Detectability Of Low-contrast Lesions Research Articles

Influence of the adaptive iterative dose reduction 3D algorithm on the detectability of low-contrast lesions and radiation dose repeatability in abdominal computed tomography: a phantom study.

The purpose of the study is to evaluate the influence of the adaptive iterative dose reduction (AIDR 3D) algorithm on the detectability of low-contrast focal liver lesions (FLLs) and the radiation dose repeatability of automatic tube current modulation (ATCM) in abdominal CT scans using anthropomorphic phantoms. Three different sizes of anthropomorphic phantoms, each with 4 low-contrast FLLs, were scanned on a 320-channel CT scanner using the ATCM technique and AIDR 3D, at different radiation doses: full-dose, half-dose, and quarter-dose. Scans were repeated three times and reconstructed with filtered back projection (FBP) and AIDR 3D. Radiation dose repeatability was assessed using the intraclass correlation coefficient (ICC). Image noise, quality, and lesion conspicuity were assessed by four reviewers and the number of invisible FLLs was compared among different radiation doses and reconstruction methods. ICCs of radiation dose among the three CT scans were excellent in all phantoms (0.99). Image noise, quality, and lesion conspicuity in the half-dose group were comparable with full-dose FBP after applying AIDR 3D in all phantoms. In small phantoms, the half-dose group reconstructed with AIDR 3D showed similar sensitivity in visualizing low-contrast FLLs compared to full-dose FBP (P = 0.77-0.84). In medium and large phantoms, AIDR 3D reduced the number of missing low-contrast FLLs [3.1% (9/288), 11.5% (33/288), respectively], compared to FBP [10.4% (30/288), 21.9% (63/288), respectively] in the full-dose group. By applying AIDR 3D, half-dose CT scans may be achievable in small-sized patients without hampering diagnostic performance, while it may improve diagnostic performance in medium- and large-sized patients without increasing the radiation dose.

Model-based iterative reconstruction for detection of subtle hypoattenuation in early cerebral infarction: a phantom study.

Model-based iterative reconstruction (MBIR) was recently shown to enable dose reduction in computed tomography (CT). The detectability of low-contrast lesions was assessed on CT images reconstructed with MBIR compared with the conventional filtered back-projection (FBP) method. A phantom simulating brain gray matter containing small lesions mimicking early cerebral infarctions was scanned at tube currents of 50, 100, 200, and 400 mA. Images were reconstructed by use of both methods. Round regions were cropped from the reconstructed images, half with a lesion, the other half without. Eight radiologists reviewed the images and scored the certainty of lesion detection on a 5-point scale. Overall performance was analyzed by use of a receiver operating characteristic curve. For the tube currents investigated, the analysis showed that the mean areas under the curves for the reviewers were 0.65, 0.70, 0.82, and 0.83 for FBP and 0.70, 0.76, 0.78, and 0.90 for MBIR. For each current, there was no significant difference between the areas under the curves for the different reconstruction methods (p = 0.32, 0.24, 0.49, and 0.17). For the small, low-contrast lesions in the phantom model used in this study, no significant difference between detectability was observed for MBIR and FBP.

Virtual single-source computed tomography using dual-source acquisition: a new technique for the dose-neutral intraindividual comparison of different scan protocols.

The objective of this study was to compare the image quality of a standard single-source (SSS) computed tomography (CT) with that of a virtual single-source CT (VSS-CT) data set reconstructed from 2 raw data sets obtained by dual-source CT acquisition in abdominal CT to establish a radiation dose-neutral approach for the intraindividual comparison of 3 acquisition protocols at different radiation dose levels (RDLs). An abdominal phantom representing an 80-kg male was imaged using dual-source CT (SOMATOM Definition; Siemens Healthcare) at 3 RDLs with 120 kV(p) and different tube currents (low, standard, and high milliampere-second protocol). For each RDL, raw data were obtained once in single-source mode using x-ray tube A only and 5 times in dual-source mode using different ratios for tube current of x-ray tubes A and B (same total radiation dose; A/B: 90%/10%, 80%/20%, 70%/30%, 60%/40%, 50%/50%). For each RDL, SSS-CT and 5 virtual single-source image data sets (VSS-CT50 - 90) were reconstructed. To compare SSS-CT and VSS-CT data sets, image quality was assessed in terms of high- and low-contrast performance by calculating the modulation transfer function, image noise, noise power spectrum, and, for low contrast lesion detectability, the modified multiscale structural similarity index (MS-SSIM*). A maximum decrease of Δ = 5% of image quality compared with SSS-CT was defined as acceptable, and a noninferiority analysis with Δ was performed. For modulation transfer function, noninferiority was observed for all VSS-CT data sets and RDL (P < 0.05). Image noise demonstrated an acceptable increase (<3.2%, P < 0.05) for each RDL and noise power spectrum showed only minor differences in the midfrequency range. The MS-SSIM* index demonstrated for the high RDL protocol a minor decrease for VSS-CT data sets (<2%, P < 0.05). For the standard and low RDL, the relative differences of the MS-SSIM* index increased and were only in 1 case above Δ (standard RDL, mean VSS-CT80 5.1%, P > 0.05). The image quality obtained by virtual and SSS reconstruction using equivalent total radiation exposure to the patient showed only negligible differences in image quality. Therefore, this technique might allow an intraindividual comparison of full and reduced radiation dose protocols within 1 image acquisition step by simply splitting the radiation dose between the 2 x-ray tubes of a dual-source CT.

Assessment of an Advanced Image-Based Technique to Calculate Virtual Monoenergetic Computed Tomographic Images From a Dual-Energy Examination to Improve Contrast-To-Noise Ratio in Examinations Using Iodinated Contrast Media

Following the trend of low-radiation dose computed tomographic (CT) imaging, concerns regarding the detectability of low-contrast lesions have been growing. The goal of this research was to evaluate whether a new image-based algorithm (Mono+) for virtual monoenergetic imaging with a dual-energy CT scanner can improve the contrast-to-noise ratio (CNR) and conspicuity of these low-contrast objects when using iodinated contrast media. Four circular phantoms of different diameter (10-40 cm) with an iodine insert at the center were scanned at a fixed radiation dose with different single- (80, 100, 120 kV) and dual-energy protocols (80/140 kV, 80/140 Sn kV, 100/140 Sn kV) using a dual-source CT system. In addition, an anthropomorphic abdominal phantom with different low-contrast lesions was scanned with the settings previously mentioned but also at only a half and a quarter of the initial dose. Dual-energy data were processed, and virtual monoenergetic images (range, 40-190 keV) were generated. Beside the established technique, a newly developed prototype algorithm to calculate monoenergetic images (Mono+) was used. To avoid noise increase at lower calculated energies, which is a known drawback of virtual monoenergetic images at low kilo electron-volt, a regional spatial frequency-based recombination of the high signal at lower energies and the superior noise properties at medium energies is performed to optimize CNR in case of Mono+ images. The CNR and low-contrast detectability were evaluated. For all phantom sizes, the Mono+ technique provided increasing iodine CNR with decreasing kilo electron-volt, with the optimum CNR obtained at the lowest energy level of 40 keV. For all investigated phantom sizes, CNR of Mono+ images at low kilo electron-volt was superior to the CNR in single-energy images at an equivalent radiation dose and even higher than the CNR obtained with 80-kV protocols. In case of the anthropomorphic phantom, low-contrast detectability in monoenergetic images was, for all settings, similar to the circular phantoms, best for the voltage combination 80/140 Sn kV, irrespective of the dose level. For all dual-energy voltage combinations, the Mono+ algorithm led to superior results compared with single-energy imaging. With regard to optimized iodine CNR, it is more efficient to perform dual-energy scans and compute virtual monoenergetic images at 40 keV using the Mono+ technique than to perform low kilovolt scans. Given the improved CNR, the Mono+ algorithm could be very useful in improving both detection and differential diagnosis of abdominal lesions, specifically low-contrast lesions, as well as in other anatomical regions where improved iodine CNR is beneficial.

Impact of Incremental Increase in CT Image Noise on Detection of Low-contrast Hypodense Liver Lesions

To determine the impact of incremental increases in computed tomography (CT) image noise on detection of low-contrast hypodense liver lesions. We studied 50 CT examinations acquired at image noise index (NI) of 15 and hypodense liver lesions and 50 examinations with no lesions. Validation of a noise addition tool to be used in the evaluation of the CT examinations was performed with a liver phantom. Using this tool, three 100-image sets were assembled: an NI of 17.4 (simulating 75% of the original patient radiation dose), 21.2 (simulating 50% dose), and 29.7 (simulating 25%). Three readers scored certainty of lesion presence using a five-point Likert scale. For original images (NI 15) plus images with NI of 17.4 and 21.2, sensitivity was >90% threshold (range, 95%-98%). For images with NI of 29.7, sensitivity was just below the threshold (89%). Reader Az values for receiver operating characteristic curves were good for original, NI 17.4, and NI 21.2 images (0.976, 0.973, and 0.96, respectively). For NI of 29.7, the Az decreased to 0.913. Detection sensitivity was <90% for both lesion size < 10 mm (85%) and lesion-to-liver contrast <60 Hounsfield units (85%) only at NI 29.7. For low-contrast lesion detection in liver CT, image noise can be increased up to NI 21.2 (a 50% patient radiation dose reduction) without substantial reduction in sensitivity.

Methods for clinical evaluation of noise reduction techniques in abdominopelvic CT.

Most noise reduction methods involve nonlinear processes, and objective evaluation of image quality can be challenging, since image noise cannot be fully characterized on the sole basis of the noise level at computed tomography (CT). Noise spatial correlation (or noise texture) is closely related to the detection and characterization of low-contrast objects and may be quantified by analyzing the noise power spectrum. High-contrast spatial resolution can be measured using the modulation transfer function and section sensitivity profile and is generally unaffected by noise reduction. Detectability of low-contrast lesions can be evaluated subjectively at varying dose levels using phantoms containing low-contrast objects. Clinical applications with inherent high-contrast abnormalities (eg, CT for renal calculi, CT enterography) permit larger dose reductions with denoising techniques. In low-contrast tasks such as detection of metastases in solid organs, dose reduction is substantially more limited by loss of lesion conspicuity due to loss of low-contrast spatial resolution and coarsening of noise texture. Existing noise reduction strategies for dose reduction have a substantial impact on lowering the radiation dose at CT. To preserve the diagnostic benefit of CT examination, thoughtful utilization of these strategies must be based on the inherent lesion-to-background contrast and the anatomy of interest. The authors provide an overview of existing noise reduction strategies for low-dose abdominopelvic CT, including analytic reconstruction, image and projection space denoising, and iterative reconstruction; review qualitative and quantitative tools for evaluating these strategies; and discuss the strengths and limitations of individual noise reduction methods.

Low-contrast lesion detection enhancement using pulse compression technique

Pulse compression methods greatly improve the quality of medical images. In comparison with standard broadband pulse techniques, these methods enhance the contrast-to-noise ratio and increase the probing depth without any perceptible loss of spatial resolution. The Golay compression technique is analyzed here in the context of ultrasonic computed tomography, first on a one-dimensional target and second, on a very low-contrast phantom probed using a half-ring array tomograph. The imaging performances were assessed based on both the point spread function properties and the image contrast-to-noise ratio. The improvement obtained in the image contrast-to-noise ratio (up to 40%) depends, however, on the number of coherently associated diffraction projections. Beyond a certain number, few advantages were observed. Advances in ultrasound computed tomography suggest that pulse compression methods should provide a useful means of optimizing the trade-off between the image quality and the probing sampling density. It could also be used to accelerate the reconstruction process during the examination of patients. The results of this study also suggest that when it is proposed to search for very low contrast lesions (such as diffuse lobar carcinomas in breast cancer), once the number of projections has been set for a given tomographic set-up, the image contrast, i.e., the probability of detection, can been enhanced by using low-power, high-energy Golay sequences

Despeckling of medical ultrasound images by wavelet threshold optimisation

Speckle noise is the primary factor that limits the contrast resolution of diagnostic ultrasound images, thereby limiting the detection of small low-contrast lesions and making the ultrasound images generally difficult for a nonspecialist to interpret. Therefore, speckle is more often considered as a dominant source of noise in ultrasound imaging and should be filtered out by a robust despeckling technique. In this paper, we compare two despeckling techniques, based on wavelet thresholding. Bayes shrinkage and genetic-algorithm-based wavelet denoising despeckling technique are implemented. Wavelet functions were studied in detail and the two algorithms were implemented using the Haar wavelet as a wavelet function. The application of this new filter on ultrasound images has shown a superior performance over the state-of-the-art wavelet-based denoising methods in terms of visual quality and structural similarity index map.

SU‐E‐I‐52: Improvement in CT Contrast‐To‐Noise Ratio in Low Contrast Spherical Simulated Liver Lesions Using Iterative Reconstruction

Purpose: Contrast‐to‐Noise ratio (CNR) is one measure to assess image quality and lesion detectability for CT exams. The purpose of this study is to calculate the change in CNR between standard filtered back‐projection (FBP) and sinogram‐affirmed iterative reconstruction (SAFIRE) by using an anthropomorphic phantom with spherical simulated liver lesions. Methods: An anthropomorphic phantom with embedded spherical liver lesions was custom‐designed to assess low contrast lesion detection at different CT doses and reconstruction techniques. This phantom consists of 36 spherical lesions with 3 different sizes (5mm, 10mm, and 15mm). Each size has three different contrast levels. All spherical lesions were embedded non‐uniformly to minimize the memory bias for a concurrent human observer study. Phantom was scanned with a MDCT (Siemens Definition Flash) at 4 dose levels and images were reconstructed with both FBP and SAFIRE using strength setting 3. CNR was computed as the ratio of the difference of the mean CT number at the center of the lesion and the liver background to the standard deviation of the CT numbers from the liver background. CNR was evaluated for the each lesion size at the center slice that intersected the middle of the lesion to minimize the error from the partial volume effect. Results: Averaged across all lesion sizes and the contrast levels (vs. the liver background), the CNR improvement from FBP to SAFIRE (strength 3) was 34% ±10% (mean ± standard deviation) at 50mAs,(p&lt;0.0001), 33%±11% at 100mAs (p&lt;0.0002), 36% ± 9% at 150mAs (p&lt;0.0001) and 36% ± 7% at 200mAs (p&lt;0.0003). Noise reduction from SAFIRE contributed almost entirely to the CNR improvement. Conclusion: Iterative reconstruction (SAFIRE) improved CNR of spherical low‐contrast liver lesions, but almost entirely due to noise reduction, with little improvement in lesion contrast. The overall CNR improvement was uniform for each dose level tested. Dr. Herts and Dr. Baker are funded by Siemens Healthcare for research in the field of abdominal CT imaging.

SU‐E‐I‐37: Lesion Circularity as a Potential Indicator for Detectability in Low Dose Abdomen CT

Purpose: Low‐contrast lesion detection is critical for abdomen CT. However, there is no standardized, objective measure of the detectability of low‐contrast objects. The purpose of this study is to investigate a new measure, the ability of a circular lesion to maintain its shape or circularity, as a potential indicator of lesion detectability. Methods: An anthropomorphic phantom with embedded spherical liver lesions was custom‐designed to assess the effectiveness of iterative reconstruction for improving the low contrast lesion detection. This phantom consists of 36 spherical lesions with 3 different sizes (5mm, 10mm, and 15mm). Each size has three different contrast levels. All spherical lesions were embedded non‐uniformly to minimize the memory bias for a concurrent human observer study. Phantom was scanned with a MDCT (Siemens Definition Flash) at multiple dose levels and images were reconstructed with both filtered back‐projection (FBP) and Iterative Reconstruction (SAFIRE). An adaptive Wiener filter was applied during the lesion detection. The circularity was calculated based on the ratio of the area to (perimeter)2 of each detected lesion. Results: The circularity was normalized to 1 when the detected lesion has a perfectly circular shape and the boundary is connected. At 50mAs, the circularity for a 10mm spherical lesion with 18HU contrast was 0.618 with FBP vs. 0.650 with SAFIRE; While at the same mAs, for a 15mm spherical lesion with 12HU contrast, the circularity was 0.666 with FBP vs. 0.763 for SAFIRE . At 200mAs, the circularity for a 10mm spherical lesion was 0.811with FBP vs. 0.870 with SAFIRE ; for a 15mm lesion, the circularity was 0.818 with FBP vs. 0.863 with SAFIRE. Conclusion: Circularity decreased with decreasing lesion size, contrast differential and dose. Iterative reconstruction techniques improved upon lesion circularity when compared to FBP at the same dose level, at similar lesion size and density differences. Dr. Herts and Dr. Baker are currently funded by Siemens Healthcare for research in low dose abdominal CT.

Influence of Sinogram Affirmed Iterative Reconstruction of CT Data on Image Noise Characteristics and Low-Contrast Detectability: An Objective Approach

ObjectivesTo utilize a novel objective approach combining a software phantom and an image quality metric to systematically evaluate the influence of sinogram affirmed iterative reconstruction (SAFIRE) of multidetector computed tomography (MDCT) data on image noise characteristics and low-contrast detectability (LCD).Materials and MethodsA low-contrast and a high-contrast phantom were examined on a 128-slice scanner at different dose levels. The datasets were reconstructed using filtered back projection (FBP) and SAFIRE and virtual low-contrast lesions (-20HU) were inserted. LCD was evaluated using the multiscale structural similarity index (MS-SIM*). Image noise texture and spatial resolution were objectively evaluated.ResultsThe use of SAFIRE led to an improvement of LCD for all dose levels and lesions sizes. The relative improvement of LCD was inversely related to the dose level, declining from 208%(±37%), 259%(±30%) and 309%(±35%) at 25mAs to 106%(±6%), 119%(±9%) and 123%(±8%) at 200mAs for SAFIRE filter strengths of 1, 3 and 5 (p<0.05). SAFIRE reached at least the LCD of FBP at a relative dose of 50%. There was no statistically significant difference in spatial resolution. The use of SAFIRE led to coarser image noise granularity.ConclusionA novel objective approach combining a software phantom and the MS-SSIM* image quality metric was used to analyze the detectability of virtual low-contrast lesions against the background of image noise as created using SAFIRE in comparison to filtered back-projection. We found, that image noise characteristics using SAFIRE at 50% dose were comparable to the use of FBP at 100% dose with respect to lesion detectability. The unfamiliar imaging appearance of iteratively reconstructed datasets may in part be explained by a different, coarser noise characteristic as demonstrated by a granulometric analysis.

A Combined Dual-Tree Complex Wavelet (DT-CWT) and Bivariate Shrinkage for Ultrasound Medical Images Despeckling

paper, an efficient DT-CWT based method for medical ultrasound images despeckling is presented. The ultrasound images are often deteriorated by speckle noise, this noise is a random granular texture that obscures anatomy in ultrasound images and degrades the detectability of low-contrast lesions. Speckle noise occurrence is often undesirable, since it affects the tasks of human interpretation and diagnosis.Different from many other schemes with wavelet transform are used on one side in which the studies have dealt more with the standard DWT case. However, the Discrete Wavelet Transform (DWT) has some disadvantages that undermine its application in image processing.In this study we investigated a performances complex wavelet transform (DT-CWT) combined with Bivariate Shrinkage.The proposed method was tested on a noisy ultrasound medical image, and the restored images show a great effectiveness of DT-CWT method compared to the classical DWT.

TU-E-217BCD-10: Dose Reduction in Digital Breast Tomosynthesis with the Dose Reduction Prior Image Constrained Compressed Sensing (DR-PICCS) Algorithm.

To reduce image noise and radiation dose in Digital Breast Tomosynthesis (DBT) reconstructions. A retrospective study was performed on clinical data sets acquired at a normal dose with Hologic Selenia Dimensions DBT systems. The Prior Image Constrained Compressed Sensing (PICCS) algorithm was used to reduce image noise. In addition, a prospective study was performed on an American College of Radiology breast phantom at various dose levels and the PICCS algorithm was used to reconstruct images at the corresponding radiation dose levels. The reconstructed images were inspected visually, and the noise levels in various regions of interest were quantitatively measured and compared between images. In the case of the clinical data, the PICCS reconstructions showed dramatic noise reduction (over 35%) with no loss of diagnostically important features such as calcifications or low contrast lesions; visibility of low contrast lesions was improved with PICCS. Dose reduction of 28% was possible with the phantom data, and the low dose PICCS reconstructions of phantom data show improved low contrast lesion detectability and lower noise. The work indicates potential dose savings in digital breast tomosynthesis. The diagnostic quality of the phantom reconstructions at 28% reduced dose was equivalent to or better than those acquired at full dose. The noise suppression in the clinical data sets improved visibility of low contrast lesions without sacrificing important diagnostic features. Support for this project was provided by a grant from Hologic Inc.

Automatic Detection of Lung and Liver Lesions in 3-D Positron Emission Tomography Images: A Pilot Study

Positron emission tomography (PET) using fluorine-18 deoxyglucose (18F-FDG) has become an increasingly recommended tool in clinical whole-body oncology imaging for the detection, diagnosis, and follow-up of many cancers. One way to improve the diagnostic utility of PET oncology imaging is to assist physicians facing difficult cases of residual or low-contrast lesions. This study aimed at evaluating different schemes of computer-aided detection (CADe) systems for the guided detection and localization of small and low-contrast lesions in PET. These systems are based on two supervised classifiers, linear discriminant analysis (LDA) and the nonlinear support vector machine (SVM). The image feature sets that serve as input data consisted of the coefficients of an undecimated wavelet transform. An optimization study was conducted to select the best combination of parameters for both the SVM and the LDA. Different false-positive reduction (FPR) methods were evaluated to reduce the number of false-positive detections per image (FPI). This includes the removal of small detected clusters and the combination of the LDA and SVM detection maps. The different CAD schemes were trained and evaluated based on a simulated whole-body PET image database containing 250 abnormal cases with 1230 lesions and 250 normal cases with no lesion. The detection performance was measured on a separate series of 25 testing images with 131 lesions. The combination of the LDA and SVM score maps was shown to produce very encouraging detection performance for both the lung lesions, with 91% sensitivity and 18 FPIs, and the liver lesions, with 94% sensitivity and 10 FPIs. Comparison with human performance indicated that the different CAD schemes significantly outperformed human detection sensitivities, especially regarding the low-contrast lesions.

Image Noise and Liver Lesion Detection With MDCT: A Phantom Study

The purpose of this study was to determine the upper limit of noise for detection of small low-contrast lesions in a liver phantom. A CT liver phantom containing 21 low-contrast, low-attenuation, circular simulated lesions ranging in size from 2.4 to 10 mm was scanned 23 times at different tube current ranges (varying noise index) on a 64-MDCT scanner with automatic tube current modulation. The attenuation of the simulated lesions was 20 HU less than that of the liver-equivalent background. Three radiologists independently reviewed the resultant CT images, which contained either a low-contrast lesion or no lesion and scored certainty of lesion detection using a 4-point Likert scale. Overall performance was evaluated by sensitivity analysis with receiver operator curve and area under the curve (A(z)) computation for ranges of noise index. The reviewers achieved 100% sensitivity with a noise index of 15 or less for lesions measuring 6.3-10.0 mm (A(z) = 0.96). Increasing noise index to the 17-21 range resulted in a minor decrease in sensitivity and overall performance (sensitivity, 92.3%; A(z) = 0.93). A further increase in noise index to the 23-27 range resulted in a moderate decrease in sensitivity (sensitivity, 81.4%; A(z) = 0.77). Beyond the noise index 23-27 range, sensitivity dropped markedly from 81.4% to 39%. Agreement between the three readers in assessing the image sets was moderate. For detection of small low-contrast lesions in the liver phantom model used in this study, the upper limit of noise index may be in the 15-21 range for sensitivity greater than 90%.

Confidence intervals for performance assessment of linear observers

This work seeks to develop exact confidence interval estimators for figures of merit that describe the performance of linear observers, and to demonstrate how these estimators can be used in the context of x-ray computed tomography (CT). The figures of merit are the receiver operating characteristic (ROC) curve and associated summary measures, such as the area under the ROC curve. Linear computerized observers are valuable for optimization of parameters associated with image reconstruction algorithms and data acquisition geometries. They provide a means to perform assessment of image quality with metrics that account not only for shift-variant resolution and nonstationary noise but that are also task-based. We suppose that a linear observer with fixed template has been defined and focus on the problem of assessing the performance of this observer for the task of deciding if an unknown lesion is present at a specific location. We introduce a point estimator for the observer signal-to-noise ratio (SNR) and identify its sampling distribution. Then, we show that exact confidence intervals can be constructed from this distribution. The sampling distribution of our SNR estimator is identified under the following hypotheses: (i) the observer ratings are normally distributed for each class of images and (ii) the variance of the observer ratings is the same for each class of images. These assumptions are, for example, appropriate in CT for ratings produced by linear observers applied to low-contrast lesion detection tasks. Unlike existing approaches to the estimation of ROC confidence intervals, the new confidence intervals presented here have exactly known coverage probabilities when our data assumptions are satisfied. Furthermore, they are applicable to the most commonly used ROC summary measures, and they may be easily computed (a computer routine is supplied along with this article on the Medical Physics Website). The utility of our exact interval estimators is demonstrated through an image quality evaluation example using real x-ray CT images. Also, strong robustness is shown to potential deviations from the assumption that the ratings for the two classes of images have equal variance. Another aspect of our interval estimators is the fact that we can calculate their mean length exactly for fixed parameter values, which enables precise investigations of sampling effects. We demonstrate this aspect by exploring the potential reduction in statistical variability that can be gained by using additional images from one class, if such images are readily available. We find that when additional images from one class are used for an ROC study, the mean AUC confidence interval length for our estimator can decrease by as much as 35%. We have shown that exact confidence intervals can be constructed for ROC curves and for ROC summary measures associated with fixed linear computerized observers applied to binary discrimination tasks at a known location. Although our intervals only apply under specific conditions, we believe that they form a valuable tool for the important problem of optimizing parameters associated with image reconstruction algorithms and data acquisition geometries, particularly in x-ray CT.

Informatics in Radiology: Sliding-Thin-Slab Averaging for Improved Depiction of Low-Contrast Lesions with Radiation Dose Savings at Thin-Section CT

Current multidetector computed tomography (CT) scanners allow volumetric data acquisition with thin-section collimations and overlapping section reconstructions. The resultant nearly isotropic data sets help minimize partial-volume averaging effects and are ideal for two- and three-dimensional postprocessing and software-assisted lesion detection and quantification. However, the section thickness, image noise, and radiation dose are closely related, and when one parameter must be altered to suit the clinical setting, the others may be affected. When the clinical purpose demands both high spatial resolution and low image noise (eg, for the detection of hypoattenuating lesions in organs such as the kidneys and liver), the necessary trade-off--an increase in the radiation dose to the patient--may be unacceptable. The application of a sliding-thin-slab averaging algorithm during image postprocessing and review helps overcome this limitation by reconstructing thicker sections with lower noise levels from thin-section data obtained with dose-saving protocols. In principle, a high noise level is acceptable in the initial reconstruction of the CT volume data set. During image review at the workstation, the section thickness can be interactively increased to minimize image noise and improve lesion detectability. The combination of thin-section scanning with thick-section display allows routine volumetric imaging without a general increase in radiation dose or a reduction in the detectability of low-contrast lesions. Supplemental material available at http://radiographics.rsna.org/lookup/suppl/doi:10.1148/rg.302096007/-/DC1.

Respiratory Gated PET Derived in a Fully Automated Manner From Raw PET Data

Respiratory motion in PET degrades image quality and limits detectability of small or low-contrast lesions. Although image quality can be improved using respiratory-gating, this adds to the complexity and expense of acquiring PET data. We aimed to develop a data-driven method, based on individual voxel signal fluctuations, for accomplishing electronic respiratory gating of clinical PET data, requiring no additional hardware or end-user input. We tested our methods using both simulated PET scans and actual human PET acquisitions. For the simulations, our methods correctly identified the start frame of each respiratory cycle defined for the phantom. Resultant gated images demonstrated improved effective resolution and increased measured uptake for lesions located in the thorax. For human PET data, we were able to recover respiratory phase information with a high signal-to-noise ratio. We report here a method to achieve fully automated voxel-based respiratory gating of PET images, without the need for gating hardware or additional user input, capable of improving effective resolution and increasing lesion detectability.

SU‐GG‐I‐82: Contrast Detail Curves in Head CT Examinations

Purpose: To generate CT based contrast detail curve using a representative head CT image from each of five clinical head CT examinations. We measured detection of 4 to 12.6 mm sized lesions using a 2 Alternate Forced Choice (2‐AFC) experimental. In each experiment the observer identifies the lesion location in one of two regions of interest; after 128 such observations, it is possible to determine the contrast (I) corresponding to a 92% accuracy (I92%) of lesion detection. For each lesion size, we performed four repeat detection experiments, and the order of the 100 AFC experiments (i.e., 5 patients × 5 lesion sizes × 4 repeats) was randomized to eliminate experimental bias (i.e., learning curve and/or observer fatigue). Average values of I92% were obtained for one observer at each lesion size. Contrast detail curves were generated with log(I92%) plotted versus log(lesion size). There were major differences in contrast detail curves for the five patients depending on the presence of anatomical structure in the image. The coefficient of variation in lesion detection, averaged over the five images and five lesion sizes, was found to be 33%. The slope of the contrast detail curve averaged over five patients was −0.14; this slope is a factor of seven lower than the value of −1.0 predicted by the Rose model of lesion detection. Results from our pilot study of low contrast lesion detection in clinical head CT suggest that that anatomical structure is of greater importance than quantum mottle.

Full-field digital mammography compared with screen-film mammography in the detection of breast cancer: rays of light through DMIST or more fog?

Breast cancer is the most common malignancy and second most common cause of cancer death in women in the US [1]. To date, screen-film mammography (SFM) has been considered the gold standard for breast cancer screening and detection [1]. Over 70% of women in the US over the age of 40 have had a mammogram within the past 2 years [2]. Prior research has investigated the efficacy of SFM at reducing mortality from breast cancer [3–17]. While the interpretation of these results has been hotly debated [18–25], the consensus in recent systematic reviews and practice guidelines is that SFM reduces breast cancer mortality among women who are 50–75 years of age, and may reduce mortality among women aged 40 years and older [26–30]. A series of complex models supports the hypothesis that SFM has contributed to a reduction in breast cancer mortality in the US [31]. Although film mammography is the established diagnostic tool, it has a number of limitations. Breast cancer screening is less effective in younger women than in older women, most likely because younger women have a higher proportion of mammographically dense breast tissue [32– 35]. Up to a third of breast cancers are missed, even though as many as 20 women are called back due to false-positive results for every one cancer identified; this increases patient anxiety and leads to additional breast imaging, radiation exposure, and biopsies. Furthermore, with SFM, image acquisition, storage, and comparison can be cumbersome. As a result, much attention has been devoted to developing improved radiographic techniques for breast cancer screening and evaluation. Full-field digital mammography (FFDM) addresses some of the limitations seen with SFM. It has a wider range of contrast resolution, which holds particular promise for improving the detection of low-contrast lesions in radiographically dense breasts. FFDM allows for the separation of image acquisition, presentation, and storage, thus enabling these tasks to be independently optimized. The implementation of computer-aided detection methods is potentially simplified with FFDM, as there is no longer the separate step of digitizing film mammograms [36]. Computer-aided enhancement of images at computer workstations may also improve the accuracy of mammographic interpretation [37]. FFDM also has the potential to improve workflow by allowing for electronic transmission, storage, and retrieval of the images. Use of this modality may facilitate the interpretation of mammograms, as the images can be obtained locally, but sent to a central location for interpretation by experts at centers that specialize in mammographic interpretation. Digital storage and retrieval may increase the likelihood that comparison images from prior mammograms would be available to aid the radiologist in the interpretation of a new mammogram. Digital images can either be printed on film for review (hard copy) or read on computer monitors (soft copy). Digital image acquisition may improve the signal-to-noise ratio of X-ray detection over a wider range of intensities, as compared to that of film acquisition over the same range [38–40]. There is also the possibility of lower radiation exposure than that required for SFM. In spite of these potential advantages, adoption of a digital approach to mammography has been slow, in part due to the high spatial resolution required by J. A. Tice (&) M. D. Feldman Division of General Internal Medicine, Department of Medicine, University of California, 1701 Divisadero Street, Suite 554, San Francisco, CA 94143-1732, USA e-mail: jtice@medicine.ucsf.edu