British Journal of Radiology (2006) 79, 239-243
© 2006 British Institute of Radiology
doi: 10.1259/bjr/24723806
Trends in image quality in high magnification digital specimen cabinet radiography
I P Birch, MSci, MPhys,
C J Kotre, PhD and
R Padgett, PhD
Regional Medical Physics Department, Newcastle General Hospital, Westgate Road, Newcastle NE4 6BE, UK
 |
Abstract
|
|---|
Advances in microfocus X-ray tube design together with the availability of high resolution charge coupled device (CCD) detectors have led to the introduction of high magnification digital specimen cabinets for the examination of tissue samples. This paper explores the effect that the high magnification geometry permitted by such units has upon image quality in terms of phase contrast edge enhancement, spatial resolution and the appearance of test phantom images. Phase contrast effects and spatial resolution were studied using a previously established method (using edge profiles) and by computing the system spatial frequency response at various geometries. It was demonstrated that the magnitude of the phase contrast enhancement effect reaches a stable maximum at a magnification of x 4. It has also been shown that a continual increase in both the spatial resolution together with an improved signal to noise ratio occurs up to the maximum permissible magnification geometry, with effects of focal spot blur being negligible. In practice, the limited size of the digital detector and the difficulty of object alignment can constrain the use of the very high magnification option.
|
Introduction
|
|---|
Radiography of excised tissue samples is usually carried out in specialized specimen cabinets. These units commonly feature a focal spot size of approximately 0.05 mm, a film–focus distance of 50 cm and contain movable shelves so that the distance between the sample and the image receptor can be varied to provide a geometric magnification up to x 1.8. Low tube currents are used, and low tube voltages in the region of 20 kVp maximize contrast. A recently introduced model, the MX20 (Faxitron, Wheeling, USA) features a nominal focal spot size of only 0.02 mm, a receptor–focus distance of 58 cm, magnification geometry of up to x 5 and a digital receptor consisting of a 5 cm x 5 cm charge coupled device (CCD) array with 1024 x 1024 pixels. The aim of this paper is to investigate the image quality trends with varying geometrical magnification on this unit in terms of spatial resolution and signal to noise ratio (SNR). In particular, the contribution of phase-contrast information is assessed.
|
Phase contrast
|
|---|
Phase-contrast enhancement occurs at interfaces between materials of differing X-ray refractive index. 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 the object where the X-ray refractive index is varying rapidly, such as the interface between two different materials. The direction of the deflection will vary from point to point within a general object, depending on the structures present, but produces 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 [1].
Although phase-contrast imaging is frequently associated with the use of monochromatic synchrotron radiation [2], a simplified scheme based on conventional microfocus X-ray tubes, with high spatial (lateral) coherence, has been demonstrated [3, 4]. The lateral coherence is enhanced by the use of low energy photons, a small focal spot size and/or a large source–object distance; many of these conditions are met by the geometry used in specimen cabinet radiography.
The visual appearance of phase contrast enhancement in the final image is edge enhancement at interfaces between materials with differing X-ray refractive indices. As there is also a change of X-ray attenuation across these interfaces, the effect of the phase contrast is to provide a subtle enhancement of the conventional attenuation image.
|
Parameters under investigation and experimental techniques
|
|---|
The investigation of the image quality trends in magnification radiography took place on a Faxitron MX20 microfocus specimen cabinet utilizing a 5 cm x 5 cm CCD detector with a 50 µm pixel pitch. The recorded pixel values from the detector were initially verified to be linear with dose using an aluminium step wedge. The focal spot was measured by the slit method as 0.02 mm x 0.02 mm and all experiments were performed at a nominal 20 kV and 0.3 mA. Some comparative measurements also took place on a film–screen Micro50 specimen cabinet (measured focal spot of 0.08 mm x 0.11 mm) using Kodak MinR2000 film and screens.
Phase contrast detection
Phase contrast enhancement was demonstrated from the imaged profiles of a low attenuation edge test object where the magnitude of the phase signal is comparable with that of the attenuation signal. A simple phase contrast test object was constructed from the edge of a standard radiography film (approximate thickness 180 µm). Thin aluminium foil (50 µm) was used to create a "non-phase contrast" edge of similar linear attenuation properties. Thin aluminium edges have been shown not to produce measurable phase enhancement effects due to the small phase signal being swamped by the larger attenuation signal [4].
Both edges were imaged at all available magnifications ( x 1, x 1.5, x 2, x 3, x 4 and x 5). In each case, the test edges were rotated by approximately 30° to the coordinate system of the CCD pixel array to allow oversampling of the edge profiles.
The phase contrast enhancement effects were further analysed using the pre-sampled modulation transform function (MTF) calculated using data from the edge profiles. By comparing the frequency response of the phase contrast edge with that of the non-phase contrast edge (which yields the conventional MTF), the effect of the contrast enhancement was quantified in frequency space.
Spatial resolution/geometric blurring
An inherent limitation of all forms of magnification radiography is the finite size of the X-ray focus, causing geometric blurring of an imaged object edge. When using the 50 µm focus, this blurring limits specimen cabinet radiography to approximately magnification x 2, after which blurring becomes unacceptable.
For digital radiography systems the spatial resolution is also limited by the Nyquist frequency of the detector defined by (2p)–1 where p is the pixel size. The theoretical maximum spatial resolution in the image plane for the MX20 system using a 50 µm pixel detector therefore is 10 cycles mm–1. As this value is low compared with that for film/screen, where over 20 line pairs mm–1 is more typical, the performance of the digital detector, in terms of limiting spatial resolution for specimen assessment, was also investigated.
The limiting spatial resolution for each magnification geometry ( x 1 to x 5) was assessed by two methods; by the 5% MTF cut-off frequency (cycles per mm), and with a Huttner line-pairs test object (Type 25a) orientated at 45° to the pixel coordinate system (line pairs per mm).
Visual appearance
The perceived SNR was visually evaluated using the Leeds TOR(MAM) phantom which is usually associated with the performance testing of mammography equipment. The phantom contains three groups of test objects: fibres, simulated microcalcification clusters and low contrast plastic discs [5].
Specimen cabinets are often used for evaluation of mammography core samples that may contain small calcification clusters associated with developing cancers. For this reason the microcalcification clusters in the TOR(MAM) phantom were used to assess the overall image quality.
The simulated microcalcification clusters in the phantom were imaged at all magnification geometries. The digital images were then rescaled (with no pixel interpolation) and windowed so that the features in each image appeared at the same size and grey level. The visual appearance of the microcalcification clusters was assessed on a standard computer monitor.
|
Results
|
|---|
Phase contrast enhancement
Figure 1
demonstrates the averaged edge profiles (pixel values) for the film and aluminium edge test objects. The distance across each edge (the x-axis) has been rescaled to account for the oversampling angle of the edge profiles. The profiles of the film edge in Figure 1a demonstrate "overshoots" that become more apparent with increasing image magnification. This is the characteristic appearance of phase contrast for this type of object [3]. The gradient of each profile also appears to increase slightly with magnification. This occurrence suggests that the phase contrast enhancement serves to counteract geometrical blurring effects.
View larger version (25K):
[in this window]
[in a new window]
|
Figure 1. (a) Pixel value profiles across image edge acquired at various magnification geometries using a phase contrast test object. (b) Pixel value profiles across image edge acquired at various magnification geometries using a non-phase contrast test object.
|
|
Figure 1b
shows that the aluminium edge profiles have no phase contrast overshoots. In addition, all these profiles are comparable for each of the magnification geometries used. This suggests that effects from geometrical blurring are small, meaning that maximum magnifications can be used for all object types, with little detectable image degradation in image spatial resolution.
The frequency response curves of Figure 2a show that phase contrast effects preferentially enhance the mid spatial frequency range for magnification geometries whilst the overall calculated limiting spatial resolution (taken as the 5% level) is left relatively unchanged at 9 cycles mm–1. For the non-phase contrast object it is seen from Figure 2b that the shape of the frequency response curve is consistent for all magnification geometries from x 1 to x 4 with magnification x 5 slightly lower, especially at the high frequency end. We draw two conclusions from this observation. First, it further demonstrates that the mid-frequency enhancement shown in Figure 2a is a true phase contrast effect and does not occur as a consequence of the magnification geometry or changes in signal to noise ratio. Second, geometrical unsharpness due to focal spot blurring is minimal up to magnification x 4 but there is some degree of blurring at magnification x 5.
View larger version (20K):
[in this window]
[in a new window]
|
Figure 2. (a) Image plane system frequency response curves (MTFs) calculated from edge profiles acquired at various magnification geometries using a phase contrast test object. (b) Image plane system frequency response curves (MTFs) calculated from edge profiles acquired at various magnification geometries using a non-phase contrast test object.
|
|
In addition, Figure 2a shows that there is little difference in mid-frequency response between the geometries of x 4 and x 5 magnification. Therefore, between these two geometries there is little to be gained from additional phase contrast signal. This is likely to be caused by the drop in lateral coherence of the polychromic X-ray beam; a consequence of shortened focus to object distance. At magnification x 4 the geometry appears optimized between the amount of phase contrast created (focus to object distance) and the capability of the detector to record the small angular phase contrast deflections (object to detector distance).
Spatial resolution/geometrical blurring
To calculate the 5% MTF cut-off frequency in the object plane (consistent with the Huttner test object) the spatial frequency axis of the frequency response curves in Figure 2b was rescaled to correct for the magnification effect. These results together with those from the Huttner test object are given in Table 1. This table shows that despite the spatial resolution of the digital detector being constrained to 10 lp mm–1, much higher object plane resolution is possible through image magnification. Note that the maximum spatial frequency measurable with the Huttner test object is 20 lp mm–1 and this was reached by magnification x 3 geometry.
View this table:
[in this window]
[in a new window]
|
Table 1. Limited spatial resolution measurements using the Huttner test object and frequency response method for digital MX20 and conventional Micro50 units
|
|
Figure 3 shows rescaled images of the 14th Huttner group (16.6 lp mm–1) for each of the 6 magnification geometries. These images demonstrate a continual improvement of image overall sharpness from magnification x 1 to magnification x 5, consistent with the estimated spatial resolution from the MTF assessment.
View larger version (60K):
[in this window]
[in a new window]
|
Figure 3. 14th Group of Huttner(Type 25a) spatial resolution test object. Images acquired at various magnifications and rescaled. (a) mag x 1, (b) mag x 1.5, (c) mag x 2, (3) mag x 3, (e) mag x 4, (f) mag x 5.
|
|
Visual appearance
Images of the 5th group (group "E", size 90–141 µm [5]) of the simulated microcalcification clusters for the TOR(MAM) phantom are presented in Figure 4. The images have been rescaled for magnification to represent them as if in the object plane (i.e. to display all features at the same size). An improvement of overall detail detectability is seen with increasing magnification.
View larger version (158K):
[in this window]
[in a new window]
|
Figure 4. Microcalcification cluster number 5 of Leeds TOR(MAM) mammography test object. Images acquired at various magnifications and rescaled. (a) mag x 1, (b) mag x 1.5, (c) mag x 2, (3) mag x 3, (e) mag x 4, (f) mag x 5.
|
|
The standard deviation in pixel value in the uniform background region was measured to be approximately equal for each image. This result is expected as the photon flux at the detector for a fixed exposure is independent of the magnification geometry selected. Increasing magnification with constant exposure factors would, however, be expected to increase the amplitude of large area signals in the plane of the object, and therefore improve the SNR in the image, due to the increased number of photons per unit area in the object plane. At high magnification more photons will interact with any given object feature, resulting in a larger difference in total number of photons recorded due to the presence of that feature. By simple geometry, the photon flux will increase as the square of the magnification factor. If quantum noise is considered to be the dominant noise source, then the SNR for large area objects would be expected to increase approximately in proportion to the magnification factor. In addition, for small objects comparable in size with the system point spread function such as the microcalcifications in Figure 4, increasing magnification will increase the size of the object projected at the plane of the detector, shifting the spatial frequencies down the system MTF (Figure 2b) so that they are imaged at a larger signal amplitude.
|
Discussion
|
|---|
The results above demonstrate that significant improvements in the overall image quality of specimen cabinet radiography can be achieved when using the high magnification geometry available with the digital Faxitron MX20 unit. We have shown that the 20 µm focus permits up to x 5 magnification with no demonstrable loss in spatial resolution in the image plane. This is compounded with the result of continual increases in resolution in the object plane with increasing magnification and a predicted maximum object resolution of 40 lp mm–1 at magnification x 5.
An improvement of SNR and spatial resolution has also been shown to occur at high geometric magnification, with the increased visibility of small, low contrast objects in the TOR(MAM) phantom. However, improvement occurs at the expense of a much reduced field size (in the object plane).
An interesting result from this study is the fact the Faxitron MX20 unit produces phase contrast image enhancement at mid to high magnification geometries. These enhancement effects produced an improvement in the mid-frequency response. The overall effect of phase contrast enhancement on clinical images will depend entirely on the object being imaged, but the most noticeable effects should be seen in the visibility of filamentous and spherical objects and at interfaces between materials of similar attenuation contrast [3].
Although we have demonstrated that a phase contrast contribution is present, the improvement in image quality is mainly governed by the increase in SNR and object plane resolution produced at high magnification geometries.
|
Conclusions
|
|---|
The results above suggest that, for a modern digital specimen cabinet, with focal spot sizes in the order of 0.02 mm, image quality in terms of spatial resolution and SNR in the object plane can be maximized by the use of the highest magnification factor ( x 5 in this case). Phase contrast is also produced at high magnification geometries with x 4 magnification producing the optimum results. However, it is appreciated that in practice, the limited size of the digital detector and the difficulty of object alignment may constrain the use of these very high magnification options.
|
Acknowledgments
|
|---|
We would like to thank staff at the Breast Screening Unit at Queen Elizabeth Hospital, Gateshead for their help with this study.
Received for publication March 23, 2005.
Revision received June 8, 2005.
Accepted for publication July 4, 2005.
|
References
|
|---|
- Ingal V, Beliaevskaya E, Brianskaya A, Merkurieva R. Phase mammography – a new technique for breast investigation. Phys Med Biol 1998;43:2555–67.[Medline]
- Lewis R. Medical applications of synchrotron radiation x-rays. Phys Med Biol 1997;42:1213–43.[CrossRef][Medline]
- Wilkins SW, Gureyev TE, Gao D, Pogany A, Stevenson AW. Phase-contrast imaging using polychromatic hard x-rays. Nature 1996;384:335–8.[CrossRef]
- Kotre CJ, Birch IP. Phase contrast enhancement of x-ray mammography: a design study. Phys Med Biol 1999;44:2853–66.[Medline]
- Cowen AR, Brettle DS, Coleman NJ, Parkin GJS. A preliminary investigation of the imaging performance of photostimuable phosphor computed radiography using a new mammographic quality control test object. Br J Radiol 1992;62:528–35.
Top
Abstract
Introduction
