GE Healthcare Designing an MR compatible Time of Flight PET Detector Floris Jansen, PhD, Chief Engineer GE Healthcare There is excitement across the industry regarding the clinical potential of a hybrid PET/MR system. Many clinicians and researchers believe that PET/MR, by bringing together two imaging modalities with very use in oncology and neurology. The modality could also help scientists probe the relationship between structure and function more deeply. Yet, building a high quality PET/MR is not as simple as bringing together two imaging devices. The technologies traditionally used in a PET detector are not compatible with the environment of an MR scanner, and this creates a series of engineering challenges. This article discusses these challenges together with engineering approaches that have been taken to develop a high resolution, high sensitivity scanner that permits simultaneous PET/ MR imaging without compromising image quality, on a platform designed to accommodate and advance future applications. The MR environment poses several challenges to the caused by gradient induced eddy currents, the vibrations due to forces on the gradient coil, the high power RF electromagnetic radiation, and the lack of information about attenuation (no CT data). Further, the MR receive system is extremely sensitive to any electronic noise, meaning that the PET detector must not radiate any noise at the frequencies of interest. Detector One goal of a PET/MR system is to reduce radiation fraction of the patient dose is a result of the CT radiation even when the CT is used for attenuation correction only. Thus, the elimination of the CT scanner is a good step towards dose reduction. To further reduce dose, it is necessary to reduce the amount of injected PET agent; it is possible to do this without adversely image quality by increasing the sensitivity of the PET detector. This can be done by reducing the diameter of the detector ring and by increase the solid angle of the detector as seen from the patient. At the same time, the scintillator annihilation photons. Conventionally, PET detectors have been made with photomultipliers (PMTs), which provide low noise and high gain. However, a PMT will not work correctly any attempt at shielding would cause unacceptable and reduce the patient opening in an integrated (simultaneous) system.
There are two alternatives to developing a highly sensitive PET detector without a photo multiplier: direct conversion radiation detectors, or solid-state photosensors. For the GE S PET/MR, the latter solution was chosen, because direct conversion materials typically used. coupled to a silicon (solid state) photomultiplier (SiPM). The scintillator can provide good stopping power, while the SiPM is very compact and can The SiPM is a relatively new technology. It is a silicon device that has been subdivided into many small cells each a Geiger-mode avalanche diode device that will break down after absorbing a single photon, and that will turn that single photon into an electrical 6 electrons. By making a large close-packed array of such devices, and summing their outputs into a single electrical signal, one can make a photosensor with consistent gain and transit time of virtually arbitrary size. For optimal performance, one needs the sensor to have the following properties: high gain: the signal from the sensor needs to subsequent stages of the electronics low noise: a low number of dark counts reduces the uncertainty in the signal, and allows one to sum signals over larger areas high speed: to get good timing resolution, rise time of the sensor output must be as short as possible probability that a incident photon is turned into an electrical signal, the higher the signal quality (high speed, low noise) linearity: in order to get good energy resolution, the number of cells must be large compared to the number of incident photons, so the probability of two photons hitting the same photo cell is small stability: performance of the sensor must be predictable over a long period of time While thinking about the above, one has to keep in mind cost, power dissipation, complexity, dead time, and other factors. Some of the optimization choices that have to be made include: Cell size: as individual cells get larger, the relative amount of dead space is reduced. This in turn time, larger cells will produce a larger signal during breakdown (good), but have a higher probability of producing cross talk (bad) and a narrower operating voltage (bad). Pixel size: individual cells can be combined into pixels. The larger the pixel, the larger the device capacitance: this makes the signal slower (bad); smaller pixels result in more dead space (bad), higher spatial decoding capability (good, if you need it) and higher electronic channel count (bad). Pixel size is further constrained by the sizes of silicon dies that can be manufactured economically and consistently. Crystal size: making a scintillator crystal smaller potentially improves spatial resolution, although radiation scatter inside the detector (over 60% of scintillator before their energy is fully absorbed) limits the usefulness of making crystals too small. Per unit volume, small crystals are also much more expensive. At the same time, keeping crystals long Based on these considerations, the GE S PET/ mm small enough for excellent spatial resolution, and long enough to have the stopping power needed for great sensitivity.
Block size: crystals are packed into blocks, with a certain amount of light sharing between crystals. Making the blocks small increases the probability of inter-block scatter (bad), but improves the system dead time (good). Smaller blocks also increase the power consumption and temperature of the detector. gain and noise (high temperature = high noise). Many experiments and simulations were performed and other considerations, resulting in a Time-Of- Flight PET detector with timing resolution of less than 400 ps FWHM. Electronics In order to make the most of the optimized SiPM detector, we had to develop highly specialized circuit) that provides pulse shaping, gain control, trigger validation, and temperature measurement. The integrated electronics corrects for crystal light output, temperature, dead time, and other parameters. Some of the key innovations in the integrated electronics include: A novel (patent pending) design of the input stage topology with feedback that achieves extremely speed despite the input capacitance presented by the SiPM. A dual trigger (time/validate) scheme that permits triggering at the level of two photons without creating excessive system dead time Gain control of individual pixels in the SiPM array Pulse shape compensation to optimize timing performance independent of count rate Active baseline restoration to maintain performance at high count rates Once the event position and energy have been turned into an electrical signal, it is important to convert this information to a digital form as quickly as possible. For Time of Flight PET, integrated electronics are essential to minimize noise and signal signals and making all corrections right in the bore: once the signals consist of only ones and zeros, any corrections can be applied easily; and the fully corrected signal can be transmitted over optical interference with the MR receive system. Compton Scatter Recovery scintillator are initially Compton scattered before their energy is fully absorbed. This has two very spatial resolution of the detector: as crystal size is reduced, the probability that events scatter between scatter from one block to an adjacent block. The larger the block, the smaller this probability. For photon from a coincidence pair is scattered, the coincidence can be lost. This means that the these adjacent coincidences and recombining them into valid events is one of the technology breakthroughs that allow the S PET/MR PET detector such high sensitivity.
S 1 S 2 by-second basis. Changes in heat impact the PET energy peak (of the PET signal) by more than 6%. Therefore it is crucial to control temperature, and to compensate for residual changes. a b c photons interact with the detector, several things can interaction: photo-electric absorption; (b) the photon is Compton scattered, but all the energy is deposited in the same block; (c) the photon is scattered to an adjacent true event, so it gets rejected. Compton Scatter Recovery restores the information about the incident photon so it can be used for event processing. In the GE S PET/MR, a sophisticated water- that are caused by gradient and RF power transients distributed throughout the PET ring. The values from these thermistors are converted to a precise map of detector temperature. This detailed map is then combined with the known behavior of the detector and electronics to compensate for the thermal variations. The result is a detector that is extremely stable, even as MR pulse sequences vary widely. Thermal challenges There are several sources of heat in a PET/MR scanner that can impact performance. While heat dissipation in the electronics and heating in the sensor should be accounted for, it is the heating caused by the gradient for - both because they substantial, and because their magnitude is variable. The pulsing of MR gradients induces eddy currents in the RF shielding that gives and time. Even a system with high thermal capacity temperature under such conditions. Since this heat load also depends on the gradients being used, it is important to know the temperature on a second-
Electrical interference challenges The receive circuits in an MR scanner are exquisitely sensitive, and any electrical noise in the environment is likely to show up in the image generated. The high speed electronics inside a PET detector inherently produce a lot of spurious electromagnetic energy. Since the high speed electronics are in proximity to the receive coils of the MRI, there are two options to prevent interference: shield the detector, or stop the noise during MRI receive cycle. as it is a goal of PET/MR to image the same part of the patient with both modalities at the same time. Therefore, excellent electromagnetic shielding is needed. This has to be done at multiple levels: in the electrical circuit itself, it requires careful consideration of layout to minimize the appearance of (ground) loops, and placing sensitive traces between ground planes; at the level of the enclosure, it implies a hermetic enclosure with multiple thin layers of conductors that shield the high frequency signals without providing a good conduction path for eddy currents; for the power cables, it means multiple layers of shielding; and for the connectors, a combination of mechanical locking mechanisms, RF gaskets, and additional decoupling is needed to eliminate any RF leakage path. Finally, all high speed communication to the detector is done optically to further reduce the potential for interference. hese precautions make the detector (RF and gradient) the MR system; and ensure no noise from the detector can interfere with the image formation on the MR side. Mechanical issues When the gradient coil is driven with large currents, vibrations that are perceived as loud noise. Since the PET detector is mounted in very close proximity to the gradient, it is potentially exposed to these vibrations, which may impact the reliability of the detector (in particular the optical coupling between scintillator and photo sensor). Special care has to be taken to develop a mounting mechanism for the detector that reduces the mechanical coupling, A further challenge is posed by the requirements on the patient table. For PET attenuation correction it is necessary to know where the receive coils are; for the coil underneath the patient table. In order to minimize the distance from the coil to the patient, this means the table must be very thin, yet strong enough to carry of view of the PET detector degrades the image quality (because of attenuation and scatter), it is not possible to include strong support elements at the edge of the be very precise and repeatable, so that images from single patient volume without errors in registration. All these challenges have been addressed with the design of a new Kevlar-reinforced ultra-thin table with a dual positioning drive and a novel positioning mechanism based on a pull string. PET/MR System PET Detector Modules MR RF Shield with dip in the center Gradient coils RF shield Body coil former electronics scintillator MR RF Body Coil Support Ribs Fig.1. Schematic of the prototype PET/MR system RF conductor End shield
Since the PET detector is on the outside of the body coil, attenuation of the body coil adds a further complication. This was overcome by creating a tube-like structure that carries the body coil on its inner surface and the RF shield on its outer surface. The PET detector is mounted on the outer surface, in an area where the tube was thinned to minimize the of the body coil (since it has a smaller diameter), and ensures good alignment of body coil and RF shield (since they are mounted on the same structure). This Conclusion Integrating a fast photosensor with a thick scintillator and fast, low-noise electronics inside a well shielded enclosure inside an MR scanner, and paying careful attention to issues of mechanical isolation, attenuation, and heating, it has been proven possible to design a Time of Flight PET/MR scanner that high quality imaging while providing simultaneous any image degradation. features shown herein, or discontinue any products described at any time without notice or obligation. Please contact your GE representative for the most current information. GE, GE Monogram and imagination at work are trademarks of General Electric Company. *Trademarks of General Electric Company. GE Healthcare, a division of General Electric Company. GE Healthcare U.S.A