Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Likelihood”

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 919 records · Page 51

Classical Simulation of Boson Sampling Based on Graph Structure

Boson sampling is a fundamentally and practically important task that can be used to demonstrate quantum supremacy using noisy intermediate-scale quantum devices. In this Letter, we present classical sampling algorithms for single-photon and Gaussian input states that take advantage of a graph structure of a linear-optical circuit. The algorithms’ complexity grows as so-called treewidth, which is closely related to the connectivity of a given linear-optical circuit. Using the algorithms, we study approximated simulations for local Haar-random linear-optical circuits. For equally spaced initial sources, we show that, when the circuit depth is less than the quadratic in the lattice spacing, the efficient simulation is possible with an exponentially small error. Notably, right after this depth, photons start to interfere each other and the algorithms’ complexity becomes subexponential in the number of sources, implying that there is a sharp transition of its complexity. Finally, when a circuit is sufficiently deep enough for photons to typically propagate to all modes, the complexity becomes exponential as generic sampling algorithms. We numerically implement a likelihood test with a recent Gaussian boson sampling experiment and show that the treewidth-based algorithm with a limited treewidth renders a larger likelihood than the experimental data.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Statistical quality indicators for electron-density maps

A likelihood-based metric for scoring the local agreement of a structure model with the observed electron density is described. The commonly used validation metrics for the local agreement of a structure model with the observed electron density, namely the real-space R (RSR) and the real-space correlation coefficient (RSCC), are reviewed. It is argued that the primary goal of all validation techniques is to verify the accuracy of the model, since precision is an inherent property of the crystal and the data. It is demonstrated that the principal weakness of both of the above metrics is their inability to distinguish the accuracy of the model from its precision. Furthermore, neither of these metrics in their usual implementation indicate the statistical significance of the result. The statistical properties of electron-density maps are reviewed and an improved alternative likelihood-based metric is suggested. This leads naturally to a χ{sup 2} significance test of the difference density using the real-space difference density Z score (RSZD). This is a metric purely of the local model accuracy, as required for effective model validation and structure optimization by practising crystallographers prior to submission of a structure model to the PDB. A new real-space observed density Z score (RSZO) is also proposed; this is a metric purely of the model precision, as a substitute for other precision metrics such as the B factor.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Weak data do not make a free lunch, only a cheap meal

Refinement and analysis of four structures with various data resolution cutoffs suggests that at present there are no reliable criteria for judging the diffraction data resolution limit and the condition I/σ(I) = 2.0 is reasonable. However, extending the limit by about 0.2 Å beyond the resolution defined by this threshold does not deteriorate the quality of refined structures and in some cases may be beneficial. Four data sets were processed at resolutions significantly exceeding the criteria traditionally used for estimating the diffraction data resolution limit. The analysis of these data and the corresponding model-quality indicators suggests that the criteria of resolution limits widely adopted in the past may be somewhat conservative. Various parameters, such as R{sub merge} and I/σ(I), optical resolution and the correlation coefficients CC{sub 1/2} and CC*, can be used for judging the internal data quality, whereas the reliability factors R and R{sub free} as well as the maximum-likelihood target values and real-space map correlation coefficients can be used to estimate the agreement between the data and the refined model. However, none of these criteria provide a reliable estimate of the data resolution cutoff limit. The analysis suggests that extension of the maximum resolution by about 0.2 Å beyond the currently adopted limit where the I/σ(I) value drops to 2.0 does not degrade the quality of the refined structural models, but may sometimes be advantageous. Such an extension may be particularly beneficial for significantly anisotropic diffraction. Extension of the maximum resolution at the stage of data collection and structure refinement is cheap in terms of the required effort and is definitely more advisable than accepting a too conservative resolution cutoff, which is unfortunately quite frequent among the crystal structures deposited in the Protein Data Bank.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

A Deep Learning Pipeline for Optimizing Large-scale Phase Field Simulations

Phase field (PF) simulations are computationally expensive but remain a key analysis tool to understand the complex mechanisms of additive manufacturing (AM) processes. Each PF simulation-aided analysis requires thousands of node hours on leadership-class supercomputers. One of the main goals of these analyses is the study of microstructure evolution during the build process which begins with the onset of nucleation. Nucleation occurs under certain thermomechanical conditions which are not known a priori and many PF simulations are required to identify ranges of input thermo-mechanical parameters that can result in the onset of nucleation. Since many of the simulations do not result in nucleation, an analysis campaign often ends up wasting tremendous amounts of precious computing resources executing nucleation-absent simulations. The goal of this work is to design and train deep learning models to inform a PF simulation about the likelihood of the occurrence of nucleation in a future simulation time-step based on the state summary over a finite number of past time-steps of a running simulation. If the prediction determines that the running simulation is unlikely to reach nucleation in the allotted time, then its execution is stopped immediately ultimately resulting in vast reduction in wasted computations when accrued over all the PF simulations typically performed in a single or multiple analysis campaign(s). The paper presents the performance of a machine learning pipeline that uses a convolutional neural network (CNN) model to learn an embedding which is then used with a self-attention network to build a multi-task deep learning model to predict the likelihood of nucleation. The model also predicts the input parameters used in a simulation. Performance is compared with a baseline pipeline that uses an off-the-shelf LeNet-5 model to learn the initial embedding. Despite their smaller size, performance results indicate significant improvement in accuracy of the proposed models compared to the larger baseline models.

Kannan, Ramakrishnan {ramki}↗

Using Accelerated Testing To Predict Module Reliability: Preprint

Long-term reliability is critical to the cost effectiveness and commercial success of photovoltaic (PV) products. Today most PV modules are warranted for 25 years, but there is no accepted test protocol to validate a 25-year lifetime. The qualification tests do an excellent job of identifying design, materials, and process flaws that are likely to lead to premature failure (infant mortality), butthey are not designed to test for wear-out mechanisms that limit lifetime. This paper presents a method for evaluating the ability of a new PV module technology to survive long-term exposure to specific stresses. The authors propose the use of baseline technologies with proven long-term field performance as controls in the accelerated stress tests. The performance of new-technology modules canthen be evaluated versus that of proven-technology modules. If the new-technology demonstrates equivalent or superior performance to the proven one, there is a high likelihood that they will survive versus the tested stress in the real world.

14 SOLAR ENERGY↗

A Robust State Estimation Framework Considering Measurement Correlations and Imperfect Synchronization

This paper develops a robust power system state estimation framework with the consideration of measurement correlations and imperfect synchronization. In the framework, correlations of SCADA and PMU measurements are calculated separately through unscented transformation and a vector autoregression (VAR) model. In particular, PMU measurements during the waiting period of two SCADA measurement scans are buffered to develop the VAR model. The latter takes into account the temporal and spatial correlations of PMU measurements and provides redundant measurements to suppress bad data and mitigate imperfect synchronization. In case of synchronization issues between SCADA and PMU measurements, either forecasted PMU measurements or prior SCADA measurements from the last estimation run are leveraged to restore system observability. After that, robust generalized maximum-likelihood (GM)-estimator is developed to integrate full measurement correlations and handle bad data in SCADA and PMU measurements. Thanks to the robustness and flexibility of the proposed robust estimator, bad data as well as imperfect measurement synchronization are mitigated, yielding high statistical efficiency. Comprehensive comparison results with other alternatives under various conditions demonstrate the benefits of the proposed framework.

Zhao, Junbo (ORCID:0000000284989666)↗

Joint Sparsity Pattern Recovery With 1-b Compressive Sensing in Distributed Sensor Networks

In this paper, we study the problem of joint sparse support recovery with 1-b quantized compressive measurements in a distributed sensor network. Multiple nodes in the network are assumed to observe sparse signals having the same but unknown sparse support. Each node quantizes its measurement vector element-wise to 1 b. First, we consider that all the quantized measurements are available at a central fusion center. We derive performance bounds for sparsity pattern recovery using 1-bit quantized measurements from multiple sensors when the maximum likelihood decoder is employed. We further develop two computationally tractable algorithms for joint sparse support recovery in the centralized setting. One algorithm minimizes a cost function defined as the sum of the likelihood function and the l-∝ quasi-norm, while the other algorithm extends the binary iterative hard thresholding algorithm to the multiple measurement vector case. Second, we consider a decentralized setting where each node transmits 1-b measurements to its one-hop neighbors. The basic idea behind the algorithms developed in the decentralized setting is to embed collaboration among nodes and fusion strategies. We show that even with noisy 1-b compressed measurements, joint support recovery can be carried out accurately in both centralized and decentralized settings. Additionally, we further show that the performance of the proposed 1-bit compressive sensing-based algorithms is very close to that of their real-valued counterparts except when the signal-to-noise ratio is very small.

97 MATHEMATICS AND COMPUTING↗

Compound fire-drought regimes promote ecosystem transitions in Mediterranean ecosystems

Understanding ecosystem responses to compound disturbance regimes and the influence of specific sequences of events in determining ecosystem shifts remains a challenge. We use a modelling framework for Mediterranean-type ecosystems to assess the effects of fire–drought interactions on long-term vegetation dynamics and to identify disturbance-driven changes in trait composition (tree seeder vs. tree resprouter prevalence) and ecosystem state (forest vs. non-forest). Changes in tree seeder and the tree resprouter dominance show nonlinear, threshold-type trends over gradients of increasing compound disturbance frequency. Vegetation composition thresholds mostly occur in a narrow range of the compound fire–drought disturbance space. Additionally, trait compositional switches and the likelihood of sudden changes in ecosystem state are promoted by fire-drought interactions. Distinct sequences of disturbance events cause vegetation transitions, disrupting ecosystem resilience, even under moderate recurrence of individual disturbances. An extreme drought year followed by one or two large fire events promotes shifts from resprouter- to seeder dominance. Contrastingly, a large crown fire followed by an extreme drought promotes changes from seeder to resprouter dominance. This disturbance sequence is also a mechanism strong enough to trigger sudden shifts in ecosystem state (from forest to non-forest). Synthesis. Thresholds of change in vegetation composition occur over a narrow range of the modelled gradients of compound fire-drought recurrence, and the loss of ecosystem resilience is contingent on particular sequences of disturbance events. Overall, our findings highlight that disturbance interactions define the relative location of tipping points in ecosystem state, and that effects and feedbacks of compound disturbance regimes increase the long-term likelihood of sudden ecosystem shifts and, therefore, uncertainty in predicting vegetation state.

54 ENVIRONMENTAL SCIENCES↗

A Selective Overview and Comparison of Robust Mixture Regression Estimators

Summary Mixture regression models have been widely used in business, marketing and social sciences to model mixed regression relationships arising from a clustered and thus heterogeneous population. The unknown mixture regression parameters are usually estimated by maximum likelihood estimators using the expectation–maximisation algorithm based on the normality assumption of component error density. However, it is well known that the normality‐based maximum likelihood estimation is very sensitive to outliers or heavy‐tailed error distributions. This paper aims to give a selective overview of the recently proposed robust mixture regression methods and compare their performance using simulation studies.

Yu, Chun↗

A Bayesian Approach for Extracting Fluorescence Lifetimes from Sparse Data Sets and Its Significance for Imaging Experiments

The measurement of fluorescence lifetimes, especially in small sample volumes, presents the dual challenge of probing a small number of fluorophores and fitting the concomitant sparse data set to the appropriate excited-state decay function. A common method of analysis, such as the maximum likelihood (ML) technique, assumes a uniform probability distribution of the parameters describing the fluorescence decay function. An improvement is thus suggested by implementing a suitable nonuniform distribution, as is provided by a Bayesian framework, where the distribution of parameters is obtained from both their prior knowledge and the evidence-based likelihood of an event for a given set of parameters. We have also considered the Dirichlet prior distribution, whose mathematical form enables analytical solutions of the fitting parameters to be rapidly obtained. If Gaussian and exponential prior distributions are judiciously chosen, they reproduce the experimental target lifetime to within 20% with as few as 20 total photon counts for the data set, as does the Dirichlet prior distribution. But because of the analytical solutions afforded by the Dirichlet prior distribution, it is proposed to employ a Dirichlet prior to search parameter space rapidly to provide, if necessary, appropriate parameters for subsequent employment of a Gaussian or exponential prior distribution

97 MATHEMATICS AND COMPUTING↗

Time-resolved in vivo luminescence dosimetry for online error detection in pulsed dose-rate brachytherapy

Purpose: The purpose of this study is to present and evaluate a dose-verification protocol for pulsed dose-rate (PDR) brachytherapy based on in vivo time-resolved (1 s time resolution) fiber-coupled luminescence dosimetry. Methods: Five cervix cancer patients undergoing PDR brachytherapy (Varian GammaMed Plus with {sup 192}Ir) were monitored. The treatments comprised from 10 to 50 pulses (1 pulse/h) delivered by intracavitary/interstitial applicators (tandem-ring systems and/or needles). For each patient, one or two dosimetry probes were placed directly in or close to the tumor region using stainless steel or titanium needles. Each dosimeter probe consisted of a small aluminum oxide crystal attached to an optical fiber cable (1 mm outer diameter) that could guide radioluminescence (RL) and optically stimulated luminescence (OSL) from the crystal to special readout instrumentation. Positioning uncertainty and hypothetical dose-delivery errors (interchanged guide tubes or applicator movements from {+-}5 to {+-}15 mm) were simulated in software in order to assess the ability of the system to detect errors. Results: For three of the patients, the authors found no significant differences (P>0.01) for comparisons between in vivo measurements and calculated reference values at the level of dose per dwell position, dose per applicator, or total dose per pulse. The standard deviations of the dose per pulse were less than 3%, indicating a stable dose delivery and a highly stable geometry of applicators and dosimeter probes during the treatments. For the two other patients, the authors noted significant deviations for three individual pulses and for one dosimeter probe. These deviations could have been due to applicator movement during the treatment and one incorrectly positioned dosimeter probe, respectively. Computer simulations showed that the likelihood of detecting a pair of interchanged guide tubes increased by a factor of 10 or more for the considered patients when going from integrating to time-resolved dose verification. The likelihood of detecting a {+-}15 mm displacement error increased by a factor of 1.5 or more. Conclusions: In vivo fiber-coupled RL/OSL dosimetry based on detectors placed in standard brachytherapy needles was demonstrated. The time-resolved dose-rate measurements were found to provide a good way to visualize the progression and stability of PDR brachytherapy dose delivery, and time-resolved dose-rate measurements provided an increased sensitivity for detection of dose-delivery errors compared with time-integrated dosimetry.

61 RADIATION PROTECTION AND DOSIMETRY↗

A modified hypoxia-based TCP model to investigate the clinical outcome of stereotactic hypofractionated regimes for early stage non-small-cell lung cancer (NSCLC)

Purpose: Stereotactic body radiotherapy (SBRT) has been applied to lung tumors at different stages and sizes with good local tumor control (LC) rates. The linear quadratic model (LQM), in its basic formulation, does not seem to be appropriate to describe the response to radiotherapy for clinical trials, based on a few fractions. Thus, the main aim of this work was to develop a model, which takes into account the hypoxic cells and their reoxygenation. Methods: A parameter named B has been introduced in a modified tumor control probability (TCP) from LQM and linear-quadratic-linear model (LQLM), and represents the fraction of hypoxic cells that survive and become oxygenated after each irradiation. Based on published trials evaluating LC at 3 yr (LC3), values of B were obtained by maximum likelihood minimization between predicted TCP and clinical LC3. Two oxygen enhancement ratio (OER) parameter sets (1 and 2) from literature have been adopted to calculate the B-factors. Initial hypoxic cell fractions ({eta}{sub h}) from 0.05 to 0.50 were assumed. Log-likelihood (L) and Akaike information criterion (AIC) were determined in an independent clinical validation dataset. Results: The B-values of modified TCPs spanned the whole interval from 0 to 1, depending on the fractionation scheme (number of fractions and dose/fraction), showing a maximum (close to 1) at doses/fraction of 8-12 Gy. The B-values calculated using the OER parameter set 1 exhibited a smoother falloff than set 2. An analytical expression was derived to describe the B-value's dependence on the fractionation scheme. The R{sup 2}-adjusted values varied from 0.63 to 0.67 for LQ models and OER set 1 and from 0.75 to 0.78 for LQ model and OER set 2. Lower values of R{sup 2}-adjusted were found for LQLM and both OER sets. L and AIC, calculated using a fraction of {eta}{sub h} = 0.15 and the B-value from the authors analytical expression were higher than for other {eta}{sub h}-values, irrespective of model or OER set. Conclusions: The authors model allows to predict the clinical outcome associated with SBRT treatment, taking into account both direct killing and indirect vasculature or stromal damage.

61 RADIATION PROTECTION AND DOSIMETRY↗

Predicting diagnostic error in radiology via eye-tracking and image analytics: Preliminary investigation in mammography

Purpose: The primary aim of the present study was to test the feasibility of predicting diagnostic errors in mammography by merging radiologists’ gaze behavior and image characteristics. A secondary aim was to investigate group-based and personalized predictive models for radiologists of variable experience levels.Methods: The study was performed for the clinical task of assessing the likelihood of malignancy of mammographic masses. Eye-tracking data and diagnostic decisions for 40 cases were acquired from four Radiology residents and two breast imaging experts as part of an IRB-approved pilot study. Gaze behavior features were extracted from the eye-tracking data. Computer-generated and BIRADS images features were extracted from the images. Finally, machine learning algorithms were used to merge gaze and image features for predicting human error. Feature selection was thoroughly explored to determine the relative contribution of the various features. Group-based and personalized user modeling was also investigated.Results: Machine learning can be used to predict diagnostic error by merging gaze behavior characteristics from the radiologist and textural characteristics from the image under review. Leveraging data collected from multiple readers produced a reasonable group model [area under the ROC curve (AUC) = 0.792 ± 0.030]. Personalized user modeling was far more accurate for the more experienced readers (AUC = 0.837 ± 0.029) than for the less experienced ones (AUC = 0.667 ± 0.099). The best performing group-based and personalized predictive models involved combinations of both gaze and image features.Conclusions: Diagnostic errors in mammography can be predicted to a good extent by leveraging the radiologists’ gaze behavior and image content.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Parametrization of textural patterns in {sup 123}I-ioflupane imaging for the automatic detection of Parkinsonism

Purpose: A novel approach to a computer aided diagnosis system for the Parkinson's disease is proposed. This tool is intended as a supporting tool for physicians, based on fully automated methods that lead to the classification of{sup 123}I-ioflupane SPECT images. Methods: {sup 123}I-ioflupane images from three different databases are used to train the system. The images are intensity and spatially normalized, then subimages are extracted and a 3D gray-level co-occurrence matrix is computed over these subimages, allowing the characterization of the texture using Haralick texture features. Finally, different discrimination estimation methods are used to select a feature vector that can be used to train and test the classifier. Results: Using the leave-one-out cross-validation technique over these three databases, the system achieves results up to a 97.4% of accuracy, and 99.1% of sensitivity, with positive likelihood ratios over 27. Conclusions: The system presents a robust feature extraction method that helps physicians in the diagnosis task by providing objective, operator-independent textural information about{sup 123}I-ioflupane images, commonly used in the diagnosis of the Parkinson's disease. Textural features computation has been optimized by using a subimage selection algorithm, and the discrimination estimation methods used here makes the system feature-independent, allowing us to extend it to other databases and diseases.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Dual-energy cone-beam CT with a flat-panel detector: Effect of reconstruction algorithm on material classification

Purpose: Cone-beam CT (CBCT) with a flat-panel detector (FPD) is finding application in areas such as breast and musculoskeletal imaging, where dual-energy (DE) capabilities offer potential benefit. The authors investigate the accuracy of material classification in DE CBCT using filtered backprojection (FBP) and penalized likelihood (PL) reconstruction and optimize contrast-enhanced DE CBCT of the joints as a function of dose, material concentration, and detail size. Methods: Phantoms consisting of a 15 cm diameter water cylinder with solid calcium inserts (50–200 mg/ml, 3–28.4 mm diameter) and solid iodine inserts (2–10 mg/ml, 3–28.4 mm diameter), as well as a cadaveric knee with intra-articular injection of iodine were imaged on a CBCT bench with a Varian 4343 FPD. The low energy (LE) beam was 70 kVp (+0.2 mm Cu), and the high energy (HE) beam was 120 kVp (+0.2 mm Cu, +0.5 mm Ag). Total dose (LE+HE) was varied from 3.1 to 15.6 mGy with equal dose allocation. Image-based DE classification involved a nearest distance classifier in the space of LE versus HE attenuation values. Recognizing the differences in noise between LE and HE beams, the LE and HE data were differentially filtered (in FBP) or regularized (in PL). Both a quadratic (PLQ) and a total-variation penalty (PLTV) were investigated for PL. The performance of DE CBCT material discrimination was quantified in terms of voxelwise specificity, sensitivity, and accuracy. Results: Noise in the HE image was primarily responsible for classification errors within the contrast inserts, whereas noise in the LE image mainly influenced classification in the surrounding water. For inserts of diameter 28.4 mm, DE CBCT reconstructions were optimized to maximize the total combined accuracy across the range of calcium and iodine concentrations, yielding values of ∼88% for FBP and PLQ, and ∼95% for PLTV at 3.1 mGy total dose, increasing to ∼95% for FBP and PLQ, and ∼98% for PLTV at 15.6 mGy total dose. For a fixed iodine concentration of 5 mg/ml and reconstructions maximizing overall accuracy across the range of insert diameters, the minimum diameter classified with accuracy >80% was ∼15 mm for FBP and PLQ and ∼10 mm for PLTV, improving to ∼7 mm for FBP and PLQ and ∼3 mm for PLTV at 15.6 mGy. The results indicate similar performance for FBP and PLQ and showed improved classification accuracy with edge-preserving PLTV. A slight preference for increased smoothing of the HE data was found. DE CBCT discrimination of iodine and bone in the knee was demonstrated with FBP and PLTV at 6.2 mGy total dose. Conclusions: For iodine concentrations >5 mg/ml and detail size ∼20 mm, material classification accuracy of >90% was achieved in DE CBCT with both FBP and PL at total doses <10 mGy. Optimal performance was attained by selection of reconstruction parameters based on the differences in noise between HE and LE data, typically favoring stronger smoothing of the HE data, and by using penalties matched to the imaging task (e.g., edge-preserving PLTV in areas of uniform enhancement)

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Predicting error in detecting mammographic masses among radiology trainees using statistical models based on BI-RADS features

Purpose: The purpose of this study is to explore Breast Imaging-Reporting and Data System (BI-RADS) features as predictors of individual errors made by trainees when detecting masses in mammograms. Methods: Ten radiology trainees and three expert breast imagers reviewed 100 mammograms comprised of bilateral medial lateral oblique and craniocaudal views on a research workstation. The cases consisted of normal and biopsy proven benign and malignant masses. For cases with actionable abnormalities, the experts recorded breast (density and axillary lymph nodes) and mass (shape, margin, and density) features according to the BI-RADS lexicon, as well as the abnormality location (depth and clock face). For each trainee, a user-specific multivariate model was constructed to predict the trainee's likelihood of error based on BI-RADS features. The performance of the models was assessed using area under the receive operating characteristic curves (AUC). Results: Despite the variability in errors between different trainees, the individual models were able to predict the likelihood of error for the trainees with a mean AUC of 0.611 (range: 0.502–0.739, 95% Confidence Interval: 0.543–0.680,p < 0.002). Conclusions: Patterns in detection errors for mammographic masses made by radiology trainees can be modeled using BI-RADS features. These findings may have potential implications for the development of future educational materials that are personalized to individual trainees.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Cavity contour segmentation in chest radiographs using supervised learning and dynamic programming

Purpose: Efficacy of tuberculosis (TB) treatment is often monitored using chest radiography. Monitoring size of cavities in pulmonary tuberculosis is important as the size predicts severity of the disease and its persistence under therapy predicts relapse. The authors present a method for automatic cavity segmentation in chest radiographs. Methods: A two stage method is proposed to segment the cavity borders, given a user defined seed point close to the center of the cavity. First, a supervised learning approach is employed to train a pixel classifier using texture and radial features to identify the border pixels of the cavity. A likelihood value of belonging to the cavity border is assigned to each pixel by the classifier. The authors experimented with four different classifiers:k-nearest neighbor (kNN), linear discriminant analysis (LDA), GentleBoost (GB), and random forest (RF). Next, the constructed likelihood map was used as an input cost image in the polar transformed image space for dynamic programming to trace the optimal maximum cost path. This constructed path corresponds to the segmented cavity contour in image space. Results: The method was evaluated on 100 chest radiographs (CXRs) containing 126 cavities. The reference segmentation was manually delineated by an experienced chest radiologist. An independent observer (a chest radiologist) also delineated all cavities to estimate interobserver variability. Jaccard overlap measure Ω was computed between the reference segmentation and the automatic segmentation; and between the reference segmentation and the independent observer's segmentation for all cavities. A median overlap Ω of 0.81 (0.76 ± 0.16), and 0.85 (0.82 ± 0.11) was achieved between the reference segmentation and the automatic segmentation, and between the segmentations by the two radiologists, respectively. The best reported mean contour distance and Hausdorff distance between the reference and the automatic segmentation were, respectively, 2.48 ± 2.19 and 8.32 ± 5.66 mm, whereas these distances were 1.66 ± 1.29 and 5.75 ± 4.88 mm between the segmentations by the reference reader and the independent observer, respectively. The automatic segmentations were also visually assessed by two trained CXR readers as “excellent,” “adequate,” or “insufficient.” The readers had good agreement in assessing the cavity outlines and 84% of the segmentations were rated as “excellent” or “adequate” by both readers. Conclusions: The proposed cavity segmentation technique produced results with a good degree of overlap with manual expert segmentations. The evaluation measures demonstrated that the results approached the results of the experienced chest radiologists, in terms of overlap measure and contour distance measures. Automatic cavity segmentation can be employed in TB clinics for treatment monitoring, especially in resource limited settings where radiologists are not available.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Sinogram restoration in computed tomography with an edge-preserving penalty

Purpose: With the goal of producing a less computationally intensive alternative to fully iterative penalized-likelihood image reconstruction, our group has explored the use of penalized-likelihood sinogram restoration for transmission tomography. Previously, we have exclusively used a quadratic penalty in our restoration objective function. However, a quadratic penalty does not excel at preserving edges while reducing noise. Here, we derive a restoration update equation for nonquadratic penalties. Additionally, we perform a feasibility study to extend our sinogram restoration method to a helical cone-beam geometry and clinical data. Methods: A restoration update equation for nonquadratic penalties is derived using separable parabolic surrogates (SPS). A method for calculating sinogram degradation coefficients for a helical cone-beam geometry is proposed. Using simulated data, sinogram restorations are performed using both a quadratic penalty and the edge-preserving Huber penalty. After sinogram restoration, Fourier-based analytical methods are used to obtain reconstructions, and resolution-noise trade-offs are investigated. For the fan-beam geometry, a comparison is made to image-domain SPS reconstruction using the Huber penalty. The effects of varying object size and contrast are also investigated. For the helical cone-beam geometry, we investigate the effect of helical pitch (axial movement/rotation). Huber-penalty sinogram restoration is performed on 3D clinical data, and the reconstructed images are compared to those generated with no restoration. Results: We find that by applying the edge-preserving Huber penalty to our sinogram restoration methods, the reconstructed image has a better resolution-noise relationship than an image produced using a quadratic penalty in the sinogram restoration. However, we find that this relatively straightforward approach to edge preservation in the sinogram domain is affected by the physical size of imaged objects in addition to the contrast across the edge. This presents some disadvantages of this method relative to image-domain edge-preserving methods, although the computational burden of the sinogram-domain approach is much lower. For a helical cone-beam geometry, we found applying sinogram restoration in 3D was reasonable and that pitch did not make a significant difference in the general effect of sinogram restoration. The application of Huber-penalty sinogram restoration to clinical data resulted in a reconstruction with less noise while retaining resolution. Conclusions: Sinogram restoration with the Huber penalty is able to provide better resolution-noise performance than restoration with a quadratic penalty. Additionally, sinogram restoration with the Huber penalty is feasible for helical cone-beam CT and can be applied to clinical data.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗