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 37 records · Page 2

WE-E-18A-10: Comparison of Patient Dose and Vessel Visibility Between Antiscatter Grid Removal and Lower Angiographic Radiation Dose Settings for Pediatric Imaging: A Preclinical Investigation

Purpose: To define an alternative to antiscatter grid (ASG) removal in angiographic systems which achieves similar patient dose reduction as ASG removal without degrading image quality during pediatric imaging. Methods: This study was approved by the local institution animal care and use committee (IACUC). Six different digital subtraction angiography settings were evaluated that altered the mAs, (100, 70, 50, 35, 25, 17.5% of reference mAs) with and without ASG. Three pigs of 5, 15, and 20 kg (9, 15, and 17 cm abdominal thickness; smaller than a newborn, average 3 yr old, and average 10 year old human abdomen respectively) were imaged using the six dose settings with and without ASG. Image quality was defined as the order of vessel branch that is visible relative to the injected vessel. Five interventional radiologists evaluated all images. Image quality and patient dose were statistically compared using analysis of variance and receiver operating curve (ROC) analysis to define the preferred dose level and use of ASG for a minimum visibility of 2nd or 3rd order branches of vessel visibility. Results: ASG grid removal reduces dose by 26% with reduced image quality. Only with the ASG present can 3rd order branches be visualized; 100% mAs is required for 9 cm pig while 70% mAs is adequate for the larger pigs. 2nd order branches can be visualized with ASG at 17.5% mAs for all three pig sizes. Without the ASG, 50%, 35% and 35% mAs is required for smallest to largest pig. Conclusion: Removing ASG reduces patient dose and image quality. Image quality can be improved with the ASG present while further reducing patient dose if an optimized radiographic technique is used. Rami Nachabe is an employee of Philips Health Care; Keith Strauss is a paid consultant of Philips Health Care.

60 APPLIED LIFE SCIENCES↗

Comparison of slot scanning digital mammography system with full-field digital mammography system

The purpose of this study was to evaluate and compare microcalcification detectability of two commercial full-field digital mammography (DM) systems. The first unit was a flat panel based DM system (FFDM) which employed an anti-scatter grid method to reject scatter, and the second unit was a charge-coupled device-based DM system (SSDM) which used scanning slot imaging geometry to reduce scatter radiation. Both systems have comparable scatter-to-primary ratios. In this study, 125-160 and 200-250 {mu}m calcium carbonate grains were used to simulate microcalcifications and imaged by both DM systems. The calcium carbonate grains were overlapped with a 5-cm-thick 50% adipose/50% glandular simulated breast tissue slab and an anthropomorphic breast phantom (RMI 165, Gammex) for imaging at two different mean glandular dose levels: 0.87 and 1.74 mGy. A reading study was conducted with seven board certified mammographers with images displayed on review workstations. A five-point confidence level rating was used to score each detection task. Receiver operating characteristic (ROC) analysis was performed and the area under the ROC curve (A{sub z}) was used to quantify and compare the performances of these two systems. The results showed that with the simulated breast tissue slab (uniform background), the SSDM system resulted in higher A{sub z}'s than the FFDM system at both MGD levels with the difference statistically significant at 0.87 mGy only. With the anthropomorphic breast phantom (tissue structure background), the SSDM system performed better than the FFDM system at 0.87 mGy but worse at 1.74 mGy. However, the differences were not found to be statistically significant.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Mammographic quantitative image analysis and biologic image composition for breast lesion characterization and classification

Purpose: To investigate whether biologic image composition of mammographic lesions can improve upon existing mammographic quantitative image analysis (QIA) in estimating the probability of malignancy. Methods: The study population consisted of 45 breast lesions imaged with dual-energy mammography prior to breast biopsy with final diagnosis resulting in 10 invasive ductal carcinomas, 5 ductal carcinomain situ, 11 fibroadenomas, and 19 other benign diagnoses. Analysis was threefold: (1) The raw low-energy mammographic images were analyzed with an established in-house QIA method, “QIA alone,” (2) the three-compartment breast (3CB) composition measure—derived from the dual-energy mammography—of water, lipid, and protein thickness were assessed, “3CB alone”, and (3) information from QIA and 3CB was combined, “QIA + 3CB.” Analysis was initiated from radiologist-indicated lesion centers and was otherwise fully automated. Steps of the QIA and 3CB methods were lesion segmentation, characterization, and subsequent classification for malignancy in leave-one-case-out cross-validation. Performance assessment included box plots, Bland–Altman plots, and Receiver Operating Characteristic (ROC) analysis. Results: The area under the ROC curve (AUC) for distinguishing between benign and malignant lesions (invasive and DCIS) was 0.81 (standard error 0.07) for the “QIA alone” method, 0.72 (0.07) for “3CB alone” method, and 0.86 (0.04) for “QIA+3CB” combined. The difference in AUC was 0.043 between “QIA + 3CB” and “QIA alone” but failed to reach statistical significance (95% confidence interval [–0.17 to + 0.26]). Conclusions: In this pilot study analyzing the new 3CB imaging modality, knowledge of the composition of breast lesions and their periphery appeared additive in combination with existing mammographic QIA methods for the distinction between different benign and malignant lesion types.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

TU-CD-BRB-09: Prediction of Chemo-Radiation Outcome for Rectal Cancer Based On Radiomics of Tumor Clinical Characteristics and Multi-Parametric MRI

Purpose: To evaluate the tumor clinical characteristics and quantitative multi-parametric MR imaging features for prediction of response to chemo-radiation treatment (CRT) in locally advanced rectal cancer (LARC). Methods: Forty-three consecutive patients (59.7±6.9 years, from 09/2013 – 06/2014) receiving neoadjuvant CRT followed by surgery were enrolled. All underwent MRI including anatomical T1/T2, Dynamic Contrast Enhanced (DCE)-MRI and Diffusion-Weighted MRI (DWI) prior to the treatment. A total of 151 quantitative features, including morphology/Gray Level Co-occurrence Matrix (GLCM) texture from T1/T2, enhancement kinetics and the voxelized distribution from DCE-MRI, apparent diffusion coefficient (ADC) from DWI, along with clinical information (carcinoembryonic antigen CEA level, TNM staging etc.), were extracted for each patient. Response groups were separated based on down-staging, good response and pathological complete response (pCR) status. Logistic regression analysis (LRA) was used to select the best predictors to classify different groups and the predictive performance were calculated using receiver operating characteristic (ROC) analysis. Results: Individual imaging category or clinical charateristics might yield certain level of power in assessing the response. However, the combined model outperformed than any category alone in prediction. With selected features as Volume, GLCM AutoCorrelation (T2), MaxEnhancementProbability (DCE-MRI), and MeanADC (DWI), the down-staging prediciton accuracy (area under the ROC curve, AUC) could be 0.95, better than individual tumor metrics with AUC from 0.53–0.85. While for the pCR prediction, the best set included CEA (clinical charateristics), Homogeneity (DCE-MRI) and MeanADC (DWI) with an AUC of 0.89, more favorable compared to conventional tumor metrics with an AUC ranging from 0.511–0.79. Conclusion: Through a systematic analysis of multi-parametric MR imaging features, we are able to build models with improved predictive value over conventional imaging or clinical metrics. This is encouraging, suggesting the wealth of imaging radiomics should be further explored to help tailor the treatment into the era of personalized medicine. This work is supported by the National Science Foundation of China (NSFC Grant No. 81201091), National High Technology Research and Development Program of China (863 program, Grant No. 2015AA020917), and Fund Project for Excellent Abroad Scholar Personnel in Science and Technology.

60 APPLIED LIFE SCIENCES↗

SU-C-201-04: Quantification of Perfusion Heterogeneity Based On Texture Analysis for Fully Automatic Detection of Ischemic Deficits From Myocardial Perfusion Imaging

Purpose: Texture-based quantification of image heterogeneity has been a popular topic for imaging studies in recent years. As previous studies mainly focus on oncological applications, we report our recent efforts of applying such techniques on cardiac perfusion imaging. A fully automated procedure has been developed to perform texture analysis for measuring the image heterogeneity. Clinical data were used to evaluate the preliminary performance of such methods. Methods: Myocardial perfusion images of Thallium-201 scans were collected from 293 patients with suspected coronary artery disease. Each subject underwent a Tl-201 scan and a percutaneous coronary intervention (PCI) within three months. The PCI Result was used as the gold standard of coronary ischemia of more than 70% stenosis. Each Tl-201 scan was spatially normalized to an image template for fully automatic segmentation of the LV. The segmented voxel intensities were then carried into the texture analysis with our open-source software Chang Gung Image Texture Analysis toolbox (CGITA). To evaluate the clinical performance of the image heterogeneity for detecting the coronary stenosis, receiver operating characteristic (ROC) analysis was used to compute the overall accuracy, sensitivity and specificity as well as the area under curve (AUC). Those indices were compared to those obtained from the commercially available semi-automatic software QPS. Results: With the fully automatic procedure to quantify heterogeneity from Tl-201 scans, we were able to achieve a good discrimination with good accuracy (74%), sensitivity (73%), specificity (77%) and AUC of 0.82. Such performance is similar to those obtained from the semi-automatic QPS software that gives a sensitivity of 71% and specificity of 77%. Conclusion: Based on fully automatic procedures of data processing, our preliminary data indicate that the image heterogeneity of myocardial perfusion imaging can provide useful information for automatic determination of the myocardial ischemia.

60 APPLIED LIFE SCIENCES↗

SU-F-T-235: Optical Scan Based Collision Avoidance Using Multiple Stereotactic Cameras During Simulation

Purpose: To demonstrate the ability to quickly generate an accurate collision avoidance map using multiple stereotactic cameras during simulation. Methods: Three Kinect stereotactic cameras were placed in the CT simulation room and optically calibrated to the DICOM isocenter. Immediately before scanning, the patient was optically imaged to generate a 3D polygon mesh, which was used to calculate the collision avoidance area using our previously developed framework. The mesh was visually compared to the CT scan body contour to ensure accurate coordinate alignment. To test the accuracy of the collision calculation, the patient and machine were physically maneuvered in the treatment room to calculated collision boundaries. Results: The optical scan and collision calculation took 38.0 seconds and 2.5 seconds to complete respectively. The collision prediction accuracy was determined using a receiver operating curve (ROC) analysis, where the true positive, true negative, false positive and false negative values were 837, 821, 43, and 79 points respectively. The ROC accuracy was 93.1% over the sampled collision space. Conclusion: We have demonstrated a framework which is fast and accurate for predicting collision avoidance for treatment which can be determined during the normal simulation process. Because of the speed, the system could be used to add a layer of safety with a negligible impact on the normal patient simulation experience. This information could be used during treatment planning to explore the feasible geometries when optimizing plans. Research supported by Varian Medical Systems.

60 APPLIED LIFE SCIENCES↗

Zone-based analysis for automated detection of abnormalities in chest radiographs

Purpose: The aim of this study was to develop an automated method for detection of local texture-based and density-based abnormalities in chest radiographs. Methods: The method was based on profile analysis to detect abnormalities in chest radiographs. In the method, one density-based feature, Density Symmetry Index, and two texture-based features, Roughness Maximum Index and Roughness Symmetry Index, were used to detect abnormalities in the lung fields. In each chest radiograph, the lung fields were divided into four zones initially and then the method was applied to each zone separately. For each zone, Density Symmetry Index was obtained from the projection profile of each zone, and Roughness Maximum Index and Roughness Symmetry Index were obtained by measuring the roughness of the horizontal profiles via moving average technique. Linear discriminant analysis was used to classify normal and abnormal cases based on the three indices. The discriminant performance of the method was evaluated using ROC analysis. Results: The method was evaluated on a database of 250 normal and 250 abnormal chest images. In the optimized conditions, the zone-based performance Az of the method for zones 1, 2, 3, and 4 were 0.917, 0.897, 0.892, and 0.814, respectively, and the case-based performance Az of the method was 0.842. Our previous method for detection of gross abnormalities was also evaluated on the same database. The case-based performance of our previous method was 0.689. Conclusions: In comparing the previous method and the new method proposed in this study, there was a great improvement by the new method for detection of local texture-based and density-based abnormalities. The new method combined with the previous one has potential for screening abnormalities in chest radiographs.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Advanced Signal Processing Methods Applied to Digital Mammography

The work reported here is on the extension of the earlier proposal of the same title, August 1994-June 1996. The report for that work is also being submitted. The work reported there forms the foundation for this work from January 1997 to September 1997. After the earlier work was completed there were a few items that needed to be completed prior to submission of a new and more comprehensive proposal for further research. Those tasks have been completed and two new proposals have been submitted, one to NASA, and one to Health & Human Services WS). The main purpose of this extension was to refine some of the techniques that lead to automatic large scale evaluation of full mammograms. Progress on each of the proposed tasks follows. Task 1: A multiresolution segmentation of background from breast has been developed and tested. The method is based on the different noise characteristics of the two different fields. The breast field has more power in the lower octaves and the off-breast field behaves similar to a wideband process, where more power is in the high frequency octaves. After the two fields are separated by lowpass filtering, a region labeling routine is used to find the largest contiguous region, the breast. Task 2: A wavelet expansion that can decompose the image without zero padding has been developed. The method preserves all properties of the power-of-two wavelet transform and does not add appreciably to computation time or storage. This work is essential for analysis of the full mammogram, as opposed to selecting sections from the full mammogram. Task 3: A clustering method has been developed based on a simple counting mechanism. No ROC analysis has been performed (and was not proposed), so we cannot finally evaluate this work without further support. Task 4: Further testing of the filter reveals that different wavelet bases do yield slightly different qualitative results. We cannot provide quantitative conclusions about this for all possible bases without further support. Task 5: Better modeling does indeed make an improvement in the detection output. After the proposal ended, we came up with some new theoretical explanations that helps in understanding when the D4 filter should be better. This work is currently in the review process. Task 6: N/A. This no longer applies in view of Tasks 4-5. Task 7: Comprehensive plans for further work have been completed. These plans are the subject of two proposals, one to NASA and one to HHS. These proposals represent plans for a complete evaluation of the methods for identifying normal mammograms, augmented with significant further theoretical work.

Stauduhar, Richard P.↗

The simulation of 3D mass models in 2D digital mammography and breast tomosynthesis

Purpose: This work proposes a new method of building 3D breast mass models with different morphological shapes and describes the validation of the realism of their appearance after simulation into 2D digital mammograms and breast tomosynthesis images. Methods: Twenty-five contrast enhanced MRI breast lesions were collected and each mass was manually segmented in the three orthogonal views: sagittal, coronal, and transversal. The segmented models were combined, resampled to have isotropic voxel sizes, triangularly meshed, and scaled to different sizes. These masses were referred to as nonspiculated masses and were then used as nuclei onto which spicules were grown with an iterative branching algorithm forming a total of 30 spiculated masses. These 55 mass models were projected into 2D projection images to obtain mammograms after image processing and into tomographic sequences of projection images, which were then reconstructed to form 3D tomosynthesis datasets. The realism of the appearance of these mass models was assessed by five radiologists via receiver operating characteristic (ROC) analysis when compared to 54 real masses. All lesions were also given a breast imaging reporting and data system (BIRADS) score. The data sets of 2D mammography and tomosynthesis were read separately. The Kendall's coefficient of concordance was used for the interrater observer agreement assessment for the BIRADS scores per modality. Further paired analysis, using the Wilcoxon signed rank test, of the BIRADS assessment between 2D and tomosynthesis was separately performed for the real masses and for the simulated masses. Results: The area under the ROC curves, averaged over all observers, was 0.54 (95% confidence interval [0.50, 0.66]) for the 2D study, and 0.67 (95% confidence interval [0.55, 0.79]) for the tomosynthesis study. According to the BIRADS scores, the nonspiculated and the spiculated masses varied in their degrees of malignancy from normal (BIRADS 1) to highly suggestive for malignancy (BIRADS 5) indicating the required variety of shapes and margins of these models. The assessment of the BIRADS scores for all observers indicated good agreement based on Kendall's coefficient for both the 2D and the tomosynthesis evaluations. The paired analysis of the BIRADS scores between 2D and tomosynthesis for each observer revealed consistent behavior for the real and simulated masses. Conclusions: A database of 3D mass models, with variety of shapes and margins, was validated for the realism of their appearance for 2D digital mammography and for breast tomosynthesis. This database is suitable for use in future observer performance studies whether in virtual clinical trials or in patient images with simulated lesions.

60 APPLIED LIFE SCIENCES↗

WE-H-BRC-05: Catastrophic Error Metrics for Radiation Therapy

Purpose: Intuitive evaluation of complex radiotherapy treatments is impractical, while data transfer anomalies create the potential for catastrophic treatment delivery errors. Contrary to prevailing wisdom, logical scrutiny can be applied to patient-specific machine settings. Such tests can be automated, applied at the point of treatment delivery and can be dissociated from prior states of the treatment plan, potentially revealing errors introduced early in the process. Methods: Analytical metrics were formulated for conventional and intensity modulated RT (IMRT) treatments. These were designed to assess consistency between monitor unit settings, wedge values, prescription dose and leaf positioning (IMRT). Institutional metric averages for 218 clinical plans were stratified over multiple anatomical sites. Treatment delivery errors were simulated using a commercial treatment planning system and metric behavior assessed via receiver-operator-characteristic (ROC) analysis. A positive result was returned if the erred plan metric value exceeded a given number of standard deviations, e.g. 2. The finding was declared true positive if the dosimetric impact exceeded 25%. ROC curves were generated over a range of metric standard deviations. Results: Data for the conventional treatment metric indicated standard deviations of 3%, 12%, 11%, 8%, and 5 % for brain, pelvis, abdomen, lung and breast sites, respectively. Optimum error declaration thresholds yielded true positive rates (TPR) between 0.7 and 1, and false positive rates (FPR) between 0 and 0.2. Two proposed IMRT metrics possessed standard deviations of 23% and 37%. The superior metric returned TPR and FPR of 0.7 and 0.2, respectively, when both leaf position and MUs were modelled. Isolation to only leaf position errors yielded TPR and FPR values of 0.9 and 0.1. Conclusion: Logical tests can reveal treatment delivery errors and prevent large, catastrophic errors. Analytical metrics are able to identify errors in monitor units, wedging and leaf positions with favorable sensitivity and specificity. In part by Varian.

60 APPLIED LIFE SCIENCES↗

Comparison of Nodal Risk Formula and MR Lymphography for Predicting Lymph Node Involvement in Prostate Cancer

Purpose: To compare the nodal risk formula (NRF) as a predictor for lymph node (LN) metastasis in patients with prostate cancer with magnetic resonance lymphography (MRL) using Ultrasmall Super-Paramagnetic particles of Iron Oxide (USPIO) and with histology as gold standard. Methods and Materials: Logistic regression analysis was performed with the results of histopathological evaluation of the LN as dependent variable and the nodal risk according to the NRF and the result of MRL as independent input variables. Receiver operating characteristic (ROC) analysis was performed to assess the performance of the models. Results: The analysis included 375 patients. In the single-predictor regression models, the NRF and MRL results were both significantly (p <0.001) predictive of the presence of LN metastasis. In the models with both predictors included, NRF was nonsignificant (p = 0.126), but MRL remained significant (p <0.001). For NRF, sensitivity was 0.79 and specificity was 0.38; for MRL, sensitivity was 0.82 and specificity was 0.93. After a negative MRL result, the probability of LN metastasis is 4% regardless of the NRF result. After a positive MRL, the probability of having LN metastasis is 68%. Conclusions: MRL is a better predictor of the presence of LN metastasis than NRF. Using only the NRF can lead to a significant overtreatment on the pelvic LN by radiation therapy. When the MRL result is available, the NRF is no longer of added value.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Predicting the fidelity of JPEG2000 compressed CT images using DICOM header information

Purpose: To propose multiple logistic regression (MLR) and artificial neural network (ANN) models constructed using digital imaging and communications in medicine (DICOM) header information in predicting the fidelity of Joint Photographic Experts Group (JPEG) 2000 compressed abdomen computed tomography (CT) images. Methods: Our institutional review board approved this study and waived informed patient consent. Using a JPEG2000 algorithm, 360 abdomen CT images were compressed reversibly (n = 48, as negative control) or irreversibly (n = 312) to one of different compression ratios (CRs) ranging from 4:1 to 10:1. Five radiologists independently determined whether the original and compressed images were distinguishable or indistinguishable. The 312 irreversibly compressed images were divided randomly into training (n = 156) and testing (n = 156) sets. The MLR and ANN models were constructed regarding the DICOM header information as independent variables and the pooled radiologists' responses as dependent variable. As independent variables, we selected the CR (DICOM tag number: 0028, 2112), effective tube current-time product (0018, 9332), section thickness (0018, 0050), and field of view (0018, 0090) among the DICOM tags. Using the training set, an optimal subset of independent variables was determined by backward stepwise selection in a four-fold cross-validation scheme. The MLR and ANN models were constructed with the determined independent variables using the training set. The models were then evaluated on the testing set by using receiver-operating-characteristic (ROC) analysis regarding the radiologists' pooled responses as the reference standard and by measuring Spearman rank correlation between the model prediction and the number of radiologists who rated the two images as distinguishable. Results: The CR and section thickness were determined as the optimal independent variables. The areas under the ROC curve for the MLR and ANN predictions were 0.91 (95% CI; 0.86, 0.95) and 0.92 (0.87, 0.96), respectively. The correlation coefficients of the MLR and ANN predictions with the number of radiologists who responded as distinguishable were 0.76 (0.69, 0.82, p < 0.001) and 0.78 (0.71, 0.83, p < 0.001), respectively. Conclusions: The MLR and ANN models constructed using the DICOM header information offer promise in predicting the fidelity of JPEG2000 compressed abdomen CT images.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

TU-C-BRE-01: KEYNOTE PRESENTATION - Emerging Frontiers in IMRT QA

As IMRT treatment processes advance and mature, so must the quality assurance processes being used to validate their delivery. In some respects, treatment delivery advancements (e.g. VMAT) have out-paced QA advancements. The purpose of this session is to describe new processes that are being implemented to bring IMRT QA up-to-date with the treatment delivery advances. It would explore emerging IMRT QA paradigms, including requirements-based IMRT QA which necessitates definition of delivery errors (e.g. patient dose error, leaf positioning error) and development of processes to ensure reliable error detection. Engineeringbased QA approaches, including use of IMRT treatment delivery process trees, fault tree analysis and failure modes effects analysis would be described. Approaches to detect errors such as (1) during treatment delivery validation using exit fluence detectors (e.g. EPIDs); (2) analysis of treatment delivery via use of machine parameter log files; (3) dose recalculation using (3a) treatment planning system; (3b) record-and-verify; or (3c) entrance and exit fluence measurement parameters would be explained. The relative advantages and disadvantages of each method would be discussed. Schemes for error classification and root cause analysis would be described – steps which are essential for future error prevention. For each QA method, testing procedures and results would be presented indicating the types of errors that can be detected, those that cannot be detected, and the reliability of the error detection method (for example determined via ROC analysis). For speakers, we are seeking to engage non-commercially biased experts. Those listed below are a sub-sample of possible qualified individuals.

60 APPLIED LIFE SCIENCES↗

Serum Gamma-Glutamyl-Transferase Independently Predicts Outcome After Transarterial Chemoembolization of Hepatocellular Carcinoma: External Validation

Purpose: An Asian study showed that gamma glutamyl transpeptidase (GGT) can predict survival after transarterial chemoembolization (TACE) of hepatocellular carcinoma (HCC). This study was designed to validate in a European population this biomarker as an independent predictor of outcome after TACE of HCC and to determine a threshold value for clinical use. Methods: In 88 consecutive patients treated by TACE for HCC, the optimal threshold for GGT serum level was determined by a ROC analysis. Endpoints were time-to-treatment failure (TTTF) and overall survival (OS). All multivariate models were internally validated using bootstrapping (90 replications). Results: Median follow-up lasted 373 days, and median overall survival was 748 days. The optimal threshold for GGT was 165 U/L (sensitivity: 89.3%; specificity: 56.7%; area under the ROC curve: 0.7515). Median TTTF was shorter when GGT was {>=}165 U/L (281 days vs. 850 days; P < 0.001). GGT {>=}165 U/L (hazard ratio (HR) = 2.06; P = 0.02), WHO PS of 2 (HR = 5.4; P = 0.002), and tumor size (HR = 1.12; P = 0.014) were independently associated with shorter TTTF. Median OS was shorter when GGT was {>=}165 U/L (508 days vs. not reached; P < 0.001). GGT {>=} 165 U/L (HR = 3.05; P = 0.029), WHO PS of 2 (HR = 12.95; P < 0.001), alfa-fetoprotein (HR = 2.9; P = 0.01), and tumor size (HR = 1.096; P = 0.013) were independently associated with shorter OS. The results were confirmed by bootstrapping. Conclusions: Our results provide in a European population the external validation of GGT as an independent predictor of outcome after TACE of HCC. A serum level of GGT {>=} 165 U/L is independently associated with both shorter TTTF and OS.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

SU-E-J-248: Contributions of Tumor and Stroma Phenotyping in Computer-Aided Diagnosis

Purpose: To gain insight into the role of parenchyma stroma in the characterization of breast tumors by incorporating computerized mammographic parenchyma assessment into breast CADx in the task of distinguishing between malignant and benign lesions. Methods: This study was performed on 182 biopsy-proven breast mass lesions, including 76 benign and 106 malignant lesions. For each full-field digital mammogram (FFDM) case, our quantitative imaging analysis was performed on both the tumor and a region-of-interest (ROI) from the normal contralateral breast. The lesion characterization includes automatic lesion segmentation and feature extraction. Radiographic texture analysis (RTA) was applied on the normal ROIs to assess the mammographic parenchymal patterns of these contralateral normal breasts. Classification performance of both individual computer extracted features and the output from a Bayesian artificial neural network (BANN) were evaluated with a leave-one-lesion-out method using receiver operating characteristic (ROC) analysis with area under the curve (AUC) as the figure of merit. Results: Lesion characterization included computer-extracted phenotypes of spiculation, size, shape, and margin. For parenchymal pattern characterization, five texture features were selected, including power law beta, contrast, and edge gradient. Merging of these computer-selected features using BANN classifiers yielded AUC values of 0.79 (SE=0.03) and 0.67 (SE=0.04) in the task of distinguishing between malignant and benign lesions using only tumor phenotypes and texture features from the contralateral breasts, respectively. Incorporation of tumor phenotypes with parenchyma texture features into the BANN yielded improved classification performance with an AUC value of 0.83 (SE=0.03) in the task of differentiating malignant from benign lesions. Conclusion: Combining computerized tumor and parenchyma phenotyping was found to significantly improve breast cancer diagnostic accuracy highlighting the need to consider both tumor and stroma in decision making. Funding: University of Chicago Dean Bridge Fund, NCI U24-CA143848-05, P50-CA58223 Breast SPORE program, and Breast Cancer Research Foundation. COI: MLG is a stockholder in R2 technology/Hologic and receives royalties from Hologic, GE Medical Systems, MEDIAN Technologies, Riverain Medical, Mitsubishi, and Toshiba. MLG is a cofounder and stockholder in Quantitative Insights.

60 APPLIED LIFE SCIENCES↗

The simulation of 3D microcalcification clusters in 2D digital mammography and breast tomosynthesis

Purpose: This work proposes a new method of building 3D models of microcalcification clusters and describes the validation of their realistic appearance when simulated into 2D digital mammograms and into breast tomosynthesis images. Methods: A micro-CT unit was used to scan 23 breast biopsy specimens of microcalcification clusters with malignant and benign characteristics and their 3D reconstructed datasets were segmented to obtain 3D models of microcalcification clusters. These models were then adjusted for the x-ray spectrum used and for the system resolution and simulated into 2D projection images to obtain mammograms after image processing and into tomographic sequences of projection images, which were then reconstructed to form 3D tomosynthesis datasets. Six radiologists were asked to distinguish between 40 real and 40 simulated clusters of microcalcifications in two separate studies on 2D mammography and tomosynthesis datasets. Receiver operating characteristic (ROC) analysis was used to test the ability of each observer to distinguish between simulated and real microcalcification clusters. The kappa statistic was applied to assess how often the individual simulated and real microcalcification clusters had received similar scores (''agreement'') on their realistic appearance in both modalities. This analysis was performed for all readers and for the real and the simulated group of microcalcification clusters separately. ''Poor'' agreement would reflect radiologists' confusion between simulated and real clusters, i.e., lesions not systematically evaluated in both modalities as either simulated or real, and would therefore be interpreted as a success of the present models. Results: The area under the ROC curve, averaged over the observers, was 0.55 (95% confidence interval [0.44, 0.66]) for the 2D study, and 0.46 (95% confidence interval [0.29, 0.64]) for the tomosynthesis study, indicating no statistically significant difference between real and simulated lesions (p > 0.05). Agreement between allocated lesion scores for 2D mammography and those for the tomosynthesis series was poor. Conclusions: The realistic appearance of the 3D models of microcalcification clusters, whether malignant or benign clusters, was confirmed for 2D digital mammography images and the breast tomosynthesis datasets; this database of clusters is suitable for use in future observer performance studies related to the detectability of microcalcification clusters. Such studies include comparing 2D digital mammography to breast tomosynthesis and comparing different reconstruction algorithms.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

SU-E-J-235: Varian Portal Dosimetry Accuracy at Detecting Simulated Delivery Errors

Purpose: To use receiver operating characteristic (ROC) analysis to quantify the Varian Portal Dosimetry (VPD) application's ability to detect delivery errors in IMRT fields. Methods: EPID and VPD were calibrated/commissioned using vendor-recommended procedures. Five clinical plans comprising 56 modulated fields were analyzed using VPD. Treatment sites were: pelvis, prostate, brain, orbit, and base of tongue. Delivery was on a Varian Trilogy linear accelerator at 6MV using a Millenium120 multi-leaf collimator. Image pairs (VPD-predicted and measured) were exported in dicom format. Each detection test imported an image pair into Matlab, optionally inserted a simulated error (rectangular region with intensity raised or lowered) into the measured image, performed 3%/3mm gamma analysis, and saved the gamma distribution. For a given error, 56 negative tests (without error) were performed, one per 56 image pairs. Also, 560 positive tests (with error) with randomly selected image pairs and randomly selected in-field error location. Images were classified as errored (or error-free) if percent pixels with γ<κ was < (or ≥) τ. (Conventionally, κ=1 and τ=90%.) A ROC curve was generated from the 616 tests by varying τ. For a range of κ and τ, true/false positive/negative rates were calculated. This procedure was repeated for inserted errors of different sizes. VPD was considered to reliably detect an error if images were correctly classified as errored or error-free at least 95% of the time, for some κ+τ combination. Results: 20mm{sup 2} errors with intensity altered by ≥20% could be reliably detected, as could 10mm{sup 2} errors with intensity was altered by ≥50%. Errors with smaller size or intensity change could not be reliably detected. Conclusion: Varian Portal Dosimetry using 3%/3mm gamma analysis is capable of reliably detecting only those fluence errors that exceed the stated sizes. Images containing smaller errors can pass mathematical analysis, though may be detected by visual inspection. This work was not funded by Varian Oncology Systems. Some authors have other work partly funded by Varian Oncology Systems.

60 APPLIED LIFE SCIENCES↗

Correlative feature analysis on FFDM

Identifying the corresponding images of a lesion in different views is an essential step in improving the diagnostic ability of both radiologists and computer-aided diagnosis (CAD) systems. Because of the nonrigidity of the breasts and the 2D projective property of mammograms, this task is not trivial. In this pilot study, we present a computerized framework that differentiates between corresponding images of the same lesion in different views and noncorresponding images, i.e., images of different lesions. A dual-stage segmentation method, which employs an initial radial gradient index (RGI) based segmentation and an active contour model, is applied to extract mass lesions from the surrounding parenchyma. Then various lesion features are automatically extracted from each of the two views of each lesion to quantify the characteristics of density, size, texture and the neighborhood of the lesion, as well as its distance to the nipple. A two-step scheme is employed to estimate the probability that the two lesion images from different mammographic views are of the same physical lesion. In the first step, a correspondence metric for each pairwise feature is estimated by a Bayesian artificial neural network (BANN). Then, these pairwise correspondence metrics are combined using another BANN to yield an overall probability of correspondence. Receiver operating characteristic (ROC) analysis was used to evaluate the performance of the individual features and the selected feature subset in the task of distinguishing corresponding pairs from noncorresponding pairs. Using a FFDM database with 123 corresponding image pairs and 82 noncorresponding pairs, the distance feature yielded an area under the ROC curve (AUC) of 0.81{+-}0.02 with leave-one-out (by physical lesion) evaluation, and the feature metric subset, which included distance, gradient texture, and ROI-based correlation, yielded an AUC of 0.87{+-}0.02. The improvement by using multiple feature metrics was statistically significant compared to single feature performance.

62 RADIOLOGY AND NUCLEAR MEDICINE↗