Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Pixel”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

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

At least 1,333 records · Page 74
All sourcesNASA NTRSDOE OSTI
DOE OSTI2015

TU-EF-204-11: Impact of Using Multi-Slice Training Sets On the Performance of a Channelized Hotelling Observer in a Low-Contrast Detection Task in CT

Purpose: To investigate using multiple CT image slices from a single acquisition as independent training images for a channelized Hotelling observer (CHO) model to reduce the number of repeated scans for CHO-based CT image quality assessment. Methods: We applied a previously validated CHO model to detect low contrast disk objects formed from cross-sectional images of three epoxy-resin-based rods (diameters: 3, 5, and 9 mm; length: ∼5cm). The rods were submerged in a 35x 25 cm2 iodine-doped water filled phantom, yielding-15 HU object contrast. The phantom was scanned 100 times with and without the rods present. Scan and reconstruction parameters include: 5 mm slice thickness at 0.5 mm intervals, 120 kV, 480 Quality Reference mAs, and a 128-slice scanner. The CHO’s detectability index was evaluated as a function of factors related to incorporating multi-slice image data: object misalignment along the z-axis, inter-slice pixel correlation, and number of unique slice locations. In each case, the CHO training set was fixed to 100 images. Results: Artificially shifting the object’s center position by as much as 3 pixels in any direction relative to the Gabor channel filters had insignificant impact on object detectability. An inter-slice pixel correlation of >∼0.2 yielded positive bias in the model’s performance. Incorporating multi-slice image data yielded slight negative bias in detectability with increasing number of slices, likely due to physical variations in the objects. However, inclusion of image data from up to 5 slice locations yielded detectability indices within measurement error of the single slice value. Conclusion: For the investigated model and task, incorporating image data from 5 different slice locations of at least 5 mm intervals into the CHO model yielded detectability indices within measurement error of the single slice value. Consequently, this methodology would Result in a 5-fold reduction in number of image acquisitions. This project was supported by National Institutes of Health grants R01 EB017095 and U01 EB017185 from the National Institute of Biomedical Imaging and Bioengineering.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2015

WE-G-204-07: Automated Characterization of Perceptual Quality of Clinical Chest Radiographs: Improvements in Lung, Spine, and Hardware Detection

Purpose: To develop and validate more robust methods for automated lung, spine, and hardware detection in AP/PA chest images. This work is part of a continuing effort to automatically characterize the perceptual image quality of clinical radiographs. [Y. Lin et al. Med. Phys. 39, 7019–7031 (2012)] Methods: Our previous implementation of lung/spine identification was applicable to only one vendor. A more generalized routine was devised based on three primary components: lung boundary detection, fuzzy c-means (FCM) clustering, and a clinically-derived lung pixel probability map. Boundary detection was used to constrain the lung segmentations. FCM clustering produced grayscale- and neighborhood-based pixel classification probabilities which are weighted by the clinically-derived probability maps to generate a final lung segmentation. Lung centerlines were set along the left-right lung midpoints. Spine centerlines were estimated as a weighted average of body contour, lateral lung contour, and intensity-based centerline estimates. Centerline estimation was tested on 900 clinical AP/PA chest radiographs which included inpatient/outpatient, upright/bedside, men/women, and adult/pediatric images from multiple imaging systems. Our previous implementation further did not account for the presence of medical hardware (pacemakers, wires, implants, staples, stents, etc.) potentially biasing image quality analysis. A hardware detection algorithm was developed using a gradient-based thresholding method. The training and testing paradigm used a set of 48 images from which 1920 51×51 pixel{sup 2} ROIs with and 1920 ROIs without hardware were manually selected. Results: Acceptable lung centerlines were generated in 98.7% of radiographs while spine centerlines were acceptable in 99.1% of radiographs. Following threshold optimization, the hardware detection software yielded average true positive and true negative rates of 92.7% and 96.9%, respectively. Conclusion: Updated segmentation and centerline estimation methods in addition to new gradient-based hardware detection software provide improved data integrity control and error-checking for automated clinical chest image quality characterization across multiple radiography systems.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

A novel approach to assess the treatment response using Gaussian random field in PET

Purpose: The assessment of early therapeutic response to anticancer therapy is vital for treatment planning and patient management in clinic. With the development of personal treatment plan, the early treatment response, especially before any anatomically apparent changes after treatment, becomes urgent need in clinic. Positron emission tomography (PET) imaging serves an important role in clinical oncology for tumor detection, staging, and therapy response assessment. Many studies on therapy response involve interpretation of differences between two PET images, usually in terms of standardized uptake values (SUVs). However, the quantitative accuracy of this measurement is limited. This work proposes a statistically robust approach for therapy response assessment based on Gaussian random field (GRF) to provide a statistically more meaningful scale to evaluate therapy effects. Methods: The authors propose a new criterion for therapeutic assessment by incorporating image noise into traditional SUV method. An analytical method based on the approximate expressions of the Fisher information matrix was applied to model the variance of individual pixels in reconstructed images. A zero mean unit variance GRF under the null hypothesis (no response to therapy) was obtained by normalizing each pixel of the post-therapy image with the mean and standard deviation of the pretherapy image. The performance of the proposed method was evaluated by Monte Carlo simulation, where XCAT phantoms (128{sup 2} pixels) with lesions of various diameters (2–6 mm), multiple tumor-to-background contrasts (3–10), and different changes in intensity (6.25%–30%) were used. The receiver operating characteristic curves and the corresponding areas under the curve were computed for both the proposed method and the traditional methods whose figure of merit is the percentage change of SUVs. The formula for the false positive rate (FPR) estimation was developed for the proposed therapy response assessment utilizing local average method based on random field. The accuracy of the estimation was validated in terms of Euler distance and correlation coefficient. Results: It is shown that the performance of therapy response assessment is significantly improved by the introduction of variance with a higher area under the curve (97.3%) than SUVmean (91.4%) and SUVmax (82.0%). In addition, the FPR estimation serves as a good prediction for the specificity of the proposed method, consistent with simulation outcome with ∼1 correlation coefficient. Conclusions: In this work, the authors developed a method to evaluate therapy response from PET images, which were modeled as Gaussian random field. The digital phantom simulations demonstrated that the proposed method achieved a large reduction in statistical variability through incorporating knowledge of the variance of the original Gaussian random field. The proposed method has the potential to enable prediction of early treatment response and shows promise for application to clinical practice. In future work, the authors will report on the robustness of the estimation theory for application to clinical practice of therapy response evaluation, which pertains to binary discrimination tasks at a fixed location in the image such as detection of small and weak lesion.

61 RADIATION PROTECTION AND DOSIMETRY↗
DOE OSTI2016

IMRT QA: Selecting gamma criteria based on error detection sensitivity

Purpose: The gamma comparison is widely used to evaluate the agreement between measurements and treatment planning system calculations in patient-specific intensity modulated radiation therapy (IMRT) quality assurance (QA). However, recent publications have raised concerns about the lack of sensitivity when employing commonly used gamma criteria. Understanding the actual sensitivity of a wide range of different gamma criteria may allow the definition of more meaningful gamma criteria and tolerance limits in IMRT QA. We present a method that allows the quantitative determination of gamma criteria sensitivity to induced errors which can be applied to any unique combination of device, delivery technique, and software utilized in a specific clinic. Methods: A total of 21 DMLC IMRT QA measurements (ArcCHECK®, Sun Nuclear) were compared to QA plan calculations with induced errors. Three scenarios were studied: MU errors, multi-leaf collimator (MLC) errors, and the sensitivity of the gamma comparison to changes in penumbra width. Gamma comparisons were performed between measurements and error-induced calculations using a wide range of gamma criteria, resulting in a total of over 20 000 gamma comparisons. Gamma passing rates for each error class and case were graphed against error magnitude to create error curves in order to represent the range of missed errors in routine IMRT QA using 36 different gamma criteria. Results: This study demonstrates that systematic errors and case-specific errors can be detected by the error curve analysis. Depending on the location of the error curve peak (e.g., not centered about zero), 3%/3 mm threshold = 10% at 90% pixels passing may miss errors as large as 15% MU errors and ±1 cm random MLC errors for some cases. As the dose threshold parameter was increased for a given %Diff/distance-to-agreement (DTA) setting, error sensitivity was increased by up to a factor of two for select cases. This increased sensitivity with increasing dose threshold was consistent across all studied combinations of %Diff/DTA. Criteria such as 2%/3 mm and 3%/2 mm with a 50% threshold at 90% pixels passing are shown to be more appropriately sensitive without being overly strict. However, a broadening of the penumbra by as much as 5 mm in the beam configuration was difficult to detect with commonly used criteria, as well as with the previously mentioned criteria utilizing a threshold of 50%. Conclusions: We have introduced the error curve method, an analysis technique which allows the quantitative determination of gamma criteria sensitivity to induced errors. The application of the error curve method using DMLC IMRT plans measured on the ArcCHECK® device demonstrated that large errors can potentially be missed in IMRT QA with commonly used gamma criteria (e.g., 3%/3 mm, threshold = 10%, 90% pixels passing). Additionally, increasing the dose threshold value can offer dramatic increases in error sensitivity. This approach may allow the selection of more meaningful gamma criteria for IMRT QA and is straightforward to apply to other combinations of devices and treatment techniques.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

An efficient computational approach to model statistical correlations in photon counting x-ray detectors

Purpose: To introduce and evaluate an increment matrix approach (IMA) describing the signal statistics of energy-selective photon counting detectors including spatial–spectral correlations between energy bins of neighboring detector pixels. The importance of the occurring correlations for image-based material decomposition is studied. Methods: An IMA describing the counter increase patterns in a photon counting detector is proposed. This IMA has the potential to decrease the number of required random numbers compared to Monte Carlo simulations by pursuing an approach based on convolutions. To validate and demonstrate the IMA, an approximate semirealistic detector model is provided, simulating a photon counting detector in a simplified manner, e.g., by neglecting count rate-dependent effects. In this way, the spatial–spectral correlations on the detector level are obtained and fed into the IMA. The importance of these correlations in reconstructed energy bin images and the corresponding detector performance in image-based material decomposition is evaluated using a statistically optimal decomposition algorithm. Results: The results of IMA together with the semirealistic detector model were compared to other models and measurements using the spectral response and the energy bin sensitivity, finding a good agreement. Correlations between the different reconstructed energy bin images could be observed, and turned out to be of weak nature. These correlations were found to be not relevant in image-based material decomposition. An even simpler simulation procedure based on the energy bin sensitivity was tested instead and yielded similar results for the image-based material decomposition task, as long as the fact that one incident photon can increase multiple counters across neighboring detector pixels is taken into account. Conclusions: The IMA is computationally efficient as it required about 10{sup 2} random numbers per ray incident on a detector pixel instead of an estimated 10{sup 8} random numbers per ray as Monte Carlo approaches would need. The spatial–spectral correlations as described by IMA are not important for the studied image-based material decomposition task. Respecting the absolute photon counts and thus the multiple counter increases by a single x-ray photon, the same material decomposition performance could be obtained with a simpler detector description using the energy bin sensitivity.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

SU-C-207B-05: Tissue Segmentation of Computed Tomography Images Using a Random Forest Algorithm: A Feasibility Study

Purpose: Develop an automated Random Forest algorithm for tissue segmentation of CT examinations. Methods: Seven materials were classified for segmentation: background, lung/internal gas, fat, muscle, solid organ parenchyma, blood/contrast, and bone using Matlab and the Trainable Weka Segmentation (TWS) plugin of FIJI. The following classifier feature filters of TWS were investigated: minimum, maximum, mean, and variance each evaluated over a pixel radius of 2n, (n = 0–4). Also noise reduction and edge preserving filters, Gaussian, bilateral, Kuwahara, and anisotropic diffusion, were evaluated. The algorithm used 200 trees with 2 features per node. A training data set was established using an anonymized patient’s (male, 20 yr, 72 kg) chest-abdomen-pelvis CT examination. To establish segmentation ground truth, the training data were manually segmented using Eclipse planning software, and an intra-observer reproducibility test was conducted. Six additional patient data sets were segmented based on classifier data generated from the training data. Accuracy of segmentation was determined by calculating the Dice similarity coefficient (DSC) between manual and auto segmented images. Results: The optimized autosegmentation algorithm resulted in 16 features calculated using maximum, mean, variance, and Gaussian blur filters with kernel radii of 1, 2, and 4 pixels, in addition to the original CT number, and Kuwahara filter (linear kernel of 19 pixels). Ground truth had a DSC of 0.94 (range: 0.90–0.99) for adult and 0.92 (range: 0.85–0.99) for pediatric data sets across all seven segmentation classes. The automated algorithm produced segmentation with an average DSC of 0.85 ± 0.04 (range: 0.81–1.00) for the adult patients, and 0.86 ± 0.03 (range: 0.80–0.99) for the pediatric patients. Conclusion: The TWS Random Forest auto-segmentation algorithm was optimized for CT environment, and able to segment seven material classes over a range of body habitus and CT protocol parameters with an average DSC of 0.86 ± 0.04 (range: 0.80–0.99).

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

SU-D-BRA-07: Applications of Combined KV/MV CBCT Imaging with a High-DQE MV Detector

Purpose: To investigate whether a high detection quantum efficiency (DQE) MV detector makes combined kV/MV CBCT clinically practical. Methods: Combined kV/MV CBCT was studied for scan time reduction (STR) and metal artifact reduction (MAR). 6MV CBCT data (dose rate = 0.017 MU/degree) were collected using 1) a novel focused pixelated cadmium tungstate (CWO) scintillator (15mm thickness, DQE(0) = 22%, 0.784mm pixel pitch) coupled to a flat panel imager, and 2) a commercial portal imager with a 133mg/cm{sup 2} gadolinium oxysulfide (GOS) screen (DQE(0) = 1.2%). The 100kVp data were acquired using a commercial imager employing a columnar cesium iodide scintillator (DQE(0) = 70%) with a dose rate of 0.0016 cGy/degree. For STR, MV and kV projections spanning 105° were combined to constitute a complete CBCT scan. Total dose was ∼2cGy and acquisition time was 18s. For MAR, only the metalcorrupted pixels in the kV projections were replaced with MV data resulting in a total dose of less than 1cGy for a 360° scan. Image quality was assessed using an 18-cm diameter electron density phantom with nine tissue inserts, some of which were replaced with steel rods for MAR studies. Results: The CWO contrast-to-noise ratio (CNR) was ∼4.0x higher than the GOS CNR and was ∼4.8x lower than the kV CNR when normalized for dose. When CWO MV data were combined with kV data for STR, all contrast inserts were visible, but only two were detectable in the composite kV/GOS image. Metal artifacts were greatly reduced using the kV/MV MAR technique with all contrast inserts clearly visible in the composite kV/CWO image but only two inserts visible in the composite kV/GOS image. Conclusion: We have demonstrated that a high DQE MV detector significantly improves kV/MV CBCT image quality thus enabling scan time reduction and metal artifact reduction without a severe dose penalty. AW and JS-L are employees of Varian, RF is an employee of Siemens.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

SU-F-SPS-06: Implementation of a Back-Projection Algorithm for 2D in Vivo Dosimetry with An EPID System

Purpose: To implement a back-projection algorithm for 2D dose reconstructions for in vivo dosimetry in radiation therapy using an Electronic Portal Imaging Device (EPID) based on amorphous silicon. Methods: An EPID system was used to calculate dose-response function, pixel sensitivity map, exponential scatter kernels and beam hardenig correction for the back-projection algorithm. All measurements were done with a 6 MV beam. A 2D dose reconstruction for an irradiated water phantom (30×30×30 cm{sup 3}) was done to verify the algorithm implementation. Gamma index evaluation between the 2D reconstructed dose and the calculated with a treatment planning system (TPS) was done. Results: A linear fit was found for the dose-response function. The pixel sensitivity map has a radial symmetry and was calculated with a profile of the pixel sensitivity variation. The parameters for the scatter kernels were determined only for a 6 MV beam. The primary dose was estimated applying the scatter kernel within EPID and scatter kernel within the patient. The beam hardening coefficient is σBH= 3.788×10{sup −4} cm{sup 2} and the effective linear attenuation coefficient is µAC= 0.06084 cm{sup −1}. The 95% of points evaluated had γ values not longer than the unity, with gamma criteria of ΔD = 3% and Δd = 3 mm, and within the 50% isodose surface. Conclusion: The use of EPID systems proved to be a fast tool for in vivo dosimetry, but the implementation is more complex that the elaborated for pre-treatment dose verification, therefore, a simplest method must be investigated. The accuracy of this method should be improved modifying the algorithm in order to compare lower isodose curves.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

SU-F-E-19: A Novel Method for TrueBeam Jaw Calibration

Purpose: A simple jaw calibration method is proposed for Varian TrueBeam using an EPID-Encoder combination that gives accurate fields sizes and a homogeneous junction dose. This benefits clinical applications such as mono-isocentric half-beam block breast cancer or head and neck cancer treatment with junction/field matching. Methods: We use EPID imager with pixel size 0.392 mm × 0.392 mm to determine the radiation jaw position as measured from radio-opaque markers aligned with the crosshair. We acquire two images with different symmetric field sizes and record each individual jaw encoder values. A linear relationship between each jaw’s position and its encoder value is established, from which we predict the encoder values that produce the jaw positions required by TrueBeam’s calibration procedure. During TrueBeam’s jaw calibration procedure, we move the jaw with the pendant to set the jaw into position using the predicted encoder value. The overall accuracy is under 0.1 mm. Results: Our in-house software analyses images and provides sub-pixel accuracy to determine field centre and radiation edges (50% dose of the profile). We verified the TrueBeam encoder provides a reliable linear relationship for each individual jaw position (R{sup 2}>0.9999) from which the encoder values necessary to set jaw calibration points (1 cm and 19 cm) are predicted. Junction matching dose inhomogeneities were improved from >±20% to <±6% using this new calibration protocol. However, one technical challenge exists for junction matching, if the collimator walkout is large. Conclusion: Our new TrueBeam jaw calibration method can systematically calibrate the jaws to crosshair within sub-pixel accuracy and provides both good junction doses and field sizes. This method does not compensate for a larger collimator walkout, but can be used as the underlying foundation for addressing the walkout issue.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

SU-F-I-08: CT Image Ring Artifact Reduction Based On Prior Image

Purpose: In computed tomography (CT) system, CT images with ring artifacts will be reconstructed when some adjacent bins of detector don’t work. The ring artifacts severely degrade CT image quality. We present a useful CT ring artifacts reduction based on projection data correction, aiming at estimating the missing data of projection data accurately, thus removing the ring artifacts of CT images. Methods: The method consists of ten steps: 1) Identification of abnormal pixel line in projection sinogram; 2) Linear interpolation within the pixel line of projection sinogram; 3) FBP reconstruction using interpolated projection data; 4) Filtering FBP image using mean filter; 5) Forwarding projection of filtered FBP image; 6) Subtraction forwarded projection from original projection; 7) Linear interpolation of abnormal pixel line area in the subtraction projection; 8) Adding the interpolated subtraction projection on the forwarded projection; 9) FBP reconstruction using corrected projection data; 10) Return to step 4 until the pre-set iteration number is reached. The method is validated on simulated and real data to restore missing projection data and reconstruct ring artifact-free CT images. Results: We have studied impact of amount of dead bins of CT detector on the accuracy of missing data estimation in projection sinogram. For the simulated case with a resolution of 256 by 256 Shepp-Logan phantom, three iterations are sufficient to restore projection data and reconstruct ring artifact-free images when the dead bins rating is under 30%. The dead-bin-induced artifacts are substantially reduced. More iteration number is needed to reconstruct satisfactory images while the rating of dead bins increases. Similar results were found for a real head phantom case. Conclusion: A practical CT image ring artifact correction scheme based on projection data is developed. This method can produce ring artifact-free CT images feasibly and effectively.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

SU-G-BRA-07: An Innovative Fiducial-Less Tracking Method for Radiation Treatment of Abdominal Tumors by Diaphragm Disparity Analysis

Purpose: To investigate the feasibility of tracking abdominal tumors without the use of gold fiducial markers Methods: In this simulation study, an abdominal 4DCT dataset, acquired previously and containing 8 phases of the breathing cycle, was used as the testing data. Two sets of DRR images (45 and 135 degrees) were generated for each phase. Three anatomical points along the lung-diaphragm interface on each of the Digital Reconstructed Radiograph(DRR) images were identified by cross-correlation. The gallbladder, which simulates the tumor, was contoured for each phase of the breathing cycle and the corresponding centroid values serve as the measured center of the tumor. A linear model was created to correlate the diaphragm’s disparity of the three identified anatomical points with the center of the tumor. To verify the established linear model, we sequentially removed one phase of the data (i.e., 3 anatomical points and the corresponding tumor center) and created new linear models with the remaining 7 phases. Then we substituted the eliminated phase data (disparities of the 3 anatomical points) into the corresponding model to compare model-generated tumor center and the measured tumor center. Results: The maximum difference between the modeled and the measured centroid values across the 8 phases were 0.72, 0.29 and 0.30 pixels in the x, y and z directions respectively, which yielded a maximum mean-squared-error value of 0.75 pixels. The outcomes of the verification process, by eliminating each phase, produced mean-squared-errors ranging from 0.41 to 1.28 pixels. Conclusion: Gold fiducial markers, requiring surgical procedures to be implanted, are conventionally used in radiation therapy. The present work shows the feasibility of a fiducial-less tracking method for localizing abdominal tumors. Through developed diaphragm disparity analysis, the established linear model was verified with clinically accepted errors. The tracking method in real time under different radiation therapy platforms will be further investigated.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

WE-AB-207A-09: Optimization of the Design of a Moving Blocker for Cone-Beam CT Scatter Correction: Experimental Evaluation

Purpose: A moving blocker based strategy has shown promising results for scatter correction in cone-beam computed tomography (CBCT). Different geometry designs and moving speeds of the blocker affect its performance in image reconstruction accuracy. The goal of this work is to optimize the geometric design and moving speed of the moving blocker system through experimental evaluations. Methods: An Elekta Synergy XVI system and an anthropomorphic pelvis phantom CIRS 801-P were used for our experiment. A blocker consisting of lead strips was inserted between the x-ray source and the phantom moving back and forth along rotation axis to measure the scatter signal. Accoriding to our Monte Carlo simulation results, three blockers were used, which have the same lead strip width 3.2mm and different gap between neighboring lead strips, 3.2, 6.4 and 9.6mm. For each blocker, three moving speeds were evaluated, 10, 20 and 30 pixels per projection (on the detector plane). Scatter signal in the unblocked region was estimated by cubic B-spline based interpolation from the blocked region. CBCT image was reconstructed by a total variation (TV) based algebraic iterative reconstruction (ART) algorithm from the partially blocked projection data. Reconstruction accuracy in each condition is quantified as CT number error of region of interest (ROI) by comparing to a CBCT reconstructed image from analytically simulated unblocked and scatter free projection data. Results: Highest reconstruction accuracy is achieved when the blocker width is 3.2 mm, the gap between neighboring lead strips is 9.6 mm and the moving speed is 20 pixels per projection. RMSE of the CT number of ROIs can be reduced from 436 to 27. Conclusions: Image reconstruction accuracy is greatly affected by the geometry design of the blocker. The moving speed does not have a very strong effect on reconstruction result if it is over 20 pixels per projection.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

WE-FG-BRA-08: Potential Role of the Glycolytic Oscillator in Acute Hypoxia in Tumors

Purpose: Oscillatory dynamics in acute hypoxia have been observed, but poorly understood. They have mostly been attributed to vascular perturbations, but no link has yet been made to metabolic causes. We set out to determine the fundamental frequencies and test for coherence in tumor oxygen dynamics and spatial properties. Methods: Severe combined immunodeficient (SCID) mice were inoculated onto bilateral flanks with human derived head and neck carcinoma (UW-SCC22) cell line xenografts. Oxygen dynamics were monitored in the tumor every minute for an hour using three modalities: blood oxygen level dependent - magnetic resonance imaging (BOLD-MRI), hemoglobin oxygen saturation photoacoustic, and locally manufactured optical probes for spectral fitting. A statistical test was used to separate fluctuating from non-fluctuating voxels and pixels in BOLD-MRI and photoacoustic data respectively. The power spectrum density (PSD) and the autocorrelation functions were calculated for the time series of each voxel, pixel and region, of the BOLD-MRI, photoacoustic or fiber optic data respectively. Results: Using all three techniques, intermittent oxygen dynamics with both coherent and incoherent signatures was observed in the tumors. Upon averaging the PSDs of fluctuating voxels and pixels, it was found that these oscillations occurred with periods of minutes to tens of minutes from all three approaches. Observations from the BOLD-MRI and photoacoustic data showed that clusters of voxels oscillated in a synchronized manner. Conclusion: We were able to use three different modalities to show that fluctuation in tumor oxygen is both coherent and incoherent, with periods of minutes to tens of minutes. These periods are very similar to those from the well-established metabolic, non-linear biomechanical phenomenon called the glycolytic oscillator. This may provide an additional explanation to the cause of cyclic hypoxia. Such dynamics could have profound implications in hypofractionated radiotherapy regiments and could help guide treatment and make it more patient specific. The authors would like to thank the University of Wisconsin Carbone Cancer Center (UWCCC) for the funds to complete this project. This work is also supported in part by NIH/NCI P30 CA014520- UW Comprehensive Cancer Center Support”.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

WE-FG-207B-04: Noise Suppression for Energy-Resolved CT Via Variance Weighted Non-Local Filtration

Purpose: The photon starvation problem is exacerbated in energy-resolved CT, since the detected photons are shared by multiple energy channels. Using pixel similarity-based non-local filtration, we aim to produce accurate and high-resolution energy-resolved CT images with significantly reduced noise. Methods: Averaging CT images reconstructed from different energy channels reduces noise at the price of losing spectral information, while conventional denoising techniques inevitably degrade image resolution. Inspired by the fact that CT images of the same object at different energies share the same structures, we aim to reduce noise of energy-resolved CT by averaging only pixels of similar materials - a non-local filtration technique. For each CT image, an empirical exponential model is used to calculate the material similarity between two pixels based on their CT values and the similarity values are organized in a matrix form. A final similarity matrix is generated by averaging these similarity matrices, with weights inversely proportional to the estimated total noise variance in the sinogram of different energy channels. Noise suppression is achieved for each energy channel via multiplying the image vector by the similarity matrix. Results: Multiple scans on a tabletop CT system are used to simulate 6-channel energy-resolved CT, with energies ranging from 75 to 125 kVp. On a low-dose acquisition at 15 mA of the Catphan©600 phantom, our method achieves the same image spatial resolution as a high-dose scan at 80 mA with a noise standard deviation (STD) lower by a factor of >2. Compared with another non-local noise suppression algorithm (ndiNLM), the proposed algorithms obtains images with substantially improved resolution at the same level of noise reduction. Conclusion: We propose a noise-suppression method for energy-resolved CT. Our method takes full advantage of the additional structural information provided by energy-resolved CT and preserves image values at each energy level. Research reported in this publication was supported by the National Institute Of Biomedical Imaging And Bioengineering of the National Institutes of Health under Award Number R21EB019597. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

60 APPLIED LIFE SCIENCES↗
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↗
DOE OSTI2026

Design optimization of MAPS-based detectors using a data-driven fast simulation approach

A parametric simulation tool for pixel sensors is presented. A realistic pixel response is simulated purely based on measurement input, without requiring detailed knowledge of the underlying manufacturing process. As such, it provides an efficient alternative to the use of Technology Computer-Aided Design simulations, which typically depend on proprietary process information. Due to its parametric approach, the package is fast and thus particularly useful for larger detector systems and high hit rate environments. This work presents measurements, simulation and its validation for the MALTA2 sensor. It is a small collection electrode monolithic active pixel sensor produced in the Tower 180 nm complementary metal-oxide-semiconductor imaging process. Modifications to the sensor’s periphery, mainly in the hit merger, are studied in order to optimize the performance for tracking and calorimetry. This optimization is of special interest as part of the MALTA3 sensor redesign in the 65 nm Tower Partners Semiconductor Co. process.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2023

Scaling Resolution of Gigapixel Whole Slide Images Using Spatial Decomposition on Convolutional Neural Networks

Gigapixel images are prevalent in scientific domains ranging from remote sensing, and satellite imagery to microscopy, etc. However, training a deep learning model at the natural resolution of those images has been a challenge in terms of both, overcoming the resource limit (e.g. HBM memory constraints), as well as scaling up to a large number of GPUs. In this paper, we trained Residual neural Networks (ResNet) on 22,528 x 22,528-pixel size images using a distributed spatial decomposition method on 2,304 GPUs on the Summit Supercomputer. We applied our method on a Whole Slide Imaging (WSI) dataset from The Cancer Genome Atlas (TCGA) database. WSI images can be in the size of 100,000 x 100,000 pixels or even larger, and in this work we studied the effect of image resolution on a classification task, while achieving state-of-the-art AUC scores. Moreover, our approach doesn't need pixel-level labels, since we're avoiding patching from the WSI images completely, while adding the capability of training arbitrary large-size images. This is achieved through a distributed spatial decomposition method, by leveraging the non-block fat-tree interconnect network of the Summit architecture, which enabled GPU-to-GPU direct communication. Finally, detailed performance analysis results are shown, as well as a comparison with a data-parallel approach when possible.

Tsaris, Aristeidis (aris)↗
DOE OSTI2015

Research on a Pulmonary Nodule Segmentation Method Combining Fast Self-Adaptive FCM and Classification

The key problem of computer-aided diagnosis (CAD) of lung cancer is to segment pathologically changed tissues fast and accurately. As pulmonary nodules are potential manifestation of lung cancer, we propose a fast and self-adaptive pulmonary nodules segmentation method based on a combination of FCM clustering and classification learning. The enhanced spatial function considers contributions to fuzzy membership from both the grayscale similarity between central pixels and single neighboring pixels and the spatial similarity between central pixels and neighborhood and improves effectively the convergence rate and self-adaptivity of the algorithm. Experimental results show that the proposed method can achieve more accurate segmentation of vascular adhesion, pleural adhesion, and ground glass opacity (GGO) pulmonary nodules than other typical algorithms.

59 BASIC BIOLOGICAL SCIENCES↗