Search NASA⌕ Search

SEARCH · Search NASA

Results for “Selective classification”

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 19 records
All sourcesNASA NTRSDOE OSTI
DOE OSTI2013

THE GALEX TIME DOMAIN SURVEY. I. SELECTION AND CLASSIFICATION OF OVER A THOUSAND ULTRAVIOLET VARIABLE SOURCES

We present the selection and classification of over a thousand ultraviolet (UV) variable sources discovered in {approx}40 deg{sup 2} of GALEX Time Domain Survey (TDS) NUV images observed with a cadence of 2 days and a baseline of observations of {approx}3 years. The GALEX TDS fields were designed to be in spatial and temporal coordination with the Pan-STARRS1 Medium Deep Survey, which provides deep optical imaging and simultaneous optical transient detections via image differencing. We characterize the GALEX photometric errors empirically as a function of mean magnitude, and select sources that vary at the 5{sigma} level in at least one epoch. We measure the statistical properties of the UV variability, including the structure function on timescales of days and years. We report classifications for the GALEX TDS sample using a combination of optical host colors and morphology, UV light curve characteristics, and matches to archival X-ray, and spectroscopy catalogs. We classify 62% of the sources as active galaxies (358 quasars and 305 active galactic nuclei), and 10% as variable stars (including 37 RR Lyrae, 53 M dwarf flare stars, and 2 cataclysmic variables). We detect a large-amplitude tail in the UV variability distribution for M-dwarf flare stars and RR Lyrae, reaching up to |{Delta}m| = 4.6 mag and 2.9 mag, respectively. The mean amplitude of the structure function for quasars on year timescales is five times larger than observed at optical wavelengths. The remaining unclassified sources include UV-bright extragalactic transients, two of which have been spectroscopically confirmed to be a young core-collapse supernova and a flare from the tidal disruption of a star by dormant supermassive black hole. We calculate a surface density for variable sources in the UV with NUV < 23 mag and |{Delta}m| > 0.2 mag of {approx}8.0, 7.7, and 1.8 deg{sup -2} for quasars, active galactic nuclei, and RR Lyrae stars, respectively. We also calculate a surface density rate in the UV for transient sources, using the effective survey time at the cadence appropriate to each class, of {approx}15 and 52 deg{sup -2} yr{sup -1} for M dwarfs and extragalactic transients, respectively.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2013

EFFICIENT SELECTION AND CLASSIFICATION OF INFRARED EXCESS EMISSION STARS BASED ON AKARI AND 2MASS DATA

The selection of young stellar objects (YSOs) based on excess emission in the infrared is easily contaminated by post-main-sequence stars and various types of emission line stars with similar properties. We define in this paper stringent criteria for an efficient selection and classification of stellar sources with infrared excess emission based on combined Two Micron All Sky Survey (2MASS) and AKARI colors. First of all, bright dwarfs and giants with known spectral types were selected from the Hipparcos Catalogue and cross-identified with the 2MASS and AKARI Point Source Catalogues to produce the main-sequence and the post-main-sequence tracks, which appear as expected as tight tracks with very small dispersion. However, several of the main-sequence stars indicate excess emission in the color space. Further investigations based on the SIMBAD data help to clarify their nature as classical Be stars, which are found to be located in a well isolated region on each of the color-color (C-C) diagrams. Several kinds of contaminants were then removed based on their distribution in the C-C diagrams. A test sample of Herbig Ae/Be stars and classical T Tauri stars were cross-identified with the 2MASS and AKARI catalogs to define the loci of YSOs with different masses on the C-C diagrams. Well classified Class I and Class II sources were taken as a second test sample to discriminate between various types of YSOs at possibly different evolutionary stages. This helped to define the loci of different types of YSOs and a set of criteria for selecting YSOs based on their colors in the near- and mid-infrared. Candidate YSOs toward IC 1396 indicating excess emission in the near-infrared were employed to verify the validity of the new source selection criteria defined based on C-C diagrams compiled with the 2MASS and AKARI data. Optical spectroscopy and spectral energy distributions of the IC 1396 sample yield a clear identification of the YSOs and further confirm the criteria defined for exploring the nature and properties of unknown excess emission sources in the infrared without optical identification.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2012

Foundations of Feature Selection and Classification for Non-Gaussian Distributed Targets

This report describes the author’s research toward creating algorithms for selecting useful statistical features in pattern/target classification problems in which the features are non-Gaussian distributed. In engineering practice, it is common to either (1) Not perform any feature selection procedure, or (2) Use a feature selection algorithm that assumes the features are Gaussian distributed. This process can be far from optimal if the data are actually non-Gaussian, as they frequently are in practice. This research has the goal of mitigating that problem by creating a useful algorithm that can be used in practice. The approach is to focus on the performance indices used in common feature selection algorithms, such as Branch and Bound algorithms and the Sequential Forward Selection Algorithm. Ordinarily, the performance index used to measure the class separation in feature space involves assuming the data are Gaussian and deriving tractable performance indices that use closed form equations that represent the probability density functions of the class data. The advantage of this approach is that it produces feature selection algorithms that have low computational complexity and do not require knowledge of the actual data densities. The disadvantage is that these algorithms may not perform reasonably when the data are non-Gaussian, as they commonly are in practice. This research examines the use of information-theoretic class separability measures that can deal with the non-Gaussian case. In particular, this work shows that the Hellinger Distance (a divergence measure), has very desirable mathematical properties and can be useful for feature selection. The work in this paper employs a multivariate kernel density estimator along with the Hellinger Distance and an optimal subset selection algorithm to do feature selection. This report (1) Defines the classification and feature selection problems, (2) Describes the continuous variable Hellinger distance in one dimension and discusses its mathematical properties in comparison with other distance measures, (3) Develops the discrete variable Hellinger distance expression for one and two dimensions, and shows that the discrete Hellinger distance for multiple dimensions is a straightforward extension, and (4) Presents experimental density estimation results using a new MATLAB implementation of a multivariate kernel density estimator. The draft of this report served as a starting point for Ensign Matthew Wilder’s MSEE thesis work at the Naval Postgraduate School (NPS) in 2010 and 2011. Ensign Wilder completed his thesis and graduated from NPS in June 2011 (M. J. Wilder, “Automatic Target Recognition: Statistical Feature Selection of Non-Gaussian Distributed Target Classes,” MSEE Thesis, Naval Postgraduate School, Thesis Advisor: Grace A. Clark, Second Reader: Monique P. Fargues, June 2011). In the thesis, Ensign Wilder created implementations of the Branch and Bound and Sequential Feature Selection algorithms, extended the multivariate kernel density estimator code, and evaluated the performance of Hellinger Distance-based feature selection algorithms using both simulated data and data from a real-world benchmark data set. The new algorithm was shown to be very effective for both Gaussian and non-Gaussian data sets. The advantages gained for non-Gaussian data come at the cost of increased computational complexity. Using a laptop computer for practical problems, we generally use initial feature vectors containing no more than about ten or twelve features. Proposed future research lies in creating faster kernel density estimator algorithms that would allow for the use of higher dimensional data.

42 ENGINEERING↗
DOE OSTI2015

Threshold selection for classification of MR brain images by clustering method

Given a grey-intensity image, our method detects the optimal threshold for a suitable binarization of MR brain images. In MR brain image processing, the grey levels of pixels belonging to the object are not substantially different from the grey levels belonging to the background. Threshold optimization is an effective tool to separate objects from the background and further, in classification applications. This paper gives a detailed investigation on the selection of thresholds. Our method does not use the well-known method for binarization. Instead, we perform a simple threshold optimization which, in turn, will allow the best classification of the analyzed images into healthy and multiple sclerosis disease. The dissimilarity (or the distance between classes) has been established using the clustering method based on dendrograms. We tested our method using two classes of images: the first consists of 20 T2-weighted and 20 proton density PD-weighted scans from two healthy subjects and from two patients with multiple sclerosis. For each image and for each threshold, the number of the white pixels (or the area of white objects in binary image) has been determined. These pixel numbers represent the objects in clustering operation. The following optimum threshold values are obtained, T = 80 for PD images and T = 30 for T2w images. Each mentioned threshold separate clearly the clusters that belonging of the studied groups, healthy patient and multiple sclerosis disease.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2023

Deep learning uncertainty quantification for clinical text classification

Machine learning algorithms are expected to work side-by-side with humans in decision-making pipelines. Thus, the ability of classifiers to make reliable decisions is of paramount importance. Deep neural networks (DNNs) represent the state-of-the-art models to address real-world classification. Although the strength of activation in DNNs is often correlated with the network’s confidence, in-depth analyses are needed to establish whether they are well calibrated. In this paper, we demonstrate the use of DNN-based classification tools to benefit cancer registries by automating information extraction of disease at diagnosis and at surgery from electronic text pathology reports from the US National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) population-based cancer registries. In particular, we introduce multiple methods for selective classification to achieve a target level of accuracy on multiple classification tasks while minimizing the rejection amount—that is, the number of electronic pathology reports for which the model’s predictions are unreliable. We evaluate the proposed methods by comparing our approach with the current in-house deep learning-based abstaining classifier. Overall, all the proposed selective classification methods effectively allow for achieving the targeted level of accuracy or higher in a trade-off analysis aimed to minimize the rejection rate. On in-distribution validation and holdout test data, with all the proposed methods, we achieve on all tasks the required target level of accuracy with a lower rejection rate than the deep abstaining classifier (DAC). Interpreting the results for the out-of-distribution test data is more complex; nevertheless, in this case as well, the rejection rate from the best among the proposed methods achieving 97% accuracy or higher is lower than the rejection rate based on the DAC. We show that although both approaches can flag those samples that should be manually reviewed and labeled by human annotators, the newly proposed methods retain a larger fraction and do so without retraining—thus offering a reduced computational cost compared with the in-house deep learning-based abstaining classifier.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2015

Luminous red galaxies: selection and classification by combining optical and infrared photometry

We describe a new method of combining optical and infrared photometry to select luminous red galaxies (LRGs) at redshifts z>0.6. We explore this technique using a combination of optical photometry from CFHTLS and Hubble Space Telescope, infrared photometry from the Wide-Field Infrared Survey Explorer satellite, and spectroscopic or photometric redshifts from the DEEP2 Galaxy Redshift Survey or COSMOS. We present a variety of methods for testing the success of our selection and present methods for optimization given a set of rest-frame color and redshift requirements. We have tested this selection in two different regions of the sky, the COSMOS and Extended Groth Strip (EGS) fields, to reduce the effect of cosmic/sample variance. We have used these methods to assemble large samples of LRGs for two different ancillary programs as a part of the Sloan Digital Sky Survey (SDSS)-III/Baryon Oscillation Spectroscopic Survey. This technique is now being used to select ∼600,000 LRG targets for SDSS-IV/Extended Baryon Oscillation Spectroscopic Survey, which began observations in Fall 2014, and will be adapted for the proposed Dark Energy Spectroscopic Instrument survey. We have found that these methods can select high-redshift LRGs efficiently with minimal stellar contamination; this is extremely difficult to achieve with selections that rely on optical photometry alone.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2015

Management of Low-Flow Vascular Malformations: Clinical Presentation, Classification, Patient Selection, Imaging and Treatment

This review article aims to give an overview of the current state of imaging, patient selection, agents and techniques used in the management of low-flow vascular malformations. The review includes the current classifications for low-flow vascular malformations including the 2014 updates. Clinical presentation and assessment is covered with a detailed section on the common sclerosant agents used to treat low-flow vascular malformations, including dosing and common complications. Imaging is described with a guide to a simple stratification of the use of imaging for diagnosis and interventional techniques.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2018

Radiation Anomaly Detection and Classification with Bayes Model Selection

We present a new method for radiation anomaly detection that is based on Bayes Model Selection (BMS), together with models for gamma-radiation measurements from benign and threat sources. The method estimates the relative odds of pairs of such models, with the aim of supporting related hypotheses about the nature of the underlying source material. We also discuss partial optimization of the parameters in the models. The method allows measurements to be broadly categorized and screened for sources of interest in real time, a property that should improve the efficiency of mobile search or unattended monitoring operations.

Anomaly detection, Gamma-ray spectroscopy, Radiati↗
DOE OSTI2020

Updates to Relevance Vector Machine: Multiclass Classification, Variable Selection, and Proof-of-Concept Application to Safeguards Fresh Fuel Verification using List-Mode Neutron Collar Data

To expand the capabilities of safeguards authorities to verify the integrity of fresh fuel assemblies, Oak Ridge National Laboratory has retrofit the existing electronics of the JCC-71 uranium neutron coincidence collar, which contains 18 3 He neutron detectors and an external 241 AmLi(α, n) neutron interrogation source arranged to surround a fresh nuclear fuel assembly. The new electronics system allows analysts to record list-mode neutron multiplicity data in addition to the singles and doubles rates that are currently measured. Based on previous proof-of-concept research, analysis of these new data will identify off-normal fuel configurations in an assembly and characterize or localize the specific partial fuel defects. The purpose of this report it to document the analysis algorithm development and then to demonstrate its capability for the safeguards verification of fresh fuel assemblies using list mode neutron collar data. To analyze the complex list-mode data collected with the upgraded uranium neutron collar, multivariate classification algorithms are being developed using a novel classification method, the relevance vector machine. This approach may be applied to multiclass problems to estimate the probability that test data belongs to one of many possible classes of data. In addition, our method identifies the most useful variables/channels for making predictions, which illuminates the basis for the model’s predictions, and this interpretability is largely unique among data analytics methods. Variable selection occurs during model training and parameter tuning and does not need any external hyperparameter tuning routines. Finally, we apply the modified relevance vector machine to a simulated dataset of list-mode neutron collar data generated with the radiation transport code MCNP. The method can correctly identify off-normal fuel configurations, categorize the data according to four fuel defect scenarios, and rank the channels in the data according to prediction utility. For nuclear safeguards applications, it is concluded that this method has the potential to increase the sensitivity and reliability to detect missing fuel rods from a standard 17 x 17 Pressurized Water Reactor (PWR) fresh fuel assembly. Within this analysis, “off-normal” (i.e., missing fuel rods) were correctly classified in 17 simulated test scenarios with one quarter (25%) of the fresh fuel rods missing using a training data set of 58 simulated measurements.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2019

The Photometric LSST Astronomical Time-series Classification Challenge PLAsTiCC: Selection of a Performance Metric for Classification Probabilities Balancing Diverse Science Goals

Classification of transient and variable light curves is an essential step in using astronomical observations to develop an understanding of the underlying physical processes from which they arise. However, upcoming deep photometric surveys, including the Large Synoptic Survey Telescope (LSST), will produce a deluge of low signal-to-noise data for which traditional type estimation procedures are inappropriate. Probabilistic classification is more appropriate for such data but is incompatible with the traditional metrics used on deterministic classifications. Furthermore, large survey collaborations like LSST intend to use the resulting classification probabilities for diverse science objectives, indicating a need for a metric that balances a variety of goals. We describe the process used to develop an optimal performance metric for an open classification challenge that seeks to identify probabilistic classifiers that can serve many scientific interests. The Photometric LSST Astronomical Time-series Classification Challenge (PLASTICC) aims to identify promising techniques for obtaining classification probabilities of transient and variable objects by engaging a broader community beyond astronomy. Using mock classification probability submissions emulating realistically complex archetypes of those anticipated of PLASTICC, we compare the sensitivity of two metrics of classification probabilities under various weighting schemes, finding that both yield results that are qualitatively consistent with intuitive notions of classification performance. We thus choose as a metric for PLASTICC a weighted modification of the cross-entropy because it can be meaningfully interpreted in terms of information content. Finally, we propose extensions of our methodology to ever more complex challenge goals and suggest some guiding principles for approaching the choice of a metric of probabilistic data products.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2015

Selecting a Classification Ensemble and Detecting Process Drift in an Evolving Data Stream

We characterize the commercial behavior of a group of companies in a common line of business using a small ensemble of classifiers on a stream of records containing commercial activity information. This approach is able to effectively find a subset of classifiers that can be used to predict company labels with reasonable accuracy. Performance of the ensemble, its error rate under stable conditions, can be characterized using an exponentially weighted moving average (EWMA) statistic. The behavior of the EWMA statistic can be used to monitor a record stream from the commercial network and determine when significant changes have occurred. Results indicate that larger classification ensembles may not necessarily be optimal, pointing to the need to search the combinatorial classifier space in a systematic way. Results also show that current and past performance of an ensemble can be used to detect when statistically significant changes in the activity of the network have occurred. The dataset used in this work contains tens of thousands of high level commercial activity records with continuous and categorical variables and hundreds of labels, making classification challenging.

Ensemble classifiers↗
DOE OSTI2020

Machine learning-based multidomain processing for texture-based image segmentation and analysis

Atomic and molecular resolved atomic force microscopy (AFM) images offer unique insights into materials' properties such as local ordering, molecular orientation, and topological defects, which can be used to pinpoint physical and chemical interactions occurring at the surface. Utilizing machine learning for extracting underlying physical parameters increases the throughput of AFM data processing and eliminates inconsistencies intrinsic to manual image analysis, thus enabling the creation of reliable frameworks for qualitative and quantitative evaluation of experimental data. Here, we present a robust and scalable approach to the segmentation of AFM images based on flexible pre-selected classification criteria. The usage of supervised learning and feature extraction allows us to retain the consideration of specific problem-dependent features (such as types of periodical structures observed in the images and the associated numerical parameters: spacing, orientation, etc.). We highlight the applicability of this approach for the segmentation of molecular resolved AFM images based on the crystal orientation of the observed domains, automated selection of boundaries, and collection of relevant statistics. Finally, we outline a general strategy for machine learning-enabled analysis of nanoscale systems exhibiting periodic order that could be applied to any analytical imaging technique

36 MATERIALS SCIENCE↗
DOE OSTI2024

Non-averaged single-molecule tertiary structures reveal RNA self-folding through individual-particle cryo-electron tomography

Large-scale and continuous conformational changes in the RNA self-folding process present significant challenges for structural studies, often requiring trade-offs between resolution and observational scope. Here, we utilize individual-particle cryo-electron tomography (IPET) to examine the post-transcriptional self-folding process of designed RNA origami 6-helix bundle with a clasp helix (6HBC). By avoiding selection, classification, averaging, or chemical fixation and optimizing cryo-ET data acquisition parameters, we reconstruct 120 three-dimensional (3D) density maps from 120 individual particles at an electron dose of no more than 168 e - Å -2 , achieving averaged resolutions ranging from 23 to 35 Å, as estimated by Fourier shell correlation (FSC) at 0.5. Each map allows us to identify distinct RNA helices and determine a unique tertiary structure. Statistical analysis of these 120 structures confirms two reported conformations and reveals a range of kinetically trapped, intermediate, and highly compacted states, demonstrating a maturation folding landscape likely driven by helix-helix compaction interactions.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Data-driven selection and spectral classification of white dwarf stars

The next generation of spectroscopic surveys is expected to provide spectra for hundreds of thousands of white dwarf (WD) candidates in the upcoming years. Currently, spectroscopic classification of white dwarfs is mostly done by visual inspection, requiring substantial amounts of expert attention. We propose a data-driven pipeline for fast, automatic selection, and spectroscopic classification of WD candidates, trained using spectroscopically confirmed objects with available Gaia astrometry, photometry, and Sloan Digital Sky Survey (SDSS) spectra with signal-to-noise ratios ≥9. The pipeline selects WD candidates with improved accuracy and completeness over existing algorithms, classifies their primary spectroscopic type with ≥ 90 % accuracy, and spectroscopically detects main sequence companions with similar performance. We apply our pipeline to the Gaia Data Release 3 cross-matched with the SDSS Data Release 17 (DR17), identifying 424 096 high-confidence WD candidates and providing the first catalogue of automated and quantifiable classification for 36 523 WD spectra. Both the catalogue and pipeline are made available online. Such a tool will prove particularly useful for the undergoing SDSS-V survey, allowing for rapid classification of thousands of spectra at every data release.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2020

Grain boundary slip transfer classification and metric selection with artificial neural networks

An artificial neural network is used to evaluate the effectiveness of six metrics and their combinations to assess whether slip transfers across grain boundaries in coarse-grained oligocrystalline Al foils. This approach extends the one- or two-dimensional projections formerly applied to analyze slip transfer. The accuracy of this binary classification reaches around 87% for the best single metric and around 90% when considering two or more metrics simultaneously. Here, the results suggest slip transfer mostly depends on the geometric relationship between grains. Training a double-layer network having 10 nodes per hidden layer with 40 measurements is sufficient to render the maximum accuracy.

36 MATERIALS SCIENCE↗
DOE OSTI2018

The STRong lensing Insights into the Dark Energy Survey (STRIDES) 2016 follow-up campaign – I. Overview and classification of candidates selected by two techniques

Here, the primary goals of the STRong lensing Insights into the Dark Energy Survey (STRIDES) collaboration are to measure the dark energy equation of state parameter and the free streaming length of dark matter. To this aim, STRIDES is discovering strongly lensed quasars in the imaging data of the Dark Energy Survey and following them up to measure time delays, high resolution imaging, and spectroscopy sufficient to construct accurate lens models. In this paper, we first present forecasts for STRIDES. Then, we describe the STRIDES classification scheme, and give an overview of the Fall 2016 follow-up campaign. We continue by detailing the results of two selection methods, the Outlier Selection Technique and a morphological algorithm, and presenting lens models of a system, which could possibly be a lensed quasar in an unusual configuration. We conclude with the summary statistics of the Fall 2016 campaign. Including searches presented in companion papers (Anguita et al.; Ostrovski et al.), STRIDES followed up 117 targets identifying 7 new strongly lensed systems, and 7 nearly identical quasars (NIQs), which could be confirmed as lenses by the detection of the lens galaxy. 76 candidates were rejected and 27 remain otherwise inconclusive, for a success rate in the range 6-35\%. This rate is comparable to that of previous searches like SQLS even though the parent dataset of STRIDES is purely photometric and our selection of candidates cannot rely on spectroscopic information.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Training data selection for event classification in a highly variable environment

A problem of interest for nuclear nonproliferation is monitoring activities at nuclear facilities, where proliferation events may only take place a few times and often under variable conditions. Machine learning has revolutionized data analytics by enabling the use of measurable signatures to generate predictive models of facility operations. However, traditional methods for training these models require large, reliable data sets with labeled observations, a challenge for nonproliferation. Highly variable conditions further complicate this as events from training data may have occurred in conditions quite different from the event of interest. Our hypothesis is that when events occur in a highly variable environment, careful training data selection for each test event could outperform the standard approach of using all available training data. We developed a method to optimize training data selection for the given test event and applied it to predicting the power level of the High Flux Isotope Reactor (HFIR) at Oak Ridge National Laboratory. In this study, the reactor startup exhibits variability between occurrences due to natural variability in environmental conditions and operational procedures. Using a combination of analysis techniques, a similitude assessment was performed on data collected from HFIR to isolate clusters that were optimal for training a predictive model. Concepts such as dynamic time warping and Jaccard similarity were used in conjunction with clustering analysis. In order to validate this approach, the model was trained on every combination of unique training events and the predictive performance was compared to the performance using a subset of the training data selected by isolated clusters found through the similitude assessment.

Iyer, A↗
DOE OSTI2016

WE-AB-209-05: Development of an Ultra-Fast High Quality Whole Breast Radiotherapy Treatment Planning System

Purpose: To enable near-real-time (<20sec) and interactive planning without compromising quality for whole breast RT treatment planning using tangential fields. Methods: Whole breast RT plans from 20 patients treated with single energy (SE, 6MV, 10 patients) or mixed energy (ME, 6/15MV, 10 patients) were randomly selected for model training. Additional 20 cases were used as validation cohort. The planning process for a new case consists of three fully automated steps:1. Energy Selection. A classification model automatically selects energy level. To build the energy selection model, principle component analysis (PCA) was applied to the digital reconstructed radiographs (DRRs) of training cases to extract anatomy-energy relationship.2. Fluence Estimation. Once energy is selected, a random forest (RF) model generates the initial fluence. This model summarizes the relationship between patient anatomy’s shape based features and the output fluence. 3. Fluence Fine-tuning. This step balances the overall dose contribution throughout the whole breast tissue by automatically selecting reference points and applying centrality correction. Fine-tuning works at beamlet-level until the dose distribution meets clinical objectives. Prior to finalization, physicians can also make patient-specific trade-offs between target coverage and high-dose volumes.The proposed method was validated by comparing auto-plans with manually generated clinical-plans using Wilcoxon Signed-Rank test. Results: In 19/20 cases the model suggested the same energy combination as clinical-plans. The target volume coverage V100% was 78.1±4.7% for auto-plans, and 79.3±4.8% for clinical-plans (p=0.12). Volumes receiving 105% Rx were 69.2±78.0cc for auto-plans compared to 83.9±87.2cc for clinical-plans (p=0.13). The mean V10Gy, V20Gy of the ipsilateral lung was 24.4±6.7%, 18.6±6.0% for auto plans and 24.6±6.7%, 18.9±6.1% for clinical-plans (p=0.04, <0.001). Total computational time for auto-plans was < 20s. Conclusion: We developed an automated method that generates breast radiotherapy plans with accurate energy selection, similar target volume coverage, reduced hotspot volumes, and significant reduction in planning time, allowing for near-real-time planning.

60 APPLIED LIFE SCIENCES↗