona fide wide binaries without the need for proper motions. Here, 105,537 visual binaries with angular separations of ∼1–20″ were identified, each with a probability of chance alignment of ≤5%. This is the largest catalog of bona fide wide binaries to date, and it contains a diversity of systems—in mass, mass ratios, binary separations, metallicity, and evolutionary states—that should facilitate follow-up studies to characterize the properties of M dwarfs and white dwarfs. There is a subtle but definitive suggestion of multiple populations in the physical separation distribution, supporting earlier findings. We suggest that wide binaries are composed of multiple populations, most likely representing different formation modes. There are 141 M7 or later wide binary candidates, representing a seven-fold increase over the number currently known. These binaries are too wide to have been formed via the ejection mechanism. Finally, we found that 6% of spectroscopically confirmed M dwarfs are not included in the SDSS STAR catalog; they are misclassified as extended sources due to the presence of a nearby or partially resolved companion. The SLoWPoKES-II catalog is publicly available to the entire community on the World Wide Web via the Filtergraph data visualization portal.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2025

Spatial profiling of the interplay between cell type- and vision-dependent transcriptomic programs in the visual cortex

How early sensory experience during “critical periods” of postnatal life affects the organization of the mammalian neocortex at the resolution of neuronal cell types is poorly understood. We previously reported that the functional and molecular profiles of layer 2/3 (L2/3) cell types in the primary visual cortex (V1) are vision-dependent [S. Chenget al.,Cell185, 311–327.e24 (2022)]. Here, we characterize the spatial organization of L2/3 cell types with and without visual experience. Spatial transcriptomic profiling based on 500 genes recapitulates the zonation of L2/3 cell types along the pial–ventricular axis in V1. By applying multitasking theory, we suggest that the spatial zonation of L2/3 cell types is linked to the continuous nature of their gene expression profiles, which can be represented as a 2D manifold bounded by three archetypal cell types. By comparing normally reared and dark reared L2/3 cells, we show that visual deprivation-induced transcriptomic changes comprise two independent gene programs. The first, induced specifically in the visual cortex, includes immediate-early genes and genes associated with metabolic processes. It manifests as a change in cell state that is orthogonal to cell-type-specific gene expression programs. By contrast, the second program impacts L2/3 cell-type identity, regulating a subset of cell-type-specific genes and shifting the distribution of cells within the L2/3 cell-type manifold. Through an integrated analysis of spatial transcriptomics with single-nucleus RNA-seq data, we describe how vision patterns cortical L2/3 cell types during the critical period.

Science & Technology - Other Topics↗

DOE OSTI2014

Diffuse interstellar bands versus known atomic and molecular species in the interstellar medium of M82 toward SN 2014J

We discuss the absorption due to various constituents of the interstellar medium (ISM) of M82 seen in moderately high-resolution, high signal-to-noise ratio optical spectra of SN 2014J. Complex absorption from M82 is seen, at velocities 45 ≲ v {sub LSR} ≲ 260 km s{sup –1}, for Na I, K I, Ca I, Ca II, CH, CH{sup +}, and CN; many of the diffuse interstellar bands (DIBs) are also detected. Comparisons of the column densities of the atomic and molecular species and the equivalent widths of the DIBs reveal both similarities and differences in relative abundances, compared to trends seen in the ISM of our Galaxy and the Magellanic Clouds. Of the 10 relatively strong DIBs considered here, 6 (including λ5780.5) have strengths within ±20% of the mean values seen in the local Galactic ISM, for comparable N(K I); 2 are weaker by 20%-45% and 2 (including λ5797.1) are stronger by 25%-40%. Weaker than 'expected' DIBs (relative to N(K I), N(Na I), and E(B – V)) in some Galactic sight lines and toward several other extragalactic supernovae appear to be associated with strong CN absorption and/or significant molecular fractions. While the N(CH)/N(K I) and N(CN)/N(CH) ratios seen toward SN 2014J are similar to those found in the local Galactic ISM, the combination of high N(CH{sup +})/N(CH) and high W(5797.1)/W(5780.5) ratios has not been seen elsewhere. The centroids of many of the M82 DIBs are shifted relative to the envelope of the K I profile—likely due to component-to-component variations in W(DIB)/N(K I) that may reflect the molecular content of the individual components. We compare estimates for the host galaxy reddening E(B – V) and visual extinction A {sub V} derived from the various interstellar species with the values estimated from optical and near-IR photometry of SN 2014J.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2015

SU-E-J-275: Review - Computerized PET/CT Image Analysis in the Evaluation of Tumor Response to Therapy

Purpose: To review the literature in using computerized PET/CT image analysis for the evaluation of tumor response to therapy. Methods: We reviewed and summarized more than 100 papers that used computerized image analysis techniques for the evaluation of tumor response with PET/CT. This review mainly covered four aspects: image registration, tumor segmentation, image feature extraction, and response evaluation. Results: Although rigid image registration is straightforward, it has been shown to achieve good alignment between baseline and evaluation scans. Deformable image registration has been shown to improve the alignment when complex deformable distortions occur due to tumor shrinkage, weight loss or gain, and motion. Many semi-automatic tumor segmentation methods have been developed on PET. A comparative study revealed benefits of high levels of user interaction with simultaneous visualization of CT images and PET gradients. On CT, semi-automatic methods have been developed for only tumors that show marked difference in CT attenuation between the tumor and the surrounding normal tissues. Quite a few multi-modality segmentation methods have been shown to improve accuracy compared to single-modality algorithms. Advanced PET image features considering spatial information, such as tumor volume, tumor shape, total glycolytic volume, histogram distance, and texture features have been found more informative than the traditional SUVmax for the prediction of tumor response. Advanced CT features, including volumetric, attenuation, morphologic, structure, and texture descriptors, have also been found advantage over the traditional RECIST and WHO criteria in certain tumor types. Predictive models based on machine learning technique have been constructed for correlating selected image features to response. These models showed improved performance compared to current methods using cutoff value of a single measurement for tumor response. Conclusion: This review showed that computerized PET/CT image analysis holds great potential to improve the accuracy in evaluation of tumor response. This work was supported in part by the National Cancer Institute Grant R01CA172638.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2017

The Impact of Streaming Data on Sensemaking with Mixed-Initative Visual Analytics

Visual data analysis helps people gain insights into data via interactive visualizations. People generate and test hypotheses and questions about data in context of the domain. This process can generally be referred to as sensemaking. Much of the work on studying sensemaking (and creating visual analytic techniques in support of it) has been focused on static datasets. However, how do the cognitive processes of sensemaking change when data are changing? Further, what implication for design does this create for mixed-initiative visual analytics systems? This paper presents the results of a user study analyzing the impact of streaming data on sensemaking. To perform this study, we developed a mixed-initiative visual analytic prototype, the Streaming Canvas, that affords the analysis of streaming text data. We compare the sensemaking process of people using this tool for a static and streaming dataset. We present the results of this study and discuss the implications on future visual analytic systems that combine machine learning and interactive visualization to help people make sense of streaming data.

visualization↗

DOE OSTI2017

Novel Super-Resolution Approach to Time-Resolved Volumetric 4-Dimensional Magnetic Resonance Imaging With High Spatiotemporal Resolution for Multi-Breathing Cycle Motion Assessment

Purpose: To develop and evaluate a super-resolution approach to reconstruct time-resolved 4-dimensional magnetic resonance imaging (TR-4DMRI) with a high spatiotemporal resolution for multi-breathing cycle motion assessment. Methods and Materials: A super-resolution approach was developed to combine fast 3-dimensional (3D) cine MRI with low resolution during free breathing (FB) and high-resolution 3D static MRI during breath hold (BH) using deformable image registration. A T1-weighted, turbo field echo sequence, coronal 3D cine acquisition, partial Fourier approximation, and SENSitivity Encoding parallel acceleration were used. The same MRI pulse sequence, field of view, and acceleration techniques were applied in both FB and BH acquisitions; the intensity-based Demons deformable image registration method was used. Under an institutional review board–approved protocol, 7 volunteers were studied with 3D cine FB scan (voxel size: 5 × 5 × 5 mm{sup 3}) at 2 Hz for 40 seconds and a 3D static BH scan (2 × 2 × 2 mm{sup 3}). To examine the image fidelity of 3D cine and super-resolution TR-4DMRI, a mobile gel phantom with multi-internal targets was scanned at 3 speeds and compared with the 3D static image. Image similarity among 3D cine, 4DMRI, and 3D static was evaluated visually using difference image and quantitatively using voxel intensity correlation and Dice index (phantom only). Multi-breathing-cycle waveforms were extracted and compared in both phantom and volunteer images using the 3D cine as the references. Results: Mild imaging artifacts were found in the 3D cine and TR-4DMRI of the mobile gel phantom with a Dice index of >0.95. Among 7 volunteers, the super-resolution TR-4DMRI yielded high voxel-intensity correlation (0.92 ± 0.05) and low voxel-intensity difference (<0.05). The detected motion differences between TR-4DMRI and 3D cine were −0.2 ± 0.5 mm (phantom) and −0.2 ± 1.9 mm (diaphragms). Conclusion: Super-resolution TR-4DMRI has been reconstructed with adequate temporal (2 Hz) and spatial (2 × 2 × 2 mm{sup 3}) resolutions. Further TR-4DMRI characterization and improvement are necessary before clinical applications. Multi-breathing cycles can be examined, providing patient-specific breathing irregularities and motion statistics for future 4D radiation therapy.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

DOE OSTI2021

Galaxy morphological classification catalogue of the Dark Energy Survey Year 3 data with convolutional neural networks

ABSTRACT We present in this paper one of the largest galaxy morphological classification catalogues to date, including over 20 million galaxies, using the Dark Energy Survey (DES) Year 3 data based on convolutional neural networks (CNNs). Monochromatic i-band DES images with linear, logarithmic, and gradient scales, matched with debiased visual classifications from the Galaxy Zoo 1 (GZ1) catalogue, are used to train our CNN models. With a training set including bright galaxies (16 ≤ i < 18) at low redshift (z < 0.25), we furthermore investigate the limit of the accuracy of our predictions applied to galaxies at fainter magnitude and at higher redshifts. Our final catalogue covers magnitudes 16 ≤ i < 21, and redshifts z < 1.0, and provides predicted probabilities to two galaxy types – ellipticals and spirals (disc galaxies). Our CNN classifications reveal an accuracy of over 99 per cent for bright galaxies when comparing with the GZ1 classifications (i < 18). For fainter galaxies, the visual classification carried out by three of the co-authors shows that the CNN classifier correctly categorizes discy galaxies with rounder and blurred features, which humans often incorrectly visually classify as ellipticals. As a part of the validation, we carry out one of the largest examinations of non-parametric methods, including ∼100 ,000 galaxies with the same coverage of magnitude and redshift as the training set from our catalogue. We find that the Gini coefficient is the best single parameter discriminator between ellipticals and spirals for this data set.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2017

Visualizations of Travel Time Performance Based on Vehicle Reidentification Data

This paper provides a visual reference of the breadth of arterial performance phenomena based on travel time measures obtained from reidentification technology that has proliferated in the past 5 years. These graphical performance measures are revealed through overlay charts and statistical distribution as revealed through cumulative frequency diagrams (CFDs). With overlays of vehicle travel times from multiple days, dominant traffic patterns over a 24-h period are reinforced and reveal the traffic behavior induced primarily by the operation of traffic control at signalized intersections. A cumulative distribution function in the statistical literature provides a method for comparing traffic patterns from various time frames or locations in a compact visual format that provides intuitive feedback on arterial performance. The CFD may be accumulated hourly, by peak periods, or by time periods specific to signal timing plans that are in effect. Combined, overlay charts and CFDs provide visual tools with which to assess the quality and consistency of traffic movement for various periods throughout the day efficiently, without sacrificing detail, which is a typical byproduct of numeric-based performance measures. These methods are particularly effective for comparing before-and-after median travel times, as well as changes in interquartile range, to assess travel time reliability.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

DOE OSTI2024

Large language model evaluation for high–performance computing software development

We apply AI-assisted large language model (LLM) capabilities of GPT-3 targeting high-performance computing (HPC) kernels for (i) code generation, and (ii) auto-parallelization of serial code in C ++, Fortran, Python and Julia. Our scope includes the following fundamental numerical kernels: AXPY, GEMV, GEMM, SpMV, Jacobi Stencil, and CG, and language/programming models: (1) C++ (e.g., OpenMP [including offload], OpenACC, Kokkos, SyCL, CUDA, and HIP), (2) Fortran (e.g., OpenMP [including offload] and OpenACC), (3) Python (e.g., numpy, Numba, cuPy, and pyCUDA), and (4) Julia (e.g., Threads, CUDA.jl, AMDGPU.jl, and KernelAbstractions.jl). Kernel implementations are generated using GitHub Copilot capabilities powered by the GPT-based OpenAI Codex available in Visual Studio Code given simple + + prompt variants. To quantify and compare the generated results, we propose a proficiency metric around the initial 10 suggestions given for each prompt. For auto-parallelization, we use ChatGPT interactively giving simple prompts as in a dialogue with another human including simple “prompt engineering” follow ups. Results suggest that correct outputs for C++ correlate with the adoption and maturity of programming models. For example, OpenMP and CUDA score really high, whereas HIP is still lacking. We found that prompts from either a targeted language such as Fortran or the more general-purpose Python can benefit from adding language keywords, while Julia prompts perform acceptably well for its Threads and CUDA.jl programming models. Finally, we expect to provide an initial quantifiable point of reference for code generation in each programming model using a state-of-the-art LLM. Overall, understanding the convergence of LLMs, AI, and HPC is crucial due to its rapidly evolving nature and how it is redefining human-computer interactions.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

Comparison of Machine Learning and Deep Learning for View Identification from Cardiac Magnetic Resonance Images

Background: Artificial intelligence is increasingly utilized to aid in the interpretation of cardiac magnetic resonance (CMR) studies. One of the first steps is the identification of the imaging plane depicted, which can be achieved by both deep learning (DL) and classical machine learning (ML) techniques without user input. We aimed to compare the accuracy of ML and DL for CMR view classification and to identify potential pitfalls during training and testing of the algorithms. Methods: To train our DL and ML algorithms, we first established datasets by retrospectively selecting 200 CMR cases. The models were trained using two different cohorts (passively and actively curated) and applied data augmentation to enhance training. Once trained, the models were validated on an external dataset, consisting of 20 cases acquired at another center. We then compared accuracy metrics and applied class activation mapping (CAM) to visualize DL model performance. Results: The DL and ML models trained with the passively-curated CMR cohort were 99.1% and 99.3% accurate on the validation set, respectively. However, when tested on the CMR cases with complex anatomy, both models performed poorly. After training and testing our models again on all 200 cases (active cohort), validation on the external dataset resulted in 95% and 90% accuracy, respectively. The CAM analysis depicted heat maps that demonstrated the importance of carefully curating the datasets to be used for training. Conclusions: Both DL and ML models can accurately classify CMR images, but DL outperformed ML when classifying images with complex heart anatomy.

artificial intelligence↗

DOE OSTI2023

Development of an accelerator-based neutron source to prototype Mo-99 production, part I: A liquid LBE windowless target

In this article, Molybdenum-99 (Mo-99)’s decay product, technetium-99 (Tc-99 m), is one of the most critical isotopes for medical diagnostics. To provide U.S. domestic supply of Mo-99 without using high-enriched uranium (HEU), a subcritical uranium target assembly (UTA) is irradiated by an accelerator-based neutron source to create Mo-99 through fission. This study discusses the development of the accelerator-based neutron source. The high-energy electrons from the accelerator irradiate a liquid lead-bismuth eutectic (LBE) target to produce neutrons. Part I of this work focuses on numerical and experimental analysis towards the development of a liquid LBE windowless target. Unlike the existing windowless targets in literature, the current design creates a vertical free surface for a beam to irradiate. First, a hydrodynamic analysis of the LBE windowless target is performed. Simplified analytical calculations are assisted by 2D computational fluid dynamics (CFD) simulations to design the target, with the focus on eliminating recirculation zones and avoiding cavitation. With the optimized geometry, the experimental study is performed to investigate the flow hydrodynamics using liquid LBE. The experiments (1) compare pressure drop in the system to correlation predictions; (2) visualize the free surface liquid LBE flow from the beam view; (3) validate the LBE flow profile using temperature sensitive paint from the side view; and (4) validate the liquid LBE film thickness using gamma densitometer measurements. Second, the power handling capability of the designed windowless target is investigated. The divider plate in the current design is susceptible to overheating due to the thin LBE film in front. As LBE erosion and corrosion is likely to occur at an LBE velocity of 2.0 m/s and temperature above 500 °C, a power limit of 10 kW of beam power was established to prevent this corrosion from occurring, which is calculated by a Nusselt number correlation. The divider plate surface temperature at 10 kW agrees well with the 3D CFD simulation results. Part I demonstrates the fundamental physics in liquid LBE windowless target design and associated testing. A companion paper, Part II will demonstrate how to couple this windowless target into the Mo-99 production system, including an accelerator system operating under an ultra-high vacuum and the UTA cooled by water at room temperature.

43 PARTICLE ACCELERATORS↗