Technologies
1: Improved PET Detector Sensitivity
ToF- Benefit: The probability to detect the two coincident gammas of a positron annihilation scales with the solid angle of the PET scintillator array and its thickness. As this single crystal material is very expensive and the available room for a PET detector inside the MR is limited, the placement of a high sensitive whole-body PET/MR system is almost impossible. But when exploring the arrival time difference of the two gammas, more information is available per event leading to a faster conversion of the image. This is equivalent to a higher sensitivity of the system and increases the noise equivalent count rate (NEC), depending on the object size and the achievable resolving time within the Line of Response (LOR). The ToF principle and benefit is shown in Figure 1.2.1.
- The left plot shows the ToF principle. The right plot shows the effective sensitivity gain, expressed by the ratio in NEC: pixelated SiPM ToF-PET readout compared to a PMT based non-ToF detector - assuming equal scintillator geometry
The count rate capability of PMT based systems is normally limited by the necessary light sharing principle, which produces signal pile-up. The NEC increases by removing the pile-up with a pixelated SiPM readout. But really dramatic improvements can be seen with the addition of the ToF benefit based on 200 ps timing resolution. The gain in sensitivity is 10 for an average patient size (27 cm diameter phantom) with an average activity (0.1 µCi/cc) used for FDG oncology studies.
In real life, this timing resolution can only be measured by having adequate scintillator, detector and electronics. As LYSO is currently the of the best scintillators for PET with high stopping power, fast decay time and very good timing characteristics, the project relies on this as a main component. Alternative scintillators like LaBr3 or LuI2 are promising, but mechanically very delicate to handle as they are hygroscopic and brittle.
Solid-State detection based on SiPMs:
The fast detection of the scintillator light is achieved by dedicated Avalanche-Photo-Diodes (APDs), so called SiPMs. Compared to APDs, which operate in the amplification mode with a gain of ≈100, SiPMs consists of parallel connected small APDs (10-50 µm), which are driven above the breakdown voltage in Geiger-Mode, see Figure 1.2.2. Each charge generating an avalanche produces very precise pulses with a constant gain of ≈106 and a very low timing jitter (<100 ps).
As the characteristics of an SiPM pixel is determined by the summing several cells, a very clean process is essential to reduce the dark-counts and increase yield. So far, no device of the desired 3-4 mm pixel is available. Also, most published SiPM designs are based on a straight forward n-on-p structure, which is not sensitive enough for the blue light of the scintillator. PDEs of <10 % are the consequence, yielding PET performance worse than PMT bases systems.
It is therefore a big challenge to establish a process with sufficient yield and develop a design with a PDE > 60 %, which is in principle possible.
- The SiPM working principle showing an electrically parallel connected set of small APD cells driven in Geiger-Mode, each having an individual quench resistor, Source: FBK
Multi-Channel Readout with TDC/ADC ASIC:
The solid-state readout for SiPMs requires ~100 times the number of readout channels compared to current PMT based systems as SiPMs are 3-4 mm compared to PMTs of 30-40 mm diameter. The challenge is to design a highly integrated low power ASIC to perform time-stamping and energy estimation for each channel. An initial two channel version of this ASIC has already proven to yield CRT = 340 ps and DE/E = 13 % for a LYSO on PMT setup with an internal jitter of 105 ps [3]. A 16-channel version of this TDC/ADC is currently being investigated with an intrinsic timing jitter of 60 ps and an intrinsic energy resolution of 0.5 %, operating at 80 mW/channel (see Fig.3)
As all noise input is translated into timing and energy jitter, the layout for very low cross talk is a real challenge. All analogue (and most digital) components are based on differential designs in order to be very insensitive to electronic noise, spikes, and MR related RF induction, as well as eddy currents due to the gradient field. Also, the analogue bandwidth has to be tailored to minimally overlap with MR frequencies. Appropriate packaging of the ASIC is essential to maintain a short analogue path to the sensor, as well as achieving a small form factor needed for whole-body PET inside the MR. Very important is the reduction of the power to reduce the effort of the cooling system and allow room temperature operation, which is essential for the SiPMs to guarantee low dark counts.
- The block diagram of existing multi-channel ToF-PET ASIC designed by Peter Fischer (UH) for Philips
2: Integration of Whole-body PET/MR for simultaneous imaging
Major modification to an existing MR system have to be made to house the PET detector, provide the infrastructure like power, communication and cooling. They all have to maintain the existing MR image quality, which requires a minimally disturbing PET detector.
On the other hand, the PET data acquisition has to be undisturbed by the MR, including magnetic field, gradient- and RF pulses. The minimization of the crosstalk for concurrent image acquisition is the biggest research challenge of the project and requires a profound understanding of the two complex systems. This includes a joint multi-disciplinary approach:
It starts with the selection of the correct of components, which have to be non-magnetic and also transparent to x-ray (depending on the position). Knowledge, experience and powerful simulation tools are necessary to design an effective shielding concept, which has be transparent for gradient and resonant for the RF. Noise sources have to be identified and quantified by dedicated test procedures like dedicated MR sequences or by putting the PET system in various modes.
Innovative solutions have to be found to minimize eddy currents within the PET electronics and the shielding, as it causes vibrations and lead to reduced MR image quality. Last but not least dynamic mechanical simulations help to find optimal mechanical suspension of the integrated PET system to guarantee a stable operation.
To ensure a fast and effective solution of concurrent PET/MR image acquisition, a straight forward three step approach is suggested here - and if will be explained in more details in work plan strategy in Section B1.3: First, a small test ring is investigated before a full animal PET/MR test system is realized in the second year. The third year addressed the whole-body ToF-PET/MR system for clinical tests.
3: Real-time motion correction
PET images allow an excellent quantitative view towards biochemical and physiological processes due to the outstanding pico-mol sensitivity, but they contains only minor anatomical information. It is thus beneficial to combine the functional PET image with an image modality delivering anatomical information. The commercial success of combined PET/CT systems are a good example of that. However, the advantages of the combination of PET and MR is much more than a replacement of an anatomic modality. Replacing CT by MR will allow to eliminate the CT dose exposure, while enabling real-time monitoring of organ motion as it provides truly simultaneous imaging. Nonetheless, organ motion still poses a severe problem in PET imaging in that it reduces sensitivity by blurring out small regions of radiotracer uptake and dramatically reducing their contrast.
Therefore, motion compensated reconstruction algorithms aim at the exploitation of all data to improve the sensitivity. Algorithms will be developed and validated using phantom and small animal data from an existing PET/MR system at KCL and compared to stand alone whole-body MR and PET/CT systems. Realistic human data will be obtained from a ‘GATE’-validated PET simulation using real dynamic human MR data as input. The real-time implementation of the motion-compensation into the PET reconstruction will be a challenging task and will require excellent knowledge of the data flows from both systems.
The MR also provides additional functional information, which is not available from CT: imaging of diffusion tensors, temperature, and tissue elasticity, image pH value or angiogenesis, up to metabolite concentrations by MR spectroscopic imaging. This data can conveniently be acquired while doing the PET imaging. A tool to display and analyse large 4D multi-functional data sets will be generated.
Advances the project will bring about
Whole body PET/MR with dramatically improved time-of-flight capability will enhance existing clinical and research applications, and will enable totally new ones. For many clinical studies (including gynaecological, prostate and brain tumours, soft tissue sarcoma, and paediatrics) MR is the preferred anatomical modality, due to the lack of radiation dose and excellent soft tissue contrast. Consequently, PET/MR is very likely to replace PET/CT in the future.
Simultaneous acquisition of complementary functional information from both PET and MR will increase the power of studies where interventions such as drug administration, sedation, experimental stimuli etc. are required. For example, in the heart, the simultaneous highly accurate measurements of both perfusion and wall motion in response to dobutamine could have a major clinical impact. In research applications PET/MR will allow direct translation between pre-clinical and human studies. Correction for subject motion, hand-in-hand with dramatically enhanced time-of-flight capability will greatly improve PET image quality, particularly in whole-body scanning, resulting in greater sensitivity and specificity for lesion detection and improved quantification of tracer uptake – these are key issues in the management of lung and many other cancers. The technical developments that we propose in this project constitute a major step towards the realisation of all these new applications.
Key aspects that will ensure advances are
- the proposed PET/MR technology is based on next generation photodetectors that are both ideally suited for PET/MR whilst simultaneously providing dramatically enhanced time-of-flight capability
- developping the novel data acquisition and processing methods necessary to ensure that we can realise the novel applications above
Literature
- Suleman Surti, Joel S. Karp, Lucretiu M. Popescu, Margaret E. Daube-Witherspoon, and Matthew Werner:
Investigation of time-of-flight benefit for fully 3-D PET, IEEE TRANSACTIONS ON MEDICAL IMAGING 25 (5): 529-538
MAY 2006 - Grazioso: PET Performance of PET-MR Brain Insert Tomograph, 2005 IEEE Nuclear Science Symposium Conference
Record M08-08 - Peter Fischer, Ivan Peric, Michael Rizert and Torsten Solf: Multi-Channel Readout ASIC for ToF-PET I, 2006 IEEE
Nuclear Science Symposium Conference Record M11-140,
