A Maidment

TH‐D‐201B‐08: An Anthropomorphic Software Breast Phantom for Tomosynthesis Simulation: Power Spectrum Analysis of Phantom Reconstructions

Purpose: To validate an anthropomorphic software breast phantom for simulation studies by analyzing breast structure in projection and reconstructed images.Method and Materials: Twenty computer breast phantoms were generated: ten 450 ml phantoms with 40% glandular tissue compressed to 5‐cm thickness and ten 1500 ml phantoms with 20% glandular tissue compressed to 7.5‐cm thickness. Monoenergetic ray tracing was done to create synthetic projection view images and maximum likelihood expectation maximization (MLEM) with 8 iterations was used to reconstruct the images. Regions of interest (ROIs) were extracted and the squared modulus 2D FFT was taken to obtain periodograms for each ROI. The radial average of each periodogram was taken and the power‐law exponent of this approximated 1D power spectrum is β a description of the amount of structure in the ROI. This was repeated for both projection view images and reconstructed slices and the average β was calculated for each projection view and for each reconstructed slice. Results: For these twenty phantoms we found that in the reconstructed images were lower than the values in the projection view images. For the 450 ml phantoms was measured to be 3.09 (σ = 0.25) in the projection and 2.82 (0.48) in the reconstruction. For the 1500 ml phantoms was measured to be 2.86 (0.43) in the projection and 2.73 (0.66) in the reconstruction. Published data show that decreases from 3.06 to 2.87 in clinical tomosynthesisimages.Conclusion: We found that the changes in between projection views and tomosynthesisreconstructed slices are comparable for the anthropomorphic phantom and clinical breast images. This supports the fact that this phantom can provide realistic breast texture. Conflict of Interest: Research sponsored by Hologic Inc. and Dexela Ltd.

SU‐C‐116‐06: Evaluation of Image Quality in Digital Mammography and Digital Breast Tomosynthesis: Phantom Observer Study From American College of Radiology Imaging Network PA 4006

PURPOSE Determine contributing factors for image quality in 2D and 3D from comparison of phantom scores and image acquisition parameters. METHODS Tomosynthesis 2D and 3D quality assurance images of the ACR Mammography Accreditation Phantom were obtained as part of the American College of Radiology Imaging Network PA-4006 trial. Imaging was performed on Selenia Dimensions (Hologic, Bedford MA) systems at the Hospital of the University of Pennsylvania (n=176) and Einstein Medical Center, Philadelphia, PA (n=70). Four trained observers scored all available images according to manufacturers instructions; image order was randomized by case number and modality for each observer. For tomosynthesis, observers selected the slice level maximizing conspicuity. Paired t-test compared FFDM and DBT cumulative scores by site, multivariate regression related device parameters to image quality, and multi-observer Kappa assessed interobserver agreement. RESULTS Fiber scores were higher for 3D than 2D at site B (p<0.001) with no difference at site A. Speck and mass scores were greater for 2D than 3D at both sites (p<0.001). At site A, regression analysis demonstrated positive correlation of exposure (mAs) with 2D mass, 3D fiber and 3D mass scores; no significant correlations were found for site B. Absolute level of interobserver agreement was highest for specks, followed by fibers and masses. There was substantially lower inter-observer agreement for specks in 3D than 2D; the agreement between 2D and 3D for fibers and masses did not change significantly. CONCLUSION Phantom scores differed significantly between 2D and 3D imaging; fibers were better visualized in 3D, while specks and masses were better visualized in 2D. Observers stated that scoring specks in 3D was more challenging; this is supported by the observation that inter-observer agreement was significantly poorer in 3D than 2D only for specks. There are many factors that influence scoring. While phantom score was partially predicted by exposure, other factors were involved.

SU‐E‐I‐153: A Method for Fast Generation of High Resolution Software Breast Phantoms

Purpose: We have developed a new method for generating anthropomorphic software breast phantoms. The method allows for faster generation of phantoms with small voxel size, and provides better control of simulated anatomical structures. Methods: In this paper we compare the simulation time and quantization artifacts between our previous and the new method. Previously, our breast anatomy simulation was based upon a region growing algorithm. The new method preserves the concept of region growing and is optimized to allow for faster simulation. The new method includes improved thickness control of the simulated skin and Coopers ligaments. For comparison, we generated phantoms simulating a 450 ml breast using both methods. The phantoms were generated with voxel sizes in the range of 25–1000 microns; we recorded the time needed for simulation. The simulation time as function of the voxel size was fitted using a power‐law regression; the regression exponents were compared between the two methods. We also performed a visual comparison of the phantoms generated with different voxel sizes. Results: For the simulated range of voxel sizes the new method yielded a regression exponent of 2.15, compared to 3.89 for the previous method. The new method is faster than the previous method for phantoms with voxels below 200 microns. Generating phantoms with small voxel size is of importance for reducing quantization artifacts in simulated images; commercial mammographysystems have detector element pitch as small as 50 microns. The visual comparison of the phantoms generated at different voxel sizes confirmed the improved quality of simulation with smaller voxels, as reflected by reduced quantization artifacts. Conclusions: Our new breast anatomy simulation is substantially improved in terms of simulation time and the quality of the simulated images. This is especially important for the simulation of phantom images with various available detector resolutions.

WE-E-330D-02: Characterization of Breast Calcifications Using X-Ray Diffraction

Purpose: Coherent scatterimaging has been developed to elucidate the chemical composition of calcifications. Breast calcifications can be divided into two broad categories. Type I are calcium oxalate dehydrate, while Type II are calcium hydroxyapatite. Type II calcifications are known to be associated with carcinoma. It is generally accepted that the exclusive finding of type I calcifications is indicative of benign lesions. Method and Materials: An imaging system has been built that utilizes a molybdenum target x‐ray tube with niobium filtration to isolate the molybdenum Kαcharacteristic radiation. The system is designed to interrogate calcifications in the field with a pencil‐beam of radiation. The transmitted beam is attenuated, and the scattered beam is recorded on an x‐ray image intensifier optically‐coupled to a CCD camera. The system is typically operated at 36 kVp and 25–100 mAs. Results: Reagent grade calcium oxalate and calcium hydroxyapatite were made into blocks of thickness 42–510 mg/cm2. Pinhole sizes varying from 0.3–2.0 mm in diameter were tested to observe the effects of irradiation area on the resolution of the diffraction patterns. All pinhole sizes produced distinctive spectra, but a pinhole size of 0.75 mm appears to be near optimal, as there is sufficient angular resolution and photon fluence to produce distinguishable diffraction patterns. Scatteringmaterials (simulating glandular tissue and fat) were placed upstream and downstream of the calcific material to probe the influence of the surrounding tissue on diffraction. Material thicknesses >1 cm dramatically degraded the measured diffraction patterns. Analysis of diffraction patterns show that calcifications are readily discernable based upon their scatteringcharacteristics. We are able to match these diffraction patterns to calculated theoretical patterns when the later is convolved with a Gaussian‐based filter. Conclusion: We have demonstrated in proof‐of‐principle that we can discern different types of calcifications using coherent scatterimaging.

TU‐FG‐209‐01: A Model of Noise Anisotropy in Amorphous Selenium Digital Detectors

PURPOSE One of the main advantages of amorphous selenium (a-Se) photoconductors is high spatial resolution. It has also been shown that the noise power spectra (NPS) is effectively white (frequency-independent) at various angles of x-ray incidence. One limitation of previous studies on NPS is that the effect of digitization has not been modeled. This work demonstrates that the NPS becomes colored (frequency-dependent) once digitization is taken into account. METHODS A mammography system is modeled by tracing rays between the focal spot and the detector. At each point on the entrance surface of a-Se, the path length of the ray through the thickness of a-Se is determined. The PSF is calculated from first principles based on the existence of ionizations at each point along the path length. To model the effect of digitization, the PSF is convolved with a product of rectangle functions. The NPS is then calculated from the discrete Fourier Transform of the autocovariance of the digitized PSF. RESULTS At the midpoint of the chest wall, the angle of x-ray incidence is normal to the detector. It is demonstrated that the NPS is white at this position, as expected. The angle of incidence becomes oblique with increasing distance from the midpoint of the chest wall. At these positions, the NPS decreases up to the alias frequency, and becomes increasingly colored as the angle of incidence deviates more considerably from the normal. In addition, the NPS is found to be orientation-dependent. Specifically, the NPS is colored along the direction of the incident ray, but is white along the direction perpendicular to the incident ray. CONCLUSION It is demonstrated that oblique x-ray incidence yields cross talk between detector elements. Hence, although the NPS of the a-Se photoconductor is white, the NPS of the digital image is colored. Support for Raymond Joseph Acciavatti was provided by the Postdoctoral Fellowship Grant PDF14302589 from Susan G. Komen. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agency.

SU-C-116-05: Using Maximum Intensity Projection in the Evaluation of Digital Breast Tomosynthesis: A Phantom Observer Study From the American College of Radiology Imaging Network PA 4006 Trial

PURPOSE Characterize the effect of slice level and maximal intensity projection (MIP) slab thickness on image quality in digital breast tomosynthesis (DBT). METHODS Eight ACRIN 4006 PA DBT acquisitions of the ACR Mammography Accreditation Phantom were selected from the Hospital of the University of Pennsylvania (n=4) and Einstein Medical Center, Philadelphia, PA (n=4). An experienced medical physicist scored each phantom by first recording scores at single slice levels in increments of increasing depth to confirm the score for each level. Scoring was next repeated with a 40-mm MIP by identical method. MIP thickness was then decreased in 10-mm increments and scored again. If a difference in score occurred between 10-mm MIP slab thickness differences, objects were rescored at the intermediate 5-mm MIP slab thickness, and so forth. RESULTS Fiber, speck and mass scores are all sensitive to slice level (single slice) or central slice level (MIP). Specks are the most sensitive. Scores can change substantially if slice level is shifted by one slice when viewing single slices or MIP slabs which are less than 10 mm thick. Scores of fibers and masses decrease with increasing MIP slab thickness, while speck scores do not change substantially. Even when all objects are ostensibly in the MIP volume, their location in the volume affects the score. Scores are degraded when the objects are nearer the top or bottom slice of the MIP slab. CONCLUSION In single slice review, slice level substantially affects phantom score, thus care in slice selection is necessary. Increasing MIP slab thickness does not appear to substantially degrade image quality of specks, supporting the clinical practice of using MIP in the review of calcifications. MIP review of phantoms may add value in visualizing artifacts outside the plane of review.

TU-A-301-07: Experimental and Theoretical Validation of Breast Tomosynthesis Reconstructions along Oblique Planes

Purpose: Digital breast tomosynthesis (DBT) is a 3D imaging modality in which tomographic sections of the breast are generated from a limited range of x‐ray projections. Conventional practice is to perform reconstructions possessing pitches within the angular range of the DBT scan, since the Central Slice theorem states that Fourier space is sampled within double‐ napped cones (DNCs) whose opening angle matches that angular range. This work investigates the possibility for both resolution and super‐ resolution (i.e., sub‐pixel resolution) outside the angular range of the DNCs.Methods: Because the image of an object is translated in sub‐pixel detector element increments with each projection, our prior work has demonstrated that DBT is capable of super‐resolution. The previous study assumed a reconstruction plane parallel to the breast support; our current work analyzes super‐resolution in oblique reconstruction planes. Experimentally, a bar pattern phantom was imaged with a commercial DBT system using a goniometry stand, and reconstruction was performed in the oblique plane of the bar patterns. Clinical images of microcalcifications were similarly reconstructed in various oblique planes. Subsequently, an analytical framework for investigating super‐resolution in oblique reconstructions was developed by calculating the filtered backprojection (FBP) reconstruction of a high frequency sine input.Results: Bar pattern reconstructions showed visibility of frequencies both less than and greater than the alias frequency of the detector at pitches well outside the angular range of the DBT scan. Visibility of microcalcifications did not differ considerably using similar oblique reconstruction planes. For analytical proof of super‐resolution in oblique reconstruction planes, we demonstrated that FBP could properly resolve a high frequency sine input whose Fourier transform is non‐vanishing outside the DNCs of frequency space. Conclusions: This work provides a platform for investigating super‐ resolution in oblique reconstruction planes whose pitches are outside the angular range of the DBT scan.

SU‐GG‐I‐137: Study of the Perception of Microcalcification Clusters in Stack‐Mode Digital Breast Tomosynthesis

Purpose: To observe human performance in viewing Digital Breast Tomosynthesis (DBT) images of microcalcification clusters (MCs) in stack‐mode. Method and Materials: A GUI program was created to simulate the presentation of DBT images in stack‐mode. The background consisted of either a uniform gray screen, a gray screen with varying amounts of Poisson noise, or simulated breast images. Each image had an average intensity value of 128, and was displayed for a total of 6 seconds. After the first 4 seconds, a MC at a specified radius from the center was shown for 1 second. The MCs were simulated after those found in the ACR mammography accreditation phantom. The radii used were: 540, 360, and 180 pixels. Contrast differences between the MC and the background were chosen randomly from a set range of about 10–35 gray values. Three test subjects were asked to look at the center of the screen, and to click where they saw a MC in their periphery after the images were displayed. The angular error of the response was calculated, and served as a measure of detection accuracy. A monotonic fit was applied to the data. Results: As contrast of the MC increases, the probability of detection increases, and the angular error decreases. With increasing radius from the center, the threshold contrast for detection increases. Similarly, as the Poisson noise in the image increases from a factor of 1 to 4, the threshold contrast for detection increases. Abnormalities that appear in the right and left sides of the screen have a lower angular error than those that appear in the top and bottom of the screen. Conclusion: These data provide insight into how radiologists perceive MCs while viewing DBT images in stack‐mode. Findings about peripheral vision may suggest a more optimal way of displaying DBT images.

SU‐GG‐I‐124: The Effect of Oblique X‐Ray Incidence on the Design of X‐Ray Detectors for Digital Tomosynthesis

Purpose: Digital breast tomosynthesis (DBT) is an emerging imaging modality in which tomographic sections of the breast are generated from x‐ray projections taken over a limited range of angles. One drawback of DBT is resolution loss in the oblique projection images. The purpose of this work is to extend Swanks formulation of the transfer functions of x‐ray fluorescent screens to oblique x‐ray incidence. This analysis generates a platform for optimizing the design of detectors used in DBT. Method and Materials: We model a non‐structured turbid granular phosphor such as Gd2O2S:Tb which is commonly used in breast imaging and which can reasonably approximate other detector materials. From first principles, the optical transfer function (OTF), noise power spectra (NPS), and detective quantum efficiency (DQE) are derived using the diffusion approximation to the Boltzmann transport equation. We use this analysis to investigate the effect of oblique x‐ray incidence on DQE. The DQE is calculated using the Nishikawa formulation, where the DQE is written as the product of the x‐ray quantum detection efficiency (QDE), the Swank information factor, and the Lubberts factor. Results: As expected, the modulation transfer function(MTF) is shown to decrease with increasing projection angle regardless of frequency. By contrast, the dependence of the DQE on projection angle is found to be frequency dependent. At low frequencies DQE increases with projection angle, but at high frequencies DQE decreases with projection angle. Furthermore, the x‐ray QDE yielding maximal DQE at high frequencies is found to be projection angle dependent, shifting to lower QDE values (thinner phosphors) at higher projection angles. Conclusion: This work develops analytical models of OTF, NPS, and DQE for a turbid granular phosphor irradiated at oblique incidence. The models have applications in the design of optimal DBT detectors.

SU‐FF‐I‐138: Computer‐Aided Image Texture Analysis as An Indicator of Image Quality: A Feasibility Study in FFDM

Purpose:Image quality is often quantified using signal‐to‐noise‐ratio (SNR). We hypothesize that clinical image texture features are also reflective of image quality. To test feasibility, we investigated the relationship between quantitative image texture features and SNR as a function of acquisition parameters for a full‐field digital mammography (FFDM) system. Method and Materials: An anthropomorphic breast phantom (Model 169, Gammex RMI, Madison, WI) was imaged with a Senographe 2000D (General Electric Medical Systems, Chalfont St. Giles, UK). Images were acquired with various target/filter, kV (25 – 32 kV) and mAs combinations. The mAs was varied from 10 to 300% of that required for a standard phototimed exposure yielding an average glandular dose (AGD) of 1.8 mGy. SNR was computed in a 2.5cm2 retroareolar region of interest (ROI) segmented from linear images after flat‐field and offset correction. Signal intensity was computed as the average pixel value in the ROI. Noise was computed as the RMS difference of two images acquired using the same acquisition parameters. Image texture features of skewness, coarseness, contrast, energy, homogeneity, and fractal dimension were computed in the corresponding ROI from the Premium View™ processed FFDM images.Results: Homogeneity decreases continuously as a function of mAs. Skewness, coarseness, and fractal dimension increase, while contrast and energy decrease, until all reach a threshold. The threshold occurs between 40–100% of the phototimed AGD, beyond which no further change is seen. Linear regressions with mAs were strong after logrithmic transformation, having R2 ⩾ 0.84, p < 0.001 for coarseness, energy, and homogeneity. Conclusions: Clinical texture features are indicative of image quality. While homogeneity correlates closely with SNR, other texture features reach a threshold, seeming to indicate optimal dose settings. Further work is underway to fully investigate the potential use of image texture as an indicator of clinically acceptable FFDM image quality.