Research in grating-based differential phase-contrast imaging (DPCI) has gained increasing momentum in the past couple of years. The first results on the potential clinical benefits of the technique for X-ray mammography are becoming available and indicate improvements in terms of general image quality, the delineation of lesions versus the background tissue and the visibility of microcalcifications. In this paper, we investigate some aspects related to the technical feasibility of DPCI for human X-ray mammography. After a short introduction to state-of-the-art full-field digital mammography in terms of technical aspects as well as clinical aspects, we put together boundary conditions for DPCI. We then discuss the implications for system design in a comparative manner for systems with two-dimensional detectors versus slit-scanning systems, stating advantages and disadvantages of the two designs. Finally, focusing on a slit-scanning system, we outline a possible concept for phase acquisition.
With the remarkable progress reported recently in the development of novel methods and instrumentation, in particular in the domain of grating-based differential phase-contrast imaging (DPCI) [1–7], the question about the clinical applicability and usefulness received growing attention from both academia and industry. The potential of X-ray imaging techniques using phase effects when compared with attenuation effects in mammographic imaging has been demonstrated both ex vivo [8–12] and in vivo [13,14], mainly by using techniques called propagation-based phase-contrast imaging and analyser-based phase-contrast imaging using laboratory X-ray tubes. For detailed literature on the potential use of the various X-ray phase-contrast techniques focusing on mammography and on medical imaging, in general, we refer to two excellent recent topical reviews [15,16].
Recent investigations [12,17] in the domain of DPCI X-ray mammography hint at a potential benefit of DPCI when compared with conventional attenuation-based imaging in terms of general image quality, lesion delineation versus background tissue as well as the visibility of microcalcifications. The latter are known to play a role in the development of breast cancer. Early detection is of paramount importance for mammographic screening. From the perspective of technical feasibility, mammography seems to be a reasonable starting point to investigate the clinical potential of the technique, in particular compared with X-ray computed tomography in which higher design energies, but similar emphasis on a compact geometrical design, impose more stringent requirements on gratings pitch and geometrical alignment during the scan.
The focus in this paper is explicitly on mammography, full-field digital mammography (FFDM) to be more precise, and the potential of DPCI in that modality. The reason why we single out DPCI over techniques such as propagation-based phase contrast, where commercial systems are already available,1 or the coded-aperture approach  is that, in our opinion, DPCI provides the best possibility to separate the three physical effects of attenuation, phase change and small-angle scatter.
The paper is organized in the following way: in §2, we briefly summarize the state of the art in digital mammography today from both clinical and technical points of view. The main system parameters will be compared for five commercial FFDM systems, including four systems with two-dimensional detectors (two-dimensional detector systems in the following) and one system with a slit-scanning design.
In §3, we briefly present one result from a study on human mastectomy samples  demonstrating the potential of the technique in FFDM. This and more results from the same study provide the main motivation for our investigations.
In §4, we derive from those system specifications potential phase-contrast specifications. The implications for system design will be discussed in §5, comparing potential advantages and disadvantages of the two-dimensional detector systems versus slit scanning.
The introduction of differential phase contrast to conventional diagnostic imaging systems, be it mammography, interventional X-ray or computed tomography, will, in general, require additional hardware measures to stabilize the system during the scan and avoid disturbance from patient motion. From an engineering perspective, the most challenging aspect might be to guarantee sufficient mechanical stability for the phase acquisition.
Finally, we mention our present work on a concept for phase acquisition within a slit-scanning geometry. As we shall show, the inherent redundancy of a slit-scanning mammography acquisition can be used to eliminate the process of phase stepping altogether and offers the possibility to keep the relative positions of the source grating, phase grating and analyser grating fixed. Discussions and conclusions will be presented in §7.
2. State of the art in digital mammography
We focus exclusively on systems with digital detectors (FFDM) in this paper and disregard computed radiography or film–screen radiography systems. FFDM is a very mature modality with stringent performance criteria. We present a short summary in the following. For a detailed overview of FFDM, we refer the reader to [19,20].
Designated X-ray tubes for mammography applications are used in systems to illuminate a woman's breast, which is placed on a breast support and is compressed by means of a compression paddle that is lowered prior to the examination. The X-rays penetrate the breast tissue and are detected either by a two-dimensional planar detector or by an arrangement of line detectors, as used in the so-called slit-scanning technique. All state-of-the-art systems allow craniocaudal (CC) views as well as mediolateral–oblique (MLO) views to be acquired of both breasts by means of a rotating gantry arm, leading to a standard four images per examination. Depending on whether the scan is performed for the purpose of breast screening or diagnosis, important acquisition parameters will change, as will workflow, dose and scan time. In table 1, a number of system parameters relevant for our purpose are collected for five different commercial FFDM systems .
From table 1, it can be seen that all systems use similar source-to-detector distances of approximately 60–70 cm and support a field of view (FOV) of approximately 24 cm × 30 cm. Standard detector pixel sizes are less than 100 μm, a prerequisite to detect small microcalcifications. Scan times range from 0.5 to 5 s per image, and dose levels averaged over all breast thicknesses are typically less than 2 mGy (figure 1) [21,22].
Four of the five systems represented in table 1 use planar integrating detector technology with the active detector area covering the entire FOV. One system, the Philips MicroDose system, uses the slit-scanning approach, in which several parallel detector lines are scanned below the breast in a circular arc motion. An inherent advantage of this system is the low patient dose owing to the implemented photon-counting technology, and the improved scatter rejection via a set of pre- and post-patient collimators designed to collimate radiation onto the detector lines [23–25].
For planar two-dimensional detector systems, the task of reducing the impact of Compton scatter on image quality is accomplished by dedicated grids on top of the detectors. A disadvantage of the slit-scanning approach is the higher demand for tube power and anode heat capacity also owing to the collimation. A more detailed comparison of the advantages and disadvantages of the two system geometries is presented in §5.
3. First results on mastectomy sample scans
Recently, the potential clinical benefits of differential phase-contrast imaging as applied to human mammography were investigated in a collaborative effort between the Kantonsspital (KSB) in Baden, Switzerland, the Paul Scherrer Institute (PSI) in Villigen, Switzerland, and Philips. A total of 33 mastectomy samples from human subjects, many of whom presented with in situ ductal carcinoma or invasive breast cancer, were imaged using two different set-ups: one conventional digital mammography set-up at KSB, and one phase-contrast laboratory prototype at PSI. Six experienced radiologists were requested to compare the phase-contrast images with images taken with state-of-the-art mammography equipment and to compare phase-contrast images with absorption images taken using the same laboratory set-up at PSI.
As the two physical set-ups naturally featured different physical parameters such as spatial resolution or spectral quality, the first part of the study was dedicated to demonstrating that this had a minor impact on image quality comparison.
Another part of the study revealed statistically significant findings in favour of the novel phase-contrast technique (p<0.001) in terms of the following quality measures:
— improvement of general image quality (figure 2),
— improved delineation and sharpness of lesions,
— improved visualization of the image periphery,
— improvement in the general visibility of microcalcifications, and
— improvement in the sharpness of microcalcifications.
All results including the outcome of the detailed statistical analysis can be found in .
4. Boundary conditions for X-ray phase-contrast imaging
In the following, we will define the quantities and corresponding notations, as outlined in figure 3, which will be used throughout the rest of this paper. These are commonly used throughout the literature. The source grating, π-phase grating and analyser grating are denoted by G0, G1 and G2 and have periodicities p0, p1 and p2, respectively. For convenience, we put together the relations that hold between these periodicities and the distances and . The nth order Talbot distance Dn for a parallel beam geometry and the correction for cone beam geometry are given by 4.1In equation (4.1), n is the Talbot order, whereas λ denotes the X-ray wavelength. The condition of the beam divergence of a conventional X-ray tube leads to 4.2with η=2 for a phase grating inducing a phase shift of Δϕ=π at the design energy, the only case we consider here. Finally, the condition of the congruency of the interference fringes generated by the various line sources of G0 demands 4.3
(b) General considerations
It is important to note that the source-to-detector distance in FFDM is the result of an intricate compromise in system design influenced by many factors. For example, the typical scanning times in mammography vary between 0.5 and 5 s, depending on compressed breast thickness, tube current and voltage. Depending on the compression force and time, compression can cause significant discomfort and pain in women and influences compliance in screening programmes. Longer compression times will thus hardly be acceptable in a future phase-contrast mammography modality, and, thus, the scanning time cannot be prolonged in practice. This simple but important consideration puts a limit on increasing the source–detector distance by significant amounts, as the flux decreases with the inverse square of the latter distance, increasing scanning time quadratically for the task of applying a given dose to the breast. In addition, increasing the distance between the focal spot and the detector requires a higher gantry, which is again undesirable. Hence, from practical considerations, the gratings G0, G1 and G2 will all have to fit within about 70 cm of distance, with only limited flexibility towards longer geometries.
A typical-sized X-ray focal spot of a commercially available mammography tube measures 100 or 300 μm for the small spot and large spot settings, respectively (table 1). Even the small spot size is still too large to coherently illuminate the period of a reasonably sized G1 grating, which is typically of the order of 2–6 μm, the transverse coherence length for 25 keV radiation at a distance of 70 cm for the 100 μm focal spot being only about 350 nm. Gratings of that size appear challenging not only from a fabrication perspective but also from the practical obstacles of alignment precision and angular acceptance. Hence, either new tube technology must be devised  or a source grating G0 must be added .
The placement of the source grating G0 is limited by the tube housing, the exit window, filtration and the required space for the mount of G0 itself. For currently available tubes, this means that at least approximately 8 cm is required for the distance between the focal spot and G0, leaving about 62 cm of total length s=l+d of the DPCI mammographic set-up. With a fixed distance between G0 and G2, the need for another compromise arises: the placement of G1. Although, in principle, the placement of the breast in between the interferometer G1–G2 is possible, this would entail a linear gradient in phase sensitivity , and sensitivity to coherently scattered radiation away from G1 towards G2 is very undesirable for imaging of thicker breasts. In addition, compression would be challenging in such an arrangement.
A more practical solution would be to have the interferometer placed on top of the detector, with the breast support on top of G1. In this configuration, the breast is positioned higher than in conventional systems by at least the Talbot distance. Higher breast positioning entails larger detector areas (and consequently higher costs) as magnification increases and truncated projections must be excluded. One would thus prefer to keep the distance G1–G2 as small as possible. Small Talbot distances dn and/or low Talbot orders n, however, both imply a small pitch for G1 in order to achieve a certain phase sensitivity, which is challenging both for fabrication and for mechanical stability. The latter being one of the main challenges for a DPCI implementation for mammography altogether. A disadvantage of having G1 below the breast would be an additional, remnant attenuation caused by the phase grating, despite the fact that, by design, G1 is supposed to be purely phase shifting.
A positive aspect to be noted here is that any technological advancement regarding the pitch of the gratings G1 (and G2) would pay off quadratically in the reduction of space used for the interferometer.
In table 2, we put together four possible design choices, denoted scenarios S1–S4, according to equations (4.1)–(4.3) for the main geometrical parameters of the mammography system. A design energy of 25 keV was chosen for all four scenarios as this was considered reasonable for a DPCI mammography system with the question of the optimal spectrum still being investigated. We note that the optimal spectra for DPCI mammography are likely to be somewhat higher in kVp than for conventional mammography [29,30]. This is a direct consequence of the E−2 energy dependence of coherent scatter when compared with the E−3 energy dependence of the photoelectric effect. For a given imaging task, optimal kVp values will hence be somewhat lower for attenuation imaging than for phase-contrast imaging. Gratings fabrication, however, becomes more challenging for higher energies as visibility losses caused by insufficient stopping power of G2 are to be avoided by higher aspect ratios. To give an example, the transmission of 30 μm of gold is 21% at 30 keV, 48% at 40 keV and 66% at 50 keV.
All four cases share the same total system length of 70 cm (focal spot to G2). Only first Talbot orders were considered (n=1). A choice of a low Talbot order leads to a compact interferometer and a large dynamic range, whereas a high Talbot order improves phase sensitivity. Another important practical restriction is imposed by the fact that the distance from the breast support to the lower part of the detector housing should be as small as possible as the scanning of women with larger body mass index gets more difficult. In addition, some women need to be seated during the examination. Hence, the entire detection unit containing grids, collimators, the interferometer unit, the detector itself and additionally some actuators will have to be as slim as possible.
In summary, we note that a significantly longer geometry than that used in current FFDM systems will be difficult to justify. The restrictions on time for the whole examination and the compression are stringent. In a high-volume screening environment in Europe, a throughput of 5 min per woman is the upper limit. Regarding the compact design of the interferometer, it is important to recall that improvements in the fabrication of structures around 1 μm would help significantly in reducing the additional space required for the system G1–G2.
5. System design considerations: two-dimensional systems versus slit-scanning systems
Here, we consider options for the design of a human DPCI mammography system. We focus on the differences in the designs based on two-dimensional detector systems versus the slit-scanning acquisition and begin by outlining the basic conceptual differences between the two designs. In subsequent sections, we will consider the details of the various aspects, including acquisition mode, FOV, patient motion and others. We summarize our findings in a table at the end of the section.
(a) Two-dimensional systems versus slit-scanning systems: basic differences
Two-dimensional FFDM systems are equipped with two-dimensional, planar, pixellated area detectors covering the entire FOV with X-ray-sensitive material. The acquisition of a single mammographic projection can be obtained from a single shot of X-rays from the source within about a second.
The X-ray detectors used in slit-scanning systems do not cover the entire mammographic FOV at any moment in time (as opposed to two-dimensional detectors), so that, for the acquisition of a single, complete two-dimensional mammographic image, a scanning of the typically much smaller detector has to be accomplished. Possible variants of the technique involve a single two-dimensional detector of small width aligned perpendicular to the scan direction or a number of line detectors with multi-slit collimators again aligned perpendicular to the scanning motion.
A disadvantage for the slit-scanning systems is that the acquisition time per projection is limited by the scanning motion and is typically longer than in two-dimensional systems. However, the collimation of the radiation into narrow X-ray fan beams proves to be beneficial in terms of the mitigation of the degrading effects of scattered radiation. For a detailed description of the geometry and the physical characteristics of a slit-scanning mammography system, see ; for the efficiency of scatter rejection of such systems, see  and references therein.
(b) Acquisition mode
The way the phase acquisition is performed is very different between the two approaches. A two-dimensional detector system requires conventional phase stepping, typically eight shots with a stepping of one of the three gratings in between. The slit-scanning technique can make use of the redundancy of the data acquisition to acquire the phase. The details of this acquisition are explained in §6 and is similar to the technique described in . In short, different detector lines in the scanning approach can be equipped with interferometers with different phase shifts between the gratings, avoiding the stepping of any of G0, G1 or G2.
(c) Field of view, interferometer design and tiling
While the two-dimensional detector systems require a two-dimensional grating array covering the entire detector, the slit-scanning system uses gratings only to cover the slits, hence using less material. For a system with several line detectors, the gratings coverage can even be restricted to a number of one-dimensional lines. The most common wafer size for the production of the DPCI gratings is 4 inches. Even with 6- or 8-inch wafers, full coverage of a typical 24×30 cm FOV is impossible. Hence, in both system designs gratings need to be cut and tiled. Tiling is obviously more challenging for the two-dimensional detector systems owing to the larger detection area. In the case of slit scanning, tiling is only one dimensional, whereas two-dimensional tiles are required for two-dimensional detectors. Requirements for tiling accuracy are stringent, because a high image quality must be maintained across tiles, and gaps between two portions of the detector array must be avoided.2
(d) Scanning time
The scanning time for a single image (CC or MLO) for the two-dimensional detector systems is given by the number of phase steps multiplied by the sum of the stepping time and acquisition time per step. Using reasonably fast actuators for the gratings, the stepping time should be small compared with the acquisition time per step, so that the total scan time should not be significantly prolonged. In the scanning system, the time for the acquisition is not limited by the phase stepping but is unchanged with respect to the conventional scanning of the system as used today. The phase acquisition during slit scanning is illustrated in §6. The scanning times to acquire phase information in two-dimensional detector systems and scanning systems will be largely unaffected by the phase acquisition itself and, hence, essentially unchanged with respect to the FFDM standard. However, this conclusion assumes the availability of sufficiently powerful X-ray sources to compensate for the large attenuation by G0.
(e) Susceptibility to patient motion
Patient motion is problematic in at least two possible ways. In the first type, patient motion can impair the alignment of the gratings with respect to one another, with a disruptive effect on image quality. In the second type, the motion of the patient itself without compromising the alignment of the gratings leads to the wrong interpretation of the corresponding phase scans and in the sequel also causes artefacts. In the two-dimensional detector system, the phase acquisition is performed via phase stepping and motion artefacts will be apparent for pixels above which the object was moved during stepping or the alignment of the gratings was changed in addition to the stepping. In the slit-scanning approach, the object motion during the scan will be superimposed on the scan motion itself. Both systems will definitely be affected by patient motion, but in somewhat different ways.
(f) Phase sensitivity
For an arrangement with a planar detector and planar gratings, the fringe visibility deteriorates whenever X-rays impinge onto the gratings at an angle with respect to the gratings normal. The visibility degrades faster when the incident beam makes an angle with the trenches than when the incident beam lies within the planes of the trenches. Hence, for linear gratings, the two-dimensional detection systems will suffer from inhomogeneous visibility over the FOV, whereas the slit-scanning approach limits the loss in visibility owing to oblique incidence onto the trench planes by the limited spatial coverage of the detector in the scan direction. To avoid visibility inhomogeneities in the two-dimensional approach and extend the FOV with homogeneous phase sensitivity, bending of the gratings was considered  but appears challenging in practice. The loss of visibility owing to the oblique incidence of X-rays onto the gratings is illustrated in a simulation in figure 4. The degradation of visibility is caused by a number of effects, including the deviations of the gold profile of G2 from the ideal rectangular shape. The degradation increases with increasing distance from the optical axis and with increasing aspect ratio, with a complete loss of visibility predicted in the simulated scenario for gratings with 35 μm deep trenches and distances as small as 5 cm from the optical axis.
(g) Anode heat capacity: X-ray flux
As can be seen from table 1, the slit-scanning system is shipped with a tube with a heat capacity significantly larger than the corresponding figures for all the other systems. This reflects the rather poor usage of the total available X-ray flux owing to collimation. When the use of a G0 grating is considered to generate sufficient coherency by collimation, an even larger fraction of photons is blocked. Typical duty cycles of 20–30% of G0 gratings reduce the available X-ray flux for both system designs in an equal manner by factors (!) of 3–5. This is a quite remarkable change and puts heavy pressure on the X-ray source. The situation will be worse for the slit-scanning approach owing to longer acquisition times. New ways of generating coherent beams of X-rays might have to be considered that use the bremsstrahlungs spectra more efficiently.
(h) Radiation dose
Typical radiation doses in FFDM mammography today are less than 2 mGy; for smaller breast thicknesses and low-dose systems, they are less than 1 mGy (figure 1). A fair comparison of radiation dose between conventional attenuation and phase-contrast imaging must take into account the losses by the post-patient gratings, i.e. G1 and G2. Although the largest concern in terms of dose results from the typical 50% duty cycle of G2, the attenuation by G1 must also be kept at an absolute minimum. This can be achieved by thinning down the wafers after processing; however, this is at the price of more delicate mounting. Owing to the relatively low dose acquisition of a slit-scanning system, the dose argument related to the attenuation of G2 might be less significant when compared with two-dimensional detector systems. To obtain the same quantum noise level in the attenuation image derived from a DPCI mammography scan as in the attenuation image from a conventional (FFDM) acquisition without gratings, twice the dose has to be used (for a 50% duty cycle of G2). It needs to be shown that the additional benefits of the differential phase and the scattering information outweigh the increased noise in the attenuation image. Radiation doses at twice the current levels of 1–2 mGy will hardly be acceptable in clinical practice, in particular in mammography screening.
(i) Summary of design aspects
A short summary of the various advantages and disadvantages of the two approaches is given in table 3.
6. Slit-scanning phase-contrast acquisition
In figure 5, we sketch a possible system design for phase acquisition in a slit-scanning phase-contrast mammography system. In a conventional system based on slit scanning, several detector lines are scanned below the static breast so that, after complete motion over the entire FOV, the successive readouts of the detectors during the scan can be used to compose a single two-dimensional image. In this process, the redundancy of the data acquisition needs to be taken into account, i.e. averages over data from detector pixels corresponding to one and the same geometrical ray acquired at different instances in time must be taken. The main idea to incorporate the phase acquisition into the scanning motion stems from this inherent redundancy in the slit-scanning acquisition, which is absent in two-dimensional detector systems.
The idea is an extension of the method described in  to the case of a cone beam geometry. In the method of Kottler et al. , a parallel beam geometry is used in order to sample one and the same geometrical ray through the object with a number of N different line detectors and grating interferometers by a linear translation of the object with respect to the detectors. The system is designed such that the lateral shifts of G1 with respect to G2 in two adjacent interferometers increases by p2/N. In this way, the subsequent sampling in time of one and the same ray by the N detector lines can be interpreted as the stepping of the phase for that ray.
Strictly speaking, the translation method works only in parallel beam geometry, as, for a source at a finite distance from the object, a linear translation will cause the different detector lines to acquire signals along slightly different rays. The extension we propose makes use of a rotational movement, as shown in figure 5, where it is irrelevant whether the object is being rotated or the scan arm as long as the rotation axis passes through the focal spot perpendicular to the plane of the drawing. Similar to the idea of Kottler et al., the interferometers are detuned by a slight misalignment in height Δl of the grating G0, generating a Moiré pattern. The pattern shown in this figure is schematic and corresponds to the case when the object is absent. The pattern is indicative of the mismatch between the periodicities of the interference fringes generated at G2 and the pitch p2. The idea is to adjust this mismatch Δl such that the entire detector width spans an integer multiple of the Moiré period. In the figure, the Moiré periodicity equals the detector width in the scan direction so that eight sampling points for the phase are provided. Deviations from the reference pattern without an object measured during the object scan will be indicative of the phase changes caused by the object.
Figure 5 shows a cylindrical volume that is passed by two different X-ray beams at different moments in time. Knowing the local lateral shifts of G1 with respect to G2, these two measurements can be interpreted as two phase steps of a stepping sequence. Note that, as in , the two measurements are taken with different X-ray detectors, which requires careful sensitivity corrections for the various detector lines prior to phase retrieval. Obviously, in the slit-scanning approach to FFDM, the circular arc motion is performed by the scan arm while the object remains static, contrary to the situation depicted in figure 5 where the breast is rotated for the purpose of illustration.
A picture of a recently built interferometer unit is shown in figure 6 with three sets of gratings mounted on a rigid Invar unit. A corresponding Moiré pattern taken with a Shad-o-box detector from Rad-icon is shown in figure 7. Visibility measurements using the interferometer shown in figure 6 yielded values of 38%, 31% and 21% for tube voltages set on a mammographic X-ray source of 26, 32 and 38 kVp, respectively, after attenuation by 17 mm of a planar polycarbonate object. The proposed method has the advantage that, for the phase acquisition itself, no actuator control is required. The correct alignment of the interferometer needs to be guaranteed by an upfront calibration procedure. During the scan, despite rigid mounts of the G0 unit and the interferometer unit onto the same arc, remaining relative positioning inaccuracies will occur, probably induced by vibrational eigenmodes of the system, in particular the scan arm. Owing to the rigid interferometer design, however, the positioning accuracy requirements will be relaxed and determined by p0 and not p1 and p2. One of the critical modes of system vibration is the relative lateral motion of G0 at the source versus G1–G2 at the detector driven by the scan arm motion. Using a simple, intuitive argument by limiting the acceptable vibrational amplitudes of this mode to about p0/16, we obtain upper limits for these amplitudes of 3.8, 2.3, 1.9 and 1.5 μm for the four scenarios discussed in §4b, respectively.
We presented an overview of boundary conditions and relations relevant for the design of a differential phase-contrast mammography system based on state-of-the-art FFDM equipment. The survey of FFDM systems showed that many clinical boundary conditions, such as throughput, scanning time, compression time and dose, affect the design choices at least equally strongly as purely technical aspects, such as system length, design energy and gratings pitch. As scan times and compression times must remain low and tube power is limited, the total system length cannot be increased by substantial amounts. Moreover, as the breast ideally is right in front of the detector, a most favourable situation would have the interferometer unit as compact and thin as possible. Challenging as the reduction in gratings periodicities might be, any advance pays off quadratically in a reduction of the required space for the interferometer.
The comparison between the two alternative solutions for designing a DPCI mammography system—two-dimensional detector systems and the slit-scanning technique—revealed advantages and disadvantages for both systems with a slight preference noted for the slit-scanning approach by a simple count of the pros and cons listed in table 3. A major concern for this approach is tube power and scanning time, a condition aggravated by the addition of a source unit G0 further reducing the amount of radiation used for imaging.
In conclusion, many obstacles remain to be overcome when trying to build a functioning DPCI mammography system. The most challenging hurdles include limited X-ray flux of conventional mammographic sources as well as stringent stability requirements during the scan. If progress in gratings fabrication continues at the pace of the past, then analyser gratings for mammographic design energies with a pitch of p2=1 μm or below might become available and would help to limit the space taken by the interferometer to a few centimetres.
We gratefully acknowledge countless discussions with Prof. Marco Stampanoni, Dr Zhentian Wang, both from PSI, Switzerland, and Dr Nataly Wieberneit, Ingrid-Schulze-Wenck and Hanns-Ingo Maack from Philips Healthcare, Diagnostic X-ray, Hamburg, Germany.
One contribution of 16 to a Discussion Meeting Issue ‘Taking X-ray phase contrast imaging into mainstream applications’ and its satellite workshop ‘Real and reciprocal space X-ray imaging’.
↵1 Konica-Minolta, Regius Pureview. See http://www.konicaminolta.com/healthcare/technology/phasecontrast/.
↵2 In the slit-scanning design, tiling can be avoided altogether as there are already gaps between the various sensors within one detection line. Their effects can be compensated by lateral offsets of the sensors from one detector line to the next.
- © 2014 The Author(s) Published by the Royal Society. All rights reserved.