The purpose is to develop and validate an automated method for detecting image unsharpness caused by patient motion blur in digital mammograms. The goal is that such a tool would facilitate immediate re-taking of blurred images, which has the potential to reduce the number of recalled examinations, and to ensure that sharp, high-quality mammograms are presented for reading. To meet this goal, an automated method was developed based on interpretation of the normalized image Wiener Spectrum. A preliminary algorithm was developed using 25 cases acquired using a single vendor system, read by two expert readers identifying the presence of blur, location, and severity. A predictive blur severity score was established using multivariate modeling, which had an adjusted coefficient of determination, R2 =0.63±0.02, for linear regression against the average reader-scored blur severity. A heatmap of the relative blur magnitude showed good correspondence with reader sketches of blur location, with a Spearman rank correlation of 0.70 between the algorithmestimated area fraction with blur and the maximum of the blur area fraction categories of the two readers. Given these promising results, the algorithm-estimated blur severity score and heatmap are proposed to be used to aid observer interpretation. The use of this automated blur analysis approach, ideally with feedback during an exam, could lead to a reduction in repeat appointments for technical reasons, saving time, cost, potential anxiety, and improving image quality for accurate diagnosis.

Poster: ECR 2016 / C-0793 / Optimizing VolparaDensity’s volumetric breast density thresholds to align with the American College of Radiology BI-RADS 5th edition updates by: C. Tromans , A. Chan, L. Johnston, R. Highnam; Wellington/NZ

Poster: ECR 2016 / C-0854 / A validation study of automated mammographic breast positioning metrics by: K. Wang, E. Ross, N. Khan, C. Tromans , L. Johnston, A. Chan, R. Highnam; Wellington/NZ

Poster: EuroSafe Imaging 2016 / ESI-0033 / Real-Time Mammography Dose Monitoring and Optimisation using Big Data by: C. Tromans1, L. Johnston2, R. Highnam2; 1Volpara Solutions Ltd Wellington/NZ, 2 Wellington/NZ

Minimising the mean glandular dose (MGD) received by the patient whilst maximising image contrast during mammographic imaging is of paramount importance due to the widespread use of the modality for screening, where subjects are for the most part healthy. The advent of digital mammography brought about a general reduction in MGD, however the introduction of tomosynthesis, particularly when used in combination with conventional projection mammography has the potential for unwanted and often unnecessary MGD increases. We describe a method to calculate the patient-specific MGD using a representation of the patient’s volumetric breast density to derive the breast glandularity. This personalises the MGD to the individual woman, rather than assuming a constant value, or one that depends solely on compressed breast thickness. The calculated patient specific MGDs are compared to those reported by the manufacturer for a database of 2D mammograms. Though agreement is generally good for dense breasts, we have found that the MGD is underestimated in fatty breasts. A separate database of 2D mammogram and 3D tomosynthesis acquisitions acquired in “combo” is also analysed. In general, the MGDs are approximately equal for dense (VDG 3 and 4) breasts, but fatty (VDG 1 and 2) breasts exhibited significant differences with tomosynthesis MGDs being higher than mammogram MGDs for these cases.

Purpose: It is essential to keep the dose to the breast as low as possible during routine mammography screening while achieving high diagnostic quality images. Currently, mammography systems report average glandular dose (AGD) assuming a 4.5 cm compressed thickness, 50% glandular and 50% adipose tissue equivalent phantom and typical x‐ray beam characteristics. In this paper, we estimate the AGD for specific breasts on specific mammography systems. Methods: For 16 digital mammograms, the technique parameters such as kVp, mAs and compressed thickness were collected from the DICOM header. An automated software system (Volpara (R), Wellington, New Zealand) estimated breast glandularity by analyzing each digital mammogram. For the specific mammography systems used, entrance air kerma and half‐value layer (HVL) was measured for various kVp and target/filter combinations at various compressed breast thicknesses. The patient‐specific AGD was calculated from dose models developed by Dance et. al. These AGD estimates were compared with those generated by the mammography system. Results: For the 16 mammograms that were analyzed, the calculated patient‐specific AGD ranged from 1.99 ± 0.20 mGy to 3.73 ± 0.37 mGy for breasts with range of calculated glandularity (approximately 3% to 27%) and compression thickness (45 mm to 68 mm). Our preliminary results show that our estimated patient‐specific AGDs differed from the AGD reported on the mammography system for each mammogram by an average of approximately 32%. The maximum difference between our estimated patient‐specific AGD and the displayed AGD was approximately 46%. Conclusion: Dose estimations implemented by manufacturers may underestimate the actual patient AGD significantly as they do not account for the specific breast tissue composition, compressed breast thickness or x‐ray beam parameters of the specific mammography system used. Patient‐specific dose estimates may give a better indication of the actual dose delivered, and therefore the risk to the patient.

Breast density is a key component of risk assessment for personalised screening, necessitating robust, repeatable measures. The Standard Attenuation Rate (SAR) enables the quantification of breast tissue radiodensity at each pixel, relative to the attenuation of a reference material, so may be used as a measure of volumetric breast density. A major complication is quantification of tissue in the periphery of the breast, the (often substantial) region between the skin boundary and the point at which the breast occupies the entire distance between the plates, since the thickness is governed by the shape of the compressed breast, rather than the separation of the plates. We present a method to measure the compressed shape from the image, hence the thickness at each point in the periphery. The method exploits the vastly different attenuation of the various breast tissues from that of air, and uses spatial smoothing to glean a signal estimating solely the underlying thickness. An iterative refinement procedure allows for variation in scatter in the periphery arising from the air boundary edge effects. The outcome of the inclusion of the periphery in breast density quantified by this method is analysed, and the importance of this region's inclusion illustrated.

Digital x-ray acquisition allows the sophisticated processing of acquired images before display to the reader, making possible such operations as the removal in software of the systematic blurring effect of scatter. A method for analysing scatter removal is presented. The scatter model incorporated within the Standard Attenuation Rate (SAR) is used, which is a method for calculating a normalised image of tissue radiodensity. The model builds on the fundamental physical relations underlying Monte Carlo techniques; but through optimal information sampling and interpolation is able to execute in a clinically realistic time. The scatter kernel arising around each primary ray is calculated, and these are superimposed to give the scatter image. An iterative refinement procedure is used to calculate the radiodensity and scatter at each ray/pixel, cyclically feeding back to each other, to yield the scatter field. Image sharpness and contrast-to-noise (CNR) analysis is presented for two tissue equivalent phantoms. The algorithm is found to be able to match image sharpness without the grid, to that with the grid present, confirmed by residual analysis using autocorrelation plots which show the difference is almost white noise within a 95% C.I. The increased fluence in the absence of the grid is shown to allow dose to be reduced by 37-49%, whilst delivering equivalent contrast and CNR.

We present an efficient method to calculate the primary and scattered x-ray photon fluence component of a mammographic image. This can be used for a range of clinically important purposes, including estimation of breast density, personalized image display, and quantitative mammogram analysis. The method is based on models of: the x-ray tube; the digital detector; and a novel ray tracer which models the diverging beam emanating from the focal spot. The tube model includes consideration of the anode heel effect, and empirical corrections for wear and manufacturing tolerances. The detector model is empirical, being based on a family of transfer functions that cover the range of beam qualities and compressed breast thicknesses which are encountered clinically. The scatter estimation utilizes optimal information sampling and interpolation (to yield a clinical usable computation time) of scatter calculated using fundamental physics relations. A scatter kernel arising around each primary ray is calculated, and these are summed by superposition to form the scatter image. Beam quality, spatial position in the field (in particular that arising at the air-boundary due to the depletion of scatter contribution from the surroundings), and the possible presence of a grid, are considered, as is tissue composition using an iterative refinement procedure. We present numerous validation results that use a purpose designed tissue equivalent step wedge phantom. The average differences between actual acquisitions and modelled pixel intensities observed across the adipose to fibroglandular attenuation range vary between 5% and 7%, depending on beam quality and, for a single beam quality are 2.09% and 3.36% respectively with and without a grid.

A method is presented which quantifies the radiodensity of lesions in projection images, providing a diagnostic indicator to better inform the decisions of both human readers and computer algorithms. The models of image formation underlying the Standard Attenuation Rate (SAR) are used to facilitate the forward simulation of the appearance of a lesion in a breast. By forming hypotheses, informed from measurements on the acquired image, virtual 3D scenes are constructed which predict the size, position and radiodensity of a suspect lesion and the surrounding breast tissue. Comparisons between simulations of this scene, and the acquired image enable both the refinement of the hypothesis, and the assessment of the likelihood of the hypothesis being correct. In the event of a high likelihood of correctness, the hypothesised lesion informs diagnosis. The application of the method to a patient image containing a cyst shows it has an attenuation corresponding to water (SAR 1.246), and an invasive carcinoma which is considerably denser at SAR 2.27. Thus the technique yields a quantitative radiodensity measure for discrimination in diagnostic decision making.

The analysis of (x-ray) mammograms remains qualitative, relying on the judgement of clinicians. We present a novel method to compute a quantitative, normalized measure of tissue radiodensity traversed by the primary beam incident on each pixel of a mammogram, a measure we term the standard attenuation rate (SAR). SAR enables: the estimation of breast density which is linked to cancer risk; direct comparison between images; the full potential of computer aided diagnosis to be utilized; and a basis for digital breast tomosynthesis reconstruction. It does this by removing the effects of the imaging conditions under which the mammogram is acquired. First, the x-ray spectrum incident upon the breast is calculated, and from this, the energy exiting the breast is calculated. The contribution of scattered radiation is calculated and subtracted. The SAR measure is the scaling factor that must be applied to the reference material in order to match the primary attenuation of the breast. Specifically, this is the scaled reference material attenuation which when traversed by an identical beam to that traversing the breast, and when subsequently detected, results in the primary component of the pixel intensity observed in the breast image. We present results using two tissue equivalent phantoms, as well as a sensitivity analysis to detector response changes over time and possible errors in compressed thickness measurement.

Scattered photons degrade mammographic image quality, so, almost universally, a physical anti-scatter grid is used to limit their effect Physical grids are not completely effective in rejecting only scattered photons, so patient dose must be increased in order to maintain low levels of quantum noise The standard attenuation rate (SAR), a quantitative normalised representation of breast tissue for image analysis applications, incorporates a model of scatter, and a software correction of the image blurring arising from scatter within the image signal A tissue equivalent phantom is used to investigate the possibility, in terms of both image sharpness and noise, of replacing the physical grid with the software correction in the SAR Encouraging results are reported, software correction almost matching the performance of the grid, whilst maintaining a superior signal-to-noise ratio.

The detection of microcalcifications is a key task in the early detection of breast cancer. Digital breast tomosynthesis (DBT) offers new opportunities to improve existing microcalcification detection methods. By utilizing the multiple projections in DBT, and a model of the DBT acquisition system, we propose the use of epipolar curves to constrain the position of a microcalcification in the multiple DBT views. We show how this can improve both the sensitivity and specification of microcalcification detection.

A common metric used to optimise digital mammography image acquisition is contrast-to-noise ratio. Using the standard attenuation rate (SAR), a quantitative normalised representation of breast tissue for image analysis applications, we demonstrate that the image contrast may be completely separated from the acquisition parameters, in particular the beam quality, used for acquisition. Optimising the contrast-to-noise ratio at acquisition is therefore suboptimal, since the contrast may be manipulated by post processing. A tissue equivalent phantom is used to investigate the variation in both signal-to-noise ratio, and image sharpness within the SAR images. The results show that the primary effect of varying the acquisition parameters through the various automated optimisation of parameter modes, and hence the mean glandular dose, is to vary the global contrast of the acquired image, an effect successfully mapped to a common normalised basis using the SAR. The signal-to-noise ratio and image sharpness are second order effects, and are therefore dominated by the global image contrast when image acquisition is optimised using the contrast-to-noise ratio.

We introduce the Standard Attenuation Rate (SAR), a quantitative, and normalised measure of radiodensity per unit distance traversed by the primary beam incident on each pixel of an x-ray mammogram is presented We sketch an algorithm to compute the SAR The calculation utilises a physics model of image formation, including consideration of photon production in the x-ray tube, photon detection within the image receptor, and photon scattering occurring within the tissues of the breast Using the model, the difference in the flux incident upon, and exiting from, the breast is quantified relative to a reference material Experimental validation of the SAR representation is presented, based on a tissue equivalent phantom designed and manufactured specifically for the purpose The observed performance across the clinical range of acquisition parameters is very promising, supporting the suitability of this approach to form the basis of a next generation of diagnostic techniques based on quantitative tissue measurement.

DBT provides significantly more information than mammography. This offers new opportunities to improve existing microcalcification detection methods. In a companion work in this volume, we showed that the use of epipolar curves can improve both the sensitivity and specificity of microcalcification detection. In this paper, we develop a clustering algorithm to form epipolar curves from candidate microcalcifications (which may be noise points), obtained after applying a detection algorithm to each individual projection. This enables the subsequent 3D analysis for the classification of microcalcification clusters.

The detection of microcalcifications, reconstruction of clusters of microcalcifications and their subsequent classification into malignant and benign are important tasks in the early detection of breast cancer. Digital breast tomosynthesis (DBT) provides new opportunities in such tasks. By utilizing the multiple projections in DBT and using the geometry of DBT, we have developed an approach to them based on epipolar curves. It improves the sensitivity and specificity in detection; provides information for estimation of 3D positions of microcalcifications; and facilitates classification. We have generated 15 simulated datasets, each with a microcalcification cluster based on an ellipsoidal shape. We estimate the 3D positions of the microcalcifications in each of the clusters and reconstruct the clusters as ellipsoids. We classify each cluster as malignant or benign based on the parameters of the ellipsoids. The classification result is compared with the ground truth. Our results show that the deviations between the actual and estimated 3D positions of the microcalcification, and the actual and estimated parameters of the ellipsoids are sufficiently small that the classification results are 100% correct. This demonstrates the feasibility in cluster classification in 3D.

We present a novel method for the detection and reconstruction in 3D of microcalcifications in digital breast tomosynthesis (DBT) image sets. From a list of microcalcification candidate regions (that is, real microcalcification points or noise points) found in each DBT projection, our method: (1) finds the set of corresponding points of a microcalcification in all the other projections; (2) locates its 3D position in the breast; (3) highlights noise points; and (4) identifies the failure of microcalcification detection in one or more projections, in which case the method predicts the image locations of the microcalcification in the images in which they are missed. From the geometry of the DBT acquisition system, an "epipolar curve" is derived for the 2D positions a microcalcification in each projection generated at different angular positions. Each epipolar curve represents a single microcalcification point in the breast. By examining the n projections of m microcalcifications in DBT, one expects ideally m epipolar curves each comprising n points. Since each microcalcification point is at a different 3D position, each epipolar curve will be at a different position in the same 2D coordinate system. By plotting all the microcalcification candidates in the same 2D plane simultaneously, one can easily extract a representation of the number of microcalcification points in the breast (number of epipolar curves) and their 3D positions, the noise points detected (isolated points not forming any epipolar curve) and microcalcification points missed in some projections (epipolar curves with less than n points).

Breast density is a well-known breast cancer risk factor. Most current methods of measuring breast density are area based and subjective. Standard mammogram form (SMF) is a computer program using a volumetric approach to estimate the percent density in the breast. The aim of this study is to evaluate the current implementation of SMF as a predictor of breast cancer risk by comparing it with other widely used density measurement methods. The case-control study comprised 634 cancers with 1,880 age-matched controls combined from the Cambridge and Norwich Breast Screening Programs. Data collection involved assessing the films based both on Wolfe's parenchymal patterns and on visual estimation of percent density and then digitizing the films for computer analysis (interactive threshold technique and SMF). Logistic regression was used to produce odds ratios associated with increasing categories of breast density. Density measures from all four methods were strongly associated with breast cancer risk in the overall population. The stepwise rises in risk associated with increasing density as measured by the threshold method were 1.37 [95% confidence interval (95% CI), 1.03-1.82], 1.80 (95% CI, 1.36-2.37), and 2.45 (95% CI, 1.86-3.23). For each increasing quartile of SMF density measures, the risks were 1.11 (95% CI, 0.85-1.46), 1.31 (95% CI, 1.00-1.71), and 1.92 (95% CI, 1.47-2.51). After the model was adjusted for SMF results, the threshold readings maintained the same strong stepwise increase in density-risk relationship. On the contrary, once the model was adjusted for threshold readings, SMF outcome was no longer related to cancer risk. The available implementation of SMF is not a better cancer risk predictor compared with the thresholding method.

A technique is presented for computing a normalised image in which each pixel directly describes the radiodensity of the underlying anatomy. Precisely, each pixel quantifies the equivalent thickness of reference material per unit traversal distance required to match the radiodensity of the breast tissues present within the traversal between the focal spot and the image receptor pixel. Measurements are computed using a model of image formation, which includes consideration of both the attenuation and scattering phenomena that occur. In view of the complexity of the underlying model, substantial computational optimisation has been made to yield clinically realistic execution times. Validation experiments are described using a purpose designed and manufactured tissue equivalent test object which allows both the assessment of the performance of the image normalisation, and a comparison with “ground truth”.