Edge on silicon microstrip detectors for medical imaging

ARTICLE IN PRESS Nuclear Instruments and Methods in Physics Research A 549 (2005) 199–204 www.elsevier.com/locate/nima Edge on silicon microstrip detectors for medical imaging A. Bergamaschi, F. Arfelli, D. Dreossi, R. Longo, A. Olivo, S. Pani, L. Rigon, E. Castelli Dipartimento di Fisica, Universita’ degli Studi di Trieste and INFN, Sezione di Trieste, Italy Available online 23 May 2005 Abstract Silicon microstrip detectors in the edge on geometry, i.e. oriented with the strips parallel to the incoming beam, allow a high absorption efficiency for X-rays in the 10–100 keV energy range. In medical imaging, this results in a significant reduction of the dose given to the patient with respect to conventional screen-film systems. This geometrical configuration was first proposed in the early 90s and since then it has been applied to several medical imaging projects. In this short review, an overview of some medical imaging applications using edge on detectors will be presented, focusing the attention on the main characteristics of the detection systems. The studies performed in Trieste in order to optimize such detectors for digital mammography with synchrotron radiation will be described in detail. In particular, novel imaging techniques (phase contrast, refraction and scatter imaging, computed tomography) implemented and tested at the Trieste synchrotron light source will be shortly discussed. r 2005 Elsevier B.V. All rights reserved. PACS: 29.40.Wk; 87.59.Hp; 87.59.Ek Keywords: Silicon microstrip; Digital mammography; Synchrotron radiation; Single photon counting 1. Introduction Silicon microstrip detectors are largely used in many fields of physical research [1]. The main problem related to the use of silicon strip detectors for medical X-ray imaging is that the absorption Corresponding author. Tel.: +39 040 375 6237; fax: +39 040 375 6258. E-mail address: [email protected] (A. Bergamaschi). length of silicon in the energy range 10–100 keV is of the order of (or larger than) one mm so that only a small fraction of the X-rays are converted in the commonly used 300 mm thick detectors when the radiation impinges on the surface of the detector (face on configuration). This low absorption efficiency means that the undetected radiation increases the dose to the patient without contributing to the image formation. The search for a new semiconductor material to replace silicon has not been successful yet because of the wide 0168-9002/$ - see front matter r 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2005.04.052 ARTICLE IN PRESS A. Bergamaschi et al. / Nuclear Instruments and Methods in Physics Research A 549 (2005) 199–204 200 p l t d Impinging beam Fig. 1. Schematic of the edge on geometric configuration. The cutting distance is indicated with d. diffusion of the techniques for the fabrication of silicon sensors and electronics devices [2]. Moreover, when using double sided detectors with crossed strip to acquire a 2D image, the read out needs to be faster than state of the art electronics in order to avoid ambiguities for multi-hit events [3–6]. These problems can be solved by orienting the detectors with the strips parallel to the incoming X-ray beam. This geometric configuration, called edge on, was first proposed by the SYRMEP collaboration and is sketched in Fig. 1 [7,8]. The pixel size is given by the strip pitch p times the detector thickness t. The absorbing thickness is therefore given by the full strip length l, that can be of several cm. The main limitation on the efficiency of the detector is given by the size of the undepleted region which is present in the entrance window of the detector between the end of the strip implants and the edge of the detector. The thickness of this region is mainly defined by the distance d between the end of the strips and the scribe line. This cutting distance is usually about 1.5 times the wafer thickness in order to limit the leakage current, but can be reduced by using innovative fabrication methods [9,10]. The size of the undepleted region is usually smaller than d and can be tuned by changing the bias voltage of the detector [11]. The efficiency of the detector in edge on orientation is much higher than in face on configuration and generally higher than the one of conventional screen-film systems, resulting in a significant reduction of the patient dose. The X-rays are converted into electric signal without the intermediate step of conversion into visible light. Thus the spatial resolution is not deteriorated by diffusion of secondary light in the phosphor screen. The charge collected for each photon is of several thousand electrons, and each X-ray can be counted directly. The photon counting technique counts as a photon every signal higher than a threshold selected in order to suppress the noise [12]. The main advantages of counting systems are: (1) the maximization of the contrast resolution, since the only noise source lays in the intrinsic Poisson statistic on the number of detected photons; (2) the perfectly linear behavior on the whole dynamic range, that can be chosen according to the specific application requirements; (3) the possibility of implementing multiple thresholds for energy discriminating techniques, that can be used for instance for dual energy radiography, K-edge subtraction or scattering discrimination. The edge on configuration determines an essentially one-dimensional detector, allowing an almost complete scattering rejection also without the use of anti-scattering grids. On the other hand, scanning is required in order to acquire a twodimensional image, leading to an increase of the duration of the examination with respect to area detectors. 2. Overview of existing edge on detectors In this section, a brief overview of some projects using edge on detectors in conjunction with an X-ray tube will be presented. The detector is aligned with a slit which is positioned before the sample; the X-ray tube, the collimator and the detector are then scanned across the object [2]. Since the divergence of the beam is not negligible, the detectors are designed with a slight strip fan out in order to avoid any parallax problem. The experiments described are active in the field of spinal radiography and mammography. The differences on the detector design resulting from ARTICLE IN PRESS A. Bergamaschi et al. / Nuclear Instruments and Methods in Physics Research A 549 (2005) 199–204 the different requirements of the examination will be discussed. 2.1. Spinal radiography Spinal cord X-ray imaging is commonly used for the diagnosis of vertebral lesions. This type of examination is usually performed with the patient standing and the area to be imaged is wide (50  120 cm2). A resolution higher than 1 mm is required and the optimum X-ray energy for the examination is about 50 keV. A ‘‘quantum X-ray radiology apparatus’’ based on a silicon microstrip detector in edge on configuration was optimized for spinal radiography by Hilt et al. [13,14]. Both the thickness of the silicon sensor and the strip pitch are equal to 500 mm. The strip length is 5 cm, corresponding to a stopping power of more than 90% for 50 keV photons. Signal processing is performed by an analog ASIC followed by a digital one [15]. Eight detectors are arranged in a linear array covering a length of about 50 cm with a dead zone of only three strips between the modules [13]. Dose measurements show that with this detection system it is possible to reduce the dose given to the patient by at least a factor of ten [14]. A second version of the apparatus with an optimized geometry, improved spatial resolution and spectroscopic capabilities is under development. 2.2. Digital mammography Mammography is the most effective examination for the diagnosis of breast cancer [16]. Optimum contrast and spatial resolution are required in order to ensure an early detection of the lesions. Since the spectrum produced by a typical mammographic tube features intensity peaks at relatively low energies, between 17.4 keV (Mo Ka) and 22.7 keV (Rh Kb), it is very important to reduce the thickness of the undepleted region in the entrance window of the detector. Several experiments of reducing the cutting distance have been performed at the Jozef Stefan Institute in Ljubljana (Slovenia) [17,18]. The prototype detectors used have a strip length of 201 4 mm, a thickness of 220 mm and a strip pitch of 100 mm. The detectors were tested on wafer and with cutting distances between 400 and 136 mm [19]. An increase of leakage current of one order of magnitude at full depletion voltage has been observed only for the detector cut at 136 mm. However, all detectors were operated without any observable increase of noise when connected to single photon counting read out electronics [20]. The resulting efficiency of the detector is of about 90% for 22 keV X-rays. Instead of reducing the cutting distance, researchers of the Royal Institute of Technology in Stockholm (Sweden) have proposed to tilt the detector in front of the beam in such a way that the device is irradiated beside the undepleted region [21–23]. The resulting effective dead layer ranges between 20 and 50 mm. Some of the higher energy photons escape from the backplane of the detector but lower energy ones are totally absorbed. The strip length of the detectors is 1 cm and, after the optimization of the tilt angle (about 4.51), the detection efficiency is of about 95% [24]. The dose can be reduced down to 15 than in a conventional mammographic examination [25]. The strip pitch of the detectors is 50 mm, and the vertical pixel size is reduced to 50 mm by means of a slit placed in front of the detector. Several detectors are arranged in a grid in order to reduce the scanning time. This detector is now produced by Mamea Imaging AB [26] and commercialized by Sectra AB [25]. 3. The SYRMEP project The main goal of the SYnchrotron Radiation for MEdical Physics (SYRMEP) collaboration is to optimize the quality of radiological examinations operating both on the source and on the detector side [27,28]. A synchrotron radiation beamline dedicated to mammography has been built at Elettra, the synchrotron radiation facility in Trieste (Italy). With respect to conventional X-ray tubes, synchrotron radiation provides much higher fluxes on a wide range of energies, with the possibility of monochromatizing the beam in order to optimize ARTICLE IN PRESS 202 A. Bergamaschi et al. / Nuclear Instruments and Methods in Physics Research A 549 (2005) 199–204 the photon energy according to the radiographed sample. The beamline is now being upgraded for clinical examinations on patients, which are expected to start in 2004. The research and development studies on edge on detectors with single photon counting read out started at the beginning of the 90s [7]. Since the synchrotron radiation beam is laminar with an area of 20  0.4 cm2, the detector is kept stationary in front of it, while the sample is scanned. The beam divergence is negligible therefore the strips of the detector are parallel. The first detector prototype had 500  500 mm2 pixel and only 15 channels were equipped with discrete read out electronics [8]. In the following paragraph, the evolution of the detection system will be described and some innovative imaging technique tested at the SYRMEP beam line will be discussed. 3.1. The detector In order to exploit the maximum beam dimension it is necessary to stack several detectors together. Fig. 2 shows how the sensors are assembled in order to obtain a multi-layer module [30,31]. The detectors were designed with an innovative trapezoidal geometry. ‘‘Full’’ structures are alternated to layers of ‘‘half’’ structures in order to allow enough space for the wire bonding to the read out electronics and power supply [11]. The sensors are connected to the readout electronics with kapton fan-outs. A three-layer prototype has been assembled and tested. The overall statistics of the image is obtained by summing up the information of corresponding pixels belonging to different layers, thereby reducing the acquisition time and the negative effects due to noisy or dead pixels [32]. The pixels in superimposed layers are aligned with a precision of about 10%, and the space between the two half modules is of only two strips [33]. A distance of 50 mm was kept between the layers by using kapton foils of the same thickness with holes for glue dripping as spacers, avoiding cross-talk effects. The width of the prototype is about 5 cm, but it would be possible to place several modules side by side in order to cover a wider area. The detector is AC coupled and FOXFET biased. The strip length is 1 cm, while, on the basis of simulations and measurements on design structures, the cutting distance has been fixed to 250 mm [34]. The average undepleted thickness has been evaluated from efficiency measurements to be less than 150 mm. The overall efficiency is about 80% at 20 keV. The pixel size is 200 mm in the horizontal direction and 300 mm in the vertical one, but the vertical resolution can be improved by scanning the sample in front of the beam with a step smaller than the wafer thickness [35]. A filtered deconvolution algorithm is then applied to the acquired data. In this way, the spatial resolution obtained is determined by the scanning step rather than by the pixel size. New detectors have been designed in order to improve the spatial resolution in the horizontal direction, decrease the dark current and optimize the efficiency. Six different types of detectors of different sizes were designed on the same set of masks. The detectors are rectangular and their width ranges between 3 and 13 cm. The strip length varies between 1 and 2 cm and the strip pitch is 50 or 100 mm [36]. In order to minimize the cutting distance the guard ring is present only on three sides of the sensors and the distance between the scribe line and the end of the strips ranges between 240 and 400 mm. The leakage current is smaller than 0.1 nA per strip and the sensors have been coupled to an application specific read out electronics [37]. 3.2. Novel imaging techniques (a) (b) Fig. 2. Scheme of the stacking of the three layers detector. New imaging modalities based on the detection of weak phase perturbation effects, namely phase ARTICLE IN PRESS A. Bergamaschi et al. / Nuclear Instruments and Methods in Physics Research A 549 (2005) 199–204 contrast and diffraction imaging, have been developed by several researchers [29,38]. The high spatial coherence requirements connected with the use of these techniques are met since a synchrotron radiation source is used. The high efficiency of the SYRMEP edge on detector allows the acquisition of medical images using phase techniques and delivering a low dose to the sample [39]. The pixel size has been reduced to 100 mm by means of a slit in order to detect the narrow interference peaks. Moreover, the use of a three layers detector makes it possible to simultaneously acquire images based on different techniques (e.g. absorption, diffraction enhanced, small angle scattering) on different detector layers by means of a specific setup [33]. This technique results in an increase of the information extracted from the sample without increasing the dose delivered. A feasibility study of breast computed tomography on in vitro tissues has been carried out with promising results [40–42]. The dose is comparable to the one delivered in clinical mammography due both to the high efficiency of the detection system and to the use of a monochromatic beam at the optimal energy for the sample. The breast structures are visible without the overlapping effects typical of two-dimensional imaging. Moreover, the reconstructed images are free from the beam hardening artifacts that arise when polychromatic X-ray beams are used. A detector optimized for synchrotron radiation breast tomography is currently being developed. It will be made of two layers and should cover a width of 20 cm with a spatial resolution of 100 mm [43]. 203 Edge on sensors play an essential role in two experiments carried-out in the field of medical applications of synchrotron radiation:   In planar radiography, the use of a three layers detector makes it possible to simultaneously acquire images based on different effects: absorption, the so-called phase contrast imaging, in free space propagation modality, and small or very small angle scattering [33]. Without increasing the dose delivered to the sample, three different images can be obtained in a single acquisition step. And a fusion of these images can be very useful for the diagnostic purposes. In tomographic radiography, in particular, in breast tomography, the edge on detection system allows the acquisition of tomographic images at doses comparables to those of the conventional planar mammographic examinations [40–42]. In order to maximize the contrast resolution in the acquired images, the single photon counting modality has been chosen for these applications. Moreover, in this geometrical setup it is necessary to rely on scanning techniques. Due to these requirements, the main limitation connected with the use of edge on detectors in clinical examinations is the long duration of the acquisition. In Trieste, this problem has been overcome by assembling arrays of detectors covering large areas [31] and by developing fast low noise read-out electronics, reducing therefore the time of each acquisition step [36,37]. In this way, the time of image acquisitions is compatible with the constraints of a clinical examination. 4. Summary Several medical applications of edge on silicon microstrip detectors have been described. In particular, they have been used for X-ray imaging applications due to their high efficiency in the energy range 10–100 keV. Interesting results regarding further efficiency enhancements can arise from novel manufacturing techniques like active edge detectors [10]. References [1] G. Lutz, Semiconductor Radiation Detectors, Springer, Berlin, 1999. [2] H.F.-W. Sadrozinski, IEEE Trans. 48 (4) (2001) 933. [3] M. Conti, Nucl. Instr. and Meth. A 360 (1995) 287. [4] A. Del Guerra, Nucl. Instr. and Meth. A 394 (1997) 191. [5] R.D. Speller, et al., Nucl. Instr. and Meth. A 457 (2001) 653. [6] G.J. Royle, et al., Nucl. Instr. and Meth. A 493 (2002) 176. ARTICLE IN PRESS 204 A. Bergamaschi et al. / Nuclear Instruments and Methods in Physics Research A 549 (2005) 199–204 [7] G. Barbiellini, et al., in: L. Andreucci, A. Schenone (Eds.), Topics on Biomedical Physics, World Scientific, Singapore, 1991, p. 79. [8] F. Arfelli, et al., Phys. Medica IX 1 (1) (1993) 229. [9] D. Krizaj, S. Amon, Nucl. Instr. and Meth. A 439 (2000) 451. [10] C.J. Kenney, et al., IEEE Trans. 48 (6) (2001) 2405. [11] F. Arfelli, et al., Nucl. Instr. and Meth. A 385 (1997) 11. [12] B. Mikulec, Nucl. Instr. and Meth. A 510 (2003) 1. [13] B. Hilt, et al., Nucl. Instr. and Meth. A 442 (2000) 355. [14] B. Hilt, et al., Nucl. Instr. and Meth. A 442 (2000) 38. [15] P. Fessler, et al., Nucl. Instr. and Meth. A 421 (1999) 130. [16] M. Yaffe, in: J. Beutel, et al. (Eds.), Digital Mammography in Handbook of Medical Imaging, SPIE Press, 2000, p. 329. [17] T. Mali, et al., Proc. IWDM (2000) 89. [18] D. Vrtacnik, et al., Sens. Act. A Phys. 85 (2000) 209. [19] T. Mali, et al., Proc. MIDEM (1998) 199. [20] G. Comes, et al., Nucl. Instr. and Meth. A 377 (1996) 440. [21] E. Beuville, et al., Nucl. Instr. and Meth. A 406 (1998) 337. [22] E. Beuville, et al., IEEE Trans. 45 (6) (1998) 3059. [23] M. Lundqvist, et al., IEEE Trans. 47 (4) (2000) 1487. [24] M. Lundqvist, et al., IEEE Trans. 48 (4) (2001) 1530. [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] Sectra AB, http://www.sectra.se/ Mamea Imaging AB, http://www.mamea.se/ F. Arfelli, et al., Rev. Sci. Instr. 66 (1995) 1325. F. Arfelli, et al., Phys. Med. XIII (1997) 7. P. Suortti, W. Thomlinson, Phys. Med. Biol. 48 (2003) R1. F. Arfelli, et al., Inf. MIDEM 29 (1999) 26. F. Arfelli, et al., Nucl. Phys. B (Proc. Suppl.) 78 (1999) 592. F. Arfelli, et al., Nucl. Instr. and Meth. A 409 (1998) 529. A. Olivo, et al., Rev. Sci. Instr. 74 (7) (2003) 3460. F. Arfelli, et al., Nucl. Instr. and Meth. A 377 (1996) 508. A. Olivo, et al., Med. Phys. 27 (11) (2000) 2609. A. Bergamaschi, et al., Nucl. Instr. and Meth. A 510 (2003) 51. M. Prest, et al., Nucl. Instr. and Meth. A 461 (2001) 435. F. Arfelli, et al., Radiology 215 (2000) 286. A. Olivo, et al., Med. Phys. 28 (8) (2001) 1610. S. Pani, et al., Proc. SPIE 4682 (2002) 228. R. Longo, et al., Nucl. Instr. and Meth. A 497 (2003) 9. S. Pani, et al., Phys. Med. Biol. 49 (2004) 1739. A. Bergamaschi, et al., Nucl. Instr. and Meth. A 535 (2004) 88.