Multi-slice quantum Computed Tomography system using a MHSP

This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 222.168.67.10

This content was downloaded on 15/10/2013 at 08:19

Please note that terms and conditions apply.

Multi-slice quantum Computed Tomography system using a MHSP

View the table of contents for this issue, or go to the journal homepage for more

2012 JINST 7 C01106

(http://iopscience.iop.org/1748-0221/7/01/C01106)

Home Search Collections Journals About Contact us My IOPscience

2012 JINST 7 C01106

PUBLISHED BY IOP PUBLISHING FOR SISSA

RECEIVED: October 4, 2011ACCEPTED: January 4, 2012

PUBLISHED: January 30, 2012

13th INTERNATIONAL WORKSHOP ON RADIATION IMAGING DETECTORS,3–7 JULY 2011,ETH ZURICH, SWITZERLAND

Multi-slice quantum Computed Tomography systemusing a MHSP

L.F.N.D. Carramate,a,1 C.A.B. Oliveira,a A.L.M. Silva,a A.M. da Silva,b

J.M.F. dos Santosc and J.F.C.A. Velosoa

aI3N, Physics Dept, University of Aveiro,3810-193 – Aveiro, Portugal

bDETI/IEETA, University of Aveiro,3810-193 - Aveiro, Portugal

cGIAN, Physics Dept, University of Coimbra,3004-516 - Coimbra, Portugal

E-mail: [email protected]

ABSTRACT: The present work reports a new system for Computed Tomography (CT) with Multi-Slice (MS) capability using the 2D-MicroHole & Strip Plate (MHSP) detector. The system is basedon the third generation of CT scanners and includes a MHSP based detector which achieves gainsabove 104 and has the capability to discriminate the interaction position of the incoming radiationin 2D. The possibility of operating in single photon counting mode with energy discriminationallows the selection of energy ranges for image reconstruction and the application of the EnergyWeighting Technique (EWT). The EWT improves the image contrast and the Signal to Noise Ratioof the images. The reconstruction of three-dimensional images of a PMMA phantom with holesfilled with chalk and air or brass are shown as a good representation of the studied objects.

KEYWORDS: X-ray detectors; Computerized Tomography (CT) and Computed Radiography (CR);Micropattern gaseous detectors (MSGC, GEM, THGEM, RETHGEM, MHSP, MICROPIC, MI-CROMEGAS, InGrid, etc)

1Corresponding author.

c© 2012 IOP Publishing Ltd and SISSA doi:10.1088/1748-0221/7/01/C01106

2012 JINST 7 C01106

Contents

1 Introduction 1

2 Experimental setup 2

3 CT reconstruction 3

4 Results 44.1 3D Image — Phantom with chalk and air 44.2 3D Image — Phantom with chalk and brass 54.3 Discussion 5

5 Conclusions 7

1 Introduction

Computed Tomography (CT) is a tomographic medical imaging modality that allows the acquisi-tion of cross-sectional images of the human body. The most common detectors used in CT ap-plications are the solid state detectors that give information about the charge deposited in a givenposition of the detector. Some gaseous detectors with high pressure xenon are also used giving thesame information [1, 2].

The development of x-ray detectors based on Micro-Pattern Gas Detectors (MPGD) hasbrought new devices with photon counting capability and energy resolution. This allows to detectand store the interaction position and the energy information of each event independently [3, 4].The MPGD have this capability and are characterized by: the possibility to define a threshold andreject the electronic noise; the absence of dead areas; an intrinsic high dynamic range; the pos-sibility of operating at moderate count rates; room temperature operation; versatility; portability;low cost and low complexity [5]. In this work, the MicroHole & Strip Plate (MHSP), a device thatprovides gains above 104, fast charge collection (10 ns), count rate capability above 0.5 MHz/mm2,capability to operate at high pressure, two-dimensional position intrinsic discrimination and goodenergy resolution (13.5% at 5.9 keV) [6], was used as the x-ray detector of this CT system. TheMHSP, being a single photon counting device, allows the construction of an energy spectrum ofthe detected X-ray photons. This information can be used to select the energy range of the pho-tons that are used for the image reconstruction and enhance some morphological structures relativeto others, depending on their attenuation coefficient. The application of the Energy WeightingTechnique (EWT) is also possible by using this information bringing an improvement of imagequality [7, 8]. The 3D reconstruction of a PMMA phantom with holes filled with chalk and air orbrass was performed using the proposed system.

The cross-sectional images in CT are obtained by acquiring views (projection images) in sev-eral angular positions of the body relative to the x-ray source-detector system coordinates. Then,

– 1 –

2012 JINST 7 C01106

the views are staked up in a sinogram and by applying an adequate transform function the cross-sectional images are constructed.

By using consecutive cross-sectional images, the reconstruction of three-dimensional (3D) im-ages by using a stack of cross-sectional images of contiguous slices is possible. The visualizationof an object surface inside the volume is obtained by selecting a window for the gray level, cor-responding to its intensity in the 3D representation. By this way, the iso-surface (correspondingto the selected window values) is displayed. As different objects have different gray values it ispossible to select different window values to visualize different surface objects [1, 2].

2 Experimental setup

The experimental setup consists of an X-ray tube with a molybdenum target, series 5000 Apogeefrom Oxford, using 30 kV acceleration voltage and filtering with a 2 mm aluminum foil, a 2D-MHSP based gaseous detector filled with xenon (at 1 atm), a PMMA [Poly(methyl methacrylate)]cylindrical phantom with two circular holes, 5 and 2 mm in diameter, and a rotation step motor,as described in reference [7] (figure 1). The aperture slit used to collimate the X-ray beam hasan area 15×25 mm2. The wide aperture allowed to acquire information of various slices of theregion of interest (ROI) in the same acquisition. As in the present setup the detector absorptiondepth is 3 mm, the detection efficiency is not high (63.5% at 6 keV, reducing to 5.75% at 17.5 keV).Nevertheless higher detection efficiencies can be achieved increasing the detector absorption depth.

Since the source-to-detector distance was about 150 cm, it was possible to consider the beamas approximately parallel due to the low angular divergence, below 0.7 [7].

The phantom has 15 mm of diameter and 30 mm of height. The holes are aligned with thelongitudinal axis of the phantom and a variety of different objects can be inserted, such as: chalk,brass or other materials. A rotational system, controlled by a stepper motor with a minimum step of1.8, provides the rotation of the phantom. It is possible to control the phantom motion time, whichis the time between view acquisitions, as well as the view acquisition time by using a dedicatedsoftware developed for this purpose [7].

The MHSP is a double-sided microstructure with a Gas Electron Multiplier (GEM) patternlike on the top side and a Micro-strip (MS) like face on the bottom side. On the cathode stripsof the bottom side, a perforated pattern of biconic holes crossing the structure is produced. Theapplication of electric fields inside the holes and between the cathode and anode strips, by applyingadequate voltage to the electrodes, enables two avalanche multiplication stages. The anode stripsof the bottom side are connected to a resistive line that allows to read the induced charge on itsboth ends. On the top side, strips perpendicular to the anode strips of the MS face are connectedto another resistive line. The 2D x-ray interaction position is then obtained by applying the chargedivision method for each of the two orthogonal resistive lines implemented in the MHSP baseddetector for this purpose. The energy of the event is determined by summing the two chargesignals collected from both ends of the resistive layer connecting the anode strips, since it providesbetter signal-to-noise ratio [9]. The electric signals from the four extremities of the two resistivelines of the 2D-MHSP were amplified by a preamplifier (Camberra 2006) and digitized with anAnalog to Digital Converter (ADC) with 4 channels, 14 bits and 100 MHz (CAEN N1728B NIM

– 2 –

2012 JINST 7 C01106

Figure 1. Schematic representation of the experimental setup. The X-ray beam is filtered through an Alfoil. Before the X-rays arrive at the detector, they are collimated into the ROI by a slit with an open area of15×25 mm2 [7].

module). The ADC was connected to a computer through an USB cable and the information aboutthe channel, amplitude and the timestamp of each pulse was stored.

A filter implemented through software (based in Matlab R©) was applied to these pulses, con-sidering that an X-ray photon interacted in the detector when four pulses (of different channels)are found to be within an appropriate time window. The coordinates of each event and its energyare determined by weighting the signals arriving from the end of each resistive line, as describedin reference [9].

3 CT reconstruction

The images presented in this work were reconstructed from 200 views that correspond to a completeturn (360) with steps of 1.8. Each view was acquired during 7 sec and the time of motion betweenthe acquisition of each view was 5 msec. The stored data corresponding to the motion of the motor

– 3 –

2012 JINST 7 C01106

was not considered, since it would introduce motion aberration effects and represents only 0.07%of the total acquisition.

The information about the x and y coordinates and the energy of the X-rays is organized andthe events are splitted according to the view to which they correspond. This is done by using theinformation about the timestamp of each event. Accordingly to the process described in refer-ence [9], for each view, the detector provides a bi-dimensional distribution of the X-ray intensitiesdetected in each region of the active area of the detector. The active area of the 2D image obtainedin this way, corresponding to the ROI, was divided in n sub-areas. Each sub-area is related with adifferent slice of the phantom. For each slice, each of the 200 2D distributions is integrated alongone of the planar directions (xx’), giving rise to 200 views. A sinogram of each slice is constructedby stacking up all the corresponding views. It was applied the EWT in the construction of theeach sinogram, as described in reference [7]. According to this method, the intensity of each pixelis obtained by summing the counts of each channel of the energy spectrum, multiplied by 1/E3,being E the energy corresponding to the channel [7]. It was demonstrated that this weighting factormaximizes the signal-to-noise ratio for energies up to 40 keV [4].

Once obtained the sinogram, the EWT cross-sectional image was reconstructed by using theiradon function available in the Image Processing Toolbox of Matlab R©. This function allows us tochoose one of the few filters in order to obtain the lowest noise for a good definition of the contours.The Shepp-Logan filter was found to be the best choice for our case. The cross-sectional images,with 90×90 pixels, were obtained using a sinogram with 128 rays×200 views.

3D images were obtained from the cross-sectional images stacked in a 3D graphic. A study ofthe window values to define the surface objects inside the phantom was done. Therefore to visualizea given surface of the represented volume, the window values representative of that surface wereselected and the corresponding iso-surface was then displayed. Since the attenuation coefficientswere known, it was possible to predict which window values are higher and which are lower.

4 Results

In this work, 3D images were computed using data acquired for the case when the chalk fills one ofthe holes of the phantom (5 mm-diameter) and air fills the other (2 mm-diameter). The case whenchalk fills the first hole and brass fills the second was also considered.

4.1 3D Image — Phantom with chalk and air

Figure 2 shows the sequence of cross-sectional images obtained for the phantom with chalk in the5 mm-diameter hole and air in the 2 mm-diameter. Since the thickness of each slice was chosen tobe 2 mm, the five images of the figure 2 correspond to 10 mm of a cross-section of the phantom,because the slices are contiguous but without overlap.

By using the cross-sectional images shown in figure 2, the method and the conditions referredabove, the 3D image presented in figure 3 was computed. For this, the window values correspond-ing to the surfaces of the PMMA, chalk and air were determined for this volume and then theiso-surface of each one of those was represented.

– 4 –

2012 JINST 7 C01106

Figure 2. Sequence of cross-sectional images obtained for the phantom with chalk and air.

Figure 3. 3D image of the phantom with chalk and air based on a stack of 5 cross-sectional images with2 mm thickness each (presented in figure 2).

4.2 3D Image — Phantom with chalk and brass

In figure 4 it is visualized the sequence of cross-sectional images now obtained for the phantomwith chalk in the 5 mm-diameter hole and brass in the 2 mm-diameter and, considering 4 slices of2 mm leading to a total of 8 mm cross-section of the phantom.

By using the cross-sectional images shown in figure 4, the 3D image presented in figure 5 wascomputed applying the same method as before.

4.3 Discussion

The 3D images computed using the data acquired for the phantom with chalk and air and of thephantom with chalk and brass show that the volumes corresponding to each of the material are wellidentified, being achieved good definition of the contours. The image quality is limited by the poorstatistics of the number of events per pixel, approximately 40 and 35, for the case of figure 2 andfigure 4, respectively. In the 3D images, shown in figures 3 and 5, the number of events per voxel

– 5 –

2012 JINST 7 C01106

Figure 4. Sequence of cross-sectional images obtained for the phantom with chalk and air.

Figure 5. 3D image of the phantom with chalk and brass based on a stack of 4 images of cross-sectionalimages with 2 mm thickness each (presented in figure 4).

is the same. Nevertheless, for both images the chalk rod is very well defined, being the first one(in figure 3) defined with better quality due to its higher statistics. Also, the small 2 mm-diameterrod of air and brass are well defined, with some artifacts in the case of air, also due to the lowstatistics in this case. The non-uniform response of the detector also introduces some artifacts,being an example of this effect the thinner diameter of the holes in top of the image than in thebottom. The artifacts resulting from the non-uniform response of the detector can be corrected byapplying the method described in reference [10]. By dividing the total active area of the detector insmall element-areas and determining the peak amplitude of the pulse height distribution obtainedfor each element-area, it is possible to construct a peak amplitude matrix. Considering the averageposition of all peaks as a reference, a correction matrix can be obtained dividing this referencefor each element of the peak amplitude matrix. The amplitude of an event occurring in a given

– 6 –

2012 JINST 7 C01106

element-area is then corrected by multiplying with the respective correction factor. This results inan improved detector energy resolution as well as in an improved global image quality [10].

Besides this, it is considered that this image has good quality and represents trustworthy theobject in study, giving rise for deeper studies relating to the image quality

5 Conclusions

The presented system provides intrinsic 2D position discrimination and gives the energy informa-tion of each single photon that interacts with the detector allowing the improvement of the imagequality without the need of extra acquisitions. The multi-slice capability of the system was demon-strated since it enables the acquisition of various cross-sectional images of the phantom simultane-ously, allowing the reduction of the acquisition time. The acquired stack of images made possibleto compute 3D images of the phantom with chalk and air and of the phantom with chalk and brassin its holes. The obtained images represent trustworthy the studied objects. It was observed thatthe statistics of the collected data is a major limitation for getting better position resolution, beingmandatory to increase it in future acquisitions, which also will allows to reduce the thickness ofthe considered slices. This first demonstration of a system that is still in its infancy indicates verygood potential for 3D CT imaging, where many improvements can be considered.

Future work will be related with the increase of the detection efficiency by using pressurizedkrypton instead of xenon as the filling gas, which will provide higher detection efficiencies forX-rays in the 15 to 30 keV range, as well as better signal-to-noise ratio of the pulses, due to thehigher gain achieved in Kr and by increasing the depth of the absorption region. The applicationof non-uniformity correction in order to minimize the image artifacts, and the optimization ofimage quality using EWT, will be considered. Systematic studies of the position resolution andapplication to biological samples are foreseen.

Acknowledgments

This work was partially supported by project CERN/FP/109283/2009 and project PTDC/FIS/113005/2009 through FEDER and COMPETE programs. L.F.N.D. Carramate, C.A.B. Oliveira andA.L.M. Silva, are supported by the FCT (Lisbon) scholarships: SFRH/BD/71429/2010, SFRH/BD/36562/2007 and SFRH/BD/61862/2009, respectively.

References

[1] T.M. Buzug, Computed Tomography: From Photon Statistics to Modern Cone-Beam CT, Springer,Berlin/Heidelberg, Germany (2008).

[2] J. Hsieh, Computed Tomography: Principles, Design, Artifacts and Recent Advances, 2a ed., SPIEPress Book and John Wiley & Sons, Inc., Washington-New Jersey (2009).

[3] P.M. Shikhaliev, Energy — Resolved Computed Tomography: first experimental results, Phys. Med.Biol. 53 (2008) 55954.

[4] J. Giersch et al., The influence of energy weighting on X-ray imaging quality, Nucl. Instrum. Meth. A531 (2004) 68.

– 7 –

2012 JINST 7 C01106

[5] J.F.C.A. Veloso et al., A proposed new microstructure for gas radiation detectors: theMicro-Hole-and-Strip Plate, Rev. Sci. Instrum. 71 (2000) 2371.

[6] J.F.C.A. Veloso et al., High-rate operation of the Micro-Hole and Strip Plate gas detector, Nucl.Instrum. Meth. A 580 (2007) 362.

[7] L.F.N.D. Carramate et al., Energy weighting technique in Quantum Computed Tomography using aMPGD, 2010 JINST 6 C02002.

[8] C.A.B. Oliveira et al., Energy Weighting in a 2D-MHSP X-Ray Single Photon Detector, IEEE Trans.Nucl. Sci. 57 (2010) 938.

[9] H.N. da Luz et al., Single photon counting x-ray imaging system using a Micro Hole and Strip Plate,IEEE Trans. Nucl. Sci. 55 (2008) 2341.

[10] A.L.M. Silva et al, EDXRF imaging of Pb in glazed ceramics using a micropattern gas detector,Anal. Bioanal. Chem. 395 (2009) 2073.

– 8 –