Imaging Technique Optimization of Tungsten Anode FFDM System Biao Chen a*, Andrew P. Smith b, Zhenxue Jing a, Elena Ingal a a Hologic, Inc. 600 Technology Drive, DE 1970 b Hologic, Inc. 35 Crosby Drive, MA 01730 ABSTRACT Single Mo target, Mo / Rh, or Mo / W bi-track targets with corresponding Mo and Rh filters have provided optimal target / filter combinations for traditional screen / film systems. In the advent of full-field digital mammography, similar target / filter combinations were adopted directly for digital imaging systems with direct and indirect conversion based detectors. To reduce the average glandular dose while maintaining the clinical image quality of FFDMs, alternative target / filter combinations have been investigated extensively to take advantages of the digital detectors with high dynamic range, high detection dose efficiency, and low noise level. This paper reports the development of a digital FFDM system that is equipped with single tungsten target and rhodium and silver filters. A mathematical model was constructed to quantitatively simulate x-ray spectra, breast compositions, contrast objects, x-ray scatter distribution, grid performance, and characteristics of a-se flat panel detector. Computer simulations were performed to select kv/filter for different breast thickness and breast compositions through maximizing the contrast object detection dose efficiency. A set of phantom experiments were employed to optimize the x-ray techniques within the constraints of exposure time and required dose levels. A 50-micrometer rhodium filter was applied for thin and average breasts and a 50-micrometer silver filter for thicker breasts. To meet our design requirements and EUREF protocol specifications, we finely adjusted x-ray techniques for 0.45, 0.75, 1.0, 1.35 mgy dose modes with regards to ACR phantom scoring and PMMA phantom SNR/CNR performance, respectively. The optimal x-ray techniques significantly reduce average glandular dose while maintaining imaging performance. Keywords: Imaging of the breast (MG), Digital x-ray imaging (DX) 1. INTRODUCTION Single molybdenum (Mo) target, Mo / rhodium (Rh), or Mo / tungsten (W) bi-track targets with corresponding Mo and Rh filters have provided optimal target / filter combinations for traditional screen / film systems 1-. In the advent of fullfield digital mammography (FFDM), similar target / filter combinations were adopted directly for digital FFDM systems with direct- and indirect-conversion based detectors. Commercial FFDM products have claimed benefits of offering significant dose reduction over screen / film system, and employ optimal x-ray techniques based on the selected existing target / filter combinations. However, the optimization processes of x-ray techniques may be vague and limited to the FFDM system design. For example, the rationales for dual-target track design may be useful for screen / film systems but need to be validated for the FFDM systems. Since more than a decade ago the target / filter combinations such as W / Rh, W / silver (Ag), and other combinations have been reported in academic articles, highlighting the imaging performance of those target / filter combinations with higher x-ray effective energies compared to traditional Mo target based target / filter combinations 4-5, 9-10. The conclusions have suggested, for digital mammographic systems, that a single tungsten target with proper filters provides the optimal imaging performance in terms of the targeted average glandular dose and image quality measures (e.g., contrast-noise-ratio (CNR ) and detection dose efficiency defined as CNR / Dose ). Therefore, it should be the time to introduce more efficient target / filter combinations to FFDM system design to improve or maintain the image quality while reducing the average breast glandular dose to patients. This paper reports the development of an FFDM system that is equipped with single tungsten target and rhodium / silver filters to significantly reduce average glandular dose while maintaining imaging performance. The optimization of x-ray * biao.chen@hologic.com; phone: 30-631-713; fax: 30-631-831; www.hologic.com Medical Imaging 009: Physics of Medical Imaging, edited by Ehsan Samei, Jiang Hsieh, Proc. of SPIE Vol. 758, 75830 009 SPIE CCC code: 1605-74/09/$18 doi: 10.1117/1.811745 Proc. of SPIE Vol. 758 75830-1
techniques has been intensively investigated through computer simulations and experimental measurements with imaging performance test phantoms on the tungsten target FFDM system. Although the clinical studies have been performed in a few beast cancer imaging centers, the clinical results of the newly developed tungsten target system is out the scope of this paper.. MATERIALS AND METHODS.1. Tungsten Target FFDM system and Test Phantoms A full field digital mammography system (Selenia, Hologic Inc., Bedford, MA) was used for this tungsten target FFDM system study. The conventional Selenia system features a high performance high transmission cellular (HTC) anti-scatter grid, a directconversion amorphous-selenium flat panel detector, and a comprehensive x-ray system (Figure 1). The prototype system had a single tungsten target x-ray tube (Varian Medical Systems Inc., CA) installed. The original 5μm Mo and Rh filters of Mo target Selenia system were replaced with a 50μm Rh filter and a 50μm Ag filter, respectively. The x-ray generator on the system was modified to accommodate the corresponding filament current changes. The tungsten target Selenia system was characterized by a beam quality (half-value-layer and tube output) that was significantly different from that of Mo target Selenia. The glandular dose level was calculated according to the target / filter combination type, kv/ mas, HVL / tube output, and breast thicknesses. By changing the dose-scaling factor, the system s automatic exposure control (AEC) can be calibrated to designated average glandular dose levels. A set of phantoms were used in the experimental study, including CIRS contrast phantom, ACR phantom (CIRS, NC), PMMA plate phantoms as described in EUREF protocols 3. W Tube Rh and Ag filters HTC grid, Selenia detector Figure 1 W-Selenia FFDM system.. FFDM System Modeling and Computer Simulations The general imaging task for FFDM screening mode is to effectively detect two targeted objects (in this optimization 5mm glandular object and 0.mm micro-calcification) in normal (50%-adipose-50% glandular) breasts. To virtually evaluate the imaging performance of the tungsten target system, a mathematical model has been constructed to quantitatively simulate x-ray spectra with different target / filter combinations 6, breast compositions and contrast objects, x-ray scatter distribution, grid performance, and characteristics of a-se flat panel detector. This model allows the calculation of contrast signal, contrast-noise ratio (CNR), and average glandular dose (AGD) of normal breasts with certain breast thickness, glandular percentage, and x-ray techniques provided that the actual beam quality is characterized on a physical FFDM prototype system. The glandular dose calculation with certain entrance skin spectrum are based on Dance and Boone s methods 7-8...1. Determination of Optimal X-Ray Photo Energy with Monochromatic X-Ray Beams CNR / Dose The detection dose efficiency ( ) is used to determine the optimal x-ray photon energy levels for various breast thicknesses with monochromatic x-ray beams to detect the targeted objects. By changing the breast thicknesses (e.g., -14cm) and monochromatic x-ray photon energy (10-40keV), we have obtained a series of figure-of-merit (FOM) curves (i.e., CNR / Dose vs. photon energy at each breast thickness). This simulation with the theoretically monochromatic x-ray beams would help us understand the optimal effective photon energy range for various breast thicknesses so that the appropriate filters with specific k-edges can be chosen to maximize the detection dose efficiency. Proc. of SPIE Vol. 758 75830-
... Optimal Filters / kvs with Multi-chromatic X-Ray Beams Practically the skin entrance x-ray beams are multi-chromatic, which are specified with the x-ray spectra that are controlled by target peak voltage and filter thickness. For each breast thickness and a selected filter, a series of computer simulations have been performed by changing kv of tungsten target from 0 through 39 kv to find the optimal kv ranges in terms of detection dose efficiency. The thicknesses of Rh and Ag filters have been varied to optimize the shape of the skin entrance x-ray spectrum, reducing the low kev photons and emphasizing the useful spectrum portion...3. Adjustment of X-Ray Techniques with Practical Constrains The computer simulations created a guideline for the system design of the tungsten target FFDM system. However, under considerations of system constraints (e.g., system maximum power, practical tube output values, and AEC method), we need to adjust the x-ray techniques to obtain the optimized ones in terms of x-ray exposure capacity, CNR values, and average glandular dose levels in AEC mode..3. Experimental Phantom Study A standard ACR phantom with a Lucite disk as a contrast object on top is used to evaluate the imaging performance in terms of ACR fibers/specks/mass scores, SNR and CNR values with the clinical techniques in 4 average glandular dose modes (i.e., 0.45, 0.75, 1.0, and 1.35 mgy). The EUREF Protocol 3 defines a procedure and a specification for CNR test with PMMA plates in the thickness range from 0 to 70mm, using the clinical AEC settings (tube voltage, target, filter and mode) with an aluminum object of 0.mm on top. The directly measured CNR values need to be adjusted to generate CNR values relative to 50mm PMMA with the threshold contrast determined based on human reviewer s observations of the images of a CDMAM phantom and a standard test block. A set of CNR values are specified as limiting values against the measured relative CNR values. 3. RESULTS 3.1. Prototype Tungsten Selenia System Based on the results of our computer simulations and fine adjustments in the following sections, we set up tungsten target FFDM test systems. With a survey of three tungsten Selenia systems, we obtained the HVL and tube output values of a typical tungsten target FFDM system (Tables 1-). Compared to Mo/Mo and Mo/Rh s data that was from a Mo Selenia system, the HVL values of W/Rh and W/Ag are much higher while the corresponding tube output values are significantly lower (Table 3). Table 1 HVL for W/Rh and W/Ag combinations KV W/50μm Rh (mm Al) W/50 μm Ag (mm Al) 4 0.481 0.495 5 0.496 0.518 6 0.511 0.540 7 0.53 0.561 8 0.534 0.58 9 0.544 0.596 30 0.553 0.610 31 0.561 0.60 3 0.569 0.630 33 0.578 0.64 34 0.586 0.653 35 0.594 0.663 36 0.60 0.673 Proc. of SPIE Vol. 758 75830-3
Table Tube outputs at 4.5cm above breast platform kv W/ 50μm Rh (mgy/mas) W/ 50 μm Ag (mgy/mas) 4 0.01 N/A 5 0.05 N/A 6 0.08 0.041 7 0.031 0.047 8 0.035 0.05 9 0.038 0.058 30 0.041 0.063 31 0.044 0.068 3 0.048 0.074 33 0.051 0.079 34 0.054 0.085 35 0.057 0.090 36 0.060 0.094 37 0.063 0.100 38 0.066 0.106 39 0.069 0.111 Table 3 HVL and tube output values at 8kV with Mo and W targets and Mo, Rh and Ag filters Target / Filter HVL (mm Al) Tube Output (mgy/mas) Mo / Mo (30μm) 0.355 0.1 Mo / Rh (5μm) 0.43 0.073 W / Rh (50μm) 0.543 0.035 W / Ag (50μm) 0.578 0.05 3.. FFDM System Modeling and Computer Simulations We built a software library (XData) and corresponding applications to compute practical spectra for various tube/filter combinations (e.g., Mo/Mo, Mo/Rh, W/Rh, W/Ag, and W/other filters), average glandular doses for -9cm breasts, contrast/cnr values, and detection dose efficiencies. 3..1. Spectra of Target / Filter Combinations With our mathematical model of the imaging system, we produced x-ray skin entrance spectra with different target/filter combinations. For example, Figure shows the x-ray fluence spectra of Mo/Mo, Mo/Rh, W/Rh, and W/Ag at 8kVp. It is clear that the low energy portion in the spectra of W/Rh and W/Ag are reduced while the x-ray photons in kev range of [0, 5.51] are boosted where 5.51 is the Ag s k-edge, resulting in more penetration of the x-ray beam and less lower energy photos imparted in the patient breast. The main spectrum portions of Mo target that is considered advantageous for screen/film systems are no longer significant in the spectra of W/Rh and W/Ag, which provides the potential ability for tungsten target system to reduce dose and maintain the imaging performance. For thicker breasts, higher energy x-ray photons that are needed can only be provided by higher k-edge filter such as Rh and Ag filter. We varied the thicknesses of Rh and Ag filters from 5μm to 60μm and confirmed that the thickness of 50μm could remove the tungsten target s inherent low energy characteristic spectrum peak around 10keV efficiently. Thicker filters would reduce the tube output values, limiting the x-ray effective exposure capability of the system. Other element filters (e.g., cadmium (Cd), tin (Tn), and aluminum (Al), may provide some benefits but the k-edge performance of Rh and Ag filters makes them favorable to our applications at this point. Proc. of SPIE Vol. 758 75830-4
Spectra of Mo/Mo, Mo/Rh, W/Rh and W/Ag at 8kVp Relative Fluence 1.8 1.6 1.4 1. 1 0.8 0.6 0.4 0. 0 W/Rh W/Ag Mo/Mo Mo/Rh 5 10 15 0 5 30 Photon Energy (kev) Figure Spectra of Mo/Mo(30μm), Mo/Rh(5μm), W/Rh(50μm), and W/Ag(50μm) at 8kVp 3... Optimal X-Ray Photo Energies with Monochromatic X-Ray Beams For a 5mm glandular object and a 0.mm calcification within -14cm 50%-glandular normal breasts, we obtained optimal photon energy levels with theoretically monochromatic x-ray beams for -8cm normal breasts, and extrapolated optimal photon energies for 10-14cm breasts (Table 4). It is worth noting that for two different targeted objects the optimal kevs are close to each other for the same breast thickness, which means we can optimize x-ray techniques at the same time for both targeted objects. We confirmed the same conclusion drawn by Fahrig and Yaffe that generally a single filter-kvp pair serves well to optimize CNR for the two major indicators of breast cancer. Table 4 Optimal photon energies with monochromatic x-ray beams Breast Thickness cm 4cm 6cm 8cm 10cm 1cm 14cm Optimal kev for a 5mm glandular 15 19 4 5 5.5 6 Optimal kev for a 0.mm calcification 16 0 4 5 5.5 6 For 4-6cm breasts, the optimal photon energies are close to Rh s k-edge of 3. kev. If no other lower element atomic number filters are used, Rh filter should be a good filter candidate to formulate a practical multi-chromatic spectrum. For thicker breasts, the optimal photon energies are close to Ag s k-edge of 5.51 kev, indicating that Ag filter is able to combine with W tube for a practical multi-chromatic spectrum. 3..3. Optimal Filters / kvs with Multi-Chromatic X-Ray Beams Again for a 5mm glandular object and a 0.mm calcification within -9cm 50%-glandular normal breasts, using the computer-simulated W/Rh spectra, we obtained optimal kvs for -9cm normal breasts (Table 5 and Figure 3). For 7-9 cm normal breasts, we obtained Table 6 and Figure 4 based on the computer-simulated W/Ag spectra. It is shown that the FOM curves around peaks for corresponding breast thicknesses are pretty flat, which means the optimal kvs can be varied in a range without sacrificing the detection dose efficiency significantly. This interesting feature allows us to adjust kv during the experimental phantom study to meet other requirements, such as CNR values and average glandular dose levels. Proc. of SPIE Vol. 758 75830-5
Table 5 Optimal kv range for W/50umRh combination Thickness (cm) Optimal kv for 5mm glandular Optimal kv for 0.mm calcification Optimal kv Range.0 0-0-3 3.0 5 6 4-6 4.0 8 8 5-9 5.0 8 8.5 6-30 6.0 9 9 8-30 7.0 9.5 30 8-3 8.0 30 30.5 30-34 9.0 31 31.5 30-35 Computer Simulated, W/50um Rh Relatibe Dose Efficiency 6 5 4 3 1 0 0 4 6 8 30 3 34 36 38 kv cm 3cm 4cm 5cm 6cm 7cm 8cm 9cm Figure 3 Optimal kv range for 5mm glandular object with W/50um Rh combination in terms of detection dose efficiency Table 6 Optimal kv range for W/50um Ag combination Optimal kv for Optimal kv for 5mm glandular 0.mm calcification 7.0 31 31 7-34 8.0 3 3 30-34 9.0 33 33 30-34 Thickness (cm) Optimal kv Range Proc. of SPIE Vol. 758 75830-6
Relative Dose Efficiency 0.5 0. 0.15 0.1 0.05 0 Computer Simulated, W/50um Ag 0 4 6 8 30 3 34 36 38 40 4 44 46 48 Figure 4 Optimal kv range for 5mm glandular object with W/50um Ag combination in terms of detection dose efficiency kv 7cm 8cm 9cm 3..4. Adjustment of X-Ray Techniques with Practical Constrains Since there are constrains / requirements on the system effective exposure capability, allowed exposure time, and average glandular dose levels (0.45, 0.75, 1.0, and 1.35mGy), we adjusted kv / filter to achieve certain CNR values against the limiting values specified in EUREF protocols, or reach certain dose levels while maintaining image quality. Table 7 Practical x-ray kv / Filter table for 1.0mGy dose mode and AEC auto-filter mode Thickness (cm) Filter kv Mas Low mas High 0.5-1.0 Rh 5 18 5 1.0-1.5 Rh 5 5 35 1.5-.0 Rh 5 35 50.0-.5 Rh 5 50 65.5-3.0 Rh 5 65 90 3.0-3.5 Rh 6 70 95 3.5-4.0 Rh 6 95 130 4.0-4.5 Rh 7 115 150 4.5-5.0 Rh 8 10 160 5.0-5.5 Rh 8 160 0 5.5-6.0 Rh 9 190 46 6.0-6.5 Rh 30 14 70 6.5-7.0 Rh 3 198 58 7.0-7.5 Ag 8 04 55 7.5-8.0 Ag 9 10 65 8.0-8.5 Ag 30 3 84 8.5-9.0 Ag 31 30 87 9.0-9.5 Ag 3 44 305 9.5-10.0 Ag 34 195 40 10.0-10.5 Ag 35 135 165 Proc. of SPIE Vol. 758 75830-7
Table 7 shows the practical x-ray kv/filter table for 1.0mGy average glandular dose level in AEC auto-filter mode. Note that the W/Ag combination steps in when the breast thickness is greater than 7.0cm. Compared to W/Rh, W/Ag provides higher tube output value at the same kv, which allows shorter exposure time to tolerate the patient s motion during the x-ray exposure. In Table 7, the mas low (e.g., 80) and mas high (e.g., 100) are actually the typical mas values for the lower thickness (e.g., 4.0cm) and upper thickness (e.g., 4.5cm) of the breast thickness step (e.g., 4.0-4.5cm). 3.3. Experimental Phantom Study 3.3.1. ACR Phantom A series of ACR phantom images were acquired on the tungsten target system in four dose modes (0.45, 0.75, 1.0, and 1.35mGy), and in 1.60mGy dose mode on molybdenum target system with a Lucite disk as the contrast object for CNR measurement (Table 8). Regarding fibers / specks / masses scores, the tungsten target system has good performance even in two lower dose modes (i.e, 0.45 and 0.75mGy). Table 8 also shows that the tungsten system s 1.0mGy dose mode is comparable to Mo system s 1.60mGy mode in terms of ACR scores and CNR value while W system s SNR value is relatively higher. Table 8 ACR phantom scores and SNR /CNR values ACR Phantom Dose Mode Fibers/Speckles/Masses Scores SNR CNR W system 0.45mGy 4.0-3.0-3.0 6.8 6.6 W system 0.75mGy 4.5-4.0-4.0 40.8 9.4 W system 1.0mGy 5.0-4.0-4.0 47.8 10.8 W system 1.35mGy 5.0-4.0-4.0 55.0 1.5 Mo system 1.60mGy 5.0-4.0-4.0 44.3 10.9 3.3.. PMMA Phantoms A series of PMMA plate images were acquired in 1.0mGy average glandular dose mode with the clinical AEC techniques. We determined the contrast threshold as 1.3μm after reviewing a set of images of CDMAM phantom and standard test block by 3 reviewers, calculated the absolute CNR values from the PMMA plate images, and eventually obtained the adjusted CNR values relative to 50mmPMMA (Table 9). It is clear the test results pass the specified limiting values in the EUREF protocol. Table 9 EUREF CNR test results in 1.0mGy average glandular dose mode with clinical AEC techniques Equivalent Breast Adjusted CNR PMMA Thickness Filter kv CNR (f = 0.73) (mm) (mm) (%) 0 1 Rh 5 8.73 178.0 30 3 Rh 6 8. 167.6 40 45 Rh 8 7.78 158.6 45 53 Rh 8 7.48 15.5 50 60 Rh 30 6.7 137.0 60 75 Ag 9 6.09 14.0 70 90 Ag 3 5.00 10.0 Proc. of SPIE Vol. 758 75830-8
4. CONCLUSIONS A mathematical model was constructed for x-ray technique optimization of FFDM systems with different target/filter combinations. The model allows the simulations of imaging performance and provides a computed FFDM platform for the x-ray technique design. We set up and tested the first single tungsten target FFDM system with a 50μm Rh filter and a 50μm Ag filter. The high imaging performance is maintained while the AGD has been significantly reduced compared to traditional Mo target based FFDM systems. REFERENCES 1. Desponds L, Depeursinge C, Grecescu M, Hessler C, Samiri A, and Valley JF, Influence of anode and filter material on image quality and glandular dose for screen-film mammography, Phys. Med. Biol., 36(9): 1165-118 (1991).. Fahrig R, Yaffe MJ, Optimization of spectral shape in digital mammography: dependence on anode material, breast thickness, and lesion type, Medical Physics, 1(9): 1473-81 (1994). 3. EUREF, European guidelines for quality assurance in breast cancer screening and diagnosis, 4 th edition, ISBN 9-79-0158-4 (006) 4. Obenauer S, et al, Dose reduction in full-field digital mammography: an anthropomorphic breast phantom study, Br J Radiol, 76(907): 478-8 (003). 5. Flynn MJ, et al, Optimal radiographic techniques for digital mammograms obtained with an amorphous selenium detector, Proceedings of the SPIE, 5030: 147-156 (003). 6. Boone J, Fewell TR, Jennings RJ, Molybdenum, rhodium, and tungsten anode spectral models using interpolating polynomials with application to mammography, medical Physics, 4(1): 1863-74 (1997). 7. Dance DR, et al. Influence of anode/filter material and tube potential on contrast, signal-to-noise ratio and average absorbed dose in mammography: a Monte Carlo study. Br. J. Radiol. 73(874): 1056-67 (000). 8. Boone J, Glandular breast dose for monoenergetic and high-energy x-ray beams: Monte Carlo Assessment, Radiology, 13: 3-37 (1999). 9. Delis H, Suitability of new anode materials in mammography: Dose and subject contrast considerations using Monte Carlo simulation, Medical Physics, 33(11): 41-435 (006). 10. Williams MB, et al, Optimization of exposure parameters in full field digital mammography, Medical Physics, 35(6): 414-3 (008). Proc. of SPIE Vol. 758 75830-9