Medical Physics and Informatics
Review
X-Ray Phase-Contrast Technology in Breast Imaging: Principles, Options, and Clinical Application
Audio Available | 
Share
OBJECTIVE. The purpose of this article is to review different x-ray phase-contrast breast imaging techniques and their potential application in clinical settings.
CONCLUSION. Phase-contrast imaging depicts not only the absorption contrast but also the refraction contrast of the transmitted x-ray beam. Early data suggest that this new modality may overcome some of the diagnostic limitations associated with current clinically available mammography systems and that it has potential for improving breast cancer detection.
Keywords: breast imaging, CT, mammography, phase-contrast, x-ray refraction
Breast cancer is the most common cancer in women worldwide with more than 1.6 million new cases diagnosed per year [1]. This constitutes more than 25% of all cancers diagnosed in women, excluding nonmelanoma skin cancer. Breast cancer is currently the leading cause of cancer death among women in less developed regions of the world and the second cause of cancer death in more developed regions, after lung cancer [1].
Early detection of breast cancer is crucial for the provision of appropriate treatment and prevention of late-stage disease. Over several decades, breast screening programs, particularly in developed countries, have considerably reduced breast cancer mortality rates. Mortality rate reductions as high as 45% among women participating in screening programs are reported in the literature [2]. However, an independent review and meta-analysis of 11 randomized trials in developed countries showed that mortality reduction from these programs is probably closer to 20% [3]. Sensitivity and specificity in the existing screening programs depend on various factors, including the imaging modality used; patient characteristics, such as breast density; and radiographers' and radiologists' expertise in performing examinations and interpreting results [4–6].
Currently, 2D digital mammography (DM) is the most common imaging modality for breast screening and diagnostic imaging of recalled patients [7]. Nevertheless, screening mammography has limitations, particularly in diagnostic efficacy. These limitations mainly result from the superimposition of breast tissue on 2D images [8]. The overlying tissue obscures cancers (lowering sensitivity) or makes the appearance of normal structures suspicious (lowering specificity) [9]. The average sensitivity of screening DM is approximately 70% [10], but it varies for different breast densities and patient ages and can be as low as 30% among women with extremely dense breasts [11]. The average specificity of screening DM is also approximately 92%, meaning that 8% of healthy women (without breast cancer) are recalled unnecessarily for further investigations [10]. In addition, there is a small radiation risk from x-ray exposure [12], and substantial patient discomfort or pain is commonly associated with mammographic breast compression, which is necessary to acquire high-contrast images [13].
Digital breast tomosynthesis (DBT) is an innovative imaging technology designed to overcome some of the limitations of mammography. DBT entails the same technology as DM, but the x-ray tube moves in an arc around the breast and produces tomographic 3D images from 2D projection images acquired from different but limited view angles [14]. DBT reduces the effects of tissue super-imposition found with 2D mammography, resulting in higher sensitivity and specificity for the combination of DBT and mammography [8]. It has been argued [9, 15] that integration of DM and DBT in a screening environment improves breast cancer detection and reduces false-positive findings in recalls. Although the specificity of DBT is higher than that of DM, some research has shown lower sensitivity of DBT alone compared with DM, particularly in detection of calcifications and calcification clusters [8, 16, 17]. DBT is currently approved by the U.S. Food and Drug Administration to be performed only as an adjunct to DM [8, 18]. The two examinations may increase patient radiation dose, but the advent of efficient synthetic 2D images should alleviate this potential problem.
Breast CT is another technology that can minimize the effects of overlying anatomic tissues by having full 3D capability for breast examination. In this technology, the woman is positioned prone and the breast is allowed to fall forward away from the chest wall uncompressed. The x-ray tube and the flat-panel detector rotate around the breast in the horizontal plane to produce many cone-beam projections, which are subsequently used in the reconstruction of the 3D breast CT image [19]. Breast CT outperforms mammography for visualization of mass lesions [19, 20], but it is inferior to mammography in spatial resolution [21] and depiction of microcalcifications [19] and delivers a larger radiation dose.
Ultrasound imaging is also used in breast cancer diagnosis as a supplement to mammography for differentiation of cysts from suspicious masses, particularly in dense parenchyma [22, 23]. Supplemental ultrasound imaging increases the number of true-positive findings (increasing sensitivity) in high-risk women but concomitantly escalates the numbers of false-positive findings (decreasing specificity) [22, 24]. Moreover, breast ultrasound is a time-consuming process and highly dependent on operator expertise [20, 25]. Another alternative modality is MRI, which has higher sensitivity than mammography, particularly in imaging of women at high risk and those with dense breasts [26, 27]. Nevertheless, breast MRI has lower specificity than mammography, which can lead to avoidable recalls and biopsies [27, 28]. Other limitations of breast MRI include higher cost, longer image acquisition time, and exclusion of patients with claustrophobia or metallic implants [8, 20, 29].
In summary, all current clinically available methods of breast imaging have limitations, particularly in terms of diagnostic performance (sensitivity and specificity), excessive recall of disease-free women, unnecessary irradiation, benign biopsy results, patient discomfort associated with breast compression, and cost. Phase-contrast imaging (PCI) is an innovative medical imaging technology for improving breast cancer diagnosis that may overcome some of the challenges of current modalities [30]. PCI is currently evolving from early theoretic research to clinical studies. The aims of this article are to briefly describe the techniques used in PCI and to review previous work in this novel technology with a focus on breast cancer imaging.
| Phase-Contrast Imaging Technology |   |
|---|
When x-rays are used in medical imaging, the differences in density and atomic number of various components of an object result in different degrees of absorption of the x-rays and are ultimately responsible for image formation and image contrast. This has been the underlying principle of x-ray imaging since Roentgen's discovery in 1895. However, the x-ray absorption contrast between different types of soft tissues in the breast is minimal, which leads to the problem of low image contrast. This is a fundamental challenge common to all absorption-based x-ray imaging methods, such as DM and DBT. To overcome this limitation the radiographic technique requires using a lower-energy x-ray beam that increases the absorbed radiation dose to radiosensitive breast tissue.
A lesser known outcome of the interaction between incident x-rays and electrons in the imaged object is altered x-ray phase patterns [31, 32]. Application of this type of process is commonly known as PCI. Whereas conventional x-ray imaging methods exploit only x-ray attenuation properties, PCI can capture the refraction of x-rays and so may overcome the challenge of minimal soft-tissue attenuation differences.
When an incident x-ray beam penetrates an object, a key quantity determining the interaction of x-rays and the object is the complex refractive index [33]. This index is a dimensionless measure that can be different at different locations inside the object and that determines how the electromagnetic wave propagates through the material. The complex refractive index (n) of an object can be described with the following expression [34]:
 |
where r denotes a position inside the object, E is the x-ray energy, the real decrement δ is responsible for the phase shift and describes how the phase pattern of the x-ray changes, i is the imaginary unit, and the imaginary part β determines the level of x-ray attenuation and describes how the transmission of x-rays through the object changes.
For the x-ray energy ranges usually used in imaging breast tissue, δ is considerably larger than β [34], and at typical mammographic x-ray energies, the phase-shift term can be up to 1000 times as large as the absorption term [35]. The difference in the absolute delta and beta values for breast tissue at typical mammographic x-ray energies provides only an indirect indication of the potential benefits of PCI compared with absorption-based imaging. To obtain more accurate estimates of this benefit, the contrast mechanism specific to each particular method of PCI must be considered. The relevant specific results in the case of propagation-based and analyzer-based PCI and CT can be found in an article by Nesterets and Gureyev [36]. For example, for propagation-based phase-contrast CT at E = 32 keV compared with absorption-based CT at the same dose with a detector that has 50-μm pixels and 2.5-m distance between the sample and the detector, the gain factor in the contrast-to-noise ratio for glandular tissue relative to adipose tissue is approximately 10. This corresponds to the dose reduction factor of 100 in propagation-based phase-contrast technology at the same image quality as in conventional CT. For analyzer-based CT with a Si(111) analyzer, compared with conventional CT at the x-ray energy and detector pixel size mentioned previously, the corresponding gain factor is approximately 21. Therefore, being able to use phase shifts as a contrast mechanism could markedly improve image information.
Several key techniques have been developed to exploit the phase-contrast information in PCI. The main ones are propagation-based imaging (PBI), analyzer-based imaging, crystal interferometry, grating interferometry, and edge illumination imaging. Each technique has different principles and experimental setups for translating the phase shift of x-rays into intensity variations in the output image. We summarize the implication of each technique in breast imaging using the literature over the last 5 years. To organize the review of relevant literature, we extracted journal articles between 2012 and 2016 that included “phase contrast” or “phase-contrast” and one of the words “breast,” “mammography,” “mammogram,” and “mammographic” in their titles. The best possible effort was made to identify and include all journal articles with these criteria in the current review.
PBI, also known as in-line PCI, is experimentally the simplest technique to exploit the x-ray phase-shift information for clinical implementation [37, 38]. To our knowledge, the only patient trial of PCI of the breast was conducted with PBI technique [39]. The PBI method is based on free-space propagation (an empty air or vacuum environment with no absorbing or reflecting optical elements between the sample and the detector) and use of phase-retrieval algorithms to fully use phase-contrast technique [40].
When an x-ray beam transmits through an object, differences in electron densities of different components of the object lead to unique changes in the phase shift of the output beam. The PBI method is used to measure the phase shift as intensity modulation at the detector by simply positioning the detector a few meters from the object [41, 42]. This is because a long sample-to-detector distance can support an improved signal-to-noise ratio in PCI, provided that the incident x-rays are sufficiently coherent. As quantitatively modeled by the transport of intensity equation [43], the longitudinal intensity derivative is related to the transverse Laplacian of the phase. Thus, when the difference between the intensity distributions in two adjacent planes orthogonal to the optical axis can be measured, the equation can be solved to retrieve the phase shifts [38, 44]. A modified method for phase retrieval based on the so-called homogeneous version of the transport of intensity equation [45] allows x-ray phase retrieval from a single in-line image.
The setup required for PBI technique is illustrated in Figure 1. The setup for this technique is simpler than alternative methods of PCI, but it does require a highly spatially coherent x-ray beam [46]. Spatial coherence of electromagnetic waves, including x-rays, means that for any point on the wavefront, the directions of propagation are limited to a narrow cone around the main local direction. Equivalently, the oscillations of the field at different spatial points are correlated with each other, provided that the points are separated by a distance less than the coherence length. In practical terms, within the considered techniques, the spatial coherence of the radiation is achieved by means of a microfocus x-ray source (producing a quasispheric x-ray wave) or a highly parallel x-ray beam (quasiplanar x-ray wave). PBI setup also requires a detector with small pixel size (a few tens of micrometers) [25] to be able to translate the differences in phase propagation to contrast in the registered image [46]. Most studies with this technique have been conducted with parallel beam with monochromatic synchrotron radiation, but a number of more recent studies have been performed with cone-beam and partially coherent polychromatic x-rays from conventional x-ray sources [47].
 View larger version (32K) | Fig. 1 —Schematic shows experimental setup of propagation-based imaging technique (horizontal cross section, not scaled). |
In the past 5 years, several studies (Table 1) have evaluated the performance of PBI for breast cancer diagnosis [37, 39, 48–53]. These studies showed that use of PBI, particularly in conjunction with the application of phase-retrieval algorithms in the reconstruction of images, has resulted in images with high quality [48] and of high diagnostic value with a dose comparable to that of 2D planar radiography [37, 39, 50, 51]. The application of phase-retrieval algorithms was found to provide the best tradeoff between spatial resolution and noise on the output images [37, 48] and to increase radiologists' ability to correctly characterize benign and malignant breast abnormalities [53]. In a 2016 study, Bliznakova et al. [49] investigated PBI technique using new algorithms for volumetric breast reconstructions (tomosynthesis) and found that phase-contrast tomosynthesis preserved enhanced edges, especially in thick and inhomogeneous phantoms. However, the study was a phantom study, and the results may not perfectly represent human tissue. The first, to our knowledge, clinical trial of 2D phase-contrast mammography with synchrotron radiation for a group of patients [39] was conducted with PBI technique. The results showed that image quality could be higher than that of DM at medically acceptable radiation doses. This result is very promising for future clinical evaluations of PBI technique.
Analyzer-based imaging (ABI), also known as diffraction-enhanced imaging, is also a common phase-contrast technique in breast imaging research, but it requires a more complicated experimental setup than does PBI. As shown in Figure 2, the ABI setup typically consists of two perfect crystals: a monochromator positioned before the object and an analyzer placed between the sample and the detector [46, 54]. The analyzer is positioned at a particular angle that allows through only the rays that are incident on the crystal at or very closely to the Bragg angle [54]. Therefore, the emergent refracted and scattered waves from the object are filtered by the analyzer. The reflectivity of a crystal is a function of the deviation of the incident beam direction from the exact Bragg angle and is normally described by a graph called the rocking curve. This dependence makes the whole setup highly sensitive to small vibrations and other instabilities [25]. With ABI technique, the propagation phase contrast (as measured in PBI technique) generally cannot be neglected because the sample-to-detector distance cannot be zero [55].
 View larger version (33K) | Fig. 2 —Schematic shows experimental setup of analyzer-based imaging technique (vertical cross section, not scaled). |
Between 2012 and 2016, three journal articles [21, 37, 56] evaluated the characteristics of ABI technique in breast imaging. In two of the studies [21, 56] (Table 2), the investigators evaluated and optimized ABI and found that in comparison with absorption-based imaging (corresponding to conventional DM), ABI yielded images with higher quality and diagnostic values. Radiation dose was greater than for dual-view DM in one study [56], but the other study [21] showed that with an equally sloped tomography reconstruction method, ABI can significantly decrease the radiation dose (to a clinically acceptable range) and further improve the image quality. One study [37] also compared CT images from both PBI and ABI techniques with absorption-based images and showed similar improvement in image-quality characteristics in the two methods when phase retrieval was implemented (Table 1). Nevertheless, the authors argued that PBI should be preferred over ABI given its easier setup, implementation (no need for analyzer crystals), and application of phase retrieval.
Crystal interferometry is one of the oldest phase-contrast techniques, having been developed in the 1960s [57, 58]. The technique requires a monochromatic parallel x-ray beam [59]. The crystal interferometry setup comprises three splitters (perfect crystals) (Fig. 3). The first splitter is used to divide the beam coherently into two identical beams, and then a second parallel crystal causes the split beams to converge. The object is placed after the second splitter in the way of one of the two beams. The two wavefronts (the one transmitted through the object and the other) are recombined at a third parallel splitter crystal. The interference pattern between the unperturbed and the transmitted wavefronts at the third splitter is recorded at a detector, and phase shifts within the object can be measured [59]. Crystal interferometry technique is highly sensitive to minor phase shifts within the object. It is less commonly used in breast imaging studies, however, because of its small FOV and experimental difficulties, such as being highly sensitive to environmental disturbances [25], including any vibrations and small temperature changes.
 View larger version (52K) | Fig. 3 —Schematic shows experimental setup of crystal interferometry technique (vertical cross section, not scaled). |
The grating interferometry experimental setup basically consists of a highly spatially coherent x-ray beam and a pair of diffraction gratings (optical elements with a large number of pairs of regions with different x-ray transmission or refraction characteristics located next to each other) that are positioned between the object and the detector to generate phase shift information [60, 61] (Fig. 4). The first grating (phase grating) has negligible absorption and generates a periodic variation in the phase of the x-rays. The diffracted x-rays interfere downstream, reproducing a periodic phase pattern at so-called fractional Talbot distances, which are inversely proportional to the wavelength of x-rays [46]. The Talbot distances are also proportional to the square of the grating period. These two dependencies give Talbot distances in tens of centimeters. The second absorption grating (analyzer grating) is positioned at a fractional Talbot distance to convert the interference patterns created by the first grating into intensity modulations that can be recorded by the detector [61]. The conventional absorption contrast (attenuation), the differential phase contrast (phase shift), and the dark-field signal (small-angle scattering) can all be measured with this setup. Grating interferometry can also be used in conjunction with a low coherent x-ray beam (generated by conventional polychromatic x-ray tubes) by adding an additional grating between the source and the object [62–64].
 View larger version (45K) | Fig. 4 —Schematic shows experimental setup of grating interferometry technique (horizontal cross section, not scaled). |
In the last 5 years, most studies on implementation of phase-contrast technology in breast imaging have focused on grating interferometry technique [63, 65–77] (Table 3). Almost all grating interferometry studies have been performed with conventional x-ray tubes in the setup to exploit differential phase and dark-field signals; one grating interferometry tomosynthesis study [74] was performed with synchrotron radiation. Grating interferometry with compact laboratory setups or with integration of diffraction gratings into clinical mammography units has been successfully used in previous research. Results have shown that dark-field and phase-contrast images are superior to clinical DM and absorption-based CT images in terms of lower anatomic noise [68], reproduction of tissue fine structures [75] and tumor features [66], depiction of type I microcalcifications [69], and overall image quality [63, 67, 72, 73]. However, one major limitation of most of these studies [66, 68, 69, 72, 73, 75] was that radiation doses used in the experiments were far greater than the clinically acceptable level (in some cases more than 10 times as great). Other work also showed that obscured details and tumor structures on absorption-based images could be reliably made visible in two-directional [70] and tomosynthesis grating interferometry [74], but yet again, applied doses were very high. Low mechanical stability and long exposure times are also important issues in grating interferometry [63, 67].
Edge illumination is an innovative PCI technique that works with both synchrotron radiation and low coherent x-ray beams produced by conventional x-ray tubes. The setup of edge illumination technique comprises a pair of aperture systems (Fig. 5). The incident beam created by the source strikes on a presample aperture, which is close to the object. The narrow slits (a few micrometers) on the aperture collimate the beam and define multiple narrow beamlets before they pass through the object. A second aperture system is placed just before the detector in a way that only approximately one-half of each beam-let can hit the detector and the other half is absorbed by this aperture system [78]. In this way, the deflection of beamlets would be immediately detectable as the increase or decrease of the x-ray intensity passing through the adjacent opening of the detector aperture. In other words, at scanning, compared with nonrefracted x-rays, the x-rays refracted in the object change the proportion of detected photons at the detector [78]. In edge illumination, the absorption, the differential phase, and the dark-field (scatter) contrasts all can be quantified. Further details about this technique and its latest developments can be found in articles by Diemoz et al. [78] and Munro et al. [79].
 View larger version (49K) | Fig. 5 —Schematic shows experimental setup of edge illumination technique (horizontal cross section, not scaled). |
Literature from 2012 to 2016 (Table 4) shows that implementation of edge illumination technique for phase-contrast breast imaging could improve overall image quality compared with that of conventional DM when assessed both objectively and subjectively and while radiologic doses are kept at clinically acceptable levels [80, 81]. Olivo et al. [81] implemented edge illumination using a conventional x-ray tube and found improved image quality in terms of contrast and detection of microcalcification. However, the samples used in that study were thin, and the results may not be transferable to thicker samples. Edge illumination technique was also successfully implemented to reduce radiation dose (to as low as 0.12 mGy mean glandular dose) without sacrificing image quality [82]. Then again, the study was conducted in a synchrotron radiation facility, and the results may not necessarily be repeated with conventional x-ray tubes. In one study [80], the researchers conducted the experiment with a simplified setup in which no detector aperture was used but the detector was aligned so that edges of the pixels themselves created the insensitive area of the detector. All current studies have been proof-of-concept experiments. For clinical application, future studies on edge illumination technique should focus on shorter exposure time [81], larger FOV [80], and detectability of different types of lesions [82].
| Conclusion | |
|---|
The most important characteristics of the PCI techniques are summarized in Table 5. Although all current x-ray–based breast imaging modalities rely on differences in x-ray attenuation (absorption contrast) from various soft tissues, PCI also depicts variations in x-ray refraction (phase contrast). PCI can depict weakly absorbing details that are obscured on absorption-based images. Over the last 5 years, grating interferometry and PBI have attracted the largest number of publications among various techniques used to exploit phase-contrast information for breast imaging. Both methods have their strengths and limitations, and there are pending questions regarding their optimization for clinical trials. Edge illumination is a newer technique in phase-contrast imaging that has potential for further consideration for application in breast imaging.
Successful implementation of the current laboratory and preclinical results with PCI techniques may contribute to overcoming the limitations of existing modalities and lead to a new paradigm in x-ray breast imaging, especially if cancer visibility is improved at an absorbed breast radiation dose equal to or less than that of current mammography systems. In the next 5 years, the first synchrotron-based patient trials of phase-contrast CT are expected to take place. Before this happens, future studies must pay special attention to the diagnostic efficacy of PCI techniques, particularly in terms of sensitivity and specificity, because a paucity of studies have focused on these parameters. The complete translation of mammographic PBI to clinical use will depend on progress with microfocus x-ray source technology.
Supported by a grant from the National Breast Cancer Foundation (Australia).
| References | |
|---|
