Alignment and resolution studies of a MARS scanner A. Gongadze, A. Zhemchugov, G. Chelkov, D. Kozhevnikov, I. Potrap, M. Demichev, P. Smolyanskiy, R. Abramishvili, S. Kotov, A.P. Butler1, P.H. Butler2, S.T. Bell3 aUniversity of Otago, Christchurch, NZ bUniversity of Canterbury, Christchurch, NZ 5 1 cMARS Bioengineering, 29a Clyde Road, Christchurch, NZ 0 2 dJoint Institute For Nuclear Research, Dubna r a M 1 Introduction 3 2 The MARS scanner[1] is designed for the x-ray spectroscopic study of samples ] t with the aid of computer tomography methods. Computer tomography allows the e d reconstruction of slices of an investigated sample using a set of shadow projections - s obtained for different angles. Projections in the MARS scanner are produced us- n i ing a cone x-ray beam geometry. Correct reconstruction in this scheme requires . s c precise knowledge of several geometrical parameters of a tomograph, such as dis- i s placement of a rotation axis, x-ray source position with respect to a camera, and y h camera inclinations. Use of inaccurate parameters leads to a poor sample recon- p [ struction. Non-ideal positioning of camera, x-ray source and cylindrical rotating 2 frame (gantry) itself on which these parts are located, leads to the need for to- v 2 mograph alignment. In this note we describe the alignment procedure that was 5 used to get different geometrical corrections for the reconstruction. Also, several 8 7 different estimations of the final spatial resolution for reconstructed images are 0 . presented. 1 0 5 1 2 The MARS scanner : v i X The MARS (Medipix All Resolution System) scanner (Figure 1) obtained by r a the Laboratory of Nuclear Problems of the Joint Institute for Nuclear Research, has a camera with two detectors based on gallium arsenide (compensated with 1 chromium) sensors[2] and equipped with the Medipix3.1[3] electronics. Each de- tector has 256 x 256 pixels of 55 microns and has dimensions of 14 x 14 mm. The Medipix3.1 electronics can simultaneously count the number of photons with energies above the threshold at each pixel for two different thresholds. Gallium arsenide sensors have higher detection efficiency of photons at high energies in comparison with silicon sensors. The advantages of the scheme with single pho- ton counting include the absence of dark currents and radiation dose reduction. Moreover, by changing user-settable parameters in the Medipix3.1 electronics, it is possible to select different ranges of photon energies for spectrometry. 2 Figure 1: The MARS scanner in the Laboratory of Nuclear Problems of the Joint Institute for Nuclear Research. SincetheMARStomographwasdesignedforbiomedicalresearch, thegantry with the scanning equipment (x-ray source and camera) attached to it is rotated around fixed scanned sample (Figure 2). The gantry rotation axis is horizontal. 3 A test sample (up to 100 mm in diameter and 300 mm length) is placed in the center and can be moved along the rotation axis. The camera and the source are placed on the gantry from the opposite sides and are able to move along the axis connecting each other, approaching or moving away from the sample. In order to enable scanning of large samples camera can also be moved in the direction transverse to the rotation axis and the axis connecting the source and the camera. Figure 2: Photo of the MARS equipment with a scanned sample. 2.1 The MARS tomograph reference frame To describe the geometry of the tomograph it is necessary to define the reference frames associated with the moving parts of the tomograph (Figure 3): Laboratory reference frame The Cartesian reference frame (X, Y, Z), associated with the tomograph support structure. The axes are as it is shown in the figure. 4 Sample reference frame The Cartesian reference frame (X , Y , Z ), associated with the sample, s s s which can be moved along the Z axis of the laboratory reference frame. Gantry reference frame The Cartesian reference frame (X , Y , Z ), associated with the gantry, and g g g differs from the laboratory frame by the rotation angle φ around the Z axis. Camera reference frame The Cartesian reference frame X , Y , Z ), associated with the camera that c c c can be moved along the Y axis of the gantry reference frame. This reference g frame is defined by the pixel detector plane. The Y axis is parallel to the c detector rows of pixels, the Z axis is parallel to the detector columns of c pixels. Figure 3: Reference frames associated with the moving parts of the tomograph 5 3 Mechanical precision of the movement of the to- mograph parts 3.1 Measurements using an electronic feeler gauge Prior to the alignment studies, the mechanical precision of the movements of the tomograph parts has been investigated. The main goal was to measure the dis- placement of the gantry rotation axis during its rotation and to measure the dis- placement of the gantry itself along the rotation axis. To perform these measure- ments an electronic feeler gauge was used. In the first case, the positions of the upper points of the left and right gantry supports (red arrows in Figure 4 left side) have been measured for different rotation angles. The mean value of these two measurements (left and right) gives an estimation of the displacement of the rotation axis in the local area of a measured sample. In the second case, position measurements have been performed from the near side and the far side of the edge of the left gantry support (red arrows in Figure 4 right side). In this case the mean value of the two measurements describes gantry movement along its rotation axis. Figure 4: Measurement of the rotation axis Z displacement along vertical Y-axis (left) and measurement of the gantry movement along the Z-axis (right). The results of the described mechanical measurements are presented in Fig- ure 5 and Figure 6 as a function of the gantry rotation angle. The results show that in the area of a scanned object the displacement of the gantry rotation axis 6 is withing 30 microns (green line in Figure 5), while the gantry itself moves along Z-axis by 100 microns (green line in Figure 6). These displacement values are not negligible with respect to the pixel size of the detector and will lead to additional smearing of a reconstructed image. The results of these measurements may be used in the future for correcting projections of scan depending on the angle of rotation. Figure 5: Relative position of upper points of the left (red) and right (blue) gantry supports as a function of the gantry rotation angle. The mean value (green) characterizes the displacement of the gentry rotation axis in the scanning area during rotation. 7 Figure 6: Position measurements of the "near" (red) and "far" (blue) sides of the left gantry support for different rotation angles. The mean value (green) characterizes the displacement of gantry along its rotation axis. 3.2 Measurements using angular encoder system Rotation of the gantry was studied using the optical angular encoder system REN- ISHAW RESR[4]. The system we used consists of the RESR ring (with the di- ameter of 200 mm) with graduations marked on it every 20 µm and the optical readhead with a resolution of 5 µm. The RESR ring is mounted on the shaft of the gantry and the readhead is attached to the tomograph frame. The angular resolution of this system is about 10 arc seconds. Inthenormalscanningprocedurethetomographequipmentrotatesarounda fixed sample making stops to get the x-ray projection images at different rotation angles. The gantry with its equipment attached has a considerable moment of inertia, which leads to vibrations of tomograph parts at the moment it stops. The behavior of the tomograph at this moment was investigated with the aid of RESR system (Figure 7). As shown in the figure, significant initial vibrations of the 8 system are observed. To get projections of good quality, there should be a time delay between the stop and the exposure of about half of a second. g] e e [d 0.06 gl n A 0.04 0.02 0 - 0.02 - 0.04 0 0.2 0.4 0.6 0.8 1 Time [s] Figure 7: Angular measurements after stopping the tomograph rotation. The next step of the investigation is to study the repeatability of angular measurements at the same positions of the step motor which rotates the gantry. During a scan the tomograph makes stops several times in the same angular po- sitions to get projections. For successful stitching of these projections, a high accuracy in the repeatability of angular positions is needed. Figure 8 shows the results of numerous angular measurements when tomograph is moved to positions 0◦ and 352◦. The results show that the design of the tomograph does provide high reproducibility of angular positions. 9 0 degree 352 degree ssttaarrtt ffiinniisshh s s entrie16 EEnnttrriieess 3300 entrie8 EEnnttrriieess 3300 # of MMeeaann 00..000011552255 # of 7 MMeeaann --335522 14 RRMMSS 00..000011660066 RRMMSS 00..000033666611 6 12 5 10 4 8 3 6 4 2 2 1 0 0 -0.01 -0.005 0 0.005 0.01 0.015 -352 -351.995-351.99-351.985-351.98-351.975-351.97 Gantry rotation angle [deg] Gantry rotation angle [deg] Figure 8: Repeatability of angular measurements when tomograph is moved to positions 0◦ and 352◦. The gantry of the MARS scanner is driven by a step motor via an automo- tive timing belt. The weight distribution of the equipment attached to the gantry is non-uniform and can lead to different tension of the belt for different rotation positions. The following measurements have been performed to check whether the tomograph angular positions correspond to actual angular positions. In Figure 9 the difference between measured and expected values of angular positions is pre- sented as a function of gantry rotation angle. Deviation of the measured angle is significant and its maximum approximately corresponds to the value of 500 mi- crons on the surface of the RESR ring. In principle, encoder measurements are affected not only by the ring rotation but also by the movement of the ring axis during its rotation. To account for this effect, independent measurements of the ring axis movement were performed depending on the rotation angle. The distance to the surface of the RESR ring was measured using SAFIBRA optical probe. The amplitude of this measurement does not exceed 80 microns and can not explain the value of the difference in measured rotation angle. The difference between the measured and expected values of the rotation angle can be taken into account to correct gantry positions and then used in the image reconstruction procedure with nonlinear rotation steps. 10