SEARCH · Search NASA

Results for “correlation 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.

SourceResearch connectionFrom yearThrough year

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

At least 127 records · Page 7

All sources NASA NTRS DOE OSTI

NASA NTRS1995

The composite classification problem in optical information processing

Optical pattern recognition allows objects to be recognized from their images and permits their positional parameters to be estimated accurately in real time. The guiding principle behind optical pattern recognition is that a lens focusing a beam of coherent light modulated with an image produces the two-dimensinal Fourier transform of that image. When the resulting output is further transformed by the matched filter corresponding to the original image, one obtains the autocorrelation function of the original image, which has a peak at the origin. Such a device is called an optical correlator and may be used to recognize the locate the image for which it is designed. (From a practical perspective, an approximation to the matched filter must be used since the spatial light modulator (SLM) on which the filter is implemented usually does not allow one to independently control both the magnitude and phase of the filter.) Generally, one is not just concerned with recognizing a single image but is interested in recognizing a variety of rotated and scaled views of a particular image. In order to recognize these different views using an optical correlator, one may select a subset of these views (whose elements are called training images) and then use a composite filter that is designed to produce a correlation peak for each training image. Presumably, these peaks should be sharp and easily distinguishable from the surrounding correlation plane values. In this report we consider two areas of research regarding composite optical correlators. First, we consider the question of how best to choose the training images that are used to design the composite filter. With regard to quantity, the number of training images should be large enough to adequately represent all possible views of the targeted object yet small enough to ensure that the resolution of the filter is not exhausted. As for the images themselves, they should be distinct enough to avoid numerical difficulties yet similar enough to avoid gaps in which certain views of the target will be unrecognized. One method that we introduce to study this problem is called probing and involves the creation of the artificial imagery. The second problem we consider involves the clasification of the composite filter's correlation plane data. In particular, we would like to determine not only whether or not we are viewing a training image, but, in the former case, we would like to determine which training image is being viewed. This second problem is investigated using traditional M-ary hypothesis testing techniques.

Hall, Eric B.↗

DOE OSTI2021

A selection function toolbox for subsets of astronomical catalogues

ABSTRACT Large catalogues are ubiquitous throughout astronomy, but most scientific analyses are carried out on smaller samples selected from these catalogues by chosen cuts on catalogued quantities. The selection function of that scientific sample – the probability that a star in the catalogue will satisfy these cuts and so make it into the sample – is thus unique to each scientific analysis. We have created a general framework that can flexibly estimate the selection function of a sample drawn from a catalogue in terms of position, magnitude and colour. Our method is unique in using the binomial likelihood and accounting for correlations in the selection function across position, magnitude, and colour using Gaussian processes and spherical harmonics. We have created a new open-source python package selectionfunctiontoolbox that implements this framework and used it to make three different estimates of the APOGEE DR16 red giant sample selection function, as a subset of 2MASS, with each estimate using an increasing amount of technical knowledge of the APOGEE targeting. In a companion paper, we applied our methodology to derive estimates of the astrometric and spectroscopic selection functions of Gaia EDR3. Our framework will make it trivial for astrophysicists to estimate the selection function that they should be using with the custom sample of stars that they have chosen to answer their scientific question.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2023

Self-calibrating optical galaxy cluster selection bias using cluster, galaxy, and shear cross-correlations

ABSTRACT The clustering signals of galaxy clusters are powerful tools for self-calibrating the mass–observable relation and are complementary to cluster abundance and lensing. In this work, we explore the possibility of combining three correlation functions – cluster lensing, the cluster–galaxy cross-correlation function, and the galaxy autocorrelation function – to self-calibrate optical cluster selection bias, the boosted clustering and lensing signals in a richness-selected sample mainly caused by projection effects. We develop mock catalogues of redMaGiC-like galaxies and redMaPPer-like clusters by applying halo occupation distribution models to N-body simulations and using counts-in-cylinders around massive haloes as a richness proxy. In addition to the previously known small-scale boost in projected correlation functions, we find that the projection effects also significantly boost three-dimensional correlation functions to scales of 100 $h^{-1} \, \rm Mpc$. We perform a likelihood analysis assuming survey conditions similar to the Dark Energy Survey and show that the selection bias can be self-consistently constrained at the 10 per cent level. We discuss strategies for applying this approach to real data. We expect that expanding the analysis to smaller scales and using deeper lensing data would further improve the constraints on cluster selection bias.

Astronomy & Astrophysics↗

NASA NTRS1977

A contribution to ULF activity in the Pc 3-4 range correlated with IMF radial orientation

The paper describes an experiment to determine whether the radial orientation of the interplanetary magnetic field (IMF) is associated with ULF activity in the Pc 3-4 range. Data are obtained from base levels, undisturbed intervals, IMF and disturbance selection, and trigonometric correlation. The results obtained are discussed, noting particularly that for low Kp, the probability of enhanced amplitude noise rises as IMF orientation with respect to the nominal solar wind flow decreases in both Pc 3 and Pc 4 channels.

Greenstadt, E. W.↗

DOE OSTI2016

Oxidation of Cyclohexene Catalyzed by Nanoporous Au(Ag) in Liquid Phase

Nanoporous gold with minor silver content has been identified as a new type of gold based catalyst for selective oxidation of cyclohexene with molecular oxygen in liquid. By oxidation of the leached nanoporous gold foils in ozone, the minor silver content was oxidized in this paper to form silver oxide nanoclusters on the surface of nanoporous gold. With further treatment in methanol, the surface silver oxide was reduced and surface alloy was formed on gold ligaments. Both nanoporous gold treated with ozone only and the one with ozone and then methanol are very active for selective oxidation of cyclohexene with molecular oxygen in liquid of cyclohexene with a turn-over-frequency (TOF) of 0.55–0.99 molecules per surface Au atom per second under a solvent-free and initiator- free condition. The total selectivity for production of 2-cyclohexene-1-one, 2-cyclohexene-1-ol, and cyclohexene oxide was increased from 57.5 % to 80.8 % by an additional treatment of nanoporous gold in methanol after activation in zone. Finally, the correlation of catalytic selectivity for the production of the three products and corresponding surface chemistry of ligament suggests that (1) the formed Au–Ag alloy surface is favorable for the formation of 2-cyclohexen-1-one, 2-cyclohexene-1-ol, and cyclohexene oxide and (2) the surface silver oxide is favorable for the production of cyclohexenyl hydroperoxide.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2021

Phenotypic effects of changes in the FTVTxK region of an Arabidopsis secondary wall cellulose synthase compared with results from analogous mutations in other isoforms

Abstract Understanding protein structure and function relationships in cellulose synthase (CesA), including divergent isomers, is an important goal. Here, we report results from mutant complementation assays that tested the ability of sequence variants of AtCesA7, a secondary wall CesA of Arabidopsis thaliana , to rescue the collapsed vessels, short stems, and low cellulose content of the irx3‐1 AtCesA7 null mutant. We tested a catalytic null mutation and seven missense or small domain changes in and near the AtCesA7 FTVTSK motif, which lies near the catalytic domain and may, analogously to bacterial CesA, exist within a substrate “gating loop.” A low‐to‐high gradient of rescue occurred, and even inactive AtCesA7 had a small positive effect on stem cellulose content but not stem elongation. Overall, secondary wall cellulose content and stem length were moderately correlated, but the results were consistent with threshold amounts of cellulose supporting particular developmental processes. Vibrational sum frequency generation microscopy allowed tissue‐specific analysis of cellulose content in stem xylem and interfascicular fibers, revealing subtle differences between selected genotypes that correlated with the extent of rescue of the collapsing xylem phenotype. Similar tests on PpCesA5 from the moss Physcomitrium (formerly Physcomitrella ) patens helped us to synergize the AtCesA7 results with prior results on AtCesA1 and PpCesA5. The cumulative results show that the FTVTxK region is important for the function of an angiosperm secondary wall CesA as well as widely divergent primary wall CesAs, while differences in complementation results between isomers may reflect functional differences that can be explored in further work.

Burris, Jason N.↗

DOE OSTI2022

Global estimates of the fitness advantage of SARS-CoV-2 variant Omicron

Abstract New variants of SARS-CoV-2 show remarkable heterogeneity in their relative fitness over both time and space. In this paper we extend the tools available for estimating the selection strength for new SARS-CoV-2 variants to a hierarchical, mixed-effects, renewal equation model. This formulation allows us to estimate selection effects at the global level while incorporating both measured and unmeasured heterogeneity among countries. Applying this model to the spread of Omicron in forty countries, we find evidence for very strong but very heterogeneous selection effects. To test whether this heterogeneity is explained by differences in the immune landscape, we considered several measures of vaccination rates and recent population-level infection as covariates, finding moderately strong, statistically significant effects. We also found a significant positive correlation between the selection advantage of Delta and Omicron at the country level, suggesting that other region-specific explanatory variables of fitness differences do exist. Our method is implemented in the Stan programming language, can be run on standard consumer-grade computing resources, and will be straightforward to apply to future variants.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2025

Taming the virtual space for incremental full configuration interaction

Incremental full configuration interaction (iFCI) closely approximates the FCI limit with polynomial cost through a many-body expansion of the correlation energy, providing highly accurate total energies within a given basis set. To extend iFCI beyond previous basis set limitations, this work introduces a novel natural orbital (NO) screening approach, incremental NO full configuration interaction (iNO-FCI). By consideration of the importance of virtual orbital selection in the convergence of iFCI, iNO-FCI maximizes the consistency between orbitals selected for each correlated body. iNO-FCI employs a principle of cancellation of errors and ensures that the same set of virtual NOs is used for interdependent terms. Here, this strategy significantly reduces computational cost without compromising precision. Computational savings of up to 95% are demonstrated, allowing access to larger basis sets that were previously computationally prohibitive. iNO-FCI is herein introduced and benchmarked for several difficult test cases involving double-bond dissociation, biradical systems, conjugated π systems, and the spin gap of a Cu-based transition metal complex.

Correlation energy↗

DOE OSTI2022

Defective Sn-Zn perovskites through bio-directed routes for modulating CO 2 RR

Sn and Zn oxides can catalyze electrochemical CO 2 reduction reactions (CO 2 RR) to produce useful chemicals. Herein, we report on perovskite-type tin-zinc oxides (TZO) synthesized by a biomineralization approach under benign aqueous conditions. The resulting TZO catalysts are disordered and possess a high quantity of defects, which serve as highly reactive active sites in CO 2 RR. Selectively toward formate or CO is tunable toward the relative amounts of Sn and Zn and their associated O defects. In-situ X-ray absorption spectroscopy measurements reveal electronic and structural transformations consistent with oxygen vacancy formation at Sn-O-Zn bridges as a function of cathodic potential that strongly correlates to CO2RR selectivity. Further, Density functional theory (DFT) calculations demonstrate that introducing an oxygen vacancy in TZO influences drastically influences formate/CO selectivity, consistent with our in-situ XAS findings and catalytic performance. Overall, these findings demonstrate the potential of biomineralization to fabricate perovskite-type metal oxides for energy conversion reactions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2022

Quantifying the Influence of Defects on Selectivity of Electrodes Encapsulated by Nanoscopic Silicon Oxide Overlayers

Encapsulation of electrocatalysts and photocatalysts with semipermeable nanoscopic oxide overlayers that exhibit selective transport properties is an attractive approach to achieve high redox selectivity. However, defects within the overlayers-such as pinholes, cracks, or particle inclusions-may facilitate local high rates of parasitic reactions by creating pathways for facile transport of undesired reactants to exposed active sites. Scanning electrochemical microscopy (SECM) is an attractive method to determine the influence of defects on macroscopic performance metrics thanks to its ability to measure the relative rates of competing electrochemical reactions with high spatial resolution over the electrode. Here, we report the use of SECM to determine the influence of overlayer defects on the selectivity of silicon oxide (SiOx) encapsulated platinum thin-film electrocatalysts operated under conditions where two competing reactions-the hydrogen evolution and Fe(III) reduction reactions-can occur. After an SECM methodology is described to determine spatially resolved selectivity, representative selectivity maps are correlated with the location of defects that are characterized by optical, electron, and atomic force microscopies. This analysis reveals that certain types of defects in the oxide overlayer are responsible for ∼60-90% of the partial current density toward the undesired Fe(III) reduction reaction. By correcting for defect contributions to Fe(III) reduction rates, true Fe(III) permeability values for the SiOxoverlayers were determined to be over an order of magnitude lower than permeabilities determined from analyses that ignore the presence of defects. Lastly, different types of defects were studied revealing that defect morphology can have varying influence on both redox selectivity and calculated permeability. This work highlights the need for spatially resolved measurements to evaluate the performance of oxide-encapsulated catalysts and understand their performance limits.

25 ENERGY STORAGE↗

NASA NTRS1989

Elevated temperature crack growth

Alloy 718 crack growth experiments were conducted to assess the ability of the selected path-independent (P-I) integrals to describe the elevated temperature crack growth behavior. These tests were performed on single edge notch (SEN) specimens under displacement control with multiple extensometers to monitor the specimen and crack mouth opening displacement (CMOD). The displacements in these tests were sufficiently high to induce bulk cyclic inelastic deformation of the specimen. Under these conditions, the linear elastic fracture mechanics (LEFM) parameter K does not correlate the crack growth data. The experimentally measured displacement gradients at the end of specimen gage length were used as the boundary conditions in elastic-plastic finite element method (FEM) analyses. These analyses were performed with a node release approach using CYANIDE, a GEAE FEM code, which included a gap element which is capable of efficiently simulating crack closure. Excellent correlation was obtained between the experimentally measured and predicted variation of stress and CMOD with crack length and the stress-CMOD loops for Alloy 718 tests conducted at 538 C. This confirmed the accuracy of the FEM crack growth simulation approach. The experimentally measured crack growth rate data correlated well the selected P-I integrals. These investigations have produced significant progress in developing P-I integrals as non-linear fracture mechanics parameters. The results suggest that this methodology has the potential of accurately describing elevated temperature crack growth behavior under the combined influence of thermal cycling and bulk elastic-inelastic deformation states.

Kim, K. S.↗

DOE OSTI2022

Effects of the method of active site characterization for determining structure-sensitivity in Ag-catalyzed ethylene epoxidation

We report catalysts have been prepared on a low surface area α-Al 2 O 3 support used commercially for previous generation olefin epoxidation catalysts. Prescreening of the low surface area alumina (0.73 m 2 /g) indicated the absence of acid catalyzed isomerization of EO at an evaluation temperature of 210 °C. A 0.1 wt% Ag base material synthesized by incipient wetness impregnation of AgNO 3 was used as a base material for electroless deposition (ED) of additional Ag to increase particle sizes by controlled reduction of Ag + directly onto the preexisting Ag surface to form weight loadings between 0.3 and 5.0 wt% metal. A 12 wt% Ag/α-Al 2 O 3 using Ag 2 C 2 O 4 as the Ag precursor was also prepared to compare performance of the ED samples with a catalyst more typical of industrial formulations. Characterization by SEM, STEM, and hydrogen titration of oxygen precovered Ag characterized before and after catalytic evaluation indicated that microscopy is required to accurately represent distributions of Ag particle sizes, but H 2 titration of O-precovered Ag gives the best representation of active sites since it directly counts the number of Ag surface sites. Larger particles >100 nm are resistant to both Ag sintering and carbon foulant; TOF values were relatively insensitive to particle size with only a 2.2 × difference between the best and worst performing samples. Selectivity, which is not a function of TOF, shows the most significant structure sensitivity effect where particle sizes follow the trend 67 nm ≈ 92 nm (58% EO) < 157 nm (67% EO) < 211–542 nm (73% EO). The lower EO selectivities were also correlated with increased fouling for the smaller Ag sizes, suggesting that more strongly bound EO precursor(s) leads to combustion and CO 2 /H 2 O formation.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2016

Carbon dioxide in an ionic liquid: Structural and rotational dynamics

Ionic liquids (ILs), which have widely tunable structural motifs and intermolecular interactions with solutes, have been proposed as possible carbon capture media. In order to inform the choice of an optimal ionic liquid system, it can be useful to understand the details of dynamics and interactions on fundamental time scales (femtoseconds to picoseconds) of dissolved gases, particularly carbon dioxide (CO 2 ), within the complex solvation structures present in these uniquely organized materials. The rotational and local structural fluctuation dynamics of CO 2 in the room temperature ionic liquid 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide (EmimNTf2) were investigated by using ultrafast infrared spectroscopy to interrogate the CO 2 asymmetric stretch. Polarization-selective pump probe measurements yielded the orientational correlation function of the CO 2 vibrational transition dipole. It was found that reorientation of the carbon dioxide occurs on 3 time scales: 0.91 ± 0.03, 8.3 ± 0.1, 54 ± 1 ps. The initial two are attributed to restricted wobbling motions originating from a gating of CO 2 motions by the IL cations and anions. The final (slowest) decay corresponds to complete orientational randomization. Two-dimensional infrared vibrational echo (2D IR) spectroscopy provided information on structural rearrangements, which cause spectral diffusion, through the time dependence of the 2D line shape. Analysis of the time-dependent 2D IR spectra yields the frequency-frequency correlation function (FFCF). Polarization-selective 2D IR experiments conducted on the CO 2 asymmetric stretch in the parallel- and perpendicular-pumped geometries yield significantly different FFCFs due to a phenomenon known as reorientation-induced spectral diffusion (RISD), revealing strong vector interactions with the liquid structures that evolve slowly on the (independently measured) rotation time scales. To separate the RISD contribution to the FFCF from the structural spectral diffusion contribution, the previously developed first order Stark effect RISD model is reformulated to describe the second order (quadratic) Stark effect—the first order Stark effect vanishes because CO 2 does not have a permanent dipole moment. Through this analysis, we characterize the structural fluctuations of CO 2 in the ionic liquid solvation environment, which separate into magnitude-only and combined magnitude and directional correlations of the liquid’s time dependent electric field. This new methodology will enable highly incisive comparisons between CO 2 dynamics in a variety of ionic liquid systems.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2023

Non-Oxidative Ethane Dehydrogenation in a Packed-Bed DBD Plasma Reactor

Abstract Plasma-assisted conversion of ethane (C 2 H 6 ) can produce value-added chemical building blocks using green electricity. Here we employ a simple packed-bed coaxial dielectric barrier discharge (DBD) reactor to convert C 2 H 6 at mild operating conditions unattainable by conventional thermocatalysis. Ethylene (C 2 H 4 ), acetylene (C 2 H 2 ), and methane (CH 4 ) are the main products along with small fractions of C 3 and C 4 hydrocarbons. Interestingly, the C 2 H 4 selectivity is primarily correlated to C 2 H 6 conversion, dominated by electron dissociation and recombination reactions irrespective of the dielectric properties of the packed bed material (SiO 2 , Al 2 O 3 , ZrO 2 , TiO 2 , and BaTiO 3 ), packing material size, supplied power, and C 2 H 6 concentration. While a distortion of the electric field and discharge propagation results in varying dissipated power as materials change, the C 2 H 4 energy yield remains constant. The particle size appears to affect conversion mainly due to pressure alterations. Pd/SiO 2 catalyst can change the selectivity, favoring saturated species by expending hydrogen.

Engineering↗

DOE OSTI2025

How does ion temperature gradient turbulence depend on magnetic geometry? Insights from data and machine learning

Magnetic geometry has a significant effect on the level of turbulent transport in fusion plasmas. Here, we model and analyse this dependence using multiple machine learning methods and a dataset of >200 000 nonlinear gyrokinetic simulations of ion-temperature-gradient turbulence in diverse non-axisymmetric geometries. The dataset is generated using a large collection of both optimised and randomly generated stellarator equilibria. At fixed gradients and other input parameters, the turbulent heat flux varies between geometries by several orders of magnitude. Trends are apparent among the configurations with particularly high or particularly low heat flux. Regression and classification techniques from machine learning are then applied to extract patterns in the dataset. Due to a symmetry of the gyrokinetic equation, the heat flux and regressions thereof should be invariant to translations of the raw features in the parallel coordinate, similar to translation invariance in computer vision applications. Multiple regression models including convolutional neural networks (CNNs) and decision trees can achieve reasonable predictive power for the heat flux in held-out test configurations, with highest accuracy for the CNNs. Using Spearman correlation, sequential feature selection and Shapley values to measure feature importance, it is consistently found that the most important geometric lever on the heat flux is the flux surface compression in regions of bad curvature. The second most important geometric feature relates to the magnitude of geodesic curvature. These two features align remarkably with surrogates that have been proposed based on theory, while the methods here allow a natural extension to more features for increased accuracy. The dataset, released with this publication, may also be used to test other proposed surrogates, and we find that many previously published proxies do correlate well with both the heat flux and stability boundary.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DOE OSTI2021

Two-Phase Turbulence Statistics from High Fidelity Dispersed Droplet Flow Simulations in a Pressurized Water Reactor (PWR) Sub-Channel with Mixing Vanes

In the dispersed flow film boiling regime (DFFB), which exists under post-LOCA (loss-of-coolant accident) conditions in pressurized water reactors (PWRs), there is a complex interplay between droplet dynamics and turbulence in the surrounding steam. Experiments have accredited particular significance to droplet collision with the spacer-grids and mixing vane structures and their consequent positive feedback to the heat transfer recorded in the immediate downstream vicinity. Enabled by high-performance computing (HPC) systems and a massively parallel finite element-based flow solver—PHASTA (Parallel Hierarchic Adaptive Stabilized Transient Analysis)—this work presents high fidelity interface capturing, two-phase, adiabatic simulations in a PWR sub-channel with spacer grids and mixing vanes. Selected flow conditions for the simulations are informed by the experimental data found in the literature, including the steam Reynolds number and collision Weber number (Wec={40,80}), and are characteristic of the DFFB regime. Data were collected from the simulations at an unprecedented resolution, which provides detailed insights into the continuous phase turbulence statistics, highlighting the effects of the presence of droplets and the comparative effect of different Weber numbers on turbulence in the surrounding steam. Further, axial evolution of droplet dynamics was analyzed through cross-sectionally averaged quantities, including droplet volume, surface area and Sauter mean diameter (SMD). The downstream SMD values agree well with the existing empirical correlations for the selected range of Wec. The high-resolution data repository from the simulations herein is expected to be of significance to guide model development for system-level thermal hydraulic codes.

Saini, Nadish↗

NASA NTRS1992

Correlation, Cost Risk, and Geometry

The geometric viewpoint identifies the choice of a correlation matrix for the simulation of cost risk with the pairwise choice of data vectors corresponding to the parameters used to obtain cost risk. The correlation coefficient is the cosine of the angle between the data vectors after translation to an origin at the mean and normalization for magnitude. Thus correlation is equivalent to expressing the data in terms of a non orthogonal basis. To understand the many resulting phenomena requires the use of the tensor concept of raising the index to transform the measured and observed covariant components into contravariant components before vector addition can be applied. The geometric viewpoint also demonstrates that correlation and covariance are geometric properties, as opposed to purely statistical properties, of the variates. Thus, variates from different distributions may be correlated, as desired, after selection from independent distributions. By determining the principal components of the correlation matrix, variates with the desired mean, magnitude, and correlation can be generated through linear transforms which include the eigenvalues and the eigenvectors of the correlation matrix. The conversion of the data to a non orthogonal basis uses a compound linear transformation which distorts or stretches the data space. Hence, the correlated data does not have the same properties as the uncorrelated data used to generate it. This phenomena is responsible for seemingly strange observations such as the fact that the marginal distributions of the correlated data can be quite different from the distributions used to generate the data. The joint effect of statistical distributions and correlation remains a fertile area for further research. In terms of application to cost estimating, the geometric approach demonstrates that the estimator must have data and must understand that data in order to properly choose the correlation matrix appropriate for a given estimate. There is a general feeling by employers and managers that the field of cost requires little technical or mathematical background. Contrary to that opinion, this paper demonstrates that a background in mathematics equivalent to that needed for typical engineering and scientific disciplines at the masters or doctorate level is appropriate within the field of cost risk.

Dean, Edwin B.↗

DOE OSTI2015

A two-stage method for microcalcification cluster segmentation in mammography by deformable models

Purpose: Segmentation of microcalcification (MC) clusters in x-ray mammography is a difficult task for radiologists. Accurate segmentation is prerequisite for quantitative image analysis of MC clusters and subsequent feature extraction and classification in computer-aided diagnosis schemes. Methods: In this study, a two-stage semiautomated segmentation method of MC clusters is investigated. The first stage is targeted to accurate and time efficient segmentation of the majority of the particles of a MC cluster, by means of a level set method. The second stage is targeted to shape refinement of selected individual MCs, by means of an active contour model. Both methods are applied in the framework of a rich scale-space representation, provided by the wavelet transform at integer scales. Segmentation reliability of the proposed method in terms of inter and intraobserver agreements was evaluated in a case sample of 80 MC clusters originating from the digital database for screening mammography, corresponding to 4 morphology types (punctate: 22, fine linear branching: 16, pleomorphic: 18, and amorphous: 24) of MC clusters, assessing radiologists’ segmentations quantitatively by two distance metrics (Hausdorff distance—HDIST{sub cluster}, average of minimum distance—AMINDIST{sub cluster}) and the area overlap measure (AOM{sub cluster}). The effect of the proposed segmentation method on MC cluster characterization accuracy was evaluated in a case sample of 162 pleomorphic MC clusters (72 malignant and 90 benign). Ten MC cluster features, targeted to capture morphologic properties of individual MCs in a cluster (area, major length, perimeter, compactness, and spread), were extracted and a correlation-based feature selection method yielded a feature subset to feed in a support vector machine classifier. Classification performance of the MC cluster features was estimated by means of the area under receiver operating characteristic curve (Az ± Standard Error) utilizing tenfold cross-validation methodology. A previously developed B-spline active rays segmentation method was also considered for comparison purposes. Results: Interobserver and intraobserver segmentation agreements (median and [25%, 75%] quartile range) were substantial with respect to the distance metrics HDIST{sub cluster} (2.3 [1.8, 2.9] and 2.5 [2.1, 3.2] pixels) and AMINDIST{sub cluster} (0.8 [0.6, 1.0] and 1.0 [0.8, 1.2] pixels), while moderate with respect to AOM{sub cluster} (0.64 [0.55, 0.71] and 0.59 [0.52, 0.66]). The proposed segmentation method outperformed (0.80 ± 0.04) statistically significantly (Mann-Whitney U-test, p < 0.05) the B-spline active rays segmentation method (0.69 ± 0.04), suggesting the significance of the proposed semiautomated method. Conclusions: Results indicate a reliable semiautomated segmentation method for MC clusters offered by deformable models, which could be utilized in MC cluster quantitative image analysis.

60 APPLIED LIFE SCIENCES↗