SEARCH · Search NASA
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.
We apply semantic boosting to the Segment Anything Model (SAM) to obtain microstructure segmentation for transmission electron microscopy. Our booster, SAM-I-Am, extracts geometric and textural features of various intermediate masks to perform mask removal and mask merging operations
Explore the source record for details and available documents.
Purpose: Although positron emission tomography (PET) images have shown potential to improve the accuracy of targeting in radiation therapy planning and assessment of response to treatment, the boundaries of tumors are not easily distinguishable from surrounding normal tissue owing to the low spatial resolution and inherent noisy characteristics of PET images. The objective of this study is to develop a generic and robust method for automatic delineation of tumor volumes using an active contour model and to evaluate its performance using phantom and clinical studies. Methods: MASAC, a method for automatic segmentation using an active contour model, incorporates the histogram fuzzy C-means clustering, and localized and textural information to constrain the active contour to detect boundaries in an accurate and robust manner. Moreover, the lattice Boltzmann method is used as an alternative approach for solving the level set equation to make it faster and suitable for parallel programming. Twenty simulated phantom studies and 16 clinical studies, including six cases of pharyngolaryngeal squamous cell carcinoma and ten cases of nonsmall cell lung cancer, were included to evaluate its performance. Besides, the proposed method was also compared with the contourlet-based active contour algorithm (CAC) and Schaefer’s thresholding method (ST). The relative volume error (RE), Dice similarity coefficient (DSC), and classification error (CE) metrics were used to analyze the results quantitatively. Results: For the simulated phantom studies (PSs), MASAC and CAC provide similar segmentations of the different lesions, while ST fails to achieve reliable results. For the clinical datasets (2 cases with connected high-uptake regions excluded) (CSs), CAC provides for the lowest mean RE (−8.38% ± 27.49%), while MASAC achieves the best mean DSC (0.71 ± 0.09) and mean CE (53.92% ± 12.65%), respectively. MASAC could reliably quantify different types of lesions assessed in this work with good accuracy, resulting in a mean RE of −13.35% ± 11.87% and −11.15% ± 23.66%, a mean DSC of 0.89 ± 0.05 and 0.71 ± 0.09, and a mean CE of 19.19% ± 7.89% and 53.92% ± 12.65%, for PSs and CSs, respectively. Conclusions: The authors’ results demonstrate that the developed novel PET segmentation algorithm is applicable to various types of lesions in the authors’ study and is capable of producing accurate and consistent target volume delineations, potentially resulting in reduced intraobserver and interobserver variabilities observed when using manual delineation and improved accuracy in treatment planning and outcome evaluation.
We predict that CeBi in the ferromagnetic state is a Weyl semimetal. Our calculations within density functional theory show the existence of two pairs of Weyl nodes on the momentum path (0,0,k z ) at 15meV above and 100meV below the Fermi level. Two corresponding Fermi arcs are obtained on surfaces of mirror-symmetric (010)-oriented slabs at E=15meV and both arcs are interrupted into three segments due to hybridization with a set of trivial surface bands. By studying the spin texture of surface states, we find the two Fermi arcs are strongly spin polarized but in opposite directions, which can be detected by spin-polarized ARPES measurements. Here, our theoretical study of quasiparticle interference (QPI) for a nonmagnetic impurity at the Bi site also reveals several features related to the Fermi arcs. Specifically, we predict that the spin polarization of the Fermi arcs leads to a bifurcation-shaped feature only in the spin-dependent QPI spectrum, serving as a fingerprint of the Weyl nodes.
Purpose: Standardized uptake values (SUV) are standard quantitative PET measures of FDG tumor uptake used,and are used as a tool to monitor response to therapy. Textural analysis is emerging as a new tool for assessing intratumoral heterogeneity which may allow better tissue characterization and improved prediction of response and survival rate.Understanding what variations may be expected in these parameters is key in order to make decisions based on how the change throughout the course of treatment. The aim of this study was to assess repeatability in SUV measures and texture parameters,and establish criteria that differentiate changes associated with treatment rather than statistical variability. Methods: Eighty patients,167 random lesions total,were scanned in a GE Discovery STE PET/CT Scanner. One field-of-view was chosen centered on the largest lesion observed in a clinical whole-body FDG PET.Immediately following,a gated 9 min scan was acquired in list mode,without changing the patient’s position between any scans. Data was replayed into 3 time bins,3 min each,in order to insure equivalent noise characteristics in each replicate.Data was reconstructed into 128×128×47 square matrices.One VOI was drawn over each lesion for each patient and used to segment all 3 replicates. The mean.max and peak SUV were calculated for each VOI and replicate. First-order textural features were also calculated (skewness and kurtosis). Repeatability was calculated as the average standard deviation over the mean for the 3 repeated measurements for each lesion. Results: The average percent error in the SUV max,peak and mean were 3.4%(0– 12.9%),1.9% (0–7.5%),2.8% (0–12.2%),respectively.For skewness and kurtosis they were 10.9% and 17.8%. Conclusion: We have shown that there is a large variation in %error in SUV measures across patients. SUVpeak is the least variable and kurtosis and skewness parameters are less reliable thatn SUVs.Higher order textures are be.
Purpose: We propose a method to examine gynecological tumor heterogeneity using texture analysis in the context of an adaptive PET protocol in order to establish if texture metrics from baseline PET-CT predict tumor response better than SUV metrics alone as well as determine texture features correlating with tumor response during radiation therapy. Methods: This IRB approved protocol included 29 women with node positive gynecological cancers visible on FDG-PET treated with EBRT to the PET positive nodes. A baseline and intra-treatment PET-CT was obtained. Tumor outcome was determined based on RECIST on posttreatment PET-CT. Primary GTVs were segmented using 40% threshold and a semi-automatic gradient-based contouring tool, PET Edge (MIM Software Inc., Cleveland, OH). SUV histogram features, Metabolic Volume (MV), and Total Lesion Glycolysis (TLG) were calculated. Four 3D texture matrices describing local and regional relationships between voxel intensities in the GTV were generated: co-occurrence, run length, size zone, and neighborhood difference. From these, 39 texture features were calculated. Prognostic power of baseline features derived from gradientbased and threshold GTVs were determined using the Wilcoxon rank-sum test. Receiver Operating Characteristics and logistic regression was performed using JMP (SAS Institute Inc., Cary, NC) to find probabilities of predicting response. Changes in features during treatment were determined using the Wilcoxon signed-rank test. Results: Of the 29 patients, there were 16 complete responders, 7 partial responders, and 6 non-responders. Comparing CR/PR vs. NR for gradient-based GTVs, 7 texture values, TLG, and SUV kurtosis had a p < 0.05. Threshold GTVs yielded 4 texture features and TLG with p < 0.05. From baseline to intra-treatment, 14 texture features, SUVmean, SUVmax, MV, and TLG changed with p < 0.05. Conclusion: Texture analysis of PET imaged gynecological tumors is an effective method for early prognosis and should be used complimentary to SUV metrics, especially when using gradient based segmentation.
Global food production in the developing world occurs within sub-hectare fields that are difficult to identify with moderate resolution satellite imagery. Knowledge about the distribution of these fields is critical in food security programs. We developed a semi-automated image segmentation approach using wall-to-wall sub-meter imagery with high-performance computing to map crop area (CA) throughout Tigray, Ethiopia that encompasses over 41,000 km (exp 2). Multiple processing streams were tested to minimize mapping error while applying five unique smoothing kernels to capture differences in land surface texture associated to CA. Typically, very-small fields (mean < 2 ha) have a smooth image roughness compared to natural scrub/shrub woody vegetation at the ~1m scale and these features can be segmented in panchromatic imagery with multi-level histogram thresholding. Multi-temporal very-high resolution (VHR) panchromatic imagery with multi-spectral VHR are sufficient in extracting critical CA information needed in food security programs. A 2011 to 2015 CA map was produced, using over 3000 WorldView-1 panchromatic images wall-to-wall in 1/2 deg mosaics for Tigray, Ethiopia. CA was evaluated with nearly 3000 WorldView-2 2m multispectral 250 X 250 m image subsets by seven expert interpretations, and with in-situ global positioning system photography. CA estimates ranged from 32 to 41% in sub regions of Tigray with median maximum per bin commission and omission errors of 11% and 1% respectively, with most of the error occurring in bins <15%. This empirical, simple, and low direct cost approach via U.S. government license agreement to access commercial VHR data, could be a viable big-data high-performance computing methodology to extract wall-to-wall CA for other regions of the world that have very-small agriculture fields with similar image texture."
Multispectral remote sensing data sets collected during the Geologic Remote Sensing Field Experiment (GRSFE) conducted during 1989 in the southwestern U.S. were used to produce thematic image maps showing details of the surface geology. LANDSAT TM (Thematic Mapper) images were used to map the distribution of clays, carbonates, and iron oxides. AVIRIS (Airborne Visible/Infrared Imaging Spectrometer) data were used to identify and map calcite, dolomite, sericite, hematite, and geothite, including mixtures. TIMS (Thermal Infrared Multispectral Scanner) data were used to map the distribution of igneous rock phases and carbonates based on their silica contents. AIRSAR (Airborne Synthetic Aperture Radar) data were used to map surface textures related to the scale of surface roughness. The AIRSAR also allowed identification of previously unmapped fault segments and structural control of lithology and minerology. Because all of the above data sets were geographically referenced, combination of different data types and direct comparison of the results with conventional field and laboratory data sets allowed improved geologic mapping of the test site.
The focus is to examine how estimates of three dimensional scene structure, as encoded in a scene disparity map, can be improved by the analysis of the original monocular imagery. The utilization of surface illumination information is provided by the segmentation of the monocular image into fine surface patches of nearly homogeneous intensity to remove mismatches generated during stereo matching. These patches are used to guide a statistical analysis of the disparity map based on the assumption that such patches correspond closely with physical surfaces in the scene. Such a technique is quite independent of whether the initial disparity map was generated by automated area-based or feature-based stereo matching. Stereo analysis results are presented on a complex urban scene containing various man-made and natural features. This scene contains a variety of problems including low building height with respect to the stereo baseline, buildings and roads in complex terrain, and highly textured buildings and terrain. The improvements are demonstrated due to monocular fusion with a set of different region-based image segmentations. The generality of this approach to stereo analysis and its utility in the development of general three dimensional scene interpretation systems are also discussed.
The mechanical behavior of partial-penetration laser welds exhibits significant variability in engineering quantities such as strength and apparent ductility. Understanding the root cause of this variability is important when using such welds in engineering designs. In Part II of this work, we develop finite element simulations with geometry derived from micro-computed tomography (μCT) scans of partial-penetration 304L stainless steel laser welds that were analyzed in Part I. We use these models to study the effects of the welds’ small-scale geometry, including porosity and weld depth variability, on the structural performance metrics of weld ductility and strength under quasi-static tensile loading. We show that this small-scale geometry is the primary cause of the observed variability for these mechanical response quantities. Additionally, we explore the sensitivity of model results to the conversion of the μCT data to discretized model geometry using different segmentation algorithms, and to the effect of small-scale geometry simplifications for pore shape and weld root texture. The modeling approach outlined and results of this work may be applicable to other material systems with small-scale geometric features and defects, such as additively manufactured materials.
Purpose: Radiomics involves the extraction of texture features from different imaging modalities with the purpose of developing models to predict patient treatment outcomes. The purpose of this study is to investigate texture feature reproducibility across [18F]FDG PET/CT and [18F]FDG PET/MR imaging in patients with primary malignancies. Methods: Twenty five prospective patients with solid tumors underwent clinical [18F]FDG PET/CT scan followed by [18F]FDG PET/MR scans. In all patients the lesions were identified using nuclear medicine reports. The images were co-registered and segmented using an in-house auto-segmentation method. Fifty features, based on the intensity histogram, second and high order matrices, were extracted from the segmented regions from both image data sets. One-way random-effects ANOVA model of the intra-class correlation coefficient (ICC) was used to establish texture feature correlations between both data sets. Results: Fifty features were classified based on their ICC values, which were found in the range from 0.1 to 0.86, in three categories: high, intermediate, and low. Ten features extracted from second and high-order matrices showed large ICC ≥ 0.70. Seventeen features presented intermediate 0.5 ≤ ICC ≤ 0.65 and the remaining twenty three presented low ICC ≤ 0.45. Conclusion: Features with large ICC values could be reliable candidates for quantification as they lead to similar results from both imaging modalities. Features with small ICC indicates a lack of correlation. Therefore, the use of these features as a quantitative measure will lead to different assessments of the same lesion depending on the imaging modality from where they are extracted. This study shows the importance of the need for further investigation and standardization of features across multiple imaging modalities.
Image segmentation is a critical enabler for tasks ranging from medical diagnostics to autonomous driving. However, the correct segmentation semantics — where are boundaries located? what segments are logically similar? — change depending on the domain, such that state-of-the-art foundation models can generate meaningless and incorrect results. Moreover, in certain domains, fine-tuning and retraining techniques are infeasible: obtaining labels is costly and time-consuming; domain images (micrographs) can be exponentially diverse; and data sharing (for third-party retraining) is restricted. To enable rapid adaptation of the best segmentation technology, we propose the concept of semantic boosting: given a zero-shot foundation model, guide its segmentation and adjust results to match domain expectations. Here, we apply semantic boosting to the Segment Anything Model (SAM) to obtain microstructure segmentation for transmission electron microscopy. Our booster, SAM-I-Am, serves as a post-processing engine that extracts geometric and textural features of various intermediate masks to perform mask removal and mask merging operations. We demonstrate a zero-shot performance increase of (absolute) +21.35%, +12.6%, +5.27% in mean IoU, and a -9.91%, -18.42%, -4.06% drop in mean false positive masks across images of three difficulty classes over vanilla SAM (ViT-L).
Several sets of LANDSAT image segments and a set of simple image processing programs operating in batch mode are being used to teach an introductory course in digital image analysis and classification. The image data sets contain only a small number of pixels and can be run quickly on most university computing systems. A number of image data sets derived from the raw spectral bands (smoothed data sets, band ratios and texture data sets) allows the student to examine and include these extracted variables in the classification process. Another data set consists of three registered LANDSAT image segments for three different dates so that the student may experiment with multitemporal data processing. The three LANDSAT scenes used in the class are described and the image data sets available for class exercises are listed.
Algorithms and techniques for use in the identification and location of large buildings in digitized copies of aerial photographs are developed and tested. The building data would be used in the simulation of objects located in the vicinity of an airport that may be detected by aircraft radar. Two distinct approaches are considered. Most building footprints are rectangular in form. The first approach studied is to search for right-angled corners that characterize rectangular objects and then to connect these corners to complete the building. This problem is difficult because many nonbuilding objects, such as street corners, parking lots, and ballparks often have well defined corners which are often difficult to distinguish from rooftops. Furthermore, rooftops come in a number of shapes, sizes, shadings, and textures which also limit the discrimination task. The strategy used linear sequences of different samples to detect straight edge segments at multiple angles and to determine when these segments meet at approximately right-angles with respect to each other. This technique is effective in locating corners. The test image used has a fairly rectangular block pattern oriented about thirty degrees clockwise from a vertical alignment, and the overall measurement data reflect this. However, this technique does not discriminate between buildings and other objects at an operationally suitable rate. In addition, since multiple paths are tested for each image pixel, this is a time consuming task. The process can be speeded up by preprocessing the image to locate the more optimal sampling paths. The second approach is to rely on a human operator to identify and select the building objects and then to have the computer determine the outline and location of the selected structures. When presented with a copy of a digitized aerial photograph, the operator uses a mouse and cursor to select a target building. After a button on the mouse is pressed, with the cursor fully within the perimeter of the building, the program scans from the position of the cursor to a perimeter position where a shift in grayscale is detected. Once at the perimeter, the process traces along it, around the building, until it eventually returns to the perimeter starting point. Spatial resolution limits cause the perimeter trace to be somewhat course so that a line straightening algorithm is employed. One result is that the building corner positions become more distinctly defined.
Robotic vision, involving the use of a vision system to control a process, is discussed. Design and selection of active sensors employing radiation of radio waves, sound waves, and laser light, respectively, to light up unobservable features in the scene are considered, as are design and selection of passive sensors, which rely on external sources of illumination. The segmentation technique by which an image is separated into different collections of contiguous picture elements having such common characteristics as color, brightness, or texture is examined, with emphasis on the edge detection technique. The IMFEX (image feature extractor) system performing edge detection and thresholding at 30 frames/sec television frame rates is described. The template matching and discrimination approach to recognize objects are noted. Applications of robotic vision in industry for tasks too monotonous or too dangerous for the workers are mentioned.
Texture is seen as an important spatial feature useful for identifying objects or regions of interest in an image. While textural features have been widely used in analyzing a variety of photographic images, they have not been used in processing radar images. A procedure for extracting a set of textural features for characterizing small areas in radar images is presented, and it is shown that these features can be used in classifying segments of radar images corresponding to different geological formations.
Endeavour crater (2.28 deg S, 354.77 deg E) is a Noachian-aged 22 km-diameter impact structure of complex morphology in Meridiani Planum. The degradation state of the crater has been studied using Mars Reconnaissance Orbiter and Opportunity rover data. Exposed rim segments rise approximately 10 m to approximately 100 m above the level of the embaying Burns Formation and the crater is 200-500 m deep with the southern interior wall exposing over approximately 300 m relief. Both pre-impact rocks (Matijevic Formation) and Endeavour impact ejecta (Shoemaker Formation) are present at Cape York, but only the Shoemaker crops out (up to approximately 140 m) along the rim segment from Murray Ridge to Cape Tribulation. Study of pristine complex craters Bopolu and Tooting, and morphometry of other martian complex craters, enables us to approximate Endeavour's pristine form. The original rim likely averaged 410 m (+/-)200 m in elevation and a 250-275 m section of ejecta ((+/-)50-60 m) would have composed a significant fraction of the rim height. The original crater depth was likely between 1.5 km and 2.2 km. Comparison between the predicted original and current form of Endeavour suggests approximately 100-200 m rim lowering that removed most ejecta in some locales (e.g., Cape York) while thick sections remain elsewhere (e.g., Cape Tribulation). Almost complete removal of ejecta at Cape York and minimal observable offset across fractures indicates current differences in rim relief are not solely due to original rim relief. Rim segments are embayed by approximately 100-200 m thickness of plains rocks outside the crater, but thicker deposits lie inside the crater. Ventifact textures confirm ongoing eolian erosion with the overall extent difficult to estimate. Analogy with degraded Noachian-aged craters south of Endeavour, however, suggests fluvial erosion dominated rim degradation in the Noachian and was likely followed by approximately 10s of meters modification by alternate processes. Such degradation is consistent with 1) the interpretation of a pediment on the rim flanks of Endeavour, 2) the formation of features such as Marathon Valley, 3) the nearly complete removal of ejecta at Cape York, 4) preservation of a thicker section of ejecta at Cape Tribulation and perhaps, 5) the origin of some gaps in the rim around the crater. A paucity of debris shed from the rim indicates most degradation occurred prior to embayment by the plains rocks.
A feasibility study of automated classification of satellite images is described. Satellite images were characterized by the textures they contain. In particular, the detection of cloud textures was investigated. The method of second-order gray level statistics, using co-occurrence matrices, was applied to extract feature vectors from image segments. Neural network technology was employed to classify these feature vectors. The cascade-correlation architecture was successfully used as a classifier. The use of a Kohonen network was also investigated but this architecture could not reliably classify the feature vectors due to the complicated structure of the classification problem. The best results were obtained when data from different spectral bands were fused.