Importance of Precise Timing for Medical Diagnostic Devices

Transcription

1 Importance of Precise Timing for Medical Diagnostic Devices M.C.S. Williams a,b, A. Zichichi a,b,c and the CERN-Bologna group. a CERN Geneva, Switzerland b INFN and Dipartimento di Fisica e Astronomia, Università di Bologna, Bologna, Italy c Museo Storico della Fisica e Centro Studi e Ricerche E. Fermi, Roma, Italy Abstract During the last decade, fast timing has played a vital role in the field of particle physics. This has also led to an impressive series of technological developments in medical physics, especially with the TOF-PET camera. In this report we examine what needs to be done to push the precise timing limits within devices used for medical diagnostics. 1 1 Introduction The development of the Multigap Resistive Plate Chamber (MRPC) in 1996 [1] revolutionised precise timing in particle physics experiments. This device operates in a magnetic field, is easy to segment and has exceptional time resolution (a record of16 ps for a 24 gap device [2]). This detector was adopted by the ALICE heavy ion experiment as the device for precise timing: i.e. the Timeof-Flight barrel [3]. During an intense period of R&D to optimise the MRPC for the ALICE experiment, in was noted that differential readout was an essential ingredient to access time resolutions better than 100 ps. Single-ended readout generated a high level of crosstalk and no reduction of common-mode noise; this extra noise manifests as time jitter and limits the time resolution. For this reason a fast Amplifier/Discriminator ASIC (the NINO ) was designed [4,5]. This ASIC has a differential input and differential architecture through-out. Charged particles pass through the MRPC, creating ionisation in each gas gap; thus the signal is the sum of the movement of charge in each gap. The small Preprint submitted to Elsevier Science 27 January 2015

2 gas gap leads to rapid avalanche growth: and thus a very prompt and precise signal that produces the good timing signal. However if there is a single gas gap, the probability for a charged particle passed through with out creating any ionisation inside the gas gap is relatively significant. Having many gas gaps ensures that the efficiency is high and also further improves the time resolution. Such a detector is fine for particle physics experiments where the particle mass can be determined by a precise measurement of the speed of the particle. For devices designed for medical diagnosis, it is necessary to detect gamma photons rather than through-going charged particles Gamma ray detectors Within the concert of diagnostic medical imaging methodologies such as planar X-ray Radiography (RX), X-ray based Computed Tomography (CT) Magnetic Resonance Imaging (MRI) and Ultrasound (US), in vivo molecular imaging methods based on the tracking of radioactive substances in the subject (Single Photon Emission Computed Tomography-SPECT and Positron Emission Tomography-PET) has significant potential in oncology, cardiovascular disease, neurology, infectious disease and inflammation research due to their higher sensitivity and the ability to directly label and track molecules of interest (justifying the term Molecular Imaging ). Molecular imaging is increasingly becoming an essential tool due to the rapid emergence of imaging-based biomarkers that can be used for both diagnosis and new treatment approaches. In this context PET is the technique of choice due to its extremely high sensitivity. The high sensitivity of PET is due to the collimator-less simultaneous detection of two back-to-back 511keV gamma photons created by a positron annihilation from the injected positron emitters attached to the molecule(s) of interest. A line of response (LOR) is constructed by connecting the detection positions of the two photons within a ring of detectors. However, if the arrival time of the 511keV photons is precisely measured, the original position of the positron annihilation can be located along the length of the LOR. This greatly reduces the statistical noise in the image, leading to an enhanced clarity. The crucial parameter is the precision of measurement of the arrival times of the two photons at the detectors; this parameter is expressed as the Coincidence Time Resolution (CTR). In order to achieve a breakthrough in this field the CTR should be reduced to 100 ps FWHM. This means that each gamma detector is operating with a time precision (sigma) of 30 ps. To detect the 511 kev gamma a scintillating crystal is needed. The most common scintillators employed in the latest generation of TOF-PET scanners are based on Lutetium Silicates (LSO, LYSO and LFS). These are fast scintillators, with good light output, high density, and large mean atomic number Z eff. If the incoming gamma interacts in the crystal, 2

3 this will be via the photoelectric effect (producing a 511 kev electron) or by a Compton scattering (one or multiple). The interaction length of the gamma within the crystal is 12 mm, thus to obtain high detection efficiency a crystal length of 20 mm is needed. Optical photons inside the crystal travel at 0.5 the speed of light, thus it is imperative to measure precisely the position of the interaction if sub 100 ps CTR is to be obtained. Our technique to enable this measurement will be discussed below Precise Time Measurement The 511 kev gamma interacts in the scintillating crystal; if this is a photoelectric interaction, a 511 kev electron is produced. This electron loses energy moving some hundreds of micron through the crystal exciting the scintillator and producing optical photons. The arrival of the optical photons will be spread in time: the scintillation process has a rise time ( 70 ps) and a decay time ( 35 ns). To obtain the best possible timing the time of arrival of the first arriving photons needs to be measured. There are effects that can make the time measurement of the first photoelectron to have higher jitter than the following photoelectrons. Our scheme allows the independent measurement of the time of arrival of the first photo-electrons The Strip Silicon PhotoMultiplier and Slab Crystal structure The Silicon PhotoMultiplier (SiPM) consists of a matrix of photosensitive diodes; these diodes are run in Geiger mode. A single incoming photon can initiate an avalanche in a photo-diode that triggers a Geiger breakdown generating a signal on the external electrodes. Since this signal is relatively large, the front end electronics is sensitive to a single photo-diode firing. Thus these diodes are known as Single Photon Avalanche Diodes (SPAD). Typically the sensitive area of the SiPM has a size between 1 1 mm 2 and 4 4 mm 2. A key ingredient of our detector modules is the geometry of the SiPM. We are designing SiPMs in the form of a strip; each strip is 18 mm long and 0.5 mm wide as shown in the photograph (fig. 1). Sixteen of these strips are mounted on a printed circuit board. Each strip is read out at both ends with a signal taken from both anode and cathode (i.e. a differential signal is derived). The time difference between these two signals locates the position along the strip of the SPAD that fired. As shown in figure 2, these readout strips will be coupled to a crystal block. This block is divided into 3 mm wide slabs. As shown, the 16 strips will read 3

4 Fig. 1. Photograph of a five strip prototype SiPM strip out the slab. Optical photons created by the scintillation process are produced isotropically, thus the amplitude of light seen by each strip will depend on the position of the interaction of the 511 kev photon. Furthermore, since each slab has many individual strips, the individual timing of the first arriving photons will be measured. Thus an excellent time resolution can be obtained. This geometry of crystal slabs coupled to SiPM strips thus allows excellent timing and good position resolution, especially for the Depth of Interaction (DoI) that is a critical measurement for sub 100 ps CTR time resolutions. However each 511 kev gamma interaction will produce a significant amount of data, since now there are 16 strips involved. The flow of data to the computers 4

5 511 kev gamma 3 mm 11 mm Interaction point 20 mm 18 mm STRIP SiPM Fig. 2. View of the Slab-Strip module. The SiPM is arranged as a series of strips. Each Strip is 0.5 mm wide; they are set on a 0.7 pitch. The scintillating crystal is in the form of a slab; each slab has a dimension of mm dedicated to the reconstruction will be considered below Front-End electronics: the SuperNINO A critical aspect of the front-end electronics designed for fast timing is whether it is differential or single-ended. This is illustrated in figure 3. The signals of all detectors are created by the movement of charge between two electrodes; this movement creates an induced charge on one electrode, and an equal but opposite charge on the other. This creates a current to flow through the input of the front end electronics. In the case of the differential readout, this current flow is from electrode to the other. However for the single-ended read-out, the 5

6 DIFFERENTIAL READ OUT SINGLE-ENDED READ OUT Channel 1 Channel 2 Channel 3 input impedance Channel 4 detector ground ground of front-end electronics Fig. 3. Schematic view highlighting the difference between the single-ended in comparison to the differential readout. The signals are produced by the movement of charge between two electrodes. This creates a current that flows from one of electrode to the other. For the single-ended read-out, there is extra noise in the common electrode (the ground in this case) current flows into the common ground and eventually back to the other electrode. Thus the ground, that is the reference level for the front-end electrode has fast current spikes fed into it, creating noise. Also, as illustrated in figure 3, a detector built with a common electrode will have a high capacitance coupling between the readout electrodes creating cross talk and addition noise. This was observed during the R&D phase of the ALICE-TOF. This resulted in the ALICE TOF detector having both anode and cathode pickup electrodes, thus creating a differential signal. Also an ultra-fast front-end ASIC (the NINO) was designed and built. This ASIC is now used to read out the SiPM arrays and is the reference for precise timing [6]. A new ASIC is being designed dedicated for the read out of the SiPM. The original NINO was designed for use with the Multigap Resistive Plate Chamber (MRPC). The capacitance of each readout cell is 30 ps. The big difference is that the SiPM has a much higher capacitance: typically a 3 3 mm 2 active area device has a capacitance of 300 to 900 pc depending on manufacturer. The original NINO also has a simple Time-over-Threshold (ToT) used to estimate the input charge; this works well for the MRPC that has a short signal. However a SiPM coupled to a crystal has a long signal (depending on the decay 6

7 3 mm Depth of interaction 20 mm SiPM (A) SiPM (B) Fig. 4. Effect of Depth of Interaction: if 511 kev gamma interacts early in the crystal (far from SiPM) the optical photons will be delayed (η = 1.8). Information regarding the Depth of Interaction is needed for precise timing time of the scintillation light). Thus this simpler ToT circuit in the original NINO is not ideal for measuring the input charge. Given this, a new ASIC, the SuperNINO, is under design. This will have a similar differential design, but a much lower input impedance that will match the high capacitances of the SiPM. There will also be improved ToT circuit optimised for the much longer signals from SiPMs attached to crystals Depth of Interaction - Whole Body TOF-PET scanner A critical measurement that is needed for precise timing is the depth of interaction. If the 511 kev gamma interacts near the beginning of the crystal, the optical photons need to propagate along the length of the crystal to reach the SiPM photosensor. These photons are delayed since (a) the refractive index of the crystal, η, is 1.8 and (b) the path of photons can be lengthened by multiple reflection. This is shown in fig 4. Another effect that is important is that for a whole body PET scanner, the crystals are pointing towards the centre of the scanner; however there are many positrons that are not created at the centre of PET, as shown in figure 5. This creates a parallax error that will smear the position resolution and in effect create extra timing jitter. For these the two reasons above it is essential to measure the DOI to within some millimetres if a CTR of 100 ps (or better) is to be exploited. 7

8 Lsinθ L θ Crystals Readout electronics 511 kev gamma Photo-sensor Data flow Fig. 5. Schematic view of a whole body scanner A typical injection used for PET imaging usually is of 370 MBq. A typical size of a TOF-PET scanner is 20 cm depth and 80 cm diameter. This covers 25 % of the solid angle. Thus the data flow from such a TOF-PET scanner is likely to be 75 MHz. Each data record will consist of 40 words leading to a data flow of 12 GB/s. A typical data fibre can transfer 1 GB/s, thus 12 fibres would be needed for the peak intensity. Such data flow are not uncommon for LHC experiments at CERN. A bigger challenge is the processing of this data to generate an online image. This task will require a high CPU usage Conclusions Precise timing will revolutionise PET imaging; however there are significant challenges to fully exploit this. We have highlighted some of the problems above. 8

9 164 Direct Costs: Requested Grant: 9 Budget Cost Category M1-12 M13-24 M25-36 M37-48 Total M1-48 Personnel: Senior Staff 45,000 45,000 45,000 45, ,000 Post docs 192, , , , ,000 Students 148, , , ,000 Other Total Personnel: 1,540,000 Other Direct Costs: Equipment 150, , , , ,000 Consumables 50,000 50,000 50,000 50, ,000 Travel 40,000 40,000 40,000 40, ,000 Publications, etc Total Other Direct Costs: 960,000 Fig. 6. Proposed budget over 4 years 2500, This budget covers the implementation of a new geometry of SiPM; a TDC system with 5 ps time resolution; a data acquisition with fast algorithms of precise position reconstruction. The publications will be within open-access publications but will be covered within the normal running costs of the group. Extra personnel will be essential both at the post-doc level and additional students. References [1] A New type of resistive plate chamber: The Multigap RPC, E. Cerron Zeballos, I. Crotty, D. Hatzifotiadou, J. Lamas Valverde, S. Neupane, M.C.S. Williams, A. Zichichi, Nucl.Instr.Meth. A374(1996)132. [2] A 20 ps timing device: A Multigap Resistive Plate Chamber with 24 gas gaps, S. An, Y.K. Jo, J.S. Kim, M.M. Kim, D. Hatzifotiadou, M.C.S. Williams, A. Zichichi, R. Zuyeuski, Nucl. Instr. Meth. A 594(2008)39 [3] Performance of the ALICE Time-Of-Flight detector at the LHC, A. Akindinov et al., Eur. Phys. J. Plus (2013) 128: 44 DOI /epjp/i x [4] F. Anghinolfi et al. Nucl. Instr. Meth. A 452 (2004) 183. [5] F. Anghinolfi et al. Proc. IEEE Nucl. Sci. Symp., Portland, OR, Oct. 2003, vol. 1, pp [6] Systematic study of new types of Hamamatsu MPPCs read out with the NINO, K. Doroud, A. Rodriguez, M.C.S. Williams, K. Yamamoto, A. Zichichi, R. Zuyeuski, Nucl. Inst. Meth. A 753(2014)149. 9

10 10