British Journal of Radiology 75 (2002),170-173 © 2002 The British Institute of Radiology
Anomalous image quality phantom scores in magnification mammography: evidence of phase contrast enhancement
C J Kotre, PhD
I P Birch, MPhys
and
K J Robson, BSc
Regional Medical Physics Department, Newcastle General Hospital, Newcastle-upon-Tyne NE4 6BE, UK
 |
Abstract
|
|---|
Anomalously high image quality scores were noted for images of the Leeds TORMAM phantom obtained using magnification mammography. Comparison of optical density profiles of fibre features in the images with non-magnified images and images previously obtained using an in-line phase contrast geometry showed the presence of phase contrast enhancement in the magnification images. The effect on the phantom score is particularly marked for this design of phantom owing to its use of fibres, which tend to enhance well. A large proportion of the phantom score is associated with fibrous features. It is concluded that direct comparison of TORMAM phantom scores from magnified images with those from non-magnified images is not valid due to the different balance of physical mechanisms forming the two kinds of image.
|
Introduction
|
|---|
As part of the continuing effort to optimize screen–film mammography for breast screening, more attention has recently been paid to maintaining image quality for magnification views. Following discussion between physicists within the UK Breast Screening Programme, a test protocol based on the popular Leeds TORMAM image quality phantom [1] was trialled as part of the latest national survey of mammographic equipment performance in 2000. For magnification views, the non-anthropomorphic half of the phantom was imaged to fit within the restricted field of view. This section contains six groups of four fibres, each arranged in a crow's foot pattern with contrast decreasing in each group, six clusters of simulated microcalcifications with decreasing average size, and patterns of low contrast discs. The phantom is conventionally scored using a 3, 2, 1, 0 system, where 3 is clearly visible on imaging, 2 is faintly visible, 1 is barely visible/at threshold and 0 is not visible [1]. This phantom, although not supporting any quantitative analysis of image quality, has been found to be very useful in making comparative measurements and has formed the basis of image quality comparisons at national level [2].
Whilst scoring TORMAM phantom films taken using magnification geometry on a number of X-ray units, suspicion was aroused at the apparent large improvement in TORMAM score over the equivalent non-magnified phantom radiographs in each case. The improved scores resulted primarily from increased visibility of the heavily scored fibre features, and were very similar in pattern to results previously obtained on this design of phantom in experiments designed to investigate the effects of phase contrast image enhancement [3]. It was decided to investigate the possibility that these improved scores arose from phase contrast effects and did not therefore reflect a true comparison between image quality for magnification and conventional views.
|
Phase contrast image enhancement
|
|---|
As a spatially coherent X-ray beam propagates through an X-ray transparent medium, the phase of the incident wavefront becomes modified in a manner related to the electron density of the medium. The resulting phase gradient across the wavefront is equivalent to a change in direction of the propagation of the wave. The angular deflections from the initial direction of propagation are small but are most pronounced in regions of an object where the X-ray refractive index is varying rapidly, such as at the interface between two different materials. The direction of deflection will vary from point to point within a general object, depending on the structures present, but will produce a net effect of edge enhancement between structures of differing X-ray refractive index when imaged using an appropriate geometry. Smoothly curved structures such as spheres and cylinders show the effect particularly strongly as they act in a manner analogous to an optical lens.
X-ray phase contrast effects have previously been explored for applications in medical imaging using monochromatic synchrotron radiation and sophisticated X-ray optics. However, a simplified scheme for phase contrast imaging, based on a conventional microfocus X-ray source with high spatial (lateral) coherence but no chromatic coherence, has been proposed by Wilkins et al [4]. This "in-line" method depends on there being a degree of lateral coherence in the beam, quantified by the lateral coherence length d. This is given by 
where 
is wavelength, r1 is the source–object distance and f is focal spot size [4]. Thus, a large coherence length suitable for phase contrast imaging will be facilitated by the use of low energy photons, a small focal spot size and a large source–object distance. In order to record phase contrast information, a distance r2 between an object and the film (or other recording device) must be introduced. This distance allows the small angular deflections of X-ray photons produced at interfaces between materials of differing X-ray refractive index to be recorded superimposed on the intensity variations owing to photoelectric and scatter mechanisms. These conditions are met to some extent by the geometry used in conventional magnification mammography where the mean photon energy is low, r1 and r2 are typically around 30 cm in current designs and focal spot size is less than 0.15 mm on a measurement axis 2 cm inwards from the chest wall edge of the image [5].
The visual appearance of phase contrast enhancement in the final image is edge enhancement at interfaces between materials with differing X-ray refractive indices. Normally these materials will also have differences in attenuation coefficient, so the effect of the phase contrast is to increase the visibility of interfaces between materials.
|
Analysis of phantom images
|
|---|
10 pairs of magnified and non-magnified phantom films were available for analysis. All the images were produced using mammographic units employed within the NHS Breast Screening Programme, and with well controlled film processing. Three different manufacturers of X-ray unit and two types of screen–film system were covered in the sample. The films were scored by two experienced observers using the conventional 3, 2, 1, 0 system. In addition, optical density profiles across the highest contrast fibres in both the magnified and non-magnified images were produced using a Joyce Loebl 3CS scanning microdensitometer (Joyce Loebl Gateshead, UK). As the features being scanned were at low contrast (close to the threshold of visibility), an average of 25 scans carried out at different points along each fibre was taken to reduce quantum noise. For comparison, profiles were also produced for test films of the TORMAM phantom previously obtained using an in-line phase contrast geometry where r1=2 m, r2=0.2 m and focal spot size=0.7 mm x 0.3 mm [3].
|
Results
|
|---|
Table 1 shows TORMAM phantom scores for magnified and non-magnified images for each unit, averaged between the two observers. The split between the fibre, microcalcification and disc scores is also shown, as well as the value of lateral coherence length d calculated for each imaging geometry based on a mean photon energy of 20 keV. For comparison, an entry in the table is also given for a film from a phase contrast experiment previously reported [3]. The scores for magnification images are all significantly higher than those for equivalent non-magnified images on the same unit, with the main enhancement being for the fibrous features. The mean scores for magnification images are similar to those previously obtained using an in-line phase contrast geometry. The lateral coherence length for magnified and phase contrast cases are also similar. The lateral coherence length for non-magnified images is about 60% of that for magnified images, but virtually no phase contrast enhancement can take place for these images owing to the small distance r2 between the phantom and the plane of the image receptor.
View this table:
[in this window]
[in a new window]
|
Table 1. TORMAM image quality scores and lateral coherence length for non-magnified, magnified and phase contrast images
|
|
Figure 1 shows averaged fibre optical density profiles for a non-magnified image and a magnification image produced on the same unit and film–screen system, and a phase contrast film from previous experiments [3]. The profiles have been scaled in the x direction to fit on a common x-axis of distance across the fibre, rather than distance in the film, and inverted for ease of comparison. Despite the averaging used, the noise level is still significant compared with the very low contrast signal, but the general shapes of the curves can still be compared. The conventional non-magnified image displays the attenuation profile expected of a cylindrical object. The magnification and phase contrast geometry images both show evidence of undershoot spikes either side of the body of the fibre and the optical density profiles have a squared-off appearance. These features are consistent with the small angular photon deflections expected where phase contrast effects are contributing.
View larger version (19K):
[in this window]
[in a new window]
|
Figure 1. Optical density profiles on an arbitrary y-axis through fibres from Group 1 of the Leeds TORMAM test object for (upper curve) non-magnified image,(middle curve) magnification image and (lower curve) phase contrast geometry. The profiles have been scaled in the x direction to fit on a common x-axis of distance across the fibre, rather than distance in the film, and inverted for ease of comparison.
|
|
|
Conclusion
|
|---|
These results show evidence of a contribution from phase contrast to the high TORMAM phantom scores obtained for conventional magnification mammography. The contribution is particularly marked for this design of phantom owing to its use of fibres, which tend to enhance well, and the large proportion of the phantom score associated with the fibrous features. It is concluded that direct comparison of TORMAM phantom scores between magnified and non-magnified images is not valid owing to the different balance of physical mechanisms forming the two kinds of image. Although some phase contrast enhancement of clinical magnification images would also be anticipated, the size of the effect would be expected to be far smaller than that seen with the phantom owing to the greater range of structure size and shape in the clinical image. The effect of phase contrast enhancement on clinical images will be the subject of further study.
|
Acknowledgments
|
|---|
We would like to thank Dr N W Marshall and Mrs J Reay for acting as additional readers for the test phantom images.
Received for publication July 10, 2001.
Revision received November 12, 2001.
Accepted for publication November 15, 2001.
|
References
|
|---|
-
Cowen AR, Brettle DS, Coleman NJ, Parkin GJS. A preliminary investigation of the imaging performance of photostimulable phosphor computed radiography using a new design of mammographic quality control test object. Br J Radiol 1992;65:528–35.[Abstract/Free Full Text]
-
National Health Service Breast Screening Programme. Performance of mammographic equipment in the UK Breast Screening Programme in 1997. NHSBSP Publication No. 41. Sheffield: NHSBSP, 1998.
-
Kotre CJ, Birch IP. Phase contrast enhancement of x-ray mammography: a design study. Phys Med Biol 1999;44:2853–66.[Medline]
-
Wilkins SW, Gureyev TE, Gao D, Pogany A, Stevenson AW. Phase-contrast imaging using polychromatic hard x-rays. Nature 1996;384:335–8.
-
Institute of Physics and Engineering in Medicine. The commissioning and routine testing of mammographic x-ray systems. IPEM Report No. 59 (2nd edn). York: IPEM, 1994.
Top
Abstract
Introduction
