British Journal of Radiology (2005) 78, 746-748
© 2005 British Institute of Radiology
doi: 10.1259/bjr/11930472
A printed image quality test phantom for mammography
C J Kotre, PhD1 and
D J T Porter, MSc2
1 Regional Medical Physics Department, Newcastle General Hospital, Newcastle upon Tyne NE4 6BE and 2 Department of Medical Physics & Bioengineering, Raigmore Hospital, Inverness IV2 3UJ, UK
 |
Abstract
|
|---|
This communication describes a novel design for a mammographic image quality test phantom, the final design of which was produced as a radiographer weekly quality assurance phantom for breast screening and symptomatic mammography. The phantom is based on low contrast test features which are built up by superimposing sheets of Mylar overhead projector transparency, on which the test features are printed using a standard LaserJet printer. The required radiation contrast at mammographic energies is produced by the approximately 50% by weight component of iron oxide (Fe3O4) present in the toner. An easily replicated design of mammographic image quality phantom based on LaserJet printed test features is described. Approximately 40 of these phantoms were constructed, and these have been used successfully for 5 years in both breast screening and symptomatic mammography. The phantom design offers a performance similar to much more expensive mammographic contrast-detail phantoms, but suffers from the disadvantage that high contrast resolution bar patterns cannot be produced using the standard printing process.
|
Introduction
|
|---|
This communication describes a novel design of mammographic image quality test phantom, the final design of which was produced as a radiographer weekly image quality assurance (QA) phantom for the breast screening and symptomatic services in the North of England. The design and method of production have not previously been published as they were subject to potential commercial interests, but this restriction no longer applies.
The phantom is based on low contrast test features which are built up by superimposing sheets of Mylar overhead projector transparency on which the test features are printed using a standard LaserJet printer, in this case a LaserJet 4M (Hewlett-Packard Corporation, Palo Alto, CA). The required radiation contrast at mammographic energies is produced by the approximately 50% by weight component of iron oxide (Fe3O4) present in the toner which is used to transport the toner magnetically before it is deposited electrostatically. Theodorakou et al [1] have recently described a related method for the production of radiographic test objects using drawing software and a modified inkjet printer printing potassium iodide solution.
|
The production of printed quantitative test features
|
|---|
Initial experiments with LaserJet printing to produce low-contrast test features established that multiple prints on the same Mylar sheet did not produce reproducible results, indicating that multiple superimposed sheets would be required in the final design. Initial mammographic images of solid-fill printed disks highlighted two further problems. First, the contrast steps produced by even a single layer of solid-fill black were too large, so that the threshold contrast was exceeded immediately, i.e. even a single layer was discernable. Second, an edge effect was noted for larger disks. Although not visually apparent on the printed disks, this was prominent on the mammograms as an enhanced contrast border to the larger disks. This was clearly not acceptable for a contrast-detail phantom as it would interfere with the detection process, alerting the observer to the presence of the edge instead of responding to the perceived signal-to-noise ratio of the whole disk [2]. The LaserJet process uses a "jumping system" where the toner is developed across a gap in response to the electric field produced by the latent image on the photoreceptor. It therefore comes under the influence of fringe fields, which occur at the boundaries between charged and uncharged areas of the photoreceptor leading to the edge enhancement effect. No edge enhancement was seen for disks below approximately 1.6 mm diameter as the spatial extent of these small disks was smaller than the range of the edge effect.
The solution to both of these problems was to use fractional fill patterns. The printer control language (PCL) allows dot-by-dot definition of fill patterns for the disks. At 600 dots per inch, each dot is approximately 42 µm in diameter. Provided the toner deposition pattern is fine enough, a uniform disk contrast should be perceived on the mammographic image, subsequent to the spatial averaging produced by the screen–film and observer visual unsharpness. Following experimentation with a number of fill patterns, it was established that a single-dot quarter fill gave a consistent increase in imaged contrast as the number of superimposed printed layers was increased. It was therefore decided to base the final design on combinations of quarter fill layers together with full-fill layers for disks smaller than 1.6 mm in diameter.
In order to establish quantitative contrasts for the test features, the mass per unit area for the quarter and full-fill patterns was established by accurate weighing. The effective thickness of each toner component was then derived from knowledge of the contribution by weight and density of each component. These thicknesses were then used to calculate the expected radiographic contrast for multiple layers, using the assumption that, as the dot pattern is not resolved by the imaging system, it can be treated for contrast calculation purposes as if the same quantity of toner was distributed evenly over the area of the disk. Air kerma contrasts were calculated using a variant of the beam spectral simulation code of Birch and Marshall [3] with modifications for the incident mammographic spectrum and inclusion of the attenuation characteristics of the toner components. The calculated toner attenuation was verified in narrow-beam, low scatter transmission experiments on multiple thicknesses of toner in the two fill patterns.
|
Phantom design
|
|---|
The phantom is of a planar threshold contrast-detail diameter type, consisting of two 5 mm thick sheets of opaque Perspex with the low contrast test features sandwiched between them. It is designed to be imaged at 28 kVp using a molybdenum anode and molybdenum filter on top of a 3 cm stack of Perspex under automatic exposure control using a standard image quality set-up for these tests. A radiograph of the phantom is reproduced in Figure 1
. The low contrast test features are produced by superimposing six sheets of Mylar overhead projector transparency on which the test features are printed.
View larger version (104K):
[in this window]
[in a new window]
|
Figure 1. Radiograph of the mammographic phantom placed on 3 cm of Perpex, taken at 28 kVp under automatic exposure control. Reproduction of the linepair groups has been degraded by digitization of the original radiograph.
|
|
The phantom design was based on the requirement to produce a pass/fail image quality test conforming to the image quality standards laid down for the National Health Service Breast Screening Programme (NHSBSP). The minimum standard (presently under revision) was defined in terms of threshold contrast at two disk diameters: 1% for 5–6 mm disks and 5% for 0.5 mm disks [4]. A system operating above this standard should produce lower threshold contrasts, i.e. more disks observed. Using only two disk diameters can produce a poor test dominated by observer variation, so we wished to produce a larger range of disks with subtle contrast steps so that a full threshold contrast-detail diameter curve could be produced. The final design (Figure 1
) was based on seven disk diameters, ranging from 0.25 mm to 6 mm, with up to 12 steps in contrast for each. This was implemented with a superimposition of six printed Mylar sheets using full and quarter fill patterns as described previously.
An attempt was made to produce a low-contrast resolution pattern using full fill and three levels of contrast. Although it was possible to fabricate these bar patterns up to a minimum spacing corresponding to 3.3 line-pairs per mm, these were found to be less useful in practice than the contrast-detail test as the results could not be related to published standards. In order to provide a high-contrast resolution bar pattern capable of testing to NHSBSP standards, it would have been necessary to incorporate a commercial test pattern based on lead foil, which would have increased the cost of each phantom by approximately 100-fold.
Early trials with the design produced the criticism from users that the disks were all of such low contrast that it took a long time for the observer to locate the start of each row and count accurately. Circles around the test disks were therefore added. This improved user acceptability, although more recent (unpublished) work we have done on the detection of low contrast objects in a cluttered background suggests that the rings for the largest (6 mm diameter) disks are too small, limiting the observer's search area and artificially elevating the threshold contrast for this diameter. In use, a trend was indeed noticed for the 6 mm diameter disks to be slightly underscored, but this did not significantly affect the pass/fail results from the phantom as these are based on a curve combining scores for all the disk diameters.
Figure 2 shows the test results sheet produced to accompany the phantom. The user counts the number of disks observed at each size, then marks the appropriate circle. The contrast-detail curve is then drawn as a smooth curve through the marked circles and compared with the dashed curve, which is a fit to the NHSPSP standard. The shape of this fit is based on the average of many mammographic contrast-detail curves measured using the CDMAM contrast-detail phantom [5]. If the resulting curve lies above the dashed curve then the image quality is acceptable.
View larger version (21K):
[in this window]
[in a new window]
|
Figure 2. Radiographers' weekly test sheet used in conjunction with the phantom. The dashed line shows the minimum performance standard. Reproduction of the linepair groups has been degraded by digitization of the original radiograph.
|
|
|
Conclusion
|
|---|
An easily replicated design of mammographic image quality phantom based on LaserJet printed test features has been described. Approximately 40 of these phantoms were constructed, and these have been used successfully for 5 years in both breast screening and symptomatic mammography. They remain in use in the symptomatic service but have recently been replaced by commercial phantoms in the screening service due to the requirement for consistency across a much-enlarged screening region. The phantom design offers a performance similar to much more expensive mammographic contrast-detail phantoms, but suffers from the disadvantage that high contrast resolution bar patterns cannot be produced using the standard printing process.
Received for publication August 25, 2004.
Revision received January 13, 2005.
Accepted for publication February 28, 2005.
|
References
|
|---|
- Theodorakou C, Horrocks JA, Marshall NW, Speller RD. A novel method for producing x-ray test objects and phantoms. Phys Med Biol 2004;49:1423–38.[Medline]
- Chesters MS, Hay GA. Quantitative relation between detectability and noise power. Phys Med Biol 1983;28:1113–25.[CrossRef]
- Birch R, Marshall M. Computation of bremsstrahlung x-ray spectra and comparison with spectra measured with a Ge(Li) detector. Phys Med Biol 1979;24:505–17.[CrossRef][Medline]
- National Health Service Breast Screening Programme. Guidance on quality assurance visits. NHSBSP Publication No. 40. Sheffield: NHSBSP Publications, 2000.
- Bijkerk KR, Lindeijer JM, Thijssen MAO. The CDMAM phantom: a contrast-detail phantom for mammography. Radiology 1993;185(P):395.
This article has been cited by other articles:
|
|
|
|
 
C J Kotre
An investigation of search pattern extent in the threshold contrast detection task
Br. J. Radiol.,
May 1, 2006;
79(941):
437 - 440.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
