Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “algorithm selection”

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.

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

At least 1,567 records · Page 87

Real-Time Anomaly Detection for Charge-Based Triggering in LArTPCs

Modern particle detectors, including liquid argon time projection chambers (LArTPCs), collect a vast amount of data, making it impractical to save everything for offline analysis. As a result, these experiments need to employ different down-selection techniques during data acquisition, referred to as triggering. In this talk, I will present a framework that would enable real-time, data-driven triggering for LArTPCs, using anomaly detection algorithms implemented on Field-Programmable Gate Arrays (FPGAs). Drawing on a study that makes use of collected charge data from the MicroBooNE LArTPC Public Dataset, I will discuss the overall performance of such algorithms and potential applications for future neutrino experiments.

43 PARTICLE ACCELERATORS↗

Recherche du boson de Higgs standard dans le canal WH a l'experience D0 aupres du Tevatron

The Higgs mechanism provides the Standard Model with an appropriate theory of the origin of the mass of gauge bosons and elementary fermions. The Higgs boson has not yet been discovered, but a lower limit on its mass has been set at 114.4 GeV at 95% con dence level by the Lep collaborations. Higgs searches are now being pursued at the Tevatron, a proton - anti-proton collider with a centreof- mass energy of 1.96TeV. At low mass the most sensitive channel is the associated production of a Higgs with a W boson. An analysis has been performed in the decay channel in which the Higgs goes into a b$\bar{b}$ pair and the W decays to a muon and a neutrino with 1 fb -1 of Run IIa data recorded by the DØ detector. The analysis relies on all the sub-detector components but most particularly on the calorimeter which is essential for reconstruction of the b$\bar{b}$ system. Good tracking and a b-identi cation neural network tool provide improved b-tagging performance which is crucial for this analysis. Because the energy resolution of jets is of paramount importance when hunting for a two-jet resonance bump, work has also been conducted on devising a better-performing calibration for jets which exhibit a muon and a neutrino in their fragmentation chain. The WH analysis is performed on a W+2 and W+3 jet event topology. Events containing a muon, missing transverse energy and two or three jets are selected; the jets are then b-tagged. The doubletagged and single-tagged channels are analysed separately so as to provide additional sensitivity. Signal-background separation via a neural network algorithme has furthermore been developed in the analysis in order to enhance sensitivity. This search for a standard Higgs boson has been conducted for Higgs masses ranging from 100GeV to 150GeV. Upper limits on production cross-section times branching ratio have been set. For a Higgs mass of 115GeV, the upper limit is set at 2.00 pb at 95% con dence level in the channel WH → μ b$\bar{b}$ and 1.37 pb when combining the two channels WH → μ b$\bar{b}$ and WH → e b$\bar{b}$, which is to be compared to the theoretical prediction of 0.136 pb. Those results have been submitted for publication, and are used in the global combination of Cdf's and DØ's Higgs analyses.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Generating An Advanced Cross-section Library For HTGR Pebble Bed Depletion Calculations Using Reduced-Order Model Generation Techniques

For code development, Advanced Reactor Technologies - Gas Cooled Reactors Program (ART-GCR) rely on a collaboration with the Nuclear Energy Advanced Modeling and Simulation (NEAMS) program, but the cross sections generation and the methodology definition is part of this program area goals. Based on previous studies in FY23, the size of microscopic cross section libraries increases rapidly with the number of tabulations, requiring significant amount of memory and drastically slowing down the Griffin calculations when evaluating cross sections via the multivariate linear interpolation approach. Rising to these challenges, this work investigates constructing Reduced-order Models (ROMs) for the multi-group microscopic cross sections to accelerate the cross section evaluation in Griffin. A database of multigroup cross sections is first collected considering all possible parameters that a designer could change for optimization. Down-selection of the ROM techniques afterward shows Deep Neural Network (DNN) as the best candidate when jointly consider memory efficiency, predictive accuracy, computational cost, scalability, flexibility and ease of implementation of the algorithms in comparison to the multidimensional interpolation. This work develops a specific interface that enables the cross section predictions using pre-trained DNN models into Griffin leveraging the existing ROM capabilities. DNNs have been trained for all isotopes for use in Griffin. Preliminary Griffin testing shows that DNNs exhibit exceptional predictive accuracy and the use of DNNs provides orders of magnitude improvement in memory efficiency compared to conventional interpolation techniques. With such ROM techniques, it holds great promise to further increase the fidelity of the Pebble Bed Reactor (PBR) simulation by increasing the number of tabulations/state variables during cross section evaluation, while maintaining the computational cost affordable in Griffin.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Search for Technicolor Particles Produced in Association with a W Boson at CDF

We present a search for the technicolor particles {rho}{sub T} and {pi}{sub T} in the process p{bar p} {yields} {rho}{sub T} {yields} W{pi}{sub T} at a center of mass energy of {radical}s = 1.96 TeV. The search uses a data sample corresponding to approximately 1.9 fb{sup -1} of integrated luminosity accumulated by the CDF II detector at the Fermilab Tevatron. The event signature we consider is W {yields} {ell}{nu} and {pi}{sub T} {yields} b{bar b}, b{bar c} or b{bar u} depending on the {pi}{sub T} charge. We select events with a single high-p{sub T} electron or muon, large missing transverse energy, and two jets. Jets corresponding to bottom quarks are identified with multiple b-tagging algorithms. The observed number of events and the invariant mass distributions are consistent with the standard model background expectations, and we exclude a region at 95% confidence level in the {rho}{sub T}-{pi}{sub T} mass plane. As a result, a large fraction of the region m({rho}{sub T}) = 180-250 GeV/c{sup 2} and m({pi}{sub T}) = 95-145 GeV/c{sup 2} is excluded.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Exploring the bulk of tidal charged micro-black holes

We study the bulk corresponding to tidal charged brane-world black holes. We employ a propagating algorithm which makes use of the three-dimensional multipole expansion and analytically yields the metric elements as functions of the five-dimensional coordinates and of the Adler-Deser-Misner mass, tidal charge, and brane tension. Since the projected brane equations cannot determine how the charge depends on the mass, our main purpose is to select the combinations of these parameters for which black holes of microscopic size possess a regular bulk. Our results could, in particular, be relevant for a better understanding of TeV-scale black holes.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

In silico discovery of metal-organic frameworks for precombustion CO 2 capture using a genetic algorithm

Discovery of new adsorbent materials with a high CO 2 working capacity could help reduce CO 2 emissions from newly commissioned power plants using precombustion carbon capture. High-throughput computational screening efforts can accelerate the discovery of new adsorbents but sometimes require significant computational resources to explore the large space of possible materials. Here in this paper we report the in silico discovery of high-performing adsorbents for precombustion CO 2 capture by applying a genetic algorithm to efficiently search a large database of metal-organic frameworks (MOFs) for top candidates. High-performing MOFs identified from the in silico search were synthesized and activated and show a high CO 2 working capacity and a high CO 2 /H 2 selectivity. One of the synthesized MOFs shows a higher CO 2 working capacity than any MOF reported in the literature under the operating conditions investigated here.

36 MATERIALS SCIENCE↗

Avoiding excess computation in asynchronous evolutionary algorithms

Abstract Asynchronous evolutionary algorithms are becoming increasingly popular as a means of making full use of many processors while solving computationally expensive search and optimization problems. These algorithms excel at keeping large clusters fully utilized, but may sometimes inefficiently sample an excess of fast‐evaluating solutions at the expense of higher‐quality, slow‐evaluating ones. We have previously introduced a steady‐state parent selection strategy, SWEET (“Selection whilE EvaluaTing”), that sometimes selects individuals that are still being evaluated and allows them to reproduce early. We perform a takeover‐time analysis that confirms that this strategy gives slow‐evaluating individuals that have higher fitnesses an increased ability to multiply in the population. We also find that SWEET appears effective at improving optimization performance on problems in which solution quality is positively correlated with evaluation time. We evaluate our approach on six simulated real‐valued optimization problems and three real‐world applications: an autonomous vehicle controller problem that involves tuning a spiking neural network and two adversarial EA problems. We further evaluate SWEET versus a basic asynchronous process in a simulated setting. We present evidence that SWEET outperforms basic asynchronous processes in a use‐case in which performance is positively correlated with evaluation time, and performs comparably (and often better) than basic asynchronous processes in several use‐cases where performance is negatively correlated with evaluation time. That said, in the cases where performance and evaluation time are negatively correlated the variance of outcomes for SWEET is notably high.

97 MATHEMATICS AND COMPUTING↗

Ligand placement based on prior structures: the guided ligand-replacement method

A new module, Guided Ligand Replacement (GLR), has been developed in Phenix to increase the ease and success rate of ligand placement when prior protein-ligand complexes are available. The process of iterative structure-based drug design involves the X-ray crystal structure determination of upwards of 100 ligands with the same general scaffold (i.e. chemotype) complexed with very similar, if not identical, protein targets. In conjunction with insights from computational models and assays, this collection of crystal structures is analyzed to improve potency, to achieve better selectivity and to reduce liabilities such as absorption, distribution, metabolism, excretion and toxicology. Current methods for modeling ligands into electron-density maps typically do not utilize information on how similar ligands bound in related structures. Even if the electron density is of sufficient quality and resolution to allow de novo placement, the process can take considerable time as the size, complexity and torsional degrees of freedom of the ligands increase. A new module, Guided Ligand Replacement (GLR), was developed in Phenix to increase the ease and success rate of ligand placement when prior protein–ligand complexes are available. At the heart of GLR is an algorithm based on graph theory that associates atoms in the target ligand with analogous atoms in the reference ligand. Based on this correspondence, a set of coordinates is generated for the target ligand. GLR is especially useful in two situations: (i) modeling a series of large, flexible, complicated or macrocyclic ligands in successive structures and (ii) modeling ligands as part of a refinement pipeline that can automatically select a reference structure. Even in those cases for which no reference structure is available, if there are multiple copies of the bound ligand per asymmetric unit GLR offers an efficient way to complete the model after the first ligand has been placed. In all of these applications, GLR leverages prior knowledge from earlier structures to facilitate ligand placement in the current structure.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

SU-D-213-04: Accounting for Volume Averaging and Material Composition Effects in An Ionization Chamber Array for Patient Specific QA

Purpose: This study explores novel methods to address two significant challenges affecting measurement of patient-specific quality assurance (QA) with IBA’s Matrixx Evolution™ ionization chamber array. First, dose calculation algorithms often struggle to accurately determine dose to the chamber array due to CT artifact and algorithm limitations. Second, finite chamber size and volume averaging effects cause additional deviation from the calculated dose. Methods: QA measurements were taken with the Matrixx positioned on the treatment table in a solid-water Multi-Cube™ phantom. To reduce the effect of CT artifact, the Matrixx CT image set was masked with appropriate materials and densities. Individual ionization chambers were masked as air, while the high-z electronic backplane and remaining solid-water material were masked as aluminum and water, respectively. Dose calculation was done using Varian’s Acuros XB™ (V11) algorithm, which is capable of predicting dose more accurately in non-biologic materials due to its consideration of each material’s atomic properties. Finally, the exported TPS dose was processed using an in-house algorithm (MATLAB) to assign the volume averaged TPS dose to each element of a corresponding 2-D matrix. This matrix was used for comparison with the measured dose. Square fields at regularly-spaced gantry angles, as well as selected patient plans were analyzed. Results: Analyzed plans showed improved agreement, with the average gamma passing rate increasing from 94 to 98%. Correction factors necessary for chamber angular dependence were reduced by 67% compared to factors measured previously, indicating that previously measured factors corrected for dose calculation errors in addition to true chamber angular dependence. Conclusion: By comparing volume averaged dose, calculated with a capable dose engine, on a phantom masked with correct materials and densities, QA results obtained with the Matrixx Evolution™ can be significantly improved. In addition, necessary correction factors are reduced, allowing for more reliable and meaningful patient-specific QA measurements.

60 APPLIED LIFE SCIENCES↗

MULTI-SCALE CLEAN: A COMPARISON OF ITS PERFORMANCE AGAINST CLASSICAL CLEAN ON GALAXIES USING THINGS

A practical evaluation of the multi-scale CLEAN algorithm is presented. The data used in the comparisons are taken from The H I Nearby Galaxy Survey. The implementation of multi-scale CLEAN in the CASA software package is used, although comparisons are made against the very similar multi-resolution CLEAN algorithm implemented in AIPS. Both are compared against the classical CLEAN algorithm (as implemented in AIPS). The results of this comparison show that several of the well-known characteristics and issues of using classical CLEAN are significantly lessened (or eliminated completely) when using the multi-scale CLEAN algorithm. Importantly, multi-scale CLEAN significantly reduces the effects of the clean 'bowl' that is caused by missing short-spacings, and the 'pedestal' of low-level un-cleaned flux (which affects flux scales and resolution). Multi-scale CLEAN can clean down to the noise level without the divergence suffered by classical CLEAN. We discuss practical applications of the added contrast provided by multi-scale CLEAN using two selected astronomical examples: H I holes in the interstellar medium and anomalous gas structures outside the main galactic disk.

79 ASTRONOMY AND ASTROPHYSICS↗

Annealing a Follow-up Program: Improvement of the Dark Energy Figure of Merit for Optical Galaxy Cluster Surveys

The precision of cosmological parameters derived from galaxy cluster surveys is limited by uncertainty in relating observable signals to cluster mass. We demonstrate that a small mass-calibration follow-up program can significantly reduce this uncertainty and improve parameter constraints, particularly when the follow-up targets are judiciously chosen. To this end, we apply a simulated annealing algorithm to maximize the dark energy information at fixed observational cost, and find that optimal follow-up strategies can reduce the observational cost required to achieve a specified precision by up to an order of magnitude. Considering clusters selected from optical imaging in the Dark Energy Survey, we find that approximately 200 low-redshift X-ray clusters or massive Sunyaev-Zel'dovich clusters can improve the dark energy figure of merit by 50%, provided that the follow-up mass measurements involve no systematic error. In practice, the actual improvement depends on (1) the uncertainty in the systematic error in follow-up mass measurements, which needs to be controlled at the 5% level to avoid severe degradation of the results; and (2) the scatter in the optical richness-mass distribution, which needs to be made as tight as possible to improve the efficacy of follow-up observations.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

TU-H-CAMPUS-TeP1-01: Variable-Beam Fractionation for SAbR

Purpose: In current conventionally-fractionated as well as hypofractionated 3D conformal radiotherapy (CRT), the same beam arrangement is employed from fraction to fraction. We challenge this notion and postulate that by varying the beam arrangement between fractions we can achieve greater sparing of organs at risk (OARs) while maintaining PTV coverage. We use an inverse planning strategy using a swarm intelligence-based global optimization algorithm to exploit the additional degree of freedom represented by inter-fractional variation in beam angles. Methods: To evaluate our variable-beam fractionation (VBF) method, a 10-beam ITV-based conformal stereotactic ablative radiotherapy (CRT-SAbR) plan was optimized. In the clinical plan, 54 Gy was delivered to a 41cc lung tumor over 3 fractions. In VBF, each original clinically-assigned beam was multiplied to a bundle of n α-degree-spaced beams, n being number of fractions. Selection of α was a compromise between retaining similar tumor irradiation and separating inline OAR sub-regions. We optimized the beam fluence weights setting an upper limit for beam delivery duration (and implicitly, monitor units) along with clinical organ-based dose-volume constraints. Zero weights were allowed so that the optimization algorithm could remove unnecessary beams. All fractions in final plan had to deliver identical monitor units (MU) while satisfying a soft constraint on having no more than one beam from every n-beam bundle in each fraction. α was 10 degrees and the dose rate was 600 MU/min. Results: The VBF plan achieved significantly superior OAR sparing compared to the clinical internal target volume (ITV)-based plan. Setting maximum beam delivery duration to 13 seconds (well within breath-hold range), Esophagus Dmax, Heart Dmax, Spinal cord Dmax and Lung V13 were improved by 25%, 81%, 0% and 27%, respectively. Conclusion: We investigated a simple approach to inter-fractional VBF planning and demonstrated its potential in reducing dose to OARs. This work was partially supported through research funding from National Institutes of Health (R01CA169102) and Varian Medical Systems, Palo Alto, CA, USA.

60 APPLIED LIFE SCIENCES↗

Development and clinical evaluation of automatic fiducial detection for tumor tracking in cine megavoltage images during volumetric modulated arc therapy

Purpose: Real-time tracking of implanted fiducials in cine megavoltage (MV) imaging during volumetric modulated arc therapy (VMAT) delivery is complicated due to the inherent low contrast of MV images and potential blockage of dynamic leaves configurations. The purpose of this work is to develop a clinically practical autodetection algorithm for motion management during VMAT. Methods: The expected field-specific segments and the planned fiducial position from the Eclipse (Varian Medical Systems, Palo Alto, CA) treatment planning system were projected onto the MV images. The fiducials were enhanced by applying a Laplacian of Gaussian filter in the spatial domain for each image, with a blob-shaped object as the impulse response. The search of implanted fiducials was then performed on a region of interest centered on the projection of the fiducial when it was within an open field including the case when it was close to the field edge or partially occluded by the leaves. A universal template formula was proposed for template matching and normalized cross correlation was employed for its simplicity and computational efficiency. The search region for every image was adaptively updated through a prediction model that employed the 3D position of the fiducial estimated from the localized positions in previous images. This prediction model allowed the actual fiducial position to be tracked dynamically and was used to initialize the search region. The artifacts caused by electronic interference during the acquisition were effectively removed. A score map was computed by combining both morphological information and image intensity. The pixel location with the highest score was selected as the detected fiducial position. The sets of cine MV images taken during treatment were analyzed with in-house developed software written in MATLAB (The Mathworks, Inc., Natick, MA). Five prostate patients were analyzed to assess the algorithm performance by measuring their positioning accuracy during treatment. Results: The algorithm was able to accurately localize the fiducial position on MV images with success rates of more than 90% per case. The percentage of images in which each fiducial was localized in the studied cases varied between 23% and 65%, with at least one fiducial having been localized between 40% and 95% of the images. This depended mainly on the modulation of the plan and fiducial blockage. The prostate movement in the presented cases varied between 0.8 and 3.5 mm (mean values). The maximum displacement detected among all patients was of 5.7 mm. Conclusions: An algorithm for automatic detection of fiducial markers in cine MV images has been developed and tested with five clinical cases. Despite the challenges posed by complex beam aperture shapes, fiducial localization close to the field edge, partial occlusion of fiducials, fast leaf and gantry movement, and inherently low MV image quality, good localization results were achieved in patient images. This work provides a technique for enabling real-time accurate fiducial detection and tumor tracking during VMAT treatments without the use of extra imaging dose.

60 APPLIED LIFE SCIENCES↗

SU-E-I-87: Automated Liver Segmentation Method for CBCT Dataset by Combining Sparse Shape Composition and Probabilistic Atlas Construction

Purpose: The aiming of this study was to extract liver structures for daily Cone beam CT (CBCT) images automatically. Methods: Datasets were collected from 50 intravenous contrast planning CT images, which were regarded as training dataset for probabilistic atlas and shape prior model construction. Firstly, probabilistic atlas and shape prior model based on sparse shape composition (SSC) were constructed by iterative deformable registration. Secondly, the artifacts and noise were removed from the daily CBCT image by an edge-preserving filtering using total variation with L1 norm (TV-L1). Furthermore, the initial liver region was obtained by registering the incoming CBCT image with the atlas utilizing edge-preserving deformable registration with multi-scale strategy, and then the initial liver region was converted to surface meshing which was registered with the shape model where the major variation of specific patient was modeled by sparse vectors. At the last stage, the shape and intensity information were incorporated into joint probabilistic model, and finally the liver structure was extracted by maximum a posteriori segmentation.Regarding the construction process, firstly the manually segmented contours were converted into meshes, and then arbitrary patient data was chosen as reference image to register with the rest of training datasets by deformable registration algorithm for constructing probabilistic atlas and prior shape model. To improve the efficiency of proposed method, the initial probabilistic atlas was used as reference image to register with other patient data for iterative construction for removing bias caused by arbitrary selection. Results: The experiment validated the accuracy of the segmentation results quantitatively by comparing with the manually ones. The volumetric overlap percentage between the automatically generated liver contours and the ground truth were on an average 88%–95% for CBCT images. Conclusion: The experiment demonstrated that liver structures of CBCT with artifacts can be extracted accurately for following adaptive radiation therapy. This work is supported by National Natural Science Foundation of China (No. 61201441), Research Fund for Excellent Young and Middle-aged Scientists of Shandong Province (No. BS2012DX038), Project of Shandong Province Higher Educational Science and Technology Program (No. J12LN23), Jinan youth science and technology star (No.20120109)

60 APPLIED LIFE SCIENCES↗

Multi-parametric analysis for mixed integer linear programming: An application to transmission upgrade and congestion management

Upgrading the capacity of existing transmission lines is essential for meeting the growing energy demands, facilitating the integration of renewable energy, and ensuring the security of the transmission system. This study focuses on the selection of lines whose capacities and by how much should be expanded from the perspective of the Independent System Operators (ISOs) to minimize the total system cost. We employ advanced multi-parametric programming and an enhanced branch-and-bound algorithm to address complex mixed-integer linear programming (MILP) problems, considering multi-period time constraints and physical limitations of generators and transmission lines. To characterize the various decisions in transmission expansion, we model the increased capacity of existing lines as parameters within a specified range. This study first relaxes the binary variables to continuous variables and applies the Lagrange method and Karush-Kuhn-Tucker (KKT) conditions to obtain optimal solutions and identify critical regions associated with active and inactive constraints. Moreover, we extend the traditional branch-and-bound (B&B) method by determining the problem’s upper and lower bounds at each node of the B&B decision tree, helping to manage computational challenges in large-scale MILP problems. Here, we compare the difference between the upper and lower bounds to obtain an approximate optimal solution within the decision-makers’ tolerable error range. In addition, the first derivative of the objective function on the parameters of each line is used to inform the selection of lines for easing congestion and maximizing social welfare. Finally, the capacity upgrades are selected by weighing the reductions in system costs against the expense of upgrading line capacities. The findings are supported by numerical simulations and provide transmission-line planners with decision-making guidance.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Using hydrogen and ammonia for renewable energy storage: A geographically comprehensive techno-economic study

Hydrogen and, more recently, ammonia have received worldwide attention as energy storage media. In this work we investigate the economics of using each of these chemicals as well as the two in combination for islanded renewable energy supply systems in 15 American cities representing different climate regions throughout the country. We use an optimal combined capacity planning and scheduling model which minimizes the levelized cost of energy (LCOE) by determining optimal unit selection and size along with unit commitments, production rates, and storage inventories for each period of system operation. These periods are aggregated from full year hourly resolution data via a consecutive temporal clustering algorithm. Ammonia is generally more economical than hydrogen as a single method of energy storage. Additionally, systems which use both hydrogen and ammonia outperform those which use only one storage option and have LCOE between $\$ 0.17$/kWh and $\$ 0.28$/kWh, including full investment in renewable generation infrastructure.

25 ENERGY STORAGE↗

Dynamic Multileaf Collimator Tracking of Respiratory Target Motion Based on a Single Kilovoltage Imager During Arc Radiotherapy

Purpose: To demonstrate and characterize dynamic multileaf collimator (DMLC) tracking of respiratory moving targets that are spatially localized with a single kV X-ray imager during arc radiotherapy. Methods and Materials: During delivery of an arc field (358 deg. gantry rotation, 72-sec duration, circular field shape), the three-dimensional (3D) position of a fiducial marker in a phantom was estimated in real time from fluoroscopic kV X-ray images acquired orthogonally to the treatment beam axis. A prediction algorithm was applied to account for system latency (570 ms) before the estimated marker position was used for DMLC aperture adaptation. Experiments were performed with 12 patient-measured tumor trajectories that were selected from 160 trajectories (46 patients) and reproduced by a programmable phantom. Offline, the 3D deviation of the estimated phantom position from the actual position was quantified. The two-dimensional (2D) beam-target deviation was quantified as the positional difference between the MLC aperture center and the marker in portal images acquired continuously during experiments. Simulations of imaging and treatment delivery extended the study to all 160 tumor trajectories and to arc treatments of 3-min and 5-min duration. Results: In the experiments, the mean root-mean-square deviation was 1.8 mm for the 3D target position and 1.5 mm for the 2D aperture position. Simulations agreed with this to within 0.1 mm and resulted in mean 2D root-mean-square beam-target deviations of 1.1 mm for all 160 trajectories for all treatment durations. The deviations were mainly caused by system latency (570 ms). Conclusions: Single-imager DMLC tracking of respiratory target motion during arc radiotherapy was implemented, providing less than 2-mm geometric uncertainty for most trajectories.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Plasma image classification using cosine similarity constrained convolutional neural network

Plasma jets are widely investigated both in the laboratory and in nature. Astrophysical objects such as black holes, active galactic nuclei and young stellar objects commonly emit plasma jets in various forms. With the availability of data from plasma jet experiments resembling astrophysical plasma jets, classification of such data would potentially aid in not only investigating the underlying physics of the experiments but also the study of astrophysical jets. In this work we use deep learning to process all of the laboratory plasma images from the Caltech Spheromak Experiment spanning two decades. We found that cosine similarity can aid in feature selection, classify images through comparison of feature vector direction and be used as a loss function for the training of AlexNet for plasma image classification. We also develop a simple vector direction comparison algorithm for binary and multi-class classification. Using our algorithm we demonstrate 93 % accurate binary classification to distinguish unstable columns from stable columns and 92 % accurate five-way classification of a small, labelled data set which includes three classes corresponding to varying levels of kink instability.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗