Search NASA⌕ Search

SEARCH · Search NASA

Results for “timing error evaluation”

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

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

At least 73 records · Page 4

The potential of positron emission tomography for intratreatment dynamic lung tumor tracking: A phantom study

Purpose: This study aims to evaluate the potential and feasibility of positron emission tomography for dynamic lung tumor tracking during radiation treatment. The authors propose a center of mass (CoM) tumor tracking algorithm using gated-PET images combined with a respiratory monitor and investigate the geometric accuracy of the proposed algorithm. Methods: The proposed PET dynamic lung tumor tracking algorithm estimated the target position information through the CoM of the segmented target volume on gated PET images reconstructed from accumulated coincidence events. The information was continuously updated throughout a scan based on the assumption that real-time processing was supported (actual processing time at each frame ≈10 s). External respiratory motion and list-mode PET data were acquired from a phantom programmed to move with measured respiratory traces (external respiratory motion and internal target motion) from human subjects, for which the ground truth target position was known as a function of time. The phantom was cylindrical with six hollow sphere targets (10, 13, 17, 22, 28, and 37 mm in diameter). The measured respiratory traces consisted of two sets: (1) 1D-measured motion from ten healthy volunteers and (2) 3D-measured motion from four lung cancer patients. The authors evaluated the geometric accuracy of the proposed algorithm by quantifying estimation errors (Euclidean distance) between the actual motion of targets (1D-motion and 3D-motion traces) and CoM trajectories estimated by the proposed algorithm as a function of time. Results: The time-averaged error of 1D-motion traces over all trajectories of all targets was 1.6 mm. The error trajectories decreased with time as coincidence events were accumulated. The overall error trajectory of 1D-motion traces converged to within 2 mm in approximately 90 s. As expected, more accurate results were obtained for larger targets. For example, for the 37 mm target, the average error over all 1D-motion traces was 1.1 mm; and for the 10 mm target, the average error over all 1D-motion traces was 2.8 mm. The overall time-averaged error of 3D-motion traces was 1.6 mm, which was comparable to that of the 1D-motion traces. There were small variations in the errors between the 3D-motion traces, although the motion trajectories were very different. The accuracy of the estimates was consistent for all targets except for the smallest. Conclusions: The authors developed an algorithm for dynamic lung tumor tracking using list-mode PET data and a respiratory motion signal, and demonstrated proof-of-principle for PET-guided lung tumor tracking. The overall tracking error in phantom studies is less than 2 mm.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Geometric and dosimetric accuracy of dynamic tumor-tracking conformal arc irradiation with a gimbaled x-ray head

Purpose: The Vero4DRT system has the capability for dynamic tumor-tracking (DTT) stereotactic irradiation using a unique gimbaled x-ray head. The purposes of this study were to develop DTT conformal arc irradiation and to estimate its geometric and dosimetric accuracy. Methods: The gimbaled x-ray head, supported on an O-ring gantry, was moved in the pan and tilt directions during O-ring gantry rotation. To evaluate the mechanical accuracy, the gimbaled x-ray head was moved during the gantry rotating according to input command signals without a target tracking, and a machine log analysis was performed. The difference between a command and a measured position was calculated as mechanical error. To evaluate beam-positioning accuracy, a moving phantom, which had a steel ball fixed at the center, was driven based on a sinusoidal wave (amplitude [A]: 20 mm, time period [T]: 4 s), a patient breathing motion with a regular pattern (A: 16 mm, average T: 4.5 s), and an irregular pattern (A: 7.2–23.0 mm, T: 2.3–10.0 s), and irradiated with DTT during gantry rotation. The beam-positioning error was evaluated as the difference between the centroid position of the irradiated field and the steel ball on images from an electronic portal imaging device. For dosimetric accuracy, dose distributions in static and moving targets were evaluated with DTT conformal arc irradiation. Results: The root mean squares (RMSs) of the mechanical error were up to 0.11 mm for pan motion and up to 0.14 mm for tilt motion. The RMSs of the beam-positioning error were within 0.23 mm for each pattern. The dose distribution in a moving phantom with tracking arc irradiation was in good agreement with that in static conditions. Conclusions: The gimbal positional accuracy was not degraded by gantry motion. As in the case of a fixed port, the Vero4DRT system showed adequate accuracy of DTT conformal arc irradiation.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Effect of Breast Volume on Treatment Reproducibility on a Tomotherapy Unit in the Treatment of Breast Cancer

Purpose: To determine whether the volume of a patient's breast is correlated with reliable daily setup in treatment of breast cancer with a helical tomotherapy treatment unit. Methods and Materials: Thirty-six consecutive patients with breast cancer were treated on a helical tomotherapy unit. During simulation, kilovoltage CT images were obtained for treatment planning. These were fused with daily megavoltage CT scans, and after setup based on skin marks and laser alignment the necessary shifts were carried out. The magnitude of daily shifts (in millimeters) was retrospectively obtained from the daily image fusions, and the breast volume was obtained from the treatment plan. A total of 873 fusion scans were reviewed. Random error for absolute and directional daily shifts was evaluated for correlation to volume. Variation over time was also evaluated. Results: Mean (SD) random shift for all patients in the lateral, longitudinal, and vertical directions was 2.7 (2.0), 3.1 (1.5), and 3.2 (2.6) mm, respectively. Mean (SD) absolute distance shifted was 6.0 (3.5) mm. There was no significant correlation between mean absolute or mean directional daily shift and breast volume (0.08, 0.08, 0.22, and 0.07, respectively). There was no correlation between setup variation and time. Conclusions: In this cohort, there was no correlation between breast volume and degree of daily shift. There was no correlation between time course and setup variation. Therefore, setup variation does not improve or degrade with repeated treatment setups.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

A simple soil mass correction for a more accurate determination of soil carbon stock changes

Abstract Agricultural soils can act as a sink for large quantities of soil organic carbon (SOC) but can also be sources of carbon to the atmosphere. The international standard for assessing SOC stock and measuring stock change stipulates fixed depth sampling to at least 30 cm. The tendency of bulk density (BD) to decrease with decreasing disturbance and increasing SOC concentration and the assumption of constant SOC and BD within this depth profile promotes error in the estimates of SOC stock. A hypothetical but realistic change in BD from 1.5 to 1.1 g cm −3 from successive fixed depth sampling to 30 cm underestimates SOC stock change by 17%. Significant effort has been made to evaluate and reduce this fixed depth error by using the equivalent soil mass (ESM) approach, but with limited adoption. We evaluate the error in SOC stock assessment and change generated from fixed depth measurements over time relative to the ESM approach and propose a correction that can be readily adopted under current sampling and analytical methods. Our approach provides a more accurate estimate of SOC stock accumulation or loss that will help incentivize management practice changes that reduce the environmental impacts of agriculture and further legitimize the accounting practices used by the emerging carbon market and organizations that have pledged to reduce their supply chain greenhouse gas (GHG) footprints.

60 APPLIED LIFE SCIENCES↗

WE-AB-BRA-01: 3D-2D Image Registration for Target Localization in Spine Surgery: Comparison of Similarity Metrics Against Robustness to Content Mismatch

Purpose: In image-guided spine surgery, mapping 3D preoperative images to 2D intraoperative images via 3D-2D registration can provide valuable assistance in target localization. However, the presence of surgical instrumentation, hardware implants, and soft-tissue resection/displacement causes mismatches in image content, confounding existing registration methods. Manual/semi-automatic methods to mask such extraneous content is time consuming, user-dependent, error prone, and disruptive to clinical workflow. We developed and evaluated 2 novel similarity metrics within a robust registration framework to overcome such challenges in target localization. Methods: An IRB-approved retrospective study in 19 spine surgery patients included 19 preoperative 3D CT images and 50 intraoperative mobile radiographs in cervical, thoracic, and lumbar spine regions. A neuroradiologist provided truth definition of vertebral positions in CT and radiography. 3D-2D registration was performed using the CMA-ES optimizer with 4 gradient-based image similarity metrics: (1) gradient information (GI); (2) gradient correlation (GC); (3) a novel variant referred to as gradient orientation (GO); and (4) a second variant referred to as truncated gradient correlation (TGC). Registration accuracy was evaluated in terms of the projection distance error (PDE) of the vertebral levels. Results: Conventional similarity metrics were susceptible to gross registration error and failure modes associated with the presence of surgical instrumentation: for GI, the median PDE and interquartile range was 33.0±43.6 mm; similarly for GC, PDE = 23.0±92.6 mm respectively. The robust metrics GO and TGC, on the other hand, demonstrated major improvement in PDE (7.6 ±9.4 mm and 8.1± 18.1 mm, respectively) and elimination of gross failure modes. Conclusion: The proposed GO and TGC similarity measures improve registration accuracy and robustness to gross failure in the presence of strong image content mismatch. Such registration capability could offer valuable assistance in target localization without disruption of clinical workflow. G. Kleinszig and S. Vogt are employees of Siemens Healthcare.

60 APPLIED LIFE SCIENCES↗

An Experiment-based Distribution Level Performance Comparison among PMUs

This paper presents a experiment-based distribution level performance comparison among three Phasor Measurement Units (PMUs). Several evaluation criteria, including the total vector error, the phase angle error, the frequency error, the rate of change of frequency, the response time, the settling time, the overshoot, and the algorithm window size, are selected to compare the static and dynamic performances of the PMUs under steady state and step response test conditions. In order to have a more realistic test environment, a test is setup in which PMUs under evaluation have exactly the same input signals. The quantitative experiment result analysis gives an end-user guideline to the PMU selection regarding distribution level applications.

Yin, He↗

Evaluating wind profiles in a numerical weather prediction model with Doppler lidar

We use Doppler lidar wind profiles from six locations around the globe to evaluate the wind profile forecasts in the boundary layer generated by the operational global Integrated Forecast System (IFS) from the European Centre for Medium-range Weather Forecasts (ECMWF). The six locations selected cover a variety of surfaces with different characteristics (rural, marine, mountainous urban, and coastal urban). We first validated the Doppler lidar observations at four locations by comparison with co-located radiosonde profiles to ensure that the Doppler lidar observations were of sufficient quality. The two observation types agree well, with the mean absolute error (MAE) in wind speed almost always less than 1 m s –1 . Large deviations in the wind direction were usually only seen for low wind speeds and are due to the wind direction uncertainty increasing rapidly as the wind speed tends to zero. Time–height composites of the wind evaluation with 1 h resolution were generated, and evaluation of the model winds showed that the IFS model performs best over marine and coastal locations, where the mean absolute wind vector error was usually less than 3 m s –1 at all heights within the boundary layer. Larger errors were seen in locations where the surface was more complex, especially in the wind direction. For example, in Granada, which is near a high mountain range, the IFS model failed to capture a commonly occurring mountain breeze, which is highly dependent on the sub-grid-size terrain features that are not resolved by the model. The uncertainty in the wind forecasts increased with forecast lead time, but no increase in the bias was seen. At one location, we conditionally performed the wind evaluation based on the presence or absence of a low-level jet diagnosed from the Doppler lidar observations. The model was able to reproduce the presence of the low-level jet, but the wind speed maximum was about 2 m s –1 lower than observed. This is attributed to the effective vertical resolution of the model being too coarse to create the strong gradients in wind speed observed. Our results show that Doppler lidar is a suitable instrument for evaluating the boundary layer wind profiles in atmospheric models.

54 ENVIRONMENTAL SCIENCES↗

Analysis framework for systematically studying ionospheric response to impulsive events from below: Analysis Framework Ionospheric Responses

We present that impulsive phenomena in the Earth's atmosphere produce acoustic and gravity waves which perturb the ionosphere. Such perturbations are often measured using total electron content fluctuations (TEC), derived from ground-based Global Positioning System data. Using TEC data from the Japanese GEONET ground network after the Tōhoku earthquake on 11 March 2011, we demonstrate capabilities of a new framework of methodologies for analyzing ionospheric perturbations. The framework consists of several new techniques: calculating velocity along a single direction to reduce error due to anisotropic propagation, producing normalized bidirectional band-pass spectra that preserve relative timing between various frequencies and allowing a more systematic determination of broadband pulses, and utilizing a wavelet-based technique that considers instantaneous wave phase changes, rather than best fit time differences, to evaluate wave characteristics (speed, direction, and wavelength) within spectral ranges of interest. Using these techniques together decreases subjectivity and reduces errors in attributing fluctuations to given sources. In validating this framework using the Tōhoku case, we consistently identify three kinds of waves: a broad-band pulse (speed: >2000 m/s, max range: >1400 km) arriving in the ionosphere 10–15 min after the quake, acoustic waves following the pulse (period: 3–5 min, speed: 700–1000 m/s, max range: <500 km), and gravity waves (period: 10–15 min, speed: 150–500 m/s, max range: >1400 km) propagating away from the epicenter, consistent with theory and demonstrated in previous studies. Lastly, this framework also can be applied to other impulsive events in the atmosphere that are more difficult to detect and attribute to sources.

58 GEOSCIENCES↗

Determining the tilt of the Raman laser beam using an optical method for atom gravimeters

The tilt of a Raman laser beam is a major systematic error in precision gravity measurement using atom interferometry. The conventional approach to evaluating this tilt error involves modulating the direction of the Raman laser beam and conducting time-consuming gravity measurements to identify the error minimum. In this work, we demonstrate a method to expediently determine the tilt of the Raman laser beam by transforming the tilt angle measurement into characterization of parallelism, which integrates the optical method of aligning the laser direction, commonly used in freely falling corner-cube gravimeters, into an atom gravimeter. A position-sensing detector (PSD) is utilized to quantitatively characterize the parallelism between the test beam and the reference beam, thus measuring the tilt precisely and rapidly. After carefully positioning the PSD and calibrating the relationship between the distance measured by the PSD and the tilt angle measured by the tiltmeter, we achieved a statistical uncertainty of less than 30 µrad in the tilt measurement. Furthermore, we compared the results obtained through this optical method with those from the conventional tilt modulation method for gravity measurement. The comparison validates that our optical method can achieve tilt determination with an accuracy level of better than 200 µrad, corresponding to a systematic error of 20 µGal in g measurement. This work has practical implications for real-world applications of atom gravimeters.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Grover-like search via a Frenkel-exciton trapping mechanism

We propose the physical implementation of a Grover-like search problem by means of Frenkel exciton trapping at a shallow isotopic impurity against a background of competing mechanisms. The search, culminating at the impurity molecule, designated the 'winner' site, is marked by its enhanced interaction with acoustic phonons at low temperatures. The quantum search proceeds with the assistance of an oracle-like exciton-phonon interaction that addresses only the impurity site via the Dyson propagator within the Green's function formalism. The optimum parameters of a graph lattice with long-range intersite interactions required to trap the exciton in the fastest time are determined, and estimates of error rates for the naphthalene-doped organic system are evaluated. We extend the analysis of the quantum search to a fluctuating long-range interacting cycle (LRIC) graph-lattice system.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

A Pulsar-Inspired Timing Framework for Power System: Optimization and Performance Evaluation

Due to their excellent stability, neutron pulsar stars are considered promising candidate timing sources for power system applications. However, the complexity of pulsar signals necessitates advanced processing algorithms to provide accurate timing references. This paper presents the foundational framework for pulsar signal processing, serving as the basis for further optimization. To enhance the timing accuracy and computation efficiency in pulsar period searches, three algorithms are proposed as the initial optimization step: wavelet de-noising, fast folding, and cross-correlation for profile evaluation. Wavelet de-noising improves signal-to-noise ratio (SNR) by 36%–70%. Fast folding reduces computation time from hundreds of seconds to mere milliseconds. Cross-correlation works better than traditional SNR-based methods by effectively identifying the optimal period. The performance of the proposed algorithms is evaluated using observation data from telescopes. Together, these algorithms significantly improve pulsar timing performance, reducing the error of the Pulse Per Second (PPS) signal from hundreds to tens of microseconds.

Wu, Ori [ORNL] (ORCID:0000000326723410)↗

Predicting Initial Trans-Membrane Pressure for Optimized Operations in UF Unit Using Random Forest

With the growing scarcity of freshwater, innovative process design mechanisms like Reverse Osmosis (RO) are increasingly gaining attention among water treatment utilities to address the rising demand. Ensuring reliable water production necessitates efficient resource utilization, minimizing downtime in (ultra-filtration) UF systems. Recent advancements in machine learning (ML) have enabled the development of accurate data-driven models for Model Predictive Control (MPC), often requiring minimal prior knowledge of underlying physical processes. In this study, we present predictive regression models based on Random Forest (RF) and Auto-Regressive (AR) approaches to forecast the initial Trans-Membrane Pressure (TMP) for each filtration cycle in data generated by Direct Potable Reuse (DPR) systems. The proposed RF-based model demonstrates superior performance compared to baseline methods, including historical mean, Last Observation Carried Forward (LOCF), and naïve AR models, across various forecasting horizons in terms of root mean square error (RMSE) metric. To evaluate how different classes of process variables contribute to TMP dynamics over time, we examine the feature importance of independent covariates across multiple forecast horizons. This analysis provides insight into the temporal relevance of operational and sensor-derived features, guiding control and monitoring strategies. Additionally, the impact of hyperparameter tuning on TMP prediction performance is studied for both direct and recursive RF modelling approaches across increasing forecast horizons. Accurate prediction of initial TMP is critical for optimizing RO operations, as it enables the development of robust modelling frameworks by accurately estimating membrane fouling trends, thereby enhancing process efficiency and long-term reliability. The demonstrated efficacy of the RF-based approach highlights its potential as a tool for real-time decision-making in water treatment systems, paving the way for advanced process optimization and sustainable water resource management.

Mukherjee, Subrata [ORNL] (ORCID:0000000309930338)↗

CAN-D: A Modular Four-Step Pipeline for Comprehensively Decoding Controller Area Network Data

Controller area networks (CANs) are a broadcast protocol for real-time communication of critical vehicle subsystems. Original equipment manufacturers of passenger vehicles hold secret their mappings of CAN data to vehicle signals, and these definitions vary according to make, model, and year. Without these mappings, the wealth of real-time vehicle information hidden in the CAN packets is uninterpretable, severely impeding vehicle-related research, including CAN cybersecurity and privacy studies, aftermarket tuning, efficiency and performance monitoring, and fault diagnosis to name a few. Guided by the four-part CAN signal definition, we present CAN-D (CAN-Decoder), a modular, four-step pipeline for identifying each signal's boundaries (start bit and length), endianness (byte ordering), signedness (bit-to-integer encoding), and by leveraging diagnostic standards, augmenting a subset of the extracted signals with meaningful, physical interpretation. En route to CAN-D, we provide a comprehensive review of the CAN signal reverse engineering research. All previous methods ignore endianness and signedness, rendering them incapable of decoding many standard CAN signal definitions. Incorporating endianness grows the search space from 128 to 4.72E21 signal tokenizations and introduces a web of changing dependencies. In response, we formulate, formally analyze, and provide an efficient solution to an optimization problem, allowing identification of the optimal set of signal boundaries and byte orderings. In addition, we provide two novel, state-of-the-art signal boundary classifiers—both of which are superior to previous approaches in precision and recall in three different test scenarios—and the first signedness classification algorithm, which exhibits a $>$ 97% F-score. Altogether, CAN-D is the only solution with the potential to extract any CAN signal that is also the state of the art. In evaluation on 10 vehicles of different makes, CAN-D's average $\ell ^1$ error is five times better (81% less) than all previous methods and exhibits lower average error, even when considering only signals that meet prior methods’ assumptions. Finally, CAN-D is implemented in lightweight hardware, allowing for an on-board diagnostic (OBD-II) plugin for real-time in-vehicle CAN decoding.

42 ENGINEERING↗

Application of asymptotic expansions for maximum likelihood estimators errors to gravitational waves from binary mergers: The single interferometer case

In this paper we apply to gravitational waves (GW) from the inspiral phase of binary systems a recently derived frequentist methodology to calculate analytically the error for a maximum likelihood estimate of physical parameters. We use expansions of the covariance and the bias of a maximum likelihood estimate in terms of inverse powers of the signal-to-noise ration (SNR)s where the square root of the first order in the covariance expansion is the Cramer Rao lower bound (CRLB). We evaluate the expansions, for the first time, for GW signals in noises of GW interferometers. The examples are limited to a single, optimally oriented, interferometer. We also compare the error estimates using the first two orders of the expansions with existing numerical Monte Carlo simulations. The first two orders of the covariance allow us to get error predictions closer to what is observed in numerical simulations than the CRLB. The methodology also predicts a necessary SNR to approximate the error with the CRLB and provides new insight on the relationship between waveform properties, SNR, dimension of the parameter space and estimation errors. For example the timing match filtering can achieve the CRLB only if the SNR is larger than the Kurtosis of the gravitational wave spectrum and the necessary SNR is much larger if other physical parameters are also unknown.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Development of real-time rotating waveplate Stokes polarimeter using multi-order retardation for ITER poloidal polarimeter

The rotating waveplate Stokes polarimeter was developed for ITER (International Thermonuclear Experimental Reactor) poloidal polarimeter. The generalized model of the rotating waveplate Stokes polarimeter and the algorithm suitable for real-time field-programmable gate array (FPGA) processing were proposed. Since the generalized model takes into account each component associated with the rotation of the waveplate, the Stokes parameters can be accurately measured even in unideal condition such as non-uniformity of the waveplate retardation. Experiments using a He-Ne laser showed that the maximum error and the precision of the Stokes parameter were 3.5% and 1.2%, respectively. The rotation speed of waveplate was 20 000 rpm and time resolution of measuring the Stokes parameter was 3.3 ms. Software emulation showed that the real-time measurement of the Stokes parameter with time resolution of less than 10 ms is possible by using several FPGA boards. Evaluation of measurement capability using a far-infrared laser which ITER poloidal polarimeter will use concluded that measurement error will be reduced by a factor of nine.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Molecular radiotherapy: The NUKFIT software for calculating the time-integrated activity coefficient

Purpose: Calculation of the time-integrated activity coefficient (residence time) is a crucial step in dosimetry for molecular radiotherapy. However, available software is deficient in that it is either not tailored for the use in molecular radiotherapy and/or does not include all required estimation methods. The aim of this work was therefore the development and programming of an algorithm which allows for an objective and reproducible determination of the time-integrated activity coefficient and its standard error.Methods: The algorithm includes the selection of a set of fitting functions from predefined sums of exponentials and the choice of an error model for the used data. To estimate the values of the adjustable parameters an objective function, depending on the data, the parameters of the error model, the fitting function and (if required and available) Bayesian information, is minimized. To increase reproducibility and user-friendliness the starting values are automatically determined using a combination of curve stripping and random search. Visual inspection, the coefficient of determination, the standard error of the fitted parameters, and the correlation matrix are provided to evaluate the quality of the fit. The functions which are most supported by the data are determined using the corrected Akaike information criterion. The time-integrated activity coefficient is estimated by analytically integrating the fitted functions. Its standard error is determined assuming Gaussian error propagation. The software was implemented using MATLAB.Results: To validate the proper implementation of the objective function and the fit functions, the results of NUKFIT and SAAM numerical, a commercially available software tool, were compared. The automatic search for starting values was successfully tested for reproducibility. The quality criteria applied in conjunction with the Akaike information criterion allowed the selection of suitable functions. Function fit parameters and their standard error estimated by using SAAM numerical and NUKFIT showed differences of <1%. The differences for the time-integrated activity coefficients were also <1% (standard error between 0.4% and 3%). In general, the application of the software is user-friendly and the results are mathematically correct and reproducible. An application of NUKFIT is presented for three different clinical examples.Conclusions: The software tool with its underlying methodology can be employed to objectively and reproducibly estimate the time integrated activity coefficient and its standard error for most time activity data in molecular radiotherapy.

61 RADIATION PROTECTION AND DOSIMETRY↗

Using an external surrogate for predictor model training in real-time motion management of lung tumors

Purpose: Precise prediction of respiratory motion is a prerequisite for real-time motion compensation techniques such as beam, dynamic couch, or dynamic multileaf collimator tracking. Collection of tumor motion data to train the prediction model is required for most algorithms. To avoid exposure of patients to additional dose from imaging during this procedure, the feasibility of training a linear respiratory motion prediction model with an external surrogate signal is investigated and its performance benchmarked against training the model with tumor positions directly. Methods: The authors implement a lung tumor motion prediction algorithm based on linear ridge regression that is suitable to overcome system latencies up to about 300 ms. Its performance is investigated on a data set of 91 patient breathing trajectories recorded from fiducial marker tracking during radiotherapy delivery to the lung of ten patients. The expected 3D geometric error is quantified as a function of predictor lookahead time, signal sampling frequency and history vector length. Additionally, adaptive model retraining is evaluated, i.e., repeatedly updating the prediction model after initial training. Training length for this is gradually increased with incoming (internal) data availability. To assess practical feasibility model calculation times as well as various minimum data lengths for retraining are evaluated. Relative performance of model training with external surrogate motion data versus tumor motion data is evaluated. However, an internal–external motion correlation model is not utilized, i.e., prediction is solely driven by internal motion in both cases. Results: Similar prediction performance was achieved for training the model with external surrogate data versus internal (tumor motion) data. Adaptive model retraining can substantially boost performance in the case of external surrogate training while it has little impact for training with internal motion data. A minimum adaptive retraining data length of 8 s and history vector length of 3 s achieve maximal performance. Sampling frequency appears to have little impact on performance confirming previously published work. By using the linear predictor, a relative geometric 3D error reduction of about 50% was achieved (using adaptive retraining, a history vector length of 3 s and with results averaged over all investigated lookahead times and signal sampling frequencies). The absolute mean error could be reduced from (2.0 ± 1.6) mm when using no prediction at all to (0.9 ± 0.8) mm and (1.0 ± 0.9) mm when using the predictor trained with internal tumor motion training data and external surrogate motion training data, respectively (for a typical lookahead time of 250 ms and sampling frequency of 15 Hz). Conclusions: A linear prediction model can reduce latency induced tracking errors by an average of about 50% in real-time image guided radiotherapy systems with system latencies of up to 300 ms. Training a linear model for lung tumor motion prediction with an external surrogate signal alone is feasible and results in similar performance as training with (internal) tumor motion. Particularly for scenarios where motion data are extracted from fluoroscopic imaging with ionizing radiation, this may alleviate the need for additional imaging dose during the collection of model training data.

60 APPLIED LIFE SCIENCES↗

Efficient and generalizable nested Fourier-DeepONet for three-dimensional geological carbon sequestration

Geological carbon sequestration (GCS) involves injecting CO2 into subsurface geological formationsfor permanent storage. Numerical simulations could guide decisions in GCS projects by predictingCO 2 migration pathways and the pressure distribution in storage formation. However, these simula-tions are often computationally expensive due to highly coupled physics and large spatial-temporalsimulation domains. Surrogate modelling with data-driven machine learning has become a promis-ing alternative to accelerate physics-based simulations. Among these, the Fourier neural operator(FNO) has been applied to three-dimensional synthetic subsurface models. Despite its good accuracyin simulating CO 2 plume migration, it requires large computational resources in training and alsolacks generalizability. Here, to further improve performance, we have developed a nested Fourier-DeepONet by combining the expressiveness of the FNO with the modularity of a deep operatornetwork (DeepONet). This new framework is twice as efficient as a nested FNO for training and has atleast 80% lower GPU memory requirement due to its flexibility to treat temporal coordinates sepa-rately. These performance improvements are achieved without compromising prediction accuracy.In addition, the generalization and extrapolation ability of nested Fourier-DeepONet beyond thetraining range has been thoroughly evaluated. Nested Fourier-DeepONet outperformed the nestedFNO for extrapolation in time with more than 50% reduced error. It also exhibited good extrapolationaccuracy beyond the training range in terms of reservoir properties, number of wells, and injectionrate.

Lee, Jonathan E. [Department of Chemical and Envir↗