7th Conference on PET/MR and SPECT/MR

20 May 2018, 08:30 → 23 May 2018, 19:30 Europe/Rome

La Biodola, Isola d'Elba

Alberto Del Guerra (PI), Georgos Loudos (Athens), N. Jon Shah (Institute of Neuroscience and Medicine - 4, Forschungszentrum Juelich)

Please go to the PSMR2018 official web page for further information.

Timetable timetable.pdf

Sunday, 20 May

08:45 → 09:30 Registration

09:30 → 12:00 Training school on PET/MR reconstruction

09:30 Basics of MR physics and image reconstruction

10:30 Hands-on session with SIRF for MR image reconstruction

12:00 → 15:00 Lunch break

15:00 → 16:00 Training school on PET/MR reconstruction

15:00 Basics of PET physics and image reconstruction

16:00 → 16:30 coffee break

16:30 → 19:00 Training school on PET/MR reconstruction

16:30 Hands-on session with SIRF for PET image reconstruction

18:00 Synergistic reconstruction for PET-MR

19:30 → 20:30 PSMR welcome cocktail

Monday, 21 May

08:30 → 09:30 Opening Session

Convener: Dr Irene Neuner (JARA Brain, Germany)

08:30 Welcome

08:50 PET/MRI and brain connectivity

Speaker: Marco Aiello (IRCCS SDN, Napoli)

Slides Aiello_psmr2018.pdf

09:30 → 10:30 Session 1 - Brain and organ specific systems

Convener: Prof. Taiga Yamaya (QST, Japan)

09:30 Evaluation of the MINDView PET insert in a 3T MRI

We report in this work the current status performance of the brain PET insert developed under the MINDView project. Final construction has been accomplished. Pilot performance studies have been carried out at the lab, partially following the NEMA protocol. It has later been installed at the nuclear medicine department in Klinikum rechts der Isar (Munich) and exhaustively tested inside the Siemens mMR, a whole body PET-MR with a 3T main magnetic field. The PET insert FOV is 154 mm axially and 240 mm transaxially, defined by 3 rings of 20 monolithic crystals each (20 mm thick), coupled to custom 12x12 SiPM arrays. The system sensitivity is above 7% (350-650 keV) with a point-like source at the CFOV. A variety of MR sequences for brain imaging have been run (EPI, ASL, T1w, T2w and UTE), and the PET response measured, without showing any deterioration. Count rates as a function of sequences were studied not also exhibiting a system malfunction. Also the MR performance has been studied, among other tests the uniformity of the B0 and B1 fields, not showing significant changes when the dedicated brain PET is inserted. Regarding performance comparison with a state-of-the-art whole body PET system, it shows an improved performance as observed through the mini-Derenzo or other phantoms. Rods of about 2 mm are clearly distinguished with standard iterative reconstruction methods and voxel/pixel sizes. Detailed analyses will be presented. Currently, the system is in Klinikum rechts der Isar (Munich) and patients recruitment is undergoing.

Speaker: Antonio Gonzalez (Institute for Instrumentation in Molecular Imaging, i3M-CSIC)

09:50 Design and detector performance of the PET component of the TRIMAGE PET/MR/EEG system

Speaker: Giancarlo Sportelli (Università di Pisa)

Slides sportelli-PSMR2018.pdf

10:10 Next-generation breast cancer PET-MR imaging with HYPMED: System and workflow design

Breast cancer is a leading cause of cancer death in women and a major socio-economic issue. With currently available methods, early diagnosis frequently fails. Moreover, beyond mere detection, there is an ever-increasing need for improved non-invasive characterisation of cancer. Targeted therapies require an in-depth analysis of cancer to select and guide appropriate treatment. Both, PET and MRI can provide molecular and functional information that may be of pivotal importance for tailoring therapy. However, current whole-body PET/MRI systems lack the necessary sensitivity and resolution for this task. HYPMED addresses this by engineering an innovative multimodal imaging tool. HYPMED will integrate a fully-digital MRI-transparent PET-detector into a novel multi-channel PET-transparent MRI surface coil. The PET-RF insert will allow unprecedented imaging of breast cancer with high-resolution/ultra-high sensitivity PET, combined with high-level structural and functional MRI, and allow minimal-invasive MRI- and PET-guided targeted biopsy. This presentation focuses on the unique system and workflow design solutions implemented in HYPMED, which will enable next-generation clinical imaging for the discovery and diagnosis of breast cancer.

Speaker: Dennis Schaart (Delft University of Technology)

10:30 → 11:00 coffee break

11:00 → 12:40 Session 2 - Software and Quantification: attenuation correction – part 1

Convener: Prof. Dimitris Visvikis (INSERM, France)

11:00 Attenuation estimation using non-TOF PET scattered photon energy information and an anatomical MRI prior

Emission-based attenuation correction methods aim to derive the attenuation information directly from the emission data. Although promising results have been obtained with this approach, joint-reconstruction of activity and attenuation suffers from several limitations. Nearly all studies report cross-talk between the estimated activity and attenuation distributions if time-of-flight information is not available. It has recently been proposed to use scattered photon data as an additional source of information that could be integrated into the reconstruction algorithm, by relying on multiple energy window measurements. In previous preliminary work, we demonstrated that stand-alone scattered data do contain useful information and can be used to estimate the attenuation-map in realistic energy resolution scenarios using maximum likelihood (ML) optimisation, if the emission is known. However, we observed that the image quality is degraded by noise over iterations. Therefore, we propose here to take advantage of the MR information available from the PET/MR scanner. We incorporated an MRI-derived prior into the reconstruction algorithm, which enforces common edges between the estimated attenuation-map and the MR image. The chosen anatomical prior relies on the concept of Parallel Level Sets (PLS). It makes use of directional information derived from the anatomical image and encourages images with aligned spatial gradients. In the case when no structural information is available, the prior reduces to the total variation prior. Simulations have been conducted on a cylindrical phantom and realistic 3D XCAT images, using the geometry and characteristics of the Siemens mMR scanner. We compared results from different initialisations of the algorithm, and with different structures in the MR-based prior. Results show that the incorporation of the anatomical prior helps in controlling the noise level in the reconstructed image, enhancing the edges and improving quantification. However, over-smoothing can occur where no anatomical information is present. In conclusion, the PLS anatomical prior shows promising results but more investigation on parameter tuning is warranted.

Speaker: Ludovica Brusaferri (Institute of Nuclear Medicine, UCL, London, UK)

Slides LB_PSMR2018.pptx

11:20 Repeatability of Deep Learning density maps of the head

The goal of the present study was to assess the longitudinal repeatability of a novel bone identification method, based on a convolutional deep network trained to convert LAVA-Flex and zero echo-time MRI data into pseudo-CT density maps. The consistency of the bone maps -with potential applications in attenuation correction of PET data and MR-based radiotherapy planning- was evaluated for a number of clinically realistic variations of the ideal acquisition conditions.

Speaker: Gaspar Delso (UniversitätsSpital Zürich)

11:40 Simultaneous emission/transmission measurements for accurate attenuation correction in PET/MRI

Attenuation correction (AC) is still a challenge in PET/MRI. Without photon transmission sources, methods for AC rely on the segmentation of MR images or the use of dedicated MR sequences. In general, these methods neglect the attenuation of cortical bone, thus, leading to a bias in quantification of tracer concentration in PET images. We have developed a head coil system with an integrated orbiting transmission source for PET/MRI to enable direct measurement of attenuation using 511 keV photons, which is the gold-standard in AC. The basic concept of our setup is based on the ”liquid drive”, presented by Jones et al [1] in 1995. Our head coil system combines a 24-channel head-and-neck MR receiver coil with an MR-compatible hydraulic system for driving a positron emitting transmission source around the patient. Using sinogram windowing, emission and transmission data of an acquisition can be separated. This allows for post-injection measurements and reduces scatter in the transmission channel. The manufactured prototype was tested in the Siemens Biograph mMR using a phantom with air cavities and a Teflon (PTFE) cylinder measured both with and without emission activity. For both measurements air, water and Teflon were clearly distinguishable and homogeneous regions of the phantom were successfully reproduced in the AC map. The linear attenuation coefficient of water was measured to be (0.096 +- 0.005) cm^-1 for both the cold and the hot measurement. The presence of emission activity did not deteriorate the quality of the attenuation map.

Speaker: Andreas Renner (Center for Medical Physics and Biomedical Engineering)

12:00 An Aristotelian view on MR-based attenuation correction (ARISTOMRAC): combining the four elements

The ancient Greeks believed that matter was made of a weighted combination of four elements: earth, water, air and fire. Using a similar simplified assumption, we formulated a multi-component MR fingerprinting framework, where every voxel was considered a weighted combination of four base tissues: bone, water, air and fat. We named our approach Aristotelian MR based attenuation correction (ARISTOMRAC). We used a 3D radial acquisition scheme at 1.5T, using a transient-state spoiled acquisition with variable flip angles and TEs, with the shortest TEs being ultra-short echo times (UTE). We simulated a multi-tissue MR signal model using the Bloch equations and used dictionary matching to extract tissue fraction maps. Compared to previous methods for MRAC, our approach takes advantage of the intrinsic multicomponent nature of MR Fingerprinting. For this reason, rather than reconstructing high resolutions images, we could acquire MR data more efficiently, directly at the resolution needed for PET attenuation maps. We could accurately estimate water, bone, fat and air fractions in a phantom containing combinations of these.

Speaker: Matteo Cencini (Università di Pisa)

Slides presentationARISTOMRAC.pptx

12:20 A lesion insertion tool for attenuation correction evaluation in simultaneous PET/MR

Attenuation Correction in simultaneous PET/MR was originally implement using a two-point DIXON which neglected the bones. More recent implementations of MR based attenuation strategy in PET/MR involve either a fast UTE or a bone atlas which allows for inclusion of bones. However, soft-tissues are still segmented to provide anatomical attenuation into five compartment identified as air, lungs, fat, water and bones. Evaluation of the quantitative performance of these recently implemented attenuation methods is needed and this abstract presents the development and performance of a lesion insertion tool which to insert a spherical lesion into human body habitus from PET/MR images and generate MonteCarlo based sinogram data and images. The tool was evaluated using digital phantoms and with FDG PET data sets from cervical cancer patient imaged by PET/MRI. For those patients, a prior CT was also registered to the MR based attenuation correction to provide a comparative CT based attenuation correction. A validated tool will be useful for efficient evaluation of MR-based attenuation correction in simultaneous PET/MR and will be faster and easier to use than complex and lengthy simulation studies.

Speaker: Richard Laforest (Washington University)

12:40 → 15:30 Lunch break

15:30 → 17:30 Session 3 - Instrumentation: systems and detectors

Convener: Prof. Sibylle Ziegler (Klinikum r. d. Isar / TUM / Dep. of Nuclear Medicine, Germany)

15:30 Initial results of a prototype brain PET system using a time-based digitizer and an FPGA-based real-time coincidence processor

Our group is developing a brain-dedicated PET insert that will be combined with an ultra-high field (7T) MRI system. Herein, we present a prototype brain PET scanner with a diameter of 254 mm and an axial length of 26 mm consisting of two time-based digitizers and FPGA-based DAQ systems. The scanner was based on dual-layer depth-of-interaction (DOI) detectors with a relative offset by half a crystal pitch in x- and y-directions. Each detector block consisted of a 14×14 array of 1.78×1.78×12 mm3 and 13×13 array of 1.78×1.78×8 mm3 LSO crystals coupled with a 2×2 array of 4×4 silicon photomultipliers. A time-based digitizer had 132 energy channels and 33 timing channels, which can support up to 33 detector blocks. The position-encoding signals were digitized using charge-to-time converters and binary counters and the timing information was acquired using a time-to-digital converter implemented in an FPGA. Single event data were sent to the master FPGA from which coincidence pairs were generated. A two-dimensional Hoffman brain phantom was scanned and the reconstructed image was evaluated by comparing with those acquired using a clinical whole-body PET/CT scanner. The average energy resolutions were 10.7±0.6% and 11.1±1.5% for the upper and lower layer crystals, respectively. The reconstructed phantom image showed better spatial resolution and contrast recovery compared to that of a clinical whole-body scanner. The detailed structures can be observed with less edge artifacts from the image acquired using the developed brain PET scanner. The next milestone will be to incorporate PSF and TOF information in the reconstruction.

Speaker: Jeong-Whan Son (Seoul National University)

15:50 Current status of the PET detector developments of a dedicated PET/MRI for the diagnosis of breast cancer

Current clinical simultaneous PET/MRI systems as well as PET/CT systems have been designed as whole-body systems. In both hybrid modalities, PET is limited in spatial resolution and sensitivity. In particular, the low spatial resolution of PET limits the ability to detect cancer at early stages and to monitor therapy responses, develop/apply radiomics and radiogenomics methods. To overcome these limitations, we are developing a PET/MRI deployment in the EU H2020 HYPMED project that combines local MR receiver coils and local PET detectors. The aim is to achieve a homogeneous spatial resolution of about 1.2 mm (FWHM) with four times higher sensitivity than existing whole-body PET systems. The PET detector is based on Philips digital SiPMs and is designed to offer PET simultaneous with 1.5-T MRI scans of both breasts to limit the overall measurement time and effective use of the injected dose. We present the current state of the PET detector design of the breast PET /MRI system that we are developing in the HYPMED project. The insert includes a newly developed RF receiver coil and a highly integrated digital PET detector. The entire PET detector is a further development of our already developed HYPERION IID detector design. To support the clinical workflow, which includes the possibility of breast fixation and PET/MRI-guided biopsy, the PET detector offers the opportunity to open the two detector rings. To maintain a homogeneous spatial resolution, a three-layer DOI crystal design with 1-mm crystal pitch was examined, which shows an excellent separation of all three layers.

Speaker: Dr David Schug (RWTH Aachen University)

16:10 The design of a wide-bore 1.5T MR body coil with integrated PET detectors

We envision a clinical workflow for the PET/MRI, in which PET/MRI is used for treatment planning and diagnostic workup for personalized cancer treatments using image-guided therapies (e.g. MR-guided radiotherapy, MR-LINAC). However, commercially available PET/MRIs have a small bore that is not suitable for scanning of patients in radiotherapy position (arms up). The aim of this study was to show the feasibility of a wide-bore body coil with integrated PET detectors that can be retrofitted into a Philips Ingenia 1.5T MR scanner, replacing the commercially available body coil. We built and simulated a wide-bore 42-rung high-pass quadrature birdcage coil (~64 MHz). The PET detector covers were designed to minimize mirror currents and to fit within the confined space between the body coil and RF shield. Simulations were done to evaluate performance with and without PET detector covers. Additionally, we studied the effect of gradient switching on coincidence resolving time (CRT). Simulations with and without the presence of PET detector covers indicate that PET covers are responsible for an attenuation of <2 dB (B1+ of 80%) and similar values were found experimentally. The PET energy resolution of the current setup was 11.57% and measurements of the CRT (333ps) showed no degrading effect of timing related to gradient switching. In conclusion, we designed a body coil with integrated PET detectors that can be fitted into a Philips Ingenia 1.5T MR scanner, replacing the commercially available body coil.

Speaker: Casper Beijst (UMC Utrecht)

16:30 Performance characterization of MPPC modules for ToF-PET applications

Current generations of PET detector designs feature arrays with large numbers of multi-pixel photon counters (MPPCs), also known as silicon photomultipliers (SiPMs), necessitating the use of data acquisition systems with either a high degree of signal multiplexing or using application specific integrated circuits (ASICs) to readout each pixel independently. In this work we evaluate the performance of two Hamamatsu C13500-4075LC-12 PET detector modules designed for time-of-flight (ToF) PET applications. These modules each have a 12 × 12 array of lutetium fine silicate (LFS) scintillator crystals (20 mm deep, 4.2 mm pitch) one-to-one coupled to a 12 × 12 array of 4 mm MPPC elements. Each detector has a total of 8 18 channel ASICs that utilize a time-over-threshold (ToT) readout method. Data were acquired to characterize the detector stability vs. temperature and event rate. At low event rates, the detectors have a FWHM energy resolution of 10.0%, measured in a linearized energy spectrum, and a coincidence timing resolution (CTR) of 275 ps FWHM with an energy window of ±1 FWHM of the 511 keV photopeak. The average change in the 511 keV photopeak amplitude in the ToT spectrum is 0.24%/˚C. There was <10% event loss up to 1.45 Mcps/detector, beyond which bandwidth limits were encountered. From a count rate of 35 kcps to 1.45 Mcps the 511 keV photopeak position varied by 1.2%, while the FWHM of the photopeak increased by 53.9%. Over this count rate range the CTR varied from 285 ps to 435 ps FWHM. The performance of the ToF-PET modules suggests they will be well suited to use in whole-body PET and PET/MR imaging applications.

Speaker: Andrew Goertzen (University of Manitoba)

Slides PSMR.2018.Goertzen.v2.pptx

16:50 Timing resolution of linearly-graded silicon photomultipliers coupled to thin monolithic crystals

This work presents a new PET detector concept for compact preclinical hybrid simultaneous MR-PET. The detector concept is based on Linearly-Graded SiPM produced with FBK RGB-HD technology. One 7.75mm x 7.75mm large sensor chip is coupled with optical grease to a black coated 8mm x 8mm large and 3 mm thick monolithic LYSO crystal. The readout is obtained from four readout channels with linear encoding based on integrated resistors and the Center of Gravity approach. The measurement setup was prepared to center, edge and corner point that all significant points of the detector were covered. A CRT of 360 to 510 ps for different points of the detector was achieved. Due to the reduced thickness, the sensitivity of this gamma detector is low but still higher than pixelated designs with the same thickness due to the monolithic crystals. Combining compact design, high spatial resolution, and high sensitivity, the detector concept is particularly suitable for applications where the scanner bore size is limited and high resolution is required - as is the case in small animal hybrid simultaneous MR-PET.

Speaker: Arne Berneking (Bruker)

Slides Timing_Resolution_of_Linearly-Graded_Silicon_Photomultipliers_Coupled_to_Thin_Monolithic_Crystals_no_effects_uploaded.pptx

17:10 RGB-HD SiPMs for fast readout of GAGG scintillation crystals in medical imaging

In this work, we present the characterization of energy and coincidence time resolution of scintillation detectors obtained by the optical coupling of gadolinium-aluminum-gallium garnet (GAGG) crystals with RGB-HD silicon photomultipliers (SiPMs) with active area of 4×4 mm2 and cell pitch of 25 μm, developed at Fondazione Bruno Kessler (FBK). The emission spectrum of GAGG:Ce, peaked at 525 nm suggests RGB-HD SiPMs as the best candidates for the scintillation light readout, with peak photodetection efficiency (PDE) around 550 nm. Moreover, the recently improved production process of RGB-HD SiPMs resulted in a substantial reduction of both dark count rate and correlated noise probability. This fact, together with an optimized electronics for leading edge discrimination allows an accurate triggering capability at low thresholds. Crystal samples with a size of 3×3×5 mm3 with different dopants were characterized for application in time-of-flight positron emission tomography (TOF-PET). Ce-activated GAGG was compared to samples codoped with Mg and to multidoped crystals containing Ce, Mg and Ti. The scintillation detectors featured an energy resolution of 6% at 511 keV at +20°C. The timing performance of the GAGG crystals was found to be improved by the presence of Mg and Ti dopants. The best coincidence resolving time of 160 ps was obtained with GAGG:Ce,Mg at +20°C.

Speaker: Alberto Mazzi (FBK - TIFPA)

Slides RGB-HD_GAGG_for_medical_imaging.pdf

17:30 → 18:00 coffee break

18:00 → 19:40 Session 4 - Preclinical systems and applications

Convener: Dr Andrew Goertzen (University of Manitoba, Canada)

18:00 MR-compatibility assessment of MADPET4: a small animal PET insert for a 7 T MRI scanner

MADPET4 (Munich Avalanche Diode PET 4) is a small animal PET insert for a 7 T MRI scanner, which has two offset layers of scintillation crystals, individually read out by silicon photomultipliers (SiPMs). To reduce the interference between the PET insert and the MRI system, high-gain SiPMs were used and the SiPM charge was transferred to outside the MRI scanner using 1.5 m coaxial cables. In this work, MR-compatibility of the PET insert was assessed by studying the interference between the PET insert and the MRI system, with a 1H/13C radio-frequency (RF) volume coil placed inside the PET insert. The volume coil was used for RF transmission and reception. The coil had a 9 µm copper layer on its outer layer for electromagnetic interference shielding, whereas the PET insert and its readout electronics were not shielded. The effects of PET on the static magnetic field, flip angle distribution, RF noise, and image quality of various MRI sequences were investigated. The effects of fast-switching gradient fields and RF pulses on PET data were studied. Simultaneous operation of the insert with standard MRI pulse sequences was possible with a maximum signal-to-noise ratio degradation of 23-30%, with no discernible artifacts. Use of RF excitation pulses and fast-switching gradient fields had negligible effects on PET data. The achieved performance allows using MADPET4 to investigate applications of simultaneous PET/MRI in future small animal studies.

Speaker: Negar Omidvari (Nuklearmedizinische Klinik und Poliklinik, Klinikum rechts der Isar, Technische Universität München)

Slides PSMR2018_NOmidvari_16_9.pptx

18:20 The SAFIR readout prototype

The aim of the SAFIR collaboration is the construction of a positron emission tomography (PET) insert for a preclinical magnetic resonance imaging (MRI) device. The device will be able to handle high source activities and the target time for one PET image acquisition is 5 s. The mechanical design is very challenging, because the inner diameter of the MRI scanner’s bore limits the diameter of our insert to 200 mm. The presented prototype consists of the same components as the full system. It fulfils the mechanical constraints, but consists of only two readout boards, each equipped with one detector module. The detector module has a matrix of 12 x 12 scintillator crystals with a pitch of 2.5 mm. We use one-to-one coupling to connect those to matching SiPM arrays. The analogue signals are digitalized with PETA6 Application Specific Integrated Circuits (ASICs) and the digital data are transmitted via an optical Ethernet link to our data acquisition computer. In this work, we evaluate the energy resolution of the prototype. After linearisation and calibration, the singles energy resolution at the photo peak is 14.7 % full width at half maximum (FWHM).

Speaker: Christian Ritzer (Institute for Particle Physics and Astrophysics, ETH Zürich, Switzerland)

Slides PSMR18CRitzer.pdf

18:40 Quantitative performance assessment of a preclinical SiPM-based SPECT/MRI insert

We present the conclusive measurements of the characterization of a preclinical SPECT insert compatible with standard MR scanners and targeting neuro-oncology. The instrument is composed of a static gantry of 10 gamma cameras based on arrays of SiPMs coupled to a 50mm by 50mm CiI(Tl) continuous scintillator. Each module features a 36-channel readout ASIC, an optimized MRI-compatible layout, optical digital output and offers an intrinsic spatial resolution of 1mm. Planar image reconstruction (with a usable FoV of 40mm by 40mm) is performed by means of iterative identification of light response function and maximum likelihood. Tomographic reconstruction is performed after linearity and uniformity corrections of single modules. The extrinsic SPECT spatial resolution, measured with 99mTc-filled capillary of 0.290mm diameter), is below 0.9mm FWHM across the whole transaxial FoV (equal to 15.6mm). The sensitivity is 1100cps/MBq and is linear up to 10MBq. SPECT energy (13%) and spatial resolution do not worsen during simultaneous MR imaging, reported here for a small Derenzo phantom and for a mouse brain. At the same time, the impact of the insert on the MR image quality is minimized by means of shielding, filtering, materials selection and shimming (for the tungsten powder multi-pinhole collimator), leading to very promising results, also in the perspective of the up-scaled clinical version of the instrument.

Speaker: Marco Carminati (Politecnico di Milano)

Slides Carminati_-_INSERT_-_PSMR_2018_2.pptx

19:00 Depth-of-interaction encoding detectors for the SAFIR system

Two different depth-of-interaction (DOI) encoding detector designs were studied for the SAFIR (Small Animal Fast Insert for mRi) PET system to improve spatial resolution in the axial direction. One is a two-layer DOI detector with two layers of crystals stacked with an offset of half crystal pitch. The other is a single-layer DOI detector using triangular shaped ESR reflectors between crystals. Both detectors utilized the same size of crystal cross-section (2.1 mm × 2.1 mm), total length of the crystal block (12 mm), and are mounted on a 4 × 4 SiPM array with 2 mm × 2 mm pixels. In this study, we report the detector performances of the two DOI-encoding designs as compared with a non-DOI detector composed of the same dimensions of crystals and SiPMs with one-to-one coupling. Energy resolutions were calculated after SiPM non-linearity correction and evaluated to be 10.6% for the non-DOI detector, 12.3% for the two-layer DOI detector, and 10.2% for the single-layer DOI detector. Coincidence resolving time resolutions at FWHM were calculated to be 336.8 ps for the non-DOI detector, 406.5 ps and 382.0 ps for the top and bottom layers of the two-layer DOI detector, and 295.6 ps for the single-layer DOI detector. The two-layer DOI detector exhibits a degradation of timing performance while the single-layer DOI detector shows excellent timing performance. For a DOI performance, the two-layer DOI detector provided a clear discrimination of events occurred in the top or bottom layer from a crystal map with a peak-to-valley of 9:1. The single-layer DOI detector provided a DOI resolution of ~ 3.1 mm FWHM, but a careful calibration is necessary for the DOI estimation.

Speaker: Mikiko Ito (Institute of Particle Physics and Astrophysics, ETH Zurich, Zurich, Switzerland)

19:20 Thallium labelled citrate coated Prussian blue nanoparticles as potential imaging agent

Powerful new high-resolution imaging tools such as optical whole body imaging (FOBI), magnetic resonance imaging (MRI) could thus nowadays enable very good anatomical detail fusion to functional image data (positron emission tomography (PET) and single photon emission computed tomography (SPECT)). With the availability of these novel multimodal imaging devices the demand for innovative multimodal contrast materials is higher than ever before. Our aim was to create a biocompatible Prussian blue nanoparticle (iron(2+);iron(3+);octadecacyanide; PBNP) platform for preclinical imaging which has four important qualities (long-term stability of the NP platform, T1 signal enhancement on MR image, stability of the bond between methylene-blue (MB) fluorescent dye and nanoparticle structure and stability of the bond between the radiolabel and nanoparticle structure. The physical properties of PBNPs were characterized with atomic force microscopy, dynamic light scattering and zeta-potencial measurement. PBNP biodistribution was determined by using FOBI, SPECT and MRI following intravenous administration into C57BL6 mice. PBNPs appeared as objects with a flat rectangular surface protruding from a rounded halo. Rectangularity of the particles was found to be 0.81±0.09. Height of the particles showed monomodal distribution (23.0±8.3 nm). The non-hydrated size of the NPs were analyzed with TEM (17.54 ± 4.56 nm, n=700). The mean Zeta potential of PBNPs at pH 7.4 was -25.7 ± 1.8 mV (n=3). The mean hydrodynamic diameter of PBNPs was 32.10 ± 0.1801, as determined by DLS. There was no significant colloidal alteration during the 6-week duration of the study (polydispersity index (PDI) 0.203 ± 0.004). T1-weighted and T2-weighted spin echo images of a PBNP containing phantom showed signal enhancement on T1-weighted image instead of decreased signal on T2-weighted image. MB-labeled PBNP accumulation peaked at 2 hours post injection in the kidneys and the bladder. After 201Tl labeling the radiochemical purity was >98 % in all experiments. The radioactive labeling yield was 99.84 % (SD: 1.01 %). PBNP accumulation peaked at 2 hours post injection predominantly in the kidneys and the liver followed by a gradual decrease in activity in later time points. In this study we successfully synthesized, characterized and demonstrated the biodistribution of citrate-coated MB-PBNP labeled using 201Tl isotope. The results show a chemically stable and biocompatible 201Tl and MB labeled nanoparticulate SPECT tracer and optical contrast material. These PBNP based particles could be applied as a drug delivery platform or a contrast agents in pre-clinical research. They could further be tailored towards clinical application, too.

Speaker: Domokos Máthé (CROmed Translational Research Centers, H-1047, Budapest, Hungary)

Tuesday, 22 May

08:30 → 10:30 Session 5 - Quantitative MR: the numbers in the picture (part 1)

Convener: Dr Michela Tosetti (IRCCS Stella Maris and IMAGO7 Research Foundation, Italy)

08:30 BURZTE: Silent Quantitative MRI using ZTE combined with BURST

ZTE image encoding was combined with Burst imaging by reversing segments of 3D radial spokes in both time and amplitude. This recalls gradient echoes for each spokes/excitation, yielding images with T2* weighting. By adding a pair of refocusing pulses with the appropriate weight period, it is also possible to obtain spin-echoes, leading to T2-weighted images. Data is reconstructed using standard 3D gridding and Fourier transformation, and fitting quantitative T2*, T2 and magnetic susceptibility (QSM) maps. In vivo feasibility was demonstrated by imaging the brain of multiple healthy volunteers.

Speaker: Rolf Schulte (GE Healthcare)

08:50 High-Resolution Reconstruction of Arterial Spin Labelling MRI Using Anatomical Priors

This work investigates super-resolution reconstruction and deconvolution of perfusion arterial spin labelling (ASL) images using a high-resolution T1-weighted MRI with the aim of reducing the partial volume effect on the estimated cerebral blood flow (CBF). The MR acquisition matrix was factorized to model the down-sampling and blurring of the underlying ASL images, as well as MR coil sensitivity, Fourier encoding and k-space undersampling. The proposed methods were evaluated using simulation and real data in comparison with the standard reconstruction method, and an anatomical non-local means filtering combined with deconvolution. In simulations, both MR-guided deconvolution and reconstruction methods achieved the lowest normalised root mean square (NRMS) errors of 20% and 18.5% respectively, compared to the standard method with NRMS error of 29.5%. For real data, the guided deconvolution gave rise to the best results in terms of contrast and detail recovery. Evaluation of the guided reconstruction of the real data is in progress. In conclusion, the proposed methods provide a promising solution for improving the quality and quantitative accuracy of estimated CBF maps.

Speaker: Abolfazl Mehranian (School of Biomedical Engineering and Imaging Sciences, King's College London, St. Thomas' Hospital, London, UK)

Slides presentation-psmr-2018-finall_22ndMay.pptx

09:10 Assessment of mesoscopic properties of brain iron through a model-based analysis of magnetic susceptibility and R2* MRI in multiple sclerosis

Inductively coupled plasma mass spectrometry (ICP-MS) of brain tissue has confirmed that both R2* and magnetic susceptibility increase linearly with the tissue iron concentration in humans without neurological diseases. However, being related to mesoscopic magnetic field inhomogeneities, R2* depends not only on the voxel-average concentration but also on the distribution of iron in the tissue. Quantitative Magnetic Susceptibility Mapping (QSM), on the other hand, is a recent MRI technique that may be regarded as a more direct measure of the iron concentration in the tissue. The goal of the present study was to exploit the dependence of both measures on physical interactions at different length-scales to retrieve information about the tissue micro-environment in vivo.

Speaker: Ferdinand Schweser (Buffalo Neuroimaging Analysis Center, University at Buffalo)

Slides schweser_PSMR2018.pptx

09:30 Quantitative Evaluation of Magnetic Susceptibility Dynamics during Brain Function

Functional Quantitative Susceptibility Mapping (fQSM) is a recent method that, based on the same acquisition technique as conventional functional Magnetic Resonance Imaging (fMRI), has two very appealing features: it is quantitative and it is less affected by non-local effects than the Blood Oxygenation Level-Dependent (BOLD) signal. Here, for the first time, the response of the auditory cortex to the presentation of relatively short (4 s) acoustic stimuli has been studied with fQSM. Activated voxels in fQSM were fewer than in BOLD fMRI, consistently with what has previously been shown in the literature, as the magnetic dipole deconvolution step removed the spurious and non-local signal changes observed in conventional fMRI. The most significant activations co-localized with veins, but both positive and negative fQSM responses were observed also in gray matter tissue, suggesting that this technique is sensitive to subtle changes in deoxyhemoglobin in the parenchyma and might be a useful tool to study small-scale functional architectures.

Speaker: Mauro Costagli (Imago7)

Slides Costagli-fQSM-PSMR2018c.pptx

09:50 MR Fingerprinting: Clinical and Scientific Rationale

Speaker: Vikas Gulani (Case Western Reserve University and University Hospitals Cleveland Medical Center)

10:30 → 11:00 coffee break

11:00 → 12:40 Session 6 - Quantitative MR: the numbers in the picture (part 2)

Convener: Dr Guido Buonincontri (INFN Pisa, Italy)

11:00 Comprehensive Quantitative Imaging: MR Fingerprinting and Beyond"

Speaker: Nicole Seiberlich (Case Western University)

11:40 Multiparametric Mapping accounting for Flow: MR Angiography in the Transient-state

Quantitative Transient-state Imaging (QTI) is a non-random, dictionary-less MR Fingerprinting alternative. Through iterative reconstructions, QTI recovers a series of contrast-weighted images from transient-state acquisitions and subsequently estimates the parameters that best describe the resulting dynamic signal evolutions. Here, we extend the QTI framework by incorporating a simple velocity model that accounts for blood flowing into and out of the imaging slice. The model, however wrong, can be very useful: it predicts signal hyperintensities in the presence of flow, allowing for the simultaneous reconstruction of MR Angiography images, hundreds of dynamic contrast-weighted images, and their corresponding parametric maps.

Speaker: Pedro Gómez (Technical University of Munich)

Slides gomezPSMR2018.pptx

12:00 Delayed Gadolinium Enhanced Relaxation Mapping of Osteoarthritis with Magnetic Resonance Fingerprinting

Mapping of quantitative MRI relaxation values is promising for improving the assessment of musculoskeletal disease. Magnetic Resonance Fingerprinting (MRF) is a new method that enables fast quantitative MRI by exploiting the transient signals caused by the variation of pseudorandom sequence parameters. This proof-of-concept work demonstrates the utility of MR Fingerprinting in the knee. Seven participants, four of which had knee osteoarthritis (Kellgren-Lawrence (KL) grade 2 or 3), were imaged approximately 80 minutes after gadolinium injection with MRF on a 3.0T MRI. The mean T1 relaxation times were shorter by 5-20% in the KL=2,3 subjects when compared to normal subjects in cartilage.

Speaker: Joshua Daniel Kaggie (University of Cambridge)

Slides PSMR_MRF_dGEMRIC.pptx

12:20 Magnetic Resonance Fingerprinting: Quantifying Relaxation Times and Fat Signal Fraction in the Breast

Magnetic Resonance Fingerprinting (MRF) promises quantitative tissue analysis based on the relaxation times. Spiral undersampling is employed with MRF to yield short scan times. However, it is prone to blurring artifacts in the presence of field inhomogeneities, especially if both aqueous and fatty tissues are present. In this contribution, we extend MRF to quantify the fat signal fraction and simultaneously correct for spiral blurring. Thereby, improved spatial information about the tissue relaxation times as well as the fat signal fraction is retrieved. We present the application to female breast imaging.

Speaker: Teresa Nolte (Philips Research Eindhoven, The Netherlands AND Department of Physics of Molecular Imaging Systems, Institute of Experimental Molecular Imaging, RWTH Aachen University, Germany)

12:40 → 15:00 Lunch break

15:00 → 16:15 Session 7 - Industrial Seminars

Convener: Prof. Maria Giuseppina Bisogni (University of Pisa and INFN Pisa, Italy)

15:00 MR solutions

15:15 Single-Crystal SiPM PET technologies for preclinical cardiac PET/MR imaging (Bruker)

15:30 Molecubes

Slides 20180522_PSMR_MOLECUBES_SN.pptx

15:45 SensL

Slides PSMR_2018_Slides_ON_Semiconductor_J_Murphy_V1p0.pptx

16:00 Hamamatsu

Slides PSMR2018_Hamamatsu_Kramer.pdf

16:15 → 17:45 Session 8 - Poster Session I

Convener: Dr Kris Thielemans (University College London, UK)

16:15 A web-based platform for multimodal data integration: the ARIANNA project

Dealing with heterogeneity is the main challenge that scientists are facing in several diseases, including Autism Spectrum Disorders (ASD). This complexity urges the implementation of dedicated analysis techniques to obtain the maximum from the interrelationship among the numerous variables that describe affected individuals, including clinical phenotypic characterization, genetic profile and multimodal brain imaging. The ARIANNA project has developed a collaborative web-based interdisciplinary research environment that is accessible to the community of researchers working on ASD (https://arianna.pi.infn.it). The main goals of the project are: to extract reliable measurements from multimodal and multicentric neuroimaging data; to identify both structural and functional brain characteristics that allow distinguishing individuals with ASD from control subjects through multivariate approaches based on machine learning and deep learning techniques; to identify neuroimaging-based criteria to stratify the population with ASD to finally support the future development of personalized treatments. The platform provides secure data handling and storage as well as the access to fast computational resources. This paper outlines the web-based architecture, the computing infrastructure and the collaborative analysis workflows at the basis of the ARIANNA interdisciplinary working environment.

Speaker: Paolo Bosco (INFN)

16:15 An Algorithm for Maximum-Likelihood Estimation of the Timing Resolution in TOF-PET

As is well known, utilization of time-of-flight (TOF) information can reduce noise and improve convergence rate in PET image reconstruction since it allows to incorporate the (Gaussian) probability density function (the ”TOF-kernel”) of the annihilation event position along the respective LOR into the image reconstruction process. In doing so, it is crucial to use the best possible estimate of the actually given timing resolution in order to achieve a realistic contrast recovery and minimize noise. In this context, it is relevant to recognize that the timing resolution of a time-of-flight PET system is count rate dependent. However, count rate dependent TOF-resolution calibration is usually not provided by the vendors. We, therefore, developed a procedure which is compatible with clinical routine and is also applicable retrospectively to existing data. We propose a novel Maximum-Likelihood Timing Resolution Estimation (MLTRE) algorithm that maximizes likelihood by updating activity image and TOF-kernel width alternately. The algorithm was evaluated using phantom and patient studies covering a large range of count rates that were acquired with a Philips Ingenuity TF PET/MR scanner. Our preliminary results indicate that the proposed algorithm is capable of realistic timing resolution estimation, while being convenient and easy to use in clinics. To the best of our knowledge, the dependency of timing resolution on singles rate of the Philips Ingenuity TF PET/MR scanner was never published before. According to our findings, the timing resolution of this scanner degrades rapidly with increasing count rate.

Speaker: Pavel Nikulin (Helmholtz-Zentrum Dresden-Rossendorf, PET Center, Institute of Radiopharmaceutical Cancer Research)

16:15 Assessment of shielding materials for the add-on PET prototype

We are developing an add-on PET system which is an MR birdcage RF-coil integrated with DOI-PET detectors. This is expected to enable realization of a high resolution, high sensitivity and low cost PET-MRI system. In the proposed system, PET detectors are extremely close to the MR RF-coil. To reduce electromagnetic interaction between the PET detectors and the MR RF-coil, the PET detectors should be covered with conductive shield boxes. However, when the magnetic field around the shield box is changed by field gradient pulses, an eddy current is generated in the shield box. In this paper, we evaluate comprehensive performance for the shield materials in terms of signal-to-noise ratio (SNR) and eddy current. The decreasing rate of SNR in the simultaneous operations with and without the power supply was measured. These results showed the decreasing rate of SNR was suppressed by 27.5% for the carbon fiber board and 32.6% for the copper foil, whereas 84.5% for the acrylic. Based on the secondary magnetic field near the shield box using copper foil, secondary magnetic field with carbon fiber board was suppressed about 47%. These results showed that carbon fiber board has higher shielding performance and suppression capacity for the eddy current though PET detectors are extremely close to the MR RF-coil.

Speaker: Mikio Suga (Chiba University)

16:15 Bayesian penalized likelihood reconstruction in PET/MR brain imaging: 11C-PE2I

The performance of the Bayesian likelihood reconstruction algorithm in PET/MR brain imaging using receptor or transporter ligands has not previously been investigated. Ten patients underwent PET/MR brain scanning with the dopamine transporter ligand 11C-PE2I and images were reconstructed using a Bayesian penalized likelihood algorithm, with a regularization term of 25, 50, and 100, in addition to TOF-OSEM. The Bayesian algorithm produced images with a higher signal-to-noise ratio and decreasing noise as the regularization term was increased. Data reconstructed using a regularization term of 50 and 100 presented a higher signal-to-noise ration that that from TOF-OSEM.

Speaker: João M. Sousa (Uppsala University)

16:15 Effect of Brain MRI Coil Attenuation Map Processing on PET Image Quantification and Uniformity for the GE SIGNA PET/MR

Attenuation correction of brain MRI coils used on PET/MR systems can be prone to error emanating from artifacts in CT-based coil attenuation maps. In this study editing was applied to coil CT images to reduce the impact of metal-induced artifacts. In addition, CT-based attenuation maps were generated for the coil mirrors and four levels of smoothing were applied to the attenuation maps. A uniform phantom experiment was performed to assess absolute quantification, and both axial and transaxial image uniformity. For the three coils tested, the edited CT attenuation maps improved absolute quantification (mean error 0.2% vs. 7.2% in the central axial 15 cm) and axial uniformity by up to 59% (33% on average) compared to the vendor-supplied attenuation maps, although transaxial uniformity deteriorated by up to 25% (11% on average). The inclusion of the mirror marginally improved axial uniformity in the majority of cases, but also slightly degraded transaxial uniformity. The impact of changing smoothing was minor, with the system default setting of 10 mm FWHM Gaussian producing the best results overall.

Speaker: Roido Manavaki (University of Cambridge)

16:15 Evaluation of template-enhanced ZTE attenuation correction for PET/MR imaging

The purpose of this study was to evaluate a new, template-enhanced zero-echo-time (ZTE) attenuation correction method for PET/MR scanners. 20 subjects underwent 18F-FDG PET/CT, followed by PET/MR on a GE SIGNA PET/MR. A ZTE sequence was used to generate two attenuation maps: one with the standard segmentation-based method; and another with an enhanced version using pre-registered anatomical templates and CT data to improve the segmentation. Reconstructed PET images were quantified in 67 volumes-of-interest. The PET bias with template-enhanced ZTE was measured to be -0.8% 0.9%, compared with -1.4% 1.1% with regular ZTE attenuation correction. Statistically significant bias reduction (of average and standard deviation) was obtained in the frontal, temporal, parietal, occipital and insula regions. These results indicate that the co-registration of pre-recorded anatomical templates to ZTE data is feasible in clinical practice and can improve the performance of segmentation-based attenuation correction.

Speaker: Gaspar Delso (UniversitätsSpital Zürich)

16:15 Feasibility of Quantitative Magnetic Resonance Fingerprinting in Ovarian Tumours for T1 and T2 Mapping

Quantitative MRI has previously shown benefits for the assessment of ovarian cancer. Magnetic Resonance Fingerprinting (MRF) is a novel technique for quantitative MRI, which exploits the transient signals caused by the variation of MRI sequence parameters. This proof-of-concept work demonstrates the utility of MRF in two patients, with low and high grade ovarian tumours on a 3.0 T MRI. The mean value for both subjects for T1 was 2464.5 ± 100.9 / 1974.8 ± 191.3 ms, and for T2 was 225.4 ± 33.9/94.1 ± 14.5 ms. The mean T1 and T2 in the tumour was higher by ~20% and ~58% in the low grade ovarian tumour in comparison with the malignant tumour.

Speaker: Joshua Daniel Kaggie (University of Cambridge)

16:15 Fractional perfusion: A simple semi-parametric measure for clinical translation of hyperpolarized 13C MR?

Hyperpolarized (HP) 13C magnetic resonance imaging (MRI) is a promising tool for in vivo metabolic interrogation of disease states and treatment efficacy assessment. However, the method is currently limited by the lack of good quantitative measures. This is particularly true in humans where large variations in transport kinetics have been reported [1]. Here we introduce a novel model-free perfusion [2] and area under the curve ratio-metric [3] combined model. We define fractional perfusion as substrate-delivery corrected metabolic conversion, for quantifying the metabolic information in hyperpolarized imaging. Healthy (normoglycemic) and diabetic (hyperglycemic) rats were subjected to unilateral ischemia (IR) with a non-traumatic clamp for 40 min, 8 days before sacrifice. The model proposed fractional perfusion was investigated using HP [1-13C]pyruvate and compared to 1H dynamic contrast enhanced (DCE) perfusion imaging. A similar perfusion assessment was derived from HP [1-13C]pyruvate and DCE (p=0.65) in both healthy contralateral (CL) and IR kidneys using the proposed model. Fractional pyruvate to lactate perfusion (FPL) was also comparable between the two methods. Alanine signal significantly increased in the hyperglycemic IR kidney (p=0.04) compared to the hyperglycemic CL kidney. However, when accounting for perfusion changes, no alteration in the metabolic conversion is seen. A significantly altered fractional lactate production is seen in the hyperglycemic animals, demonstrating a metabolic shift. Hence, we demonstrate that the use of HP 13C-metabolites could be used to quantify the energetic demand by mapping both the injected biomarker perfusion and metabolic conversion. Therefore, the pyruvate concentration curve can be a surrogate marker for perfusion, in cases where other perfusion assessment is not directly obtainable. Perfusion parameters from DCE MRI or dynamic pyruvate signal therefore enable correction of pyruvate delivery variations. This allows calculation of semi-parametric measures of metabolism per perfusion unit, which could facilitate more quantitative information from HP pyruvate in the future. [1] Granlund K, Vargas H, Lyashchenko S, DeNoble P, Laudone V, Eastham J, Sosa R, Kennedy M, Nicholson D, Guo Y, Chen A, Tropp J, Hricak H, Keshari K (2017) Utilizing hyperpolarized MRI in prostate cancer to assess metabolic dynamics and histopathologic grade. [2] Johansson E, Olsson LE, Månsson S, Petersson JS, Golman K, Ståhlberg F, Wirestam R (2004) Perfusion assessment with bolus differentiation: A technique applicable to hyperpolarized tracers. [3] Hill DK, Orton MR, Mariotti E, Boult JK, Panek R, Jafar M, Parkes HG, Jamin Y, Miniotis MF, Al-Saffar NM, Beloueche-Babari M, Robinson SP, Leach MO, Chung YL, Eykyn TR (2013) Model free approach to kinetic analysis of real-time hyperpolarized 13C magnetic resonance spectroscopy data.

Speaker: Christoffer Laustsen (MR Research Centre, Department of Clinical Medicine, Aarhus University, Aarhus, Denmark)

16:15 From SPM8-based to SPM12-based Attenuation Correction for Brain PET/MR Imaging

We evaluated whether applying SPM12 instead of SPM8 to delineate bone, soft tissue and air for head MR-based attenuation correction (MRAC) would be feasible. The differences in bone delineation between SPM8- and SPM12-based MRAC methods were studied by visual analysis and evaluation of dice coefficients against CT-based attenuation correction (CTAC). The quantitative accuracy in PET was evaluated in volume of interest (VOI) analysis using 35 regions in the gray matter. The results show slightly improved bone overlap in SPM12-based MRAC, with Dice coefficient of 0.80±0.04 compared to 0.77±0.05 in SPM8-based MRAC. The mean relative difference in PET uptake across all VOI regoins was -2.53±1.8 % in SPM8-based MRAC and -1.77±1.72 % in SPM12-based MRAC, as relative to CTAC. In conclusion, using SPM12 for head MRAC is feasible to create MR-based attenuation maps with bone, soft tissue and air and offers slight improvements over SPM8 in bone delineation and quantitative accuracy in PET.

Speaker: Jarmo Teuho (Turku PET Centre, University of Turku and Department of Medical Physics, Turku University Hospital, Turku, Finland)

16:15 Implementation and Performance of FORE-FBP for Brain Imaging on the GE SIGNA PET/MR

Fourier rebinning combined with 2D filtered backprojection (FORE-FBP) was implemented for the GE SIGNA PET/MR. The algorithm was designed and tested for brain imaging. Spatial resolution, image uniformity and signal-to-noise ratio (SNR) were assessed for a range of maximum ring differences (ΔZ). Results were compared to those for ordered subsets expectation maximum (OSEM) and time-of-flight OSEM (TOF-OSEM). Nearest neighbor axial interpolation provided the highest FORE-FBP NEMA spatial resolution (4.4 × 4.0 × 4.0 mm at 1 cm radius for maximum ΔZ=44), which was inferior to that for OSEM/TOF-OSEM (3.4 × 2.1 × 1.5 mm). Axial resolution for FORE-FBP degraded by < 0.4 mm for ΔZ=44 compared to ΔZ=27 for radii < 10 cm. Two cylindrical phantoms of disparate internal diameters (2.2 and 19.7 cm) indicated that FORE-FBP image uniformity in the axial direction in the central axial 15 cm reached optimal levels for ΔΖ=27. Image SNR for FORE-FBP improved with increasing ΔZ, with linear interpolation providing the lowest image variability. Image uniformity and SNR within the central axial 15 cm for the NEMA NU 2-1994 phantom were superior for OSEM/TOF-OSEM compared to FORE-FBP by 30% and 10-25%, respectively.

Speaker: Tim Fryer (University of Cambridge)

16:15 Implementation of cylindrical PET scanners with block detector geometry in STIR

The SAFIR collaboration is developing a PET insert for a pre-clinical MRI system, aiming at excellent temporal resolution, of ~5s time frames. Image reconstruction is performed using the Software for Tomographic Image Reconstruction (STIR). An accurate system matrix is necessary for the precise reconstruction of quantitative PET data and the exact representation of the scanners geometry is a key component of a precise system matrix. STIR models the scanner as a cylinder, i.e. it assumes the scanner is made of individual detector elements arranged uniformly on a cylinder, whereas the crystals in the SAFIR detector are assembled in blocks which are in turn arranged on the sides of a polygon. In this study, we implemented and evaluated the true block type model into the STIR library. We evaluated this new implementation with three sets of Monte Carlo simulations: a rotating plane source for normalization, a uniform cylinder, and a Derenzo phantom. We reconstructed the data using both the existing cylinder and the new block model. Attenuation and normalization corrections were included in the reconstruction algorithm. The random rate was estimated using Monte Carlo simulation and data were corrected prior to their reconstruction. Our results demonstrate a significant improvement for the new block model in terms of resolution and uniformity. The peak to valley ratio for the 2 mm Ø spheres of the Derenzo phantom increased from 1.64 to 2.53. The reconstructed images of the uniform cylinder show better uniformity already prior to normalization. The standard-deviation-to-mean ratio for the uniform cylinder is ~2.5 times smaller for the new block model.

Speaker: Parisa Khateri (Institute for Particle Physics and Astrophysics, ETH Zürich, Switzerland)

16:15 IRIS projector model for monolithic detector based systems: proof of concept using a dedicated brain PET/MR device

We present a simple but accurate model to compute the uncertainty regions of impact coordinates measured in monolithic crystals. The proposed model considers the scintillation crystal, internal reflection, crystal coating, the presence of a surface retroreflective structure, and the pixel detector layout. This model is used to estimate the uncertainty in the LOR vertex location, and incorporated through the IRIS projector for the image reconstruction of data acquired using monolithic crystal detectors. The model has been validated with Monte Carlo simulations showing good correspondence with measured data. More than 85% of the data accurately matched the predictions. The model has been tested with real data from the dedicated brain PET insert MindView. In this case, the IRIS projector implemented with the proposed model resulted in an average SNR gain of approximately 30% compared with reconstructions based on standard, experimentally decompressed, point of interaction determination. The proposed uncertainty estimation method in combination with the IRIS projector are independent of the reconstruction process and could be incorporated within any image reconstruction algorithm.

Speaker: Amadeo Iborra (LaTIM, INSERM, UMR 1101, Brest, France)

16:15 Longitudinal [18F]FDG PET-MRI at 9.4 Tesla in Twitcher mice using spatial normalization

This study presents preliminary results on the feasibility of truly simultaneous PET-MRI in live mice on a conventional 9.4T MRI system retrofitted with a miniature-PET ring. Wild-type and Twitcher mice, a model of Krabbe disease, were anaesthetized and imaged longitudinally at four time points. Simultaneously acquired MRI images allowed the generation of a model-specific brain template for voxel and atlas based analyses of the PET scans. Preliminary findings were in line with previous results from metabolomics suggestive of alterations in brain glucose metabolism in Twitcher mice. A brain region-specific analyses based on segmented MRI scans will allow the quantification and statistical analysis of the glucose uptake.

Speaker: Poonam Choudhary (Buffalo Neuroimaging Analysis Center, University at Buffalo)

16:15 Magnetic Resonance Fingerprinting with dictionary-based fat and water separation (DBFW MRF)

The purpose of this work is to obtain quantitative MRI maps of fat and water with a fast acquisition for musculoskeletal imaging applications. Conventional fat separation methods are based on steady-state signal models, relying either on relaxometry or chemical shift. A recent approach for parameter mapping, called magnetic resonance fingerprinting (MRF), uses transient-state acquisitions, comparing them with pre-computed simulations in search of the best match. Here, we build on recent MRF work by including a multi-component fat and water signal model. By using a full simulation, we can combine relaxometry and chemical shift fat separation within a single acquisition. We demonstrate robust fat fraction estimation, with a simultaneous T1, B1+ and B0 map. These were in good agreement with previous approaches, but achieving a significant speedup.

Speaker: Matteo Cencini (Università di Pisa)

16:15 Motion estimation and motion correction using EBER in simultaneous PET-MR imaging

The correction of head motion of dynamic PET in the context of PET-MR imaging can benefit from estimation either based on PET data, or on MRI data, when both modalities have been simultaneously acquired. Based on a recent method developed for a direct correction of PET list-mode data during the rebinning of PET dynamics into sinograms, this paper reports the study of the relation between modality, frame duration, smoothing strength and motion parameter accuracy in the recovery of the motion.

Speaker: Inés Mérida (CERMEP-Imagerie du vivant, Lyon, France)

16:15 NEMA tests on a whole-body TOF PET/MR scanner with and without simultaneous MR pulse sequences

In this work we present the characterization of the physical performance of the fully integrated PET/MR whole body system SIGNA PET/MRI (GE Medical Systems), following the NEMA NU 2-2012 standards. All the measurements, except count rate performance, have been performed first in the PET-only mode, without the simultaneous MR acquisition (MR-idle), then with two different MR pulse sequences on (MR-pulsing). According to the N-02 standards, the characterization of the SIGNA PET/MRI scanner yielded the following results: 1) transverse (axial) spatial resolution (FWHM) (1cm and 10cm off axis): 4,35(5,67)mm and 5.05(7,29)mm 2) sensitivity (average at 0cm and 10cm): 22,31cps/kBq; 3) scatter fraction: 42,16; 4) NEC peak rate: 225 kcps at 16,2kBq/cc using k=1 in the NEC formula for random coincidences estimation. 5) The residual error from attenuation and scatter corrections was 1.9%. The average energy resolution was 10,3%, the time resolution was 375ps. No effect of MR pulse sequences on PET measurements was observed. In conclusion, the SIGNA PET/MRI system shows very good performance.

Speaker: Paola Scifo (Nuclear Medicine Dept., IRCCS San Raffaele Scientific Institute, Milan, Italy)

16:15 On the Clinical Value of CNN-processed Ultra-low-dose Amyloid PET Reconstructions

In this study, we aimed to generate diagnostic-quality amyloid positron emission tomography (PET) images from “low-dose” PET images, reconstructed from massively undersampled raw data, as well as simultaneously-acquired multimodal magnetic resonance imaging (MRI) contrasts used as inputs in a convolutional neural network (CNN) framework. We have shown that the synthesized images generated from a model incorporating both PET and MR inputs yield images with superior image quality and diagnostic value compared to the low-dose image as well as images synthesized from a model with PET-only inputs.

Speaker: Kevin Chen (Department of Radiology, Stanford University)

16:15 Optimization of the timing and energy resolution in a staggered dual layer PET detector

Speaker: Pietro Carra (PI)

16:15 Predicting the lower count limit for adequate scatter correction using dynamic data with low count frames on Siemens mMR

AIM: Scatter correction is applied in PET imaging to reduce signal from scattered coincidence events, thereby improving image quality and quantitative accuracy. In dynamic PET imaging, predefined time frames with the aim of reflecting tissue metabolism over time contain the net coincidence count registered within a specific time window. Short time frames are often desired to obtain precise time-activity curves as the tracer accumulates. This study was motivated by a coincidental finding in a time-activity curve from a dynamic PET brain scan using Oxygen-15, which displayed peaks in several time frames that were inconsistent with tracer biology. Further, the corresponding scatter fraction factor values were noticeably low compared to the remaining frames. The objective of this study was to explore the lower limit for the amount of net true count in individual time frames while still preserving adequate scatter correction. METHODS: Image reconstruction was performed on three dynamic PET brain studies from Siemens Biograph mMR with tracer injections of Oxygen-15, Carbon-11 and Flourine-18 respectively. From each study, three patients were randomly selected. Reconstructions were performed offline using e7-tools (Siemens) with reconstruction settings identical to clinical protocol. Time frames matching clinical protocols were used as reference, and a subsequent shorter time frame designed to find a breaking point was estimated from the hypothesis that the net true count influences the scatter correction accuracy. Net true count for individual time frames was recorded and the corresponding time-activity curve extracted. RESULTS: The mean net true count for each tracer deviated ±8% at most from the total mean (1.8 mil) at the breaking point. Similar values are observed in all patients within each tracer and across different tracers. The scatter correction failure ranges from the lowest value in Oxygen-15 (1.5 mil) to the highest value in Fluorine-18 (2.1 mil). CONCLUSIONS: Results suggest net true count is related to the failure of scatter correction, independent on injected tracer, and time frames with net true counts below a certain value may provoke a failure. Predicting lower limit for frame time appears feasible based on injected tracer, dose and post injection time. Further studies and larger patient populations are necessary before definitive conclusions can be derived.

Speaker: Maya Ottilia Dalgaard (Department of Clinical Physiology, Nuclear Medicine and PET, Rigshospitalet, Copenhagen)

16:15 Quantitative evaluation of the Mediso high resolution PET/CT-system LFER-150

Background The Mediso LFER-150 is currently the only commercially available high resolution PET/CT-system mainly dedicated for non-human primate (NHP) imaging. The gantry bore diameter of the system is 26 cm. The transaxial and axial fields of view are 15 cm and 20 cm respectively. Resolution at the center of field of view, in images reconstructed with full detector modelling, is 1 mm . These characteristic make the LFER-150 particularly suitable for use in NHP brain imaging or for whole body imaging of NHPs of small size. Objective The main objective of this study was to compare the LFER imaging capabilities with that of the current gold standard system for high-resolution PET, the Siemens HRRT system, after optimizing the LFER-150 reconstruction parameters, using the improved CT-based three component (air, tissue, bone) attenuation correction. Specific goals where to compare results from the two systems by using the same NHP’s and calculate binding potential for small areas in the brain using radioligands for the dopamine D2/D3 receptor - [11C]Raclopride - and for the dopamine transporter - [18F]FE-PE2I. Method For the experiments, two cynomolgus monkeys (Macaca fascicularis) were used. The NHPs were anesthetized, using ketamine hydrochloride. Each NHP underwent two PET measurements with [11C]Raclopride and [18F]FE-PE2I. The PET measurements were performed first using the LFER-150 ([11C]Raclopride: 94 MBq, [18F]FE-PE2I: 90 MBq) and later the HRRT system ([11C]Raclopride: 147 MBq, [18F]FE-PE2I: 139 MBq). PET data was collected in list mode for 95 minutes. List mode data was reconstructed into 38 frames (10 s × 9, 15 s × 2, 20 s × 3, 30 s × 4, 60 s × 4, 180 s × 4, 360 s × 12). Data acquired with the LFER-150 were reconstructed using the latest Tera-Tomo 3D algorithm (10 iterations, 9 subsets), with high regularization, median and spike filter turned on and with detector modeling corrections and three component material map (air, tissue, bone) attenuation correction. List mode data acquired with the HRRT system were reconstructed using the 3D-OSEM algorithm (10 iterations, 16 subsets) with PSF correction applied. The reconstructed images were coregistered to the MRI of the NHP and regional VOIs were manually delineated onto the MRI images and applied to each PET data. Binding potential, BPND, was estimated using the simplified reference tissue model and the cerebellum as reference region. D2/D3 receptor availability was measured in the caudate and putamen, whereas dopamine transporter availability was measured in the caudate, putamen and substantia nigra. Results Reconstructed images of [11C]Raclopride and [18F]FE-PE2I obtained using the LFER-150 were similar to those obtained using the HRRT (Figure 1). Time-activity curves (TACs) obtained with [11C]Raclopride and [18F]FE-PE2I using the LFER-150 system were similar to those obtained using the HRRT system (Figure 2). Figure 1. Summation images (0-93 minutes) of [18F]FE-PE2I. Figure 2. Representative regional time-activity curves of [18F]FE-PE2I obtained from data acquired using the LFER-150 and the HRRT systems. The BPND values estimated from data acquired with the LFER-150 system were in close agreement with the BPND values estimated from data acquired with the HRRT system (Table 1). Table 1. BPND values for the caudate, putamen and substantia nigra, from image data acquired with the LFER-150 and the HRRT systems Conclusions The findings of this study indicate that the quantitative performance of the LFER-150 PET/CT system, for imaging the dopaminergic system in the brain of NHPs, is similar to the quantitative performance of the HRRT system. From the visual inspection of the TACs, the signal-to-noise ratio of data acquired with the LFER-150 seems to be lower than the signal-to-noise ratio of data acquired with the HRRT system, which may be due to lower amount of injected radioactivity (30-35%). Additional PET studies are ongoing to evaluate whether the signal-to-noise ratio can be improved by acquiring data after injection of higher amount of radioactivity. Other radioligands, such as [11C]Flumazenil and [18F]FDG, having a more uniform distribution in the brain, are currently being evaluated. Preliminary data from this study suggest that the LFER-150 is a PET-CT system for high-resolution imaging of the brain in NHPs.

Speaker: Göran Rosenqvist (Karolinska Institute)

16:15 Quantitative radial magnetic resonance fingerprinting in the animal brain

Magnetic resonance fingerprinting (MRF) uses non-steady state magnetic resonance acquisitions and dictionary-based matching to measure parameters such as T1, T2, B0 and B1. Standard clinical contrasts can then be calculated from these retrospectively. Here, we implement a radial acquisition scheme for MRF and apply it to the animal brain at 9.4T to map T1 and B1 in short in vivo acquisitions in rats and marmosets. We show an application of the method by calculating synthetic images at a range of repetition times and a series of inversion recovery images. The short scan times here make MRF feasible for routine animal studies, adding a huge range of potential quantitative opportunities for structural brain phenotyping in a range of models.

Speaker: Stephen Sawiak (Wolfson Brain Imaging Centre, University of Cambridge)

16:15 Reduced Acquisition Time PET/MR Quantification: a comparison of tracer delivery estimation methods

Due to the long acquisition time required for dynamic PET with pharmacokinetic (PK) modelling, the statically acquired standardised uptake value ratio (SUVR) is commonly used. However, SUVR is sensitive to blood flow changes which confound longitudinal studies. We have proposed a framework to incorporate cerebral blood flow (CBF) from arterial spin labelling (ASL) MRI into the PK modelling to halve the acquisition time from 60 to 30 mins using a simultaneous PET/MR acquisition. This framework requires the conversion of ASL-CBF into the PET tracer delivery parameter (R1), which is then fixed for PK modelling of the 30 mins of PET data. In this work we compare three methods for estimating PET-R1 from ASL-CBF data: linear regression (LR), image fusion (IF) and deep learning (DL), and assess the influence of the resulting R1 errors on the estimation of target density (BPND). This was applied to an amyloid-beta targeting PET tracer, used in the imaging of Alzheimer's Disease. Analysis on 32 subjects showed that the IF and DL methods outperformed the global LR method as they utilise local structural information to derive R1. Despite the improved R1 estimation, DL has a similar error to IF for BPND, suggesting that these methods are approaching the upper limit for this short acquisition PET/MR framework.

Speaker: Catherine Scott (UCL)

16:15 Results of a multivariate analysis of Magnetic Resonance texture parameters for rectal cancer patient staging

The therapy for locally advanced rectal cancers is: neoadjuvant chemo-radiotherapy (CRT) followed by radical surgery. Although the local pelvic recurrence rate is lower than 10%, this therapeutic approach is an over-treatment of the patients who completely respond (CR) to CRT. Indeed, CR could benefit from either less invasive surgery (ie, transanal endoscopic microsurgery) or “wait-and-watch” strategy. Also the patients who do not respond to the treatment (non-responders, NR) whose early identification (2–3 weeks after the start of neoadjuvant CRT) might help clinicians in referring them to alternative treatments. Magnetic Resonance (MR) Imaging is the gold standard for preoperative staging and re-staging of rectal cancer. However, it is hard to asses the CRT patient response only on the basis of the MR images visual inspection. With the aim of discriminating CR and NR patients prior or immediately after the start of CRT, we developed a Texture Analysis (TA) of T2 weighted MR images of the entire volume of rectal cancer acquired before, during and after the treatment of 55 patients with a 3 T MR. All the patients had a histologically confirmed rectal adenocarcinoma and locally advanced tumour stages II (cT3-4, N0, M0) and III (cT1-4, N+, M0). The extracted TA parameters have been used to build a classifier using multivariate analysis, it allows to discriminate CR with an area of the Receiver Operating Characteristic (ROC) curve of 0.94 and NR with a ROC curve area of 0.86.

Speaker: Carlo Mancini Terracciano (ROMA1)

16:15 Schedule design for parameter quantification in the transient state using Bayesian optimization

Magnetic resonance fingerprinting (MRF) is a new quanti- tative imaging paradigm, which simultaneously acquires multiple tissue parameters in an efficient experiment. MRF can map several parameters simultaneously including T1, T2, and spin density. However, it is important to underline that specific sequences may be better suited for certain parameter ranges or sampling patterns. This work aims to introduce a framework for pulse sequence optimization, individually optimizing for T1 or T2 relaxation times. We demonstrate a new method, including undersampled acquisitions, by simulating the MRI signal encoding, gridding, and pattern recognition directly in the optimization. The design framework could obtain efficient schedules for T1 and T2 acquisition.

Speaker: Giada Fallo (Dipartimento di Informatica, Università di Pisa)

16:15 SPECT and MRI evaluation increases understanding pathophysiology in a murine model of sepsis associated encephalopathy

Sepsis-associated encephalopathy (SAE) is a frequent and devastating complication of severe acute systemic inflammation that can cause both acute and long-lasting neurological dysfunction and heavily contributes to the mortality of patients with sepsis. The knowledge of the pathophysiological processes overwhelming the brain at this early stage of sepsis is far from complete. The purpose of this study was to evaluate an approach combining SPECT and anatomical MRI methods using also a set of widely used nuclear medicine imaging agents as possible methods to study the early effects of systemic inflammation on the living brain in a mouse model of sepsis associated encephalopathy (SAE). The lipopolysaccharide (LPS) induced murine systemic inflammation model was selected as a model of SAE. A multimodal imaging protocol was carried out on each mouse 4 hours following the intravenous administration of LPS using the following tracers: [99mTc]HMPAO and [125I]iomazenil to measure brain perfusion and neuronal damage respectively; [18F]FDG to measure cerebral glucose uptake. We assessed microglia activity on another group of mice using [125I]CLINME. Radiotracer uptakes were measured in different brain regions and correlated. Microglia activity was also assessed using immunohistochemistry. Brain glutathione levels were measured to investigate oxidative stress. The results of SPECT measurements illustrates significantly reduced [99mTc]HMPAO uptake in the LPS treated group. Following perfusion compensation process significantly enhanced [125I]iomazenil uptake values were registered in the LPS treated group’s cerebellum and hippocampus. Significantly enhanced [18F]FDG uptake was registered in the treated group compared to the control. Relevant [18F]FDG uptake was seen in the treated animals’ hippocampus but this difference was not significant. Significantly enhanced [125I]CLINME uptake was registered in all of the investigated brain regions in the LPS treated group compared to the control animals. In both groups [18F]FDG and [125I]iomazenil uptake showed highly negative correlation in all brain region. No significant differences were detected in the glutathione levels between the groups. The CD45 and P2Y12 double labelling immunohistochemistry showed widespread microglia activation in the LPS treated group. Our results suggest that inflammatory processes can directly contribute to the uptake of [125I]iomazenil and [18F]FDG masking the neuroinflammation-induced neuron damage and hypometabolism of neural tissue respectively. [99mTc]HMPAO and [125I]CLINME can be used to detect cerebral hypoperfusion and microglia activation respectively as early as 4h following the i.v. injection of LPS. The negative correlation between the measured uptake values (low [99mTc]HMPAO, high [125I]iomazenil and [18F]FDG uptake) could also indicate the severity of brain involvement during systemic inflammation: low perfusion coupled with massive immune cell activation indicated by high [125I]iomazenil and [18F]FDG uptake. Further investigation of the metabolic activity of different brain cells and the status of the GABA receptor system would be necessary to determine the exact source of the measured signal differences during the early phase of systemic inflammation.

Speaker: Domokos Máthé (CROmed Translational Research Centers, H-1047 Budapest, Hungary)

16:15 The quantitative performance and optimal regularization parameter in BSREM reconstructions of clinical 68Ga-PSMA PET/MR

In contrast to ordered subset expectation maximization (OSEM), block sequential regularized expectation maximization (BSREM) PET reconstruction algorithms can run until full convergence while controlling image quality and noise. Recent studies with BSREM and 18F-FDG PET reported higher signal-to-noise ratios and higher standardized uptake values (SUV). In this study we investigate the optimal regularization parameter for clinical 68Ga-PSMA PET/MR reconstructions in the pelvic region applying TOF BSREM in comparison to TOF OSEM.

Speaker: Edwin ter Voert (Department of Nuclear Medicine, University Hospital Zurich and University of Zurich)

16:15 Towards Implementing Multiparametric PET/MR for RT Planning in Head and Neck Cancer

AIM: We explore the feasibility of performing PET/MR scans of head and neck cancer (HNC) patients in radiation therapy (RT) treatment position using MR-compatible RT immobilization devices. We develop and test a proper workflow and scan protocol, which can be implemented for daily routine. Finally, to explore the use of multiparametric PET/MR in relation to personalized RT, we compare the quantitative ADC values of two different diffusion weighted imaging (DWI) methods. MATERIALS AND METHODS: The feasibility study included 3 pilot patients diagnosed with HNC referred for RT. Patients were PET/MR scanned (Siemens Biograph mMR) in the same position as the planned treatment by using an MR-compatible RT setup, allowing attachment of thermoplastic patient masks, which are also used in routine planning CT and during the RT treatment. The scan protocol included anatomical imaging, DWI, and ultrashort echotime sequences. PET images were reconstructed using E7tools (Siemens) with MR-based AC and CT-based hardware attenuation map. For comparison of DWI methods, ten patients with HNC referred for surgery were included. Each patient underwent two different DWI sequences on the same PET/MR: A single-shot EPI and a multi-shot EPI. Comparison of the resulting ADC maps was performed in a volume on interest (VOI) delineating the primary tumor. A correlation between the mean ADC values and a voxel-vise Bland-Altman analysis were performed. RESULTS: In the pilot feasibility study all three patients were scanned and data were successfully acquired. Total scan time for each patient was approximately 30 minutes excluding DCE-MRI and post contrast imaging, which is to be included in a larger future study. All there patients complained about back pain after the examination, but due to suboptimal workflow the patients spend longer time on the patient table than needed. Proper preparation of the hardware setup prior to positioning the patient is important to optimize patient comfort. For comparison of DWI methods, a strong correlation (R2=0.78, P<0.01) was seen between the mean ADC values, when excluding one clear outlier. On a voxel basis, the Bland-Altman analysis showed no apparent bias between the methods but limits of agreement of 60%. CONCLUSION: Although based on a very small cohort of patients, we have shown that it is feasible to scan patients with HNC on a PET/MR system in the same position as the planned treatment using a complete RT fixation setup. On a voxel basis, the ADC values from the two DWI methods demonstrated a rather large variation, suggesting that quantitative ADC values should be carefully evaluated, especially if RT methods such as dose-painting by numbers is considered. In future work we will investigate whether dose plans can be derived solely from PET/MR data, thus allowing for one-stop-shop PET/MR RT planning.

Speaker: Anders Olin (Dept. of Clinical Physiology, Nuclear Medicine & PET, Rigshospitalet, Denmark)

17:45 → 18:00 coffee break

18:00 → 19:00 Session 9 - Software and Quantification: attenuation correction – part 2

Convener: Dr Niccolò Camarlinghi (University of Pisa and INFN Pisa, Italy)

18:00 Perfect In-Phase ZTE Imaging for Improved MR Attenuation Correction

This work details a new approach to acquire and reconstruct zero echo-time (ZTE) data, aimed at removing chemical shift artifacts which is impairing pseudo-CT generation. In order to achieve this a multi-repetition readout scheme was implemented, where consecutive echoes were acquired with different read-out bandwidths (rBW). In the reconstruction data was separated into an in-phase, and an out-of-phase compartment using a regularized inversion of a chemical shift encoding matrix. Phantom and human scans were acquired on a time-of-flight enabled 3.0T PET/MR system.

Speaker: Gaspar Delso (GE Healthcare)

Slides Engstrom_pipZTE_PSMR18.pptx

18:20 Accuracy when using R2* as a measure of bone density in PET/MRI: Evaluation using 775 subjects

AIM: Accurate MR-based attenuation correction (MR-AC) for use in combined clinical PET/MR systems is crucial in order to achieve quantifiable PET images. The main challenge has been to represent continuous bone density. The ultra-short echo time (UTE) MR sequence has been proposed in several methods to be utilized for representation of bone, subsequently extracted using an R2*-map. The purpose of this study is to investigate a large cohort of subjects to determine whether gender or age specific cohorts are needed when learning a mapping from R2* to CT-HU, as well as the general reproducibility of the R2* measure. METHODS: We selected 775 subjects examined on PET/MRI who also had a same day low-dose CT. We subsequently calculated the R2*-map for further investigation, and used the co-registered CT as MR-AC reference. We investigated whether there is a relation between bone density and age for each gender, by calculating the mean value within bone voxels of CT-HU and R2*, respectively. We then fitted a sigmoid model to the R2*-CT data separated by age (2-18, 19-64 and 65-98) and gender. Using a leave-one-out approach for each patient, we found the median sigmoid fit within each separate cohort. For each cohort, we then calculated the root-mean-squared-error of bone density in HU between CT and each of the median sigmoid models. This allowed us to compare the effect of using age and gender specific models. RESULTS: There is a significant correlation with bone density decrease and age for the female cohort, which was present in both the CT-HU and R2* measures. The mean R2*-values were relatively higher in the pediatric cohort compared to adults, whereas it was lower with CT. The overall same results were found for the male cohort, albeit the bone density did not decrease significantly with age in any of the plots. These findings suggest that separate models for R2*-CT relationship are needed. The median sigmoid models show that separation between genders does not yield a difference in mapped HU, but the sigmoid models created with the pediatric cohort were always 100-200 HU lower than the adults. When using age-specific models to estimate CT-HU, a smaller error is found for aged-matched (pediatric versus adult) models than for alternative models. The age-related difference is important to note, since PET/MR is especially clinically relevant for use in pediatrics. CONCLUSIONS: The analysis suggest, that when modeled on large data set, the R2*-value is a usable measure of bone density in PET/MRI, when accounting for age.

Speaker: Amalie Hindsholm (Rigshospitalet)

Slides R2Accuracy_PSMR.pptx

18:40 Effect of Spatial Resolution on ZTE-based Attenuation Correction

In brain PET/MR, the zero echo-time (ZTE) pulse sequence has already been shown to be a valuable tool to derive pseudo-CT and attenuation correction maps thanks to its ability to capture the short T2* signal from bone tissue. Most studies in this area have used a moderate resolution (2.4 mm) to minimize the acquisition time in a generic clinical setting and few studies have used a high resolution (1.6 mm). The aim of the present study is to assess if a higher resolution ZTE is useful to generate more accurate pseudo-CTs and attenuation maps and hence to reduce PET quantification error. These results indicate that the increase of resolution allows to further reduce the PET quantification error compared to the moderate resolution ZTE, in particular in the cerebellum. This comes with a two-fold increase of the acquisition time. While the performance of the fast ZTE method with moderate resolution should be sufficient for most basic clinical applications, advanced PET/MR brain studies may benefit from the additional accuracy of high resolution ZTE.

Speaker: Gaspar Delso (GE Healthcare)

Slides Fernandez_ZTEres_PSMR2018.pptx

20:00 → 22:00 Conference dinner

Wednesday, 23 May

08:30 → 10:30 Session 10 - Software and Quantification: other

Convener: Prof. Pedro Almeida (Inst. Biophysics and Biom. Engineering, Faculty of Sciences, University of Lisbon, Portugal)

08:30 Comparison of motion surrogate signal derived from MR images, respiratory belt and PCA-PET for motion correction in PET/MR

Respiratory motion is one of the major causes of artefacts in chest imaging. Motion correction is therefore needed in PET/MR chest acquisitions, to restore the image quality and improve lesion detectability. In clinical practice the patient’s breathing motion is detected via the use of external devices, whereas Principal Component Analysis (PCA) can produce a signal using only the PET raw data. This allows to avoid the cost of dedicated hardware, the added dose for the staff member in charge of setting it up and the additional time required to do so. The aim of this work is to validate the quality and accuracy of the PCA signal extracted from TOF-PET, via comparison with the internal motion extracted from a fast 2D MR sequence (via edge-detection) and a respiratory belt signal. The methods are applied on 13 cancer patients, acquired on the GE SIGNA PET/MR scanner, choosing the lower lung bed position. The results show that the three signals (PCA-PET, MR-derived and belt) are very highly correlated with each other. This proves that PCA produces an accurate and reliable surrogate respiratory signal, while confirming that the external motion of the chest and the internal motion of the organs are both good measures of the breathing pattern of the patient.  The PCA-PET signal could therefore be utilised for respiratory motion extraction in PET/MR, avoiding the use of external devices and without the need for dedicated MR sequences.

Speaker: Kris Thielemans (Institute of Nuclear Medicine, UCL, London, UK)

Slides Bertolli_PSMR_2018_-_DDG-PETMR.pptx

08:50 MR image corrections for PET-system-induced gradient distortions

Combining positron emission tomography (PET) and magnetic resonance imaging (MRI) can be realized by placing a PET insert in a clinical MRI scanner. When designing the PET insert, mutual influences of both imaging modalities need to be minimized. The gradient magnetic fields induce eddy currents in all conductive components of the PET insert. Eddy currents produce superimposing magnetic fields distorting the gradient magnetic field. However, the gradient magnetic fields determine how the MRI data is acquired in the k-space. A distorted gradient shape produces a distorted k-space trajectory which then results in a distorted image. The dynamic performance of the gradient system can be characterized by measuring its gradient impulse response function (GIRF). Furthermore, knowledge of the GIRF enables to correct the k-space trajectory and thereby reduce image distortions. In this work, we characterized the influence of the preclinical PET insert, i.e. the Hyperion IID, on the gradient performance of a 3 T MRI scanner. A GIRF of up to the second order spherical harmonics was determined by measuring frequency-swept gradient pulses with an NMR probe. We used the GIRF to correct the k-space trajectory of a single-shot spiral sequence. The low-pass characteristic of the gradient system slightly increased for a measurement with the PET insert. A slight difference of the k-space trajectory of the single-shot spiral sequence with and without PET insert was predicted. We showed an improved image quality by correcting the k-space trajectory using the measured GIRF.

Speaker: Nicolas Gross-Weege (Department of Physics of Molecular Imaging Systems, Institute for Experimental Molecular Imaging, RWTH Aachen University, Germany)

Slides PSMR2018_GrossWeege.pptx

09:10 A comparison of the hybrid kernel and the parallel level-sets approaches for the inclusion of MR information into PET reconstruction

Since the introduction of integrated hybrid PET/MR scanners, several techniques have been proposed to take advantage of the combined information from the two modalities. In particular, in this work, we investigate the performance of two different approaches: the hybrid kernelised expectation maximisation (HKEM) algorithm and the parallel level sets (PLS) method. Both are implemented in the open source Software for Tomographic Image Reconstruction (STIR). Simulated phantom torso data and real patient data with long and short acquisition times were used to compare reconstructions in terms of bias, Contrast Recovery Coefficient (CRC), and Contrast to Noise Ratio (CNR) using different ROIs to study different case scenarios. The quantification performances are good for both HKEM and PLS, and both always outperform the ordered subsets expectation maximisation (OSEM) algorithm. The HKEM shows excellent quantification, noise suppression and edge-preserving results for both long acquisition and short acquisition times. Results are equally good for PLS with long acquisition times but it is more difficult to fine-tune the parameters for optimal noise suppression with the short acquisition times.

Speakers: Daniel Deidda (Biomedical Imaging Science department, Department of statistics, University of Leeds), Kris Thielemans (University College London)

Slides PSMR2018_ID21_Deidda169.pdf

09:30 High-Resolution Heterogeneous Digital PET Brain Phantom based on the BigBrain Atlas

In positron emission tomography (PET), the evaluation of image reconstruction algorithms needs realistic simulated data sets where the ground truth is known and the image quality and the quantification errors can be evaluated. In the context of brain imaging, qualitative and quantitative assessments of the radiotracer uptake in anatomical regions, such as the striatum or the cortical grey matter, are important to study brain disorders. Therefore, brain phantoms that emulate brain scans are then needed to assess the accuracy of reconstruction and post-processing algorithms. However, most of the available digital brain phantoms are usually of limited spatial resolution making them not ideal to evaluate quantification errors due to the partial volume effect (PVE). In addition, they are piece-wise constant and usually generated from segmented MRI images of the brain. As a result, quantitative errors can be underestimated when doing regularized MR-guided reconstructions. In this work, a method to create high-resolution heterogeneous digital phantoms based on the BigBrain atlas is presented. A realistic [18F]FDG digital phantom that overcomes the problems of the current PET digital brain phantoms, particularly for the simulation of simultaneous PET-MRI datasets, was created using the proposed method. In the latter, the histology images and a tissue classified volume of the BigBrain atlas are used to define the high-resolution structures of the phantom. Then, the uptake in different regions of the brain is estimated using the Hammersmith brain atlas and a reconstructed image of real data from an [18F]FDG study. The phantom was evaluated by simulating a brain scan and reconstructing the data set with MLEM and MR-guided MAP algorithms. The reconstructed images were compared with real data and a phantom created from the Brainweb atlas. The proposed phantom could account for the heterogeneities in the [18F]FDG uptake, while this was not possible with the standard phantom. Furthermore, the MR-guided reconstructions of the proposed phantom obtained a modest partial volume correction compared to the Brainweb phantom that agrees with the performance of MR-guided reconstructions for the real data. The complete dataset including the [18F]FDG phantom, a pseudo CT, a μ-map and a T1-weighted image will be available online. In addition, the simulated sinograms will be also available in order to have a realistic brain dataset that can be used as a reference to assess image reconstruction methods.

Speaker: Martin A. Belzunce (King's College London)

Slides belzunce_psmr2018.pptx

09:50 Rigid Motion Correction for Brain PET/MR Imaging Using Optical Tracking

A significant challenge during high-resolution PET brain imaging on PET/MR scanners is patient head motion, which, due to the relatively long scan-time, is often observed. This challenge is particularly significant for clinical patient populations who struggle to remain motionless in the scanner for long periods of time. Head motion also affects the MR scan data. An optical motion tracking technique, which has already been demonstrated to perform motion correction on MR data, is used with a list-mode PET reconstruction algorithm to correct the motion for each recorded event and produce a corrected reconstruction. The technique is demonstrated on real Alzheimer patient data for the Signa PET/MR scanner.

Speaker: Matthew Spangler-Bickell (Nuclear Medicine Unit, IRCCS Ospedale San Raffaele, Milan, Italy.)

10:10 Motion-Corrected PET Reconstruction with SIRF

In order to be able to fully exploit the many advantages of simultaneous PET/MR, novel synergistic reconstruction methods are being developed by several groups. Access to flexible yet powerful reconstruction software is required, which requires components that are not currently available from manufacturers. For this reason, the Synergistic Image Reconstruction Framework (SIRF) has been developed. This open source software platform for integrated PET-MR image reconstruction allows researchers to easily prototype new algorithms. SIRF has been designed so that it is both easy to use and powerful at the same time. The ease of use comes from being able to interact with SIRF from both Matlab and Python. The efficiency of SIRF comes from the underlying code being in C++. Most of the functionality relies on existing open source packages, including Gadgetron and STIR. These engines are wrapped and presented to the user in Matlab and Python via a C-interface. Two new functionalities of SIRF are highlighted in this paper: registration (for motion correction) and anatomical priors. To demonstrate these features, a reconstruction workflow for dynamic PET data incorporating motion correction was implemented in SIRF. The method used frame-by-frame realignment, repositioning the attenuation image where necessary. Motion was derived from non-attenuation corrected PET images. An example case is given of an epilepsy FDG brain study, where non-negligible motion was present. A comparison was performed between three reconstructions methods: (i) non-motion corrected, (ii) motion corrected and (iii) motion-corrected with the anatomical parallel level sets prior (PLS). It was shown that with the incorporation of motion correction, the extracted time activity curves (TACs) no longer exhibited discontinuities. Furthermore, the use of a prior helped to suppress the noise in the reconstructed image. This paper makes no attempt to find optimal parameters for the prior, rather to highlight the new functionality to SIRF's users. Future work in SIRF will be directed towards the incorporation of the registration process into the reconstruction phase in both PET and MR, allowing for truly synergistic reconstruction.

Speaker: Richard Brown (UCL)

Slides PSMR.pptx PSMR_pdf.pdf

10:30 → 11:00 coffee break

11:00 → 12:40 Session 11 - Instrumentation: detectors and electronics

Convener: Dr Giancarlo Sportelli (University of Pisa and INFN Pisa, Italy)

11:00 A Novel Method for Determining the Intrinsic Detector Resolution on Monolithic Crystals

The aim of this work is to provide a method to retrieve the intrinsic resolution of PET detector blocks based on monolithic crystals. Estimating the intrinsic detector resolution is challenging, this resolution is limited by several factors that cannot easily be isolated and studied experimentally, for this reason many works focused on exploring the resolution limits through Monte Carlo simulations or other models. The proposed method can be done in-situ, and it is validated against the approach of using bench-top set-up where the coincidence detector is moved backwards. In our set-up each detector block is composed of a monolithic LYSO crystal, two different crystal thicknesses have been studied. The detector exit face is coupled to an array of 16×16 SiPMs. The proposed method consists on a software collimation of the measured data and on an empirical equation that has been deduced to fit the experimental data in which the detector intrinsic resolution follows a Gaussian distribution whereas the source shape, given its small size, follows a Lorentzian profile. We have alternatively analyzed the measured source profiles for very tiny collimation angles, and the intrinsic detector resolution was deduced using Voigt functions (convolution of Gaussian and Lorentzian profiles). The experiments resulted on a detector intrinsic spatial resolution of 0.6±0.1 mm and 0.7±0.1 mm in the case of 10mm and the 15 mm thick crystal, respectively. We found a good agreement between both methods. These tests show a method to determine the intrinsic resolution of monolithic-based detector blocks, with high accuracy.

Speaker: Antonio Gonzalez (Institute for Instrumentation in Molecular Imaging, i3M-CSIC)

Slides Gonzalez-Montoro_PSMR2018_00.pptx

11:20 A novel DOI Positioning Algorithm for Monolithic Scintillator Crystals in PET based on Gradient Tree Boosting

Monolithic crystals are considered as an alternative for segmented scintillator arrays in positron emission tomography (PET). Monoliths provide good spatial, timing and energy resolutions as well as intrinsic depth of interaction (DOI) encoding. DOI allows reducing parallax errors (radial astigmatism) at off-center positions within a PET ring. This is especially important for high-resolution PET systems such as organ-specific applications (e.g. breast, neuro-imaging or total body PET). We present a new DOI estimation algorithm based on the supervised machine learning algorithm Gradient Tree Boosting (GTB). GTB builds a predictive regression model based on a set of sequential binary comparisons (decision trees). GTB models have been shown to be implementable in FPGA if the data usage fits the available memory resources. We propose two optimization scenarios for the best achievable positioning performance: One restricting the available data usage to enable a future FPGA implementation and one without any restrictions. The positioning performance of the GTB models is compared with a DOI estimation method based on an isotonic regression (IR) model of a DOI observable. The DOI observable is defined as the ratio of the highest photon count to the sum of the photon counts of neighboring sensor channels. The usage of a DOI observable is comparable to other methods presented in the literature. Among others, we calculate the spatial resolution (SR) as the FWHM, the bias vector and the 90-th percentile distance of the positioning error distribution to quantify the DOI positioning performance. A fan beam collimator coincidence setup was used to perform a side irradiate of a 32 mm x 32 mm x 12 mm large monolith wrapped in highly reflective Teflon tape. We achieved an averaged SR of 2.16 mm FWHM and 1.98 mm FWHM for the IR and GTB, respectively. In contrast to the IR models, the GTB models show a nearly uniform positioning performance over the whole crystal depth: The SR varies from 1.5 mm to 2.7 mm FWHM minimum to maximum for the GTB instead of 0.9 mm to 5.6 mm FWHM for the IR models.

Speaker: Florian Müller (Department of Physics of Molecular Imaging Systems, Institute for Experimental Molecular Imaging, RWTH Aachen University)

Slides presentation.pptx (Protected)

11:40 Characterisation of prototype SiGe monolithic pixel detectors in the TT-PET scanner

The TT-PET collaboration is developing a compact Time-of-Flight PET scanner for small animals with a 30ps time resolution and sub-millimeter 3D detection granularity. The sensitive element of the scanner is a fully monolithic silicon pixel detector based on state of the art SiGe BiCMOS technology. Our scanner design will be introduced along with detailed GEANT4 simulations concerning hit acceptance and expected data rate for a 50MBq F18 source. First prototypes of our ASICs and DAQ system have been produced and characterised in the laboratory, and in test beams at CERN’s SPS facility using minimum ionising particles. Pixels with capacitances of 0.8pF were measured to have a detection efficiency of 99% and the electronics exhibit an equivalent noise charge of 600 e− RMS and a pulse rise time of 2 ns which matches our simulations. The realisation of a first scanner prototype for small animals is foreseen by 2019, and has been funded by the SNSF (Swiss National Science Foundation).

Speaker: Dean Forshaw (University of Bern)

Slides DCF_PSMR18_V2.pdf DCF_PSMR18_V2.pptx

12:00 First Measurements with the PETsys TOFPET2 ASIC Evaluation Kit

We present results obtained with the PETsys TOFPET2 ASIC Evaluation Kit. The TOFPET2 ASIC is a time-of-flight capable 64-channel ASIC and is a promising candidate to build high-performance TOF-PET systems. For a first characterization, we used two KETEK-PM3325-WB SiPMs each equipped with a 3x3x5 mm³ LYSO scintillation crystal. For the measurements presented in this work, we changed the discriminator threshold t1 = 5 - 30 in steps of 5 and the bias voltage V_bias = 29 V - 35 V in steps of 0.25 V and acquired data during 60 s. For all measurements, we performed an energy calibration including a correction for saturation. We evaluated the energy resolution, the coincidence resolving time (CRT) and the coincidence rate. At the optimal operating point, we obtained an energy resolution of about 9-7% FWHM, a CRT of approximately 210 ps FWHM and 395 ps FWTM, the coincidence rate showed only small variations of about 5% in the evaluated parameter range.

Speaker: David Schug (Department of Physics of Molecular Imaging Systems, Institute for Experimental Molecular Imaging, RWTH Aachen University, Aachen, Germany)

Slides 2018-05-23_Schug_TOFPET2_Initial_Results.pptx

12:20 MRI compatible power supply for PET detectors of an integrated PET/MR scanner

Power supply of PET detectors is a serious issue for integrated PET/MRI systems, due to mutual interaction between power supply electronics and MRI magnetic fields. In this work, we developed a prototype of a magnetic compatible power supply, that can be safely placed and used in a MRI magnetic field. The power supply is a highly compact (5.2x5.2 cm) switching DC-DC converter iron-core free. The power supply works in the Very High Frequency (VHF) range, between 24.5 and 26.5 MHz, so as to avoid RF coupling between power supply signal harmonics and RF magnetic field of a 1.5 T MRI scanner, that works at 64 MHz. Power supply output parameters (power and voltage) have been chosen to match those of the PET boards of the PET/MRI TRIMAGE scanner, in which the power supply could be placed. In particular, the main goal was to convert 12 V DC input to 3.3 V DC output with 10 W output power. At present, the power supply is able to convert 12 V input to 5 V and 10 W output with a measured efficiency of 70.5% without need of a dedicated cooling system. Magnetic compatibility has been verified in a 1.5 T and 7 T MRI scanners, where no artifacts induced by the circuit were observed in MRI images and MRI scanners performances were similar in presence and absence of power supply.

Speaker: Vito Gagliardi (IRCCS Stella Maris)

Slides Neolite_-_PSRM2018.pptx

12:40 → 15:50 Lunch break

15:30 → 17:40 session 12 - Clinical systems and applications

Convener: Dr Jorge Cabello (Klinikum r. d. Isar / TUM, Germany)

15:30 PET-MR and SPECT-MR Multimodality Probes: Development and Challenges

Speaker: Yang Chang-Tong (Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore)

Slides PS-MR_2018.pptx

16:00 Design and preliminary performance of the MR component of the TRIMAGE PET/MR/EEG system

Speaker: Julien Rivoire (RS2D)

Slides P2.pptx

16:20 Clinical evaluation of 11C-Met-avid pituitary lesions using a ZTE-based AC method

Pituitary tumours account for ~16% of central nervous system tumors and they are the second most frequently reported histology in this group. Due to their small size, pituitary surgery is challenging and precise lesion localization through imaging is therefore a critical factor for a successful outcome. Simultaneous PET/MR is well suited for lesion identification and localization but it requires accurate attenuation correction to ensure optimal PET imaging. Atlas-based attenuation correction methods are often used for this purpose, as they overcome the difficulty of estimating bone tissue density with conventional MR sequences. However, atlas methods can only partially account for inter-patient variability. The goal of the present study was to investigate whether direct bone measurement, by means of a zero echo time (ZTE) MR sequence, can significantly improve the accuracy of pituitary tumor imaging with PET.

Speaker: Gaspar Delso (UniversitätsSpital Zürich)

Slides Delso2.pptx

16:40 Multi-delay arterial spin labeling MRI accurately identifies reduced perfusion in Moyamoya disease: comparison with a PET/MRI database

We directly compared standard and multi-delay arterial spin labeling (ASL) MRI measurements of cerebral blood flow (CBF) to simultaneous reference [15O]-water PET scans on hybrid PET/MRI in patients with Moyamoya disease. For these Moyamoya patients (N=15) with extremely long arterial transit times, multi-delay ASL outperforms standard ASL in regional correlation and reduces bias relative to PET. We also constructed a voxel-wise, normative CBF database from healthy controls (N=15) with PET/MRI, and identified regions of hypoperfusion in frontal and parietal regions of patients. Multi-delay ASL is more specific to areas of Moyamoya hypoperfusion (more similar to the PET reference standard), whereas standard ASL overestimates these areas due to low signal.

Speaker: Greg Zaharchuk (Stanford University)

Slides P4.pptx

17:00 PET stability measurements during simultaneous multi-nuclear (23Na) MRI/PET acquisition

PET/MR presents a unique opportunity to acquire simultaneous, complementary, functional imaging data in human and animal studies. Specifically, 23Na and 13C, termed 'x-nuclei', present novel imaging contrasts which inform upon cellular function. Due to the differing magnetic properties of x-nuclei, increased demands are placed upon gradient and RF hardware for acquisition. However, little data has been shown assessing the performance of PET detectors during x-nuclei MR imaging acquisitions. This study demonstrates negligible effect of a 3D 23Na-MRI sequence on the PET detection performance and count rate. We believe these results will help inform future studies integrating 23Na-MRI in simultaneous PET/MR acquisitions.

Speaker: Gaspar Delso (GE Healthcare)

Slides Delso1.pptx

17:20 Final Remarks

18:00 → 19:00 Farewell cocktail