Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.sao.ru/drabek/CCDP/Abstracts/2/IWDMa69.html Дата изменения: Wed Sep 9 15:52:54 1998 Дата индексирования: Tue Oct 2 09:17:25 2012 Кодировка: Поисковые слова: m 13

Application of X-ray Crystallographic Detector to Whole Breast Digital Mammography

Juanhui Xie*, Mark B. Williams**, Martin Stanton*, Daniel Rosen*, Walter Phillips*, and Laurie L. Fajardo**

*Rosenstiel Science Center, Brandeis University, Waltham, MA, USA; **Department of Radiology, The University of Virginia, Charlottesville, VA, USA
e-mail: mbw7a@virginia.edu

Introduction
We are developing a detector for whole breast digital mammography based on a design optimized during the past 10 years for use in x-ray crystallographic studies. Our system is composed of modules in which a gadolinium oxysulfide phosphor is deposited directly onto a fiber optic taper which is bonded to a CCD. In order to acquire a mammographic image of approximately 20 cm x 20 cm, two approaches are being explored. One approach is to bond together four modules with large-end dimensions of approximately 10 cm x 10 cm to form a contiguous 2x2 array. Another approach utilizes a 2x2 array of 5 cm x 5 cm modules, spaced by approximately one module's width. The array is positioned sequentially at four locations to acquire the full image, with the x-ray beam turned off during detector translations. We describe the imaging properties of an individual crystallographic detector, measured using a typical mammographic x-ray spectrum, and suggest ways in which the module's physical parameters could be optimized for mammographic imaging.

Because the phosphor is viewed distally, and because of the light losses inherent in the fiber optic minification, phosphor design considerations differ from those for single emulsion screen-film systems, where the phosphor is viewed proximally and direct coupling exists between the phosphor and the film. In particular, as the phosphor thickness is varied there is a tradeoff between spatial resolution and the total light per incident x-ray at the CCD. We present measured MTF and light output data from Gd2O2S:Tb phosphors of various thicknesses, and discuss implications for the developmental breast imaging detector.

Methods

Test module images were obtained using a molybdenum anode x-ray tube with a 25 micron molybdenum filter and a constant potential generator. The test module has a 16 mg/cm2 Gd2O2S:Tb phosphor directly deposited on a 4:1 fiber optic taper. The small end of the taper is bonded to a 1k x 1k SITe CCD with 24 micron pixels, resulting in a field of view of approximately 10 cm x 10 cm, with 100 micron pixels in the plane of the phosphor. The following detector properties were evaluated: sensitivity in ADU/mR, conversion gain in CCD electrons per incident x-ray, linearity as a function of exposure at the detector surface, dynamic range, presampling modulation transfer function (MTF), noise power spectrum (NPS), and detective quantum efficiency (DQE). The NPS and DQE were measured as functions of spatial frequency and x-ray fluence.

Individual phosphor MTFs were measured by depositing them directly onto a 1:1 fiber optic faceplate. That faceplate was then coupled with optical grease to a second faceplate bonded to a 27 micron pixel CCD. Nine different phosphor layers ranging in thickness from 10 to 30 mg/cm2 were evaluated. The light output per absorbed x-ray was obtained by measuring the average light output of a central region of the CCD in response to uniform x-ray illumination, then dividing by the calculated absorption efficiency. The MTF was measured using a 12 micron wide slit aperture, placed directly against the phosphor and angled slightly with respect to the CCD pixel rows.

Results

With the tube operated at 25 kVp (HVL=0.30 mm Al), the phosphor absorbs approximately 50% of the unattenuated incident x-rays. The average detector sensitivity is 960 ADU/mR, corresponding to a quantum gain of approximately 5.6 electrons per absorbed x-ray. The response is linear up to approximately 40,000 ADU. With 0.2 Mpixel/second readout and thermoelectric cooling to -400C, the total rms electronic noise is 2 ADU, providing a dynamic range of 20,000. The module's presampling MTF is shown in Figure 1 (solid line) along with that measured for the 16 mg/cm2 Gd2O2S:Tb phosphor alone (dashed line). Figure 2 shows the MTF at 5 cycles/mm for a range of phosphor thicknesses (circles) along with the calculated absorption efficiency for 20 keV x-rays (triangles). The light per absorbed x-ray from phosphors over the selected range of thicknesses is essentially constant for phosphor thicknesses up to ~21 mg/cm2, then falls by ~40% between 21 and 30 mg/cm2.
The DQE of the test module is relatively invariant over the measured range of x-ray fluences from about 4.0 x 105 photons/mm2 to 1.3 x 107 photons/mm2. At zero frequency the DQE is 0.40 and falls to approximately 0.2, 0.1, and 0.05 at 2, 4, and 5 cycles/mm respectively.

Discussion
The measured module MTF is significantly poorer than that of its phosphor alone (Figure 1). Contributing factors include the effects of the 100 micron sampling aperture and light scattering in the 4:1 fiber optic taper. The limited MTF is, in large part, responsible for a rapid fall off in the module DQE with increasing spatial frequency. A reduction in the fiber optic taper ratio from its current value of 4:1, or a reduction in the CCD pixel size, will reduce the size of the effective sampling aperture of the digital matrix at the phosphor. A smaller taper ratio will also reduce the optical scatter in the fiber optics. Reduction of the sampling aperture and light scatter will make the overall detector resolution more similar to that of the phosphor. In addition, the increase in quantum gain from the smaller taper ratio is likely to improve the high frequency DQE through its impact on the frequency dependence of the NPS. Figure 2 reaffirms the well known fact that an increase in the low frequency DQE will be obtained only at the expense of lower spatial in a distally viewed phosphor system. On the other hand, it is likely that resolutions intermediate between those shown in Figure 1 will provide good clinical performance. To illustrate this, the test module was used to acquire an image of the ACR accreditation phantom using an exposure that would result in a mean glandular dose of 0.57 mGy to a 50/50 glandular/adipose 4.5 cm compressed breast. This is approximately half the dose typical for a screen-film system utilizing a 25 kVp Mo/Mo beam. In the resulting image, all simulated lesions were visible with the exception of the smallest (0.16 mm diameter) speck group. Thus, even with its limited spatial resolution, the test module was able to outperform a screen-film system and provide a significant dose reduction.
Conclusions
A detector optimized for x-ray crystallography shows promise as a detector for digital mammography. Optimization for mammography is likely to result in an improvement in spatial resolution and quantum gain. Even nonoptimized, the detector exhibits a DQE that is competitive with that of current mammographic screen-film systems, but over a much larger range of input exposures. Measurements of phosphor resolution and light output indicate that the thickness of a distally viewed phosphor can be optimized for both >50% conversion efficiency and good spatial resolution.



Figure 1: Test module MTF (solid line), Figure 2: MTF at 5 cycles/mm (circles, and MTF of its 16 mg/cm2 Gd2O2S:Tb left vertical axis) and x-ray absorption phosphor alone. efficiency (triangles, right vertical axis) over a range of phosphor thicknesses.