Search NASA⌕ Search

SEARCH · Search NASA

Results for “ROC analysis”

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 109 records · Page 6

Comparative performance of multiview stereoscopic and mammographic display modalities for breast lesion detection

Purpose: Mammography is known to be one of the most difficult radiographic exams to interpret. Mammography has important limitations, including the superposition of normal tissue that can obscure a mass, chance alignment of normal tissue to mimic a true lesion and the inability to derive volumetric information. It has been shown that stereomammography can overcome these deficiencies by showing that layers of normal tissue lay at different depths. If standard stereomammography (i.e., a single stereoscopic pair consisting of two projection images) can significantly improve lesion detection, how will multiview stereoscopy (MVS), where many projection images are used, compare to mammography? The aim of this study was to assess the relative performance of MVS compared to mammography for breast mass detection. Methods: The MVS image sets consisted of the 25 raw projection images acquired over an arc of approximately 45 deg. using a Siemens prototype breast tomosynthesis system. The mammograms were acquired using a commercial Siemens FFDM system. The raw data were taken from both of these systems for 27 cases and realistic simulated mass lesions were added to duplicates of the 27 images at the same local contrast. The images with lesions (27 mammography and 27 MVS) and the images without lesions (27 mammography and 27 MVS) were then postprocessed to provide comparable and representative image appearance across the two modalities. All 108 image sets were shown to five full-time breast imaging radiologists in random order on a state-of-the-art stereoscopic display. The observers were asked to give a confidence rating for each image (0 for lesion definitely not present, 100 for lesion definitely present). The ratings were then compiled and processed using ROC and variance analysis. Results: The mean AUC for the five observers was 0.614{+-}0.055 for mammography and 0.778{+-}0.052 for multiview stereoscopy. The difference of 0.164{+-}0.065 was statistically significant with a p-value of 0.0148. Conclusions: The differences in the AUCs and the p-value suggest that multiview stereoscopy has a statistically significant advantage over mammography in the detection of simulated breast masses. This highlights the dominance of anatomical noise compared to quantum noise for breast mass detection. It also shows that significant lesion detection can be achieved with MVS without any of the artifacts associated with tomosynthesis.

60 APPLIED LIFE SCIENCES↗

Computer-aided diagnosis for classifying benign versus malignant thyroid nodules based on ultrasound images: A comparison with radiologist-based assessments

Purpose: To develop a semiautomated computer-aided diagnosis (CAD) system for thyroid cancer using two-dimensional ultrasound images that can be used to yield a second opinion in the clinic to differentiate malignant and benign lesions. Methods: A total of 118 ultrasound images that included axial and longitudinal images from patients with biopsy-confirmed malignant (n = 30) and benign (n = 29) nodules were collected. Thyroid CAD software was developed to extract quantitative features from these images based on thyroid nodule segmentation in which adaptive diffusion flow for active contours was used. Various features, including histogram, intensity differences, elliptical fit, gray-level co-occurrence matrixes, and gray-level run-length matrixes, were evaluated for each region imaged. Based on these imaging features, a support vector machine (SVM) classifier was used to differentiate benign and malignant nodules. Leave-one-out cross-validation with sequential forward feature selection was performed to evaluate the overall accuracy of this method. Additionally, analyses with contingency tables and receiver operating characteristic (ROC) curves were performed to compare the performance of CAD with visual inspection by expert radiologists based on established gold standards. Results: Most univariate features for this proposed CAD system attained accuracies that ranged from 78.0% to 83.1%. When optimal SVM parameters that were established using a grid search method with features that radiologists use for visual inspection were employed, the authors could attain rates of accuracy that ranged from 72.9% to 84.7%. Using leave-one-out cross-validation results in a multivariate analysis of various features, the highest accuracy achieved using the proposed CAD system was 98.3%, whereas visual inspection by radiologists reached 94.9% accuracy. To obtain the highest accuracies, “axial ratio” and “max probability” in axial images were most frequently included in the optimal feature sets for the authors’ proposed CAD system, while “shape” and “calcification” in longitudinal images were most frequently included in the optimal feature sets for visual inspection by radiologists. The computed areas under curves in the ROC analysis were 0.986 and 0.979 for the proposed CAD system and visual inspection by radiologists, respectively; no significant difference was detected between these groups. Conclusions: The use of thyroid CAD to differentiate malignant from benign lesions shows accuracy similar to that obtained via visual inspection by radiologists. Thyroid CAD might be considered a viable way to generate a second opinion for radiologists in clinical practice.

61 RADIATION PROTECTION AND DOSIMETRY↗

Association of Serum Bile Acid and Unsaturated Fatty Acid Profiles with the Risk of Diabetic Retinopathy in Type 2 Diabetic Patients

Aim: We aimed to identify the ability of serum bile acids (BAs) and unsaturated fatty acids (UFAs) profiles to predict the development of diabetic retinopathy (DR) in type 2 diabetes mellitus (T2DM) patients. Methods: We first used univariate and multivariate analysis to compare 15 serum BA and 11 UFA levels in healthy control (HC) group (n = 82), T2DM patients with DR (n = 58) and T2DM patients without DR (n = 60). Forty T2DM patients were considered for validation. Then, the receiver operating characteristic curve (ROC) and decision curve analysis were used to assess the diagnostic value and clinical benefit of serum biomarkers alone, clinical variables alone or in combination, and the area under the curve (AUC), integrated discrimination improvement (IDI), and net reclassification improvement (NRI) were used to further assess whether the addition of biomarkers significantly improved the predictive ability of the model. Results: Orthogonal partial least squares-discriminant analysis (OPLS-DA) of serum BAs and UFAs separated the three cohorts including HC, T2DM patients with or without DR. The difference in serum BA and UFA profiles of T2DM patients with or without DR was mainly manifested in the three metabolites of taurolithocholic acid (TLCA), tauroursodeoxycholic acid (TUDCA) and arachidonic acid (AA). Together, they had an AUC of 0.785 (0.918 for validation cohort) for predicting DR in T2DM patients. After adjusting for numerous confounding factors, TLCA, TUDCA, and AA were independent predictors that differentiated T2DM with or without DR. The results of AUC, IDI, and NRI demonstrated that adding these three biomarkers to a model with clinical variables statistically increased their predictive value and were replicated in our independent validation cohort. Conclusion: These findings highlight the association of three metabolites, TLCA, TUDCA and AA, with DR and may indicate their potential value in the pathogenesis of DR.

60 APPLIED LIFE SCIENCES↗

Performance of a machine learning-based decision model to help clinicians decide the extent of lymphadenectomy (D1 vs. D2) in gastric cancer before surgical resection

Background: Controversy still exists on the optimal surgical resection for potentially curable gastric cancer (GC). Use of radiologic evaluation and machine learning algorithms might predict extent of lymphadenectomy to limit unnecessary surgical treatment. We purposed to design a machine learning-based clinical decision-support model for predicting extent of lymphadenectomy (D1 vs. D2) in local advanced GC. Methods: Clinicoradiologic features available from routine clinical assignments in 557 patients with GCs were retrospectively interpreted by an expert panel blinded to all histopathologic information. All patients underwent surgery using standard D2 resection. Decision models were developed with a logistic regression (LR), support vector machine (SVM) and auto-encoder (AE) algorithm in 371 training and tested in 186 test data, respectively. The primary end point was to measure diagnostic performance of decision model and a Japanese gastric cancer treatment guideline version 4th (JPN 4th) criteria for discriminate D1 (pT1 + pN0) versus D2 (≥ pT1 + ≥ pN1) lymphadenectomy. Results: The decision model with AE analysis produced highest area under ROC curve (train: 0.965, 95% confidence interval (CI) 0.948–0.978; test: 0.946, 95% CI 0.925–0.978), followed by SVM (train: 0.925, 95% CI 0.902–0.944; test: 0.942, 95% CI 0.922–0.973) and LR (train: 0.886, 95% CI 0.858–0.910; test: 0.891, 95% CI 0.891–0.952). By this improvement, overtreatment was reduced from 21.7% (121/557) by treat-all pattern, to 15.1% (84/557) by JPN 4th criteria, and to 0.7–0.9% (4–5/557) by the new approach. Conclusions: The decision model with machine learning analysis demonstrates high accuracy for identifying patients who are candidates for D1 versus D2 resection. Its approximate 14–20% improvements in overtreatment compared to treat-all pattern and JPN 4th criteria potentially increase the number of patients with local advanced GCs who can safely avoid unnecessary lymphadenectomy.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Tonopah Test Range Air Monitoring: CY2017 Meteorological, Radiological, and Wind Transported Particulate Observations

In 1963, the U.S. Department of Energy (DOE) (formerly the Atomic Energy Commission [AEC]), implemented Operation Roller Coaster on the Tonopah Test Range (TTR) and an adjacent area of the Nevada Test and Training Range (NTTR) (formerly the Nellis Air Force Range). This operation resulted in radionuclide-contaminated soils at the Double Tracks site and Clean Slate I, II, and III sites. This report documents observations made during ongoing monitoring of radiological, meteorological, and dust conditions at stations installed adjacent to the Clean Slate sites and at the TTR Sandia National Laboratories (SNL) Range Operations Center (ROC). The primary objective of the monitoring effort is to support the DOE Environmental Management Nevada Program (EM NV) in the safe cleanup of the environmental legacy of nuclear device development and testing by determining if wind blowing across the Clean Slate sites is transporting particles of radionuclide-contaminated soil beyond the physical and administrative boundaries of the sites. The monitoring program underwent a significant change during 2017 with the decommissioning of one monitoring station and the establishment of three new stations. Station 400, located within TTR Area 3 and near the ROC, monitors conditions near the local workforce center and was in operation the full calendar year. Stations 401 and 403 are located on the northwest and southeast perimeter fence lines, respectively, of the Clean Slate III site. Station 401 operated throughout 2017, whereas Station 403 began operation at the end of April 2017. Stations 404 and 405 are along the northern and eastern boundary fence, respectively, at Clean Slate II and both began data collection at the end of April 2017. Also at the end of April, the Clean Slate I monitoring site, Station 402, was decommissioned. The stations are located to be generally downwind of the contaminated areas during either north-northwesterly or south-southeasterly winds, which are the predominant wind directions. The stations—which are similar in design to the Community Environmental Monitoring Program (CEMP) stations operating at locations surrounding the Nevada National Security Site and TTR—include meteorological instruments and continuous-flow, low-volume air samplers. Three pressurized ionization chambers for measuring gamma energy are in use at Stations 400, 401, and 403 (formerly at 402). Saltation sensors and saltation traps are deployed at the stations adjacent to the Clean Slate sites. Detailed meteorological data are recorded on data loggers, with periodic uploads via a satellite system to the Western Regional Climate Center (WRCC) at Desert Research Institute (DRI), to monitor instrument and site conditions. Air filter samples are collected every two weeks and the deposits in the saltation traps are collected when a sufficient sample mass for analysis has accumulated (generally a six- to eight-month interval). Dust transport by suspension and saltation is strongly dependent on wind speed. Concentrations of PM10 (particulate matter of aerodynamic diameter ≤10 micrometers [μm], a size fraction of small particles that are suspended in the air and can be easily inhaled) remain low until winds exceed approximately 32 km/hr (20 mph). Saltation particle counts also increase above the same general threshold wind speed. Wind speeds in excess of 32 km/hr (20 mph) occur less than three percent of the time. High winds are associated with two predominant directions: north-northwest and south. In 2017, the very highest winds were from the north, but all winds above 40 km/hr (25 mph) were approximately evenly split between south and northwest directions. The low-volume air samplers collect total suspended dust, and saltating particles are captured by the passive saltation traps, both for radiological analysis. Radionuclide monitoring at the ROC in TTR Area 3 (Station 400) found natural background levels of gross alpha and gross beta radiation. Similar background levels were found on filter samples analyzed using gamma and alpha spectroscopy. The gamma exposure rate data from pressurized ionization chambers (PIC) are comparable to those from the regional CEMP stations. Plutonium concentrations above background values were measured during 2017 from soil samples collected in passive saltation traps and on select air filters from the monitoring stations adjacent to the Clean Slate sites. The background for soils in the region is assumed to be 0.014 pCi/g plutonium-239+240 ( 239+240 Pu) as measured by Turner et al. (2003), whereas the background concentration of suspended 239+240Pu is assumed to be 9.5 x 10 -19 μCi/ml, which was determined by continuous air monitoring at the TTR airport in 1996 and 1997 (SNL, 1998). Background concentrations are exceeded by one to four orders of magnitude by the majority of samples from Clean Slate I, II, and III. The gross alpha and gross beta measurements of the air filters are also generally higher than those measured at the regional CEMP stations, although the gamma exposure rates determined by the PICs located at Clean Slate I and III are similar to CEMP measurements. The highest radionuclide concentrations of suspended dust are associated with individual filter samples from Clean Slate II Station 404 and Clean Slate III Station 401. Both of the stations are downwind of their respective sites during southerly winds, but they are also geographically closer to the highly contaminated detonation areas than the other monitoring stations. Considering all the analyzed samples, the 239+240Pu concentrations are generally higher for the Clean Slate II samples compared with those from Clean Slate III. The highest gross alpha and gross beta results are also observed for the Clean Slate II stations. The highest radionuclide concentrations are measured on the smaller (≤63 μm) saltation-sized particles and particulate matter collected downwind of each site. Nonetheless, traps oriented to collect dust moving toward the contamination areas also had plutonium concentrations significantly above background values, which is consistent with observations of low-level contamination dispersed beyond the fenced area. All the observed 239+240 Pu concentrations are below the risk-based action level (25 millirem per year, equivalent to a 239+240 Pu concentration of 4,120 pCi/g for an industrial worker scenario; U.S. DOE, 2014). The meteorological and particulate matter monitoring indicate that conditions for wind-borne contaminant movement exist at the Clean Slate sites, and that transport of radionuclide-contaminated soil by both suspension and saltation is occurring. The Corrective Action Unit (CAU) closure strategy uses a risk-based approach, whereby acceptable contaminant concentrations are determined as a function of anticipated human exposure. This strategy is consistent with the fence lines at the Clean Slate sites encircling areas of higher concentration and the observed presence of lower concentrations of contamination outside the fences.

54 ENVIRONMENTAL SCIENCES↗

Can MRI be used to diagnose histologic grade in T1a (< 4 cm) clear cell renal cell carcinomas?

Objective: To assess whether MRI can differentiate low-grade from high-grade T1a cc-RCC. Materials and methods: With IRB approval, 49 consecutive solid < 4 cm cc-RCC (low grade [Grade 1 or 2] N = 38, high grade [Grade 3] N = 11) with pre-operative MRI before nephrectomy were identified between 2013 and 2018. Tumor size, apparent diffusion coefficient (ADC) histogram analysis, enhancement wash-in and wash-out rates, and chemical shift signal intensity index (SI index) were assessed by a blinded radiologist. Subjectively, two blinded Radiologists also assessed for (1) microscopic fat, (2) homogeneity (5-point Likert scale), and (3) ADC signal (relative to renal cortex); discrepancies were resolved by consensus. Outcomes were studied using Chi square, multivariate analysis, logistic regression modeling, and ROC. Inter-observer agreement was assessed using Cohen’s kappa. Results: Tumor size was 24 ± 7 (13–39) mm with no association to grade (p = 0.45). Among quantitative features studied, corticomedullary phase wash-in index (p = 0.015), SI index (p = 0.137), and tenth-centile ADC (p = 0.049) were higher in low-grade tumors. 36.8% (14/38) low-grade tumors versus zero high-grade tumors demonstrated microscopic fat (p = 0.015; Kappa = 0.67). Microscopic fat was specific for low-grade disease (100.0% [71.5–100.0]) with low sensitivity (36.8% [21.8–54.6]). Other subjective features did not differ between groups (p > 0.05). A logistic regression model combining microscopic fat + wash-in index + tenth-centile-ADC yielded area under ROC curve 0.98 (Confidence Intervals 0.94–1.0) with sensitivity/specificity 87.5%/100%. Conclusion: The combination of microscopic fat, higher corticomedullary phase wash-in and higher tenth-centile ADC is highly accurate for diagnosis of low-grade disease among T1a clear cell RCC.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

An Investigation of State-Space Model Fidelity for SSME Data

In previous studies, a variety of unsupervised anomaly detection techniques for anomaly detection were applied to SSME (Space Shuttle Main Engine) data. The observed results indicated that the identification of certain anomalies were specific to the algorithmic method under consideration. This is the reason why one of the follow-on goals of these previous investigations was to build an architecture to support the best capabilities of all algorithms. We appeal to that goal here by investigating a cascade, serial architecture for the best performing and most suitable candidates from previous studies. As a precursor to a formal ROC (Receiver Operating Characteristic) curve analysis for validation of resulting anomaly detection algorithms, our primary focus here is to investigate the model fidelity as measured by variants of the AIC (Akaike Information Criterion) for state-space based models. We show that placing constraints on a state-space model during or after the training of the model introduces a modest level of suboptimality. Furthermore, we compare the fidelity of all candidate models including those embodying the cascade, serial architecture. We make recommendations on the most suitable candidates for application to subsequent anomaly detection studies as measured by AIC-based criteria.

Martin, Rodney Alexander↗

Bayesian classification of falls risk

Background. Prior research in falls risk prediction often relies on qualitative and/or clinical methods. There are two challenges with these methods. First, qualitative methods typically use falls history to determine falls risk. Second, clinical methods do not quantify the uncertainty in the classification decision. In this paper, we propose using Bayesian classification to predict falls risk using vectors of gait variables shown to contribute to falls risk. Research Questions: (1) Using a vector of risk ratios for specific gait variables shown to contribute to falls risk, how can older adults be classified as low or high falls risk? and (2) how can the uncertainty in the classifier decision be quantified when using a vector of gait variables? Methods. Using a pressure sensitive walkway, biomechanical measurements of gait were collected from 854 adults over the age of 65. In our method, we first determine low and high falls risk labels for vectors of risk ratios using the k-means algorithm. Next, the posterior probability of low or high falls risk class membership is obtained from a two component Gaussian mixture model (GMM) of gait vectors, which enables risk assessment directly from the underlying biomechanics. We classify the gait vectors using a threshold based on Youden's J statistic. Results. Through a Monte Carlo simulation and an analysis of the receiver operating characteristic (ROC), we demonstrate that our Bayesian classifier, when compared to the k-means falls risk labels, achieves an accuracy greater than 96% at predicting low or high falls risk. Significance. Our analysis indicates that our approach based on a Bayesian framework and an individual's underlying biomechanics can predict falls risk while quantifying uncertainty in the classification decision.

59 BASIC BIOLOGICAL SCIENCES↗

SU-E-J-104: Single Photon Image From PET with Insertable SPECT Collimator for Boron Neutron Capture Therapy: A Feasibility Study

Purpose: The aim of our proposed system is to confirm the feasibility of extraction of two types of images from one positron emission tomography (PET) module with an insertable collimator for brain tumor treatment during the BNCT. Methods: Data from the PET module, neutron source, and collimator was entered in the Monte Carlo n-particle extended (MCNPX) source code. The coincidence events were first compiled on the PET detector, and then, the events of the prompt gamma ray were collected after neutron emission by using a single photon emission computed tomography (SPECT) collimator on the PET. The obtaining of full width at half maximum (FWHM) values from the energy spectrum was performed to collect effective events for reconstructed image. In order to evaluate the images easily, five boron regions in a brain phantom were used. The image profiles were extracted from the region of interest (ROI) of a phantom. The image was reconstructed using the ordered subsets expectation maximization (OSEM) reconstruction algorithm. The image profiles and the receiver operating characteristic (ROC) curve were compiled for quantitative analysis from the two kinds of reconstructed image. Results: The prompt gamma ray energy peak of 478 keV appeared in the energy spectrum with a FWHM of 41 keV (6.4%). On the basis of the ROC curve in Region A to Region E, the differences in the area under the curve (AUC) of the PET and SPECT images were found to be 10.2%, 11.7%, 8.2% (center, Region C), 12.6%, and 10.5%, respectively. Conclusion: We attempted to acquire the PET and SPECT images simultaneously using only PET without an additional isotope. Single photon images were acquired using an insertable collimator on a PET detector. This research was supported by the Leading Foreign Research Institute Recruitment Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, Information and Communication Technologies (ICT) and Future Planning (MSIP)(Grant No.2009 00420) and the Radiation Technology R and D program (Grant No.2013M2A2A7043498), Republic of Korea.

07 ISOTOPE AND RADIATION SOURCES↗

Evaluating Alerting and Guidance Performance of a UAS Detect-And-Avoid System

A key challenge to the routine, safe operation of unmanned aircraft systems (UAS) is the development of detect-and-avoid (DAA) systems to aid the UAS pilot in remaining "well clear" of nearby aircraft. The goal of this study is to investigate the effect of alerting criteria and pilot response delay on the safety and performance of UAS DAA systems in the context of routine civil UAS operations in the National Airspace System (NAS). A NAS-wide fast-time simulation study was conducted to assess UAS DAA system performance with a large number of encounters and a broad set of DAA alerting and guidance system parameters. Three attributes of the DAA system were controlled as independent variables in the study to conduct trade-off analyses: UAS trajectory prediction method (dead-reckoning vs. intent-based), alerting time threshold (related to predicted time to LoWC), and alerting distance threshold (related to predicted Horizontal Miss Distance, or HMD). A set of metrics, such as the percentage of true positive, false positive, and missed alerts, based on signal detection theory and analysis methods utilizing the Receiver Operating Characteristic (ROC) curves were proposed to evaluate the safety and performance of DAA alerting and guidance systems and aid development of DAA system performance standards. The effect of pilot response delay on the performance of DAA systems was evaluated using a DAA alerting and guidance model and a pilot model developed to support this study. A total of 18 fast-time simulations were conducted with nine different DAA alerting threshold settings and two different trajectory prediction methods, using recorded radar traffic from current Visual Flight Rules (VFR) operations, and supplemented with DAA-equipped UAS traffic based on mission profiles modeling future UAS operations. Results indicate DAA alerting distance threshold has a greater effect on DAA system performance than DAA alerting time threshold or ownship trajectory prediction method. Further analysis on the alert lead time (time in advance of predicted loss of well clear at which a DAA alert is first issued) indicated a strong positive correlation between alert lead time and DAA system performance (i.e. the ability of the UAS pilot to maneuver the unmanned aircraft to remain well clear). While bigger distance thresholds had beneficial effects on alert lead time and missed alert rate, it also generated a higher rate of false alerts. In the design and development of DAA alerting and guidance systems, therefore, the positive and negative effects of false alerts and missed alerts should be carefully considered to achieve acceptable alerting system performance by balancing false and missed alerts. The results and methodology presented in this study are expected to help stakeholders, policymakers and standards committees define the appropriate setting of DAA system parameter thresholds for UAS that ensure safety while minimizing operational impacts to the NAS and equipage requirements for its users before DAA operational performance standards can be finalized.

UAS Detect-and-Avoid (DAA) System↗

Investigation of Radius of Curvature Measurements using Laser Interferometry [Slides]

Goals for this work are to: develop an application using a ZYGO Verifire for calibration of artifacts where the measurand is defined as a Radius of Curvature (ROC), validate the measurements through an uncertainty analysis, and determine if the calibration method is a viable alternative/option for calibration outside of current standard methodologies.

42 ENGINEERING↗

SU-E-T-399: Determination of the Radiobiological Parameters That Describe the Dose-Response Relations of Xerostomia and Disgeusia From Head and Neck Radiotherapy

Purpose: To estimate the radiobiological parameters that describe the doseresponse relations of xerostomia and disgeusia from head and neck cancer radiotherapy. To identify the organs that are best correlated with the manifestation of those clinical endpoints. Finally, to evaluate the goodnessof- fit by comparing the model predictions against the actual clinical results. Methods: In this study, 349 head and neck cancer patients were included. For each patient the dose volume histograms (DVH) of parotids (separate and combined), mandible, submandibular glands (separate and combined) and salivary glands were calculated. The follow-up of those patients was recorded at different times after the completion of the treatment (7 weeks, 3, 7, 12, 18 and 24 months). Acute and late xerostomia and acute disgeusia were the clinical endpoints examined. A maximum likelihood fitting was performed to calculate the best estimates of the parameters used by the relative seriality model. The statistical methods of the error distribution, the receiver operating characteristic (ROC) curve, the Pearson's test and the Akaike's information criterion were utilized to assess the goodness-of-fit and the agreement between the pattern of the radiobiological predictions with that of the clinical records. Results: The estimated values of the radiobiological parameters of salivary glands are D50 = 25.2 Gy, γ = 0.52, s = 0.001. The statistical analysis confirmed the clinical validity of those parameters (area under the ROC curve = 0.65 and AIC = 38.3). Conclusion: The analysis proved that the treatment outcome pattern of the patient material can be reproduced by the relative seriality model and the estimated radiobiological parameters. Salivary glands were found to have strong volume dependence (low relative seriality). Diminishing the biologically effective uniform dose to salivary glands below 30 Gy may significantly reduce the risk of complications to the patients irradiated for prostate cancer.

60 APPLIED LIFE SCIENCES↗

TU-D-207B-05: Intra-Tumor Partitioning and Texture Analysis of DCE-MRI Identifies Relevant Tumor Subregions to Predict Early Pathological Response of Breast Cancer to Neoadjuvant Chemotherapy

Purpose: To predict early pathological response of breast cancer to neoadjuvant chemotherapy (NAC) based on quantitative, multi-region analysis of dynamic contrast enhancement magnetic resonance imaging (DCE-MRI). Methods: In this institution review board-approved study, 35 patients diagnosed with stage II/III breast cancer were retrospectively investigated using DCE-MR images acquired before and after the first cycle of NAC. First, principal component analysis (PCA) was used to reduce the dimensionality of the DCE-MRI data with a high-temporal resolution. We then partitioned the whole tumor into multiple subregions using k-means clustering based on the PCA-defined eigenmaps. Within each tumor subregion, we extracted four quantitative Haralick texture features based on the gray-level co-occurrence matrix (GLCM). The change in texture features in each tumor subregion between pre- and during-NAC was used to predict pathological complete response after NAC. Results: Three tumor subregions were identified through clustering, each with distinct enhancement characteristics. In univariate analysis, all imaging predictors except one extracted from the tumor subregion associated with fast wash-out were statistically significant (p< 0.05) after correcting for multiple testing, with area under the ROC curve or AUCs between 0.75 and 0.80. In multivariate analysis, the proposed imaging predictors achieved an AUC of 0.79 (p = 0.002) in leave-one-out cross validation. This improved upon conventional imaging predictors such as tumor volume (AUC=0.53) and texture features based on whole-tumor analysis (AUC=0.65). Conclusion: The heterogeneity of the tumor subregion associated with fast wash-out on DCE-MRI predicted early pathological response to neoadjuvant chemotherapy in breast cancer.

60 APPLIED LIFE SCIENCES↗

Type A Verification Report for the High Flux Beam Reactor Stack and Grounds, Brookhaven National Laboratory, Upton, New York

The U.S. Department of Energy (DOE) Order 458.1 requires independent verification (IV) of DOE cleanup projects (DOE 2011). The Oak Ridge Institute for Science and Education (ORISE) has been designated as the responsible organization for IV of the High Flux Beam Reactor (HFBR) Stack and Grounds area at Brookhaven National Laboratory (BNL) in Upton, New York. The IV evaluation may consist of an in-process inspection with document and data reviews (Type A Verification) or a confirmatory survey of the site (Type B Verification). DOE and ORISE determined that a Type A verification of the documents and data for the HFBR Stack and Grounds: Survey Units (SU) 6, 7, and 8 was appropriate based on the initial survey unit classification, the walkover surveys, and the final analytical results provided by the Brookhaven Science Associates (BSA). The HFBR Stack and Grounds surveys began in June 2011 and were completed in September 2011. Survey activities by BSA included gamma walkover scans and sampling of the as-left soils in accordance with the BSA Work Procedure (BNL 2010a). The Field Sampling Plan - Stack and Remaining HFBR Outside Areas (FSP) stated that gamma walk-over surveys would be conducted with a bare sodium iodide (NaI) detector, and a collimated detector would be used to check areas with elevated count rates to locate the source of the high readings (BNL 2010b). BSA used the Mult- Agency Radiation Survey and Site Investigation Manual (MARSSIM) principles for determining the classifications of each survey unit. Therefore, SUs 6 and 7 were identified as Class 1 and SU 8 was deemed Class 2 (BNL 2010b). Gamma walkover surveys of SUs 6, 7, and 8 were completed using a 2X2 NaI detector coupled to a data-logger with a global positioning system (GPS). The 100% scan surveys conducted prior to the final status survey (FSS) sampling identified two general soil areas and two isolated soil locations with elevated radioactivity. The general areas of elevated activity identified were investigated further with a collimated NaI detector. The uncollimated average gamma count rate was less than 15,000 counts per minute (cpm) for the SU 6, 7, and 8 composite area (BNL 2011a). Elevated count rates were observed in portions of each survey unit. The general areas of elevated counts near the Building 801 ventilation and operations and the entry to the Stack were determined to be directly related to the radioactive processes in those structures. To compensate for this radioactive shine, a collimated or shielded detector was used to lower the background count rate (BNL 2011b and c). This allowed the surveyor(s) to distinguish between background and actual radioactive contamination. Collimated gamma survey count rates in these shine affected areas were below 9,000 cpm (BNL 2011a). The average background count rate of 7,500 cpm was reported by BSA for uncollimated NaI detectors (BNL 2011d). The average collimated background ranged from 4,500-6,500 cpm in the westernmost part of SU 8 and from 2,000-3,500 cpm in all other areas (BNL 2011e). Based on these data, no further investigations were necessary for these general areas. SU 8 was the only survey unit that exhibited verified elevated radioactivity levels. The first of two isolated locations of elevated radioactivity had an uncollimated direct measurement of 50,000 cpm with an area background of 7,500 cpm (BNL 2011f). The second small area exhibiting elevated radiation levels was identified at a depth of 6 inches from the surface. The maximum reported count rate of 28,000 cpm was observed during scanning (BNL 2011g). The affected areas were remediated, and the contaminated soils were placed in an intermodal container for disposal. BSA&#x27;s post-remediation walkover surveys were expanded to include a 10-foot radius around the excavated locations, and it was determined that further investigation was not required for these areas (BNL 2011 f and g). The post-remediation soil samples were collected and analyzed with onsite gamma spectroscopy equipment. These samples were also included with the FSS samples that were analyzed at an offsite facility for the primary radionuclides of concern (ROCs) (i.e., cesium-137, strontium-90, and radium-226). Analysis included full spectrum gamma spectroscopy for all samples. Sr-90 analysis was completed on all samples from SUs 6 and 7. However Sr-90 was only completed on the cores, composites, field blank and duplicate samples in SU 8. Alpha spectroscopy as well as liquid scintillation performed for tritium, carbon-14, and nickel-63 concentrations were completed on the composite samples from SUs 6 and 7. Various cores, composites, and samples from the remediated areas of SU 8 received alpha spectroscopy as well as liquid scintillation analyses for tritium, carbon-14, and nickel-63 to determine respective ROC concentrations (BNL 2011 h, i, and j). BSA submitted the FSS data and analytical results to demonstrate that remediation efforts complied with the specified cleanup goal of less than or equal to 15 millirem per year (mrem/yr) above background to a resident in 50 years (BNL 2011a through j). ORISE has reviewed the project documentation and FSS data for the HFBR Stack and Grounds: SUs 6, 7, and 8. The highest concentrations of the primary ROCs reported were 5.92 picocuries per gram (pCi/g) for Cs-137 and 2.03 pCi/g for Sr-90, with both ROCs having the qualifier for the sample result as less than the minimum detectable activity (MDA). For Ra-226, the highest detected concentration from the FSS data provided was 0.682 pCi/g. Other potential secondary contaminants were below their respective MDAs. Therefore, ORISE is of the opinion that BSA has provided sufficient evidence to demonstrate compliance with the 15 mrem/yr cleanup objectives for the final status survey data provided.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Using Receiver Operating Characteristic Curves to Optimize Discovery-Based Software with Comprehensive Two-Dimensional Gas Chromatography with Time-of-Flight Mass Spectrometry

We report a quantitative approach to optimize implementation of discovery-based software for comprehensive two-dimensional gas chromatography coupled with time-of-flight mass spectrometry (GC × GC – TOFMS). The software performs a tile-based Fisher ratio (F-ratio) analysis, and facilitates a supervised non-targeted analysis based upon the experimental design to aid in the discovery of analytes whose variances are statistically different between sample classes. The quantitative approach for software optimization uses receiver operating characteristic (ROC) curves. The area under the curve (AUC) of each ROC curve serves as a quantitative metric to optimize two key algorithm parameters: the signal-to-noise ratio (S/N) threshold of the baseline corrected and normalized data prior to calculating F-ratios at each m/z mass channel, and the number of these F-ratios per m/z to include in the calculation of the average F-ratio of a tile. A total of 25 combinations of S/N threshold by number of m/z were studied. Fifty analytes were spiked into a diesel fuel at two concentration spike levels to produce two sample classes that should in principle produce 50 positive instances in the ROC curves (30 native/20 non-native compounds, at spike levels of 200/20 and100/10 ppm native/non-native, respectively). The “sweet spot” for F-ratio analysis was determined to be a S/N threshold of 10 coupled with a maximum of the 10 most chemically selective m/z (requiring a minimum of 3 m/z), corresponding to a ~ 21% improvement in the discovery rate of true positives relative to prior studies. Furthermore, it was observed that optimization of these software parameters did not depend upon a priori determination of the statistically correct number of positive instances in the sample classes. The AUC metric appears to be suitable for the evaluation of all data analysis methods that utilize the proper experimental design.

3D-GS/TOFMS, Data discovery, ROC curves, Fisher ra↗

A Dose–Response Analysis of Biochemical Control Outcomes After {sup 125}I Monotherapy for Patients With Favorable-Risk Prostate Cancer

Purpose: To define the optimal dose for {sup 125}I prostate implants by correlating postimplantation dosimetry findings with biochemical failure and toxicity. Methods and Materials: Between 2003 and 2009, 683 patients with prostate cancer were treated with {sup 125}I prostate brachytherapy without supplemental external beam radiation therapy and were followed up for a median time of 80 months. Implant dose was defined as the D90 (the minimal dose received by 90% of the prostate) on postoperative day 1 and 1 month after implantation. Therefore, 2 dosimetric variables (day 1 D90 and day 30 D90) were analyzed for each patient. We investigated the dose effects on biochemical control and toxicity. Results: The 7-year biochemical failure-free survival (BFFS) rate for the group overall was 96.4% according to the Phoenix definition. A multivariate analysis found day 1 D90 and day 30 D90 to be the most significant factors affecting BFFS. The cutoff points for day 1 D90 and day 30 D90, calculated from ROC curves, were 163 Gy and 175 Gy, respectively. By use of univariate analysis, various dosimetric cutoff points for day 30 D90 were tested. We found that day 30 D90 cutoff points from 130 to 180 Gy appeared to be good for the entire cohort. Greater D90s were associated with an increase in late genitourinary or gastrointestinal toxicity ≥ grade 2, but the increase was not statistically significant. Conclusions: Improvements in BFFS rates were seen with increasing D90 levels. Day 30 D90 doses of 130 to 180 Gy were found to serve as cutoff levels. For low-risk and low-tier intermediate-risk prostate cancer patients, high prostate D90s, even with doses exceeding 180 Gy, achieve better treatment results and are feasible.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

SU-F-R-22: Malignancy Classification for Small Pulmonary Nodules with Radiomics and Logistic Regression

Purpose: We conducted a retrospective study of Radiomics research for classifying malignancy of small pulmonary nodules. A machine learning algorithm of logistic regression and open research platform of Radiomics, IBEX (Imaging Biomarker Explorer), were used to evaluate the classification accuracy. Methods: The training set included 100 CT image series from cancer patients with small pulmonary nodules where the average diameter is 1.10 cm. These patients registered at Chang Gung Memorial Hospital and received a CT-guided operation of lung cancer lobectomy. The specimens were classified by experienced pathologists with a B (benign) or M (malignant). CT images with slice thickness of 0.625 mm were acquired from a GE BrightSpeed 16 scanner. The study was formally approved by our institutional internal review board. Nodules were delineated and 374 feature parameters were extracted from IBEX. We first used the t-test and p-value criteria to study which feature can differentiate between group B and M. Then we implemented a logistic regression algorithm to perform nodule malignancy classification. 10-fold cross-validation and the receiver operating characteristic curve (ROC) were used to evaluate the classification accuracy. Finally hierarchical clustering analysis, Spearman rank correlation coefficient, and clustering heat map were used to further study correlation characteristics among different features. Results: 238 features were found differentiable between group B and M based on whether their statistical p-values were less than 0.05. A forward search algorithm was used to select an optimal combination of features for the best classification and 9 features were identified. Our study found the best accuracy of classifying malignancy was 0.79±0.01 with the 10-fold cross-validation. The area under the ROC curve was 0.81±0.02. Conclusion: Benign nodules may be treated as a malignant tumor in low-dose CT and patients may undergo unnecessary surgeries or treatments. Our study may help radiologists to differentiate nodule malignancy for low-dose CT.

60 APPLIED LIFE SCIENCES↗

Differentiation of pancreatic neuroendocrine tumors from pancreas renal cell carcinoma metastases on CT using qualitative and quantitative features

Purpose: To assess qualitative and quantitative imaging features on enhanced CT that may differentiate pancreatic neuroendocrine tumors (PNETs) from pancreatic renal cell carcinoma (RCC) metastases. Methods: This IRB-approved multi-center retrospective case–control study compared 43 resected PNETs and 28 resected RCC metastases with pre-operative enhanced CT identified consecutively between 2003 and 2017. Two blinded radiologists (R1/R2) independently assessed tumor location, attenuation (relative to pancreas), composition (solid/cystic/mixed), homogeneity (homogeneous/heterogeneous), calcification, multiplicity, and for main pancreatic duct (MPD) dilation. Tumors were segmented for quantitative texture analysis. Data were analyzed with Chi square, logistic regression, and receiver operating characteristic (ROC). Inter-observer agreement was assessed (Cohen’s kappa). Results: There was no difference in age, gender, location, attenuation, or composition (P > 0.05) between groups. PNETs were larger than RCC metastases (37 ± 23 mm vs. 26 ± 21 mm, P = 0.038), more frequently solitary (P < 0.001), subjectively more heterogeneous (P = 0.033/0.144, R1/R2), and associated with calcification (P = 0.002/0.004) and MPD dilation (P = 0.025/0.006). Agreement for subjective features was moderate-to-almost perfect (K = 0.4879–0.9481). Quantitative texture analysis showed higher entropy in PNETs (6.32 ± 0.49 versus 5.96 ± 0.53; P = 0.004) with no difference in other features studied (P > 0.05). Entropy had ROC area under the curve for diagnosis of PNET of 0.77 ± 0.06, with optimal sensitivity/specificity of 71.4/79.1%. Conclusions: Compared to pancreatic RCC metastases, PNETs are larger, more frequently solitary, contain calcification, show MPD dilation, and are subjectively and quantitatively more heterogeneous tumors.

62 RADIOLOGY AND NUCLEAR MEDICINE↗