Search NASA⌕ Search

SEARCH · Search NASA

Results for “ROC”

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

Optimizing Earthquake Nowcasting With Machine Learning: The Role of Strain Hardening in the Earthquake Cycle

Abstract Nowcasting is a term originating from economics, finance, and meteorology. It refers to the process of determining the uncertain state of the economy, markets or the weather at the current time by indirect means. In this paper, we describe a simple two‐parameter data analysis that reveals hidden order in otherwise seemingly chaotic earthquake seismicity. One of these parameters relates to a mechanism of seismic quiescence arising from the physics of strain‐hardening of the crust prior to major events. We observe an earthquake cycle associated with major earthquakes in California, similar to what has long been postulated. An estimate of the earthquake hazard revealed by this state variable time series can be optimized by the use of machine learning in the form of the Receiver Operating Characteristic skill score. The ROC skill is used here as a loss function in a supervised learning mode. Our analysis is conducted in the region of 5° × 5° in latitude‐longitude centered on Los Angeles, a region which we used in previous papers to build similar time series using more involved methods (Rundle & Donnellan, 2020, https://doi.org/10.1029/2020EA001097 ; Rundle, Donnellan et al., 2021, https://doi.org/10.1029/2021EA001757 ; Rundle, Stein et al., 2021, https://doi.org/10.1088/1361-6633/abf893 ). Here we show that not only does the state variable time series have forecast skill, the associated spatial probability densities have skill as well. In addition, use of the standard ROC and Precision (PPV) metrics allow probabilities of current earthquake hazard to be defined in a simple, straightforward, and rigorous way.

58 GEOSCIENCES↗

An image classification deep-learning algorithm for shrapnel detection from ultrasound images

Ultrasound imaging is essential for non-invasively diagnosing injuries where advanced diagnostics may not be possible. However, image interpretation remains a challenge as proper expertise may not be available. In response, artificial intelligence algorithms are being investigated to automate image analysis and diagnosis. Here, we highlight an image classification convolutional neural network for detecting shrapnel in ultrasound images. As an initial application, different shrapnel types and sizes were embedded first in a tissue mimicking phantom and then in swine thigh tissue. The algorithm architecture was optimized stepwise by minimizing validation loss and maximizing F1 score. The final algorithm design trained on tissue phantom image sets had an F1 score of 0.95 and an area under the ROC curve of 0.95. It maintained higher than a 90% accuracy for each of 8 shrapnel types. When trained only on swine image sets, the optimized algorithm format had even higher metrics: F1 and area under the ROC curve of 0.99. Overall, the algorithm developed resulted in strong classification accuracy for both the tissue phantom and animal tissue. This framework can be applied to other trauma relevant imaging applications such as internal bleeding to further simplify trauma medicine when resources and image interpretation are scarce.

59 BASIC BIOLOGICAL SCIENCES↗

End-to-End Jet Classification of Boosted Top Quarks with CMS Open Data

We describe a novel application of the end-to-end deep learning technique to the task of discriminating top quark-initiated jets from those originating from the hadronization of a light quark or a gluon. The end-to-end deep learning technique combines deep learning algorithms and low-level detector representation of the high-energy collision event. In this study, we use lowlevel detector information from the simulated CMS Open Data samples to construct the top jet classifiers. To optimize classifier performance we progressively add low-level information from the CMS tracking detector, including pixel detector reconstructed hits and impact parameters, and demonstrate the value of additional tracking information even when no new spatial structures are added. Relying only on calorimeter energy deposits and reconstructed pixel detector hits, the end-to-end classifier achieves a ROC-AUC score of 0.975±0.002 for the task of classifying boosted top quark jets. After adding derived track quantities, the classifier ROC-AUC score increases to 0.9824±0.0013, serving as the first performance benchmark for these CMS Open Data samples.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Superiority of compensatory reserve measurement compared with the Shock index for early and accurate detection of reduced central blood volume status

Shock index (SI) equals the ratio of heart rate (HR) to systolic blood pressure (SBP) with clinical evidence that it is more sensitive for trauma patient status assessment and prediction of outcome compared with either HR or SBP alone. We used lower body negative pressure (LBNP) as a human model of central hypovolemia and compensatory reserve measurement (CRM) validated for accurate tracking of reduced central blood volume to test the hypotheses that SI: (1) presents a late signal of central blood volume status; (2) displays poor sensitivity and specificity for predicting the onset of hemodynamic decompensation; and (3) cannot identify individuals at greatest risk for the onset of circulatory shock. We measured HR, SBP, and CRM in 172 human subjects (19–55 years) during progressive LBNP designed to determine tolerance to central hypovolemia as a model of hemorrhage. Subjects were subsequently divided into those with high tolerance (HT) (n = 118) and low tolerance (LT) (n = 54) based on completion of 60 mm Hg LBNP. The time course relationship between SI and CRM was determined and receiver operating characteristic (ROC) area under the curve (AUC) was calculated for sensitivity and specificity of CRM and SI to predict hemodynamic decompensation using clinically defined thresholds of 40% for CRM and 0.9 for SI. The time and level of LBNP required to reach a SI = 0.9 (~60 mm Hg LBNP) was significantly greater (p < 0.001) compared with CRM that reached 40% at ~40 mm Hg LBNP. Shock index did not differ between HT and LT subjects at 45 mm Hg LBNP levels. ROC AUC for CRM was 0.95 (95% CI = 0.94–0.97) compared with 0.91 (0.89–0.94) for SI (p = 0.0002). Despite high sensitivity and specificity, SI delays time to detect reductions in central blood volume with failure to distinguish individuals with varying tolerances to central hypovolemia.

60 APPLIED LIFE SCIENCES↗

Pleural homocysteine for malignant pleural effusion: A prospective and double‐blind diagnostic test accuracy study

Abstract Objective To assess the accuracy of pleural fluid homocysteine for discriminating malignant pleural effusion (MPE) and benign pleural effusion (BPE). Methods A total of 194 patients from two cohorts (Hohhot and Changshu) with undiagnosed pleural effusion were prospectively enrolled. Their pleural homocysteine was measured, and its diagnostic accuracy and net benefit for MPE were analyzed by receiver operating characteristic (ROC) curve analysis and decision curve analysis, respectively. Results In the Hohhot cohort ( n = 136) and the Changshu cohort ( n = 58), MPE patients had significantly higher homocysteine levels than BPE patients. The areas under the ROC curves of homocysteine for the diagnosis of MPE were 0.61 ( p = 0.027) and 0.59 ( p = 0.247), respectively. The decision curves of homocysteine were close to the reference line in both the Hohhot cohort and the Changshu cohort. Conclusion The diagnostic accuracy of pleural fluid homocysteine for MPE was low.

Cao, Xi‐Shan↗

Optimal band selection for target detection with a LWIR multispectral imager

Multispectral imaging can offer many benefits in cost, complexity, resolution, size, weight, and power, relative to hyperspectral imaging. When designing a multispectral system, spectral bandpasses can be selected using optimization algorithms configured to maximally separate target detection scores between target and background regions. A hyperspectral image (HSI) can serve as the source of data from which band groupings can be tested for optimality. The output of an adaptive cosine estimator target detection algorithm is used in an objective function. Three optimization algorithms are compared: particle swarm, dual annealing, and differential evolution. A global optimum is also found using a brute force approach on the Livermore Computing Syrah supercomputer. Three materials are investigated: calcite, gypsum, and limestone. This is done for 3-, 4-, and 5-band systems. The data originate from a longwave infrared HSI of a material display board. The optimization algorithms were run 30 times for every scenario. Performance statistics (maximum, minimum, mean, standard deviation, and median) based on the separation values are given. Additional characterization was performed using receiver operator characteristic (ROC) curves and the area under the ROC curve. While good performance was obtained for the three optimization algorithms, the dual annealing algorithm produced the highest and most consistent detection separation scores on average.

47 OTHER INSTRUMENTATION↗

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↗

Color Doppler quantitative measures to predict outcome of biopsies in prostate cancer

Purpose: The aim was to correlate the color Doppler flow activity pre- and postradiotherapy, using transrectal color Doppler ultrasonography (CDUS) and the 2 year positive biopsy rate after radiotherapy in patients with prostate cancer. Methods and materials: Analysis was carried out in 69 out of 160 patients who had undergone treatment with 3D-conformal radiotherapy (3D-CRT) to prostate and seminal vesicles. Patients were randomized to receive 80 Gy in 40 fractions in 8 weeks (arm A) and 62 Gy in 20 fractions in 5 weeks, 4 fractions per week (arm B). Color Doppler flow activity (CDFA) was evaluated calculating the vascularization index (VI), defined as the ratio between the colored and total pixels in the whole and peripheral prostate, delineated by a radiation oncologist on CDUS images, using EcoVasc a home-made software. The difference between the 2 year post- and pre-3D-CRT maximum VI (VI{sub max}), named {Delta}VI{sub max}, was calculated in the whole and peripheral prostate for each patient. Then, {Delta}VI{sub max} and the detected 2 year biopsy outcome were analyzed using the receiver operating characteristics (ROC) technique. Results: The VI{sub max} increased or decreased in patients with positive or negative biopsies, respectively, compared to the value before RT in both arms. The area under the ROC curve for {Delta}VI{sub max} in the whole and peripheral prostate is equal to 0.790 and 0.884, respectively. Conclusion: The {Delta}VI{sub max} index, comparing CDFA at 2 years compared to that before RT, allows the 2 year postradiotherapy positive biopsy rate to be predicted.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

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↗

Temporal subtraction in chest radiography: Mutual information as a measure of image quality

Purpose: Temporal subtraction is used to detect the interval change in chest radiographs and aid radiologists in patient diagnosis. This method registers two temporally different images by geometrically warping the lung region, or ''lung mask,'' of a previous radiographic image to align with the current image. The gray levels of every pixel in the current image are subtracted from the gray levels of the corresponding pixels in the warped previous image to form a temporal subtraction image. While temporal subtraction images effectively enhance areas of pathologic change, misregistration of the images can mislead radiologists by obscuring the interval change or by creating artifacts that mimic change. The purpose of this study was to investigate the utility of mutual information computed between two registered radiographic chest images as a metric for distinguishing between clinically acceptable and clinically unacceptable temporal subtraction images.Methods: A radiologist subjectively rated the image quality of 138 temporal subtraction images using a 1 (poor) to 5 (excellent) scale. To objectively assess the registration accuracy depicted in the temporal subtraction images, which is the main factor that affects the quality of these images, mutual information was computed on the two constituent registered images prior to their subtraction to generate a temporal subtraction image. Mutual information measures the joint entropy of the current image and the warped previous image, yielding a higher value when the gray levels of spatially matched pixels in each image are consistent. Mutual information values were correlated with the radiologist's subjective ratings. To improve this correlation, mutual information was computed from a spatially limited lung mask, which was cropped from the bottom by 10%-60%. Additionally, the number of gray-level values used in the joint entropy histogram was varied. The ability of mutual information to predict the clinical acceptability of a temporal subtraction image was evaluated through receiver operating characteristic (ROC) analysis. Results: The mean correlation coefficient obtained between mutual information computed on constituent images and the subjective rating of temporal subtraction image quality was 0.785. ROC analysis yielded an average A{sub z} value of 0.852 for the ability of mutual information to distinguish between temporal subtraction images of clinically acceptable and clinically unacceptable quality. Conclusions: The results of this study establish a relationship between mutual information and temporal subtraction registration accuracy and demonstrate the ability of mutual information to objectively indicate the presence of misregistration artifacts.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

The diagnostic capability of x-ray scattering parameters for the characterization of breast cancer

Purpose: The evaluation of the diagnostic capability of easy to measure x-ray scattering profile characterization parameters for the detection of breast cancer in excised samples. The selected parameters are the full width at half maximum (FWHM) and area under the x-ray scattering profile of breast tissue in addition to the ratio of scattering intensities (I{sub 2}/I{sub 1}%) at 1.6 nm{sup -1} to that at 1.1 nm{sup -1} (corresponding to scattering from soft and adipose tissues, respectively). Methods: A histopathologist is asked to classify 36 excised breast tissue samples into healthy or malignant. A conventional x-ray diffractometer is used to acquire the scattering profiles of the investigated samples. The values of three profile characterization parameters are calculated and the diagnostic capability of each is evaluated by determining the optimal cutoffs of scatter diagrams, calculating the diagnostic indices, and plotting the receiver operating characteristic (ROC) curves. Results: At the calculated optimal cutoff for each of the examined parameters, the sensitivity ranged from 78% (for area under curve) up to 94% (for FWHM), the specificity ranged from 94%[for I{sub 2}/I{sub 1}% and area under curve] up to 100% (for FWHM), and the diagnostic accuracy ranged from 86% (for area under curve) up to 97% (for FWHM). The area under the ROC curves is greater than 0.95 for all of the investigated parameters, reflecting a highly accurate diagnostic performance. Conclusions: The discussed tests offered a means to quantitatively evaluate the performance of the suggested breast tissue x-ray scattering characterization parameters. The performance results are promising, indicating that the evaluated parameters would be considered a tool for fast, on spot probing of breast cancer in excised tissue samples.

60 APPLIED LIFE SCIENCES↗

Fractal analysis for assessing the level of modulation of IMRT fields

Purpose: To investigate the potential of three fractal dimension (FD) analysis methods (i.e., the variation, power spectrum, and variogram methods) as metrics for quantifying the degree of modulation in planned intensity modulated radiation therapy (IMRT) treatment fields, and compare the most suitable FD method to the number of monitor units (MUs), the average leaf gap, and the 2D modulation index (2D MI) for assessing modulation. Methods: The authors implemented, validated, and compared the variation, power spectrum, and variogram methods for computing the FD. Validation of the methods was done using mathematical fractional Brownian surfaces of known FD that ranged in size from 128 x 128 to 512 x 512. The authors used a test set consisting of seven head and neck carcinoma plans (50 prescribed treatment fields) to choose an FD cut-point that ensures no false positives (100% specificity) in distinguishing between moderate and high degrees of field modulation. The degree of field modulation was controlled by adjusting the fluence smoothing parameters in the Eclipse treatment planning system (Varian Medical Systems, Palo Alto, CA). The moderate modulation fields were representative of the degree of modulation used clinically at the authors' institution. The authors performed IMRT quality assurance (QA) on the 50 test fields using the MapCHECK device. The FD cut-point was applied to a validation set consisting of four head and neck plans (28 fields). The area under the curve (AUC) from receiver operating characteristic (ROC) analysis was used to compare the ability of FD, number of MUs, average leaf gap, and the 2D MI for distinguishing between the moderate and high modulation fields. Results: The authors found the variogram FD method to be the most suitable for assessing the modulation complexity of IMRT fields for head and neck carcinomas. Pass rates as measured by the gamma criterion for the MapCHECK IMRT field measurements were higher for the moderately modulated fields, and a gamma criterion with 1 mm distance-to-agreement and 1% dose difference showed a clear separation between the 94% pass rates of the moderate and high modulation groups. From the ROC analysis of the test set, the authors found the AUC of the variogram FD, number of MUs, average leaf gap, and 2D MI methods to be 0.99 (almost perfect), 0.91 (excellent), 0.91 (excellent), and 0.92 (excellent), respectively. A cut-point of FD > 2.25 correctly identified 92.8% of the high modulation fields and 100% of the moderately modulated fields in the validation set, satisfying the condition of no false positives. Conclusions: Of the three FD methods investigated, the variogram method is the most accurate and precise metric for identifying high modulation treatment fields. It is also more accurate and precise than the number of MUs, the average leaf gap, and the 2D MI. Although MapCHECK IMRT QA does a reasonable job at identifying high modulation fields, the variogram FD method provides one with the opportunity to quantitatively and accurately assess modulation and adjust overly modulated fields at the treatment planning stage before they are sent to the treatment machine for QA or patient treatment.

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 ↗

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↗

A scaling transformation for classifier output based on likelihood ratio: Applications to a CAD workstation for diagnosis of breast cancer

Purpose: The authors developed scaling methods that monotonically transform the output of one classifier to the ''scale'' of another. Such transformations affect the distribution of classifier output while leaving the ROC curve unchanged. In particular, they investigated transformations between radiologists and computer classifiers, with the goal of addressing the problem of comparing and interpreting case-specific values of output from two classifiers. Methods: Using both simulated and radiologists' rating data of breast imaging cases, the authors investigated a likelihood-ratio-scaling transformation, based on ''matching'' classifier likelihood ratios. For comparison, three other scaling transformations were investigated that were based on matching classifier true positive fraction, false positive fraction, or cumulative distribution function, respectively. The authors explored modifying the computer output to reflect the scale of the radiologist, as well as modifying the radiologist's ratings to reflect the scale of the computer. They also evaluated how dataset size affects the transformations. Results: When ROC curves of two classifiers differed substantially, the four transformations were found to be quite different. The likelihood-ratio scaling transformation was found to vary widely from radiologist to radiologist. Similar results were found for the other transformations. Our simulations explored the effect of database sizes on the accuracy of the estimation of our scaling transformations. Conclusions: The likelihood-ratio-scaling transformation that the authors have developed and evaluated was shown to be capable of transforming computer and radiologist outputs to a common scale reliably, thereby allowing the comparison of the computer and radiologist outputs on the basis of a clinically relevant statistic.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Effect of image quality on calcification detection in digital mammography

Purpose: This study aims to investigate if microcalcification detection varies significantly when mammographic images are acquired using different image qualities, including: different detectors, dose levels, and different image processing algorithms. An additional aim was to determine how the standard European method of measuring image quality using threshold gold thickness measured with a CDMAM phantom and the associated limits in current EU guidelines relate to calcification detection. Methods: One hundred and sixty two normal breast images were acquired on an amorphous selenium direct digital (DR) system. Microcalcification clusters extracted from magnified images of slices of mastectomies were electronically inserted into half of the images. The calcification clusters had a subtle appearance. All images were adjusted using a validated mathematical method to simulate the appearance of images from a computed radiography (CR) imaging system at the same dose, from both systems at half this dose, and from the DR system at quarter this dose. The original 162 images were processed with both Hologic and Agfa (Musica-2) image processing. All other image qualities were processed with Agfa (Musica-2) image processing only. Seven experienced observers marked and rated any identified suspicious regions. Free response operating characteristic (FROC) and ROC analyses were performed on the data. The lesion sensitivity at a nonlesion localization fraction (NLF) of 0.1 was also calculated. Images of the CDMAM mammographic test phantom were acquired using the automatic setting on the DR system. These images were modified to the additional image qualities used in the observer study. The images were analyzed using automated software. In order to assess the relationship between threshold gold thickness and calcification detection a power law was fitted to the data. Results: There was a significant reduction in calcification detection using CR compared with DR: the alternative FROC (AFROC) area decreased from 0.84 to 0.63 and the ROC area decreased from 0.91 to 0.79 (p < 0.0001). This corresponded to a 30% drop in lesion sensitivity at a NLF equal to 0.1. Detection was also sensitive to the dose used. There was no significant difference in detection between the two image processing algorithms used (p > 0.05). It was additionally found that lower threshold gold thickness from CDMAM analysis implied better cluster detection. The measured threshold gold thickness passed the acceptable limit set in the EU standards for all image qualities except half dose CR. However, calcification detection varied significantly between image qualities. This suggests that the current EU guidelines may need revising. Conclusions: Microcalcification detection was found to be sensitive to detector and dose used. Standard measurements of image quality were a good predictor of microcalcification cluster detection.

61 RADIATION PROTECTION AND DOSIMETRY↗

Residual analysis of the water resonance signal in breast lesions imaged with high spectral and spatial resolution (HiSS) MRI: A pilot study

Purpose: High spectral and spatial resolution magnetic resonance imaging (HiSS MRI) yields information on the local environment of suspicious lesions. Previous work has demonstrated the advantages of HiSS (complete fat-suppression, improved image contrast, no required contrast agent, etc.), leading to initial investigations of water resonance lineshape for the purpose of breast lesion classification. The purpose of this study is to investigate a quantitative imaging biomarker, which characterizes non-Lorentzian components of the water resonance in HiSS MRI datasets, for computer-aided diagnosis (CADx). Methods: The inhomogeneous broadening and non-Lorentzian or “off-peak” components seen in the water resonance of proton spectra of breast HiSS images are analyzed by subtracting a Lorentzian fit from the water peak spectra and evaluating the difference spectrum or “residual.” The maxima of these residuals (referred to hereafter as “off-peak components”) tend to be larger in magnitude in malignant lesions, indicating increased broadening in malignant lesions. The authors considered only those voxels with the highest magnitude off-peak components in each lesion, with the number of selected voxels dependent on lesion size. Our voxel-based method compared the magnitudes and frequencies of off-peak components of all voxels from all lesions in a database that included 15 malignant and 8 benign lesions (yielding ∼3900 voxels) based on the lesions’ biopsy-confirmed diagnosis. Lesion classification was accomplished by comparing the average off-peak component magnitudes and frequencies in malignant and benign lesions. The area under the ROC curve (AUC) was used as a figure of merit for both the voxel-based and lesion-based methods. Results: In the voxel-based task of distinguishing voxels from malignant and benign lesions, off-peak magnitude yielded an AUC of 0.88 (95% confidence interval [0.84, 0.91]). In the lesion-based task of distinguishing malignant and benign lesions, average off-peak magnitude yielded an AUC 0.83 (95% confidence interval [0.61, 0.98]). Conclusions: These promising AUC values suggest that analysis of the water-resonance in each HiSS image voxel using “residual analysis” could have high diagnostic utility and could be used to enhance current CADx methods and allow detection of breast cancer without the need to inject contrast agents.

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↗