Classification of compressed breast shapes for the design of equalization filters in x-ray mammography

Automated pectoral muscle identification on MLO-view mammograms: Comparison of deep neural network to conventional computer vision

OBJECTIVES

The aim of this study was to develop a fully automated deep learning approach for identification of the pectoral muscle on mediolateral oblique (MLO) view mammograms and evaluate its performance in comparison to our previously developed texture-field orientation (TFO) method using conventional image feature analysis. Pectoral muscle segmentation is an important step for automated image analyses such as breast density or parenchymal pattern classification, lesion detection, and multiview correlation.

MATERIALS AND METHODS

Institutional Review Board (IRB) approval was obtained before data collection. A dataset of 729 MLO-view mammograms including 637 digitized film mammograms (DFM) and 92 digital mammograms (DM) from our previous study was used for the training and validation of our deep convolutional neural network (DCNN) segmentation method. In addition, we collected an independent set of 203 DMs from 131 patients for testing. The film mammograms were digitized at a pixel size of 50 μm × 50 μm with a Lumiscan digitizer. All DMs were acquired with GE systems at a pixel size of 100 μm × 100 μm. An experienced MQSA radiologist manually drew the pectoral muscle boundary on each mammogram as the reference standard. We trained the DCNN to estimate a probability map of the pectoral muscle region on mammograms. The DCNN consisted of a contracting path to capture multiresolution image context and a symmetric expanding path for prediction of the pectoral muscle region. Three DCNN structures were compared for automated identification of pectoral muscles. Tenfold cross-validation was used in training of the DCNNs. After training, we applied the ten trained models during cross-validation to the independent DM test set. The predicted pectoral muscle region of each test DM was obtained as the mean probability map by averaging the ensemble of probability maps from the ten models. The DCNN-segmented pectoral muscle was evaluated by three performance measures relative to the reference standard: (a) the percent overlap area (POA) of the pectoral muscle regions, (b) the Hausdorff distance (Hdist), and (c) the average Euclidean distance (AvgDist) between the boundaries. The results were compared to those obtained with the TFO method, used as our baseline. A two-tailed paired t test was performed to examine the significance in the differences between the DCNN and the baseline.

RESULTS

In the ten test partitions of the cross-validation set, the DCNN achieved a mean POA of 96.5 ± 2.9%, a mean Hdist of 2.26 ± 1.31 mm, and a mean AvgDist of 0.78 ± 0.58 mm, while the corresponding measures by the baseline method were 94.2 ± 4.8%, 3.69 ± 2.48 mm, and 1.30 ± 1.22 mm, respectively. For the independent DM test set, the DCNN achieved a mean POA of 93.7% ± 6.9%, a mean Hdist of 3.80 ± 3.21 mm, and a mean AvgDist of 1.49 ± 1.62 mm comparing to 86.9% ± 16.0%, 7.18 ± 14.22 mm, and 3.98 ± 14.13 mm, respectively, by the baseline method.

CONCLUSION

In comparison to the TFO method, DCNN significantly improved the accuracy of pectoral muscle identification on mammograms (P < 0.05).

Removal of pectoral muscle based on topographic map and shape-shifting silhouette

BACKGROUND

In digital mammography, finding accurate breast profile segmentation of women's mammogram is considered a challenging task. The existence of the pectoral muscle may mislead the diagnosis of cancer due to its high-level similarity to breast body. In addition, some other challenges due to manifestation of the breast body pectoral muscle in the mammogram data include inaccurate estimation of the density level and assessment of the cancer cell. The discrete differentiation operator has been proven to eliminate the pectoral muscle before the analysis processing.

METHODS

We propose a novel approach to remove the pectoral muscle in terms of the mediolateral-oblique observation of a mammogram using a discrete differentiation operator. This is used to detect the edges boundaries and to approximate the gradient value of the intensity function. Further refinement is achieved using a convex hull technique. This method is implemented on dataset provided by MIAS and 20 contrast enhanced digital mammographic images.

RESULTS

To assess the performance of the proposed method, visual inspections by radiologist as well as calculation based on well-known metrics are observed. For calculation of performance metrics, the given pixels in pectoral muscle region of the input scans are calculated as ground truth.

CONCLUSIONS

Our approach tolerates an extensive variety of the pectoral muscle geometries with minimum risk of bias in breast profile than existing techniques.

Identification and segmentation of obscure pectoral muscle in mediolateral oblique mammograms

OBJECTIVE

X-ray mammography is a widely used and reliable method for detecting pre-symptomatic breast cancer. One of the difficulties in automatically computerized mammogram analysis is the presence of pectoral muscles in mediolateral oblique mammograms because the pectoral muscle does not belong to the scope of the breast. The objective of this study is to identify the boundary of obscure pectoral muscle in mediolateral oblique mammograms.

METHODS

Two tentative boundary curves are individually created to be the potential boundaries. To find the first tentative boundary, this study finds local extrema, prunes weak extrema and then determines an appropriate threshold for identifying the brighter tissue, whose edge is considered the first tentative boundary. The second tentative boundary is found by partitioning the breast into several regions, where each local threshold is tuned based on the local intensity. Subsequently, both of these tentative boundaries are used as the reference to create a refined boundary by Hough transform. Then, the refined boundary is partitioned into quadrilateral regions, in which the edge of this boundary is detected. Finally, these reliable edge points are collected to generate the genuine boundary by curve fitting.

RESULTS

The proposed method achieves the least mean square error 4.88 ± 2.47 (mean ± standard deviation) and the least misclassification error rate (MER) with 0.00466 ± 0.00191 in terms of MER.

CONCLUSION

The experimental results indicate that this method performs best and stably in boundary identification of the pectoral muscle.

ADVANCES IN KNOWLEDGE

The proposed method can identify the boundary from obscure pectoral muscle, which has not been solved by the previous studies.

Breast peripheral area correction in digital mammograms

Digital mammograms may present an overexposed area in the peripheral part of the breast, which is visually shown as a darker area with lower contrast. This has a direct impact on image quality and affects image visualisation and assessment. This paper presents an automatic method to enhance the overexposed peripheral breast area providing a more homogeneous and improved view of the whole mammogram. The method automatically restores the overexposed area by equalising the image using information from the intensity of non-overexposed neighbour pixels. The correction is based on a multiplicative model and on the computation of the distance map from the breast boundary. A total of 334 digital mammograms were used for evaluation. Mammograms before and after enhancement were evaluated by an expert using visual comparison. In 90.42% of the cases, the enhancement obtained improved visualisation compared to the original image in terms of contrast and detail. Moreover, results show that lesions found in the peripheral area after enhancement presented a more homogeneous intensity distribution. Hence, peripheral enhancement is shown to improve visualisation and will play a role in further development of CAD systems in mammography.

Objective models of compressed breast shapes undergoing mammography

PURPOSE

To develop models of compressed breasts undergoing mammography based on objective analysis, that are capable of accurately representing breast shapes in acquired clinical images and generating new, clinically realistic shapes.

METHODS

An automated edge detection algorithm was used to catalogue the breast shapes of clinically acquired cranio-caudal (CC) and medio-lateral oblique (MLO) view mammograms from a large database of digital mammography images. Principal component analysis (PCA) was performed on these shapes to reduce the information contained within the shapes to a small number of linearly independent variables. The breast shape models, one of each view, were developed from the identified principal components, and their ability to reproduce the shape of breasts from an independent set of mammograms not used in the PCA, was assessed both visually and quantitatively by calculating the average distance error (ADE).

RESULTS

The PCA breast shape models of the CC and MLO mammographic views based on six principal components, in which 99.2% and 98.0%, respectively, of the total variance of the dataset is contained, were found to be able to reproduce breast shapes with strong fidelity (CC view mean ADE = 0.90 mm, MLO view mean ADE = 1.43 mm) and to generate new clinically realistic shapes. The PCA models based on fewer principal components were also successful, but to a lesser degree, as the two-component model exhibited a mean ADE = 2.99 mm for the CC view, and a mean ADE = 4.63 mm for the MLO view. The four-component models exhibited a mean ADE = 1.47 mm for the CC view and a mean ADE = 2.14 mm for the MLO view. Paired t-tests of the ADE values of each image between models showed that these differences were statistically significant (max p-value = 0.0247). Visual examination of modeled breast shapes confirmed these results. Histograms of the PCA parameters associated with the six principal components were fitted with Gaussian distributions. The six-component model was also used to generate CC and MLO view mammogram breast shapes, using the mean PCA parameter values of these distributions and randomly generated values based on the fitted Gaussian distributions, which resemble clinically encountered breasts. A spreadsheet with the data necessary to apply this model is provided as the supplementary material.

CONCLUSIONS

Our PCA models of breast shapes in both mammographic views successfully reproduce analyzed breast shapes and generate new clinically relevant shapes. This work can aid in research applications which incorporate breast shape modeling, such as x-ray scatter correction, dosimetry, and image registration.

Pectoral muscle segmentation: a review

Mammograms are X-ray images of breasts which are used to detect breast cancer. The pectoral muscle is a mass of tissue on which the breast rests. During routine mammographic screenings, in medio-lateral oblique (MLO) views, the pectoral muscle turns up in the mammograms along with the breast tissues. The pectoral muscle has to be segmented from the mammogram for an effective automated computer aided diagnosis (CAD). This is due to the fact that pectoral muscles have pixel intensities and texture similar to that of breast tissues which can result in awry CAD results. As a result, a lot of effort has been put into the segmentation of pectoral muscles and finding its contour with the breast tissues. To the best of our knowledge, currently there is no definitive literature available which provides a comprehensive review about the current state of research in this area of pectoral muscle segmentation. We try to address this shortcoming by providing a comprehensive review of research papers in this area. A conscious effort has been made to avoid deviating into the area of automated breast cancer detection.

Effects of exposure equalization on image signal-to-noise ratios in digital mammography: a simulation study with an anthropomorphic breast phantom

PURPOSE

The scan equalization digital mammography (SEDM) technique combines slot scanning and exposure equalization to improve low-contrast performance of digital mammography in dense tissue areas. In this study, full-field digital mammography (FFDM) images of an anthropomorphic breast phantom acquired with an anti-scatter grid at various exposure levels were superimposed to simulate SEDM images and investigate the improvement of low-contrast performance as quantified by primary signal-to-noise ratios (PSNRs).

METHODS

We imaged an anthropomorphic breast phantom (Gammex 169 "Rachel," Gammex RMI, Middleton, WI) at various exposure levels using a FFDM system (Senographe 2000D, GE Medical Systems, Milwaukee, WI). The exposure equalization factors were computed based on a standard FFDM image acquired in the automatic exposure control (AEC) mode. The equalized image was simulated and constructed by superimposing a selected set of FFDM images acquired at 2, 1, 1/2, 1/4, 1/8, 1/16, and 1/32 times of exposure levels to the standard AEC timed technique (125 mAs) using the equalization factors computed for each region. Finally, the equalized image was renormalized regionally with the exposure equalization factors to result in an appearance similar to that with standard digital mammography. Two sets of FFDM images were acquired to allow for two identically, but independently, formed equalized images to be subtracted from each other to estimate the noise levels. Similarly, two identically but independently acquired standard FFDM images were subtracted to estimate the noise levels. Corrections were applied to remove the excess system noise accumulated during image superimposition in forming the equalized image. PSNRs over the compressed area of breast phantom were computed and used to quantitatively study the effects of exposure equalization on low-contrast performance in digital mammography.

RESULTS

We found that the highest achievable PSNR improvement factor was 1.89 for the anthropomorphic breast phantom used in this study. The overall PSNRs were measured to be 79.6 for the FFDM imaging and 107.6 for the simulated SEDM imaging on average in the compressed area of breast phantom, resulting in an average improvement of PSNR by ∼35% with exposure equalization. We also found that the PSNRs appeared to be largely uniform with exposure equalization, and the standard deviations of PSNRs were estimated to be 10.3 and 7.9 for the FFDM imaging and the simulated SEDM imaging, respectively. The average glandular dose for SEDM was estimated to be 212.5 mrad, ∼34% lower than that of standard AEC-timed FFDM (323.8 mrad) as a result of exposure equalization for the entire breast phantom.

CONCLUSIONS

Exposure equalization was found to substantially improve image PSNRs in dense tissue regions and result in more uniform image PSNRs. This improvement may lead to better low-contrast performance in detecting and visualizing soft tissue masses and micro-calcifications in dense tissue areas for breast imaging tasks.

Scan equalization digital radiography (SEDR) implemented with an amorphous selenium flat-panel detector: initial experience

It is well recognized in projection radiography that low-contrast detectability suffers in heavily attenuating regions due to excessively low x-ray fluence to the image receptor and higher noise levels. Exposure equalization can improve image quality by increasing the x-ray exposure to heavily attenuating regions, resulting in a more uniform distribution of exposure to the detector. Image quality is also expected to be improved by using the slot-scan geometry to reject scattered radiation effectively without degrading primary x-rays. This paper describes the design of a prototype scan equalization digital radiography (SEDR) system implemented with an amorphous silicon (a-Si) thin-film transistor (TFT) array-based flat-panel detector. With this system, slot-scan geometry with alternate line erasure and readout (ALER) technique was used to achieve scatter rejection. A seven-segment beam height modulator assembly was mounted onto the fore collimator to regulate exposure regionally for chest radiography. The beam modulator assembly, consisting of micro linear motors, lead screw cartridge with lead (Pb) beam blockers attached, position feedback sensors and motor driver circuitry, has been tested and found to have an acceptable response for exposure equalization in chest radiography. An anthropomorphic chest phantom was imaged in the posterior-anterior (PA) view under clinical conditions. Scatter component, primary x-rays, scatter-to-primary ratios (SPRs) and primary signal-to-noise ratios (PSNRs) were measured in the SEDR images to evaluate the rejection and redistribution of scattered radiation, and compared with those for conventional full-field imaging with and without anti-scatter grid methods. SPR reduction ratios (SPRRRs, defined as the differences between the non-grid full-field SPRs and the reduced SPRs divided by the former) yielded approximately 59% for the full-field imaging with grid and 82% for the SEDR technique in the lungs, and 77% for the full-field imaging with grid and 95% for the SEDR technique in the subdiaphragm. The SEDR technique demonstrated a substantial improvement in PSNRs over the anti-scatter grid technique. The improvements of PSNRs varied with the regions and are more pronounced in heavily attenuating regions.

Identification of the breast edge using areas enclosed by iso-intensity contours

The segmentation of a mammogram into background and breast is a crucial first step in the computer aided diagnosis of mammograms that has the advantage of simplifying further processing of the image (by eliminating the background) and also provides a reference for the alignment of views when two views are being compared. A novel method of segmenting the breast from the background by analysing the area enclosed by iso-intensity contours is presented. Results are evaluated by comparison with manual borders drawn by three radiologists for a set of 25 mammograms. The effect of different pre-processing methods, on the accuracy of automated borders, is investigated. The best methods yielded average root-mean-square differences between the manual and automated iso-intensity borders of 3.0+/-0.3 mm for the image set containing clear breast edges and 4.8+/-0.5 mm for the image set containing indistinct breast edges compared to 5.1+/-0.8 and 7.9+/-0.9 mm for the two data sets with no pre-processing. Results are also compared to those obtained from global thresholding.

Digital Mammography Image Quality: Image Display

This paper on digital mammography image display is 1 of 3 papers written as part of an intersociety effort to establish image quality standards for digital mammography. The information included in this paper is intended to support the development of an American College of Radiology (ACR) guideline on image quality for digital mammography. The topics of the other 2 papers are digital mammography image acquisition and digital mammography image storage, transmission, and retrieval. The societies represented in compiling this document were the Radiological Society of North America, the ACR, the American Association of Physicists in Medicine, and the Society for Computer Applications in Radiology. These papers describe in detail what is known to improve image quality for digital mammography and make recommendations about how digital mammography should be performed to optimize the visualization of breast cancers using this imaging tool. Through the publication of these papers, the ACR is seeking input from industry, radiologists, and other interested parties on their contents so that the final ACR guideline for digital mammography will represent the consensus of the broader community interested in these topics.

A comparison of the performance of modern screen-film and digital mammography systems

This work compares the detector performance and image quality of the new Kodak Min-R EV mammography screen-film system with the Fuji CR Profect detector and with other current mammography screen-film systems from Agfa, Fuji and Kodak. Basic image quality parameters (MTF, NPS, NEQ and DQE) were evaluated for a 28 kV Mo/Mo (HVL = 0.646 mm Al) beam using different mAs exposure settings. Compared with other screen-film systems, the new Kodak Min-R EV detector has the highest contrast and a low intrinsic noise level, giving better NEQ and DQE results, especially at high optical density. Thus, the properties of the new mammography film approach those of a fine mammography detector, especially at low frequency range. Screen-film systems provide the best resolution. The presampling MTF of the digital detector has a value of 15% at the Nyquist frequency and, due to the spread size of the laser beam, the use of a smaller pixel size would not permit a significant improvement of the detector resolution. The dual collection reading technology increases significantly the low frequency DQE of the Fuji CR system that can at present compete with the most efficient mammography screen-film systems.

Computerized nipple identification for multiple image analysis in computer-aided diagnosis

Correlation of information from multiple-view mammograms (e.g., MLO and CC views, bilateral views, or current and prior mammograms) can improve the performance of breast cancer diagnosis by radiologists or by computer. The nipple is a reliable and stable landmark on mammograms for the registration of multiple mammograms. However, accurate identification of nipple location on mammograms is challenging because of the variations in image quality and in the nipple projections, resulting in some nipples being nearly invisible on the mammograms. In this study, we developed a computerized method to automatically identify the nipple location on digitized mammograms. First, the breast boundary was obtained using a gradient-based boundary tracking algorithm, and then the gray level profiles along the inside and outside of the boundary were identified. A geometric convergence analysis was used to limit the nipple search to a region of the breast boundary. A two-stage nipple detection method was developed to identify the nipple location using the gray level information around the nipple, the geometric characteristics of nipple shapes, and the texture features of glandular tissue or ducts which converge toward the nipple. At the first stage, a rule-based method was designed to identify the nipple location by detecting significant changes of intensity along the gray level profiles inside and outside the breast boundary and the changes in the boundary direction. At the second stage, a texture orientation-field analysis was developed to estimate the nipple location based on the convergence of the texture pattern of glandular tissue or ducts towards the nipple. The nipple location was finally determined from the detected nipple candidates by a rule-based confidence analysis. In this study, 377 and 367 randomly selected digitized mammograms were used for training and testing the nipple detection algorithm, respectively. Two experienced radiologists identified the nipple locations which were used as the gold standard. In the training data set, 301 nipples were positively identified and were referred to as visible nipples. Seventy six nipples could not be positively identified and were referred to as invisible nipples. The radiologists provided their estimation of the nipple locations in the latter group for comparison with the computer estimates. The computerized method could detect 89.37% (269/301) of the visible nipples and 69.74% (53/76) of the invisible nipples within 1 cm of the gold standard. In the test data set, 298 and 69 of the nipples were classified as visible and invisible, respectively. 92.28% (275/298) of the visible nipples and 53.62% (37/69) of the invisible nipples were identified within 1 cm of the gold standard. The results demonstrate that the nipple locations on digitized mammograms can be accurately detected if they are visible and can be reasonably estimated if they are invisible. Automated nipple detection will be an important step towards multiple image analysis for CAD.

Effect of area x-ray beam equalization on image quality and dose in digital mammography

In mammography, thick or dense breast regions persistently suffer from reduced contrast-to-noise ratio (CNR) because of degraded contrast from large scatter intensities and relatively high noise. Area x-ray beam equalization can improve image quality by increasing the x-ray exposure to under-penetrated regions without increasing the exposure to other breast regions. Optimal equalization parameters with respect to image quality and patient dose were determined through computer simulations and validated with experimental observations on a step phantom and an anthropomorphic breast phantom. Three parameters important in equalization digital mammography were considered: attenuator material (Z = 13-92), beam energy (22-34 kVp) and equalization level. A Mo/Mo digital mammography system was used for image acquisition. A prototype 16 x 16 piston driven equalization system was used for preparing patient-specific equalization masks. Simulation studies showed that a molybdenum attenuator and an equalization level of 20 were optimal for improving contrast, CNR and figure of merit (FOM = CNR2/dose). Experimental measurements using these parameters showed significant improvements in contrast, CNR and FOM. Moreover, equalized images of a breast phantom showed improved image quality. These results indicate that area beam equalization can improve image quality in digital mammography.

Thickness correction of mammographic images by means of a global parameter model of the compressed breast

Peripheral enhancement and tilt correction of unprocessed digital mammograms was achieved with a new reversible algorithm. This method has two major advantages for image visualization. First, the display dynamic range can be relatively small, and second, adjustment of the overall luminance to inspect details is not required in most cases. The correction is useful for preprocessing in computer-aided detection/diagnosis algorithms. The method is based on knowledge of the three-dimensional compressed breast shape to equalize thickness by adding virtual tissue, which results in intensity equalization for the mammographic image. Previously described methods implicitly estimate the contribution of thickness variations to image intensity, usually by nonparametric methods. The proposed method employs a global parametic breast shape model, which is advantageous for visualization and CAD.

Evaluation of detector dynamic range in the x-ray exposure domain in mammography: a comparison between film-screen and flat panel detector systems

Digital detectors in mammography have wide dynamic range in addition to the benefit of decoupled acquisition and display. How wide the dynamic range is and how it compares to film-screen systems in the clinical x-ray exposure domain are unclear. In this work, we compare the effective dynamic ranges of film-screen and flat panel mammography systems, along with the dynamic ranges of their component image receptors in the clinical x-ray exposure domain. An ACR mammography phantom was imaged using variable mAs (exposure) values for both systems. The dynamic range of the contrast-limited film-screen system was defined as that ratio of mAs (exposure) values for a 26 kVp Mo/Mo (HVL=0.34 mm Al) beam that yielded passing phantom scores. The same approach was done for the noise-limited digital system. Data from three independent observers delineated a useful phantom background optical density range of 1.27 to 2.63, which corresponded to a dynamic range of 2.3 +/- 0.53. The digital system had a dynamic range of 9.9 +/- 1.8, which was wider than the film-screen system (p<0.02). The dynamic range of the film-screen system was limited by the dynamic range of the film. The digital detector, on the other hand, had an estimated dynamic range of 42, which was wider than the dynamic range of the digital system in its entirety by a factor of 4. The generator/tube combination was the limiting factor in determining the digital system's dynamic range.

X-ray beam equalization: feasibility and performance of an automated prototype system in a phantom and swine

PURPOSE

To investigate the feasibility of an automated implementation of a beam equalization technique and to evaluate the experimental performance of the prototype system.

MATERIALS AND METHODS

X-ray beam equalization involved the process of low-dose image acquisition, attenuator thickness calculation, mask generation, mask positioning, equalized image acquisition, and mask reshaping. The entire equalization process was performed in approximately 7 seconds. The equalized images were assessed both qualitatively and quantitatively by using a humanoid phantom and a swine animal model. The general image quality was assessed for the ability to visualize arterial branches and other anatomic structures. The level of equalization was quantitatively assessed by segmenting the images into an 8 x 8 matrix of square regions.

RESULTS

The ratio of the root-mean-squared variance for the equalized and unequalized images from humanoid phantom and swine animal studies was 0.49 and 0.59, respectively. Furthermore, qualitative assessment of the images showed substantial improvement in image quality and visualization of arterial branches after beam equalization in phantom and animal studies.

CONCLUSION

Automated area beam equalization is feasible and improves image quality in previously underpenetrated regions of the image.

Computerized image analysis: estimation of breast density on mammograms

An automated image analysis tool is being developed for the estimation of mammographic breast density. This tool may be useful for risk estimation or for monitoring breast density change in prevention or intervention programs. In this preliminary study, a data set of 4-view mammograms from 65 patients was used to evaluate our approach. Breast density analysis was performed on the digitized mammograms in three stages. First, the breast region was segmented from the surrounding background by an automated breast boundary-tracking algorithm. Second, an adaptive dynamic range compression technique was applied to the breast image to reduce the range of the gray level distribution in the low frequency background and to enhance the differences in the characteristic features of the gray level histogram for breasts of different densities. Third, rule-based classification was used to classify the breast images into four classes according to the characteristic features of their gray level histogram. For each image, a gray level threshold was automatically determined to segment the dense tissue from the breast region. The area of segmented dense tissue as a percentage of the breast area was then estimated. To evaluate the performance of the algorithm, the computer segmentation results were compared to manual segmentation with interactive thresholding by five radiologists. A "true" percent dense area for each mammogram was obtained by averaging the manually segmented areas of the radiologists. We found that the histograms of 6% (8 CC and 8 MLO views) of the breast regions were misclassified by the computer, resulting in poor segmentation of the dense region. For the images with correct classification, the correlation between the computer-estimated percent dense area and the "truth" was 0.94 and 0.91, respectively, for CC and MLO views, with a mean bias of less than 2%. The mean biases of the five radiologists' visual estimates for the same images ranged from 0.1% to 11%. The results demonstrate the feasibility of estimating mammographic breast density using computer vision techniques and its potential to improve the accuracy and reproducibility of breast density estimation in comparison with the subjective visual assessment by radiologists.

A digital density equalization technique to improve visualization of breast periphery in mammography

In mammographic imaging, the film area corresponding to the breast periphery is overexposed, resulting in high optical density and degraded contrast in this region. A digital, model-driven density equalization technique was designed and developed to overcome this overexposure problem, taking into account the non-linear characteristic curve of the film-digitizer system. The method is based on several image processing and analysis techniques, such as thresholding, which is used to segment the pixels of the mammogram belonging to the breast region from the background, and wavelet-based fusion, which is used to equalize the pixels of breast periphery selectively while leaving the remaining breast region unaffected. Initial application of the method resulted in density-equalized mammographic images, characterized by improved contrast at the breast periphery.

Area x-ray beam equalization for digital angiography

An area beam equalization technique has been investigated in order to generate patient-specific compensating filters for digital angiography. An initial image was used to generate the compensating filter, which was fabricated using a deformable compensating material, containing CeO2, and an array of square pistons. The CeO2 attenuator thicknesses were calculated using the gray level information from the initial unequalized image. The array of pistons was pressed against a uniform thickness of attenuating material to generate a filter for x-ray beam equalization. The filter was subsequently inserted into the x-ray beam for the final equalized radiograph. It was positioned close to the focal spot (magnification of 8.0) in order to minimize edge artifacts from the filter. The equalization of x-ray transmission across the field exiting from the object significantly improved the image quality by preserving local contrast throughout the image. The contrast-to-noise ratio (CNR) in the equalized images was increased-by up to fivefold. Phantom studies indicate that equalized images using a relatively small array of pistons (e.g., 8 x 8) produce significant improvement in image quality with negligible perceptible artifacts. Animal studies showed that beam equalization significantly improved fluoroscopic and angiographic image quality. X-ray beam equalization produced an image with a relatively uniform scatter-glare intensity and it reduced the scatter-glare fraction in the previously underpenetrated region of the image from 0.65 to 0.50. Also, x-ray tube loading due to the mask assembly itself was negligible. In conclusion, area beam equalization reduces the scatter-glare fraction and significantly improves CNR in the previously underpenetrated region of the image.

A phantom-based evaluation of an exposure equalization technique in mammography

An anatomical filter based exposure equalization technique in mammography is evaluated quantitatively using a phantom. The evaluation is carried out by a comparative observer performance study, comparing the equalization technique with a conventional one based on visualization of low contrast, 6 mm circular details and high contrast, 0.5 mm and 0.25 mm small size details. These details are situated at the phantom edge, simulating the breast periphery. Visualization of these details is studied with respect to the parameters of tube voltage, optical density, detail location and phantom thickness. Phantom images are interpreted independently by three observers using a four-point grading scale. Use of the Wilcoxon signed ranks test for paired data shows statistically highly significant improvement (p < 0.0001) in the visualization of details for the equalization technique for all values of the parameters studied. The improvement is independent of tube voltage but dependent on optical density, detail location and phantom thickness. Optimal performance is obtained for detail location closer to the outer border of the simulated breast periphery and/or further away from the film, as well as for a greater phantom thickness simulating both thick and dense breast.

Design and evaluation of an external filter technique for exposure equalization in mammography

We are developing an external filter method for equalizing x-ray exposure in the peripheral region of the breast. This method requires the use of only a limited number of custom-built filters for different breast shapes in a given view. This paper describes the design methodology for these external filters. The filter effectiveness was evaluated through a simulation study on 171 mediolateral and 196 craniocaudal view digitized mammograms and through imaging of a breast phantom. The degree of match between the simulated filter and the individual 3-D exposure profiles at the breast periphery was quantified. An analysis was performed to investigate the effect of filter misalignment. The simulation study indicates that the filter is effective in equalizing exposures for more than 80% of the breast images in our database. The tolerance in filter misalignment was estimated to be about +/- 2 mm for the CC view and +/- 1 mm for the MLO view at the image plane. Some misalignment artifacts were demonstrated with simulated filtered mammograms.