Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Image Segmentation”

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 613 records · Page 34

Optical image analysis of WSe 2 − thresholding for layer detection

The fast and reliable layer identification of two-dimensional transition metal dichalcogenide (TMD), such as WSe 2 , is essential to investigating their thickness-dependent electronic and optical properties. This article presents efficient optical image thresholding methodology designed to segment the mono, bi, and tri-layer regions of WSe 2 flakes mechanically exfoliated onto a SiO 2 /Si substrate. The optical images were first preprocessed to exclude the background effect and analyzed using the pixel medians and interquartile ranges for fundamental color channels—red, green, and blue (RGB). The analysis of red channel pixel intensities yielded three distinct ranges, serving as thresholds for layer segmentation: monolayer (111.0–118.0), bilayer (103.0–110.0), and tri-layer (93.0–103.0). Similarly, thresholds were established for each color channel, facilitating a comparative study of the segmentation performances. Further, the intersection-over-union ($IoU$) calculations revealed that the red and green channels demonstrated greater than 99 % and 90 % accuracy in differentiating each layer, respectively. This approach yields remarkable results without substantial data calibration that utilizes time-intensive heuristic techniques. Moreover, the proposed methodology offers the flexibility to compare performances across different color channels, expanding the applicability for other 2D material systems.

2D Materials↗

TU-AB-BRA-01: Abdominal Synthetic CT Generation in Support of Liver SBRT Dose Calculation

Purpose: To demonstrate and validate a technique for generating MRI-derived synthetic CT volumes (MRCTs) in support of adaptive liver SBRT. Methods: Under IRB approval, ten hepatocellular carcinoma patients were scanned using a single MR sequence (T1 Dixon-VIBE), yielding inherently-registered water, fat, and T1-weighted images. Air-containing voxels were identified by intensity thresholding. The envelope of the anterior vertebral bodies was segmented from the fat image by fitting a shape model to vertebral body candidate voxels, then using level sets to expand the contour outward. Fuzzy-C-Means (FCM) was then used to classify each non-air voxel in the image as fat, water, bone, or marrow. Bone and marrow only were classified within the vertebral body envelope. The MRCT was created by integrating the product of the FCM class probability with the assigned class density for each voxel. The resulting MRCTs were deformably aligned with planning CTs and 2-ARC SBRT VMAT plans were optimized on the MRCT density maps. Fluence was copied onto the CT density grids and dose recalculated. Results: The MRCTs faithfully reproduced most of the features visible in the corresponding CT image volumes, with exceptions of ribs and posterior spinous processes. The liver, vertebral bodies, kidneys, spleen and cord all had median HU differences of less than 75 between MRCT and CT images. PTV D99% values had an average 0.2% difference (standard deviation: 0.46%) between calculations on MRCT and CT density grids. The maximum difference in dose to 0.1cc of the PTV was 0.25% (std:0.49%). OAR dose differences were similarly small (mean:0.03Gy, std:0.26Gy). The largest normal tissue complication percentage (NTCP) difference was 1.48% (mean:0.06%, std:0.54%). Conclusions: MRCTs from a single abdominal imaging sequence are promising for use in SBRT dose calculation. Future work will focus on extending models to better define bones in the upper abdomen. Supported by NIHR01EB016079 and NIH1L30CA199594-01.

60 APPLIED LIFE SCIENCES↗

Object-constrained meshless deformable algorithm for high speed 3D nonrigid registration between CT and CBCT

Purpose: High-speed nonrigid registration between the planning CT and the treatment CBCT data is critical for real time image guided radiotherapy (IGRT) to improve the dose distribution and to reduce the toxicity to adjacent organs. The authors propose a new fully automatic 3D registration framework that integrates object-based global and seed constraints with the grayscale-based ''demons'' algorithm. Methods: Clinical objects were segmented on the planning CT images and were utilized as meshless deformable models during the nonrigid registration process. The meshless models reinforced a global constraint in addition to the grayscale difference between CT and CBCT in order to maintain the shape and the volume of geometrically complex 3D objects during the registration. To expedite the registration process, the framework was stratified into hierarchies, and the authors used a frequency domain formulation to diffuse the displacement between the reference and the target in each hierarchy. Also during the registration of pelvis images, they replaced the air region inside the rectum with estimated pixel values from the surrounding rectal wall and introduced an additional seed constraint to robustly track and match the seeds implanted into the prostate. The proposed registration framework and algorithm were evaluated on 15 real prostate cancer patients. For each patient, prostate gland, seminal vesicle, bladder, and rectum were first segmented by a radiation oncologist on planning CT images for radiotherapy planning purpose. The same radiation oncologist also manually delineated the tumor volumes and critical anatomical structures in the corresponding CBCT images acquired at treatment. These delineated structures on the CBCT were only used as the ground truth for the quantitative validation, while structures on the planning CT were used both as the input to the registration method and the ground truth in validation. By registering the planning CT to the CBCT, a displacement map was generated. Segmented volumes in the CT images deformed using the displacement field were compared against the manual segmentations in the CBCT images to quantitatively measure the convergence of the shape and the volume. Other image features were also used to evaluate the overall performance of the registration. Results: The algorithm was able to complete the segmentation and registration process within 1 min, and the superimposed clinical objects achieved a volumetric similarity measure of over 90% between the reference and the registered data. Validation results also showed that the proposed registration could accurately trace the deformation inside the target volume with average errors of less than 1 mm. The method had a solid performance in registering the simulated images with up to 20 Hounsfield unit white noise added. Also, the side by side comparison with the original demons algorithm demonstrated its improved registration performance over the local pixel-based registration approaches. Conclusions: Given the strength and efficiency of the algorithm, the proposed method has significant clinical potential to accelerate and to improve the CBCT delineation and targets tracking in online IGRT applications.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

TH-AB-209-08: Next Generation Dedicated 3D Breast Imaging with XACT

Purpose: Exposure to radiation increases the risk of cancer. We have designed a new imaging paradigm, X-ray induced acoustic computed tomography (XACT). Applying this innovative technology to breast imaging, an X-ray exposure can generate a 3D acoustic image, which dramatically reduces the radiation dose to patients when compared to conventional breast CT. Methods: Theoretical calculations are done to determine the appropriate X-ray energy and ultrasound frequency in breast XACT imaging. A series of breast CT image along the coronal plane from a patient with calcifications in the breast tissue are used as the source image. HU value based segmentation is done to distinguish the skin, adipose tissue, glandular tissue, breast calcification, and chest bone from each CT image. X-ray dose deposition in each pixel is calculated based on the tissue type by using GEANT4 Monte Carlo toolkits. The initial pressure rise caused by X-ray energy deposition is calculated according to tissue properties. Then, the X-ray induced acoustic wave propagation is simulated by K-WAVE toolkit. Breast XACT images are reconstructed from the recorded time-dependent ultrasound waves. Results: For imaging a breast with large size (16cm in diameter at chest wall), the photon energy of X-ray source and the central frequency of ultrasound detector is determined as 20keV and 5.5MHz. Approximately 10 times contrast between a calcification and the breast tissue can be acquire from XACT image. The calcification can be clearly identified from the reconstructed XACT image. Conclusion: XACT technique takes the advantages of X-ray absorption contrast and high ultrasonic resolution. With the proposed innovative technology, one can potentially reduce radiation dose to patient in 3D breast imaging as compared with current x-ray modalities, while still maintaining high imaging contrast and spatial resolution.

60 APPLIED LIFE SCIENCES↗

Assessment of Tumor Volumes in Skull Base Glomus Tumors Using Gluc-Lys[{sup 18}F]-TOCA Positron Emission Tomography

Purpose: To assess a threshold for Gluc-Lys[{sup 18}F]-TOCA positron emission tomography (PET) in target volume delineation of glomus tumors in the skull base and to compare with MRI-based target volume delineation. Methods and Materials: The threshold for volume segmentation in the PET images was determined by a phantom study. Nine patients with a total of 11 glomus tumors underwent PET either with Gluc-Lys[{sup 18}F]-TOCA or with {sup 68}Ga-DOTATOC (in 1 case). All patients were additionally scanned by MRI. Positron emission tomography and MR images were transferred to a treatment-planning system; MR images were analyzed for lesion volume by two observers, and PET images were analyzed by a semiautomated thresholding algorithm. Results: Our phantom study revealed that 32% of the maximum standardized uptake value is an appropriate threshold for tumor segmentation in PET-based target volume delineation of gross tumors. Target volume delineation by MRI was characterized by high interobserver variability. In contrast, interobserver variability was minimal if fused PET/MRI images were used. The gross tumor volumes (GTVs) determined by PET (GTV-PET) showed a statistically significant correlation with the GTVs determined by MRI (GTV-MRI) in primary tumors; in recurrent tumors higher differences were found. The mean GTV-MRI was significantly higher than mean GTV-PET. The increase added by MRI to the common volume was due to scar tissue with strong signal enhancement on MRI. Conclusions: In patients with glomus tumors, Gluc-Lys[{sup 18}F]-TOCA PET helps to reduce interobserver variability if an appropriate threshold for tumor segmentation has been determined for institutional conditions. Especially in patients with recurrent tumors after surgery, Gluc-Lys[{sup 18}F]-TOCA PET improves the accuracy of GTV delineation.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Effect of threshold parameters on infrared segmentation methods for porosity detection in electron beam powder bed fusion

In-situ process monitoring has seen significant interest in additive manufacturing to address qualification and certification goals. This is especially prevalent in metal powder bed fusion processes such as electron beam powder bed fusion (PBF-EB), with layer-wise infrared imaging being commonly used to detect defects. Here, this work compares two different segmentation methods (static thresholding and statistical thresholding) used for detecting porosity from in-situ infrared imaging data for PBF-EB. Samples were manufactured at a variety of focus offset values to induce porosity. Then, the segmented infrared images were compared to ex-situ X-ray computed tomography scans, which served as a ground-truth reference for objective evaluation. Through this analysis framework, the influential parameters, static threshold and N-value (number of standard deviations above the mean pixel value), respectively, for both image segmentation methods were analyzed and compared for their effects on porosity detection. With optimal parameter settings, the two methods had similar porosity detection performance, but the statistical method performed better under a larger variety of parameter settings.

Infrared imaging↗

Proposed Analytical Methods for Determining Filter Media Properties

High Efficiency Particulate Air (HEPA) filters, commonly used in nuclear filtration applications, are an integral part of the waste management processes in nuclear plants. HEPA filters are 99.97% efficient filtration devices characterized by their high resistance to air flow, or pressure drop. From theoretical models, the initial pressure drop across a clean filter is proven to be a function of the filter media fiber diameters and porosity of the media. These filter properties are relatively difficult to obtain with traditional manual methods; therefore, there is a need to develop analytical methods to find the fiber diameter and porosity to determine the pressure drop. Typically, these values can be found by substituting a calculated representative value based upon measured values, such as equivalent fiber diameter based upon a clean pressure drop. However, these values may also be directly recorded and measured by incorporating Scanning Electron Microscope (SEM) image analysis and other measurement methods to analyze the filter media and determine its physical properties without the need for media testing. DiameterJ, an open source Java plug-in, used with ImageJ, can process images of filter media taken by an SEM to find statistical data such as mean fiber diameter and porosity. To produce the raw data, SEM images of two filter media type samples are taken and segmented in DiameterJ using the traditional and statistical region merging segmentation algorithms. Manual segmentation is necessary after the initial segmentation by the algorithms as the images tend to be too complex for the algorithms to output with the necessary accuracy. However, complications exist in the manual segmentation process as these methods can be time intensive and prone to the individual bias of the user. This in turn can skew the final mean fiber diameter result, and lead to either an over or under prediction of the pressure drop. It was also discovered that the porosity data produced by DiameterJ is inaccurate, as the SEM analyzes the three-dimensional filter media by projecting its geometry onto a plane and analyzing it as a two-dimensional binary image. Thus, the porosity is artificially inflated through the segmentation process, rendering this result incorrect. Alternatively, density determination, gravimetric analysis, and thickness testing of the filter media is collectively used to determine the filter fiber porosity. Together, the SEM image analysis and analytical lab methods produce results through direct measurements which allow for the prediction of the initial pressure drop from clean filter media without the need for prior media testing to collect pressure data. By improving upon this proposed analytical method in the future, there is potential to streamline the process of finding these filter properties into a more direct methodology for determining the pressure drop across HEPA filters.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

SU-E-J-207: Effect of Pulse Sequence Parameters On Geometric Distortions Induced by a Titanium Brachytherapy Applicator

Purpose: To investigate the effect of readout bandwidth and voxel size on the appearance of distortion artifacts caused by a titanium brachytherapy applicator. Methods: An acrylic phantom was constructed to rigidly hold a MR conditional, titanium Fletcher-Suit-Delclos-style applicator set (Varian Medical Systems) for imaging on CT (Philips Brilliance) and 1.5T MRI (Siemens Magnetom Aera). Several variants of MRI parameters were tried for 2D T2-weighted turbo spin echo imaging in comparison against the standard clinical protocol with the criteria to keep relative SNR loss less than 20% and imaging time as short as possible. Two 3D sequences were also used for comparison with similar parameters. The applicator tandem was segmented on axial CT images (0.4×0.4×1.5mm {sup 3} resolution) and the CT images were registered to the 3D MR images in Eclipse (Varian). The applicator volume was then overlaid on all MRI sets in 3D-Slicer and distances were measured from the tandem tip to the MRI artifact edge in right/left/superior and anterior/posterior/superior directions from coronal and sagittal 2D acquisitions, respectively, or 3D data reformats. Artifact regions were also manually contoured in coronal/sagittal orientations for area measurements. Results: As would be expected, reductions in voxel size and increases in readout bandwidth reduced artifact size (average max artifact length decreased by 0.95 mm and average max area decrease by 0.27 cm{sup 2}). Interestingly, bandwidth increases yielded reductions in area (0.19 cm{sup 2}) and in distance measurements (1 mm) even with voxel increases, as compared to a standard protocol. This could be useful when high performance protocols are not feasible due to long imaging times. Conclusion: We have characterized artifacts caused by cervical brachytherapy applicator across multiple sequence parameters at 1.5T. Future work will focus on finalizing an optimal protocol that balances artifact reduction with imaging time and then testing this new protocol in patients.

60 APPLIED LIFE SCIENCES↗

Vehicle track segmentation using higher order random fields

Here, we present an approach to segment vehicle tracks in coherent change detection images, a product of combining two synthetic aperture radar images taken at different times. The approach uses multiscale higher order random field models to capture track statistics, such as curvatures and their parallel nature, that are not currently utilized in existing methods. These statistics are encoded as 3-by-3 patterns at different scales. The model can complete disconnected tracks often caused by sensor noise and various environmental effects. Coupling the model with a simple classifier, our approach is effective at segmenting salient tracks. We improve the F-measure on a standard vehicle track data set to 0.963, up from 0.897 obtained by the current state-of-the-art method.

97 MATHEMATICS AND COMPUTING↗

A hybrid machine-learning approach for analysis of methane hydrate formation dynamics in porous media with synchrotron CT imaging

Fast multi-phase processes in methane hydrate bearing samples pose a challenge for quantitative micro-computed tomography study and experiment steering due to complex tomographic data analysis involving time-consuming segmentation procedures. This is because of the sample's multi-scale structure, which changes over time, low contrast between solid and fluid materials, and the large amount of data acquired during dynamic processes. Here, a hybrid approach is proposed for the automatic segmentation of tomographic data from time-resolved imaging of methane gas-hydrate formation in sandy granular media, which includes a deep-learning 3D U-Net model. To prepare a training dataset for the 3D U-Net, a technique to automate data labeling based on sample-specific information about the mineral matrix immobility and occasional fluid movement in pores is proposed. Automatic segmentation allowed for studying properties of the hydrate growth in pores, as well as dynamic processes such as incremental flow and redistribution of pore brine. Results of the quantitative analysis showed that for typical gas-hydrate stability parameters (100 bar methane pressure, 7°C temperature) the rate of formation is slow (less than 1% per hour), after which the surface area of contact between brine and gas increases, resulting in faster formation (2.5% per hour). Hydrate growth reaches the saturation point after 11 h of the experiment. Finally, the efficacy of the proposed segmentation scheme in on-the-fly automatic data analysis and experiment steering with zooming to regions of interest is demonstrated.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Few-shot leaf segmentation

In this work, we use few-shot learning to segment the body and vein architecture of Populus trichocarpa leaves from high-resolution scans obtained in the UC Davis common garden. Leaf and vein segmentation are formulated as separate tasks, in which convolutional neural networks (CNNs) are used to iteratively expand partial segmentations until reaching stopping criteria. Our leaf and vein segmentation approaches use just 50 and 8 manually traced images for training, respectively, and are applied to a set of 2,634 top and bottom leaf scans. We show that both methods achieve high segmentation accuracy, in some cases exceeding even human-level segmentation. The leaf and vein segmentations are subsequently used to extract 68 morphological traits using traditional open-source image processing tools, which are validated using real-world physical measurements. For a biological perspective, we perform a genome-wide association study using the "vein density" trait to discover novel genetic architectures associated with multiple physiological processes relating to leaf development and function. In addition to sharing all of the few-shot learning code, we are releasing all images, manual segmentations, model predictions, 68 extracted leaf phenotypes, and a new set of SNPs called against the v4 P. trichocarpa genome for 1,419 genotypes.

Lagergren, John [Oak Ridge National Lab. (ORNL), O↗

Influence of Standard Image Processing of 3D X-ray Microscopy on Morphology, Topology and Effective Properties

Estimating porous media properties is a vital component of geosciences and the physics of porous media. Until now, imaging techniques have focused on methodologies to match image-derived flows or geomechanical parameters with experimentally identified values. Less emphasis has been placed on the compromise between image processing techniques and the consequences on topological and morphological characteristics and on computed properties such as permeability. The effects of some of the most popular image processing techniques (filtering and segmentation) available in open source on 3D X-ray Microscopy (micro-XRM) images are qualitatively and quantitatively discussed. We observe the impacts of various filters such as erosion-dilation and compare the efficiency of Otsu’s method of thresholding and the machine-learning-based software Ilastik for segmentation.

58 GEOSCIENCES↗

SU-E-J-137: Incorporating Tumor Regression Into Robust Plan Optimization for Head and Neck Radiotherapy

Purpose: To develop a robust planning paradigm which incorporates a tumor regression model into the optimization process to ensure tumor coverage in head and neck radiotherapy. Methods: Simulation and weekly MR images were acquired for a group of head and neck patients to characterize tumor regression during radiotherapy. For each patient, the tumor and parotid glands were segmented on the MR images and the weekly changes were formulated with an affine transformation, where morphological shrinkage and positional changes are modeled by a scaling factor, and centroid shifts, respectively. The tumor and parotid contours were also transferred to the planning CT via rigid registration. To perform the robust planning, weekly predicted PTV and parotid structures were created by transforming the corresponding simulation structures according to the weekly affine transformation matrix averaged over patients other than him/herself. Next, robust PTV and parotid structures were generated as the union of the simulation and weekly prediction contours. In the subsequent robust optimization process, attainment of the clinical dose objectives was required for the robust PTV and parotids, as well as other organs at risk (OAR). The resulting robust plans were evaluated by looking at the weekly and total accumulated dose to the actual weekly PTV and parotid structures. The robust plan was compared with the original plan based on the planning CT to determine its potential clinical benefit. Results: For four patients, the average weekly change to tumor volume and position was −4% and 1.2 mm laterally-posteriorly. Due to these temporal changes, the robust plans resulted in an accumulated PTV D95 that was, on average, 2.7 Gy higher than the plan created from the planning CT. OAR doses were similar. Conclusion: Integration of a tumor regression model into target delineation and plan robust optimization is feasible and may yield improved tumor coverage. Part of this research is supported by Varian Medical System.

60 APPLIED LIFE SCIENCES↗