Reconstruction-free positron emission imaging | Nature Photonics
|
Measurement of the arrival times of annihilation photons in a detector with greater precision is opening the way to new direct forms of tomographic positron emission imaging that do not require back-projection-based reconstruction techniques. You have full access to this article via your institution. Download PDF Download PDF Ever since the early days of positron emission tomography (PET) in the late 1960s through 1970s it has promised an unparalleled ability to provide quantitative biomarker images from inside living organisms with signal sensitivity in the picomolar range 1 . In principle, biochemical processes can be probed without disturbing their equilibrium state by introducing trace amounts of a radiotracer that radioactively decays, emitting positrons (the anti-particle of an electron). These positrons subsequently annihilate with an electron resulting in the emission of two coincident high-energy photons (gamma rays) that travel in opposite directions and are picked up by scintillator-based detectors. While the early applications of PET were geared towards brain and cardiac imaging 2 , subsequent development of the 18 F-FDG radiotracer, whole-body PET imaging protocols and the advent of hybrid PET/CT have pushed primary clinical applications towards oncology 3 . Throughout all these developments there has been one constant: the need for detectors that provide a full azimuthal angular coverage around the object. This is commonly done with a ring of detectors fully surrounding the patient in a cylindrical system geometry with fully 3D data acquisition and image reconstruction. Now writing in Nature Photonics , Kwon et al. experimentally report a new form of direct positron emission imaging (dPEI) that can produce 3D PET images without a need for a tomographic reconstruction or detectors providing full azimuthal angular coverage around the object 4 . In PET, a valid signal is generated when the two annihilation photons are detected in time coincidence, that is, the difference in their detection times (time-of-flight, or TOF) is within a few nanoseconds to ensure that annihilation originates from within the imaging field-of-view (FOV). The precision of this TOF measurement (TOF resolution) has not been sufficient to localize the annihilation position to within a few millimetres along the coincidence line, or line-of-response (LOR). However, by measuring all such LORs around the object being imaged (full angular coverage), the data can be mathematically inverted (back-projected) to obtain a full tomographic image (Fig. 1 ). Therefore, very precise detector timing capability has not been a necessary condition for producing a PET image, and for the last 60 years this has been the guiding principle for generating 3D volumetric PET images. Fig. 1: Schematic illustrating back-projection of point source data. From left to right: a non-TOF PET scanner with full azimuthal angular coverage, a non-TOF scanner with partial angular coverage, a traditional (200–600 ps timing resolution) TOF scanner with partial angular coverage, and dPEI with no back-projection and timing resolution of 32 ps. Light-blue regions are the PET detectors, white arrows are the collected LORs, the grey circle is the imaging FOV, and the red shaded lines overlapping with the LORs indicate back-projection. The non-TOF image with full angular coverage is a point source but the spread in back-projection leads to reduced image SNR. With partial coverage, the incomplete angular back-projections lead to elongation in the non-TOF image (with poor SNR) but the level of elongation is reduced with TOF information. The dPEI image with timing resolution of 32 ps requires no back-projection and the image is accurately generated with very good SNR. Full size image Even with modern detectors in commercial scanners that can achieve a TOF resolution as good as 200–400 picoseconds, this precision is not sufficient to eliminate the need for tomographic reconstruction using the back-projection algorithm 5 ; for example, 200 ps corresponds to 3.0 cm localization precision along a LOR, which is about an order of magnitude worse than a typical detector spatial resolution. In fact, the primary advantage of TOF information has been to reduce variance in the reconstructed image, thereby improving the image signal-to-noise ratio (SNR). Ongoing research and development in the area of PET detectors has utilized scintillation crystals coupled to silicon photomultipliers (SiPMs) to achieve improved TOF resolution. While the best commercial systems have achieved close to 200 ps, the best laboratory measurements have achieved timing resolution in the range of 60 ps — but this is with very small (low stopping power) crystals and specialized, expensive electronics that are not scalable. In principle this would achieve sub-cm localization of the annihilation point along the LOR, but it is still 2–3 times worse than the spatial resolution of 3–4 mm achieved with modern PET systems. Hence, dPEI based on current scintillation detector technology would not produce images with superior image quality. The primary limitation in these measurements has been the initial photon flux rate from the scintillation mechanism and the intrinsic timing jitter of the SiPMs (single photon timing resolution, SPTR > 100 ps). Any further improvements in this performance requires large improvements in the photodetector SPTR value, together with a higher photon flux rate than that achieved by the Lu-based scintillators. In this recent work, Kwon et. al. utilize the prompt Cherenkov photons that are emitted by energetic electrons produced in a crystal (or scintillator in the case of PET) after absorption of the 511 keV photons to perform a more precise timing measurement. The Cherenkov light emission mechanism occurs much faster than the scintillation emission process and has been evaluated for precise timing with the PET scintillator bismuth germanate (BGO), showing a best coincidence timing resolution of greater than 150 ps (refs. 6 , 7 , 8 ). A major limitation in these past studies was the sensitivity of the SiPMs in detecting the very few Cherenkov photons that are emitted (due to optical absorption) as well as the SiPM’s SPTR. The key enabler in the work by Kwon et. al. is the development and application of a very fast microchannel plate photomultiplier tube (MCP-PMT) with an intrinsic SPTR of 22 ps, far superior to that of the SiPM. Besides excellent timing characteristics, the specialized MCP-PMT used in this work also has a lead glass Cherenkov light radiator integrated as the window faceplate, allowing for a very high efficiency in detecting the emitted Cherenkov photons. Combined, the detector achieves an impressive TOF resolution of 32 ps that localizes the annihilation point with 4.8 mm resolution along the LOR — similar to the detector spatial resolution. Using this detector for the first time demonstrates high-fidelity positron annihilation imaging (dPEI) with a spatial resolution of 4.8 mm, where the images are generated by directly depositing the measured events in a 3D image matrix without tomographic reconstruction. To emphasize, the system geometry modelled here consists of two parallel plate detectors where data from only one ninth of the full azimuthal angular coverage were available. From a medical and biological imaging perspective this gain in timing resolution relative to clinical PET systems — almost an order of magnitude — opens up avenues to new imaging geometries that may be suitable for dedicated organ imaging, while allowing real-time imaging with very high SNR that can have applications ranging from molecular biomarker-guided radiotherapy to surgery. Despite its tremendous potential there are limitations that can present a challenge for the adoption of the dPEI approach as conceived in this work. Detector sensitivity is still very low due to the use of thin lead as the detector as well as the use of tungsten collimators to achieve good detector spatial resolution. This limitation can be partly alleviated by using a thicker detector material like BGO, formerly used in PET as a scintillation detector. However, removal of the tungsten collimator will require some form of position estimation within the detector that does not depend on absorption of most of the signal. Additionally, it is not clear that the 32 ps timing resolution performance can be maintained if 4-mm-wide BGO pixels (the customary size in PET) are used for position estimation, since optical reflections (or photon spread) will negatively impact the timing performance. In fact, these are the same challenges that have limited the best timing resolution achieved with scintillation detectors, for example, timing resolution of 100–120 ps (refs. 9 , 10 ) is achieved with crystals of practical size instead of the 60 ps reported for very small crystals. Hence, while highly promising, the long-term evolution of this new technology is strongly dependent on the ability to maintain this timing resolution in a detector design that is practical for routine imaging. Alternatively, one could consider using thin Lu-based scintillator slabs with SiPM readout that could be stacked to build a high-sensitivity detector. With an appropriate signal readout scheme it is conceivable to achieve a 60 ps timing resolution, though getting all the way down to 30 ps, as demonstrated with the MCP-PMT detector, may be unrealistic. The paper by Kwon et. al. highlights the pursuit of applying new technology to PET imaging systems and challenging the status quo of instrument design. It also highlights the very high bar of performance that is already established with commercial PET systems that achieve accurate quantitative images through a combination of high sensitivity, spatial resolution, and TOF, together with highly refined data correction and image generation algorithms. It took 25 years from the development of the first TOF PET systems to their adoption as the standard for clinical care because of the challenges to match excellent timing performance with correspondingly high sensitivity and spatial resolution. This paper reminds us that there is always room for improvement and that progress with this new detector technology and dPEI has an exciting future and is worth following over the next few years, particularly if new imaging designs and geometries are considered. References . 1. Brownell, G. L. et al. In Symposium on Medical Radioisotope Scintigraphy (Salzburg , August 1968) 163–176 (IAEA, 1969). 2. Phelps, M. E. et al. J. Nucl. Med. 17 , 603–612 (1976). Google Scholar ? 3. Czernin, J., Allen-Auerbach, M., Nathanson, D. & Herrmann, K. Curr. Radiol. Rep. 1 , 177–190 (2013). Article ? Google Scholar ? 4. Kwon, S. I. et al. Nat. Photon . https://doi.org/10.1038/s41566-021-00871-2 (2021). 5. Surti, S. & Karp, J. S. Phys. Med. Biol. 53 , 2911–2921 (2008). Article ? Google Scholar ? 6. Brunner, S. E. & Schaart, D. R. Phys. Med. Biol. 62 , 4421–4439 (2017). Article ? Google Scholar ? 7. Brunner, S. E. et al. IEEE Trans. Nucl. Sci. 61 , 443–447 (2014). ADS ? Article ? Google Scholar ? 8. Kwon, S. I. et al. Phys. Med. Biol. 61 , L38 (2016). Article ? Google Scholar ? 9. Cates, J. W. & Levin, C. S. Phys. Med. Biol. 61 , 2255–2264 (2016). Article ? Google Scholar ? 10. Gundacker, S. et al. Phys. Med. Biol. 65 , 025001 (2020). Article ? Google Scholar ? Download references Author information . Affiliations . Department of Radiology, University of Pennsylvania, Philadelphia, PA, USA Suleman Surti?&?Joel S. Karp Authors Suleman Surti View author publications You can also search for this author in PubMed ? Google Scholar Joel S. Karp View author publications You can also search for this author in PubMed ? Google Scholar Corresponding author . Correspondence to Suleman Surti . Ethics declarations . Competing interests . The authors declare no competing interests. Rights and permissions . Reprints and Permissions About this article . Cite this article . Surti, S., Karp, J.S. Reconstruction-free positron emission imaging. Nat. Photon. 15, 873–874 (2021). https://doi.org/10.1038/s41566-021-00915-7 Download citation Published : 29 November 2021 Issue Date : December 2021 DOI : https://doi.org/10.1038/s41566-021-00915-7 Share this article . Anyone you share the following link with will be able to read this content: Get shareable link Sorry, a shareable link is not currently available for this article. Copy to clipboard Provided by the Springer Nature SharedIt content-sharing initiative .
|
From:
|
|
监测目标主题
|
(1)
|
(1)
|
(1)
|
系统抽取主题
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
(1)
|
|