Search NASA⌕ Search

SEARCH · Search NASA

Results for “Automated image quality evaluation”

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 73 records · Page 4

WE-G-204-09: Medical Physics 2.0 in Practice: Automated QC Assessment of Clinical Chest Images

Purpose: To determine whether a proposed suite of objective image quality metrics for digital chest radiographs is useful for monitoring image quality in our clinical operation. Methods: Seventeen gridless AP Chest radiographs from a GE Optima portable digital radiography (DR) unit (Group 1), seventeen (routine) PA Chest radiographs from a GE Discovery DR unit (Group 2), and sixteen gridless (non-routine) PA Chest radiographs from the same Discovery DR unit (Group 3) were chosen for analysis. Groups were selected to represent “sub-standard” (Group 1), “standard-of-care” (Group 2), and images with a gross technical error (Group 3). Group 1 images were acquired with lower kVp (90 vs. 125), shorter source-to-image distance (127cm vs 183cm) and were expected to have lower quality than images in Group 2. Group 3 was expected to have degraded contrast versus Group 2.This evaluation was approved by the institutional Quality Improvement Assurance Board (QIAB). Images were anonymized and securely transferred to the Duke University Clinical Imaging Physics Group for analysis using software previously described{sup 1} and validated{sup 2}. Image quality for individual images was reported in terms of lung grey level(Lgl); lung noise(Ln); rib-lung contrast(RLc); rib sharpness(Rs); mediastinum detail(Md), noise(Mn), and alignment(Ma); subdiaphragm-lung contrast(SLc); and subdiaphragm area(Sa). Metrics were compared across groups. Results: Metrics agreed with published Quality Consistency Ranges with three exceptions: higher Lgl, lower RLc, and SDc. Higher bit depth (16 vs 12) accounted for higher Lgl values in our images. Values were most internally consistent for Group 2. The most sensitive metric for distinguishing between groups was Mn followed closely by Ln. The least sensitive metrics were Md and RLc. Conclusion: The software appears promising for objectively and automatically identifying substandard images in our operation. The results can be used to establish local quality consistency ranges and action limits per facility preferences.

60 APPLIED LIFE SCIENCES↗

Geospatial-Temporal Semantic Graphs for Automated Wide-Area Search

We address the problem of wide-area search of overhead imagery. Given a time sequence of overhead images, we construct a geospatial-temporal semantic graph, which expresses the complex continuous information in the overhead images in a discrete searchable form, including explicit modeling of changes seen from one image to the next. We can then express desired search goals as a template graph, and search for matches using simple and efficient graph search algorithms. This produces a set of potential matches which provide cues for where to examine the imagery in detail, applying human expertise to determine which matches are correct. We include a match quality metric that scores the matches according to how well they match the stated search goal. This enables matches to be presented in sorted order with the best matches first, similar to the results returned by a web search engine. We present an evaluation of the method applied to several examples and data sets, and show that it can be used successfully for some problems. We also remark on several limitations of the method and note additional work needed to improve its scope and robustness. Approved for public release; further dissemination unlimited.

97 MATHEMATICS AND COMPUTING↗

SU-E-T-442: Sensitivity of Quality Assurance Tools to Delivery Errors On a Magnetic Resonance-Imaging Guided Radiation Therapy (MR-IGRT) System

Purpose: To test the sensitivity of the quality assurance (QA) tools actively used on a clinical MR-IGRT system for potential delivery errors. Methods: Patient-specific QA procedures have been implemented for a commercially available Cobalt-60 MR-IGRT system. The QA tools utilized were a MR-compatible cylindrical diode-array detector (ArcCHECK) with a custom insert which positions an ionization chamber (Exradin A18) in the middle of the device, as well as an in-house treatment delivery verification program. These tools were tested to investigate their sensitivity to delivery errors. For the ArcCHECK and ion chamber, a baseline was established with a static field irradiation to a known dose. Variations of the baseline were investigated which included rotated gantry, altered field size, directional shifts, and different delivery time. In addition, similar variations were tested with the automated delivery verification program that compared the treatment parameters in the machine delivery logs to the ones in the plan. To test the software, a 3-field conformal plan was generated as the baseline. Results: ArcCHECK noted at least a 13% decrease in passing rate from baseline in the following scenarios: gantry rotation of 1 degree from plan, 5mm change in field size, 2mm lateral shift, and delivery time decrease. Ion chamber measurements remained consistent for these variations except for the 5 second decrease in delivery time scenario which resulted in an 8% difference from baseline. The delivery verification software was able to detect and report the simulated errors such as when the gantry was rotated by 0.6 degrees, the beam weighting was changed by a percent, a single multileaf collimator was moved by 1cm, and the dose was changed from 2 to 1.8Gy. Conclusion: The results show that the current tools used for patient specific QA are capable of detecting small errors in RT delivery with presence of magnetic field.

60 APPLIED LIFE SCIENCES↗

Advanced SOFC quality control and the role of manufacturing defects on stack reliability

Atrex Energy rolled out first commercial SOFC generator products in 2013. Following that we embarked on implementation of mass production in order to ramp up productivity and lower the costs in the meantime. To ensure the goal of mass production, an accurate, cost effective and consistent quality control methodology needs to be in place. The aim of this DOE funded project is to devise a machine and methodology for cell quality control using modern non-destructive testing methods and cell rejection criterion based on the empirical determination of the role of manufacturing flaws on stack degradation and reliability. The work scope encompasses two goals aimed at 1) developing automatable techniques for objective flaw detection and 2) developing an understanding of the ramifications of flaws on stack and cell reliability. By accomplishing these two goals it is likely that an advanced stage Quality Control (QC) device can be developed. This project scope encompasses: 1) Invention of a high speed automated instrument for spatial mapping of surface and sub-surface imperfections in production cells (ex situ) using modern imaging techniques. 2) Identification of rejection criteria upon which to screen cells from production by: a. Magnification of the effect of production defects on cell degradation through fabrication and testing of micro-cells which encompass these imperfections. b. The long term testing of full size production cells both as single cells and in small stacks, in order to correlate the (in situ) performance and degradation with the imperfection type (identified ex-situ) and hence be able to classify imperfections as actual defects. We first built a defect library with the cells containing imperfections classified and documented. The cells with various imperfections were tested to evaluate the imperfection potential impact on cell electrochemical performance. Evaluations were conducted on minicell, single cell and stack levels. On minicell level, pinhole, scratch, crack, contamination, and pop-out were investigated. We found: (1) none of the imperfections causes calamity type failure, (2) pinholes do not show measurable effect on minicell performance, (3) imperfection such as electrolyte abrasion, scratch, and pop out cause minicell potential to fluctuate by different degrees; the cells with such imperfections exhibit sensitivity of potential to high fuel utilization and “self healing” behavior. On single cell level we evaluated the effect of long crack and high population pinholes. The cell with a long crack does not show performance compromise at start. There is no potential fluctuation or sensitivity to high fuel utilization or catastrophic failure. The voltage degradation due to thermal cycle is slightly higher than normal cells. The cells with high population pinholes perform surprisingly as well as normal cells over a period of 3,500h. Even when subjected to cyclic load condition, the cells 4 show negligible degradation. Our cells appear to have “self immunization” feature which fends off the adversary effects of certain imperfections. Defects were also evaluated on stack level. Two stacks were made each with ten defect cells. The first stack was tested for 13,2000h under steady load condition while the second was tested for 9,000h under transient load conditions. We found: (1) voltage fluctuation and sensitivity to thermal cycle were not observed in contrast to minicell test, (2) of the various cracks examined, including longitudinal cracks, circumferential crack, crack near closed and open ends, none exhibited distinguishable impact on performance, (3) defect cells exhibited higher degradation rates on average under steady load condition; Individually some defect cells performed similarly or better than reference cells, (4) under steady load condition all the cells degrade monotonically with degradation rate increasing with time, (5) under cyclic (transient) load condition, both defect and reference cells exhibit complex dynamic behaviors, (6) fuel loss (leak) was measured occasionally for stack 1, the cells with crack showed the highest fuel loss increase, nevertheless, no correlation with performance degradation could be established with certainty, and (7) no catastrophic failure ever happened other than incremental degradation. Posttest inspection and microanalysis were carried out; the observations are summarized in what follows: (1) Electrolyte pinholes, some scratches and IC porosity became undetectable after testing. (2) The cathode covering the popout defects is intact without any cracking or delamination. (3) The scratch was covered by cathode in pristine state; Ni oxide or NiO-YSZ filled the space where electrolyte was missing. After >10,000h operation, the cathode that used to cover the scratch was missing but the size of the scratch remains unchanged; there is a fairly dense Ni oxide layer on top of the scratch. (4) All of the cracks remain their original dimensions. Current collector kept longitudinal cracks closed which mitigated the adversary effect of crack on cell performance/durability. (5) The crack at open end showed some deterioration manifested by a small oxidation zone in the vicinity of the crack. Wide spectrum of imaging techniques were surveyed and evaluated to detect tubular SOFC defects. Optical methods, with and without assistance of expression technique, were applied to detect all kinds of defects. Thermograph method in combination of CO2 leak, ultra-sound and electric potential was also extensively evaluated to detect surface and under-surface defects such as short circuit. Based on these works, laser camera has been selected to build the final QC machine vision inspection system. The Gocator laser profile sensor was setup to generate 2D surface intensity images of tubular SOFC surface which were analyzed with Sherlock machine vision software supplied by Teledyne Dalsa. The Gocator output was integrated with the Sherlock software and enabled live images to be imported by a Common Vision Blox driver. The Sherlock software was programmed to detect various surface defects from the surface intensity images generated by the Gocator. These measurements are subsequently used to make pass or fail decisions.

01 COAL, LIGNITE, AND PEAT↗

Automated bone segmentation from dental CBCT images using patch-based sparse representation and convex optimization

Purpose: Cone-beam computed tomography (CBCT) is an increasingly utilized imaging modality for the diagnosis and treatment planning of the patients with craniomaxillofacial (CMF) deformities. Accurate segmentation of CBCT image is an essential step to generate three-dimensional (3D) models for the diagnosis and treatment planning of the patients with CMF deformities. However, due to the poor image quality, including very low signal-to-noise ratio and the widespread image artifacts such as noise, beam hardening, and inhomogeneity, it is challenging to segment the CBCT images. In this paper, the authors present a new automatic segmentation method to address these problems. Methods: To segment CBCT images, the authors propose a new method for fully automated CBCT segmentation by using patch-based sparse representation to (1) segment bony structures from the soft tissues and (2) further separate the mandible from the maxilla. Specifically, a region-specific registration strategy is first proposed to warp all the atlases to the current testing subject and then a sparse-based label propagation strategy is employed to estimate a patient-specific atlas from all aligned atlases. Finally, the patient-specific atlas is integrated into amaximum a posteriori probability-based convex segmentation framework for accurate segmentation. Results: The proposed method has been evaluated on a dataset with 15 CBCT images. The effectiveness of the proposed region-specific registration strategy and patient-specific atlas has been validated by comparing with the traditional registration strategy and population-based atlas. The experimental results show that the proposed method achieves the best segmentation accuracy by comparison with other state-of-the-art segmentation methods. Conclusions: The authors have proposed a new CBCT segmentation method by using patch-based sparse representation and convex optimization, which can achieve considerably accurate segmentation results in CBCT segmentation based on 15 patients.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Assessment of Spatial Uncertainties in the Radiotherapy Process With the Novalis System

Purpose: The purpose of this study was to evaluate the accuracy of a new version of the ExacTrac X-ray (ETX) system with statistical analysis retrospectively in order to determine the tolerance of systematic components of spatial uncertainties with the Novalis system. Methods and Materials: Three factors of geometrical accuracy related to the ETX system were evaluated by phantom studies. First, location dependency of the detection ability of the infrared system was evaluated. Second, accuracy of the automated calculation by the image fusion algorithm in the patient registration software was evaluated. Third, deviation of the coordinate scale between the ETX isocenter and the mechanical isocenter was evaluated. From the values of these examinations and clinical experiences, the total spatial uncertainty with the Novalis system was evaluated. Results: As to the location dependency of the detection ability of the infrared system, the detection errors between the actual position and the detected position were 1% in translation shift and 0.1{sup o} in rotational angle, respectively. As to the accuracy of patient verification software, the repeatability and the coincidence of the calculation value by image fusion were good when the contrast of the X-ray image was high. The deviation of coordinates between the ETX isocenter and the mechanical isocenter was 0.313 {+-} 0.024 mm, in a suitable procedure. Conclusions: The spatial uncertainty will be less than 2 mm when suitable treatment planning, optimal patient setup, and daily quality assurance for the Novalis system are achieved in the routine workload.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

WE-D-9A-02: Automated Landmark-Guided CT to Cone-Beam CT Deformable Image Registration

Purpose: The anatomical changes that occur between the simulation CT and daily cone-beam CT (CBCT) are investigated using an automated landmark-guided deformable image registration (LDIR) algorithm with simultaneous intensity correction. LDIR was designed to be accurate in the presence of tissue intensity mismatch and heavy noise contamination. Method: An auto-landmark generation algorithm was used in conjunction with a local small volume (LSV) gradient matching search engine to map corresponding landmarks between the CBCT and planning CT. The LSVs offsets were used to perform an initial deformation, generate landmarks, and correct local intensity mismatch. The landmarks act as stabilizing controlpoints in the Demons objective function. The accuracy of the LDIR algorithm was evaluated on one synthetic case with ground truth and data of ten head and neck cancer patients. The deformation vector field (DVF) accuracy was accessed using a synthetic case. The Root mean square error of the 3D canny edge (RMSECE), mutual information (MI), and feature similarity index metric (FSIM) were used to access the accuracy of LDIR on the patient data. The quality of the corresponding deformed contours was verified by an attending physician. Results: The resulting 90 percentile DVF error for the synthetic case was within 5.63mm for the original demons algorithm, 2.84mm for intensity correction alone, 2.45mm using controlpoints without intensity correction, and 1.48 mm for the LDIR algorithm. For the five patients the mean RMSECE of the original CT, Demons deformed CT, intensity corrected Demons CT, control-point stabilized deformed CT, and LDIR CT was 0.24, 0.26, 0.20, 0.20, and 0.16 respectively. Conclusion: LDIR is accurate in the presence of multimodal intensity mismatch and CBCT noise contamination. Since LDIR is GPU based it can be implemented with minimal additional strain on clinical resources. This project has been supported by a CPRIT individual investigator award RP11032.

60 APPLIED LIFE SCIENCES↗

Reconstruction of a time-averaged midposition CT scan for radiotherapy planning of lung cancer patients using deformable registration

Purpose: lower lobe lung tumors move with amplitudes of up to 2 cm due to respiration. To reduce respiration imaging artifacts in planning CT scans, 4D imaging techniques are used. Currently, we use a single (midventilation) frame of the 4D data set for clinical delineation of structures and radiotherapy planning. A single frame, however, often contains artifacts due to breathing irregularities, and is noisier than a conventional CT scan since the exposure per frame is lower. Moreover, the tumor may be displaced from the mean tumor position due to hysteresis. The aim of this work is to develop a framework for the acquisition of a good quality scan representing all scanned anatomy in the mean position by averaging transformed (deformed) CT frames, i.e., canceling out motion. A nonrigid registration method is necessary since motion varies over the lung. Methods and Materials: 4D and inspiration breath-hold (BH) CT scans were acquired for 13 patients. An iterative multiscale motion estimation technique was applied to the 4D CT scan, similar to optical flow but using image phase (gray-value transitions from bright to dark and vice versa) instead. From the (4D) deformation vector field (DVF) derived, the local mean position in the respiratory cycle was computed and the 4D DVF was modified to deform all structures of the original 4D CT scan to this mean position. A 3D midposition (MidP) CT scan was then obtained by (arithmetic or median) averaging of the deformed 4D CT scan. Image registration accuracy, tumor shape deviation with respect to the BH CT scan, and noise were determined to evaluate the image fidelity of the MidP CT scan and the performance of the technique. Results: Accuracy of the used deformable image registration method was comparable to established automated locally rigid registration and to manual landmark registration (average difference to both methods <0.5 mm for all directions) for the tumor region. From visual assessment, the registration was good for the clearly visible features (e.g., tumor and diaphragm). The shape of the tumor, with respect to that of the BH CT scan, was better represented by the MidP reconstructions than any of the 4D CT frames (including MidV; reduction of 'shape differences' was 66%). The MidP scans contained about one-third the noise of individual 4D CT scan frames. Conclusions: We implemented an accurate method to estimate the motion of structures in a 4D CT scan. Subsequently, a novel method to create a midposition CT scan (time-weighted average of the anatomy) for treatment planning with reduced noise and artifacts was introduced. Tumor shape and position in the MidP CT scan represents that of the BH CT scan better than MidV CT scan and, therefore, was found to be appropriate for treatment planning.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

SU-E-CAMPUS-T-01: Automation of the Winston-Lutz Test for Stereotactic Radiosurgery

Purpose: To optimize clinical efficiency and shorten patient wait time by minimizing the time and effort required to perform the Winston-Lutz test before stereotactic radiosurgery (SRS) through automation of the delivery, analysis, and documentation of results. Methods: The radiation fields of the Winston-Lutz (WL) test were created in a “machine-QA patient” saved in ARIA for use before SRS cases. Images of the BRW target ball placed at mechanical isocenter are captured with the portal imager for each of four, 2cm×2cm, MLC-shaped beams. When the WL plan is delivered and closed, this event is detected by in-house software called EventNet which automates subsequent processes with the aid of the ARIA web services. Images are automatically retrieved from the ARIA database and analyzed to determine the offset of the target ball from radiation isocenter. The results are posted to a website and a composite summary image of the results is pushed back into ImageBrowser for review and authenticated documentation. Results: The total time to perform the test was reduced from 20-25 minutes to less than 4 minutes. The results were found to be more accurate and consistent than the previous method which used radiochromic film. The images were also analyzed with DoseLab for comparison. The difference between the film and automated WL results in the X and Y direction and the radius were (−0.17 +/− 0.28) mm, (0.21 +/− 0.20) mm and (−0.14 +/− 0.27) mm, respectively. The difference between the DoseLab and automated WL results were (−0.05 +/− 0.06) mm, (−0.01 +/− 0.02) mm and (0.01 +/− 0.07) mm, respectively. Conclusions: This process reduced patient wait times by 15–20 minutes making the treatment machine available to treat another patient. Accuracy and consistency of results were improved over the previous method and were comparable to other commercial solutions. Access to the ARIA web services is made possible through an Eclipse co-development agreement with Varian Medical Systems.

60 APPLIED LIFE SCIENCES↗

SU-E-T-642: Safety Procedures for Error Elimination in Cyberknife Stereotactic Radiosurgery (SRS) and Stereotactic Body Radiation Therapy (SBRT)

Purpose: Cyberknife system is used for providing stereotactic radiosurgery (SRS) and stereotactic body radiation therapy (SBRT) hypofractionation scheme. The whole treatment delivery is based on live imaging of the patient. The minor error made at any stage may bring severe radiation injury to the patient or damage to the system itself. Several safety measures were taken to make the system safer. Methods: The radiation treatment provided thru a 6MV linac attached to Kuka robot (Cyberknife G4, Accuray Inc. Sunnyvale, CA, USA). Several possible errors were identified related to patient alignment, treatment planning, dose delivery and physics quality assurance. During dose delivery, manual and visual checks were introduced to confirm pre and intra-treatment imaging to reduce possible errors. One additional step was introduced to confirm that software tracking-tools had worked correctly with highest possible confidence level. Robotic head move in different orientations over and around the patient body, the rigidity of linac-head cover and other accessories was checked periodically. The vender was alerted when a tiny or bigger piece of equipment needed additional interlocked support. Results: As of our experience treating 525 patients on Cyberknife during the last four years, we saw on and off technical issues. During image acquisition, it was made essential to follow the site-specific imaging protocols. Adequate anatomy was contoured to document the respective doses. Followed by auto-segmentation, manual tweaking was performed on every structure. The calculation box was enclosing the whole image during the final calculation. Every plan was evaluated on slice-by slice basis. To review the whole process, a check list was maintained during the physics 2nd-check. Conclusion: The implementation of manual and visual additional checks introduced along with automated checks for confirmation was found promising in terms of reduction in systematic errors and making the system safer than before.

60 APPLIED LIFE SCIENCES↗

Multimodality Image Fusion and Planning and Dose Delivery for Radiation Therapy

Image-guided radiation therapy (IGRT) relies on the quality of fused images to yield accurate and reproducible patient setup prior to dose delivery. The registration of 2 image datasets can be characterized as hardware-based or software-based image fusion. Hardware-based image fusion is performed by hybrid scanners that combine 2 distinct medical imaging modalities such as positron emission tomography (PET) and computed tomography (CT) into a single device. In hybrid scanners, the patient maintains the same position during both studies making the fusion of image data sets simple. However, it cannot perform temporal image registration where image datasets are acquired at different times. On the other hand, software-based image fusion technique can merge image datasets taken at different times or with different medical imaging modalities. Software-based image fusion can be performed either manually, using landmarks, or automatically. In the automatic image fusion method, the best fit is evaluated using mutual information coefficient. Manual image fusion is typically performed at dose planning and for patient setup prior to dose delivery for IGRT. The fusion of orthogonal live radiographic images taken prior to dose delivery to digitally reconstructed radiographs will be presented. Although manual image fusion has been routinely used, the use of fiducial markers has shortened the fusion time. Automated image fusion should be possible for IGRT because the image datasets are derived basically from the same imaging modality, resulting in further shortening the fusion time. The advantages and limitations of both hardware-based and software-based image fusion methodologies are discussed.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Using computer-extracted image features for modeling of error-making patterns in detection of mammographic masses among radiology residents

Purpose: Mammography is the most widely accepted and utilized screening modality for early breast cancer detection. Providing high quality mammography education to radiology trainees is essential, since excellent interpretation skills are needed to ensure the highest benefit of screening mammography for patients. The authors have previously proposed a computer-aided education system based on trainee models. Those models relate human-assessed image characteristics to trainee error. In this study, the authors propose to build trainee models that utilize features automatically extracted from images using computer vision algorithms to predict likelihood of missing each mass by the trainee. This computer vision-based approach to trainee modeling will allow for automatically searching large databases of mammograms in order to identify challenging cases for each trainee. Methods: The authors’ algorithm for predicting the likelihood of missing a mass consists of three steps. First, a mammogram is segmented into air, pectoral muscle, fatty tissue, dense tissue, and mass using automated segmentation algorithms. Second, 43 features are extracted using computer vision algorithms for each abnormality identified by experts. Third, error-making models (classifiers) are applied to predict the likelihood of trainees missing the abnormality based on the extracted features. The models are developed individually for each trainee using his/her previous reading data. The authors evaluated the predictive performance of the proposed algorithm using data from a reader study in which 10 subjects (7 residents and 3 novices) and 3 experts read 100 mammographic cases. Receiver operating characteristic (ROC) methodology was applied for the evaluation. Results: The average area under the ROC curve (AUC) of the error-making models for the task of predicting which masses will be detected and which will be missed was 0.607 (95% CI,0.564-0.650). This value was statistically significantly different from 0.5 (p < 0.0001). For the 7 residents only, the AUC performance of the models was 0.590 (95% CI,0.537-0.642) and was also significantly higher than 0.5 (p = 0.0009). Therefore, generally the authors’ models were able to predict which masses were detected and which were missed better than chance. Conclusions: The authors proposed an algorithm that was able to predict which masses will be detected and which will be missed by each individual trainee. This confirms existence of error-making patterns in the detection of masses among radiology trainees. Furthermore, the proposed methodology will allow for the optimized selection of difficult cases for the trainees in an automatic and efficient manner.

60 APPLIED LIFE SCIENCES↗

Quality control and crop characterization framework for multi-temporal UAV LiDAR data over mechanized agricultural fields

Recent developments in remote sensing are enabling automatic, high resolution, and non-destructive survey of agriculture fields, providing the key basis for advancing plant breeding. Among the used remote sensing modalities, LiDAR has attracted wide attention for its ability to directly provide accurate 3D information. Despite the increasing utilization of LiDAR technology in phenotyping, there is still a lack of effective quality control strategies, in particular, quality control of LiDAR data collected on a multi-temporal basis. This study proposes a targetless framework for multi-temporal LiDAR data quality control and crop characterization in mechanized agricultural fields. Features extracted from the fields – terrain patches and row/alley locations – are utilized for evaluating the vertical and planimetric relative accuracy of the point clouds. Row/alley locations in the field are automatically identified from the point clouds based on the assumption that higher point density and/or higher elevation correspond to plant locations. The performance of the proposed quality control strategies is evaluated using multi-temporal datasets collected in agricultural fields of different sizes, orientation, crops, and growth stages. The result shows that the net vertical and planimetric discrepancies between multi-temporal point clouds are ±3 cm and ±8 cm, respectively. While the former reflects the actual accuracy of the point clouds, the latter is a combined effect of the LiDAR point cloud accuracy, rasterization artifacts, crop type, growth pattern, and wind condition during data acquisition. In terms of row and alley detection, the result shows that the proposed strategy achieves high performance and can deal with different planting orientation, crop types, growth stages, canopy cover, and planting density. In conclusion, this study presents a quality control framework for multi-temporal LiDAR data. Finally, the row and alley detection leads to automated extraction of plots, and hence facilitates the use of remotely sensed data for automated phenotyping.

54 ENVIRONMENTAL SCIENCES↗

SU-E-I-43: Pediatric CT Dose and Image Quality Optimization

Purpose: To design an approach to optimize radiation dose and image quality for pediatric CT imaging, and to evaluate expected performance. Methods: A methodology was designed to quantify relative image quality as a function of CT image acquisition parameters. Image contrast and image noise were used to indicate expected conspicuity of objects, and a wide-cone system was used to minimize scan time for motion avoidance. A decision framework was designed to select acquisition parameters as a weighted combination of image quality and dose. Phantom tests were used to acquire images at multiple techniques to demonstrate expected contrast, noise and dose. Anthropomorphic phantoms with contrast inserts were imaged on a 160mm CT system with tube voltage capabilities as low as 70kVp. Previously acquired clinical images were used in conjunction with simulation tools to emulate images at different tube voltages and currents to assess human observer preferences. Results: Examination of image contrast, noise, dose and tube/generator capabilities indicates a clinical task and object-size dependent optimization. Phantom experiments confirm that system modeling can be used to achieve the desired image quality and noise performance. Observer studies indicate that clinical utilization of this optimization requires a modified approach to achieve the desired performance. Conclusion: This work indicates the potential to optimize radiation dose and image quality for pediatric CT imaging. In addition, the methodology can be used in an automated parameter selection feature that can suggest techniques given a limited number of user inputs. G Stevens and R Singh are employees of GE Healthcare.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Monitoring leaf phenology in moist tropical forests by applying a superpixel-based deep learning method to time-series images of tree canopies

Tropical leaf phenology-particularly its variability at the tree-crown scale-dominates the seasonality of carbon and water fluxes. However, given enormous species diversity, accurate means of monitoring leaf phenology in tropical forests is still lacking. Time series of the Green Chromatic Coordinate (GCC) metric derived from tower-based red-green-blue (RGB) phenocams have been widely used to monitor leaf phenology in temperate forests, but its application in the tropics remains problematic. To improve monitoring of tropical phenology, we explored the use of a deep learning model (i.e. superpixel-based Residual Networks 50, SP-ResNet50) to automatically differentiate leaves from non-leaves in phenocam images and to derive leaf fraction at the tree-crown scale. To evaluate our model, we used a year of data from six phenocams in two contrasting forests in Panama. Here, we first built a comprehensive library of leaf and non-leaf pixels across various acquisition times, exposure conditions and specific phenocams. We then divided this library into training and testing components. We evaluated the model at three levels: 1) superpixel level with a testing set, 2) crown level by comparing the model-derived leaf fractions with those derived using image-specific supervised classification, and 3) temporally using all daily images to assess the diurnal stability of the model-derived leaf fraction. Finally, we compared the model-derived leaf fraction phenology with leaf phenology derived from GCC. Our results show that: 1) the SP-ResNet50 model accurately differentiates leaves from non-leaves (overall accuracy of 93%) and is robust across all three levels of evaluations; 2) the model accurately quantifies leaf fraction phenology across tree-crowns and forest ecosystems; and 3) the combined use of leaf fraction and GCC helps infer the timing of leaf emergence, maturation and senescence, critical information for modeling photosynthetic seasonality of tropical forests. Collectively, this study offers an improved means for automated tropical phenology monitoring using phenocams.

54 ENVIRONMENTAL SCIENCES↗

Real-time biomass feedstock particle quality detection using image analysis and machine vision

Abstract A common and costly challenge in the nascent biorefinery industry is the consistent handling and conveyance of biomass feedstock materials, which can vary widely in their chemical, physical, and mechanical properties. Solutions to cope with varying feedstock qualities will be required, including advanced process controls to adjust equipment and reject feedstocks that do not meet a quality standard. In this work, we present and evaluate methods to autonomously assess corn stover feedstock quality in real time and provide data to process controls with low-cost camera hardware. We explore the use of neural networks to classify feedstocks based on actual processing behavior and pixel matrix feature parameterization to further assess particle attributes that may explain the variable processing behavior. We used the pretrained ResNet neural network coupled with a gated recurrent unit (GRU) time-series classifier trained on our image data, resulting in binary classification of feedstock anomalies with favorable performance. The textural aspects of the image data were statistically analyzed to determine if the textural features were predictive of operational disruptions. The significant textural features were angular second moment, prominence, mean height of surface profile, mean resultant vector, shade, skewness, variation of the polar facet orientation, and direction of azimuthal facets. Expansion of these models is recommended across a wider variety of labeled feedstock images of different qualities and species to develop a more robust tool that may be deployed using low-cost cameras within biorefineries.

09 BIOMASS FUELS↗

Application of a deep learning semantic segmentation model to helium bubbles and voids in nuclear materials

Imaging nanoscale radiation-induced defects using the transmission electron microscope (TEM) is a key factor in the successful implementation of materials for nuclear energy structural applications. Analyzing each defect in a TEM micrograph is currently a manual task. To identify the defects in a single image can take anywhere from 15 min to an hour and a project can require the analysis of anywhere from tens to ≥ 100 images. Here, we use artificial intelligence (AI) models to automate this task. For simplification, we evaluated images with only a single type of defect; helium bubbles. Additionally, we performed semantic segmentation of these helium bubble defects in electron microscopy images of irradiated FeCrAl alloys using a deep learning DefectSegNet model. This model, which was previously used to classify crystal defects, is inspired by the classic DenseNet and U-Net image segmentation models. It claims high spatial resolution, but has poor performance at object boundaries. Our paper improves the DefectSegNet model’s application by adding two new features. First, the DefectSegNet model is applied not only to perform calculation pixel-wise but also object (or feature) wise. Because object-wise metrics are directly relevant to our final goal of detecting bubbles, whereas pixel-wise classification is only an intermediate step, it is an important part of our study. Second, a distance map loss (DML) function has been added to increase its performance at object boundaries. It is crucial to accurately represent defects boundaries, especially bubbles, in order to track the bubble-induced swelling caused by irradiation. The boundary-focused DML function is also compared to other loss functions like Cross-entropy, Weighted Binary Cross Entropy (WBCE), Dice and Intersection over Union (IOU). Finally, by incorporating new features, we found a marked improvement on segmentation quality and better shape preservation at the boundaries and areas of the bubbles.

42 ENGINEERING↗

Dark energy survey operations: years 4 and 5

The Dark Energy Survey is a “stage III” dark energy experiment, performing an optical imaging survey to measure cos-mological equation of state parameters using four independent methods. The scope and complexity of the survey, originallyscheduled for 525 nights of observing spread over five years and combining a 5000 degree wide survey with a 10 field weeklytime-domain survey, introduced complex strategic and tactical scheduling problems that needed to be addressed. We beginwith an overview of the process used to develop DES strategy and tactics, from the inception of the project, to task forces thatstudied and developed strategy changes over the course of the survey, to the nightly pre-observing meeting in which imme-diate tactical issues were addressed. We then summarize the strategic choices made for each sub-survey, including metrics,scheduling considerations, choice of time domain fields and their sequences of exposures, and wide survey footprint andpointing layout choices. We go on to describe the detailed process that determined which specific exposures were taken atwhich specific times. We give a chronology of the strategic and tactical peculiarities of each year of observing, including theproposal and execution of a sixth year. We give an overview ofobstac, the implementation of the DES scheduler used to sim-ulate and evaluate strategic and tactical options, and automate exposure scheduling; and describe developments inobstacfor use after DES. Appendices describe further details of data quality evaluation,τ, andteff; airmass calculation; and model-ing of the seeing and sky brightness. The significant corpus of DES data indicates that the simple scaling relations for seeingas a function of wavelength and airmass derived from the Kolmogorov turbulence model work adequately for exposureplanning purposes: deviations from these relations are modest in comparison with short time-scale seeing variations

Gruendl, Robert A.↗