Impacts of Filtration on Contrast-Detail Detectability of an X-ray Imaging System

The purpose of this study is to investigate the impacts of added filtration on the contrast-detail detectability of a digital X-ray imaging system for small animal studies. A digital X-ray imaging system specifically designed for small animal studies was used. This system is equipped with a micro X-ray source with a tungsten target and a beryllium window filtration and a CCD-based digital detector. Molybdenum filters of 0 mm, 0.02 mm, and 0.05 mm in thickness were added. The corresponding X-ray spectra and contrast-detail detectabilities were measured using two phantoms of different thicknesses simulating breast tissue under different exposures. The added Mo filters reduced the low-energy as well as the high-energy photons, hence providing a narrowband for imaging quality improvement. In the experiments with a 1.15 cm phantom, the optimal image detectability was observed using 22 kVp and the 0.05 mm Mo filter. With the 2.15 cm phantom, the best detectability was obtained with 22 kVp and the 0.02 mm Mo filter. Our experiments showed that appropriate filtrations could reduce certain low- and high-energy components of X-ray spectra which have limited contributions to image contrast. At the same time, such filtration could improve the contrast-detail detectability, particularly at relatively low kVp and high filtration. Therefore, optimal image quality can be obtained with the same absorbed radiation dose by the subjects when appropriate filtration is used.


INTRODUCTION
When X-ray images are acquired, both patient radiation dose and image quality should be considered. Low-energy X-ray photons of an entrance spectrum penetrate tissue poorly and may not reach the image detector for useful diagnostic information. These attenuated X-ray photons only contribute to patient's radiation dose and should be selectively removed from the X-ray beam by adding appropriate filtration between the X-ray source and the patients. Contributing less to the patient's radiation dose, high-energy X-ray photons may at the same time result in poor image contrast. Appropriate filtration can remove not only low but also high-energy photons in the X-ray beam. On the other hand, if the total filtration is not selected properly, overfiltration mainly reduces the number of photons that reach the image detector. Lower beam intensity will degrade the image quality as reflected by lower signal-to-noise ratio (SNR). Therefore, the material and thickness of filters need to be carefully selected to satisfy both patient radiation dose and image quality considerations so that good performance of an X-ray imaging system can be achieved [1,2].
The shape of X-ray spectra reflects the energy distribution of X-ray photons [3][4][5]. It determines the relevant radiation output characteristics and is important in X-ray imaging. Beaman et al. investigated the optimal X-ray spectra for mammography [3,5]. They analyzed the influence of kVp setting, filter material, and filter thickness on the shape of Xray spectra. The proper choice of the shape of the X-ray spectrum incident upon the breast can yield an improved image SNR for a given radiation dose [4]. The mean energy of the X-ray spectrum is not a good guide as to whether the spectrum is the appropriate one. If the spectrum is too broad, the low-energy X-rays contribute little to the image and produce Collimator kit X-ray source BR-12 phantoms X-ray and gamma detector CCD-based system detector Figure 1: Schematic of spectra acquisition system. a high radiation dose. Conversely, the high-energy X-rays degrade the image quality by reducing the contrast. Spectra with a narrowband around the effective energy are preferred.
An observer-based contrast-detail (CD) curve is one of the important attributes of the X-ray digital imaging system. Its analysis is an effective, although subjective, method of evaluating image qualities [6,7]. Combined objective and subjective studies can determine the quality of an X-ray imaging system. Based on the Rose model of human visual perception [8], an experimental technique to evaluate object detectability at the threshold of human visibility in medical images was developed [9,10]. The detectability of lowcontrast objects of various sizes is expressed graphically as a contrast-detail curve, which relates the threshold contrast necessary to perceive an object as a function of the object's size. When the threshold contrast is displayed versus the detail (object diameter), a typical contrast-detail curve begins at the upper left corner of the graph (i.e., high contrast, small detail) and declines asymptotically toward the right lower corner (i.e., low contrast, large detail) in the shape of a hyperbola [11].
A CCD-based digital imaging system was designed for small animal studies. The unique features of this system include an ultrasmall focal spot, hence providing a high spatial resolution [12]. This system has been used in studies of small animal vasculatures [13]. Preliminary observer-based analysis has shown that it is a useful tool for small animal studies [14]. In this study, both exit X-ray spectra and contrastdetail phantom images were acquired under various conditions of kVp, filter thickness, and phantom thickness. An observer-based study was conducted to evaluate the contrastdetail phantom images with different filtration conditions.

MATERIALS AND METHODS
A charge-coupled device (CCD)-based X-ray unit, designed for high-detail radiographic imaging (Faxitron X-Ray Corp, Wheeling, Ill) was used for this research. The X-ray tube of this CCD system has a focal spot of 20 µm and is designed to minimize the geometric blur introduced by magnification. The X-ray tube uses a tungsten target and a 0.254 mm  beryllium window filtration to produce radiation with energies ranging from 10 kVp to 35 kVp. The tube current is fixed at 0.3 mA, with an exposure time ranging from 0.1 second to 999 seconds. The distance between the focal spot and the detector is 57.2 cm [15]. The detector module used in this system is composed of a Min-R medium mammography intensifying screen (Eastman Kodak, Rochester, NY) and two fiber optically coupled CCDs. The pixel size on the surface of the intensifying screen is 0.048 mm × 0.048 mm, with a 1024 × 2048 array and 12-bit digitization. Observer-based measurements are conducted using a phantom specially designed for contrast-detail studies (Model 10001, MedOptics Corp., Tucson, Ariz). It is a 6 × 6 × 1.15 cm Lucite slab with forty-nine holes arranged in a 7 × 7 matrix configuration. The holes have seven different diameters ranging from 0.18 to 4.82 mm. They also have different depths ranging from 0.06 mm to 0.73 mm. The contrast of the targets, which increases with the depth of the holes, changes along the rows, and the detail of the target, which increases with the diameter of the holes, changes along the columns [16]. Two breast tissue simulating phantoms, a 1 cm BR-12 phantom, and a 1.15 cm Lucite phantom (MedOptics Corp.) were used. The two phantoms have almost identical X-ray attenuation coefficient. These two phantoms were combined together to make a phantom with an equivalent thickness of 2.15 cm.
The X-ray spectra were obtained by using a spectrometer, which includes an XR-100T-CdTe X-ray and gamma detector system with a power supply and shaping amplifier (Amptek Qirong Zhang et al.   Inc., Bedford, Mass). This unit is a high-performance X-ray and gamma ray detector equipped with a thermo-electric cooler [17]. The X-ray flux rate of the photons may exceed the capability of both the detector and the electronics that process the X-ray spectrum [18]. In order to reduce the count rate to an acceptable level, a collimator kit was used. The collimator kit includes a stainless steel collimator housing and 7 tungsten collimator discs which provide different size pinholes. The geometrical parameters of the discs are given in Table 1.
By selecting the appropriate tungsten collimators, the incoming X-ray flux can be reduced to a level at which the X-ray spectrum can be processed properly by the detector and electronics of the system [19]. In this study, appropriate collimators were used so that the absorbed radiation dose by the subject was the same under different X-ray settings.
Measurements of the spectra were acquired using phantoms of different thicknesses: 1.15 cm (MedOptics phantom) and 2.15 cm (combination of BR-12 phantom and MedOptics phantom). These measurements were conducted separately with combinations of different kVp values (22, 28, and 35) and different Mo filter thicknesses (0, 0.02, and 0.05 mm).
The Mo filters were held by a lead foil with a 10 cm diameter opening (the size of the X-ray window) and placed right beneath the X-ray source. The phantom was placed 46.2 cm above the detector plane. The detector of spectrometer was placed 28 cm below the phantom. The schematic for acquisition of X-ray spectra is shown in Figure 1.
For acquiring contrast-detail phantom images, the Xray spectrometer was removed. The contrast-detail phantom was placed in contact with the image detector to minimize Qirong Zhang et al. magnification. By adjusting the exposure time, identical radiation exposure was obtained at the entrance of the image receptor in every condition. The exposure level was measured as 0.01 R by utilizing an X-ray exposure meter with a mammographic ion chamber (Rad-Check Plus, Victoreen, Inc., Cleveland, Ohio). The chamber was placed directly on top of the imaging detector. The exposure level of 0.01 R was selected because it resulted in an imaging signal that was in the middle of the dynamic range of the X-ray imaging module. Using the same combinations of the kVp and Mo filter thickness, a series of contrast-detail phantom X-ray images were obtained under the same exposure level for the observer-based study. The schematic of the filters and the phantom locations is shown in Figure 2. The images were presented on a CRT computer monitor, one meter away from observers in a dim room. The bright background of the monitor was masked with black paper. The images were processed with window leveling individually to achieve the best perceptible displayed quality. Once the window level was set, the observers were not allowed to make further adjustment of the image appearance on the display. Ten trained observers evaluated these images and recorded the minimum perceptible hole-depths.

RESULTS
The acquired spectrum was normalized to its total counts. Normalized count percentage = (count number of a given channel/ total count number) × 100%. Figures 3 and 4 show the X-ray spectra using the 1.  recorded as functions of the diameter of the holes (the details). Ten observers participated in the study and their results were averaged and plotted as contrast-detail curves, under different conditions. The contrast-detail curves using the 1.15 cm and the 2.15 cm phantoms are given in Figures 5 and  6, respectively. The observer-based contrast details were not significantly affected by the filtration, although at lower energy (22 kVp) the detectability of the subjects was noticeably improved by the filtration. The total number of holes on the 49-hole MedOptics phantom visible to each observer was recorded. The results are shown in Figures 7 and 8 for the 1.15 cm and 2.15 cm phantoms, respectively. Using the results in these two figures, the quality of the X-ray images were compared. Using the 1.15 cm phantom, the observers detected most holes in the X-ray image acquired by 22 kVp and 0.05 mm Mo filter (average point: 27.73). Using the 2.15 cm phantom they detected most holes in the X-ray image acquired with 22 kVp and the 0.02 mm Mo filter (average point: 23.95).

DISCUSSION
In this study, Mo filters of different thicknesses were used in a digital X-ray imaging system for small animals research. Using this system, images of phantom simulating tissue of different thicknesses were acquired. The Mo filters effectively removed the low-energy portion of X-ray spectra, as shown in Figures 3 and 4. Since the low-energy X-ray photons only increase the patient radiation dose without much contribution to the contrast of the X-ray images, the reduction of such photons will reduce the radiation risks. Furthermore, the Mo filters provided a low cut-off around 10 keV, particularly when a 2.15 cm phantom was used (see Figure 4)  the higher-energy region, the Mo filters reduced the X-ray photons higher than 20 keV because of the K-edge of the Mo filters. Again, strong attenuation (2.15 cm phantom) significantly enhanced the cut-off above 20 keV (see Figure 4). By comparing the results in Figure 3 (using the 1.15 cm phantom) and Figure 4 (using the 2.15 cm phantom), it is easy to notice that the thicker the filter and/or the phantom, the narrower the spectrum. The results in the spectrum study show that with appropriate filtration, the shape of the spectra become a narrowband. Furthermore, using the phantom of thickness in the range of tissue thickness of small animals, better image quality can be obtained with reduced radiation dose to the subjects. The results presented in this study provide a guide for selecting adequate filtration for subjects of different thicknesses.
The overall contrast details indicate that the filtration enhances detectability with the same absorbed radiation dose by the subject, particularly when the phantom is thin and the energy is low (see Figures 5(a) and 7). When the phantom is thick, the impact of filtration on detectability is not as significant, as shown in Figures 6 and 8.
Using the 49-hole phantom for contrast-detail analysis, the image acquired using the 1.15 cm phantom, with

ACKNOWLEDGMENTS
This work was supported in part by Grants from the National Institutes of Health (RO1 EB-002604, RO1 CA104773), by the College of Graduate Studies and Research, University of Central Oklahoma, and by a Grant from the National Institutes of Health (P20 RR016478 from the INBRE Program of the National Center for Research Resources). The authors would like to acknowledge the support of Charles and Jean Smith Chair endowment fund as well. The authors would also like to thank the help from observers Qiong Wang, Ben Steele, Yiyang Zhou, Yongsheng Ni, Xingwei Wang, and Da Zhang.