SEARCH · Search NASA

Results for “Image Processing”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 793 records · Page 44

All sources NASA NTRS DOE OSTI

DOE OSTI2016

MO-DE-207A-12: Toward Patient-Specific 4DCT Reconstruction Using Adaptive Velocity Binning

Purpose: While 4DCT provides organ/tumor motion information, it often samples data over 10–20 breathing cycles. For patients presenting with compromised pulmonary function, breathing patterns can change over the acquisition time, potentially leading to tumor delineation discrepancies. This work introduces a novel adaptive velocity-modulated binning (AVB) 4DCT algorithm that modulates the reconstruction based on the respiratory waveform, yielding a patient-specific 4DCT solution. Methods: AVB was implemented in a research reconstruction configuration. After filtering the respiratory waveform, the algorithm examines neighboring data to a phase reconstruction point and the temporal gate is widened until the difference between the reconstruction point and waveform exceeds a threshold value—defined as percent difference between maximum/minimum waveform amplitude. The algorithm only impacts reconstruction if the gate width exceeds a set minimum temporal width required for accurate reconstruction. A sensitivity experiment of threshold values (0.5, 1, 5, 10, and 12%) was conducted to examine the interplay between threshold, signal to noise ratio (SNR), and image sharpness for phantom and several patient 4DCT cases using ten-phase reconstructions. Individual phase reconstructions were examined. Subtraction images and regions of interest were compared to quantify changes in SNR. Results: AVB increased signal in reconstructed 4DCT slices for respiratory waveforms that met the prescribed criteria. For the end-exhale phases, where the respiratory velocity is low, patient data revealed a threshold of 0.5% demonstrated increased SNR in the AVB reconstructions. For intermediate breathing phases, threshold values were required to be >10% to notice appreciable changes in CT intensity with AVB. AVB reconstructions exhibited appreciably higher SNR and reduced noise in regions of interest that were photon deprived such as the liver. Conclusion: We demonstrated that patient-specific velocity-based 4DCT reconstruction is feasible. Image noise was reduced with AVB, suggesting potential applications for low-dose acquisitions and to improve 4DCT reconstruction for irregular breathing patients. The submitting institution holds research agreements with Philips Healthcare.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

TU-A-201-01: Introduction to In-Room Imaging System Characteristics

Recent years have seen a widespread proliferation of available in-room image guidance systems for radiation therapy target localization with many centers having multiple in-room options. In this session, available imaging systems for in-room IGRT will be reviewed highlighting the main differences in workflow efficiency, targeting accuracy and image quality as it relates to target visualization. Decision-making strategies for integrating these tools into clinical image guidance protocols that are tailored to specific disease sites like H&N, lung, pelvis, and spine SBRT will be discussed. Learning Objectives: Major system characteristics of a wide range of available in-room imaging systems for IGRT. Advantages / disadvantages of different systems for site-specific IGRT considerations. Concepts of targeting accuracy and time efficiency in designing clinical imaging protocols.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

TU-A-201-02: Treatment Site-Specific Considerations for Clinical IGRT

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

TU-AB-BRA-02: An Efficient Atlas-Based Synthetic CT Generation Method

Purpose: A major obstacle for MR-only radiotherapy is the need to generate an accurate synthetic CT (sCT) from MR image(s) of a patient for the purposes of dose calculation and DRR generation. We propose here an accurate and efficient atlas-based sCT generation method, which has a computation speed largely independent of the number of atlases used. Methods: Atlas-based sCT generation requires a set of atlases with co-registered CT and MR images. Unlike existing methods that align each atlas to the new patient independently, we first create an average atlas and pre-align every atlas to the average atlas space. When a new patient arrives, we compute only one deformable image registration to align the patient MR image to the average atlas, which indirectly aligns the patient to all pre-aligned atlases. A patch-based non-local weighted fusion is performed in the average atlas space to generate the sCT for the patient, which is then warped back to the original patient space. We further adapt a PatchMatch algorithm that can quickly find top matches between patches of the patient image and all atlas images, which makes the patch fusion step also independent of the number of atlases used. Results: Nineteen brain tumour patients with both CT and T1-weighted MR images are used as testing data and a leave-one-out validation is performed. Each sCT generated is compared against the original CT image of the same patient on a voxel-by-voxel basis. The proposed method produces a mean absolute error (MAE) of 98.6±26.9 HU overall. The accuracy is comparable with a conventional implementation scheme, but the computation time is reduced from over an hour to four minutes. Conclusion: An average atlas space patch fusion approach can produce highly accurate sCT estimations very efficiently. Further validation on dose computation accuracy and using a larger patient cohort is warranted. The author is a full time employee of Elekta, Inc.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

TU-AB-202-03: Prediction of PET Transfer Uncertainty by DIR Error Estimating Software, AUTODIRECT

Purpose: Deformable image registration (DIR) is a powerful tool, but DIR errors can adversely affect its clinical applications. To estimate voxel-specific DIR uncertainty, a software tool, called AUTODIRECT (automated DIR evaluation of confidence tool), has been developed and validated. This work tests the ability of this software to predict uncertainty for the transfer of standard uptake values (SUV) from positron-emission tomography (PET) with DIR. Methods: Virtual phantoms are used for this study. Each phantom has a planning computed tomography (CT) image and a diagnostic PET-CT image set. A deformation was digitally applied to the diagnostic CT to create the planning CT image and establish a known deformation between the images. One lung and three rectum patient datasets were employed to create the virtual phantoms. Both of these sites have difficult deformation scenarios associated with them, which can affect DIR accuracy (lung tissue sliding and changes in rectal filling). The virtual phantoms were created to simulate these scenarios by introducing discontinuities in the deformation field at the lung rectum border. The DIR algorithm from Plastimatch software was applied to these phantoms. The SUV mapping errors from the DIR were then compared to that predicted by AUTODIRECT. Results: The SUV error distributions closely followed the AUTODIRECT predicted error distribution for the 4 test cases. The minimum and maximum PET SUVs were produced from AUTODIRECT at 95% confidence interval before applying gradient-based SUV segmentation for each of these volumes. Notably, 93.5% of the target volume warped by the true deformation was included within the AUTODIRECT-predicted maximum SUV volume after the segmentation, while 78.9% of the target volume was within the target volume warped by Plastimatch. Conclusion: The AUTODIRECT framework is able to predict PET transfer uncertainty caused by DIR, which enables an understanding of the associated target volume uncertainty.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

TU-FG-209-05: Demonstration of the Line Focus Principle Using the Generalized Measured-Relative Object Detectability (GM-ROD) Metric

Purpose: Demonstrate and quantify the augmented resolution due to focalspot size decrease in images acquired on the anode side of the field, for both small and medium (0.3 and 0.6mm) focal-spot sizes using the experimental task-based GM-ROD metric. Theoretical calculations have shown that a medium focal-spot can achieve the resolution of a small focal-spot if acquired with a tilted anode, effectively providing a higher-output small focal-spot. Methods: The MAF-CMOS (micro-angiographic fluoroscopic complementary-metal-oxide semiconductor) detector (75µm pixel pitch) imaged two copper wire segments of different diameter and a pipeline stent at the central axis and on the anode side of the beam, achieved by tilting the x-ray C-arm (Toshiba Infinix) to 6° and realigning the detector with the perpendicular ray to correct for x-ray obliquity. The relative gain in resolution was determined using the GM-ROD metric, which compares images on the basis of the Fourier transform of the image and the measured NNPS. To emphasize the geometric unsharpness, images were acquired at a magnification of two. Results: Images acquired on the anode side were compared to those acquired on the central axis with the same target-area focal-spot to consider the effect of an angled tube, and for all three objects the advantage of the smaller effective focal-spot was clear, showing a maximum improvement of 36% in GM-ROD. The images obtained with the small focal-spot at the central axis were compared to those of the medium focal-spot at the anode side and, for all objects, the relative performance was comparable. Conclusion: For three objects, the GM-ROD demonstrated the advantage of the anode side focal-spot. The comparable performance of the medium focal-spot on the anode side will allow for a high-output small focal-spot; a necessity in endovascular image-guided interventions. Partial support from an NIH grant R01EB002873 and an equipment grant from Toshiba Medical Systems Corp.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

TU-FG-209-08: Distribution of the Deviation Index (DI) in Digital Radiography Practices Across the United States

Purpose: To characterize the distribution of the deviation index (DI) in digital radiography practices across the United States. Methods: DI data was obtained from 10 collaborating institutions in the United States between 2012 and 2015. Each institution complied with the requirements of the Institutional Review Board at their site. DI data from radiographs of the body parts chest, abdomen, pelvis and extremity were analyzed for anteroposterior, posteroanterior, lateral, and decubitus views. The DI data was analyzed both in aggregate and stratified by exposure control method, image receptor technology, patient age, and participating site for each body part and view. The number of exposures with DI falling within previously published control limits for DI and descriptive statistics were calculated. Results: DI data from 505,930 radiographic exposures was analyzed. The number of exposures with DI falling within published control limits for DI varied from 10 to 20% for adult patients and 10 to 23% for pediatric patients for different body parts and views. Mean DI values averaged over other parameters for radiographs of the abdomen, chest, pelvis, and extremities ranged from 0.3 to 1.0, −0.6 to 0.5, 0.8, and −0.9 to 0.5 for the different adult views and ranged from −1.6 to −0.1, −0.3 to 0.5, −0.1, −0.2 to 1.4 for the different pediatric views, respectively (DI data was solicited only for anteroposterior view of pelvis). Standard deviation values of DI from individual sites ranged from 1.3 to 3.6 and 1.3 to 3.0 for the different adult and pediatric views, respectively. Also of interest was that target exposure indicators varied by up to a factor of 6 between sites for certain body parts and views. Conclusion: Previously published DI control limits do not reflect the state of clinical practice in digital radiography. Mean DI and target exposure indicators are targets for quality improvement efforts in radiography.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

WE-AB-BRA-07: Operating Room Quality Assurance (ORQA) for Spine Surgery Using Known-Component 3D-2D Image Registration

Purpose: Intraoperative x-ray radiography/fluoroscopy is commonly used to qualitatively assess delivery of surgical devices (e.g., spine pedicle screws) but can fail to reliably detect suboptimal placement (e.g., breach of adjacent critical structures). We present a method wherein prior knowledge of the patient and surgical components is leveraged to match preoperative CT and intraoperative radiographs for quantitative assessment of 3D pose. The method presents a new means of operating room quantitative quality assurance (ORQA) that could improve quality and safety, and reduce the frequency of revision surgeries. Methods: The algorithm (known-component registration, KC-Reg) uses patient-specific preoperative CT and parametrically defined surgical component models within a robust 3D-2D registration method to iteratively optimize gradient similarity using the covariance matrix adaptation evolution strategy. Advances from previous work address key challenges to clinical translation: i) absolving the need for offline geometric calibration of the C-arm; and ii) solving multiple component bodies simultaneously, thereby allowing QA in a single step (e.g., spinal construct with 4–20 screws), rather than sequential QA of each component. Performance was tested in a spine phantom with 10 pedicle screws, and first results from clinical studies are reported. Results: Phantom experiments demonstrated median target registration error (TRE) of (1.0±0.3) mm at the screw tip and (0.7°±0.4°) in angulation. The simultaneous multi-body registration approach improved TRE from the previous (sequential) method by 42%, reduced outliers, and fits into the natural workflow. Initial application of KC-Reg in clinical data shows TRE of (2.5±4.5) mm and (4.7°±0.5°). Conclusion: The KC-Reg algorithm offers a potentially valuable method for quantitative QA of the surgical product, using radiographic systems that are already within the surgical arsenal. For spine surgery, the method offers a near-real-time independent check on the quality of surgical product, facilitating immediate revision if necessary and potentially avoiding postoperative morbidity and/or revision surgery. Gerhard Kleinszig and Sebastian Vogt are employees of Siemens Healthcare.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

WE-AB-202-06: Correlating Lung CT HU with Transformation-Based and Xe-CT Derived Ventilation

Purpose: Regional lung ventilation is useful to reduce radiation-induced function damage during lung cancer radiation therapy. Recently a new direct HU (Hounsfield unit)-based method was proposed to estimate the ventilation potential without image registration. The purpose of this study is to examine if there is a functional dependence between HU values and transformation-based or Xe-CT derived ventilation. Methods: 4DCT images acquired from 13 patients prior to radiation therapy and 4 mechanically ventilated sheep subjects which also have associated Xe-CT images were used for this analysis. Transformation-based ventilation was computed using Jacobian determinant of the transformation field between peak-exhale and peak-inhale 4DCT images. Both transformation and Xe-CT derived ventilation was computed for each HU bin. Color scatter plot and cumulative histogram were used to compare and validate the direct HU-based method. Results: There was little change of the center and shape of the HU histograms between free breathing CT and 4DCT average, with or without smoothing, and between the repeated 4DCT scans. HU of −750 and −630 were found to have the greatest transformation-based ventilation for human and sheep subjects, respectively. Maximum Xe-CT derived ventilation was found to locate at HU of −600 in sheep subjects. The curve between Xe-CT ventilation and HU was noisy for tissue above HU −400, possibly due to less intensity change of Xe gas during wash-out and wash-in phases. Conclusion: Both transformation-based and Xe-CT ventilation demonstrated that lung tissues with HU values in the range of (-750, −600) HU have the maximum ventilation potential. The correlation between HU and ventilation suggests that HU might be used to help guide the ventilation calculation and make it more robust to noise and image registration errors. Research support from NIH grants CA166703 and CA166119 and a gift from Roger Koch.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

WE-AB-202-08: Feasibility of Single-Inhalation/Single-Energy Xenon CT for High-Resolution Imaging of Regional Lung Ventilation in Humans

Purpose: To demonstrate the efficacy of a novel functional lung imaging method that utilizes single-inhalation, single-energy xenon CT (Xe-CT) lung ventilation scans, and to compare it against the current clinical standard, ventilation single-photon emission CT (V-SPECT). Methods: In an IRB-approved clinical study, 14 patients undergoing thoracic radiotherapy received two successive single inhalation, single energy (80keV) CT images of the entire lung using 100% oxygen and a 70%/30% xenon-oxygen mixture. A subset of ten patients also received concurrent SPECT ventilation scans. Anatomic reproducibility between the two scans was achieved using a custom video biofeedback apparatus. The CT images were registered to each other by deformable registration, and a calculated difference image served as surrogate xenon ventilation map. Both lungs were partitioned into twelve sectors, and a sector-wise correlation was performed between the xenon and V-SPECT scans. A linear regression model was developed with forced expiratory volume (FEV) as a predictor and the coefficient of variation (CoV) as the outcome. Results: The ventilation comparison for five of the patients had either moderate to strong Pearson correlation coefficients (0.47 to 0.69, p<0.05). Of these, four also had moderate to strong Spearman correlation coefficients (0.46 to 0.80, p<0.03). The patients with the strongest correlation had clear regional ventilation deficits. The patient comparisons with the weakest correlations had more homogeneous ventilation distributions, and those patients also had diminished lung function as assessed by spirometry. Analysis of the relationship between CoV and FEV yielded a non-significant trend toward negative correlation (Pearson coefficient −0.60, p<0.15). Conclusion: Significant correlations were found between the Xe-CT and V-SPECT ventilation imagery. The results from this small cohort of patients indicate that single inhalation, single energy Xe-CT has the potential to quantify regional lung ventilation volumetrically with high resolution using widely accessible radiologic equipment. Bill Loo and Peter Maxim are founders of TibaRay, Inc. Bill Loo is also a board member. Bill Loo and Peter Maxim have received research grants from Varian Medical Systems, Inc. and RaySearch Laboratory.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

WE-AB-207A-07: A Planning CT-Guided Scatter Artifact Correction Method for CBCT Images

Purpose: Cone beam computed tomography (CBCT) imaging is on increasing demand for high-performance image-guided radiotherapy such as online tumor delineation and dose calculation. However, the current CBCT imaging has severe scatter artifacts and its current clinical application is therefore limited to patient setup based mainly on the bony structures. This study’s purpose is to develop a CBCT artifact correction method. Methods: The proposed scatter correction method utilizes the planning CT to improve CBCT image quality. First, an image registration is used to match the planning CT with the CBCT to reduce the geometry difference between the two images. Then, the planning CT-based prior information is entered into the Bayesian deconvolution framework to iteratively perform a scatter artifact correction for the CBCT mages. This technique was evaluated using Catphan phantoms with multiple inserts. Contrast-to-noise ratios (CNR) and signal-to-noise ratios (SNR), and the image spatial nonuniformity (ISN) in selected volume of interests (VOIs) were calculated to assess the proposed correction method. Results: Post scatter correction, the CNR increased by a factor of 1.96, 3.22, 3.20, 3.46, 3.44, 1.97 and 1.65, and the SNR increased by a factor 1.05, 2.09, 1.71, 3.95, 2.52, 1.54 and 1.84 for the Air, PMP, LDPE, Polystryrene, Acrylic, Delrin and Teflon inserts, respectively. The ISN decreased from 21.1% to 4.7% in the corrected images. All values of CNR, SNR and ISN in the corrected CBCT image were much closer to those in the planning CT images. The results demonstrated that the proposed method reduces the relevant artifacts and recovers CT numbers. Conclusion: We have developed a novel CBCT artifact correction method based on CT image, and demonstrated that the proposed CT-guided correction method could significantly reduce scatter artifacts and improve the image quality. This method has great potential to correct CBCT images allowing its use in adaptive radiotherapy.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

WE-AB-207A-08: BEST IN PHYSICS (IMAGING): Advanced Scatter Correction and Iterative Reconstruction for Improved Cone-Beam CT Imaging On the TrueBeam Radiotherapy Machine

Purpose: To improve CBCT image quality for image-guided radiotherapy by applying advanced reconstruction algorithms to overcome scatter, noise, and artifact limitations Methods: CBCT is used extensively for patient setup in radiotherapy. However, image quality generally falls short of diagnostic CT, limiting soft-tissue based positioning and potential applications such as adaptive radiotherapy. The conventional TrueBeam CBCT reconstructor uses a basic scatter correction and FDK reconstruction, resulting in residual scatter artifacts, suboptimal image noise characteristics, and other artifacts like cone-beam artifacts. We have developed an advanced scatter correction that uses a finite-element solver (AcurosCTS) to model the behavior of photons as they pass (and scatter) through the object. Furthermore, iterative reconstruction is applied to the scatter-corrected projections, enforcing data consistency with statistical weighting and applying an edge-preserving image regularizer to reduce image noise. The combined algorithms have been implemented on a GPU. CBCT projections from clinically operating TrueBeam systems have been used to compare image quality between the conventional and improved reconstruction methods. Planning CT images of the same patients have also been compared. Results: The advanced scatter correction removes shading and inhomogeneity artifacts, reducing the scatter artifact from 99.5 HU to 13.7 HU in a typical pelvis case. Iterative reconstruction provides further benefit by reducing image noise and eliminating streak artifacts, thereby improving soft-tissue visualization. In a clinical head and pelvis CBCT, the noise was reduced by 43% and 48%, respectively, with no change in spatial resolution (assessed visually). Additional benefits include reduction of cone-beam artifacts and reduction of metal artifacts due to intrinsic downweighting of corrupted rays. Conclusion: The combination of an advanced scatter correction with iterative reconstruction substantially improves CBCT image quality. It is anticipated that clinically acceptable reconstruction times will result from a multi-GPU implementation (the algorithms are under active development and not yet commercially available). All authors are employees of and (may) own stock of Varian Medical Systems.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

WE-DE-BRA-07: Megavoltage Spectral Imaging with a Layered Detector

Purpose: The aim of the current work is to investigate the feasibility of megavoltage spectral imaging using a multiple layered detector for enhancement of low contrast detectability through material segmentation and discrimination (such as bone, markers and metal implants). Potentially the technique can be applied to improve detection and reduce dose in Megavoltage Cone Beam Computed Tomography (MV-CBCT). Methods: Experiments were performed with a prototype multi-layer imager (MLI) which has higher detective efficiency and lower noise characteristics than conventional Electronic Portal Imaging Devices (EPIDs). Images of a solid water phantom were acquired at 2.5 MV, 6MV and 6MV without flattening filter (FFF). The following materials were placed within a stack of solid water: aluminum, copper and gold. Material separation was assessed based on Contrast-to-Noise Ratio (CNR) of the weighted image, formed by a weighted subtraction of the images from two layers of the MLI. A range of weighting factors were investigated for material separation. Results: CNR can be minimized for each material by appropriate selection of the subtraction weighting factor. This is equivalent to a selective subtraction of specific materials from the image. Using multiple layers simultaneously also decreases the dose requirement and removes any registration errors. The minimum CNR for aluminum, copper and gold at the weighted image formed with 2.5MV was obtained at weighting factors equal to 0.92, 0.76 and 0.64 respectively. The corresponding values at 6MVFFF were 0.99, 0.92 and 0.78 respectively. Conclusion: In the current work, an MV spectral imaging feasibility study was attempted using a novel multi-layer prototype EPID imager. Initial results suggest that material separation based on spectral differences between different layers is possible. This spectral imaging technique has potential advantages in MV-CBCT for real-time target tracking, patient set-up imaging and adaptive radiotherapy. The project was supported, partially, by a grant from Varian Medical Systems, Inc., and Award No. R01CA188446-01 from the National Cancer Institute. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

TH-CD-202-02: A Preliminary Study Evaluating Beam-Hardening Artifact Reduction On CT Direct Electron-Density Images

Purpose: A prototype reconstruction algorithm that can provide direct electron density (ED) images from single energy CT scans is being currently developed by Siemens Healthcare GmbH. This feature can eliminate the need for kV specific calibration curve for radiation treatemnt planning. An added benefit is that beam-hardening artifacts are also reduced on direct-ED images due to the underlying material decomposition. This study is to quantitatively analyze the reduction of beam-hardening artifacts on direct-ED images and suggest additional clinical usages. Methods: HU and direct-ED images were reconstructed on a head phantom scanned on a Siemens Definition AS CT scanner at five tube potentials of 70kV, 80kV, 100kV, 120kV and 140kV respectively. From these images, mean, standard deviation (SD), and local NPS were calculated for regions of interest (ROI) of same locations and sizes. A complete analysis of beam-hardening artifact reduction and image quality improvement was conducted. Results: Along with the increase of tube potentials, ROI means and SDs decrease on both HU and direct-ED images. The mean value differences between HU and direct-ED images are up to 8% with absolute value of 2.9. Compared to that on HU images, the SDs are lower on direct-ED images, and the differences are up to 26%. Interestingly, the local NPS calculated from direct-ED images shows consistent values in the low spatial frequency domain for images acquired from all tube potential settings, while varied dramatically on HU images. This also confirms the beam -hardening artifact reduction on ED images. Conclusions: The low SDs on direct-ED images and relative consistent NPS values in the low spatial frequency domain indicate a reduction of beam-hardening artifacts. The direct-ED image has the potential to assist in more accurate organ contouring, and is a better fit for the desired purpose of CT simulations for radiotherapy.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

TH-CD-202-10: Tumor Metabolic Control Probability & Dose Response Mapping for Adaptive Dose Painting by Number

Purpose: Adaptive dose-painting-by-number (DPbN) requires a dose-response-mapping (DRM) obtained early in the treatment course. To obtain DRM, voxel-by-voxel tumor dose response needs to be quantified. Our recent study has demonstrated that voxel-by-voxel radio-sensitivity of patient tumor can be determined using tumor-metabolic-ratio measured early during the treatment using FDG-PET images. In this study, the measurements were utilized to construct tumor metabolic control probability (TMCP) and DRM for DPbN. Methods: FDG-PET/CT images of 18 HN cancer patients obtained pre- and weekly during the treatment were used. Spatial parametric images of tumor-metabolic-ratio (dSUV) were constructed following voxel-by-voxel deformable image registration. Each voxel value in dSUV was a function of baseline SUV and delivered dose. Utilizing all values of dSUV in the controlled tumor group at the last treatment week, a cut-off function between the baseline SUV and dSUV was formed, and applied in early treatment days on dSUV of all tumors to model the TMCP. At the treatment week k, TMCP was constructed with respect to the tumor voxel dSUV appeared at the week using the maximum likelihood estimation for all dose levels, and used for DRM construction. Results: TMCPs estimated in the week 2 & 3 have D{sub 50}=11.1∼47.6Gy; γ{sub 50}=0.55∼0.92 respectively with respect to dSUV=0.3∼1.2. The corresponding DRM between tumor voxel dSUV and the expected treatment dose has sigmoid shape. The expected treatment dose are 26∼40Gy (for 95% TMCP) for high sensitive tumor voxels with dSUV=0.3∼0.5; and 65∼110Gy for low sensitive tumor voxels with the dSUV>1.0 depending on the time of the estimation. Conclusion: TMCP can be constructed voxel-by-voxel in human tumor using multiple FDG-PET imaging obtained in early treatment days. TMCP provides a potential quantitative objective of tumor DRM for DPbN to plan the best dose, escalate or de-escalate, in tumor adaptively based on its own radio-sensitivity.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

TH-EF-BRA-04: Individually Optimized Contrast-Enhanced 4D-CT for Radiotherapy Simulation in Pancreatic Ductal Adenocarcinoma

Purpose: To develop an individually optimized contrast-enhanced (CE) 4D-CT for radiotherapy simulation in pancreatic ductal adenocarcinomas (PDA). Methods: Ten PDA patients were enrolled. Each underwent 3 CT scans: a 4D-CT immediately following a CE 3D-CT and an individually optimized CE 4D-CT using test injection. Three physicians contoured the tumor and pancreatic tissues. We compared image quality scores, tumor volume, motion, tumor-to-pancreas contrast, and contrast-to-noise ratio (CNR) in the 3 CTs. We also evaluated interobserver variations in contouring the tumor using simultaneous truth and performance level estimation (STAPLE). Results: Average image quality scores for CE 3DCT and CE 4D-CT were comparable (4.0 and 3.8, respectively; P=0.47), and both were significantly better than that for 4D-CT (2.6, P<0.001). Tumor-to-pancreas contrast results were comparable in CE 3D-CT and CE 4D-CT (15.5 and 16.7 HU, respectively; P=0.71), and the latter was significantly higher than in 4D-CT (9.2 HU, P=0.03). Image noise in CE 3D-CT (12.5 HU) was significantly lower than in CE 4D-CT (22.1 HU, P<0.001) and 4D-CT (19.4 HU, P=0.005). CNRs were comparable in CE 3D-CT and CE 4DCT (1.4 and 0.8, respectively; P=0.23), and the former was significantly better than in 4D-CT (0.6, P = 0.04). Mean tumor volumes were smaller in CE 3D-CT (29.8 cm{sup 3}) and CE 4D-CT (22.8 cm{sup 3}) than in 4D-CT (42.0 cm{sup 3}), although these differences were not statistically significant. Mean tumor motion was comparable in 4D-CT and CE 4D-CT (7.2 and 6.2 mm, P=0.23). Interobserver variations were comparable in CE 3D-CT and CE 4D-CT (Jaccard index 66.0% and 61.9%, respectively) and were worse for 4D-CT (55.6%) than CE 3D-CT. Conclusion: CE 4D-CT demonstrated characteristics comparable to CE 3D-CT, with high potential for simultaneously delineating the tumor and quantifying tumor motion with a single scan. Supported in part by Philips Healthcare.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2018

Optimal structure and parameter learning of Ising models

Reconstruction of the structure and parameters of an Ising model from binary samples is a problem of practical importance in a variety of disciplines, ranging from statistical physics and computational biology to image processing and machine learning. The focus of the research community shifted toward developing universal reconstruction algorithms that are both computationally efficient and require the minimal amount of expensive data. Here, we introduce a new method, interaction screening, which accurately estimates model parameters using local optimization problems. The algorithm provably achieves perfect graph structure recovery with an information-theoretically optimal number of samples, notably in the low-temperature regime, which is known to be the hardest for learning. Here, the efficacy of interaction screening is assessed through extensive numerical tests on synthetic Ising models of various topologies with different types of interactions, as well as on real data produced by a D-Wave quantum computer. Finally, this study shows that the interaction screening method is an exact, tractable, and optimal technique that universally solves the inverse Ising problem.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

Femtojoule optical nonlinearity for deep learning with incoherent illumination

Optical neural networks (ONNs) are a promising computational alternative for deep learning due to their inherent massive parallelism for linear operations. However, the development of energy-efficient and highly parallel optical nonlinearities, a critical component in ONNs, remains an outstanding challenge. Here, we introduce a nonlinear optical microdevice array (NOMA) compatible with incoherent illumination by integrating the liquid crystal cell with silicon photodiodes at the single-pixel level. We fabricate NOMA with more than half a million pixels, each functioning as an optical analog of the rectified linear unit at ultralow switching energy down to 100 femtojoules per pixel. With NOMA, we demonstrate an optical multilayer neural network. Our work holds promise for large-scale and low-power deep ONNs, computer vision, and real-time optical image processing.

36 MATERIALS SCIENCE↗