Ultra-Time-Of-Flight Positron Emission Tomography with multi-channel digital Silicon photomultipliers and photonic crystals


UTOFPET is a European H2020 research project within the PHOTONIC SENSING call. The international consortium is lead by the University of Pisa (Italy) and consists further of Ghent University (Belgium), AGE Scientific s.r.l (Italy), Ecole polytechnique federal de Lausanne (Suisse) and MOLECUBES NV (Belgium).

Multi-parametric molecular imaging (MMI) is an imaging modality largely based on clinical PET that allowed many of the recent discoveries on new molecular probes and the associated models. MMI is a fundamental pillar of the so-called personalised medicine that is expected to allow the delivery of “the right treatment to the right patient at the right time”. There is a general consensus on the necessity of increasing MMI sensitivity by at least a factor of ten in order to exploit its full potential. This project aims at providing a new type of time-of-flight (TOF) PET demonstrator that is expected to drive such a big leap in sensitivity. In TOF-PET, the timing information of each annihilation photon coming from the source under examination is used to improve the statistical information regarding the position of the source itself. It can be demonstrated that TOF can introduce a gain in the PET sensitivity that is approximately proportional to the inverse of the resolving time of its photon detectors. Therefore, one way to increase sensitivity is by improving the TOF resolution.

This project includes developments on both TOF-PET hardware and software aimed at improving the TOF resolution beyond the level achievable today with analogue SiPMs. A PET prototype will be constructed based on the emerging type of photosensor devices known as multichannel digital SiPM and the latest scintillating materials and photonic crystals technologies. A scalable acquisition system will be developed to handle the high data rate produced by each detector.

The system will provide timing performances beyond the state of the art, well below 300-400 ps, which will translate into higher imaging sensitivity, higher contrast, lower noise, faster convergence and improved robustness with respect to state of the art PET and TOF-PET alternatives.


Background and vision

Integrated positron emission tomography (PET) and magnetic resonance imaging (MRI) represent the state of the art in multi-parametric in-vivo molecular imaging (MMI), i.e., one of the emerging technologies in medical applications and “personalised medicine”. New PET molecular probes and associated small animal imaging models are under development to find and quantify subtle molecular and cellular processes such as protein interactions, cancer cell interactions with the immune system, therapeutic stem cells proliferation, gene delivery and expression in living subjects [1].

PET provides unparalleled sensitivity, especially when capable of measuring the difference in time of flight (TOF) of the two positron-electron annihilation photons. MRI provides morphological images with high resolution and exceptionally high soft-tissue contrast. When fully integrated, TOF-PET/MRI gives the most valuable molecular information available today on a living subject. By providing the physiological characteristics and biological behaviour of the diseased lesion, in addition to its location and extent, it is believed that MMI is close to taking diagnostics beyond the early detection of diseases. It is also expected that MMI will allow treating the disease not only on the basis of the symptoms but also of the biochemical profile of an individual’s disease state [2]. To achieve these goals, the sensitivity offered by current TOF-PET/MRI needs a leap ahead of one order of magnitude [3]. PET sensitivity depends on the detectors geometry: it increases with the detectors size and with their proximity to the subject. In the U.S., the NIH-funded project Explorer is bringing this aspect to its limits with a detector that is 2 meters long, 80 cm inner diameter, and whose expected sensitivity is 43 kcps/MBq – roughly 2 to 7 times higher than state of the art whole-body TOF-PET scanners [4,5]. Such a system is said to be “totalbody”, to hint at the fact that the axial extent of the scanner is much longer than usual, as opposed to the more conventional “whole-body” designation. The retained system concept is effective but intrinsically expensive, as it is primarily based on a multiplication of conventional block detectors.

A European research approach to the same problem follows the alternative direction of improving the TOF technology by bringing its coincidence time resolution (CTR) to the 10 ps limit [6]. The principle is that the signal-to-noise ratio gain in a TOF-PET system is inversely proportional to the square root of the CTR [7].

This signal-to-noise gain translates to a gain in sensitivity. A CTR of 10 ps would be roughly 30 times better than any commercial system available today [5], and 50 times better than the Explorer PET CTR up to specs (530 ps simulated [8]). Therefore, a 10 ps TOF-PET system could have virtually the same sensitivity as the Explorer, with 7 times fewer detectors, while still keeping the MR-compatibility. The feasibility of sub-100 ps CTR has been already demonstrated with small scintillating calcium/cerium codoped L(Y)SO:Ca,Ce scintillating crystals [9]. Modern Silicon photomultipliers (SiPM) proved in laboratory that timing resolutions below 100 ps can be achieved in PET. Digital SiPMs (dSiPMs) brought to laboratory practice the estimation of 511 keV photons arrival time from the triggering time of the first few scintillation photons [10]. This practice led to the proposal of building networks of single photon avalanche diodes (SPAD-Nets) to acquire every single scintillation photon independently and thus pulling the timing resolution down to the fundamental physical limit of the scintillator [11,12]. This project intends to leverage this latest type of photodetectors to push the state of the art of PET technology towards the 10 ps limit. In order to do so, we intend to develop and validate in laboratory a novel concept photodetector design based on single photon sensing technologies, distributed on-chip real-time photon processing and advanced TOF-grade reconstruction techniques. This development effort is planned for a time frame of three years.

[1] Gu E. et al., Theranostics. 2012;2:335–345.
[2] White paper, ESR, Insights Imaging. 2015;6:141–155.
[3] Levin C.S., Proc. IEEE. 2008;96:439–467.
[4] Cherry S.R. et al., http://dx.doi.org/10.1117/12.2227151.
[5] Vandenberghe S. et al., EJNMMI Phys. 2016;3:3.
[6] Gundacker S. et al., PMB 2016;61:2802.
[7] Conti M., EJNMMI. 2011;38:1147–1157.
[8] Zhang X. et al., J. Nucl. Med. 2014;55:269–269.
[9] Nemallapudi M.V. et al., PMB 2015;60:4635.
[10] Seifert S. et al., PMB 2012;57:1797.
[11] Fishburn M.W. et al.,IEEE TNS 2010;57:2549–2557.
[12] Venialgo E. et al., PMB 2015;60:2435.

Recent News

19-20/09/2019 Consortium Meeting University of PISA

12-13/09/2019 UTOFPET @ Total Body PET Workshop Brussels

13/12/2018 Consortium Meeting Ghent University/MOLECUBES

3/7/2018 Project Kick-off meeting University of PISA


Scientific Challenges

The timely introduction of techniques for increasing the sensitivity of PET is mandatory today to have the immediate impact of better image quality, reduced radiation dose and/or reduced scanning time. The scientific challenge of this project will be to demonstrate that this problem can be tackled by pushing the TOF technique to its limit, exploiting the effect of sensitivity gain the TOF-PET can offer.

It can be demonstrated that in a TOF-PET scanner with Δt timing uncertainty, the gain in sensitivity when imaging an object of diameter D is G = 2⋅D / (c⋅Δt) (c is the speed of light). This estimate is approximate, but it gives a clear indication of the importance of timing resolution . More optimistic estimations of G include the beneficial role of TOF on randoms rejection. TOF-PET image reconstruction is also faster and its convergence is more stable. Moreover, a higher robustness against mismatched attenuation correction, erroneous normalisation and poorly estimated scatter correction has been observed.

Achieving the objective of a CTR lower than 200 ps will result in a sensitivity gain of about 5 with respect to conventional PET in objects of the size of the brain making a major step forward in the direction of the ambitious goal of increasing the sensitivity of PET by one order of magnitude. In order to achieve this result, however, the scintillation process, the light transport and the photosensing characteristics will have to be fully understood. It is important to note that these processes are today still object of research and precise timing models have been started being published only in the last decade.

Technical Challenges

There are two major technical difficulties in achieving timing resolutions less than a few hundreds of picoseconds. The first is related to the scintillation process: in order to derive precisely the arrival time of the 511 keV annihilation photon, the scintillator must have a high light yield (more than 30 ph/keV) a low decay time (less than 40 ns) and most of the scintillation light must reach the photosensor in the smallest interval as possible. Investigation on new scintillating materials is being carried out by several research groups worldwide and it is expected to result in new products in the next few years. In this project, we will focus in optimising the scintillating light transport so as to reduce the time spread of scintillation photons from the crystal to the photosensor. To achieve a CTR lower than 200 ps, the scintillator needs to be simulated against several parameters, such as photon absorption efficiency, photon travel path distribution, intrinsic spatial resolution and depth-of-interaction capabilities. Moreover, it might be necessary to improve the optical surface between the scintillator itself and the photosensor by using advanced coupling materials such as the photonic crystals. The second major technical difficulty is related to the time discrimination of the scintillation pulse. It has been shown in laboratory and with small scintillators (still unsuitable for a real world PET system) that it is possible to pull down the single photon timing resolution to roughly 100 ps with highly integrated discriminators at the photosensor level. This has been achieved with silicon multipliers (SiPM) technologies. SiPMs, in their analogue embodiment, are large arrays of single-photon avalanche diodes (SPADs) whose avalanche currents are combined together in a single output. SiPMs can typically detect optical photons generated in scintillators with an increased time resolution with respect to conventional photomultiplier tubes. A new type of “digital” SiPMs (dSiPMs) brought to laboratory practice the estimation of 511 keV photons arrival time from the triggering time of the first few scintillation photons. The capability of discriminating the arrival time of each of the first scintillation photons improves the timing resolution [21], [22]. However, integrating time-to-digital converters introduces additional difficulties: 1) the dead area of the photosensor increases due to the space required by the TDCs; 2) the power consumption grows sensibly, thus leading to cooling difficulties to keep the dark counts low; 3) the improvement in timing resolution is subject to a correct statistical model of the light transport from the scintillator to the sensor. A third difficulty comes as a consequence of the first two. If a PET detector is made of scintillators with high light yield, for each scintillation event we can expect a relatively high number of input photons. However, a photosensor with embedded discriminating electronics will require a fixed amount of space (dead space) in the Silicon die for each SPAD that composes the SiPM. The ratio between sensitive area and the total space including dead space is often referred to as fill factor. Photosensors with relatively low fill factor have lower detection efficiencies. Therefore, one might be tempted to increase the size of the SPAD to increase the fill factor. This would result in smaller input dynamic ranges, because the number of SPADs per unit active area decreases thus producing an early saturation effect, which in turn limits the use of high light yield scintillators.


The UTOFPET project comprises a consortium capable of achieving all its ambitious and multidisciplinary objectives due to its complementarity which covers all the necessary roles and expertise. The partners have solid scientific backgrounds from academia and highly qualified engineering and commercial expertise from SME.

Consistent with our scientific objectives, we have assembled a consortium of 5 expert and highly complementary partners, with a good balance between scientific and industrial partners, in six key R&D areas:

  1. Integrated photosensor technology and VLSI design,
  2. Nuclear instrumentation electronics and digital data acquisition,
  3. PET detector design and testing,
  4. PET system simulation and scintillator modelling,
  5. TOF image reconstruction and
  6. Dissemination, commercial exploitation and outreach.

The partners from academia and research institutions represent excellence in the research in Europe in their respective fields of expertise at this time, with an optimal balance between physics, applied technology, and system engineering. In addition to the academic research community represented by 3 participating institutions, 2 SMEs characterized by a high level of innovation and excellent research attitude are also active partners in the project. The role of enterprises in UTOFPET is not those of mere observers but all of them have strong leading responsibilities in various aspects of the project.


The research group at UNIPI is involved in several research projects covering various fields of functional imaging, medical physics and detector technology, such as PET, CT and MRI.

The research group worked to the development of a preclinical multimodality system IRIS PET/CT that is commercialised by Inviscan s.a.s., France. The group has strong expertise on SiPM technology and applications. Other active projects include the development of SiPM-based PET system for dose

monitoring in hadrontherapy (DoPET and INSIDE projects) and for brain imaging PET/MR/EEG (FP7- Cooperation TRIMAGE project, which UNIPI coordinates). The group has several years of experience in applied physics for PET, FPGA-based PET data acquisition design and EU-funded projects management.

Principal Investigator: Prof. Nicola Belcari (Male), Associate Professor of Medical Physics at UNIPI. He received the M.S. from University of Pisa in 1999, and the Ph.D. in Applied Physics in 2003. Since 2000, he has conducted his research activity in the field of Medical Physics. His research interests include the development of new imaging detectors for preclinical molecular imaging, hadrontherapy monitoring and brain PET/MR. He was responsible for the development of a preclinical PET/CT system that is now commercialized by Inviscan s.a.s. (France). He is now responsible, as WP leader, for the design and construction of the SiPM-based PET component of the TRIMAGE brain scanner. He is author/co-author of about 100 papers on international journals and conference proceeding on nuclear physics and medical physics, and he holds two patents. Prof. Belcari is a member of the Physics Committee of the EANM.

Co-PI: Dr. Giancarlo Sportelli (Male), Researcher in Medical Physics at UNIPI. He received his M.Eng. from the Technical University of Bari in 2006, and his Ph.D. in 2010 from the Technical University of Madrid. Since 2011, he is working at UNIPI on PET for dedicated applications, such as mammography, PET monitoring in hadron therapy, high-resolution PET for small animal imaging and PET/MR for brain imaging. He has more than ten-years expertise in DAQ systems design and image reconstruction for PET. His main skills are related to the fields of digital electronics, FPGA architectures and parallel computation.

Get in touch

If you want to know more about the UTOFPET consortium or its partners, feel free to contact us here