Image processing can cause some malignant soft-tissue lesions to be missed in digital mammography images.

AIM: To investigate the effect of image processing on cancer detection in mammography. METHODS AND MATERIALS: An observer study was performed using 349 digital mammography images of women with normal breasts, calcification clusters, or soft-tissue lesions including 191 subtle cancers. Images underwent two types of processing: FlavourA (standard) and FlavourB (added enhancement). Six observers located features in the breast they suspected to be cancerous (4,188 observations). Data were analysed using jackknife alternative free-response receiver operating characteristic (JAFROC) analysis. Characteristics of the cancers detected with each image processing type were investigated. RESULTS: For calcifications, the JAFROC figure of merit (FOM) was equal to 0.86 for both types of image processing. For soft-tissue lesions, the JAFROC FOM were better for FlavourA (0.81) than FlavourB (0.78); this difference was significant (p=0.001). Using FlavourA a greater number of cancers of all grades and sizes were detected than with FlavourB. FlavourA improved soft-tissue lesion detection in denser breasts (p=0.04 when volumetric density was over 7.5%) CONCLUSIONS: The detection of malignant soft-tissue lesions (which were primarily invasive) was significantly better with FlavourA than FlavourB image processing. This is despite FlavourB having a higher contrast appearance often preferred by radiologists. It is important that clinical choice of image processing is based on objective measures.

Further factors for the estimation of mean glandular dose using the United Kingdom, European and IAEA breast dosimetry protocols.

The United Kingdom, European and IAEA protocols for breast dosimetry in mammography make use of s-factors which allow for the use of different target/filter combinations. To supplement the existing protocols, a Monte Carlo computer program has been used to calculate s-factors for mammography using a tungsten target with silver filters of thicknesses 50-75 microm and for the same target filtered with 0.5 mm aluminium. The dosimetry protocols use slabs of polymethyl methacrylate (PMMA) of specified thicknesses to simulate the exposure of typical breasts. The equivalent thickness of PMMA has been calculated using a simplified approach for a wider range of x-ray spectra and for breast thicknesses of 2-11 cm. The results show that for the tungsten/silver target/filter combination, a single s-factor of 1.042 can be used with the protocols, but when the tungsten target is filtered with 0.5 mm of aluminium, it is necessary to select from a tabulation of s-factors against breast thickness. The equivalent thicknesses of PMMA for a given breast thickness show some dependence on beam quality and the values obtained differ from those presently used in the dosimetry protocols by an amount which depends upon breast thickness and half value layer (HVL). For the extreme case of an 11 cm breast and an HVL of 0.62 mm Al, the use of the protocol thickness would give rise to an error of 10%, but for breast thicknesses of 6 cm or less, the error is typically 2-3%.

Amorphous silicon EPID calibration for dosimetric applications: comparison of a method based on Monte Carlo prediction of response with existing techniques.

For EPID dosimetry, the calibration should ensure that all pixels have a similar response to a given irradiation. A calibration method (MC), using an analytical fit of a Monte Carlo simulated flood field EPID image to correct for the flood field image pixel intensity shape, was proposed. It was compared with the standard flood field calibration (FF), with the use of a water slab placed in the beam to flatten the flood field (WS) and with a multiple field calibration where the EPID was irradiated with a fixed 10x10 field for 16 different positions (MF). The EPID was used in its normal configuration (clinical setup) and with an additional 3 mm copper slab (modified setup). Beam asymmetry measured with a diode array was taken into account in MC and WS methods. For both setups, the MC method provided pixel sensitivity values within 3% of those obtained with the MF and WS methods (mean difference<1%, standard deviation<2%). The difference of pixel sensitivity between MC and FF methods was up to 12.2% (clinical setup) and 11.8% (modified setup). MC calibration provided images of open fields (5x5 to 20x20 cm2) and IMRT fields to within 3% of that obtained with WS and MF calibrations while differences with images calibrated with the FF method for fields larger than 10x10 cm2 were up to 8%. MC, WS and MF methods all provided a major improvement on the FF method. Advantages and drawbacks of each method were reviewed.

Can the channelized Hotelling observer including aspects of the human visual system predict human observer performance in mammography?

PURPOSE: In mammography, images are processed prior to display. Model observers (MO) are candidates to objectively evaluate processed images if they can predict human observer performance for detail detection. The aim of this study was to investigate if the channelized Hotelling observer (CHO) can be configured to predict human observer performance in mammography like images. METHODS: The performance correlation between human observers and CHO has been evaluated using different channel-sets and by including aspects of the human visual system (HVS). The correlation was investigated for the detection of disk-shaped details in simulated white noise (WN) and clustered lumpy backgrounds (CLB) images, representing respectively quantum noise limited and mammography like images. The images were scored by the MO and five human observers in 2-alternative forced choice experiments. RESULTS: For WN images the most useful formulation of the CHO to predict human observer performance was obtained using three difference of Gaussian channels without adding HVS aspects (RLR2=0.62). For CLB images the most useful formulation was the partial least square channel-set without adding HVS aspects (RLR2=0.71). The correlation was affected by detail size and background. CONCLUSIONS: This study has shown that the CHO can predict human observer performance. Due to object size and background dependency it is important that the range of object sizes and allowed variability in background are specified and validated carefully before the CHO can be implemented for objective image quality assessment.

Design and application of a structured phantom for detection performance comparison between breast tomosynthesis and digital mammography.

This paper introduces and applies a structured phantom with inserted target objects for the comparison of detection performance of digital breast tomosynthesis (DBT) against 2D full field digital mammography (FFDM). The phantom consists of a 48 mm thick breast-shaped polymethyl methacrylate (PMMA) container filled with water and PMMA spheres of different diameters. Three-dimensionally (3D) printed spiculated masses (diameter range: 3.8-9.7 mm) and non-spiculated masses (1.6-6.2 mm) along with microcalcifications (90-250 µm) were inserted as targets. Reproducibility of the phantom application was studied on a single system using 30 acquisitions. Next, the phantom was evaluated on five different combined FFDM & DBT systems and target detection was compared for FFDM and DBT modes. Ten phantom images in both FFDM and DBT modes were acquired on these 5 systems using automatic exposure control. Five readers evaluated target detectability. Images were read with the four-alternative forced-choice (4-AFC) paradigm, with always one segment including a target and 3 normal background segments. The percentage of correct responses (PC) was assessed based on 10 trials of each reader for each object type, size and imaging modality. Additionally, detection threshold diameters at 62.5 PC were assessed via non-linear regression fitting of the psychometric curve. The reproducibility study showed no significant differences in PC values. Evaluation of target detection in FFDM showed that microcalcification detection thresholds ranged between 110 and 118 µm and were similar compared to the detection in DBT (range of 106-158 µm). In DBT, detection of both mass types increased significantly (p = 0.0001 and p = 0.0002 for non-spiculated and spiculated masses respectively) compared to FFDM, achieving almost 100% detection for all spiculated mass diameters. In conclusion, a structured phantom with inserted targets was able to show evidence for detectability differences between FFDM and DBT modes for five commercial systems. This phantom has potential for application in task-based assessment at acceptance and commissioning testing of DBT systems.

Can the non-pre-whitening model observer, including aspects of the human visual system, predict human observer performance in mammography?

PURPOSE: In mammography, images are processed prior to display. Current methodologies based on physical image quality measurements are however not designed for the evaluation of processed images. Model observers (MO) might be suitable for this evaluation. The aim of this study was to investigate whether the non-pre-whitening (NPW) MO can be used to predict human observer performance in mammography-like images by including different aspects of the human visual system (HVS). METHODS: The correlation between human and NPW MO performance has been investigated for the detection of disk shaped objects in simulated white noise (WN) and clustered lumpy backgrounds (CLB), representing quantum noise limited and mammography-like images respectively. The images were scored by the MO and five human observers in a 2-alternative forced choice experiment. RESULTS: For WN images it was found that the log likelihood ratio (RLR2), which expresses the goodness of fit, was highest (0.44) for the NPW MO without addition of HVS aspects. For CLB the RLR2 improved from 0.46 to 0.65 with addition of HVS aspects. The correlation was affected by object size and background. CONCLUSIONS: This study shows that by including aspects of the HVS, the performance of the NPW MO can be improved to better predict human observer performance. This demonstrates that the NPW MO has potential for image quality assessment. However, due to the dependencies found in the correlation, the NPW MO can only be used for image quality assessment for a limited range of object sizes and background variability.

SIMULATING LOCAL DENSE AREAS USING PMMA TO ASSESS AUTOMATIC EXPOSURE CONTROL IN DIGITAL MAMMOGRAPHY.

Current digital mammography (DM) X-ray systems are equipped with advanced automatic exposure control (AEC) systems, which determine the exposure factors depending on breast composition. In the supplement of the European guidelines for quality assurance in breast cancer screening and diagnosis, a phantom-based test is included to evaluate the AEC response to local dense areas in terms of signal-to-noise ratio (SNR). This study evaluates the proposed test in terms of SNR and dose for four DM systems. The glandular fraction represented by the local dense area was assessed by analytic calculations. It was found that the proposed test simulates adipose to fully glandular breast compositions in attenuation. The doses associated with the phantoms were found to match well with the patient dose distribution. In conclusion, after some small adaptations, the test is valuable for the assessment of the AEC performance in terms of both SNR and dose.

Estimation of the peak entrance surface air kerma for patients undergoing computed tomography-guided procedures.

The purpose of this work was to develop a method for estimating the patient peak entrance surface air kerma from measurements using a pencil ionisation chamber on dosimetry phantoms exposed in a computed tomography (CT) scanner. The method described is especially relevant for CT fluoroscopy and CT perfusion procedures where the peak entrance surface air kerma is the risk-related quantity of primary concern. Pencil ionisation chamber measurements include scattered radiation, which is outside the primary radiation field, and that must be subtracted in order to derive the peak entrance surface air kerma. A Monte Carlo computer model has therefore been used to calculate correction factors, which may be applied to measurements of the CT dose index obtained using a pencil ionisation chamber in order to estimate the peak entrance surface air kerma. The calculations were made for beam widths of 5, 7, 10 and 20 mm, for seven positions of the phantom, and for the geometry of a GE HiSpeed CT/i scanner. The program was validated by comparing measurements and calculations of CTDI for various vertical positions of the phantom and by directly estimating the peak ESAK using the program. Both validations showed agreement within statistical uncertainties (standard deviation of 2.3% or less). For the GE machine, the correction factors vary by approximately 10% with slice width for a fixed phantom position, being largest for the 20 mm beam width, and at that beam width range from 0.87 when the phantom surface is at the isocentre to 1.23 when it is displaced vertically by 24 cm.

CT dosimetry: getting the best from the adult Cristy phantom.

The use of geometrical phantoms for computed tomography (CT) dosimetry can incur errors in the calculation of effective dose due to the anatomically incorrect organ shapes and distributions, and unrepresentative body dimensions. A Monte Carlo program that makes use of an anatomically correct voxel phantom has been developed to calculate effective doses in CT and to compare with conventional dosimetric techniques. The code was validated against the latter by matching the phantom dimensions and simulating whole-body irradiation; agreement to within 6% was found. Effective doses were calculated for brain, lung, abdomen and pelvis CT scans for voxel phantom sizes corresponding to those of standard-sized adult, a teenager and 10% greater than those of standard-sized adult. Errors incurred by using the conventional techniques are minimised if the scan range is set by matching the fractions of radiosensitive organs that are irradiated directly. Under these circumstances, the conventional techniques will underestimate the dose to a 15 y old by up to 22% while the dose to a large subject is overestimated by up to 11%.

Breast dosimetry using high-resolution voxel phantoms.

A computer model of X-ray mammography has been developed, which uses quasi-realistic high-resolution voxel phantoms to simulate the breast. The phantoms have 400 microm voxels and simulate the three-dimensional distributions of adipose and fibro-glandular tissues, Cooper's ligaments, ducts and skin and allow the estimation of dose to individual tissues. Calculations of the incident air kerma to mean glandular dose conversion factor, g, were made using a Mo/Mo spectrum at 28 kV for eight phantoms in the thickness range 40-80 mm and of varying glandularity. The values differed from standard tabulations used for breast dosimetry by up to 43%, because of the different spatial distribution of glandular tissue within the breast. To study this further, additional voxel phantoms were constructed, which gave variations of between 9 and 59% compared with standard values. For accurate breast dosimetry, it is therefore very important to take the distribution of glandular tissues into account.

Calculation of the properties of digital mammograms using a computer simulation.

A Monte Carlo computer model of mammography has been developed to study and optimise the performance of digital mammographic systems. The program uses high-resolution voxel phantoms to model the breast, which simulate the adipose and fibroglandular tissues, Cooper's ligaments, ducts and skin in three dimensions. The model calculates the dose to each tissue, and also the quantities such as energy imparted to image pixels, noise per image pixel and scatter-to-primary (S/P) ratios. It allows studies of the dependence of image properties on breast structure and on position within the image. The program has been calibrated by calculating and measuring the pixel values and noise for a digital mammographic system. The thicknesses of two components of this system were unknown, and were adjusted to obtain a good agreement between measurement and calculation. The utility of the program is demonstrated with the calculations of the variation of the S/P ratio with and without a grid, and of the image contrast across the image of a 50-mm-thick breast phantom.

Monte Carlo simulation of a mammographic test phantom.

A test phantom, including a wide range of mammographic tissue equivalent materials and test details, was imaged on a digital mammographic system. In order to quantify the effect of scatter on the contrast obtained for the test details, calculations of the scatter-to-primary ratio (S/P) have been made using a Monte Carlo simulation of the digital mammographic imaging chain, grid and test phantom. The results show that the S/P values corresponding to the imaging conditions used were in the range 0.084-0.126. Calculated and measured pixel values in different regions of the image were compared as a validation of the model and showed excellent agreement. The results indicate the potential of Monte Carlo methods in the image quality-patient dose process optimisation, especially in the assessment of imaging conditions not available on standard mammographic units.

Dose assessment in contrast enhanced digital mammography using simple phantoms simulating standard model breasts.

Slabs of polymethyl methacrylate (PMMA) or a combination of PMMA and polyethylene (PE) slabs are used to simulate standard model breasts for the evaluation of the average glandular dose (AGD) in digital mammography (DM) and digital breast tomosynthesis (DBT). These phantoms are optimized for the energy spectra used in DM and DBT, which normally have a lower average energy than used in contrast enhanced digital mammography (CEDM). In this study we have investigated whether these phantoms can be used for the evaluation of AGD with the high energy x-ray spectra used in CEDM. For this purpose the calculated values of the incident air kerma for dosimetry phantoms and standard model breasts were compared in a zero degree projection with the use of an anti scatter grid. It was found that the difference in incident air kerma compared to standard model breasts ranges between -10% to +4% for PMMA slabs and between 6% and 15% for PMMA-PE slabs. The estimated systematic error in the measured AGD for both sets of phantoms were considered to be sufficiently small for the evaluation of AGD in quality control procedures for CEDM. However, the systematic error can be substantial if AGD values from different phantoms are compared.

Average glandular dose in digital mammography and digital breast tomosynthesis: comparison of phantom and patient data.

For the evaluation of the average glandular dose (AGD) in digital mammography (DM) and digital breast tomosynthesis (DBT) phantoms simulating standard model breasts are used. These phantoms consist of slabs of polymethyl methacrylate (PMMA) or a combination of PMMA and polyethylene (PE). In the last decades the automatic exposure control (AEC) increased in complexity and became more sensitive to (local) differences in breast composition. The question is how well the AGD estimated using these simple dosimetry phantoms agrees with the average patient AGD. In this study the AGDs for both dosimetry phantoms and for patients have been evaluated for 5 different x-ray systems in DM and DBT modes. It was found that the ratios between patient and phantom AGD did not differ considerably using both dosimetry phantoms. These ratios averaged over all breast thicknesses were 1.14 and 1.15 for the PMMA and PMMA-PE dosimetry phantoms respectively in DM mode and 1.00 and 1.02 in the DBT mode. These ratios were deemed to be sufficiently close to unity to be suitable for dosimetry evaluation in quality control procedures. However care should be taken when comparing systems for DM and DBT since depending on the AEC operation, ratios for particular breast thicknesses may differ substantially (0.83-1.96). Although the predictions of both phantoms are similar we advise the use of PMMA + PE slabs for both DM and DBT to harmonize dosimetry protocols and avoid any potential issues with the use of spacers with the PMMA phantoms.

Schemes for the optimization of chest radiography using a computer model of the patient and x-ray imaging system.

A computer program has been developed to model chest radiography. It incorporates a voxel phantom of an adult and includes antiscatter grid, radiographic screen, and film. Image quality is quantified by calculating the contrast (deltaOD) and the ideal observer signal-to-noise ratio (SNR(I)) for a number of relevant anatomical details at various positions in the anatomy. Detector noise and system unsharpness are modeled and their influence on image quality is considered. A measure of useful dynamic range is computed and defined as the fraction of the image that is reproduced at an optical density such that the film gradient exceeds a preset value. The effective dose is used as a measure of the radiation risk for the patient. A novel approach to patient dose and image quality optimization has been developed and implemented. It is based on a reference system acknowledged to yield acceptable image quality in a clinical trial. Two optimizations schemes have been studied, the first including the contrast of vessels as measure of image quality and the second scheme using also the signal-to-noise ratio of calcifications. Both schemes make use of our measure of useful dynamic range as a key quantity. A large variety of imaging conditions was simulated by varying the tube voltage, antiscatter device, screen-film system, and maximum optical density in the computed image. It was found that the optical density is crucial in screen-film chest radiography. Significant dose savings (30%-50%) can be accomplished without sacrificing image quality by using low-atomic-number grids with a low grid ratio or an air gap and more sensitive screen-film system. Dose-efficient configurations proposed by the model agree well with the example of good radiographic technique suggested by the European Commission.

Quantitative measurement of small-angle gamma ray scattering from water, nylon, and Lucite.

The small-angle (2 degrees-10 degrees) total-differential-scattering cross sections of water, methyl methacrylate, and nylon 6 have been measured at a photon energy of 59.54 keV. An annular target geometry has been used with an acceptance of +/- 0.5 degrees and an experimental uncertainty of approximately 4% achieved. The experimental methods and data corrections necessary to obtain this accuracy are discussed in detail. The results demonstrate clearly the contribution of intermolecular effects to the scattering process and allow critical comparison with calculations using molecular form factors for water obtained from interpolated x-ray diffractometer data; agreement is obtained within the experimental uncertainties. They also support earlier Monte Carlo predictions of scattering in water at angles of the order of 2 degrees which are in disagreement with certain published experimental data.

Segmentation of mammograms using multiple linked self-organizing neural networks.

A possible first stage in the analysis of the mammographic scene is its segmentation into four major components: background (the nonbreast area), pectoral muscle, fibroglandular region (parenchyma), and adipose region. An algorithm has been developed for this task. It is based on the classification of a feature vector constructed from statistical measures of texture calculated at two window sizes. Separate self-organizing neural networks are trained on sample data taken from each of the four regions. The feature vectors from the entire mammogram are then classified with the trained networks linked via a decision logic. To overcome the variability of texture between mammograms the algorithm uses data from a mammogram to classify itself in a staged approach consisting of several binary decisions. The training regions for each successive stage are determined from geometric information produced by the previous stages. The dataset in the study consisted of thirty (fifteen pairs) digitized normal mammograms of variable radiographic appearance. As a measure of performance, the outlines of the parenchyma were compared to those drawn by a radiologist experienced in reading mammograms. Comparison of the areas and perimeters generated by the human and computer observers gives a relationship with correlation coefficients of 0.74 and 0.59 for each measure, respectively. The overlapping areas of the parenchymas segmented by the observers normalized by the combined area was also calculated for each case. The mean and standard deviation of this measure was 0.69 +/- 0.12.

The Monte Carlo calculation of integral radiation dose in xeromammography.

A Monte Carlo computer program has been developed for the computation of integral radiation dose to the breast in xeromammography. The results are given in terms of the integral dose per unit area of the breast per unit incident exposure. The calculations have been made for monoenergetic incident photons and the results integrated over a variety of X-ray spectra from both tungsten and molybdenum targets. This range incorporates qualities used in conventional and xeromammography. The program includes the selenium plate used in xeroradiography; the energy absorbed in this detector has also been investigated. The latter calculations have been used to predict relative values of exposure and of integral dose to the breast for xeromammograms taken at various radiation qualities. The results have been applied to recent work on the reduction of patient exposure in xeromammography by the addition of aluminium filters to the X-ray beam.

Monte Carlo calculation of conversion factors for the estimation of mean glandular breast dose.

The IPSM report on the commissioning and routine testing of mammographic x-ray systems recommends that breast dose be specified as the mean dose to the glandular tissues within the breast and gives the size and compositions of a standard breast phantom for the comparison of doses. The dose to this standard breast phantom can be determined by measuring the incident air kerma to a Perspex phantom and applying appropriate multiplicative conversion factors. These conversion factors have been evaluated by Monte Carlo calculations for a wide range of mammographic x-ray spectra. Some factors are provided for a range of breast thicknesses to supplement existing tabulations. Results are also given for equivalent thicknesses of Perspex and breast tissue.

The computation of scatter in mammography by Monte Carlo methods.

A Monte Carlo computer program has been used to calculate the ratio of the scatter-to-primary radiation recorded by a range of image receptors used in mammography. The dependence of this ratio on breast size and photon energy has been investigated and the contributions of different sources of scattered radiation examined. The use of magnification and grid techniques have been included in the program and the performance of the Philips mammographic grid examined in detail. The results of the calculations are in good agreement with the limited experimental data available.