Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “LABELLING”

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 1,441 records · Page 80

Interactive Visual Study of Multiple Attributes Learning Model of X-Ray Scattering Images

Existing interactive visualization tools for deep learning are mostly applied to the training, debugging, and refinement of neural network models working on natural images. However, visual analytics tools are lacking for the specific application of x-ray image classification with multiple structural attributes. In this paper, we present an interactive system for domain scientists to visually study the multiple attributes learning models applied to x-ray scattering images. It allows domain scientists to interactively explore this important type of scientific images in embedded spaces that are defined on the model prediction output, the actual labels, and the discovered feature space of neural networks. Users are allowed to flexibly select instance images, their clusters, and compare them regarding the specified visual representation of attributes. The exploration is guided by the manifestation of model performance related to mutual relationships among attributes, which often affect the learning accuracy and effectiveness. The system thus supports domain scientists to improve the training dataset and model, find questionable attributes labels, and identify outlier images or spurious data clusters. Case studies and scientists feedback demonstrate its functionalities and usefulness.

97 MATHEMATICS AND COMPUTING↗

Deep learning multiphysics network for imaging CO 2 saturation and estimating uncertainty in geological carbon storage

Multiphysics inversion exploits different types of geophysical data that often complement each other and aims to improve overall imaging resolution and reduce uncertainties in geophysical interpretation. Despite the advantages, traditional multiphysics inversion is challenging because it requires a large amount of computational time and intensive human interactions for preprocessing data and finding trade-off parameters. These issues make it nearly impossible for traditional multiphysics inversion to be applied as a real-time monitoring tool for geological carbon storage. In this paper, we present a deep learning (DL) multiphysics network for imaging CO 2 saturation in real time. The multiphysics network consists of three encoders for analysing seismic, electromagnetic and gravity data and shares one decoder for combining imaging capabilities of the different geophysical data for better predicting CO 2 saturation. The network is trained on pairs of CO 2 label models and multiphysics data so that it can directly image CO 2 saturation. Here we use the bootstrap aggregating method to enhance the imaging accuracy and estimate uncertainties associated with CO 2 saturation images. Using realistic CO 2 label models and multiphysics data derived from the Kimberlina CO 2 storage model, we evaluate the performance of the deep learning multiphysics network and compare its imaging results to those from the deep learning single-physics networks. Our modelling experiments show that the deep learning multiphysics network for seismic, electromagnetic, and gravity data not only improves the imaging accuracy but also reduces uncertainties associated with CO 2 saturation images. Our results also suggest that the deep learning multiphysics network for the non-seismic data (i.e., electromagnetic and gravity) can be used as an effective low-cost monitoring tool in between regular seismic monitoring.

58 GEOSCIENCES↗

Interactions among roots, mycorrhizas and free-living microbial communities differentially impact soil carbon processes

Plant roots, their associated microbial community and free-living soil microbes interact to regulate the movement of carbon from the soil to the atmosphere, one of the most important and least understood fluxes of terrestrial carbon. Our inadequate understanding of how plant-microbial interactions alter soil carbon decomposition may lead to poor model predictions of terrestrial carbon feedbacks to the atmosphere. Roots, mycorrhizal fungi and free-living soil microbes can alter soil carbon decomposition through exudation of carbon into soil. Exudates of simple carbon compounds can increase microbial activity because microbes are typically carbon limited. When both roots and mycorrhizal fungi are present in the soil, they may additively increase carbon decomposition. However, when mycorrhizas are isolated from roots, they may limit soil carbon decomposition by competing with free-living decomposers for resources. We manipulated the access of roots and mycorrhizal fungi to soil insitu in a temperate mixed deciduous forest. We added 13 C-labelled substrate to trace metabolized carbon in respiration and measured carbon-degrading microbial extracellular enzyme activity and soil carbon pools. We used our data in a mechanistic soil carbon decomposition model to simulate and compare the effects of root and mycorrhizal fungal presence on soil carbon dynamics over longer time periods. Contrary to what we predicted, root and mycorrhizal biomass did not interact to additively increase microbial activity and soil carbon degradation. The metabolism of 13 C-labelled starch was highest when root biomass was high and mycorrhizal biomass was low. These results suggest that mycorrhizas may negatively interact with the free-living microbial community to influence soil carbon dynamics, a hypothesis supported by our enzyme results. Our steady-state model simulations suggested that root presence increased mineral-associated and particulate organic carbon pools, while mycorrhizal fungal presence had a greater influence on particulate than mineral-associated organic carbon pools.Synthesis. Our results suggest that the activity of enzymes involved in organic matter decomposition was contingent upon root-mycorrhizal-microbial interactions. Using our experimental data in a decomposition simulation model, we show that root-mycorrhizal-microbial interactions may have longer-term legacy effects on soil carbon sequestration. Lastly, our study suggests that roots stimulate microbial activity in the short term, but contribute to soil carbon storage over longer periods of time.

54 ENVIRONMENTAL SCIENCES↗

In situ visualization of newly synthesized proteins in environmental microbes using amino acid tagging and click chemistry

Summary Here we describe the application of a new click chemistry method for fluorescent tracking of protein synthesis in individual microorganisms within environmental samples. This technique, termed bioorthogonal non‐canonical amino acid tagging ( BONCAT ), is based on the in vivo incorporation of the non‐canonical amino acid L ‐azidohomoalanine ( AHA ), a surrogate for l ‐methionine, followed by fluorescent labelling of AHA ‐containing cellular proteins by azide‐alkyne click chemistry. BONCAT was evaluated with a range of phylogenetically and physiologically diverse archaeal and bacterial pure cultures and enrichments, and used to visualize translationally active cells within complex environmental samples including an oral biofilm, freshwater and anoxic sediment. We also developed combined assays that couple BONCAT with ribosomal RNA (rRNA)‐targeted fluorescence in situ hybridization ( FISH ), enabling a direct link between taxonomic identity and translational activity. Using a methanotrophic enrichment culture incubated under different conditions, we demonstrate the potential of BONCAT‐FISH to study microbial physiology in situ . A direct comparison of anabolic activity using BONCAT and stable isotope labelling by nano‐scale secondary ion mass spectrometry ( 15 NH 3 assimilation) for individual cells within a sediment‐sourced enrichment culture showed concordance between AHA ‐positive cells and 15 N enrichment. BONCAT‐FISH offers a fast, inexpensive and straightforward fluorescence microscopy method for studying the in situ activity of environmental microbes on a single‐cell level.

59 BASIC BIOLOGICAL SCIENCES↗

Phenazine‐1‐carboxylic acid and soil moisture influence biofilm development and turnover of rhizobacterial biomass on wheat root surfaces

Summary Phenazine‐1‐carboxylic acid (PCA) is produced by rhizobacteria in dryland but not in irrigated wheat fields of the Pacific Northwest, USA. PCA promotes biofilm development in bacterial cultures and bacterial colonization of wheat rhizospheres. However, its impact upon biofilm development has not been demonstrated in the rhizosphere, where biofilms influence terrestrial carbon and nitrogen cycles with ramifications for crop and soil health. Furthermore, the relationships between soil moisture and the rates of PCA biosynthesis and degradation have not been established. In this study, expression of PCA biosynthesis genes was upregulated relative to background transcription, and persistence of PCA was slightly decreased in dryland relative to irrigated wheat rhizospheres. Biofilms in dryland rhizospheres inoculated with the PCA‐producing (PCA + ) strain Pseudomonas synxantha 2‐79RN 10 were more robust than those in rhizospheres inoculated with an isogenic PCA‐deficient (PCA ‐ ) mutant strain. This trend was reversed in irrigated rhizospheres. In dryland PCA + rhizospheres, the turnover of 15 N‐labelled rhizobacterial biomass was slower than in the PCA ‐ and irrigated PCA + treatments, and incorporation of bacterial 15 N into root cell walls was observed in multiple treatments. These results indicate that PCA promotes biofilm development in dryland rhizospheres, and likely influences crop nutrition and soil health in dryland wheat fields.

LeTourneau, Melissa K.↗

Lignin decomposition is sustained under fluctuating redox conditions in humid tropical forest soils

Abstract Lignin mineralization represents a critical flux in the terrestrial carbon (C) cycle, yet little is known about mechanisms and environmental factors controlling lignin breakdown in mineral soils. Hypoxia is thought to suppress lignin decomposition, yet potential effects of oxygen (O 2 ) variability in surface soils have not been explored. Here, we tested the impact of redox fluctuations on lignin breakdown in humid tropical forest soils during ten‐week laboratory incubations. We used synthetic lignins labeled with 13 C in either of two positions (aromatic methoxyl or propyl side chain C β ) to provide highly sensitive and specific measures of lignin mineralization seldom employed in soils. Four‐day redox fluctuations increased the percent contribution of methoxyl C to soil respiration relative to static aerobic conditions, and cumulative methoxyl‐C mineralization was statistically equivalent under static aerobic and fluctuating redox conditions despite lower soil respiration in the latter treatment. Contributions of the less labile lignin C β to soil respiration were equivalent in the static aerobic and fluctuating redox treatments during periods of O 2 exposure, and tended to decline during periods of O 2 limitation, resulting in lower cumulative C β mineralization in the fluctuating treatment relative to the static aerobic treatment. However, cumulative mineralization of both the C β ‐ and methoxyl‐labeled lignins nearly doubled in the fluctuating treatment relative to the static aerobic treatment when total lignin mineralization was normalized to total O 2 exposure. Oxygen fluctuations are thought to be suboptimal for canonical lignin‐degrading microorganisms. However, O 2 fluctuations drove substantial Fe reduction and oxidation, and reactive oxygen species generated during abiotic Fe oxidation might explain the elevated contribution of lignin to C mineralization. Iron redox cycling provides a potential mechanism for lignin depletion in soil organic matter. Couplings between soil moisture, redox fluctuations, and lignin breakdown provide a potential link between climate variability and the biochemical composition of soil organic matter.

Hall, Steven J.↗

Fine-root mortality rates in a temperate forest: Estimates using radiocarbon data and numerical modeling

We used an inadvertent whole-ecosystem {sup 14}C label at a temperate forest in Oak Ridge, Tennessee, USA to develop a model (Radix1.0) of fine-root dynamics. Radix simulates two live-root pools, two dead-root pools, non-normally distributed root mortality turnover times, a stored carbon (C) pool, and seasonal growth and respiration patterns. We applied Radix to analyze measurements from two root size classes (< 0.5 and 0.5-2.0 mm diameter) and three soil-depth increments (O horizon, 0-15 cm and 30-60 cm). Predicted live-root turnover times were < 1 yr and 10 yr for short- and long-lived pools, respectively. Dead-root pools had decomposition turnover times of 2 yr and 10 yr. Realistic characterization of C flows through fine roots requires a model with two live fine-root populations, two dead fine-root pools, and root respiration. These are the first fine-root turnover time estimates that take into account respiration, storage, seasonal growth patterns, and non-normal turnover time distributions. The presence of a root population with decadal turnover times implies a lower amount of belowground net primary production used to grow fine-root tissue than is currently predicted by models with a single annual turnover pool.

54 ENVIRONMENTAL SCIENCES↗

RANKL, Osteopontin, and Osteoclast Homeostasis in a Hyper-Occlusion Mouse Model

The biological mechanisms that maintain the position of teeth in their sockets establish a dynamic equilibrium between bone resorption and apposition. In order to reveal some of the dynamics involved in the tissue responses towards occlusal forces on periodontal ligament (PDL) and alveolar bone homeostasis, we developed the first mouse model of hyperocclusion. Swiss-Webster mice were kept in hyperocclusion for 0, 3, 6, and 9 d. Morphological and histological changes in the periodontium were assessed using micro-computed tomography (micro-CT) and ground sections with fluorescent detection of vital dye labels. Sections were stained for tartrate-resistant acid phosphatase, and the expression of receptor activator of nuclear factor-{kappa}B ligand (RANKL) and osteopontin (OPN) was analyzed by immunohistochemistry and real-time polymerase chain reaction (PCR). Traumatic occlusion resulted in enamel surface abrasion, inhibition of alveolar bone apposition, significant formation of osteoclasts at 3, 6 and 9 d, and upregulation of OPN and RANKL. Data from this study suggest that both OPN and RANKL contribute to the stimulation of bone resorption in the hyperocclusive state. In addition, we propose that the inhibition of alveolar bone apposition by occlusal forces is an important mechanism for the control of occlusal height that might work in synergy with RANKL-induced bone resorption to maintain normal occlusion.

59 BASIC BIOLOGICAL SCIENCES↗

Routes to roots: direct evidence of water transport by arbuscular mycorrhizal fungi to host plants

Summary Arbuscular mycorrhizal fungi (AMF) can help mitigate plant responses to water stress, but it is unclear whether AMF do so by indirect mechanisms, direct water transport to roots, or a combination of the two. Here, we investigated if and how the AMF Rhizophagus intraradices transported water to the host plant Avena barbata , wild oat. We used two‐compartment microcosms, isotopically labeled water, and a fluorescent dye to directly track and quantify water transport by AMF across an air gap to host plants. Plants grown with AMF that had access to a physically separated compartment containing 18 O‐labeled water transpired almost twice as much as plants with AMF excluded from that compartment. Using an isotopic mixing model, we estimated that water transported by AMF across the air gap accounted for 34.6% of the water transpired by host plants. In addition, a fluorescent dye indicated that hyphae were able to transport some water via an extracytoplasmic pathway. Our study provides direct evidence that AMF can act as extensions of the root system along the soil–plant–air continuum of water movement, with plant transpiration driving water flow along hyphae outside of the hyphal cell membrane.

59 BASIC BIOLOGICAL SCIENCES↗

Effect of image resolution on automated classification of chest X-rays

Deep learning (DL) models have received much attention lately for their ability to achieve expert-level performance on the accurate automated analysis of chest X-rays (CXRs). Recently available public CXR datasets include high resolution images, but state-of-the-art models are trained on reduced size images due to limitations on graphics processing unit memory and training time. As computing hardware continues to advance, it has become feasible to train deep convolutional neural networks on high-resolution images without sacrificing detail by downscaling. This study examines the effect of increased resolution on CXR classification performance. We used the publicly available MIMIC-CXR-JPG dataset, comprising 377,110 high resolution CXR images for this study. We applied image downscaling from native resolution to 2048 × 2048 pixels, 1024 × 1024 pixels, 512 × 512 pixels, and 256 × 256 pixels and then we used the DenseNet121 and EfficientNet-B4 DL models to evaluate clinical task performance using these four downscaled image resolutions. We find that while some clinical findings are more reliably labeled using high resolutions, many other findings are actually labeled better using downscaled inputs. We qualitatively verify that tasks requiring a large receptive field are better suited to downscaled low resolution input images, by inspecting effective receptive fields and class activation maps of trained models. Lastly, we show that stacking an ensemble across resolutions outperforms each individual learner at all input resolutions while providing interpretable scale weights, indicating that diverse information is extracted across resolutions.

47 OTHER INSTRUMENTATION↗

One-shot gas detection with transformer paired neural networks in Mako collected longwave infrared hyperspectral imagery

To date, careful data treatment workflows and statistical detectors are used to perform hyperspectral image (HSI) detection of any gas contained in a spectral library, which is often expanded with physics models to incorporate different spectral characteristics. In general, surrounding evidence or known gas-release parameters are used to provide confidence in or confirm detection capability, respectively. This makes quantifying detection performance difficult as it is nearly impossible to develop an absolute ground truth for gas target pixel presence in collected HSI. Consequently, developing and comparing new detection methods, especially machine learning (ML)-based methods, is susceptible to subjectivity in derived detection map quality. Here, in this work, we demonstrate the first use of transformer-based paired neural networks (PNNs) for one-shot gas target detection for multiple gases while providing quantitative classification and detection metrics for their use on labeled data. Terabytes of training data are generated from a database of long-wave infrared HSI obtained from historical Mako sensor campaigns over Los Angeles. By incorporating labels, singular signature representations, and a model development pipeline, we can tune and select PNNs to detect multiple gas targets that are not seen in training on a quantitative basis. We additionally assess our test set detections using interpretability techniques widely employed with ML-based predictors, but less common with detection methods relying on learned latent spaces.

Hyperspectral imaging↗

Land cover classification in multispectral imagery using clustering of sparse approximations over learned feature dictionaries

We present results from an ongoing effort to extend neuromimetic machine vision algorithms to multispectral data using adaptive signal processing combined with compressive sensing and machine learning techniques. Our goal is to develop a robust classification methodology that will allow for automated discretization of the landscape into distinct units based on attributes such as vegetation, surface hydrological properties, and topographic/geomorphic characteristics. We use a Hebbian learning rule to build spectral-textural dictionaries that are tailored for classification. We learn our dictionaries from millions of overlapping multispectral image patches and then use a pursuit search to generate classification features. Land cover labels are automatically generated using unsupervised clustering of sparse approximations (CoSA). We demonstrate our method on multispectral WorldView-2 data from a coastal plain ecosystem in Barrow, Alaska. We explore learning from both raw multispectral imagery and normalized band difference indices. We explore a quantitative metric to evaluate the spectral properties of the clusters in order to potentially aid in assigning land cover categories to the cluster labels. In this study, our results suggest CoSA is a promising approach to unsupervised land cover classification in high-resolution satellite imagery.

42 ENGINEERING↗

Design of a dynamic biofilm imaging cell for white-light interferometric microscopy

In microbiology research there is a strong need for next generation imaging and sensing instrumentation that will enable minimally invasive and label-free investigation of soft, hydrated structures such as in bacterial biofilms. White light interferometry (WLI) can provide high resolution images of surface topology without the use of fluorescent labels but is not typically used to image biofilms because there is insufficient refractive index contrast to induce reflection from the biofilm’s interface. The soft structure and water-like bulk properties of hydrated biofilms make them difficult to characterize in situ , especially in a non-destructive manner. In this report, we build on our prior description of static biofilm imaging and describe the design of a dynamic imaging flow cell that enables monitoring the thickness and topology of live biofilms over time using a WLI microscope. The microfluidic system is specifically designed to create a reflective interface on the surface of biofilms while minimizing disruption of fragile structures. The imaging cell was also designed to accommodate limitations imposed by the depth of focus of the microscope’s objective lens. Example images of live biofilm samples are shown in order to illustrate the ability of the flow cell and WLI instrument to 1) support bacterial growth and biofilm development, 2) image biofilm structure that reflects growth in flow conditions, and 3) monitor biofilm development over time non-destructively. In future work, the apparatus described here will enable surface metrology measurements (roughness, surface area, etc.) of biofilms and may be used to observe changes in biofilm structure in response to changes in environmental conditions (e.g., flow velocity, availability of nutrients, and presence of biocides). Furthermore, this development will open new opportunities for the use of WLI in bioimaging.

47 OTHER INSTRUMENTATION↗

User-Driven Sampling Strategies in Image Exploitation

Visual analytics and interactive machine learning both try to leverage the complementary strengths of humans and machines to solve complex data exploitation tasks. These fields overlap most significantly when training is involved: the visualization or machine learning tool improves over time by exploiting observations of the human-computer interaction. This paper focuses on one aspect of the human-computer interaction that we call user-driven sampling strategies. Unlike relevance feedback and active learning sampling strategies, where the computer selects which data to label at each iteration, we investigate situations where the user selects which data is to be labeled at each iteration. User-driven sampling strategies can emerge in many visual analytics applications but they have not been fully developed in machine learning. We discovered that in user-driven sampling strategies suggest new theoretical and practical research questions for both visualization science and machine learning. In this paper we identify and quantify the potential benefits of these strategies in a practical image analysis application. We find user-driven sampling strategies can sometimes provide significant performance gains by steering tools towards local minima that have lower error than tools trained with all of the data. Furthermore, in preliminary experiments we find these performance gains are particularly pronounced when the user is experienced with the tool and application domain.

97 MATHEMATICS AND COMPUTING↗

Production and dosimetric aspects of the potent Auger emitter {sup 58m}Co for targeted radionuclide therapy of small tumors

Purpose: Based on theoretical calculations, the Auger emitter {sup 58m}Co has been identified as a potent nuclide for targeted radionuclide therapy of small tumors. During the production of this isotope, the coproduction of the long-lived ground state {sup 58g}Co is unfortunately unavoidable, as is ingrowth of the ground state following the isomeric decay of {sup 58m}Co. The impact of {sup 58g}Co as a {beta}{sup +}- and {gamma}-emitting impurity should be included in the dosimetric analysis. The purpose of this study was to investigate this critical part of dosimetry based on experimentally determined production yields of {sup 58m}Co and {sup 58g}Co using a low-energy cyclotron. Also, the cellular S-values for {sup 58m}Co have been calculated and are presented here for the first time. Methods: {sup 58m}Co was produced via the {sup 58}Fe(p,n){sup 58m}Co nuclear reaction on highly enriched {sup 58}Fe metal. In addition, radiochemical separations of produced radio-cobalt from {sup nat}Fe target material were performed. The theoretical subcellular dosimetry calculations for {sup 58m}Co and {sup 58g}Co were performed using the MIRD formalism, and the impact of the increasing ground state impurity on the tumor-to-normal-tissue dose ratios (TND) per disintegration as a function of time after end of bombardment (EOB) was calculated. Results: 192 {+-} 8 MBq of {sup 58m}Co was produced in the irradiation corresponding to a production yield of 10.7 MBq/{mu}Ah. The activity of {sup 58g}Co was measured to be 0.85% {+-} 0.04% of the produced {sup 58m}Co activity at EOB. The radio-cobalt yields in the rapid separations were measured to be >97% with no detectable iron contaminations in the cobalt fractions. Due to the unavoidable coproduction and ingrowth of the long-lived ground state {sup 58g}Co, the TND and the potency of the {sup 58m}Co decrease with time after EOB. If a future treatment with a {sup 58m}Co labeled compound is not initiated before, e.g., 21 h after EOB, the resulting TND will be approximately 50% of the TND of 'pure'{sup 58m}Co as a result of the increased normal tissue dose from the ground state. Conclusions: The Auger emitter {sup 58m}Co is a potent radioisotope for targeted radionuclide therapy, and the production of therapeutic quantities should be achievable using a small biomedical cyclotron. However, the unavoidable coproduction and ingrowth of the long-lived ground state {sup 58g}Co requires fast radiochemical processing and use of future {sup 58m}Co-labeled radiopharmaceuticals in order to exploit the high achievable TND of {sup 58m}Co.

61 RADIATION PROTECTION AND DOSIMETRY↗

Physiologically guided approach to characterizing respiratory motion

Purpose: To characterize radiation therapy patient breathing patterns based on measured external surrogate information. Methods: Breathing surrogate data were collected during 4DCT from a cohort of 50 patients including 28 patients with lung cancer and 22 patients without lung cancer. A spirometer and an abdominal pneumatic bellows were used as the surrogates. The relationship between these measurements was assumed to be linear within a small phase difference. The signals were correlated and drift corrected using a previously published method to convert the signal into tidal volume. The airflow was calculated with a first order time derivative of the tidal volume using a window centered on the point of interest and with a window length equal to the CT gantry rotation period. The airflow was compared against the tidal volume to create ellipsoidal patterns that were binned into 25 ml × 25 ml/s bins to determine the relative amount of time spent in each bin. To calculate the variability of the maximum inhalation tidal volume within a free-breathing scan timeframe, a metric based on percentile volume ratios was defined. The free breathing variability metric (κ) was defined as the ratio between extreme inhalation tidal volumes (defined as >93 tidal volume percentile of the measured tidal volume) and normal inhalation tidal volume (defined as >80 tidal volume percentile of the measured tidal volume). Results: There were three observed types of volume-flow curves, labeled Types 1, 2, and 3. Type 1 patients spent a greater duration of time during exhalation withκ = 1.37 ± 0.11. Type 2 patients had equal time duration spent during inhalation and exhalation with κ = 1.28 ± 0.09. The differences between the mean peak exhalation to peak inhalation tidal volume, breathing period, and the 85th tidal volume percentile for Type 1 and Type 2 patients were statistically significant at the 2% significance level. The difference between κ and the 98th tidal volume percentile for Type 1 and Type 2 patients was found to be statistically significant at the 1% significance level. Three patients did not display a breathing stability curve that could be classified as Type 1 or Type 2 due to chaotic breathing patterns. These patients were classified as Type 3 patients. Conclusions: Based on an observed volume-flow curve pattern, the cohort of 50 patients was divided into three categories called Type 1, Type 2, and Type 3. There were statistically significant differences in breathing characteristics between Type 1 and Type 2 patients. The use of volume-flow curves to classify patients has been demonstrated as a physiological characterization metric that has the potential to optimize gating windows in radiation therapy.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Low-complexity atlas-based prostate segmentation by combining global, regional, and local metrics

Purpose: To improve the efficiency of atlas-based segmentation without compromising accuracy, and to demonstrate the validity of the proposed method on MRI-based prostate segmentation application. Methods: Accurate and efficient automatic structure segmentation is an important task in medical image processing. Atlas-based methods, as the state-of-the-art, provide good segmentation at the cost of a large number of computationally intensive nonrigid registrations, for anatomical sites/structures that are subject to deformation. In this study, the authors propose to utilize a combination of global, regional, and local metrics to improve the accuracy yet significantly reduce the number of required nonrigid registrations. The authors first perform an affine registration to minimize the global mean squared error (gMSE) to coarsely align each atlas image to the target. Subsequently, atarget-specific regional MSE (rMSE), demonstrated to be a good surrogate for dice similarity coefficient (DSC), is used to select a relevant subset from the training atlas. Only within this subset are nonrigid registrations performed between the training images and the target image, to minimize a weighted combination of gMSE and rMSE. Finally, structure labels are propagated from the selected training samples to the target via the estimated deformation fields, and label fusion is performed based on a weighted combination of rMSE and local MSE (lMSE) discrepancy, with proper total-variation-based spatial regularization. Results: The proposed method was applied to a public database of 30 prostate MR images with expert-segmented structures. The authors’ method, utilizing only eight nonrigid registrations, achieved a performance with a median/mean DSC of over 0.87/0.86, outperforming the state-of-the-art full-fledged atlas-based segmentation approach of which the median/mean DSC was 0.84/0.82 when applying to their data set. Conclusions: The proposed method requires a fixed number of nonrigid registrations, independent of atlas size, providing desirable scalability especially important for a large or growing atlas. When applied to prostate segmentation, the method achieved better performance to the state-of-the-art atlas-based approaches, with significant improvement in computation efficiency. The proposed rationale of utilizing jointly global, regional, and local metrics, based on the information characteristic and surrogate behavior for registration and fusion subtasks, can be extended naturally to similarity metrics beyond MSE, such as correlation or mutual information types.

62 RADIOLOGY AND NUCLEAR MEDICINE↗