Search NASA⌕ Search

SEARCH · Search NASA

Results for “Gaussian Process Models”

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 1,135 records · Page 63

Learning effective stochastic differential equations from microscopic simulations: Linking stochastic numerics to deep learning

We identify effective stochastic differential equations (SDEs) for coarse observables of fine-grained particle- or agent-based simulations; these SDEs then provide useful coarse surrogate models of the fine scale dynamics. We approximate the drift and diffusivity functions in these effective SDEs through neural networks, which can be thought of as effective stochastic ResNets. The loss function is inspired by, and embodies, the structure of established stochastic numerical integrators (here, Euler–Maruyama and Milstein); our approximations can thus benefit from backward error analysis of these underlying numerical schemes. They also lend themselves naturally to “physics-informed” gray-box identification when approximate coarse models, such as mean field equations, are available. Existing numerical integration schemes for Langevin-type equations and for stochastic partial differential equations can also be used for training; we demonstrate this on a stochastically forced oscillator and the stochastic wave equation. Our approach does not require long trajectories, works on scattered snapshot data, and is designed to naturally handle different time steps per snapshot. We consider both the case where the coarse collective observables are known in advance, as well as the case where they must be found in a data-driven manner.

97 MATHEMATICS AND COMPUTING↗

Electron dynamics in complex environments with real-time time dependent density functional theory in a QM-MM framework

This article presents a time dependent density functional theory (TDDFT) implementation to propagate the Kohn-Sham equations in real time, including the effects of a molecular environment through a Quantum-Mechanics Molecular-Mechanics (QM-MM) hamiltonian. The code delivers an all-electron description employing Gaussian basis functions, and incorporates the Amber force-field in the QM-MM treatment. The most expensive parts of the computation, comprising the commutators between the hamiltonian and the density matrix—required to propagate the electron dynamics—, and the evaluation of the exchange-correlation energy, were migrated to the CUDA platform to run on graphics processing units, which remarkably accelerates the performance of the code. The method was validated by reproducing linear-response TDDFT results for the absorption spectra of several molecular species. Two different schemes were tested to propagate the quantum dynamics: (i) a leap-frog Verlet algorithm, and (ii) the Magnus expansion to first-order. These two approaches were confronted, to find that the Magnus scheme is more efficient by a factor of six in small molecules. Interestingly, the presence of iron was found to seriously limitate the length of the integration time step, due to the high frequencies associated with the core-electrons. This highlights the importance of pseudopotentials to alleviate the cost of the propagation of the inner states when heavy nuclei are present. Finally, the methodology was applied to investigate the shifts induced by the chemical environment on the most intense UV absorption bands of two model systems of general relevance: the formamide molecule in water solution, and the carboxy-heme group in Flavohemoglobin. In both cases, shifts of several nanometers are observed, consistently with the available experimental data.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Mechanism of the G-protein mimetic nanobody binding to a muscarinic G-protein-coupled receptor

Protein–protein binding is key in cellular signaling processes. Molecular dynamics (MD) simulations of protein–protein binding, however, are challenging due to limited timescales. In particular, binding of the medically important G-protein-coupled receptors (GPCRs) with intracellular signaling proteins has not been simulated with MD to date. Here, we report a successful simulation of the binding of a G-protein mimetic nanobody to the M 2 muscarinic GPCR using the robust Gaussian accelerated MD (GaMD) method. Through long-timescale GaMD simulations over 4,500 ns, the nanobody was observed to bind the receptor intracellular G-protein-coupling site, with a minimum rmsd of 2.48 Å in the nanobody core domain compared with the X-ray structure. Binding of the nanobody allosterically closed the orthosteric ligand-binding pocket, being consistent with the recent experimental finding. In the absence of nanobody binding, the receptor orthosteric pocket sampled open and fully open conformations. The GaMD simulations revealed two low-energy intermediate states during nanobody binding to the M 2 receptor. The flexible receptor intracellular loops contribute remarkable electrostatic, polar, and hydrophobic residue interactions in recognition and binding of the nanobody. As a result, these simulations provided important insights into the mechanism of GPCR–nanobody binding and demonstrated the applicability of GaMD in modeling dynamic protein–protein interactions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

PROBABILISTIC MODEL OF BEAM–PLASMA INTERACTION IN RANDOMLY INHOMOGENEOUS PLASMA

We propose a new model that describes beam–plasma interaction in the presence of random density fluctuations with a known probability distribution. We use the property that, for the given frequency, the probability distribution of the density fluctuations uniquely determines the probability distribution of the phase velocity of waves. We present the system as discrete and consisting of small, equal spatial intervals with a linear density profile. This approach allows one to estimate variations in wave energy density and particle velocity, depending on the density gradient on any small spatial interval. Because the characteristic time for the evolution of the electron distribution function and the wave energy is much longer than the time required for a single wave–particle resonant interaction over a small interval, we determine the description for the relaxation process in terms of averaged quantities. We derive a system of equations, similar to the quasi-linear approximation, with the conventional velocity diffusion coefficient D and the wave growth rate γ replaced by the average in phase space, by making use of the probability distribution for phase velocities and by assuming that the interaction in each interval is independent of previous interactions. Functions D and γ are completely determined by the distribution function for the amplitudes of the fluctuations. For the Gaussian distribution of the density fluctuations, we show that the relaxation process is determined by the ratio of beam velocity to plasma thermal velocity, the dispersion of the fluctuations, and the width of the beam in the velocity space.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Λ Baryon Production in ν¯µ Interactions in the MicroBooNE Detector

The Cabibbo suppressed production of $\Lambda$ baryons in anti-neutrino interactions with nuclei is a rare process that is yet to be measured with a modern neutrino detector with automated reconstruction. The cross section for this process is sensitive to a number of unique nuclear effects, most notably the secondary interactions of the produced hyperon while attempting to escape from the nucleus. Other interactions within the nuclear remnant can impact the estimation of neutrino energy in oscillation measurements, and thus an accurate description of the nuclear environment is required. The strangeness violating hyperon production process is only available to anti-neutrinos. The model of this interaction is implemented into the NuWro neutrino interaction Monte Carlo simulation, and some predictions are presented, focusing on the role of nuclear effects. This model introduces a hyperon-nucleus potential, which calculations from hypernuclear theory permit to be strongly repulsive in th e case of $\Sigma$ baryons. The presence of this potential is found to sculpt the shape of the differential cross section in some variables. The MicroBooNE detector will be described, followed by a description of a measurement of the flux averaged, restricted phase space cross section of Cabibbo suppressed $\Lambda$ baryon production. A sophisticated event selection is employed, as a very large quantity of background neutrino interactions must be removed to perform the measurement with any sensitivity. This selection introduces some novel techniques such as the island finding method, and achieves a background reduction of $\sim 10^6$, with an efficiency of around 7\%. The calculation of the systematic uncertainties will be explained, including two procedures explored to handle sources of background with extremely poor simulation statistics: an in-situ constraint using data from sidebands, and a visual inspection of the data and simulation to remove the troublesome background events. The sensitivity to the $\Lambda$ baryon production cross section is calculated in the form of Bayesian posterior probability distributions, combining the systematic uncertainties with data and simulation statistical uncertainties. As a rare process, the statistical uncertainties are highly non-Gaussian, and the Bayesian approach is applied to include the full shapes of these uncertainties. Data corresponding to $2.2 \times 10^{20}$ protons on target of neutrino mode running and $4.9 \times 10^{20}$ protons on target of anti-neutrino running is analysed. When the data was unblinded, five $\Lambda$ production candidates were selected from the data, consistent with the MC simulation prediction of $5.3 \pm 1.1$ events. The final estimated cross section is $1.8^{+2.0}_{-1.6} \times 10^{-40}$cm$^2/$Ar when employing the sideband constraint procedure. A similar result of $2.0^{+2.2}_{-1.8} \times 10^{-40}$cm$^2/$Ar is obtained when performing the visual scan instead. The methods used in t his analysis are intended to be easily exported to other LArTPC detectors such as the Short Baseline Near Detector.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

A New Criterion to Improve Operational Drizzle Detection with Ground-Based Remote Sensing

Light shallow precipitation in the form of drizzle is one of the mechanisms for liquid water removal, affecting cloud lifetime and boundary layer dynamics and thermodynamics. The early formation of drizzle drops is of particular interest for quantifying aerosol–cloud–precipitation interactions. In models, drizzle initiation is represented by the autoconversion, that is, the conversion of liquid water from a cloud liquid water category (where particle sedimentation is ignored) into a precipitating liquid water category. Various autoconversion parameterizations have been proposed in recent years, but their evaluation is challenging due to the lack of proper observations of drizzle development in the cloud. This work presents a new algorithm for Classification of Drizzle Stages (CLADS). CLADS is based on the skewness of the Ka-band radar Doppler spectrum. Skewness is sensitive to the drizzle growth in the cloud: the observed Gaussian Doppler spectrum has skewness zero when only cloud droplets are present without any significant fall velocity. Defining downward velocities positive, skewness turns positive when embryonic drizzle forms and becomes negative when drizzle starts to dominate the spectrum. CLADS identifies spatially coherent structures of positive, zero, and negative skewness in space and time corresponding to drizzle seeding, drizzle growth/nondrizzle, and drizzle mature, respectively. We test CLADS on case studies from the Jülich Observatory for Cloud Evolution Core Facility (JOYCE-CF) and the Barbados Cloud Observatory (BCO) to quantitatively estimate the benefits of CLADS compared to the standard Cloudnet target categorization algorithm. We suggest that CLADS can provide additional observational constraints for understanding the processes related to drizzle formation better.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Impact of Uncertainty from Load-Based Reserves and Renewables on Dispatch Costs and Emissions

Aggregations of controllable loads are considered to be a fast-responding, cost-efficient, and environmental-friendly candidate for power system ancillary services. Unlike conventional service providers, the potential capacity from the aggregation is highly affected by factors like ambient conditions and load usage patterns. Previous work modeled aggregations of controllable loads (such as air conditioners) as thermal batteries, which are capable of providing reserves but with uncertain capacity. A stochastic optimal power flow problem was formulated to manage this uncertainty, as well as uncertainty in renewable generation. In this paper, we explore how the types and levels of uncertainty, generation reserve costs, and controllable load capacity affect the dispatch solution, operational costs, and CO2 emissions. We also compare the results of two methods for solving the stochastic optimization problem, namely the probabilistically robust method and analytical reformulation assuming Gaussian distributions. Case studies are conducted on a modified IEEE 9-bus system with renewables, controllable loads, and congestion. We find that different types and levels of uncertainty have significant impacts on dispatch and emissions. More controllable loads and less conservative solution methodologies lead to lower costs and emissions.

24 POWER TRANSMISSION AND DISTRIBUTION↗

RBFNN-based Minimum Entropy Filtering for a Class of Stochastic Nonlinear Systems

This paper presents a novel minimum entropy filter design for a class of stochastic nonlinear systems which are subjected to non-Gaussian noises. Motivated by stochastic distribution control, an output entropy model is developed using RBF neural network while the parameters of the model can be identified by the collected data. Based upon the presented model, the filtering problem has been investigated while the system dynamics have been represented. As the model output is the entropy of the estimation error, the optimal nonlinear filter is obtained based on the Lyapunov design which makes the model output minimum. Moreover, the entropy assignment problem has been discussed as an extension of the presented approach. To verify the presented design procedure, a numerical example is given which illustrates the effectiveness of the presented algorithm. The contributions of this paper can be included as 1) an output entropy model is presented using neural network; 2) a nonlinear filter design algorithm is developed as the main result and 3) a solution of entropy assignment problem is obtained which is an extension of the presented framework.

42 ENGINEERING↗

OBSERVATIONS OF SOLAR ENERGETIC PARTICLES FROM {sup 3}He-RICH EVENTS OVER A WIDE RANGE OF HELIOGRAPHIC LONGITUDE

A prevailing model for the origin of {sup 3}He-rich solar energetic particle (SEP) events attributes particle acceleration to processes associated with the reconnection between closed magnetic field lines in an active region and neighboring open field lines. The open field from the small reconnection volume then provides a path along which accelerated particles escape into a relatively narrow range of angles in the heliosphere. The narrow width (standard deviation <20 Degree-Sign ) of the distribution of X-ray flare longitudes found to be associated with {sup 3}He-rich SEP events detected at a single spacecraft at 1 AU supports this model. We report multispacecraft observations of individual {sup 3}He-rich SEP events that occurred during the solar minimum time period from 2007 January through 2011 January using instrumentation carried by the two Solar Terrestrial Relations Observatory spacecraft and the Advanced Composition Explorer. We find that detections of {sup 3}He-rich events at pairs of spacecraft are not uncommon, even when their longitudinal separation is >60 Degree-Sign . We present the observations of the {sup 3}He-rich event of 2010 February 7, which was detected at all three spacecraft when they spanned 136 Degree-Sign in heliographic longitude. Measured fluences of {sup 3}He in this event were found to have a strong dependence on longitude which is well fit by a Gaussian with standard deviation {approx}48 Degree-Sign centered at the longitude that is connected to the source region by a nominal Parker spiral magnetic field. We discuss several mechanisms for distributing flare-accelerated particles over a wide range of heliographic longitudes including interplanetary diffusion perpendicular to the magnetic field, spreading of a compact cluster of open field lines between the active region and the source surface where the field becomes radial and opens out into the heliosphere, and distortion of the interplanetary field by a preceding coronal mass ejection. Statistical studies of additional {sup 3}He-rich events detected at multiple spacecraft will be needed to establish the relative importance of the various mechanisms.

79 ASTRONOMY AND ASTROPHYSICS↗

POINT-SPREAD FUNCTIONS FOR THE EXTREME-ULTRAVIOLET CHANNELS OF SDO/AIA TELESCOPES

We present the stray-light point-spread functions (PSFs) and their inverses we characterized for the Atmospheric Imaging Assembly (AIA) EUV telescopes on board the Solar Dynamics Observatory (SDO) spacecraft. The inverse kernels are approximate inverses under convolution. Convolving the original Level 1 images with them produces images with improved stray-light characteristics. We demonstrate the usefulness of these PSFs by applying them to two specific cases: photometry and differential emission measure (DEM) analysis. The PSFs consist of a narrow Gaussian core, a diffraction component, and a diffuse component represented by the sum of a Gaussian-truncated Lorentzian and a shoulder Gaussian. We determined the diffraction term using the measured geometry of the diffraction pattern identified in flare images and the theoretically computed intensities of the principal maxima of the first few diffraction orders. To determine the diffuse component, we fitted its parameterized model using iterative forward-modeling of the lunar interior in the SDO/AIA images from the 2011 March 4 lunar transit. We find that deconvolution significantly improves the contrast in dark features such as miniature coronal holes, though the effect was marginal in bright features. On a percentage-scattering basis, the PSFs for SDO/AIA are better by a factor of two than that of the EUV telescope on board the Transition Region And Coronal Explorer mission. A preliminary analysis suggests that deconvolution alone does not affect DEM analysis of small coronal loop segments with suitable background subtraction. We include the derived PSFs and their inverses as supplementary digital materials.

79 ASTRONOMY AND ASTROPHYSICS↗

Non-Markovian dynamics of quantum open systems embedded in a hybrid environment

Quantum systems of interest are typically coupled to several quantum channels (more generally environments). In this paper, we develop an exact stochastic Schrödinger equation for an open quantum system coupled to a hybrid environment containing both bosonic and fermionic particles. Such a stochastic differential equation may be obtained directly from a microscopic model through employing a classical complex Gaussian noise and a non-commutative fermionic noise to simulate the hybrid bath. As an immediate application of our developed stochastic approach, we show that the evolution of the reduced density matrix can be derived by taking the average over both the bosonic noise and the fermionic noise. Three specific examples are given in this paper to illustrate that the hybrid quantum trajectory is fully consistent with the standard quantum mechanics. Our examples also shed new light on the special features exhibited by the fermionic bath and bosonic bath. - Highlights: • Exact dynamic equation for quantum open system coupled to a hybrid bath is derived. • The equation is consistent with standard quantum mechanics. • The dynamic equation derived is useful in exploring non-Markovian features.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

The 1600 Å Emission Bump in Protoplanetary Disks: A Spectral Signature of H{sub 2}O Dissociation

The FUV continuum spectrum of many accreting pre-main sequence stars, Classical T Tauri Stars (CTTSs), does not continue smoothly from the well-studied Balmer continuum emission in the NUV, suggesting that additional processes contribute to the short-wavelength emission in these objects. The most notable spectral feature in the FUV continuum of some CTTSs is a broad emission approximately centered at 1600 Å, which has been referred to as the “1600 Å Bump.” The origin of this feature remains unclear. In an effort to better understand the molecular properties of planet-forming disks and the UV spectral properties of accreting protostars, we have assembled archival FUV spectra of 37 disk-hosting systems observed by the Hubble Space Telescope -Cosmic Origins Spectrograph. Clear 1600 Å Bump emission is observed above the smooth, underlying 1100–1800 Å continuum spectrum in 19/37 Classical T Tauri disks in the HST -COS sample, with the detection rate in transition disks (8/8) being much higher than that in primordial or non-transition sources (11/29). We describe a spectral deconvolution analysis to separate the Bump (spanning 1490–1690 Å) from the underlying FUV continuum, finding an average Bump luminosity L (Bump) ≈ 7 × 10{sup 29} erg s{sup −1}. Parameterizing the Bump with a combination of Gaussian and polynomial components, we find that the 1600 Å Bump is characterized by a peak wavelength λ {sub o} = 1598.6 ± 3.3 Å, with FWHM = 35.8 ± 19.1 Å. Contrary to previous studies, we find that this feature is inconsistent with models of H{sub 2} excited by electron -impact. We show that this Bump makes up between 5%–50% of the total FUV continuum emission in the 1490–1690 Å band and emits roughly 10%–80% of the total fluorescent H{sub 2} luminosity for stars with well-defined Bump features. Energetically, this suggests that the carrier of the 1600 Å Bump emission is powered by Ly α photons. We argue that the most likely mechanism is Ly α -driven dissociation of H{sub 2}O in the inner disk, r ≲ 2 au. We demonstrate that non-thermally populated H{sub 2}O fragments can qualitatively account for the observed emission (discrete and continuum) and find that the average Ly α -driven H{sub 2}O dissociation rate is 1.7 × 10{sup 42} water molecules s{sup −1}.

79 ASTRONOMY AND ASTROPHYSICS↗

A novel conditional generative model for efficient ensemble forecasts of state variables in large-scale geological carbon storage

Integrating monitoring data to efficiently update reservoir pressure and CO 2 plume distribution forecasts presents a significant challenge in geological carbon storage (GCS) applications. Inverse modeling techniques are commonly used to fuse observational data and refine reservoir model parameters, thereby improving state variable forecasts. However, these techniques often rely on linear or Gaussian assumptions, which can limit their effectiveness in accurately predicting state variables. Moreover, simulating large-scale three-dimensional (3D) GCS problems is computationally expensive, making iterative runs in inverse problems prohibitive. To address these challenges, we propose a conditional generative model utilizing the score-based diffusion method for real-time 3D pressure and saturation field distribution predictions. Our approach involves solving the score function with a mini-batch-based Monte Carlo estimator to generate labeled data. This data is subsequently employed to train a fully connected neural network, enabling it to learn the conditional sample generator within a supervised learning framework. This method enables the rapid generation of a large ensemble of predictions, facilitating comprehensive uncertainty quantification of state variables. Here we applied our method to forecast the dynamic 3D distributions of pressure and saturation fields over a 30-year injection period. The statistical assessment with low root mean square error (RMSE) values demonstrates that our method can accurately predict the spatiotemporal distributions of both pressure and saturation fields. Moreover, the developed conditional generative model shows high computational efficiency by generating 100 ensemble forecasts of 3D state variables in less than 10 min. The consistency between ensemble averages and ground truth values further illustrates the model’s capability to capture state variable dynamics during the CO 2 plume injection process. Notably, the ground truth values fall within the ensemble forecasts, indicating that our uncertainty quantification effectively captures variability and potential noise in the observations. Thus, the developed conditional generative model proves to be a more efficient, accurate, and practical tool for GCS applications, facilitating timely risk analysis and informed decision-making.

58 GEOSCIENCES↗

SU-D-9A-03: STAMP: Simulator for Texture Analysis in MRI/PET

Purpose: To develop a convenient simulation platform to facilitate PET/MR image analysis with the prospect of gaining a better understanding of the influence of acquisition parameters on PET/MRI textural features. The simulation platform is demonstrated by showing textural variations of a representative case study using different image acquisition parameters. Methods: The simulation platform is composed of MRI simulators JEMRIS and SIMRI to achieve simulations of customized MRI sequences on sample tumor models. The PET simulator GATE is used to get 2D and 3D Monte Carlo acquisitions of voxelized PET sources using a phantom geometry and a customized scanner architecture. The platform incorporates a series of graphical user interfaces written in Matlab. Two GUIs are used to facilitate communication with the simulation executables installed on a computer cluster. A third GUI is used to collect and display the clinical and simulated images, as well as fused PET/MRI images, and perform computation of textural features.To illustrate the capabilities of this platform, one FDG-PET and T1-weighted (T1w) digitized tumor models were generated from clinical images of a soft-tissue sarcoma patient. Numerically simulated MR images were produced using 3 different echo times (TE) and 5 different repetition times (TR). PET 2D images were simulated using an OSEM algorithm with 1 to 32 iterations and a post-reconstruction Gaussian filter of 0, 2, 4 or 6 mm width. Results: STAMP was successfully used to produce numerically simulated FDG-PET and MRI images, and to calculate their corresponding textures. Three typical textures (GLCM-Contrast, GLSZM-ZSV and NGTDM-Coarseness) were found to vary by a range of 45% on average compared to reference scanning conditions in the case of FDG-PET, and by a range of 40% in the case of T1w MRI. Conclusion: We have successfully developed a Matlab-based simulation platform to facilitate PET/MRI texture image analysis for outcome prediction.

60 APPLIED LIFE SCIENCES↗

An unsupervised machine-learning checkpoint-restart algorithm using Gaussian mixtures for particle-in-cell simulations

We propose an unsupervised machine-learning checkpoint-restart (CR) algorithm for particle-in-cell (PIC) algorithms using Gaussian mixtures (GM). The algorithm compresses the particle population per spatial cell by constructing a velocity distribution function using GM. Particles are reconstructed at restart time by local resampling of the Gaussians. To guarantee fidelity of the CR process, we ensure the exact preservation of invariants such as charge, momentum, and energy for both compression and reconstruction stages, everywhere on the mesh. We also ensure the preservation of Gauss' law after particle reconstruction by exactly matching the density profile at restart time. As a result, the GM CR algorithm is shown to provide a clean, conservative restart capability while potentially affording orders of magnitude savings in input/output requirements. Here, we demonstrate the algorithm using a recently developed exactly energy- and charge-conserving PIC algorithm using both electrostatic and electromagnetic tests. The tests demonstrate not only a high-fidelity CR capability, but also its potential for enhancing the fidelity of the PIC solution for a given particle resolution.

97 MATHEMATICS AND COMPUTING↗

Prediction of atomization energy using graph kernel and active learning

Data-driven prediction of molecular properties presents unique challenges to the design of machine learning methods concerning data structure/dimensionality, symmetry adaption, and confidence management. In this paper, we present a kernel-based pipeline that can learn and predict the atomization energy of molecules with high accuracy. The framework employs Gaussian process regression to perform predictions based on the similarity between molecules, which is computed using the marginalized graph kernel. To apply the marginalized graph kernel, a spatial adjacency rule is first employed to convert molecules into graphs whose vertices and edges are labeled by elements and interatomic distances, respectively. We then derive formulas for the efficient evaluation of the kernel. Specific functional components for the marginalized graph kernel are proposed, while the effects of the associated hyperparameters on accuracy and predictive confidence are examined. We show that the graph kernel is particularly suitable for predicting extensive properties because its convolutional structure coincides with that of the covariance formula between sums of random variables. Using an active learning procedure, we demonstrate that the proposed method can achieve a mean absolute error of 0.62 ± 0.01 kcal/mol using as few as 2000 training samples on the QM7 dataset.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Development and testing of texture discriminators for the analysis of trabecular bone in proximal femur radiographs

Purpose: Texture analysis of femur radiographs may serve as a potential low cost technique to predict osteoporotic fracture risk and has received considerable attention in the past years. A further application of this technique may be the measurement of the quality of specific bone compartments to provide useful information for treatment of bone fractures. Two challenges of texture analysis are the selection of the best suitable texture measure and reproducible placement of regions of interest (ROIs). The goal of this in vitro study was to automatically place ROIs in radiographs of proximal femur specimens and to calculate correlations between various different texture analysis methods and the femurs' anchorage strength. Methods: Radiographs were obtained from 14 femoral specimens and bone mineral density (BMD) was measured in the femoral neck. Biomechanical testing was performed to assess the anchorage strength in terms of failure load, breakaway torque, and number of cycles. Images were segmented using a framework that is based on the usage of level sets and statistical in-shape models. Five ROIs were automatically placed in the head, upper and lower neck, trochanteric, and shaft compartment in an atlas subject. All other subjects were registered rigidly, affinely, and nonlinearly, and the resulting transformation was used to map the five ROIs onto the individual femora. Results: In each ROI, texture features were extracted using gray level co-occurence matrices (GLCM), third-order GLCM, morphological gradients (MGs), Minkowski dimensions (MDs), Minkowski functionals (MFs), Gaussian Markov random fields, and scaling index method (SIM). Coefficients of determination for each texture feature with parameters of anchorage strength were computed. In a stepwise multiregression analysis, the most predictive parameters were identified in different models. Texture features were highly correlated with anchorage strength estimated by the failure load of up to R{sup 2}=0.61 (MF and MG features, p<0.01) and were partially independent of BMD. The correlations were dependent on the choice of the ROI and the texture measure. The best predictive multiregression model for failure load R{sub adj}{sup 2}=0.86 (p<0.001) included a set of recently developed texture methods (MF and SIM) but excluded bone mineral density and commonly used texture measures. Conclusions: The results suggest that texture information contained in trabecular bone structure visualized on radiographs may predict whether an implant anchorage can be used and may determine the local bone quality from preoperative radiographs.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

A Theoretical Framework for Calibration in Computer Models: Parametrization, Estimation and Convergence Properties

Calibration parameters in deterministic computer experiments are those attributes that cannot be measured or available in physical experiments. Here, an approach to estimate them by using data from physical experiments and computer simulations. A theoretical framework is given which allows us to study the issues of parameter identifiability and estimation. We define the L 2 -consistency for calibration as a justification for calibration methods. It is shown that a simplified version of the original KO method leads to asymptotically L 2 -inconsistent calibration. This L 2 -inconsistency can be remedied by modifying the original estimation procedure. A novel calibration method, called the L 2 calibration, is proposed and proven to be L 2 -consistent and enjoys optimal convergence rate. Furthermore a numerical example and some mathematical analysis are used to illustrate the source of the L 2 -inconsistency problem.

97 MATHEMATICS AND COMPUTING↗