| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, University of Texas Medical Branch, 2.448 Clinical
Sciences Bldg., 301 University Blvd., Galveston, TX 77555-0709.
2 Department of Radiology, University of Alabama at Birmingham, Birmingham,
AL.
Received September 21, 2004;
accepted after revision November 8, 2004.
Address correspondence to T. K. Nishino.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. We studied how the CNR between breast and a 5-mm simulated infiltrating ductal carcinoma (IDC) embedded in a 5-cm-thick breast changes with Mo filter thickness. We performed phantom imaging experiments by modifying the filter wheel of a Senographe 2000D unit with Mo filters that ranged from 15 to 90 µm in thickness. A 5-cm-thick 50% glandular-50% adipose breast phantom with a 5-mm insert simulating IDC was used as the phantom for all the cases. The CNRs between the breast phantom and the IDC insert were measured, and average glandular doses were calculated using a filtration-dependent X-ray spectra model and a breast dosimetry model based on a validated Monte Carlo simulation.
RESULTS. The lesion-tissue CNR at a given dose level increases with increasing Mo filter thickness from 15 to 90 µm. The measured squared CNR per dose increased by 8%, 14%, 17%, and 17% for 45-, 60-, 75-, and 90-µm Mo filters, respectively, compared with the standard 30-µm Mo filter. Meanwhile, the exposure times were increased by 35% (45 µm), 71% (60 µm), 177% (75 µm), and 229% (90 µm).
CONCLUSION. Increasing Mo filter thickness from 30 to 60 µm can increase lesion-tissue squared CNR per dose by 14% with a tolerable increase in the duration of exposure.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
X-ray tube filtration is an essential component for generating optimal image contrast and reducing radiation dose in all X-ray imaging techniques. The filter shapes the incident X-ray spectrum and therefore greatly affects the image contrast-to-noise ratio (CNR) and radiation dose. The radiation dose is especially important in mammography because breast tissue is highly sensitive to radiation and there are strict guidelines concerning acceptable dose levels. Hence, the selection of X-ray tube filtration plays an important role in achieving good image quality with reasonable average glandular dose [4-8].
Currently, all digital mammography units still incorporate the same X-ray tube filtration (0.03-mm Mo and 0.025-mm rhodium [Rh]) inherited from their screen-film predecessors. This level of X-ray tube filtration, although roughly adequate for screen-film mammography systems, may not be optimal for digital mammography systems for a variety of reasons. First, because the detectors in screen-film mammography (a gadolinium-based intensifying screen and film) and digital mammography (a layer of cesium iodide on a photodiode array on amorphous silicon substrate) are physically different, their respective response to incident X-ray spectra and fluence will also be quite different. Second, in digital imaging, the CNR is the image quality descriptor, but the contrast itself is an important image quality descriptor in screen-film mammography. Finally, in conventional screen-film mammography, one has to consider the reciprocity law of failure for each filter selection, whereas in digital mammography one does not have this problem. This suggests that the absorption efficiency, conversion efficiency, dose efficiency, signal-to-noise level, CNR, and detective quantum efficiency (DQE) found in digital mammography systems will differ from those in screen-film systems. These quantities and a host of other image parameters are highly dependent on the shape and spectrum of the initial X-ray beam.
Although several works address X-ray spectra and filter optimization for digital mammography, most are theoretic optimizations using computer modeling and simulations, whereas others just compare the two filters available with the full-field digital mammography systems [9-15]. None of these published works specifically addresses the optimization of the thickness of the Mo filter combined with the amorphous silicon-based digital mammography detector, such as the currently prevalent full-field digital mammography system from GE Healthcare (Senographe 2000D). Different from published studies, our study was performed to determine if the current filtration standard of 30 µm of molybdenum coupled with an amorphous silicon-based digital mammography detector is optimal or if a different thickness of Mo filtration would be better.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In our experiment, we used a custom-made breast phantom (Computerized Imaging Reference Systems [CIRS]). The breast phantoms were constructed from custom-made BR12 material. The breast compositions we selected included 50% glandular tissue and 50% adipose tissue. To assess the performance of different filters, we designed two lesion-simulating inserts embedded in the breast phantom. Because detection of infiltrating ductal carcinoma (IDC) is an important task of mammography, we decided to make two inserts to simulate IDC. The material of the simulated IDC inserts was developed exclusively for this experiment by CIRS, a leading vendor in tissue simulation technology. The size of the two cylinder inserts was 5 mm in diameter and 5 mm in thickness, corresponding to a common size when an IDC can be detected on mammography.
The Senographe 2000D digital mammography unit is fitted with a STATORIX M52.2 tube housing that contains a Maxiray 70 T.1 X-ray tube (GE Healthcare). This tube is equipped with two targets (vanadium-doped molybdenum and rhodium), and each target has two focal spots (0.15 and 0.3 mm). The Senographe 2000D collimator is equipped with two sets of removable X-ray tube filters, in addition to the permanent 0.69-mm-thick beryllium window, that are placed in a filter wheel. These filters, 30-µm-thick molybdenum and 25-µm-thick rhodium, were carefully removed with the aid of a qualified service engineer. In place of the standard amounts of filtration, custom-made Mo filters, ranging from 30 to 90 µm in thickness, were inserted into the filter wheel in increments of 15 µm for phantom image acquisitions. All the Mo filters for our study have a purity of 99.9% (Goodfellow). The X-ray beam quality at the various filtration thicknesses was quantified in terms of half-value layer (HVL) and radiation output at 27 kVp. The X-ray exposure reproducibility and exposure time-tube current linearity have been checked with a coefficient of variance less than 1%.
To evaluate filter performance on digital mammography image quality, we
compared the lesion-tissue CNR and radiation dose to the breast. In our
experiment, the lesion-tissue CNRs were measured for images acquired using
settings of 27 kVp and 200 mAs with different Mo filters. Comparing these
exposure techniques with the exposure conditions with a previous experiment
for screen-film mammography
[5], we know that under these
exposure techniques the detector exposures are larger than 3 mR for all the Mo
filters used in the experiments. Under these exposure conditions, the detector
of the Senographe 2000D exhibits a quantum-limited response
[17]. To evaluate the
performance in terms of lesion-tissue CNR and the radiation dose to the
breast, we calculated a figure of merit (FOM) for each filter on the
basis of the measurements as follows:
 |
 |
To determine all the DgN values associated with other nonstandard Mo filters, we performed Monte Carlo simulations. The same Monte Carlo algorithm and X-ray spectra modeling have been used to compute the DgN data tables associated with a 30-µm Mo filter for combinations of different peak kilovoltage settings, X-ray beam HVLs, and different breast compositions and thicknesses [18, 19]. These DgN tables for the standard 30-µm Mo filter have since been adopted as the standard DgN tables by the radiology community. The Monte Carlo code used is based on the a general Monte Carlo N-particle transport code (MCNP code [version 4C], Diagnostics Applications Group, Los Alamos National Laboratory), and simulations were performed on both an SV1 supercomputer (Cray) and a desktop computer (Macintosh, Apple Computer). The X-ray spectra (with the various filters) to be used in the simulation were calculated from a semiempiric spectral model [20]. The DgN(i) values for different filters computed in this work, as well other measured values, are listed in Table 1.
|
Once the FOM values have been calculated from the methods described, each filter's performance is judged by comparing the FOM values. Thick filters reduced the output X-ray photon flux due to the filter's large X-ray attenuation. To compare the performance of different filters, it is meaningful to compare the exposure times needed to achieve the same CNR for different filters. Therefore, we rescaled the exposure times associated with different filters such that they correspond to the same CNR based on the fact that although the image contrast is independent of exposure time, squared noise is inversely proportional to the exposure time.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The improvement of the FOM with a 60-µm filter compared with that for a 30-µm is comparable with the improvement brought by the current full-field digital mammography compared with screen-film mammography. This improvement in CNR per dose is achieved without any increase in the cost of the full-field digital mammography equipment. In addition, as illustrated in Table 1, the exposure time for the 60-µm Mo filter is increased by 71% compared with the standard 30-µm Mo filter. This increase in exposure time is moderate because it will lead to an exposure time of approximately 3 sec for a 5-cm breast with a composition of 50% adipose tissue-50% glandular tissue at 27 kVp.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We have shown that the standard 30-µm Mo filtration, currently used in GE Healthcare's full-field digital mammography systems, is not optimal in terms of CNR and radiation dose to the breast. Our experiments have shown that at 27 kVp, the optimal filtration is more than the current standard of 30 µm. For a 5-cm breast composed of 50% glandular tissue and 50% fatty tissue, an improvement of 14% in the FOM is seen with a Mo filter thickness of 60 µm. Considering the prevalence of GE Healthcare's full-field digital mammography systems in the digital mammography market, our findings are truly significant. We hope that our results will stimulate more work in filtration optimization research from the radiology community and the imaging equipment industry.
Our results indicate that increasing filtration results in dose savings without an accompanying loss of image contrast. In mammography, the X-ray spectrum is shaped by the K-edge filter (Mo filtration) and a spectrum shaped by an appropriate amount of filtration may increase the lesion-tissue contrast due to spectral squeezing [5-9, 18]. Note that the effects of spectral shaping depend on the imaging detector. We expect to see similar dose savings with other detectors (e.g., amorphous selenium-based detectors), although the magnitude of the dose savings will vary uniquely with each detector type. Apparently, more work needs to be done to take advantage of spectrum-shaping in full-field digital mammography. We understand the limitations in our experimental design (the use of one breast thickness, one size for the simulated IDC insert, one type of simulated IDC insert, one breast composition, and one peak kilovoltage setting) and will address these issues in future investigations. For our research with the Senographe 2000D full-field digital mammography system, we will use different peak kilovoltage settings, different breast densities and thicknesses, and different filter materials when we perform future investigations.
The significance of this finding can be further appreciated by noting the following two facts. First, the Senographe 2000D full-field digital mammography model is the most prevalent full-field digital mammography model (> 90% market share) used in current mammography practices in the United States and all over the world. This dominance motivates our efforts to make improvements in its performance by optimization of its X-ray tube filters. Second, an optimal filter depends on breast thickness and composition. Note that 5 cm is the most common breast thickness encountered in breast imaging. Furthermore, 50% adipose-50% glandular breast composition is a standard in mammography research. This is why we have directed our efforts to the filter optimization for imaging a 5-cm breast phantom with 50%-50% composition.
Finally, we would like to point out that although the results presented are for a Senographe 2000D full-field digital mammography system, the results would be similar for other full-field digital mammography systems because all commercial full-field digital mammography systems use the standard 30-µm Mo filter for their Mo target. Some systems provide more powerful X-ray tubes so that these systems may be able to use even thicker Mo filters such as 75-µm filters. Using even thicker filters may increase further the FOM, (CNR)2 / Dg, as our results suggest.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
