Program Listings

Session Chair: Andrew Goertzen, University of Manitoba, Canada; Patrick Hallen, RWTH Aachen University, Germany

PET and SPECT are widely used for medical imaging. However, radio isotopes (RI) available for SPECT (RIs with its energy + decay RIs) are limited. Under these circumstances, it is expected the appearance of the new gamma imaging detector which can measure more various kinds of gamma-ray sources in order to develop new biomarkers using new RIs. We have developed an Electron-Tracking Compton Camera (ETCC) for medical imaging due to its wide energy dynamic range (0.1-2MeV) and abilities of both background rejection and clear imaging using the tracking information of the recoil electron since 2005. Thus this camera has a potential of developing the new reagent for molecular imaging. Until now we have carried out several imaging reagent studies such as: F-18, I-131, Zn-65, etc in mouse for imaging test with high energy gamma emitting RIs. In 2013 we have improved readout electronics system of ETCC, by which tracking efficiency and maximum counting rate were dramatically improved, and eventually its sensitivity has been improved by a factor of >10. Here we present the performance results using this new ETCC such as the imaging test using F-18 in point-like and rod-like phantoms with varying the intense of radiation. The position resolution achieves less than about 8 degrees from 10 degrees at 511keV by this improvement. Further improvement of the angular (position) resolution will be presented until 2015 spring. Also, we are developing the next ETCC by increasing the GSO scintillator thickness from 1 rad. to 2 rad. and the gas pressure from 1 atm to 2 atm which improves the detection efficiency by a factor of >5 at 511keV. By these improvements, the imaging time of mouse is expected to be reduced from several hours with to <20 minutes for lots of kinds of RIs with the energy band from 0.1-2MeV. In addition, we will setup two ETCCs with the above improvements, and take a tomographic image of a rod phantom by using them.

According to the technical requirement of the time coincident measurement for TOF-PET systems, an eight-channel wave-union time-to-digital convertor (TDC) based on a Spartan-6 Xilinx FPGA has been implemented and tested. The root-mean-square (RMS) error of the two channels on the same FPGA measuring a fixed time interval is less than 14 ps, while the RMS error of the two channels inter-FPGA is less than 26.5 ps. The TDC measurement dead time is only 8 ns, which is conducive to the trigger mechanism combining a low threshold for timing and a high threshold for event confirmation. Because the time coincidence measurements in PET scanners always occur between the opposing TDC circuits, the variation of the RMS error and measured mean value by the two inter-FPGA TDC channels in different temperature environments is also presented.

The digital silicon photomultiplier (dSiPM or DPC) has gained popularity for researchers developing advanced PET detectors due its simple readout structure and excellent intrinsic timing resolution without the need for additional high-performance electronics. Although high spatial resolution detectors have been developed with dSiPMs using monolithic scintillators and statistical algorithms for event positioning, the readout of these devices is best suited for one-to-one coupling of scintillator crystals with 4 x 4 mm² cross-section to the dSiPM pixels for minimal sensor dead time and event processing requirements. To effectively double the dSiPM spatial sampling and resolve smaller crystals, as well as introduce depth-encoding capability, we propose to employ dual-ended readout where the top detector module is offset in both directions by half the dSiPM pixel pitch. In this way, we can resolve 2 x 2 x 20 mm³ crystals using the same readout as with 4 x 4 x 20 mm³ crystals (no neighbor logic or light sharing is required) as each crystal will have a unique die - pixel combination when both dSiPMs are considered. Crystal identification accuracy, energy resolution and timing resolution were measured with two 2 x 2 x 20 mm³ polished LYSO crystals. On average, >99% of events were assigned to the correct crystal, an energy resolution of 14.7% was measured and the best coincidence timing resolution achieved was 165 ps. Using a 14 x 14 array of polished LYSO crystals separated by ESR (2.07 mm pitch), we measured event rates in each crystal and energy resolution. The average energy resolution in the array was 14.7+/–3.0%. The event map showed loss of events for the 42 crystals that were positioned over the 1.6 mm dead-spaces between rows of dSiPM dies. In future work, we will address event losses and optimize the crystal array (surface finish, inter-crystal reflector) for dual-ended DOI performance, and measure the DOI resolution.

Recently, several studies have been carried in order to determine potential capabilities of BGO scintillator crystals coupled to SiPM photosensor devices for PET and SPECT applications. The prior studies have been typically done on small size BGO samples. Despite the fact that timing capabilities for devices based on BGO are worse than those based in commonly used fast scintillators, such as LYSO and LSO, the price of BGO material is considerably lower and, thus, BGO could be an option for systems where the required scintillator volume is significantly higher. In this report we present results of further studies using 12x12 SiPM arrays of the 3 mm C-Series SensL sensors, placed at a pitch of 4.2 mm, and coupled to the readout recording the 12 row and 12 column SiPM array signals. Two types of pixellated BGO crystals were tested: an array of 10x10 elements with 2.5 mm pitch and 10 mm thickness and an array of 30x30 pixels with 1.67 mm pitch but only 3 mm thick. A staggered depth-of-interaction (DOI) configuration was also evaluated using two layers of 2.5 mm pitch BGO pixels, with 10x10 (top) and 11x11 (bottom) elements with a total thickness of 20 mm. An energy resolution as good as 12% FWHM has been obtained. Our new results confirm that either single layer crystal arrays with pitch values as low as 1.67 mm or two staggered layers of 2.5 mm pitch could be well suited for PET applications, especially for large systems or low cost dedicated PET systems

Lu_1.8Gd_0.2SiO₅(LGSO) scintillators have high light output, high density, fast decay time and also have enough timing resolution to use for TOF-PET system. We have already shown that phoswich LGSO with two different decay times is promising for use in TOF-DOI-PET detector. However, it is unclear what phoswich LGSO configuration is better to obtain good timing resolution; i.e. which type of scintillator should be arranged in the upper side. In this study, we measured the energy spectra and the timing resolutions of the two different configurations of phoswich LGSO detectors. We used two types of LGSO scintillators with different Ce concentration, one is LGSO-fast (0.025 mol% Ce, decay time: 36 ns) and the other is LGSO-slow (0.75 mol% Ce, decay time: 48 ns). They were optically coupled in depth direction for phoswich configurations. The phoswich LGSO configurations were “LGSO-fast/LGSO-slow” arranged with LGSO-fast on the upper side and LGSO-slow at lower the side, and “LGSO-slow/LGSO-fast” arranged with LGSO-slow on upper side and LGSO-fast at the lower side. The phoswich LGSOs were optically coupled to silicon photomultipliers (Si-PMs) and measured the timing resolution with a high bandwidth digital oscilloscope. The best timing resolution was 260 ps FWHM at 5 mm upper position from the vertical center of phoswich LGSO-fast/LGSO-slow. The timing resolution at the vertical center of LGSO-fast/LGSO-slow was much better than that of LGSO-slow/LGSO-fast. The timing resolution in head-on condition was slightly better for LGSO-fast/LGSO-slow. Therefore, we conclude that LGSO-fast/LGSO-slow is a better use for TOF-DOI-PET detector.

Objective
Recent small PET or Compton camera applies depth-of-interaction (DOI) information to achieve higher resolution in image reconstruction. We have developed a novel discrimination method for decay time of scintillation signal using peak-to-charge ratio (Vp/Q) of output waveform from a photomultiplier tube (PMT). Since the discrimination method can recognize 1 ns difference of decay time, multi-layer DOI discrimination can be achieved by using adequate scintillators with different decay constant. Moreover the Vp/Q discrimination method can suppress a crosstalk event because the value of Vp/Q is quite different from true value when crosstalk event is occurred.
Materials and Methods
Feasibility study on the multi-layer DOI detector system, GEANT4 based Monte Carlo simulation was performed including optic transport. Scintillators for the simulation were GSO:Ce and LuAG:Pr scintillators with different decay constants, the scintillator size was assumed 2.5 mm cube each. The decay constants for each scintillator were assuming a realistic range taken from published research, the best combination of the decay constant was determined.
Results and Discussions
By consideration of decay-constant combination, we assumed the decay constants for GSO:Ce scintillator were 30, 36.8, 44.7, 56.5 and 73.4 ns, and that for LuAG:Pr scintillator were 15, 16.6, 18.6, 21.1 and 24.3 ns, respectively. Larger deviation from true Vp/Q value for cross talk event can be expected by using more different decay constant in neighboring scintillators, when GSO:Ce and LuAG:Pr were arranged alternately, the presented method can recognize 99.15 % event correctly and 89.5% crosstalk-event was suppressed.
Conclusion
This novel DOI technique can increase the number of DOI layers by simply combining several scintillators with different decay constant. This method has also advantage to suppress crosstalk event, improvement in better spatial resolution will be expected.

To obtain good timing resolution with SiPM-based time of flight (TOF) PET detectors, one-to-one coupling of crystals to SiPM elements, and the use of discriminators to extract timing information for each SiPM signal are typically required. In this study, a simple readout method is introduced and evaluated using 4 x 4 arrays of 3.16 mm x 3.16 mm x 20 mm LYSO crystals directly coupled to 4 x 4 arrays of 3.16 mm x 3.16 mm through-silicon via (TSV) SiPMs. In our proposed readout, the timing information was extracted from each SiPM signal, and the crystal information that detects gamma photon was identified using a discriminator and encoder. The preliminary results show that a global timing resolution of 475.9 ± 12.6 ps was obtained using two identical detectors working at a bias voltage of 28.0 V and a temperature of 5 °C, indicating that the proposed method works well. Further experiments, including obtaining timing resolution for each crystal element and optimizing the bias voltage and threshold for the comparators, are ongoing to further evaluate the readout method. The proposed method was evaluated using a 4 x 4 array of SiPMs, but can also work for larger SiPM arrays with more elements by scaling the electronics.

Although I-131 is used for radionuclide therapy, high resolution image is difficult to obtain because of the high energy of I-131 gamma photons (364keV) with the conventional gamma cameras. Cerenkov-light imaging is a possible method for beta emitting radionuclides and I-131 (606MeV maximum beta energy) is a candidate to obtain a high resolution image. Thus we developed a high energy gamma camera system for I-131 radionuclide and combined it with a Cerenkov-light imaging system to form a gamma-photon/Cerenkov-light hybrid imaging system to compare the simultaneously measured images of these two modalities. The high energy gamma imaging detector used 0.85mm x 0.85mm x10mm thick GAGG scintillator pixels arranged in 44 x 44 with 0.1mm thick reflector and optical coupled to a Hamamatsu 2 inch square position sensitive photomultiplier tube (PSPMT: H12700 MOD). The gamma imaging detector was encased in a 2cm thick tungsten shield and pinhole collimator was mounted on its top to form a gamma camera. The Cerenkov-light imaging system was made of a high sensitivity CCD camera (Hamamatsu photonics, ORCA-II ER). The Cerenkov-light imaging system was combined with the gamma camera using optical mirrors to image the same area of the subject. With this configuration, we could simultaneously image the gamma photons and Cerenkov-light from the I-131 in the subjects. The spatial resolution and sensitivity of the gamma camera system for I-131 was 3.5mm FWHM and 10cps/MBq for high sensitivity collimator at 10 cm from the collimator surface, respectively. The spatial resolution of the Cerenkov-light imaging system was 0.64mm FWHM at 10 cm from the collimator surface. The thyroid phantom and nude rat images were successfully obtained with the developed gamma-photon/Cerenkov-light hybrid imaging system. The developed gamma-photon /Cerenkov-light hybrid imaging system is useful to evaluate the advantages and disadvantages of these two modalities.

Stable production techniques for high lutetium content LGSO (Lu_1.8Gd_0.2SiO₅:Ce) with decay times varying from ~30-40 ns have been established over the past decade, and the decay time can now be accurately controlled with varying cerium concentration (0.025-0.075 mol%). This material is promising for time-of-flight positron emission tomography (TOF-PET), as it has similar light output and equivalent stopping power for 511 keV annihilation photons compared to industry standard LSO:Ce and LYSO:Ce, and the decay time is improved by more than 30% with proper Ce concentration. This work investigates the achievable coincidence timing resolution (CTR) with 90%Lu LuGSO:Ce(0.025 mol%) when coupled to recently developed and emerging silicon photomultipliers. As one example, the CTR between two 3x3x20 mm³ 90%Lu LuGSO:Ce crystals coupled to SensL MicroFC-30035 silicon photomultipliers was measured to be 150±2 ps FWHM at a bias of V_br+8 V. Measurements of light yield, energy resolution, and intrinsic rise and decay times are also presented for a thorough investigation into the timing performance with this scintillator.

Occurrence of pileup is a significant concern in PET systems, especially when scanning high radioactivity. It decreases the number of events acquired, degrades the statistics of the data, leads to incorrect event time and coincidences, and created mis-positioned events in block detectors. In this paper, we propose a digital pileup correction method for the Trans-PET system that employs the MVT (multi-voltage threshold) sampling DAQ for obtaining event time and free-run ADC (analog-to-digital converters) for obtaining pulse heights. The basic idea of the method is to terminate the integration of the ADC samples for calculating pulse heights when a second pulse is detected. The resulting pulse height is then corrected in accordance with the known pulse shape and the integration time. On the other hand, the pulse height of the second pulse is corrected by subtracting the contribution from the first pulse, which is equal to the difference between the corrected and uncorrected pulse heights of the first pulse. This correction method has been implemented in hardware and our initial experimental results demonstrate that application of the correction method can significantly recover count loss, mitigate degradation of the energy resolution, and improve the position histogram.

Spatial resolution of clinical PET scanners is approximately 4 mm, limited mainly by detector intrinsic resolution and also by non-colinearity of 511 keV photons. However this level of resolution is insufficient for small animal imaging, which normally requires a dedicated high resolution scanner. High resolution 511 keV SPECT is an alternative, but this approach requires significant shielding and is not really practical as a low-cost insert for clinical cameras. Combining pinhole collimation with coincidence detection offers a practical way to achieve high resolution (< 1.5 mm) with clinical PET detectors. Each event is associated with three locations (two crystals, one pinhole), thereby overdetermining the line of response and providing additional information over coincidence PET or pinhole SPECT alone. Coincidence data are used to determine whether a 511 keV photon likely passed through a specific pinhole. If not, the event is rejected. If so, the photon is assumed to pass through that pinhole, and spatial resolution then is limited by the effective pinhole diameter. Other advantages of this approach include reduced shielding requirements (versus pinhole SPECT), elimination of non-colinearity error (versus PET), and elimination of overlapping multi-pinhole projections. Monte Carlo simulations were performed for a 54-pinhole mouse wholebody collimator with clinical PET detectors. Reconstructed images demonstrated the ability to resolve 1.4 mm rods in a hot rods phantom. Sensitivity was 0.10% and 0.075% for pinholes having 1.4 mm and 1.0 mm diameter, respectively. These simulations indicate acceptable performance for small animal imaging with clinical PET detectors. Performance improvements are expected through further optimization of pinhole and collimator design.

The main advantage of the radio-guided surgery (RGS) technique exploiting beta- radiation, compared to the traditional RGS using gamma radiation, is a more favorable tumor-to-non-tumor activity ratio (TNR). Operating in a low background environment allows the extension of the RGS applicability also to cases with large uptake of the tracer from healthy organs next to the lesion, minimizing the radiotracer activity to be administered with obvious extremely positive implications for the patients. Our first study cases are meningioma brain tumors since an appropriate beta- emitting tracer is already available (90Y-DOTATOC) but the goal is to apply the technique to gliomas. Feasibility studies on meningioma and gliomas show high potentialities of this new treatment, since TNR=10 has been observed for 8/10 patients affected by meningiomas and TNR=4 for 9/12 patients affected by gliomas. To validate the technique, we developed a prototype of the intraoperative probe detecting beta- decays with millimetric scintillator core made of para-terphenyl, adopted due to its high light yield and low density. The probe is tested in laboratory with specific phantoms reproducing finite size meningioma remnants embedded in healthy brain tissue and saturated with a liquid source diluted to reach the mass and concentration (10 kBq/ml) and the TNR as expected in a real clinical case. Clinical tests on patients affected by meningiomas are foreseen in the very near future.

To develop a small animal PET scanner for mouse-whole-body imaging, not only structures of detector blocks but also signal readout schemes are important factors to achieve sub-millimeter spatial resolution because both specifications affect the intrinsic spatial resolution. A sub-millimeter resolution PET detector composed of a 1.2 mm pitch MPPC array with TSV structure one-to-one coupled to LFS scintillator crystals has been constructed and evaluated. The one-to-one coupling scheme allows us to read out individual MPPC signals of the MPPC array corresponding to each crystal element independently. The smaller the crystal element, the more frequently inter-crystal scatter events occur. Therefore, the individual signal readout enables to be better spatial resolution compared to the centroid calculation readout, since, a first interacted crystal element by an incident gamma-ray can be more accurately detected by using energy information of not only a channel given the maximum energy but also channels around it. We constructed prototype detectors that were composed of an 8 × 8 TSV-MPPC array, each having a 1.0 mm × 1.0 mm active area with 1.2 mm pitch, and a LFS scintillator array of 8 × 8 with 1.2 mm pitch. The dimensions of each crystal element are 1.13 mm × 1.13 mm × 10 mm. To evaluate an intrinsic spatial resolution, coincidence response functions (CRFs) of the detector pair located at parallel with 80 mm distance were measured by scanning a 0.25 mm diameter 22Na point source with 0.1 mm steps. The data were acquired by the CAMAC system. The average FWHM of the CRFs was obtained 0.80 mm without corrections of the point source size. Additionally, to reduce positioning errors, an interaction-order defining algorithm based on Compton kinematics was adapted to the same data and we discussed on inter-crystal scatter effects comparing errors arising from the interaction-order defining algorithm and the centroid calculation.

For laparoscopic surgery, an intra-operative Compton camera aimed to detect metastatic lymph node with a radiopharmaceutical is being developed in our group. The GATE Monte Carlo simulation platform based on the Geant4 toolkit was used for Compton camera simulations with assumption of using HAMAMATSU TSV MPPC S12892PA-50 (2.4 x 2.4 mm2, 1504 micro cells of 50 x 50 um2) and 2 x 2 x 3 mm3 GAGG scintillator crystals. Simulation results of 8 layers of 12 pixel arranged module show that it has 3 mm FWHM of spatial resolution for 10 mm distance from the first layer, and has about 0.5 % of intrinsic efficiency. Preliminary experiments result showed 6.3 % FWHM of energy resolution for 662 keV, which is sufficient for imaging requirements of our geometry.

Ultrahigh resolution small animal PET system requires small pixel size scintillators. We developed a ultrahigh resolution small animal PET system using fine LYSO pixels. The size of the LYSO pixels used was 0.32mm x 0.5mm x 5.0mm. The LYSO pixels were combined into 22 x 15 matrix with 0.1mm thick BaSO4 reflector between the pixels. The size of the LYSO block was 9.24mmx 9.0mm x 5mm. The LYSO block was optically coupled to a 4 x 4 silicon photomultiplier (Si-PM) array with small gaps (Hamamatsu, S12642-050) with a 1-mm thick light guide. We made 8 Si-PM based block detectors and arranged in octagonal shape to form a PET detector ring. The detector ring diameter is 22.3mm. The analog signals from the detectors were fed to the weight summing boards with 1-m long, small diameter coaxial cables. The weighed sum signals were fed to 100-MHz analog to digital (A-D) converters of the data acquisition system and integrated for 320 ns; the positions were calculated using the Anger principle by FPGA. The coincidence was also measured digitally among the eight block detectors. The spatial resolution of the developed PET system at center measured with 100µm size Na-22 point source and reconstructed by filtered back projection (FBP) was 0.6mm FWHM. The sensitivity at axial center was 0.5%. The peak noise equivalent count rate (NECR) was 12.5k cps. We could obtain high resolution images of phantoms and small animals with the developed PET system. In the images of small animals, small structures of the head were observed. With these results, we conclude that high resolution PET system is possible with 0.32mm pixel LYSO scintillators.

Positron emission tomography (PET) has high sensitivity for imaging the radioactive tracer distribution in subjects. However it is not possible to image the free radical distribution in subjects by PET. The Overhauser enhanced MRI (OMRI) is so far an only method to image the free radical distribution in vivo. Combining PET and OMRI may become possible to simultaneously image the radioactive tracer and free radical distributions. For this purpose, we developed a PET/OMRI combined system for small animals. PET/OMRI combined system consists of an optical fiber based small animal PET system combined with a permanent magnet based OMRI system. The optical fiber based PET system used flexible optical fiber bundle. Eight optical fiber based block detectors (16 LGSO blocks) were arranged in a 56mm diameter ring to form a PET system. The LGSO blocks were located inside the field-of-view (FOV) of OMRI and the PSPMTs were positioned outside the OMRI to minimize the interference between PET and OMRI. The OMRI system used 0.0165T permanent magnet with resonance frequency of 0.7 MHz for nuclear magnetic resonance (NMR) and 670MHz for electron spin resonance (ESR). The system has an ESR coil to enhance the MRI signal by the use of Overhause effect to image the free radical in the field-of-view of the PET/OMRI system. For measuring the OMRI image, RF pulses for ESR are irradiated just before the MRI sequences to the subject. The ESR coil has a 20mm diameter single or multiple ring surface coils located inside the MRI receiver coil. Spatial resolution and sensitivity of the optical fiber based PET system were 1.2mmFWHM and 1.2% at the central field of view (FOV), respectively. The OMRI system could image the distribution of nitroxyl radical solution. Simultaneous imaging of positron radiotracer and nitroxyl radical solution was possible with the developed PET/OMRI system for phantom and small animal studies. The developed system has potentials to provide interesting results in molecular imaging research.

The major advantages of a novel SPECT system with a Variable Pinhole (VP) collimator are its high sensitivity and spatial resolution for the Region of Interest (ROI). The shape of the collimator is optimized to the ROI for each rotation angle. A previous collimator simulation study showed a marked improvements in, the collimator sensitivity when compared to a conventional pinhole system, and a submillimeter phantom image was well discriminated. Realization of this SPECT system requires the development of a gamma ray detector with a submillimeter level intrinsic resolution, wider detector size, and enhanced Signal to Noise Ratio (SNR). The scintillator array consisted of a pixelated Ce:GAGG (Furukawa Co., Ltd.). The size of the scintillator array is 26.7 mm×26.7 mm (31×31 pixels), and the pixel dimension is 0.7 mm×0.7 mm×5 mm. Silicon photomultiplier is tiled into 2 × 2, and the dimension is 28.6 mm × 27.2 mm. The readout analog circuit of prototype system is composed of charge-sensitive preamplifier, a resistive chain, and a CR-RC shaping circuit. The flood map image is acquired by Na-22 (57.99 µCi) and Ba-133 (86.20 µCi). The peak-to-peak distance between the pixels is 0.8 mm, and the separations are sufficient to set the position boundaries for both RIs. In this study, we developed a thin and light gamma ray detector for VP SPECT system with a high resolution, and wide area.

Objectives: Quantitative small-animal SPECT of the combined ß- and ?-emitter I-131 is highly important for development of e.g. new sodium-iodine symporter (NIS) based radioiodine therapies of cancer. To quantify I-131 concentrations in organs of mice it is essential to have a high image resolution and to limit effects of pinhole wall photon penetration of the high energy (364 keV) photons emitted by I-131. Here we test I-131 SPECT capabilities when a dedicated high energy cluster multi-pinhole collimator is used in a SPECT system with large stationary NaI detectors. Methods: A VECTor/CT system (MILabs, The Netherlands) was equipped with a collimator with 162 clustered pinholes (0.7 mm). An energy window of 20% around 364 keV was applied with 14.5 keV scatter windows on each side. Data were reconstructed with pixel-based OSEM with a system matrix that models specifically for 364 keV the distant-dependent resolution and sensitivity of each pinhole, as well as the intrinsic detector blurring and depth-of-interaction effects. The system performance was measured with a uniformly filled cylinder and a hot-rod Derenzo resolution phantom. Capabilities of in vivo imaging were tested in a total body and a focused scan of the mouse thyroid with I-131-NaI. Results: The resolution reached 0.5 mm in terms of minimal visible hot rod diameter. Quantitative accuracy measured with a uniformly filled cylinder scan results in 98.5% ± 4% of the gold standard value. In vivo mouse scans illustrated a clear separation between the thyroid lobes and detailed anatomical structure of the glands, even in total body scans. Conclusion: Performing quantitative 0.5 mm resolution I-131 SPECT in live mice is possible using dedicated high energy collimation. This is beneficial for preclinical development of i.e. targeted cancer therapies with NIS I-131-labeled compounds.

Scintillator is an important component which determines the overall performances of PET detectors, and scintillators with fast timing in addition to high light yield are required to achieve better performances, especially for TOF PET scanners. Recently, there have been several reports on Lutetium Fine Silicate (LFS) scintillator showing its improved timing and energy resolution than LYSO (LSO), which has been commonly used for TOF PET. In this paper, we present the results from a comparison study to measure the energy and timing performances of LYSO and LFS scintillators. In our test, the scintillators are coupled to Hamamatsu R9800 PMTs, and the detector signals are readout by a DRS4 waveform digitizer at 5 GS/s sampling. The test are carried out using scintillators with three different lengths (10, 15, and 20 mm) to measure their effect on timing resolution. Preliminary results show that LFS has a slightly shorter decay time (38-40 vs. 42-43 ns) and smaller time resolution (252-263 vs 257-275 ps FWHM) than LYSO; the energy resolution at 511 keV is measured to be ~10% FWHM for both scintillators. The measurement using a new Hamamatsu MPPC, S12572-050P, is on going, and the details of the study are presented in the paper.

We have developed a new detector for PET using multi-pixel photon counter (MPPC) for MRI compatible application. This module has an 8 x 8 MPPC array, each segment has 3 mm x 3 mm active area and the pitch of the array is 4.1 mm in both directions. The MPPC array is connected to the front-end circuit with a detachable flexible printed circuit cable, it would provide flexibility for the detector arrangement. The front-end circuit consists a preamplifier ASIC and a two-dimensional register network, which multiplexes the 64 signals from the MPPC array into four position-encoding signals. The preamplifier ASIC also has a sum output, which is used for timing pick-off and energy discrimination. Coupled with four-layered LYSO arrays and covered with a copper shielding, basic performances were evaluated with a 3T-MRI. Very little interferences were observed in the flood images with MRI sequences.

Higher sensitivity in PET systems would allow for a higher effective image spatial resolution, better lesion detectability, shorter scans or scans at lower radiotracer doses. However, a significant increase in the sensitivity can only be achieved by increasing the solid angle coverage as current detectors based on 2-3 cm LYSO scintillators already achieve 70%-90% stopping power. Therefore it is proposed to design a PET scanner with an extended axial field of view (AFOV). However, long AFOV PET scanners face multiple factors that cause image degradation compared to conventional PET: higher number of random and scattered events, higher attenuation along oblique LORs, and more influence of depth of interaction (DOI). In this work we investigated the design of a PET detector based on 20 mm thick monolithic scintillators with single side readout for a long AFOV PET scanner. The goal is to simultaneously achieve good energy resolution, high spatial resolution and provides both good TOF and DOI capabilities. The detector consists of 32x32x20 mm3 LSYO crystals, all sides polished, wrapped in Teflon tape and coupled to the Digital Photon Counter (DPC). Event processing used the Mean Nearest Neighbor (MNN) method. Calibrations with a collimated pencil beam at known positions were performed to obtain reference data for the MNN method. The detector performance was evaluated in terms of energy resolution, coincidence resolving time (CRT), intrinsic spatial resolution and DOI decoding capabilities. The measured energy resolution of the detector was 13.4% and the CRT was 270 +/- 19 ps. An average intrinsic spatial resolution of 3.24 mm FWHM in the x- and 2.77 mm FWHM in the y- direction was obtained with the MNN method. The mean positioning bias was 0.19 mm. The achieved DOI was at the level of 4 mm. The achieved results are highly promising, however further optimization of the detector's design is still required to improve its performance.

We are developing a new PET-MRI system based on our novel concept: 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 by minimizing the size of the PET detector ring, i.e., minimizing the number of PET detectors. In the proposed system, PET detectors are extremely close to the MR head coil. To reduce electromagnetic interaction between the PET detectors and the MRI 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. Moreover, the electromagnetic wave noises from the power supply are not negligible. Following our previous studies, we are developing the second prototype of our PET-MRI system. In this paper, we develop shielding techniques for the second prototype. And we evaluate comprehensive performance for the shield materials and a filter circuit for the power supply in terms of signal-to-noise ratio (SNR) and eddy current that has a relatively short time constant for attenuation. The decreasing rate of SNR in the simultaneous operations with and without the filter circuit for the power supply was measured. These results showed the decreasing rate of SNR was suppressed by a maximum of 7% for the Cu foil shield as against 26% without it. When the filter circuit for the power supply is used with carbon roving, the decreasing rate of SNR is improved from 15.5% to 3.8%. Based on the secondary magnetic field (?B0) near the shield box using Cu foil, ?B0 with carbon roving was suppressed about 80%. These results showed that carbon roving with the filter circuit for the power supply has higher shielding performance and suppression capacity for the eddy current though PET detectors are extremely close to the MR head coil.

There is a strong potential demand for high-sensitivity and low-cost brain positron emission tomography (PET) imaging that is applicable to early diagnosis of Alzheimer’s disease. Therefore, we have proposed a high-sensitivity dedicated brain PET geometry composed of a helmet detector having a hemisphere shape and a chin detector, which we call helmet-chin PET. Because the shape of a human head is a sphere, the hemispherical arrangement of the detectors allows closer positioning of detectors and better sensitivity than the conventional cylindrical arrangement. In addition, adding detectors around the chin position significantly improves the sensitivity at the center where the cerebellum is located. For a proof-of-concept of the helmet-chin PET, we developed the first prototype of the helmet-chin PET using 4-layer depth-of-interaction (DOI) detectors. The helmet detector for the prototype system was realized by multiple rings having different numbers of detectors and a cross-shaped part covering the top. We used in total 54 DOI detectors, each of which consisted of 1,024 GSOZ crystals with dimensions of 2.8×2.8×7.5 mm³ and a 64-ch flat-panel photomultiplier tube. In performance evaluations, we determined there were uniform spatial resolutions of 3.0 mm by an analytical method and 1.4 mm by an iterative method. Peak sensitivity was measured as 10 % at a region near the top of the head, which was almost equivalent to the central sensitivity of the cylindrical PET composed of 120 DOI detectors. Also, we performed an initial imaging test with a brain phantom and we reconstructed the images with and without the chin detector. We found the slice near the bottom of the helmet detector had strong noise without the chin detector, while the slice had good imaging performance with it, and the overall image quality was improved. Therefore, we concluded that the helmet-chin PET had high potential for realizing high-sensitivity, low-cost, and accurate brain imaging.

We developed a motorized variable angle slant-hole collimator (VASH) for use a breast-specific gamma imaging detector. The collimator system consists of a stack of 48 tungsten sheets, four push blocks, and two motorized actuators for moving the push blocks. User controlled shearing of the stack of tungsten sheets can provide multiple projection angles allowing the implementation limited angle tomography also known as tomosysnthesis. Each individual sheet has an edge profile with a different angled cut on each side. Face matching push blocks, driven by two servo motors provide accurate positioning of each sheet resulting in the desired collimation angle. A prototype plastic model was built using a 3D printer to verify the design of the drive mechanism. The final product was built using 48 collimator 0.25 mm tungsten sheets. The achieved angular range of the VASH collimator is ± 28 degrees.

A new multi-modality imaging tool is under development in the framework of the INSERT (Integrated SPECT/MRI for Enhanced Stratification in Radio-chemo Therapy) project, supported by the European Community. The final goal is to develop a custom SPECT apparatus that can be used as an insert for commercially available MRI systems. INSERT is expected to offer more effective and earlier diagnosis with potentially better outcome in survival for the treatment of brain tumors, primarily glioma. Two SPECT prototypes are being developed, one dedicated to pre-clinical imaging (7 and 9.4 T), the second one dedicated to clinical imaging (3 T). The fundamental unit is a 5 cm x 5 cm gamma camera, based on the well-established Anger architecture with a continuous CsI:Tl scintillator readout by an array of Silicon PhotoMultipliers (SiPMs). In order to improve the noise performance of the overall SPECT system, the SiPM arrays need to operate at temperatures around 0 °C to 5 °C to reduce their Dark Count Rate (DCR). Thereby, the detection modules require mild to moderate cooling essential to achieve the final goals of energy and spatial resolutions. However, the presence of static magnetic field B0, rapidly switching gradients and radio-frequency pulses in the bore of the MRI machine, make use of traditional cooling blocks with bulky metallic components unsuitable. In this work we describe the study of a MRI-compatible cooling unit, thermal simulations of their expected performance and the results of a test prototype.

ClearPEM is a PET scanner dedicated to breast imaging, developed in the frame of the Crystal Clear Collaboration. The two prototypes, currently installed at ICNAS (Coimbra, PT) and Ospedale San Gerardo (Monza, IT), both show very good performances in terms of spatial resolution (< 1.5 mm) and sensitivity (1-2 %). The high production costs, however, prevents a large diffusion of this scanner in hospitals. The Collaboration has started a complete redesign of the scanner, aiming to develop a detector that can reach the same level of performance, but with reduced complexity and costs. Improved timing performance would be also desirable, to achieve better image quality and reduce the dose rate exposure to the patient. The development is based on SiPMs, single-sided readout scheme, and 4-to-1 coupling between crystals and detectors, as opposite to the 1-to-1 coupling and double-sided readout (with APDs) currently implemented into the prototypes. Despite reducing the electronic channels by a factor 8, the aim is to reach the same levels of energy, spatial and DOI resolutions while improving the timing performance. The developed module couples a 8x8 scintillator matrix to a 4x4 MPPC array. Crystal array consists in 64 LYSO pixels, each 1.53x1.53x15 mm3, arranged in a 8x8 grid, separated by foils of 70 microns thick ESR reflector. The matrix is read out by a 4x4 MPPC array matching the scintillator matrix in such a way that 4 crystals are coupled to each SiPM. Different types of crystal surface treatments and several coupling configurations between scintillators and MPPC have been tested. In the best configuration, using different combinations of the information collected by the 16 detectors, photoelectric events can be discriminated, with an energy resolution around 20% FWHM, interaction position can be correctly sorted among the 64 crystals, with efficiencies in the 80%-90% range, and continuous DOI information can be obtained, with an average resolution of 3.4 mm sigma.

We have demonstrated the use of strip-line method and waveform sampling for reading 2x8 silicon photomultipliers (SiPMs) with two strip-lines, each connecting 8 SiPMs. In this multiplexed readout, 16 SiPMs are read by using 4 data acquisition (DAQ) channels. The resulting PET detectors are compact and the DAQ can be placed away from the SiPMs, making them favorable for developing MR and other inserts. Subject to count-rate limitation, potentially the design is also scalable as one can connect multiple strip-lines to increase the SiPM/DAQ channel ratio. In the present work, we develop new 4x8 strip-line boards for reading two Hamamatsu 4x4 MPPC arrays by using 4 strip-lines. This MPPC array is selected to address the issues with gain non-uniformity and inferior coincidence time resolution with our previous design. Also, the MPPC array has a 3.2 mm pixel pitch, which is smaller than the 5.2 mm pitch and equal to the resolution on the strip-line of our previous design. Hence, we also add design features to increase path length between adjacent pixels, making it significantly larger than their physical separation. Therefore, smaller pixel pitches can be resolved and requirement on the timing accuracy alleviated. Two coupling methods between the SiPMs and strip-line are also studied. The MPPC pixels are coupled 1:1 to LYSO crystals, the detector is exposed to a Cs-137 source, and the output pulses are digitally acquired at 10 GSps by using an oscilloscope. For both coupling methods we obtain an energy resolution of 12.1-16.6% @511 keV and we can clearly separate adjacent MPPC pixels. Scalability is tested by connecting two strip-lines and all 16 pixels on the line are clearly resolved. We also obtain a coincidence time resolution of ~280 ps when the MPPC pixels are individually readout. We will apply the DRS4 for DAQ of the board and test the detector in a small-animal MR scanner. We will give a full report of the design and performance measurements at the conference.

Simultaneous PET/MRI shows great promise for improved detection and diagnosis of disease. Commercially available whole body combined PET/MRI systems are currently prohibitively expensive, limiting access to the technology. MRI-compatible PET inserts present a lower-cost alternative, adding PET capabilities to existing MRI installations. A brain sized radio-frequency (RF)-penetrable PET insert has been designed for simultaneous operation within MRI systems. This insert takes advantage of electro-optical coupling and battery power to electrically float the PET insert relative to the MRI ground, permitting RF signals to be transmitted and received through small gaps between the PET modules. Non-magnetic silicon photomultipliers are used in conjunction with a compressed sensing signal multiplexing scheme, and optical fibers transmit analog PET detector signals out of the MRI room for decoding, processing, and image reconstruction. Results of the first simultaneous PET/MR imaging studies with the full RF-penetrable PET ring are presented in this work. The PET insert was placed within a 3T whole body MRI system, and tomographic images of a 3D-printed resolution phantom with hot and cold rods of varying sizes were acquired both with only the B₀ field present, and under continuous pulsing from three MR imaging sequences: gradient echo (GRE), fast spin echo (FSE), and echo planar imaging (EPI). The resulting PET images have comparable contrast-to-noise ratios (CNR) under all pulsing conditions, with an average (of all rod types) maximum CNR difference of 14.0% among images. MR images of the resolution phantom were successfully acquired through the RF-penetrable PET shielding using only the built in MR body coil, with mean CNR decreases of 19.5% and 29.8% for GRE and FSE, respectively. These results suggest that simultaneous imaging is possible with the RF-penetrable insert, and show promise for this technology as an alternative to costly integrated PET/MRI scanners.

PURPOSE Blood flow quantification during perctaneous coronary intervention enables better patient outcomes. Recent imaging advances allow quantitative 2D perfusion images and morphological information to be produced together. To achieve a linear relationship between the contrast concentration in the body and the 2D image signal intensity, scatter and beam hardening correction are required. But the correction procedures involve complicated preliminary calibration or cumbersome special phantoms. We propose a method for easy scatter and beam hardening correction using only angiography images without phantom calibration. METHOD Based on a theoretical x-ray absorption model, our method generates the correction function using the catheter intensity measured in angiography images and the known contrast concentration and catheter diameter. The only unknown parameters in the model are the scatter ratio and body thickness. In typical clinical situations, the body thickness can be limited to 10-30 cm, allowing generation of a function for correction of signal intensity in angiography images. For verification, 370 mg/ml contrast agent in a 2-mm-diameter silicone tube was imaged, and the image signal intensity was measured to generate the correction function, which was used to correct the signal intensity of 15 tubes of different concentrations and sizes. The estimated values were compared with the actual concentrations for 6 sets of geometrical and x-ray conditions. RESULTS The mean error between the actual and estimated concentrations was 6%. The linear regression line slope was 1.032 and correlation coefficient was R=0.997. A color-coded 2D perfusion image was successfully generated after this correction. CONCLUSION The proposed method was confirmed to correct scatter and beam hardening without a phantom and to achieve a good linear relationship between the contrast concentration in the body and the 2D image signal intensity.

During X-ray diagnosis, occasionally when different materials are measured with a traditional X-ray detector such as a computed radiography (CR) system, the same image densities (digital values) are obtained. This situation occurs, when the X-ray attenuation factors which are determined by thickness, linear attenuation factor and density, have the same value. The aim of this study is to develop a simple and effective substance identification method which can be applied to X-ray inspections by means of general X-ray equipment. In our method, we newly proposed a universal curve which can transform “normalized” linear attenuation coefficient for 25 keV to the atomic number Z. Initially, using a published database of the linear attenuation coefficients, we made the universal curve theoretically. Next, to investigate the method experimentally, X-ray spectra after penetrating four different substances (aluminum, bone, acrylic, and soft tissue) were measured by a single probe CdTe detector and they were analyzed by our method. Normalized linear attenuation coefficient for each substance was determined using our method. Then, the experimentally determined value was plotted against the theoretical universal curve, and the effective atomic number Z was derived. The determined atomic number for each substance was in good agreement with the database value. We concluded that our method worked well to identify low-Z substances (Z<13) such as living material with an accuracy of Z +/-1.

The use of Rb-82 positron emission tomography (PET) for myocardial perfusion imaging (MPI) has increased but the effects of time of flight (TOF) and point spread function (PSF) modeling on large cardiac patients have not been investigated in-depth. This study reports our preliminary work in investigating the effect of TOF and PSF modeling in PET image reconstruction for large cardiac patients using a custom-built large anthropomorphic cardiac-torso phantom.The phantom mimicking a large patient was filled with F-18-FDG to model Rb-82 PET; the target activity levels were based on Rb-82 PET patient studies. The phantom was imaged using a TOF PET-CT scanner (Siemens Biograph mCT). The images were reconstructed in 6 ways: Filtered Back- Projection (FBP), FBP with TOF, ordered-subsets expectation-maximization (OSEM) algorithm (4 iterations and 21 subsets) with: ordinary Poisson (OP), OP with TOF, PSF, and PSF with TOF. In each case, a 3-D Gaussian post-reconstruction smoothing filter with FWHM of 0 to 10 mm in increments of 2 mm was applied. The resultant reconstructed images were then reoriented using PMOD software to create short-axis slices of the left ventricle and to generate polar plots. The effect of different reconstruction schemes and Gaussian smoothing filter sizes was analyzed for recovery coefficient and RMS noise on the polar plot data. The modeling of PSF resulted in higher recovery coefficient at the expense of increased RMS noise and the application TOF reduced image noise, while the combined use of PSF and TOF yielded better recovery coefficient values with moderate RMS noise. Further studies will investigate the effect of TOF and PSF on different phantom sizes and large clinical patients.

Objective: We studied the impact of using a constant, uniform attenuation coefficient to perform attenuation correction of breast PET data for heterogeneously dense breasts. Methods: We used breast phantoms with uniform and heterogeneous tissue densities, as well as with uniform and heterogeneous tracer uptakes in Monte Carlo simulations (SimSET). We inserted hot spheres (5-40mm diam, with 2:1 and 8:1 contrast) to represent lesions. The scanner, PET/X, was a rectangular dedicated breast PET scanner (20cm x 15cm x 8 cm). Attenuation correction was performed using both constant and matched heterogeneous attenuation coefficients. Photons scattered in the object and detectors were ignored. Image reconstruction was done with a fully 3D MLEM algorithm. Results: For the uniform density and activity breast phantom (using a constant attenuation coefficient), the relative contrast recovery (RCR) converged to 97.3% and 96.4% in a 40mm sphere for 8:1 and 2:1 contrast. For uniform activity images with heterogeneous density, reconstructed with uniform or matched attenuation correction, we found negligible differences (<1%) in RCR and RMSE for 10mm spheres with 8:1 contrast. Partial volume effects reduced the RCR in the 10mm sphere (~88%), with some variation with lesion location. For the breast phantom with heterogeneous activity and density, the RCR differed by up to 2% when uniform or matched attenuation coefficients were used. Conclusions: Our simulation results indicate that using a uniform attenuation coefficient for PET attenuation correction in a heterogeneously dense breast causes negligible errors (<1%) for uniform activity distribution, while differences of 2% were seen in RCR when the activity distribution was also heterogeneous.

PET quantitation is limited by the accuracy of the CT-derived attenuation correction map. In the lung, respiration leads to both positional and density mismatches, causing PET quantitation errors at lung borders but also within the whole lung. The aim of this work is to determine the extent of errors on the measured time activity curves (TACs) and the associated kinetic parameter estimates. 5 patients with idiopathic pulmonary fibrosis underwent dynamic 18F-FDG PET and cine-CT imaging as part of an ongoing study. The cine-CT was amplitude gated using PCA techniques to produce end expiration (EXP), inspiration (INS) and mid-breathing cycle (MID) gates representative of a short clinical CT acquisition. The ungated PET data were reconstructed with each CT gate and the TACs and kinetic parameters compared. Patient representative XCAT simulations with varying lung density, both with and without motion, were also produced to represent the above study allowing comparison of true to measured results. In all cases, the obtained PET TACs differed with each CT gate. For ROIs internal to the lung, the effect was dominated by changes in density, as opposed to motion. The errors in the TACs varied with time providing evidence that errors due to attenuation mismatch depend on activity distribution. In the simulations, kinetic parameters were over and underestimated by a factor of 2 in the INS and EXP gates respectively. For the patients, the maximum variation in kinetic parameters was 20%. Whole lung density changes during the respiratory cycle have a significant impact on PET quantitation. This is especially true of the kinetic parameter estimates as the extent of the error is dependent on tracer distribution which varies with time. It is therefore vital to use matched PET/CT for attenuation correction.

Siemens recently introduced xSPECT Quant (xQ) and xSPECT Bone (xB), both enabling quantitative SPECT for Tc-99m. In this work, we evaluate the quantitative accuracy of xSPECT. We acquired a Hot-Cold sphere phantom with a clinical representative Symbia Intevo test system in our lab and calibrated using a 3% NIST traceable Co-57 source (Calibrated Sensitivity Source (CSS)). Projections are acquired in step-and-shoot mode with 120 views over 360 degrees, and are also subsampled to 60 views (6-degree angular step). We used three reconstruction methods: xQ, xB and xEMAS (xE) at varying updates. xE is based upon OSEM but uses the projection operator employed in xSPECT, and is currently not commercially available. Mean activity concentration (in kBq/ml) is measured for the background and the 6 hot spheres. For the background, a large uniform VOI is chosen. We chose a reduced sphere VOI with 80% of the physical diameter of each sphere for xQ and xE, however we use the interior sphere volume as VOI for xB, as the boundary delinaeation stems from CT and is used in the reconstruction. For the background, all methods give estimations within 1% of error margin compared to truth. For spheres, xE and xQ have almost overlapping curves. Their accuracy gradually improves with the increasing size of spheres, and reach over 90% accuracy for the largest sphere (16ml). xB gives much higher recovery (>60%) even for the smallest sphere (0.5ml), and rises close to 100% for the largest sphere (16ml). This suggests that xB not only results in considerably higher resolution, but can also yield more accurate quantitation estimation. We did not find significant difference on quantitation between 3-degree and 6-degree sampling.

For PET imaging, CT scans are often used for PET attenuation correction and can be acquired at greatly reduced CT radiation dose levels. Techniques as low tube voltage/current have been used to obtain adequate attenuation maps in medium size patients. These techniques usually employ a smooth filter before backprojection to reduce CT image noise which can introduce bias in the conversion from HU to attenuation values. Due to the heaviness of the smooth filter and advancement of CT reconstruction algorithm, the CT dose can be further reduced while providing the same attenuation estimation. In this work, we propose an ultra-low dose CT technique for PET attenuation correction based on sparse-view acquisition. That is, instead of an acquisition of full amount of views, only a fraction of views are acquired. We tested this technique on a 256-slice GE Revolution CT scanner using a whole-body anthropomorphic phantom. An FBP reconstruction with Q.AC on 492 views (10 mA, 120kV, 0.5s) and an FBP reconstruction with standard filter on 984 views (150 mA, 120 kV, 0.5s) produced similar attenuation uniformity. We also simulated sparse-view acquisition by skipping views in an interleaved manner from the fully-acquired data. FBP reconstruction with Q.AC filter on simulated 246 views (20mA, 120 kV, 0.5s) looks similarly to the reconstruction on acquired 984 views (20 mA, 120 kV, 0.5s), showing a further potential for dose reduction compared to the full acquisition. With the proposed sparse-view method, this work can bring at least 2x more CT dose reduction to the current Ultra-Low Dose (ULD) PET/CT protocol.

X-ray fluorescence computed tomography (XFCT) can perform elemental imaging of an object, but its results often suffer from attenuation due to the low-energy nature of X-ray fluorescence. Here we propose an analytical reconstruction method based on solving a partial differential equation precisely describing a forward-scattering XFCT system. This method is able to carry out fast quantitative reconstruction with arbitrary attenuation correction for both source and fluorescence simultaneously. It incorporates the prior knowledge of the attenuation of scanned object, thus demands an extra scan. However, this can be avoided if the attenuation of source is similar to that of fluorescence or the attenuation is not significant. Monte Carlo simulation results show that this method is capable of providing more accurate quantitative results than traditional algorithms such as filtered backprojection or some simple attenuation compensation techniques.

In this work we proposed a new respiratory and cardiac motion correction method for cardiac PET/MR imaging by using 2D-MRI image navigator technique and novel dynamic MRI reconstruction method. In the PET/MR acquisition, ECG trigger is used. After trigger starts, 2D-MRI image navigator and golden angle cardiac imaging sequence are applied. After acquisition, cardiac MRI data and PET data are binned to different motion phases according to 2D-MRI image navigator and ECG trigger. We use the high down sampling acquisition and novel reconstruction method based on Low-rank and Sparsity decomposition to reduce MRI acquisition time and improve the time resolution. Finally, we estimate the motion field with both cardiac and respiratory motion based on the reconstructed MRI image series. And PET images are reconstructed by OSEM algorithm based on the binned data. After reconstruction, the PET images are deformed to the same motion phase by accurate motion field from MRI. Finally, sum up all the deformed images to get corrected PET image. PET data is simulated via analytical projection method using segmented MRI images. The results of the proposed method are compared to methods with only respiratory and without motion correction. The simulation shows that the proposed method could improve the image quality and the accuracy of lesion estimation especially on the ventricular wall with cardiac motion. In conclusion, the simulation result demonstrated the feasibility and improvement of the new method.

For medical applications such as cell tracking and angiography the detection of low concentrations of superparamagnetic iron oxide nanoparticles (SPIOs) is desired. Magnetic particle imaging (MPI) is a new technology which enables quantitatively and time-resolved the localization of a SPIO distribution. In contrast to magnetic resonance imaging (MRI) the SPIOs generate a positive contrast in MPI and a direct quantification of the amount of SPIOs is possible. The minimum concentration for which SPIOs can be reconstructed by MPI with a suitable quality is still an open question. In this work we investigate the background signal that is present in the measured signal even when no nanoparticles are located in the scanner tube. We show that a background subtraction in combination with a frequency cutoff for the dynamic part of the background signal lowers the detection limit for SPIOs in MPI up to a factor of ten. In-vivo mouse experiments show that for early time points when the tracer enters the vena cava a reconstructed image with sufficient quality can only be obtained when a background subtraction is performed. A fusion of the MPI image with data obtained by MRI shows the anatomical distribution of the SPIOs in the blood stream of the living mouse. Accordingly, structures with a lower concentration of SPIOs are visible in reconstructed images when no background correction is performed.

We present an image-domain point spread function (PSF) modeling approach for resolution recovery where spatially varying PSF kernel widths are adjusted based on data quality. This approach attempts to maximize contrast recovery while minimizing edge artifacts (ringing) associated with PSF modeling. We choose broader PSF kernels for noisier datasets where the extent of ringing is comparable to noise standard deviation levels and therefore result in minimal visible ringing artifacts. Similarly, we choose narrower, undermodeled PSF kernels for high count datasets to avoid edge artifacts associated with broader kernels which would have been visible at low image noise levels. We quantify ringing by measuring the difference between the highest overshoot and lowest undershoot levels around the mean background activity level. We define “visible ringing” by the difference between ringing and twice the background standard deviation and use it as our primary metric for measuring artifact levels. We show through simulations that broader PSF kernels can be used for noisier datasets with minimal visible ringing for improved contrast recovery. We use these results to determine the broadest PSF kernel widths to be used at each count level to achieve the highest level of contrast recovery with minimal visible ringing.

In a low contrast lesion detection experiment, large amounts of repeated measurements on the same object are typically required in order to achieve a good estimation of statistics. One way to obtain these many realizations can be through repeatedly scanning the low contrast object, which however, is a challenge with Positron Emission Tomography (PET) imaging due to the decaying nature of radionuclides using radiopharmaceuticals like Fluorodeoxyglucose (FDG). In order to calculate lesion detectability from a limited number PET FDG scans, the bootstrap resampling method was used to create n realizations. That is, from the same sinogram data, we randomly draw n bootstrap samples of size k with replacement for prompt and random coincidence events. To validate the proposed method, 60 parallel scans (3 min each) of a NEMA phantom filled with FDG were obtained by using listmode data from a gated acquisition on a PET/CT scanner. We resampled the sinogram of full data into 60 replicates with each having the same total number of prompts and random events as the replayed scan data. PET images were reconstructed from both scan data and bootstrap data using Time-of-Flight (TOF) and non-TOF Ordered-Subset Expectation Maximization (OSEM) algorithms. A 10-channel dense Difference-of-Gaussian (DoG) Channelized-Hotelling Observer (CHO) was developed with internal noise added to channel outputs and applied to the PET images. CHO showed close detectability on the 4 hot spheres (10mm, 13mm, 17mm and 22mm) of the NEMA phantom between scan data and bootstrap data, on both TOF and non-TOF PET images. We applied the bootstrap method to another scan of NEMA phantom with fixed contrast 4:1 and smaller spheres (5mm, 6mm, 8mm and 10mm), and created 150 realizations (10 min each). We conclude the bootstrap method is useful for low contrast lesion detection study using CHO.

In a functional PET/MR study, it is difficult to get good temporal resolution of activity distribution from PET images because of the need to image for a certain length of time to get sufficient count statistics (image SNR). Time-of-flight (TOF) reconstruction can be used to increase PET images SNR and therefore increase the temporal resolution. Five patients were injected with 410±80 MBq of FDG and scanned 140±30 minutes post-injection on a simultaneous TOF-enabled PET/MR scanner. TOF reconstruction shows faster convergence while it achieves a SNR improvement of 5-45% compared to non-TOF reconstruction. In another experiment Hoffman brain phantom was injected with 92 MBq of FDG and scanned for 60 minutes. TOF reconstruction shows faster convergence and additional 6-8% SNR improvement in gray matter. With this additional SNR gain, frame durations as short as 30s are possible while preserving reasonable image quality. This in turn effectively increases the temporal resolution of dynamic brain studies using simultaneous PET/MR imaging.

Clinical objective evaluation of methods for quantitative nuclear imaging is difficult since the true value of the quantitative parameters, also referred to as the gold-standard value, is rarely available in patient studies. Previously developed techniques to evaluate quantitative methods in the absence of the gold standard assume linear relationship between the true and estimated quantitative values and unimodal distribution of true values. These assumptions are restrictive and often not accurate. To overcome these limitations and widen the applicability of no-gold-standard (NGS) evaluation, we propose a NGS technique that models non-linear relationships between the true and estimated values and accounts for possible multimodal distribution of the true values. To estimate the parameters of the non-linear relationship and the distribution of true values, we designed an expectation-maximization (EM) algorithm. From the estimated parameters, we can compute two figures of merit (FoMs) that can rank the methods on the basis of accuracy and precision. The technique was validated using realistic simulated single-photon emission-computed tomography (SPECT) image data from an object database consisting of five phantom anatomies with all possible combinations of five sets of organ uptakes, where each anatomy included eight different organ volumes of interest (VOIs). The projection data were reconstructed using three different compensation methods. The NGS technique evaluated the performance of these three methods on the task of accurately estimating the activity concentration. Results indicated that the proposed NGS method provided accurate ranking of the reconstruction methods for multiple noise realizations using both the FoMs. Further, the proposed NGS method yielded improved performance compared to the existing NGS techniques. The generality of this method enables the use of patient data in the evaluation of a variety of quantitative imaging methods.

Contrast enhanced digital mammography (CEDM) has been generally used for dual energy subtraction images with iodine contrast agent to enhance the contrast between breast lesions and glandular tissues. However, the radiation dose remains a significant problem due to the need for two exposures. To overcome these drawbacks, we have proposed the contrast enhanced spectral mammography system based on photon counting detector. This detector allows for the energy resolved imaging with single exposure. In this study, the spectral mammography system based on CZT photon counting detector was simulated using realistic 3D XCAT breast phantom. The XCAT breast phantom obtained from human subject data consists of a realistic skin, adipose tissue, and three classes of fibroglandular tissue according to density. The breast images of photon counting detector were acquired from single exposure and separated two energy bins; one below and the other above K-edge absorption energy of Iodine (33.2 keV). For the comparison, amorphous selenium-based flat panel detector was used as a conventional detector. To evaluate lesion detectability, the breast phantom was simulated with the calcifications embedded in the mid-depth. As a result, the measured CNR of CZT-based spectral mammography was higher than that of the conventional mammography system. The images of amorphous selenium detector were higher noise than CZT detector. The dual energy subtracted images based on photon counting detector show the iodine uptake in the lesion while maintaining the image quality. The photon counting detector with energy discrimination capabilities can be possible to subtract the breast tissue in high energy images for contrast enhancement using single exposure. This study demonstrated the feasibility of photon counting detector in contrast enhanced mammography system. Compared with the conventional mammography system, the photon counting detector can reduce exposure dose while maintaining the image quality.

Digital breast tomosynthesis (DBT) is most commonly used in three-dimensional (3D) mammography because it provides a 3D view, so suspected tumors and masses in the breast can be detected with a higher degree of accuracy. Conventional DBT reconstruction methods are based on the filtered-backprojection (FBP) with an additional deblurring filter. However, this approach usually requires dense projection data with low noise levels for acceptable reconstruction quality. In this work, instead, we investigated a state-of-the-art image reconstruction based on the compressed-sensing (CS) theory for potential application to accurate, low-dose DBT. We implemented a CS-based algorithm as well as a FBP-based one for DBT reconstruction and performed a systematic experiment to verify the usefulness of the algorithm by comparing its reconstruction quality to the FBP-based one. In the experiment, a tomographic angle of ? = 20^o and an angle step of ?? = 2^o were used and the detector remained stationary during the projection data acquisition. We successfully obtained DBT images of substantially very high accuracy by using the CS-based algorithm. In addition, we synthesized a 2D breast image from the FBP-reconstructed DBT images which showed heightened details retained from DBT images, indicating superior performance compared to traditional 2D breast image alone. More details of the experimental results will be described in the paper.

The goal of the work was to create inexpensive, task-configurable fillable phantom features for radionuclide imaging. The primary design goals included simplicity of filling, feature size/shape/contrast control and having a realistic nature for the features and background. 3D-printed exoskeleton features of regular polyhedrons have been constructed and tested, with the idea derived from a soccer-ball like structure with sticks at the seams. Placing these structures into a container filled with solid spheres of a size that are below the resolution of the imaging system but that are excluded from the inside of the polyhedral features allows full-concentration radionuclide solution to leak into the interior of the printed features while the solution around the solid spheres makes for a reduced background concentration. The design thus creates a virtually wall-less contrast feature. A prototype phantom with 24 features of various sizes was created and imaged in PET/CT. A design for a full-scale detectability phantom is described and is planned for imaging.

Most PET scanners are designed to work well over a wide dynamic range of patient conditions. Clinical constraints, however, often impact the realization of the optimal capability for a particular scanner design. For whole-body FDG PET/CT, the operating constraints are most often dictated by injected dose (=370MBq or as low as possible) and imaging time (=30 minutes or as low as possible). Given the acquired data, image reconstruction settings are usually chosen to push the limit of acceptable image noise yet maintain detectability, quantitation and measurement accuracy. This work focuses on PET system capability to recover accurate feature profiles under identical constrained scan conditions (same detected count levels), by comparing the performance of two PET detector sampling designs. This quantitative measurement can be used in conjunction with other system design metrics (detectability, quantitation accuracy) to help in both optimizations of system design as well as optimization of imaging protocols for a given design. Methods: A hot rods phantom was imaged in two PET/CT systems, one with a 10x10 [x,z] and one with a 9x6 detector block (both blocks with the same form factor). PET transaxial images with 40M prompts and identical reconstruction settings were generated. 1cm thick PET images were registered to an image template using the mechanical rod positions. Profiles through the mask image center of mass per feature were measured and compared. The sum-squared difference (SSD) between the average mask profile (per rod size) and the measured data was calculated. Results: At this count level, the scanner with finer sampling had consistently higher SSD, which is consistent with the need for higher count levels necessary to take advantage of the finer sampling. Further measurements will compare axial SSD with the phantom rotated, use of cold rods (aligned and rotated) and the use of both rod types inside a body-size phantom.

The use of time-of-flight (TOF) information in positron emission tomography (PET) can significantly improve image quality. Many analytical expressions have been proposed to predict the variance of PET images. However, these methods become impractically time-consuming for TOF-PET. Our work is motivated by the goal of deriving a fast algorithm that can accurately predict image variance of TOF-PET. We first present an elegant expression for the improvement of image variance achieved by post-smoothed maximum likelihood expectation maximization (MLEM). Then we apply this expression to the variance prediction for TOF-PET. The proposed algorithm was validated by Monte Carlo simulation. Our results come close to achieving the goal, in that we can predict the image variance of TOF-PET for both uniform disk and realistic phantoms accurately with a significant reduction of computation time.

The current standard for measuring the spatial resolution performance of positron emission tomography (PET) systems is defined by the NEMA Standards Publication NU 2-2007, wherein a point source suspended in air and filtered back projection (FBP) are used to generate the image used for spatial resolution measurements. Clinical settings, in contrast, routinely use iterative algorithms, and lesions would be surrounded by other tissue that attenuates and scatters radiation, and would have radiopharmaceutical uptake of its own. We compared the results of measuring spatial resolution in both FBP and iterative images, by means of NU 2-2007 methodology, by fitting of Gaussian curves, and by an alternative method using a series of images with increasing smoothing Gaussian filters from 5.0 to 14.0 mm combined with a model to extrapolate the intrinsic spatial resolution of the scanner. A second experiment was also performed using the 10 mm sphere of the image quality phantom described in NU 2-2007, and using the abovementioned alternative method with modification to evaluate spatial resolution. For a point source, the NU 2-2007 method gave numbers of 4.04 and 4.48 mm in the X and Y directions respectively. Other approaches to measuring the spatial resolution typically yielded lower resolution measurements. The closest results were obtained with the multiple filter approach, which gave 3.79 mm in the X and 3.78 mm in the Y directions. When considering the 10 mm sphere in water, results were 3.58 mm in the X and 3.21 mm in the Y. Differences might be attributed to difficulties in estimating the equivalent width of the sphere.

Respiratory and cardiac motion during PET lead to unwanted blurring effects in reconstructed images. The anatomical information provided by MRI in combined PET/MRI scanners can be used to overcome this limitation. In this work we present a novel approach for creating precise motion models as basis for respiratory motion correction. The model is based on the combination of high-resolution sagittal and coronal image slices that cover the torso. Besides their acquisition we obtain data on a single diagonally oriented slice that intersects with all other slices. On the one hand, this allows us to make the connection between model and actual scan by deriving a navigator signal with high temporal resolution and comparing it with a global, PET data-driven respiratory signal. On the other hand, we can utilize the pixel intensity profiles along the intersection lines to identify and assign sagittal and coronal images to different breathing phases. Gaps due to spacing between slices are filled by optical flow interpolation. By combining the data we estimate the motion between breathing phases based on volumes with high information content. We chose the model approach to bypass the necessity to obtain 3D volumes in almost real-time and to leave room for additional acquisition protocols.

We describe an evaluation of the accuracy of the average µ-value within patients from MR derived µ-maps. Recently, a more optimal initial attenuation image or µ-map estimate in TOF-MLAA reconstruction has been proposed for hybrid PET/MR imaging. The proposed initial µ-map estimate is an image of the object filled with the average µ-value, and the proposed Initial Average Mu-value approach is referred to as IAM-TOF-MLAA. The average µ-value within the object is prior information which can be extracted from MR and patient database. In this work, the accuracy of the average µ-values within patients estimated from simulated MR µ-maps was evaluated using CT µ-maps as references. Clinically acquired human CT µ-maps (10 dedicated brain and 10 whole body scans) were randomly selected and used as the gold-standard. Dixon and non-Dixon MR µ-maps were simulated by replacing bone (and fat) with water as well as assigning a single µ-value for the lungs in the CT µ-maps. The average µ-value was calculated within the head, chest, and abdomen regions from the reference CT, Dixon MR, non-Dixon MR µ-maps for all patients. The error in the average µ-value obtained from the simulated MR µ-maps was evaluated, and correction factors which account for bone, fat, and uniform lungs were derived based on patient data to improve the accuracy of the estimated average µ-value. It was observed that the average µ-value within humans can be accurate within 10% from MR derived µ-map without any correction and within 5% with corrections for bone, fat and uniform lungs. It is also possible to further improve the accuracy of the average µ-value by applying the corrections derived from patients with a more specific range of body mass index and a more sufficient population size.

Among numerous scatter correction (SC) methods of cone-beam computed tomography (CBCT), the use of lead-strip blockers is low-cost and easy to implement and holds potential to significantly lower patient radiation dose. In blocker-based SC methods, the signal detected in the blocked region is deemed scatter through an ideal projection assumption and used to estimate scatter in the unblocked region. However, since the signal in the blocked region is not pure scatter, meticulous adjustment of working parameters has to be done to avoid over- and under-correction of scatter. In this work, we propose to model blocker-based CBCT projections as ideal projections convolved by a point spread function and to use a blind deconvolution method to recover true scatter in blocked regions. Combined with blockers’ motion and compressed sensing reconstruction, the proposed method can yield better image quality and more accurate CT numbers as demonstrated by physical phantom data.

The variation of gray levels across slices (named as DC shifting) is observed at outer slices in CT images acquired from circular cone-beam CT (CBCT) scan. The major reason is the approximation characteristic of analytical reconstruction algorithm for circular CBCT imaging. To eliminate the gray level variation in circular CBCT scan, we propose a practical image-domain correction scheme. Therein, histogram-based clustering is applied to achieve an adaptive image segmentation and related parameter extraction. The parameters extracted from the reference and corrected images are subtracted from each other to evaluate the DC shifting for several discrete gray levels. Spline-based interpolating is applied to estimate the DC shifting for the whole-range gray levels from these discrete ones. Forbild head mathematical phantom is used to evaluate the proposed algorithm. Preliminary results show that even a DC shifting less than 8 HU can be corrected effectively. This method provides an effective solution to suppress the gray level variation and is promising for advanced clinical applications.

Background and Purpose: We have been developing a practical and reliable calibration and evaluation method based on the use of traceable point-like sources. In this report, a novel traceable Ge-68/Ga-68 point-like source with a spherical acrylic absorber is proposed as a useful tool for calibrating and evaluating PET scanners. Materials and Method: The traceable point-like Ge-68/Ga-68 source consists of a spherical acrylic part and an incorporated small radioactive bead. The spherical design is required to keep the angular distribution of the emitted annihilation photons uniform. Physical characteristics of the source were evaluated by Monte Carlo simulations. For validation of its usefulness, the source was placed in the field of view of a clinical PET scanner. Circular region of interests (ROI’s) were defined in reconstructed images to obtain total ROI values, which were compared with the source radioactivity to determine the overall calibration factor (CF). Results and Conclusion: On the basis of Monte Carlo simulations, assuming possible realistic geometrical uncertainty, the uniformity of the emitted annihilation photons was better than 0.2%; the probability of positron escape was less than 0.01%. The calibration factors determined with the point-like source were reasonably consistent with those determined by the standard cross-calibration method, considering the uncertainty of the standard method. When this point-like source is used with a specially designed acrylic phantom, being practically free from ? ray background, it offers a unique way to evaluate quantitative aspects of PET images as a response to a traceable point-like source in a water-equivalent medium. In addition, the uncertainty in scatter and attenuation correction and that in the calibration factor can be evaluated independently. In conclusion, the calibration and evaluation scheme based on the use of this traceable point-like source is useful for calibrating and evaluating PET scanners.

Photoacoustic imaging (PAI) is a fast-developing biomedical imaging technology suitable for in vivo imaging. Recently, the compressive sensing (CS) theory has been adopted in PAI so that the image can be accurately recovered from many fewer samples than required by Nyquist theory. During the iteration process decided by these CS based methods, sparse matrix-vector multiplication (SpMxV) kernel always plays a dominant role. However, the huge number of trivial operations with zero elements in the measurement matrices will surely bring about latency, while on the other hand, cause load imbalance. To improve the performance of the kernel, a lot of SpMxV partitioning algorithms have been proposed, but all failed to consider the sparsity pattern of matrices. In this paper, we proposed a greedy approach based heuristics to recursively merge the nonzeros of the row vectors in a sparse matrix. Different from the algorithm based on the nnz of rows, the row vectors were then no longer considered in their original order in the matrix. As for the 30 measurement matrices related to the 30 sampling points in the total variation based gradient descent (TV-GD) algorithm, our proposed algorithm yielded a highest mean density of partition of 98.28%, which was more than 16% compared with all the other three strategies widely adopted.

Previously we have proposed a method to quantify the measurement completeness for a helical-CT reconstruction after motion correction. For every voxel, the “Tuy value” was computed, as a measure of the degree to which the local Tuy condition is satisfied. This voxel-based Tuy map can be used to predict artifacts due to data insufficiency. We designed it to analyze sampling completeness of motion corrected helical CT, but it can be applied to any source-detector geometries and orbits. When the Tuy values are low everywhere, the sampling is complete and exact and stable reconstruction is predicted. However, when the Tuy values are low inside a particular region, it cannot be excluded that artefacts in that region could still occur, due to incomplete sampling of neighboring regions. To study this, we created a 3D interior problem such that inside the interior region the Tuy values are low (sufficient samplings), whereas they are high outside the region (incomplete samplings). We simulated two scenarios to generate the desired truncations, one by motion and one by reducing the number of transaxial detector elements. The results showed that in both cases, the reconstruction of the interior region does not have a unique solution. Thus, even if the Tuy values are low in a particular region, exact reconstruction may not be possible due to incomplete sampling in neighboring regions.

An accurate quantification of the images in positron emission tomography (PET) requires knowing the actual sensitivity at each voxel, which represents the probability that a positron emitted in that voxel is finally detected as a coincidence of two gamma rays in a pair of detectors in the PET scanner. This sensitivity depends on the characteristics of the acquisition, as it is affected by the attenuation of the annihilation gamma rays in the body, and possible variations of the sensitivity of the scanner detectors. In this work we propose a new approach to handle time-of-flight (TOF) PET data, which allows performing either or both, a self-attenuation correction, and self-normalization correction. Our method is based on a fully Bayesian statistical model of complete data, and it provides different and complementary insights in the problem with respect to previously proposed methods based on a maximum-Poisson likelihood theory. We performed an initial evaluation of the theory and algorithms proposed in this work using numerical 2D simulations with different TOF resolutions and total number of detected coincidences. With moderate number of counts (500k counts per slice) and TOF resolution of 400 ps, the quality of the images obtained without knowing the attenuation and normalization factors were similar to the ones obtained when all the sensitivity factors were known. Therefore, this method can be very useful in cases in which standard attenuation and/or normalization corrections are not available.

In nuclear medicine motion-corrected image recon- struction approaches receive more and more attention. With new hybrid scanners like PET/MR the combination of multiple sources for motion information are possible. In current practice, separate gated data is reconstructed. Single gated reconstructions have inferior quality, in which the motion effect is negligible. The low quality is caused by bad signal-to-noise ratio in gated PET data. The motion in these gated images, in comparison to one reference gate, can be corrected successively and then obtained by overlaying. Disadvantages are artefacts caused by unclear convergence properties of the alternating iteration method of the reconstruction and the motion estimation. To solve this, we use a novel variational approach for motion estimation and image reconstruction of the PET data, which is based on a Bayesian model. With MR informations we receive more precise motion information, so that we can improve the PET motion estimation with the MR motion data. To verify this approach, the first experiments were done with the well known XCAT thorax phantom to receive evaluable results. In conclusion, the results in our approach shows a more reasonable tracer distribution and a better reconstruction of the edges of the heart. The outlook deals with the problem of attenuation correction in motion- corrected PET reconstruction, which has severe impact on the reconstructed images.

We describe an evaluation of effects of mis-matched boundary conditions caused by MR detection or PET emission being smaller than the attenuating medium in TOF-MLAA for PET/MR. Effects of different boundary conditions: (a) mis-matched boundary between the initial µ-map estimate and the true µ-map and (b) mis-matched boundary between the emission image and the true µ-map were investigated using 2D and 3D GATE simulations. It was observed that the estimated µ-value decreases as the size of the initial µ-map decreases and vice versa. Correct µ-values were recovered when the boundary of the initial µ-map estimate matched with that of the true µ-map. These trends were observed independent of the size of the emission. When the emission was smaller than the attenuating medium, edge overshoot was observed in the estimated µ-map independent of the size of the initial µ-map. On the other hand, the edge overshoot was barely visible when the emission covered the whole attenuating medium. Furthermore, the edge overshoot was observed to be more severe in the axial direction as compared to the radial direction. The edge overshoot was not very apparent in the emission image since the overshoot in the µ-map was located mainly outside the emission. However, the intensity or magnitude of the emission image was overestimated due to the overestimated attenuation correction factors for the lines-of-response which passed through the overshoot region. Other than having the emission covered the whole attenuating medium, it was observed that constraining the wall region in the estimated µ-map can also reduce the edge overshoot and improve the accuracy of the emission image. In summary, the size of the initial µ-map in TOF-MLAA affects the estimated µ-value. Furthermore, when the boundary of the emission is smaller than that of the attenuating medium, the estimated µ-map contains edge overshoot which causes overestimated emission image.

The purpose of our research is to develop a fast image reconstruction algorithm with a ray-driven method using a general-purpose computing on graphics processing units (GPGPU). The ray-driven method based on a projection bin uses sample points that are located on the center of the bin. In the implementation with a ray-driven method using a GPU, a collision of memory accesses sometimes reduces the performance of the calculation. To avoid the collision of memory accesses, two methods were used in our proposed algorithm: one was the grouping of sample points that were assigned to threads in the GPU, and the other was the calculation order (the order of access to memories, which corresponded to pixels in an image matrix). The performance of the proposed method was compared with an image reconstruction with a CPU and that with the GPU using an atomic function, which was prepared to avoid collision. The results of the simulations confirmed the feasibility of our proposed reconstruction algorithm in the applications of the filtered-backprojection method, ML-EM method and OS-EM method.

Direct 4D image reconstruction methods have been shown to generate parametric maps of improved precision and accuracy in dynamic PET imaging. However, due to the inability to construct a common single kinetic model for the entire FOV, the interleaving between tomographic and kinetic modelling steps causes errors in erroneously modelled regions to spatially propagate. Adaptive models have been used to mitigate the problem though they are complex and difficult to optimize. In this work, we demonstrate a new way to minimize errors, focusing on the tomographic step rather than the kinetic modelling step, limiting the actual propagation of bias by incorporating time-of-flight (TOF) information within a direct 4D reconstruction framework. Using ever improving TOF time resolutions (580ps, 440ps, 300ps and 160ps), we demonstrate that TOF direct 4D image reconstruction can substantially prevent error propagation from erroneous kinetic models with errors constrained ever so closely to the vicinity of erroneously modelled regions and bias reduction of up to 60% in well modelled regions compared to non-TOF direct 4D reconstruction. By combining such TOF direct 4D image reconstruction with adaptive models, further improvement could possibly be achieved in the future.

Normalization of all lines of response (LORs) or sinogram bins is necessary to avoid artifacts in fully 3D PET imaging. Component-based normalization (CBN) is an effective strategy to generate normalization correction factors (NCFs) from short time scans of known emission sources. In the CBN, the NCFs can be factorized into time-invariant and time-variant components. The effective crystal efficiencies are the main time-variant component, and a regular normalization scan is needed to update their values. However, the effective crystal efficiencies during an emission scan can be different to those estimated from the last normalization scan. Therefore, it would be advantageous to be able to estimate unique crystal efficiencies to account for this time-variant component. In this work, we present a self-normalization algorithm to estimate the crystal efficiencies directly from any emission acquisition. The algorithm is based on the principle that if the true image were known, the mismatch between its projections, corrected for the time-invariant NCFs, and the acquired data could be used to estimate the effective crystal efficiencies. We show that the algorithm successfully estimates the effective crystal efficiencies for simulated sinograms with different levels of Poisson noise and for different distributions of crystals efficiencies. This algorithm permits the reconstruction of good quality images without the need for an independent, separate, normalization scan. A key advantage of the method is the estimation of relatively few parameters (~10^4) compared to the number of NCFs for 3D data (~10^8).

Spectral CT plays an important role in medicine, security and other field because of its ability of material discrimination. The development of photon counting detectors brings better material discrimination with more separable spectrums. Since material decomposition is a noise- amplification process, the reconstructed image can be noisy and suffers from severe ring artifacts due to pixel inconsistency in photon counting detectors. Meanwhile, the images reconstructed directly from the polychromatic projections have much less artifacts and share the same structural information with the material images, which can be utilized to suppress the noises in the material images. An iterative reconstruction algorithm is proposed in this work, where a weighted non-local total-variation object function is designed. The weighting is inversely proportional to the difference between corresponding pixels in the reference images, which are reconstructed from polychromatic projections. The optimization program is solved via the ASD-POCS algorithm for the decomposed material projections. Experiments with a Pilatus 200K photon counting detector to discriminate Teflon and polyethylene are carried out to verify the algorithm. The results show that the proposed method is able to suppress ring artifacts and preserve the edges.

We describe an evaluation of a more optimal initial attenuation image or µ-map estimate in TOF-MLAA for PET/MR. Typically, the initial µ-map estimate used in TOF-MLAA is an image filled with the µ-value of water uniformly within the object, and additional calibration is required during or post reconstruction to produce a quantitative µ-map since TOF PET data determine the attenuation sinogram up to a constant offset. In this work, a more optimal initial µ-value was selected to fill the object such that the forward-projection of the initial µ-map is already very close to that of the reference µ-map thus reducing/minimizing the offset during the early iterations of TOF-MLAA. Consequently, the estimated µ-map is expected to reach the reference quickly and naturally without any calibration. A more optimal initial µ-value which can be practically obtained is the average µ-value within the object (prior information from MR and patient database). The Initial Average Mu (µ)-value approach is referred to as the IAM approach, and the IAM-TOF-MLAA was evaluated using 2D simulations. The performances of other initial µ-map estimates were also compared. It was observed that the estimated µ-map reached the reference more quickly and naturally for IAM-TOF-MLAA as compared to all other cases. Furthermore, starting with the correct ‘magnitude’ (i.e. average µ-value) was observed to be more important than starting with the correct spatial µ-distribution in terms of reaching the reference µ-map with a low number of TOF-MLAA iterations. In addition, the estimated µ-value from IAM-TOF-MLAA was still reasonably quantitative even with a 5% error in the IAM. However, overestimation bias in estimated µ-value was observed for noisy data sets. Nevertheless, noise reduction was demonstrated to be able to decrease the bias as expected. In conclusion, the proposed IAM-TOF-MLAA can produce quantitative µ-map without any calibration, and the accuracy of the average µ-value can be further improved.

In this work, we performed a feasibility study on the three-dimensional (3D) image reconstruction in a truncated Archimedean-spiral geometry with a long-rectangular detector for application to high-accurate, cost-effective dental x-ray imaging. Here an x-ray tube and a detector rotate together around the rotational axis several times and concurrently the detector moves horizontally in the detector coordinate at a constant speed to cover the whole imaging volume during the projection data acquisition. We established a table-top setup which mainly consists of an x-ray tube (60 kV_p, 5 mA), a CMOS-type detector (198-µm pixel size, 36.4 × 232.8 mm² active area), and a rotational stage for object mounting. We performed systematic experiment to demonstrate the viability of the proposed approach to volumetric dental x-ray imaging. For the image reconstruction, we employed a compressed-sensing (CS)-based algorithm, rather than a common filtered-backprojection (FBP) one, for more accurate reconstruction. We successfully reconstructed 3D images of considerably high quality and investigated the image characteristics in terms of the image profile, the contrast-to-noise ratio (CNR), and the spatial resolution.

Spectral x-ray computed tomography with photon counting detectors involves the estimation of basis material projections from energy-resolved measurements, which is also known as sinogram material decomposition. The statistically optimal image reconstruction requires knowledge of the full covariance matrix of the basis material projection estimates. The covariance matrix is block diagonal (assuming no crosstalk), with each block having dimensions equal to the number of basis materials. The maximum likelihood estimate is efficient, meaning it will attain the Cramer-Rao lower bound so that the covariance matrix is equal to the inverse of the Fisher information matrix. Using the Fisher information matrix, we derive an analytical formula for the covariance matrix of the estimates of the basis material projections from photon counting measurements. The derivation assumes Poisson statistics, which is valid if pulse pileup is negligible or the detector electronics possesses a pileup rejection mechanism. We compare the derived analytical formula to the covariance calculated using multiple realizations of maximum likelihood estimation. The detector response model in this study assumes perfect pileup rejection and a FWHM energy resolution of 9 keV. Several flux levels with two different sets of basis material lengths were simulated. We found good agreement (less than 4% relative error) between the formula and simulations.

We investigate the correlation between the clinical severity of neurodegenerative disease and texture metrics computed using PET images of the brain. Specifically, we explore how the parameters of feature computation - such as the region of interest definition method, and the direction used for texture quantification - affect the correlation between the Haralick features (and other texture metrics) and clinical disease. The advantage of such metrics is that they do not require kinetic modelling or the tracer input function. The analysis was based on an ongoing Parkinson's disease imaging study, with co-registered PET and MRI images, and tracer predominantly concentrated in the striatum. Disease duration was used as the primary clinical metric. We found that the region of interest placement method determined which texture metrics were correlated with disease duration. Using the (PET-defined) regions of high tracer uptake resulted in relatively low correlation values (Spearman's ? ˜ 0.6); on the other hand, using the MRI-defined anatomical regions resulted in higher correlation strength (? ˜ 0.9). When simple rectangular regions defined over the striatum were used, the correlation strength was comparable to that obtained with the MRI regions, although in this case different textural features were correlated with the disease (? ˜ 0.9). The direction chosen for feature computation only had a pronounced effect when rectangular regions were used; the anatomy-defined direction generally produced highest values of the correlation coefficient. These results suggest that the Haralick features and other texture metrics that do not require kinetic modeling could potentially be used for the analysis of PET images for which the corresponding MRI data is not available. The results also show that the region of interest definition method and the direction along which metrics are computed may strongly affect metric performance.

In 3D PET an axial compression is often applied to reduce the data size and the computation times during image reconstruction. This compression scheme can achieve good results in the centre of the FOV. However, there is a loss in the spatial resolution at off-centre positions and this effect is increased in scanners with a larger FOV. This is the case of the Siemens Biograph mMR, which by default uses an axial compression of span 11. An assessment of the improvement in the spatial resolution, that would be achieved in a reconstruction without axial compression, is necessary to evaluate if the additional computational burden is justified for routine image reconstruction. In this work, we present an implementation of the ordinary Poisson ordered subsets expectation maximization (OP-OSEM) algorithm without axial compression for the mMR, and evaluate its performance for span 1 and span 11. We show that an improvement of 3 mm FWHM (i.e. an improvement of 40%) can be achieved when span 11 compression is avoided and the source is at a distance greater than 100 mm from the centre of the FOV. In addition, the general image quality properties of the algorithm were evaluated with a NEMA image quality phantom acquisition and contrasted with its reconstruction via the STIR open source reconstruction software.

CT protocol design and quality control would benefit from automated tools to estimate the quality of generated CT images. These tools could be used to identify erroneous CT acquisitions or refine protocols to achieve certain signal to noise characteristics. This paper investigates blind estimation methods to determine global signal strength and noise levels in chest CT images. Methods: We propose novel performance metrics corresponding to the accuracy of noise and signal estimation. We implement and evaluate the noise estimation performance of six spatial- and frequency- based methods, derived from conventional image filtering algorithms. Algorithms were tested on patient data sets from whole-body repeat CT acquisitions performed with a higher and lower dose technique over the same scan region. Results: The proposed performance metrics can evaluate the relative tradeoff of filter parameters and noise estimation performance. The proposed automated methods tend to underestimate CT image noise at low-flux levels. Initial application of methodology suggests that anisotropic diffusion and Wavelet-transform based filters provide optimal estimates of noise. Furthermore, methodology does not provide accurate estimates of absolute noise levels, but can provide estimates of relative change and/or trends in noise levels.

Objectives: To establish an accurate and fast image segmentation method for extracting the tumor from the background of healthy tissue using the dedicated ¹⁸F-FDG MAMmography with Molecular Imaging (MAMMI) Positron Emission Tomography (PET) dynamic images of breast in vivo.
Methods: The dedicated breast MAMMI PET imager is a 12 detectors head ring with trapezoidal LYSO scintillators of 12 mm thickness, black painted, each coupled to a Hamamatsu H8500 PSPMT. With a total dose of 215-230 MBq FDG injected intravenously, ten 60-second frames and three 300-second frames after radiotracer injection were reconstructed for selective visualization of the breast lesions in a patient. Time activity curve (TAC) of each pixel in PET reconstructed images reflects the characteristics of ¹⁸F-FDG uptake rate. A feed-forward Artificial Neural Network (ANN) with error back-propagation (BP) algorithm were exploited to classify the TACs features of tissue pixels explicitly and then segment the tumor region in the MAMMI PET images.
Results: TAC curves of tumor region have higher value of FDG uptake with time and the curves have steep slopes. In contrast, for non-tumor regions a gentler positive slope and sometimes a downward trend is observed. After filtering and smoothing to improve the SNR, the tumor region are clearly distinguishable from the healthy tissue using ANN algorithm.
Conclusions: An innovative approach for image segmentation of breast cancer has been proposed in this work. Our main purpose was to demonstrate an automatic method for accurately and quickly diagnosing breast tumors. The initial pilot validation of the proposed method has shown promising results that the method can successfully segment patient lesion and identify the tumor region in 3D space based on segmentation obtained in different 2D slices.

Image-domain dual-energy CT (DECT) is convenient and critical for medical diagnosis and treatment evaluation. The difficulty of this work is that a good balance between signal-to-noise ratio (SNR) and spatial resolution is not readily achievable. The major reason is the noise-amplifying characteristic using direct matrix inversion for the dual-energy material decomposition. Noise suppression is thus included implicitly or explicitly, resulting in the visible degradation of spatial resolution. To solve this problem, we propose a novel image-domain material decomposition algorithm within multiscale framework for DECT. High- and low-energy CT images are separated into scale space to fully explore the flexibility in signal processing in each scale. The scale images are then decomposed into two material images using penalized weighted least-squares (PWLS) algorithm with adjustable regularization parameters in different scales. The final material images are generated by accumulating the decomposed images of all scales. We evaluate the proposed method using an anthropomorphic head phantom. Compared with single-scale PWLS decomposition, the proposed multiscale method increases the contrast by 19.4% and 17.4% and spatial resolution by 20% and 33% in bone and soft-tissue images, respectively. The multiscale DECT scheme is thus attractive for advanced clinical applications, including bone separation in CT angiography, iodine quantification and calculation of pseudo-monochromatic images and virtual nonenhanced images.

While SPECT is widely used for in vivo diagnostic imaging, it has much worse spatial resolution than other imaging modalities such as MRI and CT. Resolution recovery reconstruction incorporating collimator-detector response model is an advanced way to overcome this problem. However, additional low-pass filters are commonly applied to the resolution recovery reconstruction images to reduce the noise and enhance the convergence property, evoking the ironical situation of "blurring images after the resolution recovery". In this respect, the more reasonable approach to fully utilize the benefits of resolution recovery reconstruction would be the replacement of low-pass filters with edge preserving ones, such as non-local means (NLM) filter. In this study, we examined the usefulness of NLM filters in the resolution recovery reconstruction of SPECT data in bone SPECT/CT studies. The NLM filters reduced noise remarkably in bone SPECT images reconstructed with resolution recovery, without loss of edge information. By incorporating CT side information into NLM filters, we could achieve both the reduction of noise and the enhancement of anatomical information. Applying thresholding strategies provided additional improvement of the performance of NLM filters. The proposed method will facilitate more accurate qualitative and quantitative result of resolution recovery reconstruction of SPECT in bone SPECT/CT studies.

For the diagnosis of FDG-PET images, standardized uptake value (SUV) is one of the most significant criteria to diagnose the normal case or the abnormal case that requires further testing. But, it is not sufficient to diagnose only by SUV. FDG-PET image has the characteristics that the organs of which glucose metabolism is active are bright shade. On the other hand, the winding blood circulations such as colon have also bright shade because the FDG remains these areas. The former is the typical pattern of malignant, and these seem to be isolated shadow. And the latter is the normal shadow, and these areas have linear structure look like tube. It is important to understand the three dimensional organs shape for the cancer diagnosis. In this research, we propose the support system of interpretation of radiogram by offering the additional information based on FDG-PET images. The additional information means two attributes, one is the representation of structural information on FDG-PET images, and other is the three dimensional sequential view inside the human body. To represent the structural information, we calculate the curvatures of fourth dimensional hyper-surface of FDG- PET images. There are three curvatures, and each curvature can express the different structures such as the connectivity and the isolation degree. Using these curvatures, we extract the doubtful regions, and make quantitative evaluation of malignancy from the viewpoints of such as isolation degree, connectivity, thee dimensional shape, maximum SUV and SUV average. To the recognition of the spatial information inside the human body, we make the sequential view of human body based on MIP (Maximum Intensity Projection) and the perspective projection. From the sequential images, we can get the sense that we walk through inside the human body. We applied these methods for some cases, and we investigated the effectiveness of our methods for the cancer diagnosis.

In dynamic PET, to quantitatively analyze functional changes or estimate physiological parameters, ROIs are applied to the images to extract the underlying tissue time-activity curves (TACs). Manual delineation of ROIs is commonly used, but is time consuming and operator dependent. Also, most semi- or automated segmentation methods proposed are based on static images. However, with dynamic PET images available, the temporal information can also be employed to differentiate different tissues. Clustering analysis has been used for dynamic studies, in which voxel TACs are classified into a number of clusters. Clustering algorithms may depend on the Amplitude of the TACs (Method-A), or a similarity metric can be based on the Shape of the TACs (Method-S). To integrate spatial information with temporal information, in this study, we proposed a Spatial Continuity-Enhanced cluster algorithm (Method-SCE). We evaluated the performance of this proposed Method-SCE method, as well as Method-A and Method-S, using both simulated dynamic PET data and human data of beta cells in the pancreas with 18F FP-DTBZ. In both simulation and human data, Method-A successfully segmented pancreas and liver, but mis-assigned voxels around the edge of organs, due to respiratory motion and partial volume effect. Method-S failed to differentiate pancreas from background tissue and liver with similar TAC shapes, and assigned many non-pancreas voxels incorrectly. Method-SCE gave similar or slightly better performance than Method-A, which successfully differentiated pancreas from the background and liver. This approach once fully optimized has the potential to provide a more accurate segmentation, properly accounting for edge effects, by inclusion of the spatial continuity term.

We propose a spatially adaptive Non-Local Means (NLM) post-filtering approach for whole-body clinical PET imaging. Our approach is aimed at avoiding different effective smoothing strengths in different organs that result from with traditional, non-adaptive NLM. We vary the smoothing strength according to the intensity level around a given voxel such that regions with low absolute noise levels are smoothed less and those with high absolute noise levels are smoothed more. We evaluated this approach and compared it to alternative filtering techniques by inserting lesions of known size and activity into clinical datasets acquired on a Toshiba TOF-PET/CT Large Bore scanner. Images were reconstructed with list-mode OSEM with all physical corrections inside the system model prior to post-filtering. We qualitatively evaluated the techniques by comparing the differences between filtered and original images and quantitatively evaluated them using lesion contrast vs. background variability curves. Both evaluations showed that the proposed method could better accommodate varying noise statistics in whole-body PET images and better preserved lesion contrast across different regions.

A simple, generic and unified model for summing either step-and-shoot or continuous bed motion (CBM) dynamic sinogram time frames is formulated. The mathematical model and its preliminary validation with phantom study is presented. There are recent interests and research efforts in the potential merits of parametric image processing in clinical Positron Emission Tomography (PET) whole-body oncology. Doing so requires dynamic sequence data acquisition and image reconstruction. Currently, static standardized uptake value (SUV) imaging is the gold standard and mainstream clinical practice for diagnoses and reimbursement. To balance between patient volume, daily clinical workflow and continued dynamic imaging research requires 2 scans - a static acquisition and an additional dynamic sequence acquisition. This current effort focuses on quantitative summing of dynamic frames to recover the corresponding static data for SUV analysis. It may open up the possibility of doing one scan instead of two. In addition to dynamic whole-body imaging, another recent development in clinical whole-body PET is continuous bed motion (CBM). In light of the growing trends mentioned, a single unified model to quantitatively sum dynamic sinograms from either conventional step-and-shoot or CBM mode can reduce complexity, simplify effort and enhance future scalability

The noise reduction in x-ray computed tomography (XCT) is a critical technique for improving the diagnostic quality of images and saving the radiation dose of imaging. Here, we propose an image-domain multiscale decomposition and anisotropic diffusion based noise reduction method for clinical applications. As used in the projection domain before, a practical multiscale decomposition of CT image is carried out by using isotropic diffusion partial differential equation (PDE) in the image domain, followed by the image-domain anisotropic diffusion to reduce noise in each scales. The performance of the proposed method for noise reduction is experimentally evaluated and verified using the real CT image data. The preliminary result shows that, as compared to single-scale one, the proposed method increases the spatial resolution (full width at half maximum, FWHM) by 8.12%.

A study is described that investigates the capacity for mathematical observer models to mimic the performance of human observers in a PET lesion detection task. FDG-PET data from seventeen tuberculosis patients presenting diffuse hyper-metabolic lung lesions were selected for the study, to include a wide range of lesion sizes and contrasts. All subjects were scanned on a simultaneous PET/MR system (Siemens mMR) with one bed position over the lungs, after a 4.8±1.6 mCi injection and 60-minute uptake period. Various noise levels were simulated by randomly discarding events in the PET list mode according to 10 predefined fractions, from 5×10-1 to 5×10-4. Thirty-three lesions were selected in the 17 patients, as well as one background region in each. A lesion detection task including these 550 images ((33 lesions + 17 backgrounds) × (1 original image + 10 simulated noise levels)) was developed and presented to 5 experienced image viewers (2 nuclear medicine radiologists and 3 postdoctoral researchers). The observers’ responses classified each lesion into 1 of 2 classes, depending on whether it was detected or not-detected. The lesions were characterized by 4 parameters: lesion metabolic volume, lesion-to-background contrast, lesion SUV, and lesion-to-background SNR, and each was represented by its associated 4-element vector. The binary detectability data were used to train a linear observer model in the 4D space. Various fractions of the observers’ decisions were used for training the model, and the accuracy was evaluated for the model’s ability to predict the remaining decisions, i.e. those not used for training. This was performed randomly for 100 realizations at every training level. The findings show that good performance, in terms of matching human accuracy, was achieved with only 20% training. The model was robust, with similar trends for all human observers.

Current dosimetry for radionuclide therapy is based on calculations only, or on poor activity assessment from current emission tomography systems that have severe limitations for imaging and determining the dose of radiotherapeutics. Continuous-spectrum Emission Tomography (CET) is a new in vivo molecular imaging modality for early visualization of the absorbed dose delivered by in vivo beta decaying radionuclides. CET uses emission by Bremsstrahlung similarly to CT imaging, but X-rays are generated by the interaction of electrons from radiotherapeutic isotopes within the body of the patient rather than an X-ray source. These X-rays have a continuous spectrum ranging from lower energies up to 1-2 MeV (depending on the isotope), and will be collimated by narrow pinholes suitable for high energy. They need to be detected by high-density detectors similar to those used in PET. Lutetium-based scintillators that are widely used in current systems are not optimal because of intrinsic radioactivity from Lu-176, which results in a continuous background spectrum which cannot be filtered out as the system will be working in singles mode. BGO is a promising alternative for this application as it is a very low cost scintillator that offers high stopping power. The potential of monolithic BGO read out by digital photon counters (DPC, Philips) for CET was investigated in this work using Monte Carlo simulations and experiments. We modeled and experimentally characterized 32 x 32 x 30 mm³ BGO crystals, and showed that roughening the crystal surfaces increased the light output and improved the energy resolution at low energies. Light spread was not affected by the energy, rather by the depth-of-interaction. Our future work will focus on further studying the light spread function and the different factors that could optimize it (type of reflector, dual-ended readout of the scintillator, surface treatment).

The objective of this study is to validate the in-house GATE simulations of the new Philips Digital Vereos PET/CT scanner and evaluate their accuracy by comparing results with experimental data obtained according to the National Electrical Manufacturers Association (NEMA) NU2-2012 standards. Vereos is the first fully-digital PET system and is based on Philips Digital Photon Counter (DPC) sensors, using a crystal-to-sensor coupling of 1:1, with a total of 23040 channels, yielding a time-of-flight (TOF) resolution around 320 ps and an energy resolution of 11%. A detailed implementation of the geometrical and functional models of this scanner and the NEMA phantoms was conducted, allowing the evaluation of the simulated absolute sensitivity, spatial resolution, count rates and the image quality. All Monte Carlo data production was performed according to the NEMA protocols. Simulated data were converted into the Philips list-mode format, reconstructed and analyzed using the same software tools as in the quality control step of the production line. Absolute sensitivity results show a maximum deviation of 2% between the simulation results and the reference data. A good agreement was also obtained for the spatial resolution. Maximum differences of 500 microns for the FWHM and 570 microns for the FWTM were found. Preliminary work on the validation of the count rates and quantification of image quality has started already and will be presented at the conference as well.

In any domain, simulations are of utmost importance for assessing and predicting the performance of post-processing methods by controlling the truth and all model's parameters. In PET, two simulation methods exist: analytical and Monte Carlo. While not being as realistic as Monte Carlo ones, analytical simulations are fast enough for generating statistical replicates and realistic enough for assessing data-processing methods, especially in regard with noise. In this work, our aim was to produce replicated simulations of dynamic PET brain scans with realistic count rates properties. We extended the methodology of the analytical simulator ASIM to get (i) a different modeling of the partial volume effect (down-sampling and image-based PSF modeling), (ii) a projection model based on real crystal coordinates, (iii) a fan-sum based random distribution and (iv) an intrinsic modeling of the decay and count rates through the dynamic acquisition. Attenuation, normalization (from real crystal efficiencies), scatters (convolution-based) and randoms (fan-sum based) were taken into account. The Zubal brain phantom and the geometry of the Siemens Biograph Hirez 4-rings scanner were used. All other input parameters were extracted from real acquisitions of volunteers with the HRRT: frame durations, time-activity curves, scatter and random fractions and total number of prompts for the whole acquisition. Simulated count rates were compared to those from the acquisitions on a frame-by-frame basis, for three different tracers. Results showed good agreements and realistic reconstructed images. These data will be used for assessing reconstruction and post-processing methods in the context of dynamic PET brain scans.
S Stute, C Leroy, M Bottlaender, V Brulon and C Comtat are with UIMIV, Service Hospitalier Frédéric Joliot, Orsay, France. C Tauber is with UMR Inserm U930 - Université de Tours, Tours, France.

Circular pinholes are most commonly used for small-animal pinhole collimators in SPECT, but some systems have moved toward utilizing square or rectangular projections from square or lofted apertures in order to increase system sensitivity due to more efficient use of detector area. We are designing a dual-resolution system based on square-pinhole apertures for use in small-FOV SPECT applications, such as cardiac imaging. Aperture fabrication techniques cannot produce a perfect square, giving the square pinholes some amount of roundedness at the corners. This work investigates how this roundedness affects projection image quality in terms of packing fraction and spatial resolution. Five different pinhole full-acceptance angles (a = 10°, 20°, 30°, 40°, 50°) and 11 different roundedness values were simulated. Aperture size and opening angle were varied such that the FOV and the sensitivity were matched between the square pinhole and subsequent rounded pinholes. Using the sensitivity-matched and FOV-matched apertures, spatial resolution and hexagonal packing fractions were determined. Packing fractions started at 1 for square pinholes and decreased until reaching the packing fraction of a circle (0.907). For the 1-mm apertures studied, the full-width half-maximum (FWHM) widened as pinhole shape changed from square to circle, while full-width tenth-maximum (FWTM) narrowed. These results show that for packing fraction, the perfect square shape has the best packing fraction, while FWHM and FWTM resolution are best when the shape is a perfect square and perfect circle, respectively. In summary, for these apertures, the shape does affect projection-image quality, and preliminary results indicate that the true square may have some advantage.

Positron Emission Mammography (PEM) imaging systems with the ability in detection of millimeter-sized tumors were developed in recent years. And some of them have been well used in clinical applications. In consideration of biopsy application, a double plane detector configuration is more practical than a ring detector for the convenience of breast immobilization. In this study, a real-time imaging reconstruction solution is provided for imaging-guided intervention in a double plane designed PEM system. The biopsy under PEM is processed step by step. The time for each step could be 20 seconds for example. Each step perform once data acquisition and reconstruction. The real-time reconstruction is supported with GPU. The distance between the planes is changeable with different breast size. And the system matrix used in the reconstruction is real-time computing. For the reconstruction process, both projection and back-projection procedures use tube models. Each tube is distributed with a thread and calculated in parallel in GPU. Estimated image is bound to 3D texture memory. A double plane PET system is simulated above Geant4 Application for Emission Tomography (GATE) software based on MC methods. Each plane consists of 100 × 75 crystal LYSO arrays with a pixel size of 1.9 mm × 1.9 mm × 15 mm. Images are reconstructed with a pixel size of 0.5 mm × 0.5 mm × 1mm. Biopsy procedure is simulated with a point source and a rod source. The series of biopsy images are listed as follows. The point and rod source both could be imaged in the simulated biopsy process. With all the span tubes calculated, the GPU accelerated reconstruction time is 6.9 seconds while the CPU reconstruction time is 243 seconds.

A conventional LINAC had a single head and the head rotated around a patient as predefined treatment planning. However, the single LINAC system requires significant time cost to perform treatment and to reach the expected dose. In this study, a newly proposed LINAC which is employed two heads simultaneously was investigated. The dual head LINAC can contribute to reduce time for treatment and have possibility for various treatment plan. In order to design the dual head LINAC, the LINAC head was modeled using GATE as a preliminary study. The LINAC head was designed with VARIAN manufacturer's information according to VARIAN 2300EX. 6 MV photons were generated from the head and the photons were irradiated to a water phantom for beam evaluation. GATE simulation was segmented by two stages, the one was to generate X-ray spectrum and the other one was for irradiation X-ray to the water phantom. The quantitative results were described in PDD and beam profile. Two field size conditions were employed as 5x5 and 10x10 cm. At 10x10 cm of field size, field flatness and symmetry were 4.56% and 0.115%, respectively. The simulated LINAC head showed acceptable beam quality result for radiotherapy. After beam quality was verified, dual LINAC heads were simulated and the dual system was compared to the single head LINAC system in terms of dose deposition within the phantom. The comparison showed about 40 to 60% of efficiency increase at dual heads depending on phantom shape and position of the heads. Form the result, this study demonstrated that the simulated dual head LINAC could reduce the treatment time. The proposed dual head LINAC will be built and the system will contribute to the faster and more precise radiotherapy.

Digital radiography (DR) is replacing the analogue and computed radiography (CR) systems. In particular, scintillation flat panel detectors are used for medical applications requiring high speed and high spatial resolution. Advance quality metrics are used to assess image quality of digital detectors such as modulation transfer function (MTF) and detector quantum efficiency (DQE). For an imaging system, the MTF is a standard measure of spatial resolution, which describes the transfer of sinusoidal inputs through the system. Choosing the proper system and setup parameters for the vast range of scintillators different applications can be a time consuming task, especially when developing new detector systems. Monte-Carlo (MC) simulations is the tool to gain further knowledge of system components and their behaviour for different imaging conditions. In this work, we used the MCNP6.1.1, a Monte Carlo based model to examine the spatial resolution of an indirect converting flat panel detector, the Hamamatsu C9312SK. To calculate the MTF we used the edge image procedure, i.e. placing a highly absorbing edge directly in front of the detector, so that it creates a hard border in the grey value image. From that image, the derivation of the edge spread function (ESF) generates the line spread function (LSF). The MTF can then be calculated through Fourier transformation of the LSF. The MTF of scintillator based detectors is mainly influenced by the pixel size, the scattering of X-rays within the detector volume and the diffusion of optical light. The first two effects can be easily calculated by MCNP6. Moreover, since MCNP6 is capable of simulating optical processes (the new lower limits for energy cutoffs are 1 eV for photons and 10 eV for electrons), the light propagation was also computed. The good agreement between measurement and simulation in the MTF proves that the simulated model describes the reality quite well.

Pixelated direct-conversion detectors are poised to replace scintillators in single photon emission computed tomography (SPECT) cameras with pixel-matching parallel-hole collimators, to compensate for lost resolution while yielding higher sensitivity. However, smaller pixel sizes to improve resolution leads to increased inter-crystal signal sharing, resulting in a drop in sensitivity. It is, therefore, evident that the pixel pitch of the detector plays a prime role in tuning the trade-off between sensitivity and resolution. Our objective in this simulation study is to determine an acceptable combination of detector pixel and collimator hole pitch for improved performance of the SPECT gamma camera in detecting small lesions. Using the Geant4 Monte Carlo toolkit, a family of CZT detectors, with varying pixel size, mounted with corresponding parallel-hole pitch-matched collimators were simulated. For the entire study, the septal thickness of the collimators was fixed at 0.16 mm keeping in line with the Siemens LEHR collimator specifications, and the ratio of hole length to hole pitch was maintained at 16 to obtain approximately 7 mm planar collimator resolution when the source-to-collimator distance is 10 cm. From the simulations, the figures of merit such as sensitivity, resolution, signal sharing fraction (SSF) have been compared for pixel size ranging from 1 to 2 mm in steps of 0.2 mm. Recovery ratio calculated from a hot rod phantom filled with ^99mTc helped ascertain the optimum pixel design. Based on observations from the Monte Carlo simulations, we concluded that smaller pixels do not necessarily imply improved performance since they are more susceptible to lateral signal sharing. Pitch-matched collimators with larger holes exhibited higher detection efficiency but with deteriorated resolution, supporting our conclusion. The best recovery ratio was attained at 1.4 mm pixel pitch with matching collimator pitch and 19.84 mm hole length.

The objective of this work is to evaluate the shielding requirements of a SPECT insert for installation in the Siemens Biograph mMR in order to perform simultaneous SPECT/MR imaging of the human brain. We intend to use the radionuclides Tc-99m, I-123 and In-111. The main photopeaks of these radionuclides have the following energies: 140.51, 158.97, 171.28 and 245.35 keV. There is also a small percentage (3%) of emission at about 528.96 keV for I-123. Considering that the SPECT insert will be installed in the PET/MR system, there might be some backscatter from the bore, especially due to the high energy gamma photons. In addition, the LSO PET crystals have intrinsic radioactivity due to Lu-176 (88.36, 201.83 and 306.84 keV), which could be detected by the SPECT system.
We used GATE to simulate the SPECT acquisition. For simplification purposes and to shorten the running time, we define only one detection unit, a cylindrical aluminium shell to represent the MRI bore, and a thinner shell of LSO to simulate the PET crystals. The PET crystals and patient sources were simulated with 5.59 MBq of Lu-176 and 185 MBq of I-123, respectively. We ran four simulations with different shielding thicknesses, 1, 2.5, 5 and 10 mm, and one simulation without shielding at the back of the detector, and evaluated the energy spectra.
Results show that the peak height reduces for simulations with thicker shielding, for energies below 350 keV. Although the Lu-176 gamma emissions are almost eliminated from the spectrum with 10 mm shielding, the contribution of the Lu-176 gamma photons is lower than 5% with for a shielding thickness greater than 2.5 mm.
The results suggest that a shielding of 2.5 mm is enough to reduce the detected photons from Lu-176 to an acceptable level.