Contrast-detail phantom scoring methodology

Analysis of the threshold image contrast obtained with the CDMAM 3.4 and CDMAM 4.0 phantoms

Quality control in mammography is a very important element. One of the parameters indicating the appropriate image quality is the threshold image contrast. The CDMAM phantom is used to measure this parameter. It is currently available in two versions 3.4 and 4.0. The aim of this work is to compare the threshold image contrast readings obtained with the CDMAM 3.4 and CDMAM 4.0 phantoms. In the measurements, 9 CDMAM 4.0 phantoms were used to check the difference in indications of individual copies. The phantom whose readings were closest to the average of all readings was used for comparative measurements with the CDMAM 3.4 phantom. Measurements were made on 40 mammography devices. The obtained images were read with the software provided by the phantom manufacturer and the CDMAM Analysis v2.3.0 (NCCPM) software. The average percentage difference between the minimum and maximum values indicated by the CDMAM 4.0 phantoms was 10.09%. Using the CDMAM Analysis v2.3.0 (NCCPM) software, the average difference in readings between the CDMAM 3.4 and CDMAM 4.0 phantoms is 7.93%, and when using the software provided by the phantom manufacturer, it is as much as 60.15%. The obtained results of the threshold image contrast are affected by the type of software used for reading and the accuracy of the execution of individual elements of the phantom. It is recommended to use CDMAM Analysis v2.3.0 (NCCPM) software or the latest software provided by the phantom manufacturer to read the phantom images.

Explainability for deep learning in mammography image quality assessment

The application of deep learning has recently been proposed for the assessment of image quality in mammography. It was demonstrated in a proof-of-principle study that the proposed approach can be more efficient than currently applied automated conventional methods. However, in contrast to conventional methods, the deep learning approach has a black-box nature and, before it can be recommended for the routine use, it must be understood more thoroughly. For this purpose, we propose and apply a new explainability method: the oriented, modified integrated gradients (OMIG) method. The design of this method is inspired by the integrated gradientsmethod but adapted considerably to the use case at hand. To further enhance this method, an upsampling technique is developed that produces high-resolution explainability maps for the downsampled data used by the deep learning approach. Comparison with established explainability methods demonstrates that the proposed approach yields substantially more expressive and informative results for our specific use case. Application of the proposed explainability approach generally confirms the validity of the considered deep learning-based mammography image quality assessment (IQA) method. Specifically, it is demonstrated that the predicted image quality is based on a meaningful mapping that makes successful use of certain geometric structures of the images. In addition, the novel explainability method helps us to identify the parts of the employed phantom that have the largest impact on the predicted image quality, and to shed some light on cases in which the trained neural networks fail to work as expected. While tailored to assess a specific approach from deep learning for mammography IQA, the proposed explainability method could also become relevant in other, similar deep learning applications based on high-dimensional images.

Introduction of a New Parameter for Evaluation of Digital Radiography System Performance

Background

The aim of this study was to compare the image quality and radiation doses in various digital radiography systems using contrast-detail radiography (CDRAD) phantom.

Methods

The image quality and radiation dose for seven different digital radiography systems were compared using the CDRAD phantom. Incident air kerma (IAK) values were measured for certain exposure settings in all digital radiography systems. The images from the CDRAD phantom were evaluated by three observers. The results were displayed in the form of a contrast-detail (CD) curve. In addition, the inverse image quality figure (IQFinv)-to-IAK ratios were used for quantitative comparison of different digital radiography system performance.

Results

Results of this study showed that the CD curves cannot be suitable criterion for determining the performance of digital radiography systems. For this reason, IQFinv-to-radiation dose (IAK) ratios in a fixed radiation condition were used. The highest performance in terms of producing high-quality images and low radiation dose was related to X-ray unit 1 and the lowest performance was for X-ray unit 5.

Conclusion

The ratio of IQFinv to IAK for performance evaluation of digital radiography systems is an innovation of this study. A digital radiography system with a higher IQFinv-to-IAK ratio is associated with lower patient dose and better image quality. Therefore, it is recommended to equip the new imaging centers with the systems that have higher IQFinv-to-IAK ratios.

Mammography Image Quality Assurance Using Deep Learning

OBJECTIVE

According to the European Reference Organization for Quality Assured Breast Cancer Screening and Diagnostic Services (EUREF) image quality in mammography is assessed by recording and analyzing a set of images of the CDMAM phantom. The EUREF procedure applies an automated analysis combining image registration, signal detection and nonlinear fitting. We present a proof of concept for an end-to-end deep learning framework that assesses image quality on the basis of single images as an alternative.

METHODS

Virtual mammography is used to generate a database with known ground truth for training a regression convolutional neural net (CNN). Training is carried out by continuously extending the training data and applying transfer learning.

RESULTS

The trained net is shown to correctly predict the image quality of simulated and real images. Specifically, image quality predictions on the basis of single images are of similar quality as those obtained by applying the EUREF procedure with 16 images. Our results suggest that the trained CNN generalizes well.

CONCLUSION

Mammography image quality assessment can benefit from the proposed deep learning approach.

SIGNIFICANCE

Deep learning avoids cumbersome pre-processing and allows mammography image quality to be estimated reliably using single images.

COMPARISON OF RADIATION EXPOSURE TO THE PATIENT AND CONTRAST DETAIL RESOLUTIONS ACROSS LOW DOSE 2D/3D SLOT SCANNER AND TWO CONVENTIONAL DIGITAL RADIOGRAPHY X-RAY IMAGING SYSTEMS

PURPOSE

To assess and compare the radiation dose and image quality of the low dose 2D/3D EOS slot scanner (LDSS) to conventional digital radiography (DR) X-ray imaging systems for chest and knee examination protocols.

METHODS AND MATERIALS

The effective doses (ED) to the patient in the chest and knee clinical examination protocols for LDSS and DR X-ray imaging systems were determined using the dose area product and PCXMC Monte Carlo simulation software. The CDRAD phantom was imaged with 19 cm, and 13 cm thick Polymethyl Methacrylate (PMMA) blocks to simulate the chest and knees respectively of a patient of average adult size. The contrast detail resolution was calculated using image analysis software.

RESULTS

The EDs for the LDSS default setting were up to 69% and 51% lower than for the DR systems for the chest (speed 4) and knee (speed 6) protocols, respectively, while for the increased dose level setting then the EDs were up to 42% and 35% lower than for the DR systems for the chest (speed 6) and knee (speed 8) protocols respectively. At the default setting, the contrast detail was lowest for the default setting of the 2D/3D low dose slot scanner (LDSS) for both chest and knee examinations, but at the highest dose levels then the threshold were equal or higher than the contrast resolution of DR imaging systems.

CONCLUSION

The LDSS has the potential to be used for clinical diagnosis of chest and knee examinations using the higher dose level. For speed 6 in chest protocol and speed 8 in knee protocol, the measured contrast detail resolution was comparable with the DR systems but at a lower effective dose.

Model and human observer reproducibility for detection of microcalcification clusters in digital breast tomosynthesis images of three-dimensionally structured test object

We compare the reproducibility of the human observers and a channelized Hotelling observer (CHO), when reading digital breast tomosynthesis (DBT) images of a physical phantom containing a breast simulating structured background and calcification clusters at three dose levels. The phantom is scanned 217 times on a Siemens Inspiration DBT system. Volumes of interest, with and without the calcification targets, are extracted and the human observers' percentage of correct (PC) scores is evaluated using a four-alternative forced choice method. A two-layer CHO is developed using the human observer results. The first layer consists of a localizing CHO that identifies the most conspicuous calcifications using two Laguerre-Gauss channels. Then a CHO with eight Gabor channels estimates the PC score for the calcification cluster. Observer reproducibility is estimated by bootstrapping, and the standard deviation (SD) is used as a figure of merit. The CHO closely approximated the human observer results for all the three dose levels with a correlation of > 0.97 . For the larger calcification cluster sizes, both observers have similar reproducibility, whereas the CHO is more reproducible for the smaller calcifications, with a maximum of 5.5 SD against 13.1 SD for the human observers. The developed CHO is a good candidate for automated reading of the calcification clusters of the structured phantom, with better reproducibility than the human readers for small calcifications.

Imaging and Dosimetric Study on Direct Flat-Panel Detector-Based Digital Mammography System

INTRODUCTION

Image quality of digital mammography system is generally defined by three primary physical parameters, namely, contrast, resolution, and noise. Quantification of these metrics can be done by measuring objective image quality parameters defined as contrast-to-noise ratio (CNR), modulation transfer function (MTF), and noise power spectra (NPS).

MATERIALS AND METHODS

In the present study, various imaging metrics such as CNR, contrast detail resolution, MTF, and NPS were evaluated for a direct flat-panel detector-based digital mammography system following the European Guidelines. Furthermore, system performance relating to both image quality and doses were evaluated using figure of merit (FOM) in terms of CNR2/mean glandular dose (MGD) under automatic exposure control (AEC) and clinically used OPDOSE operating mode.

RESULTS AND CONCLUSION

Under AEC mode, FOM values for the 4.5 cm thick BARC polymethyl methacrylate (PMMA) phantom were found to be 15.02, 15.88, and 19.82 at Mo/Mo, Mo/Rh, and W/Rh target/filter (T/F), respectively. Under OPDOSE mode, FOM values were found to 65.32, 11.80, and 1.14 for the BARC PMMA phantom thickness of 2, 4.5, and 8 cm, respectively. Under OPDOSE mode, the calculated MGD values for three Computerized Imaging Reference Systems slab phantoms having total thickness of 7 cm were observed to be 3.03, 2.32, and 1.75 mGy with glandular/adipose tissue compositions of 70/30, 50/50, and 30/70, respectively, whereas for the 2-8-cm thick BARC PMMA phantom, the calculated MGDs were found to be in the range of 0.57-3.32 mGy. All the calculated MGDs values were found to be lower than the acceptable level of dose limits provided in European Guidelines.

Evaluating clinical implications of 15-mega-sub-pixel liquid-crystal display in phase-contrast mammography

BACKGROUND

Phase-contrast mammography (PCM) systems characteristically yield sharp images with edge enhancement and high-resolution 25-μm/pixel mammograms. However, not all PCM image information can be shown on the display at a resolution of 5-megapixel (5-MP), although 5-MP monitors are recommended for interpretation of digital mammograms. Therefore, we investigated the potential utility of a 15-mega-sub-pixel (15-MsP) display for PCM images.

METHODS

We used a monitor that offered both 5-MP and 15-MsP displays by using a sub-pixel drive (SPD) technique to increase the spatial resolution of the monitor by threefold in the direction of the sub-pixels. Contrast-detail mammography phantom images were evaluated visually by four radiologic technologists. In this study, four display magnification ratios were used and the calculated image quality figures (IQFs) were compared with those of a 5-MP display.

RESULTS

The detection capability of the 15-MsP display was significantly better than that of the 5-MP display at magnification ratios of 49 and 100 %. At other magnification ratios, the detection capability of the 15-MsP display was higher than that of the 5-MP display, but the difference was not significant.

CONCLUSIONS

A 15-MsP display has the potential to provide better detection than that provided by conventional 5-MP displays. A 15-MsP display using SPD technology is suitable for high-resolution digital mammograms, such as those produced by PCM systems.

Comparison of signal to noise ratios from spatial and frequency domain formulations of nonprewhitening model observers in digital mammography

PURPOSE

Image quality indices based upon model observers are promising alternatives to laborious human readings of contrast-detail images. This is especially appealing in digital mammography as limiting values for contrast thresholds determine, according to some international protocols, the acceptability of these systems in the radiological practice. The objective of the present study was to compare the signal to noise ratios (SNR) obtained with two nonprewhitening matched filter model observer approaches, one in the spatial domain and the other in the frequency domain, and with both of them worked out for disks as present in the CDMAM phantom.

METHODS

The analysis was performed using images acquired with the Siemens Novation and Inspiration digital mammography systems. The spatial domain formulation uses a series of high dose CDMAM images as the signal and a routine exposure of two flood images to calculate the covariance matrix. The frequency domain approach uses the mathematical description of a disk and modulation transfer function (MTF) and noise power spectrum (NPS) calculated from images.

RESULTS

For both systems most of the SNR values calculated in the frequency domain were in very good agreement with the SNR values calculated in the spatial domain. Both the formulations in the frequency domain and in the spatial domain show a linear relationship between SNR and the diameter of the CDMAM discs.

CONCLUSIONS

The results suggest that both formulations of the model observer lead to very similar figures of merit. This is a step forward in the adoption of figures of merit based on NPS and MTF for the acceptance testing of mammography systems.

Preliminary investigation of the clinical usefulness of super-high-resolution LCDs with 9 and 15 mega-sub-pixels: observation studies with phantoms

Our purpose in this study was to evaluate the preliminary clinical efficacy of soft-copy reading of digital mammography, for a 15-mega-sub-pixel (MsP) and a 9-MsP super-high-resolution liquid-crystal display (SHR-LCD) by use of an independent sub-pixel driving technology. We performed three kinds of phantom observation studies by six radiological technologists. Detectability of a contrast-detail phantom and simulated small objects (SSOs) resembling microcalcifications (MCLs), and shape discrimination ability of SSOs with round and square shapes, were examined and compared with a 5-MP conventional LCD (5-MP LCD). In each study, four types of display magnification ratio were used. The detectability and the shape discrimination ability of the 15-MsP SHR-LCD were highest among the three LCDs of most of the display magnification ratios. The 9-MsP SHR-LCD indicated a higher or equal performance as compared with the 5-MP LCD in the SSO detection and shape studies. The results of our study demonstrated that the SHR-LCDs had good potential to detect MCLs and to evaluate the shape in high-resolution digital mammography.

Using a NPWE model observer to assess suitable image quality for a digital mammography quality assurance programme

A method of objectively determining imaging performance for a mammography quality assurance programme for digital systems was developed. The method is based on the assessment of the visibility of a spherical microcalcification of 0.2 mm using a quasi-ideal observer model. It requires the assessment of the spatial resolution (modulation transfer function) and the noise power spectra of the systems. The contrast is measured using a 0.2-mm thick Al sheet and Polymethylmethacrylate (PMMA) blocks. The minimal image quality was defined as that giving a target contrast-to-noise ratio (CNR) of 5.4. Several evaluations of this objective method for evaluating image quality in mammography quality assurance programmes have been considered on computed radiography (CR) and digital radiography (DR) mammography systems. The measurement gives a threshold CNR necessary to reach the minimum standard image quality required with regards to the visibility of a 0.2-mm microcalcification. This method may replace the CDMAM image evaluation and simplify the threshold contrast visibility test used in mammography quality.

Analysis of image quality parameter of conventional and dental radiographic digital images

The image quality obtained by a radiographic equipment is very useful to characterize the physical properties of the image radiographic chain, in a quality control of the radiographic equipment. In the radiographic technique it is necessary that the evaluation of the image can guarantee the constancy of its quality to carry out a suitable diagnosis. In this work we have designed some radiographic phantoms for different radiographic digital devices, as dental, conventional, equipments with computed radiography (phosphor plate) and direct radiography (sensor) technology. Additionally, we have developed a software to analyse the image obtained by the radiographic equipment with digital processing techniques, as edge detector, morphological operators, statistical test for the detected combinations‥ The design of these phantoms let the evaluation of a wide range of operating conditions of voltage, current and time of the digital equipments. Moreover, the image quality analysis by the automatic software, let study it with objective parameters.

Physical characteristics of GE Senographe Essential and DS digital mammography detectors

The purpose of this study was to investigate physical characteristics of two full field digital mammography (FFDM) systems (GE Senographe Essential and DS). Both are indirect conversion (x ray to light) alpha-Si flat panels coupled with a CsI(Tl) scintillator. The examined systems have the same pixel size (100 microm) but a different field of view: a conventional size 23 x 19.2 cm2 and a large field 24 X 30.7 cm2, specifically designed to image large breasts. In the GE Senographe Essential model relevant improvements in flat panel design were implemented and new deposition tools for metal, alpha-Si, and CsI(Tl) were introduced by GE. These changes in detector design are expected to be beneficial for advanced applications such as breast tomosynthesis. The presampling modulation transfer function (MTF), normalized noise power spectrum (NNPS), and detective quantum efficiency (DQE) were measured for a wide range of exposure (25-240 microGy) with a RQA-M2 technique (28 kVp with a Mo/Mo target/filter combination and 2 mm of additional aluminum filtration). At 1, 2, and at 4 lp/mm MTF is equal to 0.9, 0.76, and 0.46 for the conventional field detector and to 0.85, 0.59, and 0.24 for the large field detector. The latter detector exhibits an improved NNPS due to a lower electronic noise and a better DQE that reaches 60%. In addition a contrast-detail analysis was performed with CDMAM 3.4 phantom and CDCOM software: GE Senographe DS showed statistically significant poorer detection ability in comparison with the GE Senographe Essential. These results could have been expected, at least qualitatively, considering the relative DQE of the two systems.

Comparison of different commercial FFDM units by means of physical characterization and contrast-detail analysis

The purpose of this study was to perform a complete evaluation of three pieces of clinical digital mammography equipment. Image quality was assessed by performing physical characterization and contrast-detail (CD) analysis. We considered three different FFDM systems: a computed radiography unit (Fuji "FCR 5000 MA") and two flat-panel units, the indirect conversion a-Si based GE "Senographe 2000D" and the direct conversion a-Si based IMS "Giotto Image MD." The physical characterization was estimated by measuring the MTF, NNPS, and DQE of the detectors with no antiscatter grid and over the clinical range of exposures. The CD analysis was performed using a CDMAM 3.4 phantom and custom software designed for automatic computation of the contrast-detail curves. The physical characterization of the three digital systems confirms the excellent MTF properties of the direct conversion flat-panel detector (FPD). We performed a relative standard deviation (RSD) analysis, for investigating the different components of the noise presented by the three systems. It turned out that the two FPDs show a significant additive component, whereas for the CR system the statistical noise is dominant. The multiplicative factor is a minor constituent for all the systems. The two FPDs demonstrate better DQE, with respect to the CR system, for exposures higher than 70 microGy. The CD analysis indicated that the three systems are not statistically different for detail objects with a diameter greater than 0.3 mm. However, the IMS system showed a statistically significant different response for details smaller than 0.3 mm. In this case, the poor response of the a-Se detector could be attributed to its high-frequency noise characteristics, since its MTF, NEQ, and DQE are not inferior to those of the other systems. The CD results were independent of exposure level, within the investigated clinical range. We observed slight variations in the CD results, due to the changes in the visualization parameters (window/level and magnification factor). This suggests that radiologists would benefit from viewing images using varied window/level and magnification.

Amorphous selenium flat panel detectors for digital mammography: Validation of a NPWE model observer with CDMAM observer performance experiments

Model observers have been developed which incorporate a specific imaging task, system performance, and human observer characteristics and can potentially overcome some of the limitations in using detective quantum efficiency for optimization and comparison of detectors. In this paper, a modified nonprewhitening matched filter (NPWE) model observer was developed and validated to predict object detectability for an amorphous selenium (a-Se) direct flat-panel imager (FPI) where aliasing is severe. A preclinical a-Se digital mammography FPI with 85 microm pixel size was used in this investigation. Its physical imaging properties including modulation transfer function (MTF), noise power spectrum, and DQE were fully characterized. An observer performance study was conducted by imaging the CDMAM 3.4 contrast-detail phantom designed specifically for digital mammography and presenting these images to a panel of seven observers. X-ray attenuation and scatter due to the phantom were determined experimentally for use in development of the model observer. The observer study results were analyzed via threshold averaging and signal detection theory (SDT) based techniques to produce contrast-detail curves where threshold contrast is plotted as a function of disk diameter. Validity of the model was established using SDT analysis of the experimental data. The effect of aliasing on the detectability of small diameter disks was determined using the NPWE model observer. The signal spectrum was calculated using the presampling MTF of the detector with and without including the aliased terms. Our results indicate that the NPWE model based on Fourier domain parameters provides reasonable prediction of object detectability for the signal-known-exactly task in uniform image noise for a-Se direct FPI.

Toward objective and quantitative evaluation of imaging systems using images of phantoms

The use of imaging phantoms is a common method of evaluating image quality in the clinical setting. These evaluations rely on a subjective decision by a human observer with respect to the faintest detectable signal(s) in the image. Because of the variable and subjective nature of the human-observer scores, the evaluations manifest a lack of precision and a potential for bias. The advent of digital imaging systems with their inherent digital data provides the opportunity to use techniques that do not rely on human-observer decisions and thresholds. Using the digital data, signal-detection theory (SDT) provides the basis for more objective and quantitative evaluations which are independent of a human-observer decision threshold. In a SDT framework, the evaluation of imaging phantoms represents a "signal-known-exactly/background-known-exactly" ("SKE/ BKE") detection task. In this study, we compute the performance of prewhitening and nonprewhitening model observers in terms of the observer signal-to-noise ratio (SNR) for these "SK E/BKE" tasks. We apply the evaluation methods to a number of imaging systems. For example, we use data from a laboratory implementation of digital radiography and from a full-field digital mammography system in a clinical setting. In addition, we make a comparison of our methods to human-observer scoring of a set of digital images of the CDMAM phantom available from the internet (EUREF-European Reference Organization). In the latter case, we show a significant increase in the precision of the quantitative methods versus the variability in the scores from human observers on the same set of images. As regards bias, the performance of a model observer estimated from a finite data set is known to be biased. In this study, we minimize the bias and estimate the variance of the observer SNR using statistical resampling techniques, namely, "bootstrapping" and "shuffling" of the data sets. Our methods provide objective and quantitative evaluation of imaging systems with increased precision and reduced bias.