Search NASA⌕ Search

SEARCH · Search NASA

Results for “distributed estimation”

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 1,099 records · Page 61

Experimental study of residence time distributions of ball-mill circuits grinding coal-water mixtures

Residence time distributions (RTDs) were estimated by water tracing in a number of wet overflow ball mills (diameters 0.38 to 4.65 m) producing dense, coal-water slurries. In open-circuit mills of 0.38 m diameter and various length-diameter (LID) ratios, the mean residence times of solid were also determined from measured mill holdups. Holdup increased with increased mill feed rate, but the mean residence times of coal and water were still equal to each other. The experimental residence time distributions were fitted to the Mori-Jimbo-Yamazaki semi-infinite, axial mixing model, and the dimensionless mixing coefficient was determined for each of 25 tests in single- and two-compartment mills. This coefficient was found to be independent to the feed rate but linearly proportional to the D/L ratio. The mixing coefficient was smaller for two-compartment mills than for single-compartment mills, showing that there was reduced mixing introduced by the diaphragm separating the compartments. Equations are given to scale residence time distributions for changes in mill diameter and length.

01 COAL, LIGNITE, AND PEAT↗

A Multiscale Bayesian Data Integration Approach for Mapping Air Dose Rates around the Fukushima Daiichi Nuclear Power Plant - 17166

This paper presents a multiscale data integration method to estimate the spatial distribution of air dose rates in the regional scale around the Fukushima Daiichi Nuclear Power Plant. We integrate various types of datasets, such as ground-based walk and car surveys, and airborne surveys, all of which have different scales, resolutions, spatial coverage, and accuracy. This method is based on geo-statistics to represent spatial heterogeneous structures, and also on Bayesian hierarchical models to integrate multiscale, multi-type datasets in a consistent manner. The Bayesian method allows us to quantify the uncertainty in the estimates, and to provide the confidence intervals that are critical for robust decision-making. Although this approach is primarily data-driven, it has great flexibility to include mechanistic models for representing radiation transport or other complex correlations. We demonstrate our approach using three types of datasets collected at the same time over Fukushima City, Japan: (1) coarse-resolution airborne surveys covering the entire area, (2) car surveys along major roads, and (3) walk surveys in multiple neighborhoods. Results show that the method can successfully integrate three types of datasets in a consistent manner, and create an integrated map (including the confidence intervals) of air dose rates over the domain in high resolution. (authors)

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Selection biases in empirical p(z) methods for weak lensing

To measure the mass of foreground objects with weak gravitational lensing, one needs to estimate the redshift distribution of lensed background sources. This is commonly done in an empirical fashion, i.e. with a reference sample of galaxies of known spectroscopic redshift, matched to the source population. In this paper, we develop a simple decision tree framework that, under the ideal conditions of a large, purely magnitude-limited reference sample, allows an unbiased recovery of the source redshift probability density function p(z), as a function of magnitude and colour. We use this framework to quantify biases in empirically estimated p(z) caused by selection effects present in realistic reference and weak lensing source catalogues, namely (1) complex selection of reference objects by the targeting strategy and success rate of existing spectroscopic surveys and (2) selection of background sources by the success of object detection and shape measurement at low signal to noise. For intermediate-to-high redshift clusters, and for depths and filter combinations appropriate for ongoing lensing surveys, we find that (1) spectroscopic selection can cause biases above the 10 per cent level, which can be reduced to ≈5 per cent by optimal lensing weighting, while (2) selection effects in the shape catalogue bias mass estimates at or below the 2 per cent level. Finally, this illustrates the importance of completeness of the reference catalogues for empirical redshift estimation.

79 ASTRONOMY AND ASTROPHYSICS↗

Estimation of soil classes and their relationship to grapevine vigor in a Bordeaux vineyard: advancing the practical joint use of electromagnetic induction (EMI) and NDVI datasets for precision viticulture

Working within a vineyard in the Pessac Léognan Appellation of Bordeaux, France, this study documents the potential of using simple statistical methods with spatially-resolved and increasingly available electromagnetic induction (EMI) geophysical and normalized difference vegetation index (NDVI) datasets to accurately estimate Bordeaux vineyard soil classes and to quantitatively explore the relationship between vineyard soil types and grapevine vigor. First, co-located electrical tomographic tomography (ERT) and EMI datasets were compared to gain confidence about how the EMI method averaged soil properties over the grapevine rooting depth. Then, EMI data were used with core soil texture and soil-pit based interpretations of Bordeaux soil types (Brunisol, Redoxisol, Colluviosol and Calcosol) to estimate the spatial distribution of geophysically-identified Bordeaux soil classes. A strong relationship (r = 0.75, p < 0.01) was revealed between the geophysically-identified Bordeaux soil classes and NDVI (both 2 m resolution), showing that the highest grapevine vigor was associated with the Bordeaux soil classes having the largest clay fraction. The results suggest that within-block variability of grapevine vigor was largely controlled by variability in soil classes, and that carefully collected EMI and NDVI datasets can be exceedingly helpful for providing quantitative estimates of vineyard soil and vigor variability, as well as their covariation. The method is expected to be transferable to other viticultural regions, providing an approach to use easy-to-acquire, high resolution datasets to guide viticultural practices, including routine management and replanting.

54 ENVIRONMENTAL SCIENCES↗

Investigation of the effect of noise parameters of a 3D lidar on the error in estimating relief signatures of distant objects from 2D field intensity distributions of reflected radiation

We consider the influence of noise parameters of a 3D lidar and external background radiation on the error in estimating the relief signatures of distant objects from 2D field intensity distributions of the reflected radiation. A physical and mathematical model is developed to analyse the effect of noise. The dependence of the error in estimating the relief signatures at various signal-to-noise ratios is investigated taking into account the parameters of real multi-element photodetectors (MEPDs). The obtained results indicate the possibility of evaluating relief signatures of distant objects with high accuracy based on standard MEPDs simultaneously for a large number of points on the object’s surface with a signal-to-noise ratio of more than 4. (paper)

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Delineating Facies Spatial Distribution by Integrating Ensemble Data Assimilation and Indicator Geostatistics with Level Set Transformation.

A new approach is developed to delineate the spatial distribution of discrete facies (geological units that have unique distributions of hydraulic, physical, and/or chemical properties) conditioned not only on direct data (measurements directly related to facies properties, e.g., grain size distribution obtained from borehole samples) but also on indirect data (observations indirectly related to facies distribution, e.g., hydraulic head and tracer concentration). Our method integrates for the first time ensemble data assimilation with traditional transition probability-based geostatistics. The concept of level set is introduced to build shape parameterization that allows transformation between discrete facies indicators and continuous random variables. The spatial structure of different facies is simulated by indicator models using conditioning points selected adaptively during the iterative process of data assimilation. To evaluate the new method, a two-dimensional semi-synthetic example is designed to estimate the spatial distribution and permeability of two distinct facies from transient head data induced by pumping tests. The example demonstrates that our new method adequately captures the spatial pattern of facies distribution by imposing spatial continuity through conditioning points. The new method also reproduces the overall response in hydraulic head field with better accuracy compared to data assimilation with no constraints on spatial continuity on facies.

97 MATHEMATICS AND COMPUTING↗

The Spatial Distribution of Forest Biomass in the Brazilian Amazon: A Comparison of Estimates

The amount of carbon released to the atmosphere as a result of deforestation is determined, in part, by the amount of carbon held in the biomass of the forests converted to other uses. Uncertainty in forest biomass is responsible for much of the uncertainty in current estimates of the flux of carbon from land-use change. We compared several estimates of forest biomass for the Brazilian Amazon, based on spatial interpolations of direct measurements, relationships to climatic variables, and remote sensing data. We asked three questions. First, do the methods yield similar estimates? Second, do they yield similar spatial patterns of distribution of biomass? And, third, what factors need most attention if we are to predict more accurately the distribution of forest biomass over large areas? Amazonian forests (including dead and below-ground biomass) vary by more than a factor of two, from a low of 39 PgC to a high of 93 PgC. Furthermore, the estimates disagree as to the regions of high and low biomass. The lack of agreement among estimates confirms the need for reliable determination of aboveground biomass over large areas. Potential methods include direct measurement of biomass through forest inventories with improved allometric regression equations, dynamic modeling of forest recovery following observed stand-replacing disturbances (the approach used in this research), and estimation of aboveground biomass from airborne or satellite-based instruments sensitive to the vertical structure plant canopies.

Houghton, R. A.↗

Exploring Alternate Parameterizations for Snowfall with Validation from Satellite and Terrestrial Radars

Increases in computational resources have allowed operational forecast centers to pursue experimental, high resolution simulations that resolve the microphysical characteristics of clouds and precipitation. These experiments are motivated by a desire to improve the representation of weather and climate, but will also benefit current and future satellite campaigns, which often use forecast model output to guide the retrieval process. Aircraft, surface and radar data from the Canadian CloudSat/CALIPSO Validation Project are used to check the validity of size distribution and density characteristics for snowfall simulated by the NASA Goddard six-class, single-moment bulk water microphysics scheme, currently available within the Weather Research and Forecast (WRF) Model. Widespread snowfall developed across the region on January 22, 2007, forced by the passing of a midlatitude cyclone, and was observed by the dual-polarimetric, C-band radar King City, Ontario, as well as the NASA 94 GHz CloudSat Cloud Profiling Radar. Combined, these data sets provide key metrics for validating model output: estimates of size distribution parameters fit to the inverse-exponential equations prescribed within the model, bulk density and crystal habit characteristics sampled by the aircraft, and representation of size characteristics as inferred by the radar reflectivity at C- and W-band. Specified constants for distribution intercept and density differ significantly from observations throughout much of the cloud depth. Alternate parameterizations are explored, using column-integrated values of vapor excess to avoid problems encountered with temperature-based parameterizations in an environment where inversions and isothermal layers are present. Simulation of CloudSat reflectivity is performed by adopting the discrete-dipole parameterizations and databases provided in literature, and demonstrate an improved capability in simulating radar reflectivity at W-band versus Mie scattering assumptions.

Molthan, Andrew L.↗

Exploring Alternative Parameterizations for Snowfall with Validation from Satellite and Terrestrial Radars

Increases in computational resources have allowed operational forecast centers to pursue experimental, high resolution simulations that resolve the microphysical characteristics of clouds and precipitation. These experiments are motivated by a desire to improve the representation of weather and climate, but will also benefit current and future satellite campaigns, which often use forecast model output to guide the retrieval process. The combination of reliable cloud microphysics and radar reflectivity may constrain radiative transfer models used in satellite simulators during future missions, including EarthCARE and the NASA Global Precipitation Measurement. Aircraft, surface and radar data from the Canadian CloudSat/CALIPSO Validation Project are used to check the validity of size distribution and density characteristics for snowfall simulated by the NASA Goddard six-class, single moment bulk water microphysics scheme, currently available within the Weather Research and Forecast (WRF) Model. Widespread snowfall developed across the region on January 22, 2007, forced by the passing of a mid latitude cyclone, and was observed by the dual-polarimetric, C-band radar King City, Ontario, as well as the NASA 94 GHz CloudSat Cloud Profiling Radar. Combined, these data sets provide key metrics for validating model output: estimates of size distribution parameters fit to the inverse-exponential equations prescribed within the model, bulk density and crystal habit characteristics sampled by the aircraft, and representation of size characteristics as inferred by the radar reflectivity at C- and W-band. Specified constants for distribution intercept and density differ significantly from observations throughout much of the cloud depth. Alternate parameterizations are explored, using column-integrated values of vapor excess to avoid problems encountered with temperature-based parameterizations in an environment where inversions and isothermal layers are present. Simulation of CloudSat reflectivity is performed by adopting the discrete-dipole parameterizations and databases provided in literature, and demonstrate an improved capability in simulating radar reflectivity at W-band versus Mie scattering assumptions.

Molthan, Andrew L.↗

A Scalable Gradient Free Method for Bayesian Experimental Design with Implicit Models

Bayesian experimental design (BED) is to answer the question that how to choose designs that maximize the information gathering. For implicit models, where the likelihood is intractable but sampling is possible, conventional BED methods have difficulties in efficiently estimating the posterior distribution and maximizing the mutual information (MI) between data and parameters. Recent work proposed the use of gradient ascent to maximize a lower bound on MI to deal with these issues. However, the approach requires a sampling path to compute the pathwise gradient of the MI lower bound with respect to the design variables, and such a pathwise gradient is usually inaccessible for implicit models. In this paper, we propose a novel approach that leverages recent advances in stochastic approximate gradient ascent incorporated with a smoothed variational MI estimator for efficient and robust BED. Without the necessity of pathwise gradients, our approach allows the design process to be achieved through a unified procedure with an approximate gradient for implicit models. Several experiments show that our approach outperforms baseline methods, and significantly improves the scalability of BED in high-dimensional problems.

Zhang, Jiaxin↗

Spatial and temporal prediction of radiation dose rates near Fukushima Daiichi Nuclear Power Plant

In this paper, we have developed a methodology to estimate the spatiotemporal distribution of radiation air dose rates around the Fukushima Daiichi Nuclear Power Plant (FDNPP). In our exploratory data analysis, we found that (1) the temporal evolution of dose rates is composed of a log-linear decay trend and fluctuations of air dose rates that are spatially correlated among adjacent monitoring posts; and (2) the slope of the log-linear environmental decay trend can be represented as a function of the apparent initial dose rates, coordinate position, land-use type, and soil type. From these observations, we first estimated the log-linear decay trend at each location based on these predictors, using the random forest method. We then developed a modified Kalman filter coupled with a Gaussian process model to estimate the dose-rate time series at a given location and time. We applied this method to the Fukushima evacuation zone (as of March 2017), which included 17 monitoring post locations (with monitoring datasets collected between 2014 and 2018) and generated a time series of dose-rate maps. Our results show that this approach allows us to produce accurate spatial and temporal predictions of radiation dose-rate maps using limited spatiotemporal measurements.

61 RADIATION PROTECTION AND DOSIMETRY↗

Patterns of Failure After Proton Therapy in Medulloblastoma; Linear Energy Transfer Distributions and Relative Biological Effectiveness Associations for Relapses

Purpose: The pattern of failure in medulloblastoma patients treated with proton radiation therapy is unknown. For this increasingly used modality, it is important to ensure that outcomes are comparable to those in modern photon series. It has been suggested this pattern may differ from photons because of variations in linear energy transfer (LET) and relative biological effectiveness (RBE). In addition, the use of matching fields for delivery of craniospinal irradiation (CSI) may influence patterns of relapse. Here we report the patterns of failure after the use of protons, compare it to that in the available photon literature, and determine the LET and RBE values in areas of recurrence. Methods and Materials: Retrospective review of patients with medulloblastoma treated with proton radiation therapy at Massachusetts General Hospital (MGH) between 2002 and 2011. We documented the locations of first relapse. Discrete failures were contoured on the original planning computed tomography scan. Monte Carlo calculation methods were used to estimate the proton LET distribution. Models were used to estimate RBE values based on the LET distributions. Results: A total of 109 patients were followed for a median of 38.8 months (range, 1.4-119.2 months). Of the patients, 16 experienced relapse. Relapse involved the supratentorial compartment (n=8), spinal compartment (n=11), and posterior fossa (n=5). Eleven failures were isolated to a single compartment; 6 failures in the spine, 4 failures in the supratentorium, and 1 failure in the posterior fossa. The remaining patients had multiple sites of disease. One isolated spinal failure occurred at the spinal junction of 2 fields. None of the 70 patients treated with an involved-field-only boost failed in the posterior fossa outside of the tumor bed. We found no correlation between Monte Carlo-calculated LET distribution and regions of recurrence. Conclusions: The most common site of failure in patients treated with protons for medulloblastoma was outside of the posterior fossa. The most common site for isolated local failure was the spine. We recommend consideration of spinal imaging in follow-up and careful attention to dose distribution in the spinal junction regions. Development of techniques that do not require field matching may be of benefit. We did not identify a direct correlation between lower LET values and recurrence in medulloblastoma patients treated with proton therapy. Patterns of failure do not appear to differ from those in patients treated with photon therapy.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Recovering non-Maxwellian particle velocity distribution functions from collective Thomson-scattered spectra

Collective optical Thomson scattering (TS) is a diagnostic commonly used to characterize plasma parameters. These parameters are typically extracted by a fitting algorithm that minimizes the difference between a measured scattered spectrum and an analytic spectrum calculated from the velocity distribution function (VDF) of the plasma. However, most existing TS analysis algorithms assume that the VDFs are Maxwellian, and applying an algorithm that makes this assumption does not accurately extract the plasma parameters of a non-Maxwellian plasma due to the effect of non-Maxwellian deviations on the TS spectra. We present new open-source numerical tools for forward modeling analytic spectra from arbitrary VDFs and show that these tools are able to more accurately extract plasma parameters from synthetic TS spectra generated by non-Maxwellian VDFs compared to standard TS algorithms. Estimated posterior probability distributions of fits to synthetic spectra for a variety of example non-Maxwellian VDFs are used to determine uncertainties in the extracted plasma parameters and show that correlations between parameters can significantly affect the accuracy of fits in plasmas with non-Maxwellian VDFs.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

A dosimetric uncertainty analysis for photon-emitting brachytherapy sources: Report of AAPM Task Group No. 138 and GEC-ESTRO

This report addresses uncertainties pertaining to brachytherapy single-source dosimetry preceding clinical use. The International Organization for Standardization (ISO) Guide to the Expression of Uncertainty in Measurement (GUM) and the National Institute of Standards and Technology (NIST) Technical Note 1297 are taken as reference standards for uncertainty formalism. Uncertainties in using detectors to measure or utilizing Monte Carlo methods to estimate brachytherapy dose distributions are provided with discussion of the components intrinsic to the overall dosimetric assessment. Uncertainties provided are based on published observations and cited when available. The uncertainty propagation from the primary calibration standard through transfer to the clinic for air-kerma strength is covered first. Uncertainties in each of the brachytherapy dosimetry parameters of the TG-43 formalism are then explored, ending with transfer to the clinic and recommended approaches. Dosimetric uncertainties during treatment delivery are considered briefly but are not included in the detailed analysis. For low- and high-energy brachytherapy sources of low dose rate and high dose rate, a combined dosimetric uncertainty <5% (k=1) is estimated, which is consistent with prior literature estimates. Recommendations are provided for clinical medical physicists, dosimetry investigators, and source and treatment planning system manufacturers. These recommendations include the use of the GUM and NIST reports, a requirement of constancy of manufacturer source design, dosimetry investigator guidelines, provision of the lowest uncertainty for patient treatment dosimetry, and the establishment of an action level based on dosimetric uncertainty. These recommendations reflect the guidance of the American Association of Physicists in Medicine (AAPM) and the Groupe Europeen de Curietherapie-European Society for Therapeutic Radiology and Oncology (GEC-ESTRO) for their members and may also be used as guidance to manufacturers and regulatory agencies in developing good manufacturing practices for sources used in routine clinical treatments.

61 RADIATION PROTECTION AND DOSIMETRY↗

A 31 GHz SURVEY OF LOW-FREQUENCY SELECTED RADIO SOURCES

The 100 m Robert C. Byrd Green Bank Telescope and the 40 m Owens Valley Radio Observatory telescope have been used to conduct a 31 GHz survey of 3165 known extragalactic radio sources over 143 deg{sup 2} of the sky. Target sources were selected from the NRAO VLA Sky Survey in fields observed by the Cosmic Background Imager (CBI); most are extragalactic active galactic nuclei (AGNs) with 1.4 GHz flux densities of 3-10 mJy. The resulting 31 GHz catalogs are presented in full online. Using a maximum-likelihood analysis to obtain an unbiased estimate of the distribution of the 1.4-31 GHz spectral indices of these sources, we find a mean 31-1.4 GHz flux ratio of 0.110 +- 0.003 corresponding to a spectral index of alpha = -0.71 +- 0.01 (S {sub n}u propor to nu{sup a}lpha); 9.0% +- 0.8% of sources have alpha > - 0.5 and 1.2% +- 0.2% have alpha > 0. By combining this spectral-index distribution with 1.4 GHz source counts, we predict 31 GHz source counts in the range 1 mJy < S {sub 31} < 4 mJy, N(>S {sub 31}) = (16.7 +- 1.7) deg{sup -2}(S {sub 31}/1 mJy){sup -0.80+}-{sup 0.07}. We also assess the contribution of mJy-level (S {sub 1.4GHz} < 3.4 mJy) radio sources to the 31 GHz cosmic microwave background power spectrum, finding a mean power of l(l + 1)C {sup src} {sub l}/(2pi) = 44 +- 14 muK{sup 2} and a 95% upper limit of 80 muK{sup 2} at l = 2500. Including an estimated contribution of 12 muK{sup 2} from the population of sources responsible for the turn-up in counts below S {sub 1.4GHz} = 1 mJy, this amounts to 21% +- 7% of what is needed to explain the CBI high-l excess signal, 275 +- 63 muK{sup 2}. These results are consistent with other measurements of the 31 GHz point-source foreground.

79 ASTRONOMY AND ASTROPHYSICS↗

Accurate Characterization of Real Networks from Inaccurate Measurements

Our nation's dependence on information networks makes it vital to anticipate disruptions or find weaknesses in these networks. But networks like the Internet are vast, distributed, and there is no mechanism to completely collect their structure. We are restricted to specific data collection schemes (like traceroute samples from router interfaces) that examine tiny portions of such a network. It has been empirically documented and theoretically proven that these measurements have significant biases, and direct inferences from them will be wrong. But these data collection mechanisms have limited flexibility and cannot be easily modified. Moreover, in many applications there are limits on how much data can be collected. How do we make accurate inferences of network properties with biased and limited measurements? The general problem this report deals with is how to work with incompletely observed networks. We will present several different approaches to this problem. First we will present an approach to estimate the degree distribution of a graph by sampling only a small portion of the vertices. This algorithm provides provably accurate results with sublinear samples. An alternative approach would be to try to enhance the information in the by selective collecting new information by probing for neighbors of a vertex or presence of individual edges. A different setting for working with incomplete arises when we have full access to local information, but do not have any global version of the graph. Can we still identify critical nodes in such a graph? We present an approach to identify such nodes efficiently. Finally, how can we put these ideas together to identify the structure of a network? We present an approach that can complement the existing approaches for network mapping. We start with an estimate of network structure based on existing network mapping methods. Then we find a critical router in the network, use the traffic through this network to selectively collect new data to enhance our prediction.

97 MATHEMATICS AND COMPUTING↗

Impact of the MLC on the MRI field distortion of a prototype MRI-linac

Purpose: To cope with intrafraction tumor motion, integrated MRI-linac systems for real-time image guidance are currently under development. The multileaf collimator (MLC) is a key component in every state-of-the-art radiotherapy treatment system, allowing for accurate field shaping and tumor tracking. This work quantifies the magnetic impact of a widely used MLC on the MRI field homogeneity for such a modality.Methods: The finite element method was employed to model a MRI-linac assembly comprised of a 1.0 T split-bore MRI magnet and the key ferromagnetic components of a Varian Millennium 120 MLC, namely, the leaves and motors. Full 3D magnetic field maps of the system were generated. From these field maps, the peak-to-peak distortion within the MRI imaging volume was evaluated over a 30 cm diameter sphere volume (DSV) around the isocenter and compared to a maximum preshim inhomogeneity of 300 μT. Five parametric studies were performed: (1) The source-to-isocenter distance (SID) was varied from 100 to 200 cm, to span the range of a compact system to that with lower magnetic coupling. (2) The MLC model was changed from leaves only to leaves with motors, to determine the contribution to the total distortion caused by MLC leaves and motors separately. (3) The system was configured in the inline or perpendicular orientation, i.e., the linac treatment beam was oriented parallel or perpendicular to the magnetic field direction. (4) The treatment field size was varied from 0 × 0 to 20×20 cm{sup 2}, to span the range of clinical treatment fields. (5) The coil currents were scaled linearly to produce magnetic field strengths B{sub 0} of 0.5, 1.0, and 1.5 T, to estimate how the MLC impact changes with B{sub 0}.Results: (1) The MLC-induced MRI field distortion fell continuously with increasing SID. (2) MLC leaves and motors were found to contribute to the distortion in approximately equal measure. (3) Due to faster falloff of the fringe field, the field distortion was generally smaller in the perpendicular beam orientation. The peak-to-peak DSV distortion was below 300 μT at SID≥130 cm (perpendicular) and SID≥140 cm (inline) for the 1.0 T design. (4) The simulation of different treatment fields was identified to cause dynamic changes in the field distribution. However, the estimated residual distortion was below 1.2 mm geometric distortion at SID≥120 cm (perpendicular) and SID≥130 cm (inline) for a 10 mT/m frequency-encoding gradient. (5) Due to magnetic saturation of the MLC materials, the field distortion remained constant at B{sub 0}>1.0 T.Conclusions: This work shows that the MRI field distortions caused by the MLC cannot be ignored and must be thoroughly investigated for any MRI-linac system. The numeric distortion values obtained for our 1.0 T magnet may vary for other magnet designs with substantially different fringe fields, however the concept of modest increases in the SID to reduce the distortion to a shimmable level is generally applicable.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

An adaptive Gaussian process-based method for efficient Bayesian experimental design in groundwater contaminant source identification problems: ADAPTIVE GAUSSIAN PROCESS-BASED INVERSION

Surrogate models are commonly used in Bayesian approaches such as Markov Chain Monte Carlo (MCMC) to avoid repetitive CPU-demanding model evaluations. However, the approximation error of a surrogate may lead to biased estimations of the posterior distribution. This bias can be corrected by constructing a very accurate surrogate or implementing MCMC in a two-stage manner. Since the two-stage MCMC requires extra original model evaluations, the computational cost is still high. If the information of measurement is incorporated, a locally accurate approximation of the original model can be adaptively constructed with low computational cost. Based on this idea, we propose a Gaussian process (GP) surrogate-based Bayesian experimental design and parameter estimation approach for groundwater contaminant source identification problems. A major advantage of the GP surrogate is that it provides a convenient estimation of the approximation error, which can be incorporated in the Bayesian formula to avoid over-confident estimation of the posterior distribution. The proposed approach is tested with a numerical case study. Without sacrificing the estimation accuracy, the new approach achieves about 200 times of speed-up compared to our previous work using two-stage MCMC.

Zhang, Jiangjiang↗