Search NASA⌕ Search

SEARCH · Search NASA

Results for “Linearization error”

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 559 records · Page 31

Implications of CT noise and artifacts for quantitative {sup 99m}Tc SPECT/CT imaging

Purpose: This paper evaluates the effects of computed tomography (CT) image noise and artifacts on quantitative single-photon emission computed-tomography (SPECT) imaging, with the aim of establishing an appropriate range of CT acquisition parameters for low-dose protocols with respect to accurate SPECT attenuation correction (AC). Methods: SPECT images of two geometric and one anthropomorphic phantom were reconstructed iteratively using CT scans acquired at a range of dose levels (CTDI{sub vol} = 0.4 to 46 mGy). Resultant SPECT image quality was evaluated by comparing mean signal, background noise, and artifacts to SPECT images reconstructed using the highest dose CT for AC. Noise injection was performed on linear-attenuation (μ) maps to determine the CT noise threshold for accurate AC. Results: High levels of CT noise (σ ∼ 200–400 HU) resulted in low μ-maps noise (σ ∼ 1%–3%). Noise levels greater than ∼10% in 140 keV μ-maps were required to produce visibly perceptible increases of ∼15% in {sup 99m}Tc SPECT images. These noise levels would be achieved at low CT dose levels (CTDI{sub vol} = 4 μGy) that are over 2 orders of magnitude lower than the minimum dose for diagnostic CT scanners. CT noise could also lower (bias) the expected μ values. The relative error in reconstructed SPECT signal trended linearly with the relative shift in μ. SPECT signal was, on average, underestimated in regions corresponding with beam-hardening artifacts in CT images. Any process that has the potential to change the CT number of a region by ∼100 HU (e.g., misregistration between CT images and SPECT images due to motion, the presence of contrast in CT images) could introduce errors in μ{sub 140} {sub keV} on the order of 10%, that in turn, could introduce errors on the order of ∼10% into the reconstructed {sup 99m}Tc SPECT image. Conclusions: The impact of CT noise on SPECT noise was demonstrated to be negligible for clinically achievable CT parameters. Because CT dose levels that affect SPECT quantification is low (CTDI{sub vol} ∼ 4 μGy), the low dose limit for the CT exam as part of SPECT/CT will be guided by CT image quality requirements for anatomical localization and artifact reduction. A CT technique with higher kVp in combination with lower mAs is recommended when low-dose CT images are used for AC to minimize beam-hardening artifacts.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Nonlinear Observers for Gyro Calibration

High precision estimation and control algorithms, to achieve unprecedented levels of pointing accuracy, will be required to support future formation flying missions such as interferometry missions. Achieving high pointing accuracy requires precise knowledge of the spacecraft rotation rate. Typically, the rotation rate is measured by a gyro. The measured rates can be corrupted by errors in alignment and scale factor, gyro biases, and noise. In this work, we present nonlinear observers for gyro calibration. Nonlinear observers are superior to extended or pseudo-linear Kalman filter type approaches for large errors and global stability. Three nonlinear gyro calibration observers are developed. The first observer estimates a constant gyro bias. The second observer estimates scale factor errors. The third observer estimates the gyro alignment for three orthogonal gyros. The convergence properties of all three observers are discussed. Additionally, all three observers are coupled with a nonlinear control algorithm. The stability of each of the resulting closed loop systems is analyzed. The observers are then combined, and the gyro calibration parameters are estimated simultaneously. The stability of the combined observers is addressed, as well as the stability of the resulting closed loop systems. Simulated test results are presented for each scenario. Finally, the nonlinear observers are compared to a pseudo-linear Kalman filter.

Thienel, Julie↗

Sequential Monte Carlo sampling in hidden Markov models of nonlinear dynamical systems

We investigate the issue of which state functionals can have their uncertainty estimated efficiently in dynamical systems with uncertainty. Because of the high dimensionality and complexity of the problem, sequential Monte Carlo (SMC) methods are used. We investigate SMC methods where the proposal distribution is computed by maximum likelihood or by a linearization approach. We prove that the variance of the SMC method is bounded linearly in the number of time steps when the proposal distribution is truncated normal distribution. We also show that for a moderate large number of steps the error produced by approximation of dynamical systems linearly accumulates on the condition that the logarithm of the density function of noise is Lipschitz continuous. This finding is significant because the uncertainty in many dynamical systems, in particular, in chemical engineering systems, can be assumed to have this nature. We demonstrate our findings for a simple test case from chemical engineering. The theoretical findings provide a foundation for the parallel software SISTOS.

97 MATHEMATICS AND COMPUTING↗

A Bayesian approach to nonlinear inversion

Powerful methods are now available for solving linear parametric inverse problems. However, many inverse problems which arise in geohysics are nonlinear. Fortunately, it is possible to treat most of these with the air of linear perturbation theory and liner inversion. But a convenient method is needed for assessing the importance of nonlinearity in these quasi-linear problems. The present paper provides such a method. Matsu'ura and Jackson (1984) have presented a simple algorithm for evaluating the asymptotic covariance matrix fo estimation errors. In the present investigation, aspects of linear inversion are discussed, taking into account linear parametric inverse problems, nonuniqueness, prior information, confidence limits, conditional and marginal statistics, the relative importance of the prior and observational data, and standardized variables. Attention is also given to nonlinear inversion, and the application of the considered approaches to a number of examples.

Jackson, D. D.↗

An efficient explicit implementation of a near-optimal quantum algorithm for simulating linear dissipative differential equations

We propose an efficient block-encoding technique for the implementation of the Linear Combination of Hamiltonian Simulations (LCHS) for simulating dissipative initial-value problems. This algorithm approximates a target nonunitary operator as a weighted sum of Hamiltonian evolutions, thereby emulating a dissipative problem by mixing various time scales. We introduce an efficient encoding of the LCHS into a quantum circuit based on a simple coordinate transformation that turns the dependence on the summation index into a trigonometric function. Classically, this method is equivalent to the use of a highly accurate Fejér-Clenshaw-Curtis quadrature formula. Quantumly, this significantly simplifies block-encoding of a dissipative problem and allows one to perform an exponential number of Hamiltonian simulations by a single Quantum Signal Processing (QSP) circuit. The resulting LCHS circuit has high success probability and the selector scales logarithmically with the number of terms in the LCHS sum and linearly with time. Careful analysis of error convergence proves that this method is more efficient than other LCHS circuits that have recently appeared in the literature. We verify the quantum circuit and its scaling by simulating it on a digital emulator of fault-tolerant quantum computers and, as a test problem, solve the advection-diffusion equation. The proposed algorithm can be used for simulating a wide class of nonunitary initial-value problems including the Liouville equation with added dissipation and linear embeddings of nonlinear systems, such as the Koopman-von Neumann and Carleman embeddings.

Novikau, I [Lawrence Livermore National Laboratory↗

Growth Curve Parameterization of Metabolic Activity of Yeast Cells for BioSentinel

The goal of the BioSentinel small satellite payload is to measure the effect of deep space radiation on the growth and metabolic activity of yeast cells. Raw test data is generated by fluidics cards containing yeast cells rehydrated at different periods, with metabolic activity measured by the reduction of alamarBlue. Each card well has a sensor array that measures the amount of red, green, and infrared light transmitted through the yeast culture. This illumination data is then converted to absorbance values, which are further converted into concentrations. The ultimate objective is to convert these concentrations into biologically-relevant metrics that can be compared against one another to determine changes due to differential radiation exposure. Beginning with IR absorbance data (corresponding to cell density) from ground studies, three parameters from a sigmoidal growth curve were extracted and analyzed: 𝜆 (lag phase), 𝜇 (max growth rate), and A (max cell growth). The data was fit to the Gompertz model of microbial growth using non-linear regression (Minitab), as the fit error was reduced compared to the simpler logistic growth curve. Graphs showed that the data contained a discrepancy (drift) in the lag phase that is attributable to a slow, constant loss of moisture. Correcting this discrepancy by fitting the first 25 hours of the data to a power function and subtracting these values from the absorbance readings obtained a better statistical fit to the growth curve in the lag phase. A power fit was selected over a linear fit because it reflected the effects of constant volume loss. This correction to the BioSentinel data analysis pipeline will enable quantitative statistical analysis of the effect of different levels of deep space radiation on yeast cells. Future work includes automation of drift correction and curve modeling to extract these parameters directly from data.

Growth Curve↗

Nonlinear analysis and redesign of the mixed-mode bending delamination test

The Mixed Mode Bending (MMB) test uses a lever to simultaneously apply mode I and mode II loading to a split beam specimen. An iterative analysis that accounts for the geometric nonlinearity of the MMB test was developed. The analysis accurately predicted the measured load displacement response and the strain energy release rate, G, of an MMB test specimen made of APC2 (AS4/PEEK). The errors in G when calculated using linear theory were found to be as large as thirty percent in some cases. Because it would be inconvenient to use a nonlinear analysis to analyze MMB data, the MMB apparatus was redesigned to minimize the nonlinearity. The nonlinear analysis was used as a guide in redesigning the MMB apparatus. With the redesigned apparatus, loads were applied through a roller attached to the level and loaded just above the midplane of the test specimen. The redesigned apparatuus has geometric nonlinearity errors of less than three percent, even for materials substantially tougher than APC2. This apparatus was demonstrated by measuring the mixed mode delamination fracture toughness of APC2.

Reeder, J. R.↗

Prospects for in vivo estimation of photon linear attenuation coefficients using postprocessing dual-energy CT imaging on a commercial scanner: Comparison of analytic and polyenergetic statistical reconstruction algorithms

Purpose: Accurate patient-specific photon cross-section information is needed to support more accurate model-based dose calculation for low energy photon-emitting modalities in medicine such as brachytherapy and kilovoltage x-ray imaging procedures. A postprocessing dual-energy CT (pDECT) technique for noninvasivein vivo estimation of photon linear attenuation coefficients has been experimentally implemented on a commercial CT scanner and its accuracy assessed in idealized phantom geometries. Methods: Eight test materials of known composition and density were used to compare pDECT-estimated linear attenuation coefficients to NIST reference values over an energy range from 10 keV to 1 MeV. As statistical image reconstruction (SIR) has been shown to reconstruct images with less random and systematic error than conventional filtered backprojection (FBP), the pDECT technique was implemented with both an in-house polyenergetic SIR algorithm, alternating minimization (AM), as well as a conventional FBP reconstruction algorithm. Improvement from increased spectral separation was also investigated by filtering the high-energy beam with an additional 0.5 mm of tin. The law of propagated uncertainty was employed to assess the sensitivity of the pDECT process to errors in reconstructed images. Results: Mean pDECT-estimated linear attenuation coefficients for the eight test materials agreed within 1% of NIST reference values for energies from 1 MeV down to 30 keV, with mean errors rising to between 3% and 6% at 10 keV, indicating that the method is unbiased when measurement and calibration phantom geometries are matched. Reconstruction with FBP and AM algorithms conferred similar mean pDECT accuracy. However, single-voxel pDECT estimates reconstructed on a 1 × 1 × 3 mm{sup 3} grid are shown to be highly sensitive to reconstructed image uncertainty; in some cases pDECT attenuation coefficient estimates exhibited standard deviations on the order of 20% around the mean. Reconstruction with the statistical AM algorithm led to standard deviations roughly 40% to 60% less than FBP reconstruction. Additional tin filtration of the high energy beam exhibits similar pDECT estimation accuracy as the unfiltered beam, even when scanning with only 25% of the dose. Using the law of propagated uncertainty, low Z materials are found to be more sensitive to image reconstruction errors than high Z materials. Furthermore, it is estimated that reconstructed CT image uncertainty must be limited to less than 0.25% to achieve a target linear-attenuation coefficient estimation uncertainty of 3% at 28 keV. Conclusions: That pDECT supports mean linear attenuation coefficient measurement accuracies of 1% of reference values for energies greater than 30 keV is encouraging. However, the sensitivity of the pDECT measurements to noise and systematic errors in reconstructed CT images warrants further investigation in more complex phantom geometries. The investigated statistical reconstruction algorithm, AM, reduced random measurement uncertainty relative to FBP owing to improved noise performance. These early results also support efforts to increase DE spectral separation, which can further reduce the pDECT sensitivity to measurement uncertainty.

61 RADIATION PROTECTION AND DOSIMETRY↗

On linear structure and phase rotation invariant properties of block 2(sup l)-PSK modulation codes

Two important structural properties of block 2(l)-ary PSK (phase shift keying) modulation codes, linear structure and phase symmetry, are investigated. For an additive white Gaussian noise (AWGN) channel, the error performance of a modulation code depends on its squared Euclidean distance distribution. Linear structure of a code makes the error performance analysis much easier. Phase symmetry of a code is important in resolving carrier phase ambiguity and ensuring rapid carrier phase resynchronization after temporary loss of synchronization. It is desirable for a code to have as many phase symmetries as possible. A 2(l)-ary modulation code is represented here as a code with symbols from the integer group. S sub 2(l) PSK = (0,1,2,...,2(l)-1), under the modulo-2(l) addition. The linear structure of block 2(l)-ary PSK modulation codes over S sub 2(l)-ary PSK with respect to the modulo-2(l) vector addition is defined, and conditions under which a block 2(l)-ary PSK modulation code is linear are derived. Once the linear structure is developed, phase symmetry of a block 2(l)-ary PSK modulation code is studied. It is a necessary and sufficient condition for a block 2(l)-PSK modulation code, which is linear as a binary code, to be invariant under 180 deg/2(l-h) phase rotation, for 1 is less than or equal to h is less than or equal to l. A list of short 8-PSK and 16-PSK modulation codes is given, together with their linear structure and the smallest phase rotation for which a code is invariant.

Lin, Shu↗

SU-D-BRD-07: Evaluation of the Effectiveness of Statistical Process Control Methods to Detect Systematic Errors For Routine Electron Energy Verification

Purpose: To evaluate the ability of statistical process control methods to detect systematic errors when using a two dimensional (2D) detector array for routine electron beam energy verification. Methods: Electron beam energy constancy was measured using an aluminum wedge and a 2D diode array on four linear accelerators. Process control limits were established. Measurements were recorded in control charts and compared with both calculated process control limits and TG-142 recommended specification limits. The data was tested for normality, process capability and process acceptability. Additional measurements were recorded while systematic errors were intentionally introduced. Systematic errors included shifts in the alignment of the wedge, incorrect orientation of the wedge, and incorrect array calibration. Results: Control limits calculated for each beam were smaller than the recommended specification limits. Process capability and process acceptability ratios were greater than one in all cases. All data was normally distributed. Shifts in the alignment of the wedge were most apparent for low energies. The smallest shift (0.5 mm) was detectable using process control limits in some cases, while the largest shift (2 mm) was detectable using specification limits in only one case. The wedge orientation tested did not affect the measurements as this did not affect the thickness of aluminum over the detectors of interest. Array calibration dependence varied with energy and selected array calibration. 6 MeV was the least sensitive to array calibration selection while 16 MeV was the most sensitive. Conclusion: Statistical process control methods demonstrated that the data distribution was normally distributed, the process was capable of meeting specifications, and that the process was centered within the specification limits. Though not all systematic errors were distinguishable from random errors, process control limits increased the ability to detect systematic errors using routine measurement of electron beam energy constancy.

07 ISOTOPE AND RADIATION SOURCES↗

Poster — Thur Eve — 56: Design of Quality Assurance Methodology for VMAT system on Agility System equipped with CVDR

The aim of this study was to analyze the feasibility of designing comprehensive QA plans using iComCAT for Elekta machines equipped with Agility multileaf collimator and continuously variable dose rate. Test plans with varying MLC speed, gantry speed, and dose rate were created and delivered in a controlled manner. A strip test was designed with three 1 cm MLC positions and delivered using dynamic, StepNShoot and VMAT techniques. Plans were also designed to test error in MLC position with various gantry speeds and various MLC speeds. The delivery fluence was captured using the electronic portal-imaging device. Gantry speed was found to be within tolerance as per the Canadian standards. MLC positioning errors at higher MLC speed with gravity effects does add more than 2 mm discrepancy. More tests need to be performed to evaluate MLC performance using independent measurement systems. The treatment planning system with end-to-end testing necessary for commissioning was also investigated and found to have >95% passing rates within 3%/3mm gamma criteria. Future studies involve performing off-axis gantry starshot pattern and repeating the tests on three matched Elekta linear accelerators.

07 ISOTOPE AND RADIATION SOURCES↗

Predicting Chemical Reaction Barriers with a Machine Learning Model

In the past few decades, tremendous advances have been made in the understanding of catalysis at solid surfaces. Despite this, most discoveries of materials for improved catalytic performance are made by a slow trial and error process in an experimental laboratory. Computational simulations have begun to provide a way to rationally design materials for optimizing catalytic performance, but due to the high computational expense of calculating transition state energies, simulations cannot adequately screen the phase space of materials. In this work, we attempt to mitigate this expense by using a machine learning approach to predict the most expensive and most important parameter in a catalyst’s affinity for a reaction: the reaction barrier. Previous methods which used the step reaction energy as the only parameter in a linear regression had a mean absolute error (MAE) on the order of 0.4 eV, too high to be used predictively. Here in our work, we achieve a MAE of about 0.22 eV, a marked improvement towards the goal of computational prediction of catalytic activity.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

One-step error correction for multipartite polarization entanglement

We present two economical one-step error-correction protocols for multipartite polarization-entangled systems in a Greenberger-Horne-Zeilinger state. One uses spatial entanglement to correct errors in the polarization entanglement of an N-photon system, resorting to linear optical elements. The other uses frequency entanglement to correct errors in the polarization entanglement of an N-photon system. The parties in quantum communication can obtain a maximally entangled state from each N-photon system transmitted with one step in these two protocols, and both of their success probabilities are 100%, in principle. That is, they both work in a deterministic way, and they do not largely consume the less-entangled photon systems, which is far different from conventional multipartite entanglement purification schemes. These features may make these two protocols more useful for practical applications in long-distance quantum communication.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

SU-E-T-545: MLC Distance Travelled as a Predictor for Motor Failure

Purpose: To study the frequency of Multi-Leaf Collimator (MLC) leaf failures, investigate methods to predict them and reduce linac downtime. Methods: A Varian HD120 MLC was used in our study. The hyperterminal MLC errors logged from 06/2012 to 12/2014 were collected. Along with the hyperterminal errors, the MLC motor changes and all other MLC interventions by the linear accelerator engineer were recorded. The MLC dynalog files were also recorded on a daily basis for each treatment and during linac QA. The dynalog files were analyzed to calculate root mean square errors (RMS) and cumulative MLC travel distance per motor. An in-house MatLab code was used to analyze all dynalog files, record RMS errors and calculate the distance each MLC traveled per day. Results: A total of 269 interventions were recorded over a period of 18 months. Of these, 146 included MLC motor leaf change, 39 T-nut replacements, and 84 MLC cleaning sessions. Leaves close to the middle of each side required the most maintenance. In the A bank, leaves A27 to A40 recorded 73% of all interventions, while the same leaves in the B bank counted for 52% of the interventions. On average, leaves in the middle of the bank had their motors changed approximately every 1500m of travel. Finally, it was found that the number of RMS errors increased prior to an MLC motor change. Conclusion: An MLC dynalog file analysis software was developed that can be used to log daily MLC usage. Our eighteen-month data analysis showed that there is a correlation between the distance an MLC travels, the RMS and the life of the MLC motor. We plan to use this tool to predict MLC motor failures and with proper and timely intervention, reduce the downtime of the linac during clinical hours.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

On the Relationship Between Global Land-Ocean Temperature and Various Descriptors of Solar-Geomagnetic Activity and Climate

Examined are sunspot cycle- (SC-) length averages of the annual January-December values of the Global Land-Ocean Temperature Index ( ) in relation to SC-length averages of annual values of various descriptors of solar-geomagnetic activity and climate, incorporating lags of 0-5 yr. For the overall interval SC12-SC23, the is inferred to correlate best against the parameter incorporating lag = 5 yr, where the parameter refers to the resultant aa value having removed that portion of the annual aa average value due to the yearly variation of sunspot number (SSN). The inferred correlation between the and is statistically important at confidence level cl > 99.9%, having a coefficient of linear correlation r = 0.865 and standard error of estimate se = 0.149 degC. Excluding the most recent cycles SC22 and SC23, the inferred correlation is stronger, having r = 0.969 and se = 0.048 degC. With respect to the overall trend in the , which has been upwards towards warmer temperatures since SC12 (1878-1888), solar-geomagnetic activity parameters are now trending downwards (since SC19). For SC20-SC23, in contrast, comparison of the against SC-length averages of the annual value of the Mauna Loa carbon dioxide ( ) index is found to be highly statistically important (cl >> 99.9%), having r = 0.9994 and se = 0.012 degC for lag = 2 yr. On the basis of the inferred preferential linear correlation between the and , the current ongoing SC24 is inferred to have warmer than was seen in SC23 (i.e., >0.526 degC), probably in excess of 0.68 degC (relative to the 1951-1980 base period).

Wilson, Robert M.↗

DETECTING CHANGING POLARIZATION STRUCTURES IN SAGITTARIUS A* WITH HIGH FREQUENCY VLBI

Sagittarius A* is the source of near infrared, X-ray, radio, and (sub)millimeter emission associated with the supermassive black hole at the Galactic Center. In the submillimeter regime, Sgr A* exhibits time-variable linear polarization on timescales corresponding to <10 Schwarzschild radii of the presumed 4 x 10{sup 6} M {sub sun} black hole. In previous work, we demonstrated the potential for total-intensity (sub)millimeter-wavelength very long baseline interferometry (VLBI) to detect time-variable-and periodic-source structure changes in the Sgr A* black hole system using nonimaging analyses. Here, we extend this work to include full polarimetric VLBI observations. We simulate full-polarization (sub)millimeter VLBI data of Sgr A* using a hot spot model that is embedded within an accretion disk, with emphasis on nonimaging polarimetric data products that are robust against calibration errors. Although the source-integrated linear polarization fraction in the models is typically only a few percent, the linear polarization fraction on small angular scales can be much higher, enabling the detection of changes in the polarimetric structure of Sgr A* on a wide variety of baselines. The shortest baselines track the source-integrated linear polarization fraction, while longer baselines are sensitive to polarization substructures that are beam-diluted by connected-element interferometry. The detection of periodic variability in source polarization should not be significantly affected even if instrumental polarization terms cannot be calibrated out. As more antennas are included in the (sub)millimeter-VLBI array, observations with full polarization will provide important new diagnostics to help disentangle intrinsic source polarization from Faraday rotation effects in the accretion and outflow region close to the black hole event horizon.

79 ASTRONOMY AND ASTROPHYSICS↗

Sliding Mode Control of the X-33 with an Engine Failure

Ascent flight control of the X-3 is performed using two XRS-2200 linear aerospike engines. in addition to aerosurfaces. The baseline control algorithms are PID with gain scheduling. Flight control using an innovative method. Sliding Mode Control. is presented for nominal and engine failed modes of flight. An easy to implement, robust controller. requiring no reconfiguration or gain scheduling is demonstrated through high fidelity flight simulations. The proposed sliding mode controller utilizes a two-loop structure and provides robust. de-coupled tracking of both orientation angle command profiles and angular rate command profiles in the presence of engine failure, bounded external disturbances (wind gusts) and uncertain matrix of inertia. Sliding mode control causes the angular rate and orientation angle tracking error dynamics to be constrained to linear, de-coupled, homogeneous, and vector valued differential equations with desired eigenvalues. Conditions that restrict engine failures to robustness domain of the sliding mode controller are derived. Overall stability of a two-loop flight control system is assessed. Simulation results show that the designed controller provides robust, accurate, de-coupled tracking of the orientation angle command profiles in the presence of external disturbances and vehicle inertia uncertainties, as well as the single engine failed case. The designed robust controller will significantly reduce the time and cost associated with flying new trajectory profiles or orbits, with new payloads, and with modified vehicles

Shtessel, Yuri B.↗

Forecasting longitudinal changes in oropharyngeal tumor morphology throughout the course of head and neck radiation therapy

Purpose: To create models that forecast longitudinal trends in changing tumor morphology and to evaluate and compare their predictive potential throughout the course of radiation therapy. Methods: Two morphology feature vectors were used to describe 35 gross tumor volumes (GTVs) throughout the course of intensity-modulated radiation therapy for oropharyngeal tumors. The feature vectors comprised the coordinates of the GTV centroids and a description of GTV shape using either interlandmark distances or a spherical harmonic decomposition of these distances. The change in the morphology feature vector observed at 33 time points throughout the course of treatment was described using static, linear, and mean models. Models were adjusted at 0, 1, 2, 3, or 5 different time points (adjustment points) to improve prediction accuracy. The potential of these models to forecast GTV morphology was evaluated using leave-one-out cross-validation, and the accuracy of the models was compared using Wilcoxon signed-rank tests. Results: Adding a single adjustment point to the static model without any adjustment points decreased the median error in forecasting the position of GTV surface landmarks by the largest amount (1.2 mm). Additional adjustment points further decreased the forecast error by about 0.4 mm each. Selection of the linear model decreased the forecast error for both the distance-based and spherical harmonic morphology descriptors (0.2 mm), while the mean model decreased the forecast error for the distance-based descriptor only (0.2 mm). The magnitude and statistical significance of these improvements decreased with each additional adjustment point, and the effect from model selection was not as large as that from adding the initial points. Conclusions: The authors present models that anticipate longitudinal changes in tumor morphology using various models and model adjustment schemes. The accuracy of these models depended on their form, and the utility of these models includes the characterization of patient-specific response with implications for treatment management and research study design.

60 APPLIED LIFE SCIENCES↗