Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Serialization”

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 397 records · Page 22
All sourcesNASA NTRSDOE OSTI
DOE OSTI2013

Crystallographic data processing for free-electron laser sources

A processing pipeline for diffraction data acquired using the ‘serial crystallography’ methodology with a free-electron laser source is described with reference to the crystallographic analysis suite CrystFEL and the pre-processing program Cheetah. A processing pipeline for diffraction data acquired using the ‘serial crystallography’ methodology with a free-electron laser source is described with reference to the crystallographic analysis suite CrystFEL and the pre-processing program Cheetah. A detailed analysis of the nature and impact of indexing ambiguities is presented. Simulations of the Monte Carlo integration scheme, which accounts for the partially recorded nature of the diffraction intensities, are presented and show that the integration of partial reflections could be made to converge more quickly if the bandwidth of the X-rays were to be increased by a small amount or if a slight convergence angle were introduced into the incident beam.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2015

Acoustic transfer of protein crystals from agarose pedestals to micromeshes for high-throughput screening

An acoustic high-throughput screening method is described for harvesting protein crystals and combining the protein crystals with chemicals such as a fragment library. Acoustic droplet ejection (ADE) is an emerging technology with broad applications in serial crystallography such as growing, improving and manipulating protein crystals. One application of this technology is to gently transfer crystals onto MiTeGen micromeshes with minimal solvent. Once mounted on a micromesh, each crystal can be combined with different chemicals such as crystal-improving additives or a fragment library. Acoustic crystal mounting is fast (2.33 transfers s{sup −1}) and all transfers occur in a sealed environment that is in vapor equilibrium with the mother liquor. Here, a system is presented to retain crystals near the ejection point and away from the inaccessible dead volume at the bottom of the well by placing the crystals on a concave agarose pedestal (CAP) with the same chemical composition as the crystal mother liquor. The bowl-shaped CAP is impenetrable to crystals. Consequently, gravity will gently move the crystals into the optimal location for acoustic ejection. It is demonstrated that an agarose pedestal of this type is compatible with most commercially available crystallization conditions and that protein crystals are readily transferred from the agarose pedestal onto micromeshes with no loss in diffraction quality. It is also shown that crystals can be grown directly on CAPs, which avoids the need to transfer the crystals from the hanging drop to a CAP. This technology has been used to combine thermolysin and lysozyme crystals with an assortment of anomalously scattering heavy atoms. The results point towards a fast nanolitre method for crystal mounting and high-throughput screening.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2015

Determination of crystallographic intensities from sparse data

X-ray serial microcrystallography involves the collection and merging of frames of diffraction data from randomly oriented protein microcrystals. The number of diffracted X-rays in each frame is limited by radiation damage, and this number decreases with crystal size. The data in the frame are said to be sparse if too few X-rays are collected to determine the orientation of the microcrystal. It is commonly assumed that sparse crystal diffraction frames cannot be merged, thereby setting a lower limit to the size of microcrystals that may be merged with a given source fluence. The EMC algorithm [Loh & Elser (2009), Phys. Rev. E , 80 , 026705] has previously been applied to reconstruct structures from sparse noncrystalline data of objects with unknown orientations [Philipp et al. (2012), Opt. Express , 20 , 13129–13137; Ayyer et al. (2014), Opt. Express , 22 , 2403–2413]. Here, it is shown that sparse data which cannot be oriented on a per-frame basis can be used effectively as crystallographic data. As a proof-of-principle, reconstruction of the three-dimensional diffraction intensity using sparse data frames from a 1.35 kDa molecule crystal is demonstrated. The results suggest that serial microcrystallography is, in principle, not limited by the fluence of the X-ray source, and collection of complete data sets should be feasible at, for instance, storage-ring X-ray sources.

Chemistry↗
DOE OSTI2026

Towards time-resolved MicroED grid preparation using mix-and-inject gas dynamic virtual nozzles

Recent progress in gas dynamic virtual nozzle (GDVN) technologies in combination with high-brilliance synchrotron and X-ray free-electron lasers (XFELs) has allowed the visualization of protein dynamics in crystallo by mixing macromolecular protein crystals with a substrate using tunable mixing times on the order of milliseconds to seconds prior to serial X-ray diffraction data collection. This has become the method of choice for high-resolution structure determination of intermediate states. However, such experiments require large counts of crystals of proper sizes for high-resolution data collection, and premium beam times for screening efforts. Cryogenic microcrystal electron diffraction (MicroED) represents a complementary technique that may be a more accessible avenue for time-resolved nanocrystallography compared with serial X-ray diffraction experiments. MicroED can produce full diffraction datasets from just a few submicrometre-thick crystals, and the approach is more readily accessible, requiring standard cryogenic transmission electron microscopy (TEM) equipment available at many universities and institutes. Cryogenic MicroED, like other forms of cryo-EM, begins with rapidly freezing biological material on electron microscopy grids. In the case of MicroED, micro- to nano-crystals (<500 nm thick) are deposited onto electron microscopy grids and plunge-frozen for subsequent electron diffraction data collection. Here, we have incorporated GDVN technology developed originally for XFEL experiments into the freezing process as a first step towards time-resolved studies. We describe the limited deposition efficiency of the model MicroED protein proteinase K on TEM grids using GDVNs, preceding sample vitrification and successful MicroED data collection. We discuss both the initial results from such experiments and the methodological challenges in developing this approach into a reliable workflow for millisecond-to-second time-resolved structural studies of macromolecules. Our results promise a strategy to deposit crystals on grids using GDVNs and determine high-resolution structures by MicroED, constituting a first step towards development of time-resolved MicroED experiments.

MicroED↗
DOE OSTI2015

A comprehensive review of the lipid cubic phase or in meso method for crystallizing membrane and soluble proteins and complexes

A comprehensive and up-to-date review of the lipid cubic phase or in meso method for crystallizing membrane and soluble proteins and complexes is reported. Recent applications of the method for in situ serial crystallography at X-ray free-electron lasers and synchrotrons are described. The lipid cubic phase or in meso method is a robust approach for crystallizing membrane proteins for structure determination. The uptake of the method is such that it is experiencing what can only be described as explosive growth. This timely, comprehensive and up-to-date review introduces the reader to the practice of in meso crystallogenesis, to the associated challenges and to their solutions. A model of how crystallization comes about mechanistically is presented for a more rational approach to crystallization. The possible involvement of the lamellar and inverted hexagonal phases in crystallogenesis and the application of the method to water-soluble, monotopic and lipid-anchored proteins are addressed. How to set up trials manually and automatically with a robot is introduced with reference to open-access online videos that provide a practical guide to all aspects of the method. These range from protein reconstitution to crystal harvesting from the hosting mesophase, which is noted for its viscosity and stickiness. The sponge phase, as an alternative medium in which to perform crystallization, is described. The compatibility of the method with additive lipids, detergents, precipitant-screen components and materials carried along with the protein such as denaturants and reducing agents is considered. The powerful host and additive lipid-screening strategies are described along with how samples that have low protein concentration and cell-free expressed protein can be used. Assaying the protein reconstituted in the bilayer of the cubic phase for function is an important element of quality control and is detailed. Host lipid design for crystallization at low temperatures and for large proteins and complexes is outlined. Experimental phasing by heavy-atom derivatization, soaking or co-crystallization is routine and the approaches that have been implemented to date are described. An overview and a breakdown by family and function of the close to 200 published structures that have been obtained using in meso-grown crystals are given. Recommendations for conducting the screening process to give a more productive outcome are summarized. The fact that the in meso method also works with soluble proteins should not be overlooked. Recent applications of the method for in situ serial crystallography at X-ray free-electron lasers and synchrotrons are described. The review ends with a view to the future and to the bright prospects for the method, which continues to contribute to our understanding of the molecular mechanisms of some of nature’s most valued proteinaceous robots.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2020

DeepFreeze: Towards Scalable Asynchronous Checkpointing of Deep Learning Models

In the age of big data, deep learning has emerged as a powerful tool to extract insight and exploit its value, both in industry and scientific applications. One common pattern emerging in such applications is frequent checkpointing of the state of the learning model during training, needed in a variety of scenarios: analysis of intermediate states to explain features and correlations with training data, exploration strategies involving alternative models that share a common ancestor, knowledge transfer, resilience, etc. However, with increasing size of the learning models and popularity of distributed data-parallel training approaches, simple checkpointing techniques used so far face several limitations: low serialization performance, blocking I/O, stragglers due to the fact that only a single process is involved in checkpointing. This paper proposes a checkpointing technique specifically designed to address the aforementioned limitations, introducing efficient asynchronous techniques to hide the overhead of serialization and I/O, and distribute the load over all participating processes. Experiments with two deep learning applications (CANDLE and ResNet) on a pre-Exascale HPC platform (Theta) shows significant improvement over state-of-art, both in terms of checkpointing duration and runtime overhead.

checkpointing↗
DOE OSTI2017

Parallel Reconstruction of Three Dimensional Magnetohydrodynamic Equilibria in Plasma Confinement Devices

Fast, accurate three dimensional reconstructions of plasma equilibria, crucial for physics interpretation of fusion data generated within confinement devices like stellarators/ tokamaks, are computationally very expensive and routinely require days, even weeks, to complete using serial approaches. Here, we present a parallel implementation of the three dimensional plasma reconstruction code, V3FIT. A formal analysis to identify the performance bottlenecks and scalability limits of this new parallel implementation, which combines both task and data parallelism, is presented. The theoretical findings are supported by empirical performance results on several thousands of processor cores of a Cray XC30 supercomputer. Parallel V3FIT is shown to deliver over 40X speedup, enabling fusion scientists to carry out three dimensional plasma equilibrium reconstructions at unprecedented scales in only a few hours (instead of in days/weeks) for the first time.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2017

Preemptive resource management for dynamically arriving tasks in an oversubscribed heterogeneous computing system

We design resource management heuristics that assign serial tasks to the nodes of a heterogeneous high performance computing (HPC) system. The value of completing these tasks is modeled using monotonically decreasing utility functions that represent the time-varying importance of the task. The value of completing a task is equal to its utility function at the time of its completion. The overall performance of this system is measured using the total utility earned by all tasks during some interval of time. To maximize the performance of such a system where the preemption of tasks is possible, we have designed, analyzed, and compared a set of resource allocation heuristic techniques. We combine two utility-aware heuristics with three different preemption techniques to create six preemption-capable heuristics. We also consider the two utility-aware heuristics without preemption. We use simulation studies to evaluate this set of eight heuristics and compare them with an FCFS heuristic, which is often used in real systems, and random assignments. In general, our set of eight heuristics is able to significantly outperform the comparison heuristics, and the preemption-capable heuristics are able to significantly increase the utility earned compared to the heuristics that do not use preemption. We analyze the performance tradeoffs among the different preemption-capable heuristics under a variety of oversubscribed workload environments.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2017

Parallel peak pruning for scalable SMP contour tree computation

As data sets grow to exascale, automated data analysis and visualisation are increasingly important, to intermediate human understanding and to reduce demands on disk storage via in situ analysis. Trends in architecture of high performance computing systems necessitate analysis algorithms to make effective use of combinations of massively multicore and distributed systems. One of the principal analytic tools is the contour tree, which analyses relationships between contours to identify features of more than local importance. Unfortunately, the predominant algorithms for computing the contour tree are explicitly serial, and founded on serial metaphors, which has limited the scalability of this form of analysis. While there is some work on distributed contour tree computation, and separately on hybrid GPU-CPU computation, there is no efficient algorithm with strong formal guarantees on performance allied with fast practical performance. Here in this paper, we report the first shared SMP algorithm for fully parallel contour tree computation, withfor-mal guarantees of O(lgnlgt) parallel steps and O(n lgn) work, and implementations with up to 10x parallel speed up in OpenMP and up to 50x speed up in NVIDIA Thrust.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Mooring Load Monitoring of a Wave Energy Converter Using a Self-Synchronizing Underwater Acoustic Network

A self-synchronizing underwater acoustic network, designed for remote monitoring of mooring loads in Wave Energy Converters (WEC), has been developed and tested. This network uses Time Division Multiple Access and operates self-contained with the ability for users to remotely transmit commands to the network as needed. Each node is a self-contained unit, consisting of a protocol adaptor board, an underwater acoustic modem and a battery pack. A node can be connected to a load cell, to a topside user or to the WEC. Every node is swapable. The protocol adaptor board, named Protocol Adaptor for Digital LOad Cell (PADLOC) supports a variety of digital load cell message formats (CAN, MODBUS, custom ASCII) and underwater acoustic modem serial formats. PADLOC enables topside users to connect to separate load cells through a user-specific command.

acoustic↗
DOE OSTI2020

PANTHER: A Programmable Architecture for Neural Network Training Harnessing Energy-Efficient ReRAM

The wide adoption of deep neural networks has been accompanied by ever-increasing energy and performance demands due to the expensive nature of training them. Additionally, numerous special-purpose architectures have been proposed to accelerate training: both digital and hybrid digital-analog using resistive RAM (ReRAM) crossbars. ReRAM-based accelerators have demonstrated the effectiveness of ReRAM crossbars at performing matrix-vector multiplication operations that are prevalent in training. However, they still suffer from inefficiency due to the use of serial reads and writes for performing the weight gradient and update step. A few works have demonstrated the possibility of performing outer products in crossbars, which can be used to realize the weight gradient and update step without the use of serial reads and writes. However, these works have been limited to low precision operations which are not sufficient for typical training workloads. Moreover, they have been confined to a limited set of training algorithms for fully-connected layers only. To address these limitations, we propose a bit-slicing technique for enhancing the precision of ReRAM-based outer products, which is substantially different from bit-slicing for matrix-vector multiplication only. We incorporate this technique into a crossbar architecture with three variants catered to different training algorithms. To evaluate our design on different types of layers in neural networks (fully-connected, convolutional, etc.) and training algorithms, we develop PANTHER, an ISA-programmable training accelerator with compiler support. Our design can also be integrated into other accelerators in the literature to enhance their efficiency. Our evaluation shows that PANTHER achieves up to 8.02×, 54.21×, and 103× energy reductions as well as 7.16×, 4.02×, and 16× execution time reductions compared to digital accelerators, ReRAM-based accelerators, and GPUs, respectively.

42 ENGINEERING↗
DOE OSTI2026

Versatile High-Gain Low-Noise Readout ASIC for Silicon Microstrip Tracking Detectors

This work presents Turpial, a custom-designed low- power front-end readout ASIC for microstrip silicon sensors. Implemented in 130 nm CMOS technology, the chip integrates 64 identical readout channels, each including a configurable charge-sensitive amplifier, a bipolar pulse shaper, a 32-sample 50 Msps analog memory, and a 12-bit RC-hybrid SAR ADC operating at 1 Msps. To satisfy the target power budget of 5 mW per channel, the architecture employs a time-decoupled readout scheme in which fast transient signals are first captured in the analog memory and subsequently digitized at a lower rate. Turpial supports a wide dynamic range from 1 kℎ+ to 1 Mℎ+ while maintaining low noise performance, targeting an equivalent noise charge (ENC) below 200 𝑒−including the sensor, and providing a maximum gain of 1500 mV/fC. A digital block manages slow control, data acquisition, and data serialization through dual CML 300 Mb/s serializers. In addition, an on-chip reference circuit, based on a sub-1 V bandgap reference and an integrated LDO regulator, eliminates the need for external reference circuitry. Experimental results demonstrate that both the individual building blocks and the fully integrated ASIC meet the design specifications.

Hernandez, Hugo [Stanford University] (ORCID:00000↗
DOE OSTI2026

Efficient Probabilistic Visualization of Local Divergence of 2D Vector Fields with Independent Gaussian Uncertainty

This work focuses on visualizing uncertainty of local divergence of two-dimensional vector fields. Divergence is one of the fundamental attributes of fluid flows, as it can help domain scientists analyze potential positions of sources (positive divergence) and sinks (negative divergence) in the flow. However, uncertainty inherent in vector field data can lead to erroneous divergence computations, adversely impacting downstream analysis. While Monte Carlo (MC) sampling is a classical approach for estimating divergence uncertainty, it suffers from slow convergence and poor scalability with increasing data size and sample counts. Thus, we present a two-fold contribution that tackles the challenges of slow convergence and limited scalability of the MC approach. (1) We derive a closed-form approach for highly efficient and accurate uncertainty visualization of local divergence, assuming independently Gaussian-distributed vector uncertainties. (2) We further integrate our approach into Viskores, a platform-portable parallel library, to accelerate uncertainty visualization. In our results, we demonstrate significantly enhanced efficiency and accuracy of our serial analytical (speed-up up to 1946×) and parallel Viskores (speed-up up to 19698×) algorithms over the classical serial MC approach. We also demonstrate qualitative improvements of our probabilistic divergence visualizations over traditional mean-field visualization, which disregards uncertainty. We validate the accuracy and efficiency of our methods on wind forecast and ocean simulation datasets.

Ouermi, Timbwaoga [University of Utah]↗
DOE OSTI2024

Readout optimization of multi-amplifier sensing charge-coupled devices for single-quantum measurement

The non-destructive readout capability of the Skipper Charge Coupled Device (CCD) has been demonstrated to reduce the noise limitation of conventional silicon devices to levels that allow single-photon or single-electron counting. The noise reduction is achieved by taking multiple measurements of the charge in each pixel. These multiple measurements come at the cost of extra readout time, which has been a limitation for the broader adoption of this technology in particle physics, quantum imaging, and astronomy applications. This work presents recent results of a novel sensor architecture that uses multiple non-destructive floating-gate amplifiers in series to achieve sub-electron readout noise in a thick, fully-depleted silicon detector to overcome the readout time overhead of the Skipper-CCD. This sensor is called the Multiple-Amplifier Sensing Charge-Coupled Device (MAS-CCD) can perform multiple independent charge measurements with each amplifier, and the measurements from multiple amplifiers can be combined to further reduce the readout noise. We will show results obtained for sensors with 8 and 16 amplifiers per readout stage in new readout operations modes to optimize its readout speed. The noise reduction capability of the new techniques will be demonstrated in terms of its ability to reduce the noise by combining the information from the different amplifiers, and to resolve signals in the order of a single photon per pixel. The first readout operation explored here avoids the extra readout time needed in the MAS-CCD to read a line of the sensor associated with the extra extent of the serial register. The second technique explore the capability of the MAS-CCD device to perform a region of interest readout increasing the number of multiple samples per amplifier in a targeted region of the active area of the device.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2014

SU-E-T-41: Analysis of GI Dose Variability Due to Intrafraction Setup Variance

Purpose: Proton SBRT (stereotactic body radiation therapy) can be an effective modality for treatment of gastrointestinal tumors, but limited in practice due to sensitivity with respect to variation in the RPL (radiological path length). Small, intrafractional shifts in patient anatomy can lead to significant changes in the dose distribution. This study describes a tool designed to visualize uncertainties in radiological depth in patient CT's and aid in treatment plan design. Methods: This project utilizes the Shadie toolkit, a GPU-based framework that allows for real-time interactive calculations for volume visualization. Current SBRT simulation practice consists of a serial CT acquisition for the assessment of inter- and intra-fractional motion utilizing patient specific immobilization systems. Shadie was used to visualize potential uncertainties, including RPL variance and changes in gastric content. Input for this procedure consisted of two patient CT sets, contours of the desired organ, and a pre-calculated dose. In this study, we performed rigid registrations between sets of 4DCT's obtained from a patient with varying setup conditions. Custom visualizations are written by the user in Shadie, permitting one to create color-coded displays derived from a calculation along each ray. Results: Serial CT data acquired on subsequent days was analyzed for variation in RPB and gastric content. Specific shaders were created to visualize clinically relevant features, including RPL (radiological path length) integrated up to organs of interest. Using pre-calculated dose distributions and utilizing segmentation masks as additional input allowed us to further refine the display output from Shadie and create tools suitable for clinical usage. Conclusion: We have demonstrated a method to visualize potential uncertainty for intrafractional proton radiotherapy. We believe this software could prove a useful tool to guide those looking to design treatment plans least insensitive to motion for patients undergoing proton SBRT in the GI tract.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2014

TU-A-12A-08: Computing Longitudinal Material Changes in Bone Metastases Using Dual Energy Computed Tomography

Purpose: This study demonstrates a methodology for tracking changes in metastatic bone disease using trajectories in material basis space in serial dual energy computed tomography (DECT) studies. Methods: This study includes patients with bone metastases from breast cancer that had clinical surveillance CT scans using a General Electric CT750HD in dual energy mode. A radiologist defined regions-of-interested (ROI) for bone metastasis, normal bone, and marrow across the serial DECT scans. Our approach employs a Radon transform to forward-projection the basis images, namely, water and iodine, into sinogram space. This data is then repartitioned into fat/bone and effective density/Z image pairs using assumed energy spectrums for the x-ray energies. This approach both helps remove negative material densities and avoids adding spectrum-hardening artifacts. These new basis data sets were then reconstructed via filtered back-projection to create new material basis pair images. The trajectories of these pairs were then plotted in the new basis space providing a means to both visualize and quantitatively measure changes in the material properties of the tumors. Results: ROI containing radiologist defined metastatic bone disease showed well-defined trajectories in both fat/bone and effective density/Z space. ROI that contained radiologist defined normal bone and marrow did not exhibit any discernible trajectories and were stable from scan to scan. Conclusions: The preliminary results show that changes in material composition and effective density/Z image pairs were seen primarily in metastasis and not in normal tissue. This study indicates that by using routine clinical DECT it may be possible to monitor therapy response of bone metastases because healing or worsening bone metastases change material composition of bone. Additional studies are needed to further validate these results and to test for their correlation with outcome.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2014

TH-A-9A-02: BEST IN PHYSICS (THERAPY) - 4D IMRT Planning Using Highly- Parallelizable Particle Swarm Optimization

Purpose: We present a particle swarm optimization (PSO)-based 4D IMRT planning technique designed for dynamic MLC tracking delivery to lung tumors. The key idea is to utilize the temporal dimension as an additional degree of freedom rather than a constraint in order to achieve improved sparing of organs at risk (OARs). Methods: The target and normal structures were manually contoured on each of the ten phases of a 4DCT scan acquired from a lung SBRT patient who exhibited 1.5cm tumor motion despite the use of abdominal compression. Corresponding ten IMRT plans were generated using the Eclipse treatment planning system. These plans served as initial guess solutions for the PSO algorithm. Fluence weights were optimized over the entire solution space i.e., 10 phases × 12 beams × 166 control points. The size of the solution space motivated our choice of PSO, which is a highly parallelizable stochastic global optimization technique that is well-suited for such large problems. A summed fluence map was created using an in-house B-spline deformable image registration. Each plan was compared with a corresponding, internal target volume (ITV)-based IMRT plan. Results: The PSO 4D IMRT plan yielded comparable PTV coverage and significantly higher dose—sparing for parallel and serial OARs compared to the ITV-based plan. The dose-sparing achieved via PSO-4DIMRT was: lung Dmean = 28%; lung V20 = 90%; spinal cord Dmax = 23%; esophagus Dmax = 31%; heart Dmax = 51%; heart Dmean = 64%. Conclusion: Truly 4D IMRT that uses the temporal dimension as an additional degree of freedom can achieve significant dose sparing of serial and parallel OARs. Given the large solution space, PSO represents an attractive, parallelizable tool to achieve globally optimal solutions for such problems. This work was supported through funding from the National Institutes of Health and Varian Medical Systems. Amit Sawant has research funding from Varian Medical Systems, VisionRT Ltd. and Elekta.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2015

Effect of deformable registration on the dose calculated in radiation therapy planning CT scans of lung cancer patients

Purpose: To characterize the effects of deformable image registration of serial computed tomography (CT) scans on the radiation dose calculated from a treatment planning scan. Methods: Eighteen patients who received curative doses (≥60 Gy, 2 Gy/fraction) of photon radiation therapy for lung cancer treatment were retrospectively identified. For each patient, a diagnostic-quality pretherapy (4–75 days) CT scan and a treatment planning scan with an associated dose map were collected. To establish correspondence between scan pairs, a researcher manually identified anatomically corresponding landmark point pairs between the two scans. Pretherapy scans then were coregistered with planning scans (and associated dose maps) using the demons deformable registration algorithm and two variants of the Fraunhofer MEVIS algorithm (“Fast” and “EMPIRE10”). Landmark points in each pretherapy scan were automatically mapped to the planning scan using the displacement vector field output from each of the three algorithms. The Euclidean distance between manually and automatically mapped landmark points (d{sub E}) and the absolute difference in planned dose (|ΔD|) were calculated. Using regression modeling, |ΔD| was modeled as a function of d{sub E}, dose (D), dose standard deviation (SD{sub dose}) in an eight-pixel neighborhood, and the registration algorithm used. Results: Over 1400 landmark point pairs were identified, with 58–93 (median: 84) points identified per patient. Average |ΔD| across patients was 3.5 Gy (range: 0.9–10.6 Gy). Registration accuracy was highest using the Fraunhofer MEVIS EMPIRE10 algorithm, with an average d{sub E} across patients of 5.2 mm (compared with >7 mm for the other two algorithms). Consequently, average |ΔD| was also lowest using the Fraunhofer MEVIS EMPIRE10 algorithm. |ΔD| increased significantly as a function of d{sub E} (0.42 Gy/mm), D (0.05 Gy/Gy), SD{sub dose} (1.4 Gy/Gy), and the algorithm used (≤1 Gy). Conclusions: An average error of <4 Gy in radiation dose was introduced when points were mapped between CT scan pairs using deformable registration, with the majority of points yielding dose-mapping error <2 Gy (approximately 3% of the total prescribed dose). Registration accuracy was highest using the Fraunhofer MEVIS EMPIRE10 algorithm, resulting in the smallest errors in mapped dose. Dose differences following registration increased significantly with increasing spatial registration errors, dose, and dose gradient (i.e., SD{sub dose}). This model provides a measurement of the uncertainty in the radiation dose when points are mapped between serial CT scans through deformable registration.

60 APPLIED LIFE SCIENCES↗