Digital mammography (DM) and digital breast tomosynthesis, the gold standards for breast cancer screening, requires correct breast positioning to ensure accuracy. Improper positioning can result in missed cancers, or can lead to additional imaging. We propose an automated deep learning (DL) segmentation approach to perform multi-class identification of regions of interest (ROI) commonly used for identification of poor positioning in mediolateral oblique (MLO) breast views. We hypothesize that by leveraging the capabilities of DL through the use of the well-founded U-Net model architecture, multi-class DL-based segmentation approaches can accurately identify air, parenchyma, pectoralis, and nipple locations within MLO images. In this study, we employed model hyperparameter searches to determine optimal model parameters for our proposed DL architecture, including the optimal loss function configuration; our best model achieved an average Sørensen-Dice coefficient of 0.919 ± 0.061 on the held-out test set. We identified high levels of localization performance in the nipple ROI. We believe our proposed segmentation model can be a foundational step in further mammogram analysis, such as for breast positioning and localized image processing tools.
KEYWORDS: Digital breast tomosynthesis, Image quality, Breast, Image restoration, Prototyping, 3D image reconstruction, Sensors, X-rays, X-ray sources, Super resolution
A next generation tomosynthesis (NGT) prototype has been developed to investigate alternative scanning geometries for digital breast tomosynthesis (DBT). Performance of NGT acquisition geometries is evaluated to validate previous phantom experiments. Two custom NGT acquisition geometries were compared to a conventional DBT geometry. Noise power spectra are used to describe features of specimen image reconstructions and compare acquisition geometries. NGT acquisition geometries improve high-frequency performance with superior isotropic super resolution, reduced out-of-plane blurring, and better overall reconstruction quality. NGT combines benefits of narrow- and wide-angle tomosynthesis in a single scan improving high-frequency spatial resolution and out-of-plane blurring, respectively.
Noiseless digital mammograms (DM) are unobtainable in clinical screening environments, limiting the development of deep learning-based (DL) denoising applications. Virtual clinical trials (VCTs) allow the precise simulation of noise levels in DM images for controlled training of DL models. We evaluated a set of DL denoising models, trained using VCT data, that showcases the trade-offs between denoising strength and fine structure preservation. Our results show that metrics, such as peak signal-to-noise ratio (PSNR), are improved with the use of our trained residual convolutional neural network. This quantifiable improvement indicates that our proposed DL methodology can accurately denoise simulated mammograms.
Specimen x-ray imaging provides important information on the margin of surgically excised tissue as well as radiologic and pathologic correlation of the lesion. Similar to breast imaging, where mammograms are digitally processed to enhance readability and lesion conspicuity, specimen images are also processed and enhanced. However, specimen image processing is made challenging by the diversity of specimen containers that are commercially available, compounded by variations in specimen size. In this work, we demonstrate our specimen container and size classification system based on a simple convolutional neural network (CNN), trained to identify the container type. This system allows for automated image processing of the supported container types. A dataset consisting of 1428 HIPAA and IRB-complaint anonymized specimen images were collected. We prepared a simple CNN for image classification with 3 convolutional and 3 fully connected layers, and evaluated the performance based on three comparison metrics. Each network was analyzed in terms of accuracy, multi-class AUC, and via a confusion matrix. The best performing classifier, determined via cross validation, was then used for testing, and evaluated with the same three metrics. The results of training and tuning within cross validation showed that the specimen classes are easily differentiable with this simple convolutional neural network structure. During testing, the network was able to achieve an accuracy of 95.8±4.0%, and an AUC of 0.9763±0.0001.
Simultaneous digital breast tomosynthesis (DBT) and mechanical imaging (MI) offer the potential to combine anatomic information from DBT with functional information from MI. This makes it possible to associate tissue stiffness with specific anatomic structures in the breast, a combination that can reduce false-positive findings by using the MI data to discriminate between ambiguous lesions in DBT. This, in turn, will reduce the frequency of negative biopsies. Simultaneous imaging requires that the MI sensor array be present during DBT acquisition. This introduces artifacts, since the sensor is attenuating. Previously, we demonstrated that the DBT reconstruction could be modified to reduce sensor conspicuity in DBT images. In this paper, we characterize the relative attenuation of the breast and the sensor, to calculate the artifact reduction in DBT reconstruction. We concentrate on pre-processing DBT projections prior to reconstruction. Using commercially available a DBT system, we have confirmed that the sensor array does not completely attenuate the x-rays. This suggests that a pre-processing method based upon flat fielding can be used to reduce artifacts. In a proof-ofconcept study, we performed flat fielding by combining DBT projections of the MI sensor with and without an anthropomorphic breast phantom. Visual evaluation confirmed substantially improved image quality. The artifacts were reduced throughout the image for all sensor elements. Few residual artifacts are noticeable where the phantom thickness decreases. The investigation of additional pre-processing, including beam hardening correction is ongoing. Future work includes quantitative validation, noise stabilization, and method optimization in virtual clinical trials and subsequent patient studies.
Mechanical imaging (MI) uses a pressure sensor array to estimate the stiffness of lesions. Recent clinical studies have suggested that MI combined with digital mammography may reduce false positive findings and negative biopsies by over 30%. Digital breast tomosynthesis (DBT) has been adopted progressively in cancer screening. The tomographic nature of DBT improves lesion visibility by reducing tissue overlap in reconstructed images. For maximum benefit, DBT and MI data should be acquired simultaneously; however, that arrangement produces visible artifacts in DBT images due to the presence of the MI sensor array. We propose a method for reducing artifacts during the DBT image reconstruction. We modified the parameters of a commercial DBT reconstruction engine and investigated the conspicuity of artifacts in the resultant images produced with different sensor orientations. The method was evaluated using a physical anthropomorphic phantom imaged on top of the sensor. Visual assessment showed a reduction of artifacts. In a quantitative test, we calculated the artifact spread function (ASF), and compared the ratio of the mean ASF values between the proposed and conventional reconstruction (termed ASF ratio, RASF). We obtained a mean RASF of 2.74, averaged between two analyzed sensor orientations (45° and 90°). The performance varied with the orientation and the type of sensor structures causing the artifacts. RASF for wide connection lines was larger at 45° than at 90° (5.15 vs. 1.00, respectively), while for metallic contacts RASF was larger at 90° than at 45° (3.31 vs. 2.21, respectively). Future work will include a detailed quantitative assessment, and further method optimization in virtual clinical trials.
KEYWORDS: Modulation transfer functions, Digital breast tomosynthesis, Imaging systems, X-ray sources, X-rays, Image quality, Signal to noise ratio, Mammography, Sensors, Carbon nanotubes
The stationary Digital Breast Tomosynthesis System (s-DBT) has the advantage over the conventional DBT systems as
there is no motion blurring in the projection images associated with the x-ray source motion. We have developed a
prototype s-DBT system by retrofitting a Hologic Selenia Dimensions rotating gantry tomosynthesis system with a
distributed carbon nanotube (CNT) x-ray source array. The linear array consists of 31 x-ray generating focal spots
distributed over a 30 degree angle. Each x-ray beam can be electronically activated allowing the flexibility and easy
implementation of novel tomosynthesis scanning with different scanning parameters and configurations. Here we report
the initial results of investigation on the imaging quality of the s-DBT system and its dependence on the acquisition
parameters including the number of projections views, the total angular span of the projection views, the dose
distribution between different projections, and the total dose. A mammography phantom is used to visually assess image
quality. The modulation transfer function (MTF) of a line wire phantom is used to evaluate the system spatial resolution.
For s-DBT the in-plan system resolution, as measured by the MTF, does not change for different configurations. This is
in contrast to rotating gantry DBT systems, where the MTF degrades for increased angular span due to increased focal
spot blurring associated with the x-ray source motion. The overall image quality factor, a composite measure of the
signal difference to noise ratio (SdNR) for mass detection and the z-axis artifact spread function for microcalcification
detection, is best for the configuration with a large angular span, an intermediate number of projection views, and an
even dose distribution. These results suggest possible directions for further improvement of s-DBT systems for high
quality breast cancer imaging.
Improvement in metrology performance when using a combination of multiple optical channels vs. standard single
optical channel is studied. Two standard applications (gate etch 4x and STI etch 2x) are investigated theoretically
and experimentally. The results show that while individual channels might have increased performance for few
individual parameters each - it is the combination of channels that provides the best overall performance for all
parameters.
Shrinking design rules and reduced process tolerances require tight control of CD linewidth, feature shape, and profile of
the printed geometry. The Holistic Metrology approach consists of utilizing all available information from different
sources like data from other toolsets, multiple optical channels, multiple targets, etc. to optimize metrology recipe and
improve measurement performance. Various in-line critical dimension (CD) metrology toolsets like Scatterometry OCD
(Optical CD), CD-SEM (CD Scanning Electron Microscope) and CD-AFM (CD Atomic Force Microscope) are typically
utilized individually in fabs. Each of these toolsets has its own set of limitations that are intrinsic to specific
measurement technique and algorithm. Here we define "Hybrid Metrology" to be the use of any two or more metrology
toolsets in combination to measure the same dataset. We demonstrate the benefits of the Hybrid Metrology on two test
structures: 22nm node Gate Develop Inspect (DI) & 32nm node FinFET Gate Final Inspect (FI). We will cover
measurement results obtained using typical BKM as well as those obtained by utilizing the Hybrid Metrology approach.
Measurement performance will be compared using standard metrology metrics for example accuracy and precision.
KEYWORDS: Digital breast tomosynthesis, 3D image processing, Tissues, Diagnostics, Sensors, Image processing, Tomography, Mammography, Visualization, 3D image reconstruction
Dynamic Reconstruction and Rendering (DRR) is a fast and flexible tomosynthesis image reconstruction and display
implementation. By leveraging the computational efficiency gains afforded by off-the-shelf GPU hardware,
tomosynthesis reconstruction can be performed on demand at real-time, user-interactive frame rates. Dynamic
multiplanar reconstructions allow the user to adjust reconstruction and display parameters interactively, including axial
sampling, slice location, plane tilt, magnification, and filter selection. Reconstruction on-demand allows tomosynthesis
images to be viewed as true three-dimensional data rather than just a stack of two-dimensional images. The speed and
dynamic rendering capabilities of DRR can improve diagnostic accuracy and lead to more efficient clinical workflows.
A digital breast tomosynthesis (DBT) reconstruction algorithm has been optimized using an anthropomorphic software
breast phantom. The algorithm was optimized in terms of preserving the x-ray attenuation coefficients of the simulated
tissues. The appearance of the reconstructed images is controlled in the algorithm using three input parameters related
to the reconstruction filter. We varied the input parameters to maximally preserve the attenuation information. The
primary interest was to identify and to distinguish between adipose and non-adipose (dense) tissues. To that end, a
software voxel phantom was used which included two distinct attenuation values of simulated breast tissues. The
phantom allows for great flexibility in simulating breasts of various size, glandularity, and internal composition.
Distinguishing between fatty and dense tissues was treated as a binary decision task quantified using ROC analysis. We
defined the reconstruction geometry to enable voxel-to-voxel comparison between the original and reconstructed
volumes. Separate histograms of the reconstructed pixels corresponding to simulated adipose and non-adipose tissues were computed. ROC curves were generated by varying the reconstructed intensity threshold; pixels above the threshold were classified as dense tissue. The input parameter space was searched to maximize the area under the ROC curve. The reconstructed phantom images optimized in this manner better preserve the tissue x-ray attenuation properties; concordant results are seen in clinical images. Use of the software phantom was successful and practical in this task-based optimization, providing ground truth information about the simulated tissues and providing flexibility in defining anatomical properties.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.