Program Listings

PET is the most sensitive imaging technique in the field of nuclear medicine. However, the non-coincident single count is typically 10 times higher than the coincidence count. This means that there are a lot of pairs of undetected annihilation photons. For single gamma imaging, Compton imaging is known as the only method to obtain directional information of an incoming gamma ray without using collimators. Compton kinematics collimation suffers from poor angular resolution due to Doppler broadening, energy blurring, and position blurring. Therefore, the spatial resolution of Compton imaging is worse than PET imaging in general. However, non-coincident single photons have the potential to extend the FOV beyond the scope of the PET FOV and to contribute to improved imaging performance. In this paper, we propose a system combining PET and Compton imaging by DOI detectors, and we show the feasibility of this system using a Monte Carlo simulator. The simulated system consisted of 48 DOI detectors arranged in five rings with a ring diameter of 80 cm. Detector materials of the scatter layer were selected from among Si, LaBr₃ and GAGG. The absorber layer was only made of GAGG crystal to have both high stopping power of annihilation photons and high energy resolution. The thickness of the scatter layer was changed from 1 mm to 10 mm, and the thickness of the absorber layer was fixed at 2 cm. For the 70-cm NEMA phantom measurements, LaBr₃ and GAGG had a backscatter ratio of about 15% without an energy window, but had higher Compton sensitivity than PET sensitivity even after back scatter reduction. For all detector materials of the scatter layer, single events were up to 30 times higher than coincidence events, and ratios of Compton events in single events were below 20%. Finally, Compton sensitivities were up to 3 times higher than PET sensitivities over 5 mm in thickness. The proposed system promises higher Compton sensitivity than PET sensitivity.

The use of monolithic crystals to detect the incident gamma photons in PET has the flexibility to define different virtual pixel sizes. These pixels determine the Lines-Of-Response (LORs) needed for the reconstruction process. Currently, virtual pixels with all identical sizes are used to form a regular grid on the crystal. In this work we present a Pixel Size Gradient method (PSG) to improve the detector response model. This approach takes into account the spatial resolution degradation caused by the edge effect and, therefore, larger pixel sizes are considered near to the crystal edges. The PSG has been implemented in a PET ring prototype. The ring is formed by 8 modules and each of them is composed by a continuous LYSO crystal coupled to a SiPM array of (12x12) elements [4]. The proposed gradient gradually decreases the pixels size from the edges to the center of the crystal. It consisted of (4x4)mm2 pixels size at the borders, followed by three rows and columns of pixels size of (2x2)mm2 and the remaining pixels with (1x1)mm2 over the (40x40)mm2 original detector area. This new approach was implemented for LMEM algorithm using the Tube-Of-Response (TOR) backprojector. It identifies the emission probabilities through an approximation of the intersection volume of the TOR within each voxel. In this work, this operator has been modified to calculate the emission probabilities with variable TOR sizes. In order to evaluate the effect of PSG in the crystal, the eccentricity and spatial resolution on the point source reconstruction were measured. These results were compared with LMEM with pixels of (1x1)mm2 but also with MLEM (12 iterations) with pixel sizes of (1.6x1.6)mm2 and (2x2)mm2 and voxels sizes of (0.5 x0.5x0.5)mm3 and (0.4x0.4x0.4)mm3, respectively. The results show similar values in volumetric resolution when LMEM (1x1)mm2 pixels size is considered, and an increase of more than 50% in the reconstructed source eccentricity for the most axially shifted sources

Measurement of the Time-of-Flight (TOF) of the 511 keV gammas brings an important reduction of statistical noise in the PET image, with higher precision time measurements producing clearer images. The common method of coupling a photodetector to scintillating crystals is to have two matching matrices, with a one-to-one coupling between the crystal and the photodetector. We propose a new geometry based on analogue SiPMs reading out a scintillator cut into slabs. This allows us to obtain (a) the depth of interaction in the crystal and (b) an improved CTR. We will present the results from the first ‘slab module’.

In this study, we propose two lower-cost high-resolution depth-of-interaction (DOI) PET-detector designs for PET/CT and PET/MR systems. These designs would use even fewer silicon photomultipliers (SiPM) than do the current non-DOI PET detectors in current clinical systems, thus clinical PET/MR and PET/CT with DOI capability may become practical while reducing the cost of such systems. The designs employ the dual-end signal read-out from a scintillation crystal pixel. However, unlike the regular dual-end read-out scheme whereby the signal from both crystal ends are only used for finding the DOI, we propose a dichotomous approach whereby the signals exiting both crystal ends are used to decode both the firing crystal location (X-Y) and the depth-of-interaction (Z), resulting in using multiple times fewer SiPM than crystal pixels. The design would reduce the cost while adding DOI capability for increasing detection sensitivity by increasing crystal depth, axial field of view, improving imaging resolution and decreasing detector-ring diameter. Two dichotomous implementations are proposed: (a) the Dichotomous Offset-Quadrant-Sharing (DO-QS) DOI design that reduces the number of SiPM used to N*N/2 for an NxN array of crystals, thus employing half the number of SiPM of the conventional non-DOI detector design, and (b) The Dichotomous Orthogonal Symmetry (DOS) DOI design that reduces the SiPM number to 4N from 2N*N. The DO-QS design is also able to decode 1x1 mm crystal-pixel array using the regular 3x3 mm SiPM to make ultrahigh resolution PET DOI detectors, in a deterministic way without using analog position-decoding. Monte Carlo simulation studies (GATE) were performed for both dichotomous designs for both BGO and LSO crystals in different array sizes and crystal depths. The DOI (Z) resolution ranged 4.8-6.5 mm (for crystal depths of 20-30 mm); the energy resolution ranged 20-25%, and the X-Y decoding error ranged 0.4-0.8 mm (for 2.4 x 2.4 mm pixels).

Particle tracking technology is currently being explored with Monte Carlo simulation studies as well as first experimental prototypes as a method to improve the planning and delivery accuracy of hadron therapy. Advanced particle tracking technology is required to provide large-area detectors capable of single-particle registration at high data rates for applications such as particle imaging and pencil beam monitoring. One such candidate large-area particle tracking detector is based on plasma panel sensors (PPS). The PPS is an inherently digital, high gain, novel variant of micropattern gas detectors inspired by many operational and fabrication principles common to plasma display panels (i.e. plasma-TV?s). The initial results of a simulation study of a realistic PPS telescope design shows that this technology is comparable to, or better than, existing silicon sensors in terms of both particle energy loss through the detector and spatial resolution for particle imaging. Further improvement of the track reconstruction accuracy is possible by reducing the PPS substrate thickness. Fabrication of the simulated PPS telescope is currently underway including thinner devices than those simulated.

Depth-of-interaction (DOI) measurement is essential for PET to reduce the parallax error, resulting in improved spatial resolution while maintaining sensitivity. One of the ultimate goals of improvement in DOI detectors is achieving isotopic spatial resolution. For this purpose, we have developed an isotropic 3D position sensitive PET detector named X’tal cube detector. The X’tal cube detector is composed of a scintillation crystal block segmented into cubes and arrays of multi pixel photon detectors (MPPCs) which cover all surfaces of the crystal block. Segment responses are made as a result of the 3D Anger-type calculation and are drawn in a 3D position histogram. In our previous study, although we achieved (0.8 mm)³ isotropic spatial resolution in the central part of the detector, we were not able to discriminate the outermost responses from the second outermost responses, in other words, crystal identification was not achieved at the edge of the detector. In this study, we improved the crystal identification performance at the edge and also got 10% higher volume resolution than that of the previous detector without any degradation. The scintillation crystal block used was a 13.1 x 13.1 x 13.1 mm³ monolithic LYSO having 3D segmentation in 17 x 17 x 17 array of 0.77 mm x 0.77 mm x 0.77 mm grids fabricated by laser processing. For each face of the crystal block, the monolithic 4 x 4 array of MPPCs (S12642-0404PB-50(X)) was coupled. We irradiated 511keV annihilation gamma rays from ²²Na sources to the detector and acquired the data with NIM and CAMAC. According to the profile of the response, 17 segments in a row were identified clearly. Its peak to valley ratio was 7.2 at the edge and about 10 % energy resolution was observed in each response.

Depth encoding detectors are required to improve the spatial resolution and spatial resolution uniformity of small diameter positron emission tomography (PET) scanners, especially for dedicated breast and brain PET. A novel transparent ceramic garnet scintillator, (Gd_xLu_1-x)₃(Ga_yAl_1-y)₅O₁₂:Ce abbreviated as GLuGAG:Ce, has been recently developed. The RGB-HD SiPM was developed based on high-density (HD) technology with the red-blue-green (RGB) SiPM and has a good spectral match for the emission wavelength of GLuGAG:Ce. In this study, DOI performance of polished GLuGAG:Ce pixels with different reflectors was measured using a dual-ended readout setup using two RGB-HD SiPMs and compared with the results obtained with polished LYSO crystals. Dual-ended readout of polished GLuGAG:Ce wrapped with enhanced specular reflector (ESR) discriminated three DOI positions better than equivalent readout of polished LYSO wrapped with ESR. Photopeak amplitude and energy resolution were also greatly improved in comparison with LYSO wrapped with ESR. Depth discrimination capability of dual-ended readout of polished GLuGAG:Ce further improved using polytetrafluoro-ethylene tape (Teflon) as the reflector. Energy resolution at each depth was under 10%. Overall, RGB-HD SiPM coupled with polished GLuGAG:Ce pixels showed outstanding performance in terms of photopeak amplitude and energy resolution (<10%). Moreover, good DOI performance using polished GLuGAG:Ce scintillators is also possible due to the self-absorption of scintillation light within GLuGAG:Ce that leads to a strong depth response. We will measure DOI performance using a range of reflectors and then assemble the first GLuGAG:Ce arrays for testing.

The use of semiconductor detectors such as a pad detector or pixel detector consisting of segmentalized detection elements in medical scanners is effective because it improves both spatial resolution and system sensitivity by providing downsized detection elements and their dense arrangement. We report on the preliminary results of a CdTe pad detector fabricated with electrode formation based on printable electronics (PE) technologies. PE technologies are alternatives to conventional electronics technologies based on lithography processes and are characterized by lower cost in small-scale production, as opposed to the large-scale production of consumer products. In addition, the PE inkjet method has a remarkable capability to cope with the redesign of circuits. There is also the possibility that the inkjet method shortens the time it takes to prototype and fabricate medical scanners’ detectors. CdTe detectors were prototyped by patterning electrodes of 14 × 5.0 mm on both sides of CdTe wafers with gold nanoparticle ink and using an inkjet machine. The process conditions, mainly for pattern writing and ink metallization, were researched and the detectors’ performances were examined. Three types of wafers—with different surface roughnesses—were prepared. It was confirmed that good flatness contributes to precise pattern writing. On one hand, the ink is metallized by heat treatment at 200–350°C; higher temperatures achieve good conductivity. On the other hand, a CdTe detector’s performance is known to deteriorate because of heat. We experimentally obtained an applicable temperature for detector fabrication. Successful detectors fabricated under appropriate conditions achieved 2.7–5.7% of energy resolution for 662 keV ?. In addition, the detector of smaller electrodes of 2.0 × 2.0 mm was fabricated. The detector also achieved a good resolution of 1.9% and adequately low leak. These results suggest the applicability of PE to radiation detector fabrication.

In high resolution positron emission tomography (PET) light sharing elements are incorporated in typical detector modules to readout scintillator crystals, which are smaller than the optical sensors. Then positioning algorithms are required to estimate the first hit crystal. The most commonly applied positioning algorithm is the center of gravity (CoG) algorithm. Conventional CoG algorithms are limited in spatial resolution, due to noise or inter crystal Compton scatter. One alternative is the maximum likelihood (ML) algorithm, which assumes a certain light distribution model, whereby uncertainties like noise or scatter can be implemented. Thus it is promising that it can handle these uncertainties or even correct for them. In this work we present a ML algorithm with a single hit model. The single hit probability densities are generated from a measured data set with a restricted CoG algorithm. Validations for detecting noise or Compton scattering and a linear photon correction model are implemented in the ML search. We apply the developed algorithm for examination and evaluation on a data set of a phantom, measured with the Hyperion-IID preclinical PET insert. We compare the results concerning efficiency, energy resolution and spatial resolution with a restricted CoG algorithm. A gain of 19% / 10% in sensitivity is measured (trigger restriction: pixel with most photon count and all/direct neighbors). Energy resolution is in the same order of magnitude and differences in spatial resolution are not visible. Hence the ML maximizes sensitivity without a loss in image quality. We also examine the effects of likelihood filtering on spatial resolution. An improvement of image quality is observed with increasing amount of filtered events. An increase in peak to valley is measured up to a factor of two. Thus badly positioned events can be successfully detected and rejected.

A large-area silicon photomultiplier (SiPM) array was evaluated for PET applications using LSO arrays with different pitch sizes. The SiPM array, with a surface area 50.2 mm x 50.2 mm, consists of 12 x 12 SiPM pixels. Each pixel has a 3.0 mm x 3.0 mm active area on a 4.2 mm pitch. Custom front-end electronics was designed to reduce the 144 SiPM signals to five signals, four for position information and one for timing information. Schottky diodes were used to block noise from SiPMs that did not detect a significant number of scintillation photons following a gamma interaction. Using three different polished LSO scintillator arrays (pitch sizes: 1.0 mm, 1.2 mm and 2.06 mm) coupled to the center of the SiPM arrays, we evaluated the performance of this SiPM array with a focus for its use in high-resolution PET applications. Measurements of energy resolution and flood histograms were obtained at a bias voltage of 28.0 V and at two different temperatures (5 °C and 20 °C). The results show that all the crystal arrays can be clearly resolved. The average energy resolution and flood histogram quality were 22.0 ± 2.7 % and 1.9 ± 0.1, respectively, for the 1.0 mm pitch array (temperature of 5 °C). In summary, this large-area SiPM array can resolve small crystal elements and high-resolution small-animal PET and PET/MRI systems based on this detector are feasible.

A new design of the small animal PET Albira system which detectors are based on arrays of SiPMs and monolithic LYSO scintillator is presented here. The system is made out of 3 rings of 8 detectors each resulting on an axial and transaxial FOVs of 148 and 80 mm, respectively. The scanner is thermally stabilized using air temperature controlled cooling. Working temperatures around 22-25C are reached with individual detector variations around 0.2oC minimizing dark count rates and gain drifts. One of the most significant improvements of the current system is the accurate photon impact determination within the crystal volume, specially its depth of interaction. This allows the calculation of the true line of response and, therefore, minimizes the parallax error at the image FOV edges. As a result, the system returns a homogeneous volumetric spatial resolution FWHM for the entire FOV, of about 1 mm3 (MLEM algorithm). Applying the NEMA protocol, this prototype reaches a Noise Equivalent Count Rate of 576 kcps at 986 uCi using the mouse-like phantom and 330 kcps at 1275 uCi using the rat-like phantom. The current geometry, similar to the former design based on PSPMTs but with higher active detection area, shows a NEMA sensitivity of 9% using a 50% energy window.

A phoswich depth of interaction (DOI) detector composed by two layers of scintillator arrays (LYSO and BGO) has been proposed. To use this design for building practical systems, a tapered multiple-element glass lightguide is necessary to couple the scintillator arrays to a PSPMT. The complete individual detector module offers an overall dimension of 52x52 mm2 that matches the external dimensions of the PSPMT package, allowing continuous positioning of the scintillator arrays for creating flat panel detectors without introducing gaps between detector modules. In this work, we define the elements adjacent to the edge in an array as “edge+” elements. Both of the layers of the phoswich detector and the lightguide have such edge+ elements. The lightguide was modified according to the photodetector geometry, aiming at improving edge and corner crystal identification and energy resolution: In an assembled detector, the optimized edge+ lightguide elements superimpose on top of both edge and edge+ anodes of the PSPMT. In that case, the edge+ lightguide elements distribute light onto both the edge and edge+ anodes, while edge lightguide elements primarily distribute light onto the edge anodes. This different light sharing scheme improves the capability of identifying edge+ crystals from edge crystals. In addition, the edge lightguide elements were enlarged to improve the light collection and corresponding energy resolution for the edge crystals. Flood images were acquired and energy resolution was calculated for individual crystals. All the edge and edge+ crystals were clearly resolved. Energy resolution was improved from 24.4 ± 4.4% to 20.0 ± 1.7% for edge crystals. The new proposed lightguide successfully improved edge crystal identification and energy resolution in the proposed pixelated scintillator detector.

Molecular imaging is now recognized as an essential tool, whether to establish new therapeutic regimens or to develop new radiopharmaceuticals. Among all imaging techniques, Single Photon Emission Computed Tomography (SPECT) settled in the area of small animal imaging and led to numerous technological and imaging protocol improvements. Today, scintigraphic, conventional SPECT and the helical acquisitions are the acquisition protocols the most commonly used. This preliminary work evaluates the AMISSA (A Multimodal Imaging System for Small Animal) SPECT system performance obtained through the different conventional acquisition protocols we implemented. The Ultra-Micro Defrise Phantom (Data Spectrum Corporation) was used to assess the reconstructed images. The conventional SPECT acquisition led to the best SNR measured in the reconstructed image. Although 6.7% and 15.4% absolute differences in terms of SNR were measured in images obtained through two helical acquisition protocols presenting different axial steps, the total acquisition time was reduced by a factor 4 and 8, respectively. The next steps will consist in evaluating the system performance in terms of sensitivity, spatial resolution and image quality for the conventional SPECT acquisitions before assessing a linear acquisition protocol based on the limited angle tomography approach. Images will be reconstructed using an iterative MAP-EM reconstruction algorithm using the TV as the energy function and compared the results with the conventional approaches. Here we are aiming at further reducing the acquisition time while maintaining a relative quantitative accuracy.

The purpose of this study was to develop a MR compatible small animal PET insert and to evaluate the performance of the developed PET and the influence of the PET on MRI performance. A PET detector block was composed of 2 × 3 detector modules, each of which consisted of a 8 × 8 array of 1.52 × 1.52 × 6 mm³ LYSO and a 4 × 4 array MPPC. The PET insert consisted of 16 detector blocks with 66 mm transverse FOV and 41 mm axial FOV. The PET gantry was shielded with gold-plated conductive fabric tapes, which had 5 axial segments and a thickness of 0.1 mm. The output signals from the PET detectors were transmitted to readout electronics using 4 m long flexible ?at cables to minimize mutual interference between PET and MRI by separately locating PET detector and subsequent electronics. The PET readout electronics was located 1 m apart from the MR bore. The energy resolution, time resolution, spatial resolution and axial sensitivity of PET were measured outside MRI. The whole-body mouse imaging of eight 21-24 g mice injected with 10-20 MBq FDG or F-18 was performed to assess the capability of small animal imaging of the PET insert. The developed PET scanner was inserted inside 9.4 T MRI to evaluate the effect of the PET on MRI performance. Energy and time resolutions were measured 13.2 ± 1.2% and 8.1 ns, respectively. Sensitivity at the center of axial FOV was 4.5% and spatial resolution was 1.1 mm at the center of FOV. The acquired PET images demonstrated high quality FDG uptake patterns and fine bony structures of mice. No considerable degradation of MRI performance was observed and SNR and homogeneity of MR image were only slightly degraded 5% and 3% by the inserted PET, respectively. Experimental results indicated that high quality mouse images could be obtained using the PET developed in this study and the PET demonstrated good MR compatibility allowing simultaneous PET-MRI of small experimental animal.

Realtime near-infrared (NIR) fluorescence imaging system could be exploited for intraoperative imaging guidance. However, the NIR fluorescence imaging technique attains only small penetration depths (a few centimeters) due to the absorption and scattering of light. The purpose of this study was to develop a realtime intraoperative imaging system using annihilation gamma-ray detectors for localization and laser projectors for visualization of surgical region embedded deep in tissue. The proof-of-principle gamma-ray detectors consisted of two front detector blocks and one backside detector block. The front detector block consisted of 8 detector modules arranged in a blank square shape and the backside detector block was composed of 4 × 8 detector modules. Each detector module was composed of a 4 × 4 array LYSO coupled to a 4 × 4 GAPD array. Front and backside detectors were located at opposite sides of each other. Commercial portable LED laser projector with a luminous flux of 500 ANSI lumens was used for the intra-operative system. Two laser projectors were positioned behind each front detector. The design parameters of the gamma-ray imaging system were adjusted and optimized using a Monte Carlo simulation tool, GATE. The performance of the system spatial resolution and count rate were estimated by simulation and by experiment using 450 kBq Na-22 point source. Experimentally measured spatial resolution and count rate were 4.1 mm and 300 counts per second which were similar to those estimated by the simulation. The beam projected by laser projectors was accurately localized the position of radiation source. Experimental results indicate that realtime intraoperative imaging system combining annihilation gamma-ray detectors and laser projectors proposed in this study would be useful to localize and to visualize surgical region embedded deep in tissue.

The aim of this project is to develop a novel handheld gamma image probe for intraoperative use in radioguide cancer surgery. This camera is defined to have highly compact size and light weight while offering high spatial resolution and high sensitivity in order to resolve small tumour lesions and sentinel lymph nodes. A highly compact detector head has been developed using new detector technology based on position-sensitive silicon photomultiplier (SiPM) array and the newly available cerium doped Gadolinium Aluminum Gallium Garnet (GAGG:Ce) scintillation crystal. A 12x12 GAGG:Ce array from the C&A Corporation in Japan with 1x1x5mm3 pixel elements coupled with a 4x4 SiPM array (ArraySL-4, SensL Inc.) was used in this study. Compact readout boards with multiplexing readout scheme have been developed for this detector module. The performance of the detector was measured using 137Cs and 57Co sources. A pinhole collimator with different aperture diameters was designed to provide high spatial resolution and high sensitivity for this camera. The preliminary test results show that this compact detector module exhibits excellent performance with an intrinsic spatial resolution around 1mm. The detector also shows good energy resolutions of ~6.9% @ 662 keV and ~18.8% @122 keV. These results indicate that this compact detector module is well suited for the desired gamma imaging probe. Further quantitative characterization of this gamma imaging probe’s performance is under study.

Aims: This report reviews low-dose tissue substitutes for micro-CT imaging and describes a method for producing tissue mimicking, durable phantoms of mouse thorax . It is presented an implementation that involves representing lungs with polyurethane, tuberculosis nodules with a mixture of resin and contrast agent and the surrounding tissues with bee wax and deprotenized bovine bone. Methods: A micro computed tomography (micro-CT) image volume of a mouse with normal thorax anatomy was use for size and layering reference. The size and shape of the phantom was a trade of between true mouse modelling and machining simplification. The polyurethane lung acorn was embedded in bee wax as a soft tissue component and with rings of deprotenized bone granules around. Micro-CT images of the phantom were acquired and assessed for the HU ranges. Results: The contrast measurements of the phantom reconstructed images compared favorably with the ranges observed in real tuberculosis studies. Conclusions: These phantoms with selectable nodule enhancement, which simulate the absorption of the multiple surrounding tissues, have strong potential as calibration tools for validating optimized acquisition protocols in micro-CT scanners. They can also be highly appropriated for setting up procedures dealing with infectious diseases, such as tuberculosis, since they will tackle the need of mouse models at the calibration/protocol design stage.

Continuous crystals represent a cost-effective alternative to segmented crystals for PET detectors, to increase sensitivity, spatial resolution and time resolution. Additionally, the light distribution within the crystal and the subsequent pattern conformed by the optical photons detected in the photodetector can potentially be used to extract the depth in the crystal where the photon interacted. Not correcting for depth of interaction (DoI) is responsible for the shift-variant spatial resolution in a PET system, in the form of parallax effects. Therefore, this is one of the most important problems to address, specially in pre-clinical systems. In this work we have characterized an 18x18x18 mm³ continuous LYSO crystal coupled to two 4x4 pixels SiPM arrays. The characterization was performed regarding spatial resolution and DoI accuracy. Since the DoI accuracy is highly influenced by the reflected light on the crystal sides, two different reflectors were used to wrap the crystal: a diffusive reflector (Teflon) and a specular reflector (Vikuiti). The speed of the algorithm used to determine the 3D interaction position of the photon was prioritised in this work. Therefore we used a method based on an analytic model, combined with a random optimizer with adaptive step size, which has shown to produce accurate results in previous studies. The resulting spatial resolution was 1.5-2.5 mm for both reflectors, not showing important differences in resolution nor bias. However, Vikuiti produced better DoI resolution (3-3.8 mm) compared to Teflon (4-4.5 mm).

This works shows the feasibility study performed with a new small animal PET design based on SiPM arrays in front of a high field MR scanner. The PET ring is made out of 8 detector blocks each containing a single monolithic 50x50x10 mm3 LYSO block and coupled to an array of 12x12 SiPMs. The PET system performance has been evaluated both alone and in front of the MR with about 400-500 mT magnetic field. We evaluated two important parameters such as the energy and the spatial resolution. We did not observe, on overage, any energy resolution degradation when the system was inside the residual magnetic field. A slight shift to lower energies of about 7 keV on average was determined. The use of monolithic blocks and the high readout granularity has allowed us to explore novel approaches to precisely determine the 511 keV photon depth of interaction (DOIs). An accurate DOI resolution permitted to efficiently correct the parallax error for all events within the FOV, even those near the edges, resulting on a homogenous spatial resolution across the entire FOV. The spatial resolution performance did not degrade in the presence of the residual magnetic field.

The phenoPET system is a PET scanner dedicated for plant research. It was developed and is being established at the Research Center in Jï¿½lich. The scanner employs LYSO scintillators of 1.85ï¿½1.85ï¿½10 mm3 and digital SiPM arrays as photo detectors (DPC3200-22-44, Philips Digital Photon Counting) which are arranged in three stacked rings. The field of view measures 18 cm in diameter and 20 cm of axial height. The electronic hardware allows coincidence detection in realtime performed on a Kintex-7 FPGA (Xilinx). The system was introduced in 2014 [1] but at this time consisted of a single detector ring only. Now we will present results obtained from measurements with the complete scanner.

Performance measurements were conducted on a prototype PET/CT system under development by GE Healthcare. The system utilizes 4.0 x 5.3 x 25 mm LYSO crystals that are coupled to SiPMs. A light guide channels a 4x3 array of crystals onto a 3x2 SiPM. The ring diameter is 744 mm and the transaxial field of view (FOV) is 700 mm. The axial FOV is 248 mm and is comprised of 89 2.78 mm thick slices. The scanner has time-of-flight (TOF) capability. The coincidence timing window is 5.25 ns and the energy window is 425 to 650 keV. Methods: Performance measurements of spatial resolution, sensitivity, image quality, scatter fraction and count rate were obtained using NEMA NU2-2012 methodology. The image quality (IQ) test was performed with a sphere to background ratio of 4:1 and images were reconstructed with a fully-3D OSEM algorithm with and without TOF; otherwise the same reconstruction parameters were used for all images. Timing and energy resolution were also measured. Results: The radial and tangential transaxial and axial resolutions measured from FBP reconstructed images were 4.5, 4.3 mm and 5.2 mm FWHM at 1 cm from gantry center respectively. The system sensitivity was measured as 22.0 cps/kBq at the gantry center. The peak noise equivalent count rate was 283.4 kcps at 19.7 kBq/ml and the scatter fraction was 38.8%. The contrast of the 37, 22, and 10 mm spheres in the IQ phantom for TOF (and non-TOF) were 87.5 (82.8), 68.4 (67.2), and 34.6% (31.1%) respectively. The background variability of the same spheres for TOF (and non-TOF) were 3.6 (4.3), 4.6 (5.6) and 7.1% (8.7%) respectively. The lung error was 6% for TOF and 10% for non-TOF. A timing resolution of < 380ps and energy resolution of < 9.2% were measured. Conclusions: The prototype PET/CT system combines high sensitivity and high count rate capability with TOF capability.

Combined positron emission tomography (PET) and magnetic resonance imaging (MRI) has shown potential to provide a powerful tool for disease characterization as it enables the simultaneous measurement of molecular, functional and anatomical information of the body. However, the availability of whole-body simultaneous PET/MRI has been limited by its high cost. To address this issue, we have developed an RF-penetrable PET technology that can be inserted into an MRI system without requiring modifications to the MR hardware. The prototype PET insert consists of 16 PET detector modules in a 32 cm ring pattern with 1 mm inter-modular gaps. By using electro-optical coupling technology, and batteries for power, the PET insert is electrically floating relative to the MRI RF ground which allows the RF field transmitted from the built-in body coil to penetrate through the PET ring. We performed experiments with various configurations to study the RF-penetrability of the PET insert in the MR system: (a) No PET insert, (b) with PET insert, (c) with PET ring ends blocked with copper conductor but gaps open, and (d) with PET inserted, ends open, but inter-modular gaps blocked with copper tape. With these configurations, we acquired B1 maps and MR images with gradient echo (GRE) and fast spin echo (FSE) sequences. We then compared the magnitude, SNR, homogeneity, and the radial/axial variations of the acquired B1 maps and MR images. For cases (a) and (b), the B1 magnitude and the image SNR dropped by 3.4 dB and 3.9 dB. The FSE SNR and homogeneity in (d), 13.4 and 42.1%, were lower than in (c), 15.4 and 54.8%. Radial variations for SNR and homogeneity were also deteriorated from (c) to (d), suggesting that more RF penetrates through the gaps than the ends. Our results have shown that RF power penetrating through the inter-modular gaps is more important for overall B1 uniformity and MR image SNR than RF power penetration through the ends of the PET cylinder.

Using time of flight (ToF) measurements for positron emission tomography (PET) is an attractive avenue for increasing the signal to noise (SNR) ratio of PET images. However, achieving fast timing resolution required for high SNR gain using silicon photomultipliers (SiPM) requires many resource heavy high speed readout channels. A method of multiplexing many SiPM signals into a single electronic channel would greatly simplify ToF PET readout systems. In this work the effect of multiplexing on the timing resolution of analog SiPMs is modeled, simulated and tested. The simulations and experiments show that baseline fluctuations from cumulative uncorrelated dark noise are the most important cause of timing degradation, but their effects can be mitigated with baseline correction. A charge sharing network for position sensitive multiplexing is proposed and tested. Results show a full width at half maximum (FWHM) coincidence timing resolution of 232 +/- 2ps for single 3mm x 3mm x 20mm LYSO scintillation crystals with 16 SiPMs multiplexed to a single timing channel (in addition to 4 position channels). Measurements with a 4x4 array of 3mm x 3mm x 20mm LFS crystals show excellent crystal separation with a minimum ratio of distance between crystal peaks to standard deviation of crystal peaks of 11.9. The mean FWHM coincidence timing resolution of the 4x4 LFS array was 278 +/- 7ps. All experiments were performed at room temperature with no thermal regulation. These results show that fast timing resolution for ToF can be achieved with highly multiplexed analog readout.

We are developing a brain dedicated PET-MRI system in which our original, four-layer DOI-PET detectors are integrated with a birdcage RF-coil. The main point is making the PET ring diameter as small as possible while avoiding interference with the RF-coil, in order to realize a highly sensitive PET system with high spatial resolution by making full use of DOI measurements. Previously, we reported 1.6 mm spatial resolution with prototype PET detectors integrated with the RF-coil. However the PET system had only a 12 mm axial FOV and a large electrical circuit was needed.
We developed a new four-layer DOI-PET detector for the full system of the proposed PET-MRI; it has an extended axial FOV (96 mm) compared to the previous system which is achieved by arranging three 16 × 16 × 4-layer crystal blocks along the axial direction. The LYSO crystal dimensions are 2.0 mm × 2.0 mm × 4.0 mm. The single-channel multi-pixel photon counters (MPPCs, S12641PA-050) are arranged in an 8 × 8 array with 4.1 mm pitch and 1 mm dead spaces between the MPPCs which have a 3 × 3 mm² sensitive area. The 64 channel signals are read out through an ASIC functioning as an amplifier and sum circuit to reduce the circuit board size.
We evaluated performance of the new PET detector and influence of the detector on MRI images during simultaneous measurements. As a result, for PET performance, we found the averaged peak-to-valley ratios were 2.60 (without MRI measurement), 2.60 (fast spin echo, FSE) and 2.63 (echo planar imaging, EPI). Energy resolutions of 14.0 % (without MRI), 14.6 % (GRE) and 14.7 % (SE) were obtained for single crystal elements. For the MRI, the B0 drop was smaller than 5.0 ppm and the SNR degradation was lower than 7% by the PET measurement.

A modality for real-time simultaneous x-ray and nuclear imaging is currently being developed by our group, by combining a c-arm and a gamma camera. The interventional availability of hybrid x-ray and nuclear images could benefit image guided (oncological) procedures. The purpose of this study is to develop a calibration method which improves the resolution and co-registration of simultaneously acquired hybrid x-ray and nuclear images by estimating the geometric parameter set used for reconstruction and co-registration. Resolution loss and co-registration errors are the potential consequence of incorrect description of the acquisition geometry, defined by the relative position of the gamma camera, collimator, x-ray tube and x-ray detector. These parameters can be determined by direct measurement or using a calibration method. A calibration method was developed using acquisitions of point sources visible on x-ray and nuclear images. Subsequently, direct physical measurement was compared with the calibration procedure by determining the resolution and co-registration error of hybrid images for both methods.
The resolution and co-registration of simultaneously acquired x-ray and nuclear images was substantially improved by using a calibration method that optimizes the geometric parameter set as compared with a parameter set acquired by direct physical measurement. The mean FWHM was lower for the parameter set estimated by the calibration method (1.13 cm) than for the parameter set acquired by direct measurement (1.18 cm). The mean co-registration error was lower for the parameter set estimated by the calibration method (0.08 cm) than for the parameter set acquired by direct physical measurement (0.67 cm). A calibration method was presented that improves the resolution and co-registration of simultaneously acquired hybrid x-ray and nuclear images by optimizing the geometric parameter set as compared with a parameter set acquired by direct physical measurement.

A MR compatible PET insert has been developed in our group employing SiPM and a charge signal transmission method that transmits the charge signal of SiPM to subsequent electronics through 4 m long cables. The previously developed PET/MRI system, however, had several limitations, such as the instability of PET performance due to temperature variations inside PET gantry and the degradation of SNR in the MR image due to the inserted PET. The purpose of this study was to optimize the design of PET/MRI system to maintain stable temperature inside the PET gantry and to minimize mutual interference between PET and MRI. The PET insert was composed of 9 PET detector blocks that were arranged in a circle around the subject with 390 mm diameter and 60 mm axial FOV. The PET detectors were cooled using a water chiller system to compensate temperature fluctuations caused by eddy currents. PET analog electronics were enclosed in an aluminum box and located 50 cm apart from the MR bore. The FPGA-embedded DAQ and power supply were installed at outside MR room. PET performance was measured inside and outside MRI using spin echo and gradient echo sequences. MRI performance was evaluated with and without the PET insert. Human brain imaging was also acquired to evaluate the performance of simultaneous PET and MR imaging. No significant degradation of the PET performance caused by MR was observed when the PET was operated under various MR imaging sequences. The homogeneities and SNRs of MR images obtained with and without the PET insert installed inside the MR bore were similar and changed by only 1% and 3%, respectively. High quality simultaneous PET and MR images of human brain were successfully acquired. Experimental results indicate that simultaneous PET and MR imaging is feasible without considerable mutual interference between PET and MRI by employing the PET/MRI proposed in this study.

The development of MRI compatible detectors based on compact solid state photomultipliers has recently led to the development of PET/MRI systems for human imaging in the form of brain PET insert or whole body, fully integrated solutions. In this work we present the PET detector design that is being developed for the TRIMAGE project. The TRIMAGE consortium aims to develop a cost-effective, fully-integrated PET/MR/EEG brain scanner dedicated to the early diagnosis of schizophrenia and other mental health disorders. In this work we present the PET detector design that is being developed for the TRIMAGE project. The PET component of the TRIMAGE system features a full ring made up of 18 rectangular detectors, approximately 55 mm (transversal) × 163 mm (axial) comprising twelve independent tiles each. The tile is built with a two staggered layers of LYSO matrices for DOI uncertainty reduction, 7x7 and 8x8 for the top and bottom matrices , respectively, the latter being one-to-one coupled to 64 SIPMs that are, in turn, read out by a dedicated 64 channels ASIC. One of the main features of the PET design is the solution proposed for the layer and pixel identification that is performed in the front-end FPGA. In this way, it would be possible to compress the data packet generated by each event, thus allowing the data transmission of all single events to a motherboard where coincidence are detected and acquired. The feasibility of this concept is here validated with Monte Carlo simulations of a single detector tile using GAMOS. Energy resolution of each layer as well as layer identification capability are evaluated. Experimental results on the read-out of a tile will be also presented.

We have reported the feasibility of Multi-Voltage Threshold (MVT) based PET DAQ system using field-programmable-gate-array (FPGA) without ADC and TDC. In these sampling methods, a multi-voltage threshold (MVT) method has been employed to generate samples of an event waveform of PET with respect to the user defined threshold to extract energy and time information of a PET event. The goal of this study was to extend MVT based multi-channel front-end circuit and DAQ with FPGA to utilize for the development of a functional PET system. The PET detector modules were composed of 4×4 matrix of 3×3×20 mm³ LYSO, a 4×4 array GAPD. Output charge signals of PET detector modules were amplified and fed into 4 comparators to generate trigger signals in analog circuit. The energy of the detected gamma ray was calculated by integrating the digitized pulse and the arrival time was determined from the time stamp of each trigger signal by FPGA. The calculated data packet containing energy, time and position information was stored in list mode on the host computer. The performance of multi-channel analog circuit and DAQ system was evaluated by measuring the energy resolution and time resolution and the values were 15% and 880 ps, respectively. These results verify the feasibility of the proposed PET detector module, MVT based front-end circuit and DAQ system with only FPGA for the development of PET. The main advantage of this approach is simple design of electronics and DAQ, and scalable design for PET system. Further study will be performed to improve the performance and to develop a functional PET system using the proposed method.

Maintaining excellent timing resolution in time-of-flight positron emission tomography (TOF-PET) at system scale requires an enormous amount of high speed, high bandwidth electronic channels and components. Therefore, to minimize cost and complexity of a system's back-end architecture and data acquisition, many analog signals are multiplexed to a single channel using techniques that encode timing, energy, and position information. With progress in the development silicon photomultipliers (SiPMs) having lower dark noise, after pulsing, and crosstalk along with higher photodetection efficiency, a coincidence timing resolution (CTR) well below 200 ps FWHM is now easily achievable using single pixels of 20 mm long, lutetium-based inorganic scintillators. However, multiplexing many SiPM channels to a single readout will unavoidably degrade CTR. This work seeks to optimize trade offs between high analog multiplexing ratios and achievable CTR. Measurements with a fast scope indicate that for a 16:1 multiplexing ratio CTRs of 186±3 ps FWHM and 160±1 ps FWHM CTRs are achievable with 3x3x20 mm³ LYSO:Ce and 90%LuGSO:Ce(0.025 mol%) crystals 1:1 coupled to SiPMs, respectively. With a scalable waveform digitizer (DRS4) and 16:1 multiplexing ratio, the same scintillators yielded 211±2 and 202±2 ps FWHM CTR. Using information gained from test boards that multiplex many channels, a prototype detector module comprising a 4x4 array of SiPMs and readout with a 16:2 multiplexing ratio (one channel for timing, one for energy and position information) is presented.

Recent studies suggested that [NaF]-PET provides adequate markers of plaque inflammation in the coronary tree. However, PET imaging of plaque inflammation in the coronary arteries is challenging because of the respiratory and cardiac motion and the partial volume effects, particularly for small lesions. The objective of this project is to obtain reliable motion-compensated (MoCo) and partial volume corrected (PVC) cardiac PET images for the Biograph TPTV PET/CT and the Biograph mMR PET/MR systems. We evaluated the effect of motion compensation (MoCo) and partial volume correction (PVC) on the detectability and quantification of vulnerable plaques in the coronary arteries by means of realistic simulations. Both, the MoCo and PVC method, were then applied in a pilot [NaF]-PET patient study. In all cases MoCo-based image reconstruction was performed using the STIR software. The PVC was obtained by the clinical implementation of a local projection (LP) method, previously evaluated in preclinical PET. The results obtained with the simulated datasets show a significant improvement in the detectability and quantification of small lesions when MoCo is incorporated into the system matrix. Moreover, improvement in the quantification of the vulnerable plaques was observed when applying the LP method for PVC. When using the LP method on MoCo images SUV values were within 10% of the true value in all cases. Further, we observed a significant increase of the lesion-to-background ratio of the vulnerable plaque in the pilot patient study when MoCo and PVC were incorporated into the reconstruction framework: from 1.3 in the static reconstruction to 1.8 in the MoCo+PVC images. The LP method enhanced these ratios further. In conclusion, we have built a MoCo and PVC image reconstruction framework for the Biograph mMR and TPTV systems using STIR. First results from simulations and pilot patient indicated an improved detectability and quantification of small, plaque-type lesions.

Data-driven gating (DDG) is an elegant way to overcome respiratory motion blur in Positron Emission Tomography (PET) scans of thoracic and abdominal structures. In contrast to conventional gating approaches, e.g. using video cameras or pressure belts to monitor respiratory phases of the patient during the scan, no additional hardware is needed in DDG. Instead, a respiratory signal is derived directly from the measured PET raw data. However, it is clear that the quality of DDG methods is directly influenced by the statistical quality of the acquired coincidence data. In this work we analyze the relation of data statistics and data-driven gating success in terms of motion resolution of PET-active structures like tumors. For this, we analyzed high-statistic [18F]FDG patient data by artificially degrading data statistics. We found a consistent behaviour of resolved respiratory displacements of PET-positive structures as a function of data quality. This leads to the possibility to give reasonable estimations of the real respiratory displacement of structures even in cases of limited data statistics. This may be of particular interest in the context of using gated PET image data for radiotherapy planning where precise information of target motion is vital for successful treatments.

In SPECT, diagnoses based on quantitative measurements may be uncertain due to high noise levels and low spatial resolution. I-123 DAT SPECT has been shown to have a relatively high sensitivity and specificity, but improving image quality could potentially increase these values even further, especially for early cases with parkinsonian syndromes. The aim of the study was to optimize the reconstruction protocol for I-123 DAT SPECT with a LEGP collimator, with a resolution recovery algorithm (Evolution Toolkit, GE) included in the iterative reconstruction, and compare to images reconstructed without resolution recovery. The optimization concentrated on critical frequency of the post-reconstruction Butterworth filter and the number of reconstruction iterations. Monte Carlo simulations of a morphological brain phantom were used, and from contrast-to-noise diagrams it was found that a critical frequency of 0.50 cm-1 (power factor: 8) was most optimal of the studied filters. The optimal number of OSEM iterations was evaluated by a radiologist, specialized in nuclear medicine, and 8 iterations with 6 subsets were chosen. A group of 20 subjects diagnosed with Parkinson’s disease (PD) were then compared to a group of 20 healthy controls, with respect to uptake ratios for caudate nucleus, putamen and the whole striatum (background region: the occipital lobe or the whole cortex), calculated using the software Exini DAT. It was found that the group differences were highly significant both with and without resolution recovery. In putamen, where early stages of PD first manifests, the average uptake ratios increased from (5.21 +/- 0.97) to (5.47 +/- 0.93) for healthy controls and (1.98 +/- 0.72) to (2.08 +/- 0.79) for PD when including resolution recovery (background: occipital lobe). The already high group difference significance improved slightly. The images were also found to be more visually appealing, with a higher spatial resolution, by the nuclear medicine specialist.

The partial volume effect (PVE) is a significant cause of quantitative inaccuracy in PET. In this work we evaluate the local projection (LP) method for partial volume correction (PVC) in clinical PET imaging. The method requires the segmentation of only a few tissues within a small volume approximately centered on the lesion to be quantified. Activity estimates for each tissue within the small volume of interest (VOI) are obtained by fitting an emission model to the raw projection data. The activity in each voxel within the VOI is substituted with the tissue-activities obtained with the LP method and the resulting image is forward projected. The obtained sinogram is reconstructed into a PVC image using STIR software. We used a Biograph True-Point True-View (TPTV) PET/CT system for the clinical evaluation of the LP method. First, we performed acquisitions of a NEMA-2012 IQ phantom, in order to validate the performance and the accuracy of the method. Then, we applied the method to selected [NaF]-PET studies of patients with tracer accumulation in atherosclerotic plaques of carotid arteries. Hot sphere recovery coefficients (RC) were chosen as the figure of merit in evaluating the effects of the LP-based PVC. The RC values obtained in the PVC image were significantly better than the ones obtained in the uncorrected image, while the RC values obtained from the LP method were close to the experimental measured values (RC close to 1.0), except for the smaller sphere (10 mm), where the total recovery of the PVE was not achieved. Regarding the first patient data analyzed in this work, we observed in the CT image two regions with active calcification, which were not separated in the standard PET image due to the PVE. In the PVC PET image, the two calcifications were clearly visible. In summary, we confirmed the good performance of the method in clinical cases, provided that an accurate tissue-segmentation from a high-resolution anatomical image is available.

Due to the varying position and shape of flexible MRI RF surface coils, the creation of attenuation maps for these coils is a challenging task. Nevertheless, coil material (metal, plastic, rubber) attenuates the PET signal to a considerable amount. Thus, including a coil attenuation into the existing attenuation map is essential. In this work, we present a method to determine the position and shape of flexible coils with the help of the Microsoft Kinect depth camera. For the creation of the attenuation map template, CT scans of the flexible 32-channel coils are performed. Since the anterior part of the coil is flexible in shape, it is equipped with 15 cod liver oil markers visible for both Kinect camera and CT. PET of a cylindrical phantom is performed with and without the coil for comparison. Prior to the PET scan, Kinect depth data of the phantom with the coil on top are acquired. MRI scans using the integrated body coil are performed to validate the position determination via Kinect camera. Using the corresponding marker position from CT and Kinect, the CT dataset is warped to the shape of the coil in the Kinect data with a 3D thin-plate splines method and converted into an attenuation map using a bilinear conversion from Hounsfield units to attenuation values. Subsequently, the attenuation map is placed in relation to the patient table according to the Kinect-derived marker positions. First results show that the transformation of the template CT dataset according to Kinect marker positions during PET/MRI leads to appropriate results. Furthermore, the position of the coil can also be determined for an accurate placement of the attenuation map in relation to the patient table. The determination of position and shape of flexible surface coils using the Kinect camera can be a way to include the CT-based coil attenuation map in PET/MRI acquisitions without the need for additional MRI scans.

One of the main benefits of PET/CT imaging is its ability for absolute quantification. Calibration according to the manufacturer‘s procedure specifies an accuracy of about 10%, whereas especially in dynamic clinical studies a higher quantification accuracy is desired. Therefore a more accurate calibration is needed. At the Gemini TF, a scanner with large area detectors, there are differences in the measurement set-up between calibration and clinical acquisition. This study intends to evaluate the influence of those differences with the aim to increase the accuracy of quantification. The major difference herein is the acquisition format, as calibration is performed in sinogram-mode and clinical acquisition in list-mode format. Using the list-mode format for the calibration, increases the activity recovery coefficient for the sinogram-based acquisition from 0.93±0.08 up to 1.00±0.03. This is however only valid for the calibration set-up but not for clinical situations. Considering more realistic situations like lesions outside of the center of the FOV or additional random events coming from highly accumulating regions outside the FOV (like bladder or brain), a different calibration can be found. In the evaluation of clinical oncological datasets with low count rates, no significant improvement of the reconstructed mean activity concentration can be achieved. The mean recovery in the bladder was about 0.91±0.08 in comparison to 0.92±0.09 for the manufacture‘s calibration. For very high count rates the normalization of the scanner has to be adapted to improve the quantification accuracy.

Respiratory motion correction in PET imaging is of crucial importance in the thorax. On current scanners, respiratory gating is performed based on the signal of an external device. Recent methods extract a respiratory signal from raw PET data exploiting data driven (DD) methods, avoiding the use of external equipment and having potential increased fidelity to internal motion. However, many of these DD methods determine the signal up to an arbitrary scale factor: it is not know if maxima and minima in the signal are related to in- or expiration states, possibly causing inaccurate motion correction. The aim of this work is to propose methods based on PCA and compare their performance on clinical data with other existing methods based on registration of reconstructed gates. The methods rely on the assumption that respiratory motion is mostly in the axial direction. When using PCA for DD gating, the respiratory Principal Component should therefore correspond to ï¿½up and downï¿½ motion. We propose two methods that determine the direction of motion by comparing similarities between the PC and the gradient of the data in the axial direction. We tested the methods on chest patient datasets acquired in 3D PET listmode, using activity levels according to routine clinical protocols, with monitoring with an RPM camera. For each patient different time intervals were evaluated ranging from 100 to 300 seconds in duration. The sign obtained by the 4 methods was compared to the ground truth provided by the RPM. The proposed methods showed lower failure rates compared to the registration-based methods. Their performance worsens as the correlation between the PCA signal and the RPM decreases (nevertheless remaining superior compared to the other methods) suggesting that the methods are more likely to fail when the PCA signal is poorly correlated with respiratory motion, presumably due to the presence of other types of patient movements.

Background: The Argus (AKA BioPET) scanner manufactured by Sedecal requires a lot of user intervention in order for the PET scanner images to be quantitative. Furthermore, there is no easy procedure by which one can maintain calibration factors for different positron emitting isotopes like 64Cu, 76Br and 89Zr. A java application by the name DICOM Tool has been developed to address this deficiency. Materials and Methods: The DICOM Tool was written in Java for ease of portability between operating systems. Java also provides the libraries needed to develop a GUI front end for the application. For the backend DICOM manipulation the open source DCM4CHE 3.1 was chosen because of its breadth of DICOM utilities. DICOM Tool implements a branching fraction quantitation algorithm in which the calibration factor which converts the native PET DICOM unit of milli-counts per sec (mcps) to Bq/ml is boot strapped from a Ge68 sealed source, and then converted to the quantitation scale factor for 18F, 64Cu, 76Br and 89Zr using the branching fraction ratio between 68Ge and the target positron emitter. A second order correction scale factor is further applied which brings the accuracy of quantitation to within +/- 5%. The DICOM Tool also allows for the editing of several key DICOM fields like injection time, activity and the specimen weight so that top tiered imaging packages like MIMSoftware’s image viewer package will convert PET units to SUV. Results: The DICOM Tool has been validated with 18F, 64Cu, and 89Zr with quantitation accuracy less than +/- 5%. It is also in production use on the NCI/MIP BioPET and has proven to save time in converting images from mcps to Bq/ml and providing a mechanism by which users can archive their imaging data to PACS.

Purpose: Imaging Y-90 microspheres with PET/MRI following hepatic radioembolization has the potential for predicting treatment outcome and, in turn, improving patient care. In liver imaging, respiratory motion induces blurring in the PET image and results in poor quantification. The advent of simultaneous PET/MR scanners has promoted the use structural MR imaging to extract real-time motion information and incorporate it in the PET reconstruction. In this study, we evaluate the feasibility of MR guided motion correction for Y90 PET imaging on simultaneous PET/MR in both phantom and human studies. Methods: A uniform water phantom was injected with Y-90 microspheres and placed on a custom-built motion stage designed to simulate respiratory motion. The phantom was then imaged on the Siemens mMR both under static (ground truth) and moving conditions. Throughout the PET acquisition, a 2-D MR navigator was used to track the motion of the phantom. The emission data from the moving scan was either reconstructed with or without including the MR motion estimates into an OSEM reconstruction algorithm. Similarly, a human subject that was treated with Y-90 was scanned using the same imaging protocol as the phantom scan. Results: Visually, significant blurring was observed in the phantom scans that were collected with the motion phantom as compared to the ground truth static scan. Quantitatively, an error of up to 48% was measured, but was corrected for when motion information was used in the reconstruction. Similar results were also observed in the human scans. Conclusions: The use of MR motion information in the PET reconstruction of Y-90 biodistribution results in large reduction in blurring and in quantitative errors. Simultaneous PET/MRI might be better suited for such studies compared to sequential PET/CT as it could provide real-time motion information that could be used to improve PET reconstruction.

The capability of SPECT imaging developed on an preclinical PET scanner can provide a combined PET/SPECT dual modality imaging environment, potentially opening the opportunity for many new clinical and preclinical applications, However, an obstacle to the implementation of the protocol is the interference of signal between two isotope, particularly in the down-scatter from high energy gamma-ray into low energy window. In this work, we developed a new down-scatter correction method for simultaneous dual isotopes PET/SPECT imaging. A Siemens Inveon preclinical PET with a slit-slat collimator insert was modeled using the GATE/MPHG Monte Carlo simulation software developed by our laboratory. For dual imaging capability, dual energy window settings at 120-160 keV and 350-650keV were used to acquired SPECT and PET imaging, respectively. The procedure of the proposed method includes two steps: (1) A 18F uniform phantom needs to be acquired for establishing a transform function and a scaling factor between the two energy windows beforehand. (2) During dual isotopes simultaneous acquisition (DISA), the down-scatter distribution can be estimated by multiplying the acquired projections at 350-650keV with the transform function, and then the absolute scatter amount can be finally obtained by scaling the resulting scatter distribution by the scaling factor. Various phantoms were conducted to compare the image quality using the proposed method and triple energy window method, which is common used for scatter correction in DISA. The results indicated that images generated by our method approach pure 99mTc isotope imaging, and always outperformed the triple energy window (TEW) method in terms of coefficient of variation (15.08% vs. 21.04%) and contrast recovery coefficient (0.848 vs. 0.58). In conclusion, we have developed a novel down-scatter correction method for DISA imaging. It is expected that the method can also be applied to dual isotopes SPECT imaging with high energy peaks that down-scattered to low photopeak data, such as Tc-99m/Tl-101.

In small-animal SPECT, the multi-pinhole collimators are widely applied to improve the resolution and sensitivity. However, the axial distortion arises in circular-orbit pinhole SPECT due to incomplete sampling. In this study, a sampling evaluation model is proposed based on the Tuy’s condition: every plane intersected with the object space should also intersect the sampling orbit at least once. The sampling completeness coefficient (SCC) of each voxel within the field of view (FOV) is estimated by the proposed model. According to the reconstructions of 7-disk Defrise phantom with the circular-orbit 4-pinhole SPECT systems, the axial distortion exists in the regions where SCCs are less than 0.9. Therefore, the criterion of sufficient sampling is set at SCC greater than 0.9. For mouse whole-body imaging, a helical 4-pinhole SPECT system with a cylindrical FOV enclosed by a 40 × 40 × 100 cubic mm cuboid is proposed based on the sampling completeness evaluations. The helix pitch is determined by a modified pitch limitation model. The pitch limitation of the SPECT system with 1.52X magnification and 49 × 49 square mm detection area is about 46 mm. Moreover, two helical scanning modes, the single helix and double helix, are compared with their SCC maps. According to the SCC evaluations, the helical 4-pinhole SPECT system with the double helix and a 1/2 extended pitch at both ends of the FOV is the appropriate configuration for mouse whole-body imaging. Further verifications of the sampling evaluation model and the proposed helical 4-pinhole SPECT system are in progress.

Radioiodine therapy with I-131 is used for the treatment of suspected recurrence of differentiated thyroid carcinoma. Pre-therapy I-124 PET/CT with a low administered activity (~1% of I-131 activity) can be performed to determine whether uptake of I-131, and thereby the desired therapeutic effect, may be expected. However, false negative low activity I-124 PET/CT scans as compared with the post-therapy high activity I-131 SPECT/CT have been reported by several groups. The purpose of this study is to quantitatively compare the detectability of lesions on I-124 PET/CT and I-131 SPECT/CT.
Phantom measurements were performed using the NEMA-2007 image quality phantom. The contrast-to-noise ratio (CNR) was calculated at different activity concentrations as a measure of detectability for each sphere size (10, 13, 17, 22, 28 and 37 mm). The I-124 activity concentration was expressed as a percentage of the I-131 activity concentration required to achieve the same CNR. This metric was defined as the detectability equivalence percentage (DEP). The DEP was calculated for time-of-flight (TOF) and non-TOF PET.
The DEP was 1.0, 1.2, 1.0, 2.3, 4.0 and 7.2% for the spheres with a diameter of 10, 13, 17, 18, 25 and 37 mm respectively, for TOF PET. For non-TOF PET, the DEP was 3.9, 5.0, 5.5, 14.9, 22.3 and 27.8% for the spheres with a diameter of 10, 13, 17, 18, 25 and 37 mm respectively.
For small lesions (=17mm) on TOF PET images, a low activity of I-124 (~1% of I-131 activity) is sufficient to achieve a similar detectability of lesions on I-124 PET/CT and I-131 SPECT/CT. False negative low activity I-124 PET/CT scans as compared with the post-therapy high activity I-131 SPECT/CT may be ascribed to differences in detectability for large lesions (>17 mm) and for non-TOF PET.

PET reconstruction results in images with correlations between neighboring image voxels. Alternative reconstruction algorithms, such as OSEM reconstruction incorporating resolution modelling (RM) can significantly alter this voxel covariance. While RM has been demonstrated to reduce voxel variance, it has been suggested that the increased covariance results in increased region-based ensemble variance (EV). The aim of this work was to develop and evaluate methods to estimate covariance kernels and evaluate their utility in predicting EV. The end goal of the work is to enable tools for comparing regional EV from reconstruction algorithms with very different correlation structures and in a way that can be derived from routine clinical PET data. Fifty sequential 2.5 minute images of a uniform Ge-68 phantom were acquired on a Siemens Biograph mCT. Images were reconstructed: with and without RM (Siemens HD-PET); and with isotropic voxel dimensions of 2 mm and 4 mm. A 40 mm × 40 mm × 40 mm cube of voxels was extracted from the centre of the image and used to estimate covariance kernels. Spherical regions of different diameters (40 mm max) drawn on the uniform phantom were used to estimate EV of the mean regional concentration using the 50 image replicates. This was compared against estimates of EV derived from the covariance kernels. Estimation involved convolving a mask of unit volume defining the region shape, with the covariance kernel before sampling the mean value of the resulting image within the region. As expected estimated covariance kernels when RM was used were broader. Good agreement (-40% to +30%) was observed between the measured EV and values estimated from the covariance kernels. RM and larger image voxels showed reduced EV values for small spherical regions. For large regions only very small differences in EV were observed between reconstruction algorithms. The use of Fourier methods to derive voxel covariance kernels appears to show promise.

In chest digital tomosynthesis (CDT), algorithm development is challenging in that a limited number of projections are acquired over a limited angular range. To confirm the feasibility of back-projection (BP) algorithm, filtered-backprojection (FBP) algorithm, algebraic reconstruction technique (ART), and maximum-likelihood (ML) algorithm in newly developed CDT system, quality metrics were conducted using ball phantom, Shepp-Logan phantom and chest phantom. The reconstructed image was acquired from the 21 projections over a 30? angular range using a prototype CDT system. To investigate comparison of the limited-angle cone beam reconstruction methods for chest tomosyntehsis system, the projection data were reconstructed using BP, FBP, ART, and ML algorithm. We investigated the quality metrics, such as contrast-to-noise ratio (CNR), and artifact spread function (ASF), to evaluate the quality of tomosynthesis image in the reconstruction plane (XY) and in vertical dimension (Z). The ART and ML methods could provide better depth resolution information than FBP method. The ART and ML methods resulted in the superior quality of reconstructed image with both high-contrast bones and low-contrast heart. The FBP method gave the best CNR for spine images by enhancing their contrast and sharpness, while the quality was poor for heart image.

In List-mode PET iterative image reconstruction two major approaches exist to model the detector response. One is based on measuring or estimating the detector response before the reconstruction process. Although such an approach has the advantage of accurately accounting for the detector response it requires large memory for storing the system response matrix (SRM). The alternative approach is based on using an on-the-fly projector to compute the required parts of the SRM. Until recently the on-the-fly projector approach was limited to over-simplistic detector models largely due to its computational burden. Following the growth of the general-purpose processing on graphics processing units (GPGPU) that allows large computational power for scientific applications, the use of complex on-the-fly projectors has now become feasible. In this work we compared the two reconstruction approaches in terms of both image quantitative and qualitative performance as well as reconstruction speed. Three different statistical quality stored SRMs and three on-the-fly projectors (with increasing complexity: Siddon, Gaussian, IRIS: iterative random intrinsic detector response function sampling) were considered in this comparison. Worst performance was obtained with the Siddon on-the-fly projector and the low statistical quality SRM reconstructions. Results show that the most accurate on-the-fly projector modeling the intrinsic detector response function demonstrated superior performance in terms of image contrast, noise and resolution compared to the reconstructed images using the stored SRM (even of the highest statistical quality). At the same time the need for substantial memory in the case of the stored SRM precludes an efficient GPGPU implementation, which in the case of the most accurate on-the-fly projector (IRIS) leads to reduced computational cost reconstructions (29secs / iteration).

In this work, an experimental investigation was carried out to study the effect that positron range has over microPET scans of line-sources embedded in tissue-equivalent phantoms. Line-sources consisted of 0.65 mm internal diameter plastic capillary-tubes filled with F-18 or Ga-68 solutions (with 635 keV and 1.9 MeV maximum ß+ energies, respectively). These were inserted along the axis of symmetry of cylindrical phantoms constructed with tissue-equivalent materials including lung (inhale and exhale), adipose tissue, water, trabecular and cortical bone. PET scans (15 min) were performed with a microPET Focus 120 (Concorde Microsystems) and image reconstruction with FBP and OSEM2D algorithms. Line-source distributions were analyzed using radial profiles taken on axial slices from which the spatial resolution was determined through the FWHM and FWTM. The activity concentration was determined and compared to the known values for each material. Results show wider distributions with longer tails for Ga-68 compared to F-18. A plot of the FWHM and FWTM as a function of physical density of each material was produced for each isotope. Differences between activity concentration, as seen by the microPET and direct measurement using a well-counter, were as large as 80% and 58% for Ga-68 in lung-inhale equivalent material reconstructed with FBP and OSEM2D, respectively. Our experiments indicate that positron range in different materials has a strong effect on both spatial resolution and activity concentration quantification in PET scans. These results point out that extra care should be taken when computing standard-uptake values (SUV) in PET scans, in particular when the radiopharmaceutical is taken up by different tissues in the body, and more even so with high-energy positron emitters such as Ga-68.

In Emission tomography, maximum a posteriori (MAP) reconstruction is one of the strategies to impose desired property into the images. It can be achieved by various optimization algorithms. Since each algorithm takes its unique way to approach the solution, the convergence rate varies accordingly. The aim of this study is to compare the convergence rate of different MAP algorithms, including one-step-late MAP expectation maximization (OSL-MAPEM), the relaxed ordered subsets separable paraboloidal surrogate (OSSPS) and the limited-memory BFGS with boundary constraints (LBFGS-B). The convergence rate of all 3 algorithms was insensitive to the type of prior. OSL-MAPEM and relaxed OSSPS converged at different rates for different strength of the penalty and activity distribution. In contrast, the LBFGS-B converged in approximately the same number of iterations for all cases, which is desirable for clinical application.

In this work, we investigate and compare the use of different priors in the maximum-a-posteriori (MAP) reconstruction of doubly-gated (noisy) cardiac PET datasets. OSEM3D reconstructions of the thorax, with and without resolution modelling, are compared to MAP reconstructions regularized with the total variation and the relative difference prior. Anatomy-based priors with either matching or mismatched anatomical information (wrong cardiac phase, mis-registered) are also included in the comparison. The aim is to verify whether the use of anatomy-based priors during reconstruction is worth the additional computational and acquisition burden, in the case where cardiac datasets (subject to motion and prone to misregistration) are of interest. Simulated datasets, obtained with the XCAT software, are reconstructed with the different algorithms and quantitatively analysed. The first results of the simulation study show a slight superiority of the anatomy-based prior when an ideal, perfectly matching anatomy is used. In case of mismatched or missing anatomical information, the quality of the anatomy-based MAP reconstructions decreases. The edge-preserving priors, on the other hand, produce reconstructions with good noise properties and recovery of activity, with the advantage of not relying on an external, additional scan for anatomy.

Yttrium-90 (Y-90) microsphere radioembolization is used in the treatment of both primary and metastatic cancer in the liver. PET/MRI has the potential for better prediction of treatment outcome due to improved anatomical images and potentially more accurate dose delivery assessment. However, the positron branching ratio is only 32 ppm and yields images with poor statistics. The purpose of this study is to find optimal PET reconstruction parameters for PET/MRI that provide images optimized for count recovery and minimal noise. Y-90 chloride solution was used to fill an ACR phantom imaged with a PET/MRI scanner (Siemens Biograph mMR) at two institutions. Four hot cylinders and warm background activity of the phantom were filled with a 10:1 ratio. The phantom was imaged at a range of Y-90 activities (7.5–1.0 GBq total activity) on a PET/MRI and PET/CT (either Siemens Biograph 40 or mCT scanner). PET images were reconstructed via OP-OSEM (1–5 iterations, 21 subsets), post-reconstruction Gaussian filter size (5–10 mm), with either absolute or relative scatter, and with or without point spread function (PSF) compensation. Time-of-flight (TOF) and non-TOF measurements were also compared for scans from mCT. Recovery coefficients and noise properties were measured as well as total activity in the phantom. For the 120 different reconstructions, recovery coefficients ranged from 0.1–0.6 (within 15% between each institution) for both PET/CT and PET/MRI and improved with increasing iteration number and decreased with post-reconstruction filter size. Little improvement in recovery coefficient was observed beyond 3 iterations. Recovery coefficients with TOF measurements typically improved slightly compared to non-TOF measurements and were the most significant with the smallest cylinder on the 1 GBq scan. With PSF, 3 iterations, 21 subsets, and applying a 5 mm filter provided optimal recovery coefficients at a moderate noise level over a wide range of activity levels.

Extracting prior knowledge from previous high quality normal-dose computed tomography (NdCT) data for Bayesian reconstruction of current low-dose CT (LdCT) images has attracted great research interests recently. While most efforts focused on registering the previous NdCT data to the current LdCT reconstruction, this work investigated an alternative strategy by extracting the local structure-specific spectral information of the NdCT data to preserve the local image textures in the LdCT reconstruction, because many clinical studies have revealed the image textures are clinically desirable. In implementation, we adapted a Gaussian Markov random field (GMRF) model to incorporate the spectral information for knowledge-based penalized weighted least squares (PWLS) LdCT reconstruction. Specifically, we explored a novel idea of extracting the spectral information by statistically estimating the weighting coefficients of the GMRF prior model. We adopted an iteratively-reweighted least squares regression (IRLSR) method to statistically estimate the linear relationship between the central voxel intensity and adjacent neighboring voxel intensities, which overcomes the limitations of ordinary least square regression. This novel idea led to very encouraging results as demonstrated by outcomes of the Bayesian reconstruction of LdCT image data.

Metallic objects in the field of view (FOV) cause metal artifacts in medical imaging. In X-ray computed tomography (CT), there are several ways to eliminate the effects of these artifacts. These methods are called metal artifact reduction methods. The iterative MAR algorithms are mainly based on identifying the metallic regions and using this information in the reconstruction algorithm for modeling of the system matrix. This paper aims to use a novel reconstruction algorithm in order to reconstruct images with clearly defined metal/tissue boundaries and reduction of metal artifacts around the metallic regions. We present a novel metal artifact reduction algorithm using maximum a-posteriori expectation maximization (amMAP-EM), together with a multilevel segmentation algorithm based on Otsu’s threshold. Adaptive multiresolution method for maximum a-posteriori expectation maximization (amMAP-EM) has been used for reconstruction of missing wedges problem in emission tomography earlier. It was selected for metal artifact removal problem as the identified metallic regions can be regarded as missing information from the data. The segmentation method improves the Otsu’s threshold performance, which performs non-ideally for two very different class sizes. In order to evaluate the effect of different sizes, shapes and locations of metals, several phantoms contaminated with Gaussian noise were used. The performance of the algorithm was compared against filtered backprojection (FBP) and maximum likelihood expectation maximization (MLEM). According to the results, it is possible to reconstruct the images with clear and sharp metal/tissue boundaries, while avoiding the undesired artifacts such as blurring, streak artifacts or ringing.

Due to the nature of the Compton Camera operation, the detection efficiency for distributed gamma-sources is strongly affected by the angular acceptance and the interaction probability for a given tracer energy. In order to overcome imposed artifacts and other deformation effects in the final image, an optimization method for the reconstruction procedure is proposed. It is based on a GEANT4/Gate simulation study and the parametrization of the recorded planar distortion of homogeneous plane sources placed at different distances inside the camera’s field of view. The efficiency of this method is presented for a variety of 3D reconstructed phantoms and the achieved improvement in spatial resolution is discussed.

We have developed an analytic approximation for the image pixel covariance in CT images reconstructed with a total-variation (TV) penalty using the iteratively-reweighted least-squares algorithm. Use of the proposed approximation is demonstrated by investigating the dependence of noise correlations on the strength of TV regularization. By minimizing similarities between the image noise correlation structure and signals of interest (e.g. microcalcifications in dedicated breast CT), we hypothesize that the reconstruction algorithm can be optimized to improve signal detection performance. In order to illustrate this application, a simulated microcalcification cluster is reconstructed using different strengths of TV-regularization and the corresponding images and noise correlation structures are investigated. In the future, the proposed method could also potentially form the basis for image quality metrics based on the Hotelling observer.

We introduce the first ever ordered subset expectation maximization (OSEM) method for joint estimation of emission activity distribution and photon attenuation map from positron emission tomography (PET) emission data alone. It differs from standard OSEM in the use of distinct- as opposed to identical subsets for emission activity distribution- and photon attenuation map update. The method accelerates EMAA (expectation maximization activity and attenuation estimation) by 50x, reducing the two dimensional estimation time from 14 hours to 17 minutes. EMAA with and without OSEM results in similar emission activity distribution- and photon attenuation map estimation errors.

Low-dose X-ray computed tomography (CT) has garnered much recent interest as it provides a method to lower patient dose and simultaneously reduce scan time. In non-medical applications the possibility of preventing sample damage makes low-dose CT desirable. Reconstruction in low-dose CT poses a significant challenge due to the high level of noise in the data. Here we propose an iterative method for reconstruction which maximizes the transmission Poisson likelihood subject to a total-variation constraint. This formulation accommodates efficient methods of parameter selection because the choice of TV constraint can be guided by an image reconstructed by filtered backprojection (FBP). We apply our algorithm to low-dose synchrotron X-ray CT data from the Advanced Photon Source (APS) at Argonne National Labs (ANL) to demonstrate its potential utility. We find that the algorithm provides a means of edge-preserving regularization with the potential to generate useful images at low iteration number in low-dose CT.

In recent years, panel PET is an interesting field in medical imaging. But the major obstacle of panel PET is the limited view problem. Images reconstructed by traditional algorithms will suffer serious artifacts. Time-of-flight (TOF) information can be incorporated into the image reconstruction to remove the artifacts. However, the best system timing resolution available in current commercial PET scanners is about 500 ps, which is not precise enough to achieve satisfactory images. In this paper, a list mode reconstruction algorithm coupled with non-local means regularizer was formulated to improve the panel PET image quality. It incorporates the anatomical information into the non-local means regularizer and modifies the reconstructed image while it is updating. The results show that our proposed method could remove the distortions effectively without TOF information.

Image reconstruction for Y-90 SPECT is challenging due to wide continuous energy spectrum of the bremsstrahlung photons. Previously, forward projection models with narrow single-energy window or wide multi-energy windows were used for image reconstruction with a single acquisition window. We propose a new Y-90 SPECT image reconstruction method referred to as joint spectral reconstruction (JSR) using multi-energy window forward model and multiple acquisition windows. Monte Carlo (MC) simulation using SIMIND coupled with XCAT phantom (two tumors with 80, 41 mL) generated two sets of 6 acquisition window measurements (narrow and wide windows). Each measurement was scaled to the count level of a typical patient bremsstrahlung SPECT scan following Y-90 microsphere radioembolization of the liver. 3D OSEM with 6 window data and a single image was implemented with 6 energy-dependent collimator detector response, attenuation, and known scatter and was run with 100 iterations, 6 subsets. The reconstructed image was evaluated using the recovery coefficient (RC), the standard deviation (SD), the bias over tumors and liver. Results were compared with those of the single spectral reconstruction (SSR) method that models the projector with a single-energy window with a single acquisition window. JSR yielded significantly higher RC and lower SD than SSR in all cases. For narrow window case, JSR achieved up to 11.19% increase in RC and up to 46.9% decrease in SD compared to SSR at the 100th iteration. For wide window case, JSR achieved up to 19.23% increase in RC and up to 19.01% decrease in SD compared to SSR at the 100th iteration. Therefore, our JSR was able to considerably improve Y-90 SPECT image reconstruction quality over SSR. The improvement in noise while maintaining comparable RC is particularly significant because the sensitivity of Y-90 SPECT is low and count levels for applications in targeted radionuclide therapy are low.

The finite focal spot size is one of the major physical limitations to the spatial resolution for CT reconstruction. In this paper, a projection restoration based method is proposed to improve the spatial resolution. First of all, the focal spot size is considered into the projection model. Then the equivalence relation between the measured projection and the ideal projection is deduced. Based on such relation, the equivalent ideal projection can be recovered from the measured data before reconstruction. Finally, a high-resolution image is reconstructed using the recovered projection. Furthermore, an improved Non-local Mean method is presented to suppress the noise in the high-resolution image. Both numerical simulation and experiment results have shown that the proposed method with the outstanding performance and efficiency characteristics can significantly enhance the spatial resolutions

In the reconstruction, the misaligned geometry will cause streak artifacts which influence image quality. Here we proposed a method to calibrate all the parameters simultaneously with the matrix form. In many cases, the relative position between the X-ray source and the detector is known and invariant, and then we can build a Cartesian coordinate system for the source and the detector. At the same time, we build another reconstruction coordinate system for the phantom. Space coordinate transformation can be achieved by the rotation matrix M and the translation matrix T. In our method, firstly we get non-collinear three points’ coordinate in both coordinate systems, and then calculate the two matrixes. Using the above matrixes, the projection data of each point at each angular position can be calculated. Our mathematical expressions are simple and the advantage of our method is its feasibility and stability. The calibration procedure has been tested on an actual dental-CT system.

Multi-modality medical imaging has found its increasing application for better healthcare. The main advantage of multi-modality imaging is that the weakness of each modality are offset by the others. MRI and X-ray CT simultaneous imaging is a possible multi-modality imaging mode in clinical applications of the further. Both MRI and CT are widely used imaging modalities. Though both of them both belong to structure imaging, MRI and CT can provide complementary information about patient anatomy. Usually, MRI may offer good soft tissue information while CT depicts better quality image on bones. Hence, the idea of CT-MRI hybrid imaging integrated in one machine was put forward, which is expected to provide high image quality in both soft tissue and bones. Moreover, because CT scanning is much fast than MRI, it is possible to greatly shorten the scanning time of MRI by reconstructing the MRI image from the highly under-sampled data and the information of the CT image at the same tissues. This paper focuses on the problem of multi-modalities medical image reconstruction problem (MRI-CT), and proposes a self-adaptive mask-enhanced dual-dictionary learning (DDL) for MRI image reconstruction with highly under-sampled measurement data and training data of well-registered MRI and CT images. Its main feature is to establish a one-to-one mapping relationship between MRI image and CT image of the same tissues, and obtain a good initial value of MRI imaging from the CT data. Numerical simulations were performed to evaluate and validate the proposed algorithm.

The combined effort of several laboratories at our institution resulted in the building of a second generation dedicated, high resolution PET/CT prototype for imaging the human breast and extremities, referred to as Davis breast PET/CT (DbPET/CT 2.1), that has comparable performance to the best breast PET systems currently under clinical evaluation in addition to a state of the art breast CT with resolution that approaches that of mammography. To partially overcome the limitations of the limited axial Field-of-View (AFOV) and sensitivity of the PET scanner we developed an helical acquisition geometry similar to that of third-generation CT scanners. In this work, we investigated the challenges that a helical acquisition trajectory poses to the completeness of data sampling, developed multiple helical acquisition protocols for DbPET/CT 2.1 and reconstructed images of complex phantoms with step and shoot as well as helical acquisitions. The information about gantry angular and vertical position throughout the acquisition is used to generate a 4D sensitivity map that for each plane in the FOV determines for how many seconds as well as at which angles that plane was sampled. The reciprocal of the sensitivity map is used as a weighting factor for each event in the list-mode dataset and fed into a customized filtered back-projection (FBP) algorithm. This weighting factor contributes to the normalization which is performed on a event-by-event basis rather than on a LOR-by-LOR due to the rotation of the system. We showed that we can obtain artifact-free images of complex phantoms with helical acquisition protocols and slightly better contrast recovery coefficients. The first patient has been scanned on DbPET/CT 2.1 with a helical protocol and implementation of a list-mode, iterative reconstruction algorithm for arbitrary trajectories is under way.

Time-of-flight (TOF) PET imaging enables a more precise localization of an emitted event along a given line-of-response (LOR). The additional temporal information could be used to restrict the effective coincidence window in the scanner to conform to the object being imaged. This can be achieved by truncating the sinogram bins along the TOF direction, keeping only those that contain LORs passing through the object, while further truncation can be done in the transaxial direction. When the region of interest (ROI) is localized and known, TOF sinogram truncation could then be used to exclude data not only outside the patient outline but further inside as well. One specific type of imaging which could substantially benefit from such an approach is dynamic imaging. By truncating the sinograms to conform to a specific ROI which is often needed in dynamic imaging, significant computation benefits could be obtained. Moreover, additional benefits could be realized through reduced propagation of errors specific to dynamic imaging. Using simulated and clinical studies, we investigate the feasibility and benefits of limited FOV imaging from truncated TOF sinograms in dynamic imaging.

Nuclear imaging with low-energy gamma-emitting tracers for diagnostic purposes, such as for pathologies of the thyroid, is usually performed with whole-body scintigraphy and SPECT systems. Interventional radioguidance, such as for targeted biopsy, mostly relies on single-channel gamma probes, multi-channel 2D mini gamma cameras or the reconstruction of a 3-dimensional freehand SPECT image based on acquisitions with a tracked handheld gamma probe. In this work, we evaluate the performance of a system for flexible and patient-specific 3-dimensional gamma imaging based on a tracked mini gamma camera. A major challenge that needs to be overcome is the reconstruction of data with low statistics, due to the relatively low activity uptake in the thyroid of patients, and the continuous nature of such a handheld acquisition. We propose recording data in a list-mode format and reconstruction with a list-mode expectation-maximisation (LM-EM) algorithm. Results are acquired with the mini gamma camera mounted on a robotic arm for precise tracking and repeatable acquisition trajectories. Images of a custom 3D thyroid phantom containing hot and cold nodules confirm the feasibility of the approach. We further simulate different uptake values by statistically thinning out the acquired data, showing that different radioactivity uptakes can be reliably reconstructed.

Linac-mounted cone-beam computed tomography (CBCT) imaging systems help with patient setup and treatment volume verification in image guide radiation therapy. However, analytic-based reconstruction algorithms, such as FDK, restrict the imaging isocenter to the mechanical isocenter of the gantry. Unfortunately, this makes it difficult to implement a virtual isocenter, or a treatment isocenter not at the mechanical isocenter of the device. This is a technique often used in head and neck cancer as well as breast cancer patients. New optimization-based reconstruction algorithms have the potential to reconstruct a virtual isocenter. We demonstrated this by acquiring a CBCT scan of such a virtual isocenter with Varian's TrueBeam Developer Mode and subsequently reconstructing it.

Background: Simultaneous reconstruction of activity and attenuation with MLAA offers an interesting option to solve the problem of attenuation correction in hybrid PET/MR scanners. The algorithm maximizes the joint likelihood by alternating MLEM and MLTR steps. Any scheme where a number of MLTR iterations follow any number of MLEM ones converges to the true solution, but convergence speed and computation time might differ significantly. We investigate which alternating strategy is expected to give optimal results as a function of TOF coincidence timing resolution (CTR), given different computation time and convergence speed between TOF-MLEM and non-TOF MLTR. Following, acceleration strategies exploiting ordered subset (OS) techniques are proposed and tested. Methods: A slice of a 2D digital body phantom was simulated and forward projected with CTR of 200, 400, 600 and 800 ps. Reconstructions were performed using the corresponding CTR using 1, 2, 4 and 8 MLTR iterations for each MLEM iteration, using native-geometry, ray-driven forward and back-projectors. Convergence curves were plotted as a function of computation time. Different strategies of OS acceleration were then tested and compared using simulations both noiseless and featuring Poisson noise, using 32 subsets. Results: Convergence was achieved with the least computation time when 4 or 8 MLTR iterations were performed for each MLEM one. All proposed OS strategies had results similar to reference non-accelerated versions and performed equivalently also in presence of noise, despite the high subsets number. Conclusions: Performing 4 or 8 MLTR iterations for each MLEM one appears to be the optimal solution for current implementations of the MLAA algorithm. OS techniques are effective in reducing computation time and perform comparably to non-accelerated versions.

Given the complexity of the image reconstruction procedures for Compton Camera events, especially when a 3D image is required for distributed sources in space, a simple, direct algorithmic approach is presented in this work. The developed ComptonRec package carefully handles the geometry of the conic sections to accumulate ray density distribution in a user defined voxelized volume inside the specified field of interest. Prior to planar reconstruction, the event selection part of the program filters out misidentified coincidence events and other physical background events with unbalanced total energy or inverse interaction sequence with the camera’s subsystems. For each accepted event a series of planar reconstructions is performed, where the density distribution is the accumulation product of the conic intersection with all the affected pixels. A 3D image is finally reconstructed by assembling the partial planar information and by taking into account volume effects. The efficiency of this reconstruction method is checked with a plethora of simulated phantoms and results are presented and discussed.

In this work a new method for image reconstruction in computed tomography (CT) using QR-decomposition is proposed. As the image reconstruction problem can be modelled by a large linear system of equations, QR-decomposition could be used to solve this system. Current advances in computer science enable the use of direct methods for solving such large linear system (10⁶ unknowns). Techniques as Singular Value Decomposition (SVD) and QR-decomposition [1] have already begun to be investigated as a new arising alternative to traditional methods like Filtered Back Projection (FBP) and Expectation Maximization (EM) methods. QR-decomposition method let us speed up image reconstruction because heavy computational cost are precalculated and stored once for a given CT system, and from then on, each image reconstruction only involves a backward substitution process [2]. Three-dimensional (3D) images obtained with QR-decomposition have been analyzed and compared with those obtained with FBP and MLEM methods. Results show that QR-decomposition process achieves competitive advantages compared to FBP and MLEM reconstruction with the same voxel size. Comparisons using sharpness and noise power spectrum (NPS) between QR with FBP and MLEM show that QR-decomposition provides higher image quality images using identical voxels. This paper highlights the possibilities of QR-decomposition for CT image reconstruction. However, QR-decomposition method can be used not only for CT image reconstruction, but also for any imaging technique involving tomography data, as those used in Nuclear Medicine reconstruction (i. e. Positron Emission Tomography). [1]MJ. Rodríguez-Álvarez et al., Sparse Givens resolution of large system of linear equations:Applications to image reconstruction,Mathl. Comput.,52,258–1264,2010. [2]A. Iborra et al., Noise analysis in computed tomography (CT) image reconstruction using QR-Decomposition algorithm. IEEE T.Nucl. Sci.,DOI:10.1109/TNS.2015.242221.

Positron Emission Tomography (PET) is an important tool for nuclear medicine imaging, because of its ability to provide quantitative information of probe distribution. However, PET works only for a single probe, because positron-electron annihilation ?-rays utilized for PET imaging have inevitably constant energy of 511 keV irrespective of radionuclides. In order to overcome this restriction, we have been developing Polychrome-PET (P-PET). The P-PET identifies different probes based on detection of de-excitation ?-ray, which is emitted after the positron. For a feasibility study of the P-PET, we performed experimental studies and computer simulation. In the experimental study, we employed a planer type PET system integrated with additional ?-ray detectors and succeeded in multiple probes imaging. In the computer simulation, we used a Monte Carlo simulator GATE and simulate multiple probes imaging by a ring type PET system with additional ?-ray detectors. These experimental and simulation studies revealed that a specific image reconstruction algorithm is needed for multiple molecular imaging using the P-PET. Specifically, because of limited detection efficiency for de-excitation ?-ray, reconstructed image without detection of de-excitation ?-ray contains both pure positron emitter and positron-? emitter. Therefore, in order to reconstruct isolated image of the pure positron emitter, image subtraction between images with and without detection of the de-excitation ?-ray is indispensable. Moreover, the image reconstructed from list-data with detection of the de-excitation ?-ray contains distribution of the pure positron emitter as the noise of accidental coincidence. This noise can be canceled by background subtraction.

Time-of-flight (TOF) technology applied to positron emission tomography (PET) not only can improve image quality of PET reconstruction, but also can be utilized to estimate attenuation maps. In this work, we aim to reconstruct both attenuation and activity images of the dual-head small-animal PET (DHAPET) system, and developed an advanced algorithm based on the proximal point algorithm and the alternating direction method of multipliers (ADMM) to solve the composite proximal problem. The activity and attenuation maps can be estimated and updated at each iteration despite the measurement data are incomplete. Additionally, the proposed method imposed the anatomical prior that is imposed by adding the well-known median root prior (MRP) to the ADMM algorithm to achieve better attenuation estimation

Positron emission tomography (PET) is an essential imaging modality to quantitatively recover the tracer distribution and hence the underlying kinetic mechanism for clinical or research purposes. Combined with computed tomography (CT) or magnetic resonance imaging (MRI), PET/CT or PET/MRI can gain complementary information from different perspectives. CT or MRI images can be used as an anatomical prior, attenuation correction, image segmentation template, etc. In this work, we incorporated CT or MRI image as the anatomical prior to improve the reconstruction of PET images. The priors were computed based on simple edge preserving, i.e., the median root prior (MRP) and label mean prior (LMP) that aim to retain the true intensity edges without blurring the edges by replacing the pixel values among neighboring pixels with the median or mean value in a predefined anatomical region [1]. In this work, we employed the splitting-based fast iterative shrinkage-thresholding algorithm (FISTA) to solve the weighted sum of LMP and MRP. However, when anatomical boundaries fail to correspond to or align well with the functional regions, reconstruction errors will occur. To circumvent these, we proposed a run-time segmentation scheme, in which anatomical boundaries from CT or MR image will be utilized as the prior knowledge in the PET reconstruction. The prior boundaries will be recomputed after ten iterations, and could lead to better activity reconstruction results.

Abstract--- In recent years, the clinical application of C-arm cone-beam computed tomography (CBCT) has been widened from the original field of interventional neuroradiology to the field of peripheral procedures. Since the C-arm CBCT can provide the 3D information about the liver cancer and its feeder, creating a vascular roadmap to guide catheters for embolization procedures, it has been demonstrated to be a valuable tool in the treatment of liver cancer. However, current reconstruction algorithms employed in the C-arm CBCT system, FDK-based reconstruction algorithms, are susceptible to cone-beam artifacts, metal artifacts, and truncation artifacts. In this work, we have investigated an optimization-based algorithm, the adaptive-steepest-descent-projection-onto-convex-sets (ASD-POCS) algorithm, tailored to reconstructing C-arm CBCT images. The results show that the optimization-based algorithm can yield the images with significant reduction of image artifacts compared to the conventional FDK algorithm.

Multi-modality imaging raises issues in how to exploit common information from the different modalities. Joint reconstruction methods link common features of the images such as edges, promoting joint gradient sparsity. Thus reconstructions are improved or, in case of highly corrupted data, at least reasonable and useful reconstructions are permitted. Promising recent advances successfully coupled PET and MR data \cite{Knoll, Erhardt}, but bring disadvantages, such as penalizing image intensities. We propose an adaption of an iterative multichannel Bregman reconstruction method \cite{Moeller}, originally applied to images with multiple color channels, to overcome these issues and provide a more versatile framework for multi-modality imaging. In the case of total variation regularization, it is capable of linking common edge information in form of common subgradients, while maintaining the well-known properties of Bregman iterations such as contrast preservation. We present promising results on artificial PET-MR data.

In practical cone-beam computed tomographic applications, it is challenging to employ optimization-based (e.g., iterative) algorithms to reconstruct 3D high-resolution images because of the heavy computational load. It is worth noting that one is often interested in detailed information within a region of interest (ROI), while rough knowledge outside the ROI may be sufficient. Therefore, it is of practical merit to develop algorithms that are capable of reconstructing an image with variable resolution: an image consisting of high-resolution voxels within the ROI and low-resolution voxels outside the ROI. In this work, we investigate and develop optimization-based algorithms for 3D image reconstruction with variable resolution and apply them to real patient data collected with a dental CBCT scanner. The results of our study demonstrate that optimization-based algorithms can be developed for yielding an image can have different levels of spatial resolution. The work may have implications for the reduction of computation memory and acceleration of computation speed. In the presence of data truncation, it may also provide an approach to obtaining an ROI image of high resolution with minimized truncation artifacts.

It is an important cost consideration to reduce detectors in a PET-system design, while not significantly compromising the PET utility. Recently developed optimization-based algorithms may be used for enabling the design of innovative PET systems. In this work, we investigate PET configurations with reduced number of detectors. We consider an optimization problem combining Kullback-Leibler (KL) data fidelity with an image TV penalty, and solve it by using a primal-dual optimization algorithm developed by Chambolle and Pock. We carry out IEC phantom data studies that demonstrate the potential of the algorithm in enabling the design of advanced PET imaging configurations with reduced number of detectors.

In particle beam therapy, sudden drainages or filling of cavities that located across the beam track in the body cause unexpected shifts of the Bragg-peak positions. Therefore, detection techniques of such changes of the states of the cavities are of major importance to avoid mistakes of irradiations. The purpose of this work is to study the feasibility of detection of a gap which is located across a beam track by measuring low energy (63-68 keV) photons generated by the beam irradiation. An experiment was performed with the heavy-ion medical accelerator in Chiba (HIMAC) in the National Institute of Radiological Sciences (NIRS). A carbon-12 beam having 290 MeV/n was injected on a target consisting of two acryl blocks. These two blocks were placed with a 10 mm gap along the beam axis. A detection system consisting a CdTe semiconductor detector and a lead collimator having a 2.4-mm-width slit was placed on a movable stage to detect the low energy photons emitted perpendicularly to the beam axis. The position of the detection system was moved in twelve positions with 2 mm pitch along the beam axis. The yields of 63-68 keV photons were clearly correlated with the positions of the detection system. The yields had the minimum value at the center of the gap and the gap position could be estimated by the result. We are also going to report a comparison of the experimental result with that of a Monte Carlo simulation.

Proton imaging can improve dose delivery compared to that provided by traditional X-ray computed tomography, while also requiring lower doses for image production. We describe the novel proton CT scanner that is based on a fiber tracker and a scintillator-stack range detector, which has been developed at Northern Illinois University in conjunction with Fermi National Accelerator Laboratory. The scanner can operate at 2 MHz and provides large areal coverage, 20x24 cm2,for imaging an adult head phantom. We report initial results from tests of the major scanner components: the Range Stack Calorimeter, the Tracker, and the Data Acquisition system, and present the analysis of scanner tests conducted at the CDH Proton therapy center in Warrenville, IL.

Calculation of a dose distribution inside a human body is based on CT images and electron densities which are converted from CT numbers in radiation treatment planning. Although tissue types are necessary for accurate dose calculation, most existent calculation software in treatment planning systems recognize all human tissues including bone as water. Alternative index for tissue identification for CT images are necessary. The effective atomic numbers (Zeff) of materials are attracted attention as alternative to CT numbers. Zeff is a hypothetical atomic number which explains the degree of x-ray attenuation in a composite. Zeff can be measured using two monochromatic x-rays from synchrotron. For the Zeff measurement at hospitals, we have proposed the energy-resolved CT (ERCT) with a current-mode “transXend” detector system which can give energy distribution. For the purpose of tissue identification in clinical CT images, we measured Zeff of eleven different human tissue equivalent materials by the ERCT with the transXend detector system. Originally developed transXend detector was only applicable for the first generation CT. To apply the transXend detector system to the third generation CT, we developed a transXend detector system with a flat panel detector (FPD) and grid filters. The FPD with GOS scintillator is used. The pixel size is 48 µm x48 µm. Tissue types are classified into three groups; lung, soft tissue, and bone. Response functions are obtained from calculated current values for 0-50 mm each material thickness using Lambert-Beer’s law. X-ray tube conditions are 120 kVp, 2.3 mA and 1 s. Transmission measurements for phantoms are conducted. All Zeff are underestimated systematically, 93-95 % of theoretical values. The standard deviations of Zeff of all tissue types are within 0.15. Normalizing a measured value to one tissue, Zeff obtained by the present method can be practical.

One of the challenging applications of PET is implementing it for in-beam PET, which is an in situ monitoring method for charged particle therapy. For this purpose, we have previously proposed two geometries for our original open-type PET scanners named OpenPET. Following our initial proposal of the dual-ring OpenPET (DROP) in 2008, we developed a whole-body prototype of DROP (WBDROP) in 2014 to show a proof-of-concept. On the other hand, in 2011, we also proposed the single-ring OpenPET (SROP), which is more efficient than DROP in terms of manufacturing cost and its open gap width. In this paper, we developed the whole-body prototype of the SROP (WBSROP). The WBSROP prototype is designed with 4 axially shifted detector rings of 40 depth-of-interaction (DOI) detectors. The detector rings are each 66 cm in diameter, and they are slanted by 45 deg from the axial direction to obtain a large open gap width of 43 cm. The DOI detectors consist of 1024 Zr-doped GSO crystals which are arranged in 4 layers of 16 × 16 arrays, coupled to a 64-ch flat panel position sensitive PMT with a super-bialkali photocathode, which has a 30% higher quantum efficiency. Each crystal element is 2.8 × 2.8 × 7.5 mm³. In order to enable stable in-beam PET measurement even under high background radiations, voltage divider circuits are designed so as to have 5 times higher linearity. The front-end circuit equipped only with a resister chain and amplifier and the data acquisition system are separated from the gantry by 7-m coaxial cables to protect the electronics circuits from radiation damage by secondary particles. For sensitivity comparison between WBSROP and WBDROP prototypes, we simulated both scanners using Monte Carlo simulation. Predicted sensitivity of the WBSROP prototype was 3.1%. The WBSROP prototype promises high sensitivity although it has a wide open gap.

In-beam positron emission tomography (PET) is expected to enable in situ noninvasive confirmation of treatment delivery. For accurate range and dose verification or 3D volume imaging, it is necessary to correct the biological washout effect in a living body. The final aim of this study is to establish appropriate washout modeling not only for washout correction but also tumor viability estimation. Here, we measured the washout rate in a rabbit using oxygen ion beams as well as carbon ion beams. To measure components of washout, three species of nuclides, ¹⁰C, ¹¹C and ¹⁵O, which were generated as secondary beams, were irradiated on the rabbit brain and thigh under two conditions, live and dead. In-beam data were acquired by our OpenPET prototypes, which enable 3D in-beam imaging. Regions of interest (ROIs) were set as a 3D positron distribution and the time activity curves (TACs) of the irradiated field were acquired. We employed multiple washout components model according to the analysis of washout rate of carbon and oxygen ions. Significant diffusion of the positron distribution was observed in the live brain irradiation. Fitting experimental TACs of ¹¹C in the live condition showed that there are three components in the washout process (fast, medium and slow decay). The observed medium and slow decay rates of ¹¹C in brain were 0.30 min^-1 and 0.004 min^-1, respectively. Our present work suggested the TAC of ¹⁵O could be fitted to the two exponential functions, the observed medium and slow decay rates were 0.98 min^-1 and 0.01 min^-1, respectively. The ratio of medium and slow decay components were 80% and 20% for brain, and 40 % and 60 % for thigh muscle. The difference of washout speed in the carbon beams and the oxgeon beam was observed. This difference would be a great help to specify the chemical form of compounds with positron emitters, because unknown chemical forms are a major issue which makes washout modeling difficult.

To realize the accurate irradiation targeted to a small volume, confirmation of the target position is very important, and a temporal location of the target can help to improve the accuracy of the target localization. A kilo-voltage cone-beam computed tomography (CBCT) system mounted on a linear accelerator can verify the target on each treatment day, but this system requires a long data acquisition time, and therefore reconstructed images are affected by moving of a target. In this study, we proposed a method to reconstruct target motions obtained with the 4DCBCT using the sorted projection data according to the phase and displacement of the extracorporeal infrared monitor signal, and evaluated the proposed method with a moving phantom. In our method, for the phase binning, motion cycle of the marker were sorted by each 20 % phase of the cycle, and in the displacement binning, the positions of the signal were divided into five equally spaced sections. As a result, the reconstructed image using the phase binning and displacement binning method, in which blurrings were reduced. The results showed that the blurring of the target caused by the motion was reduced, and we could obtain the reconstructed images incorporating the target motion with our proposed method.

To investigate the feasibility of clinical on-line proton beam-range verification with PET, we simulated proton irradiation and PET acquisition using GATE, reconstructed images, and measured positron activity range (AR). The dependence of AR measurement precision and accuracy on count statistics was studied with different proton beams, phantoms, and PET imaging systems. Proton beams with 110-MeV energy, 50-mm and 10-mm diameters, and 0.3-Gy dose (for imaging only) at Bragg peak were simulated to irradiate 7 geometric phantoms with different materials and patterns, as well as a human brain phantom. A box-shaped (consisted of 4 detector panels) and a stationary dual-panel PET systems (both with 2X2X30 mm^3 crystals, 5 mm DOI resolution and 300X300X100 mm^3 FOV) were simulated for imaging. For each imaging acquisition, 50 repeated simulations were conducted for calculating the means and variations of AR values. Results showed that, when no smoothing technique was applied, AR measurement accuracy and precision were proportional to the reciprocal of square root of image count, indicating Poisson noise characteristics of PET data with different phantoms. Brain phantom study showed that <1 mm accuracy and precision of AR measurement can be achieved by both PET systems with the 50-mm diameter beam and 100-sec or longer acquisition time. With smoothing techniques, the required PET acquisition time can be substantially reduced to 20 to 50 seconds. For the 10-mm diameter beam, count statistics was significantly reduced with the same acquisition time, however, it was still feasible to achieve <1 mm accuracy/precision by the box-shape PET with 80-sec acquisition and applying smoothing techniques. In conclusion, this simulation study indicates that with a high-performance brain PET and improved methods of imaging and AR measurement, it is feasible to achieve <1 mm accuracy and precision of on-line AR measurement.

Proton computed tomography has been suggested as an imaging technique alternative to x-ray CT for proton therapy treatment planning. It provides better proton range accuracy because the relative stopping power is directly calculated from proton energy loss. The most likely path of each proton is an essential element for performing reconstruction of the relative stopping power, using iterative reconstruction techniques. Several approaches have been formulated to calculate the most likely path (MLP) but all based on the assumption that the protons are traversing a uniform medium. Here we report on the integration of a pixelated detector (Medipix) with the phase II proton CT scanner, built by the pCT collaboration, to study the accuracy of the MLP estimation. The Medipix consists of a square silicon semiconductor sensor chip bonded to a readout chip; the phase II proton CT scanner consists of a silicon-based particle tracking system and a multi (5) stage scintillator (MSS). The DAQ systems of the Medipix and pCT scanner were linked so that the Medipix measurements were triggered by the signal provided by the first stage of the MSS. The Medipix was placed between the pCT scanner telescopes, embedded between slabs of different tissue equivalent materials. Coordinates and time of proton hits in the Medipix were recorded in coincidence with the hits recorded by the pCT scanner. On average, about 5000 matching events were recorded for each experimental run. The study of the effect of heterogeneities on the MLP accuracy using the acquired data is currently ongoing. To improve accuracy and spatial resolution of pCT reconstruction, future work will include creating a new formalism to calculate the MLP taking into account the different scattering power of heterogeneous materials.

In recent years, we have developed precise carbon-interspaced antiscatter grids having superhigh grid strip densities in the range of 200 – 300 lines/inch by adopting the precision sawing process for the demands of specific imaging techniques. However, because grid strips of the recently developed grids are usually invisible through a nondestructive x-ray testing with a flat-panel detector having an ordinary pixel resolution (> 100 µm), a thorough inspection in the grid manufacturing has not yet satisfactorily performed. In this study, we proposed a useful method to evaluate actual grid strip densities over the Nyquist sampling rate based on the moiré pattern analysis. We performed systematic simulation with sample grids having a wide range of strip densities (i.e., 70 – 400 lines/inch) and a detector having a 143-µm pixel resolution to demonstrate the viability of the proposed method. According to our simulation results, the relative differences between the original and the evaluated grid strip densities were within 0.2%, demonstrating that the proposed method is applicable for quality assurance in the grid manufacturing.

Various Monte Carlo (MC) codes exist for imaging simulations, including in the field of Single Photon Emission Computed Tomography (SPECT). Recently, new strategies exploiting the computing capabilities of graphical processing units (GPU) have been proposed. This work aims at evaluating the accuracy of such GPU implementation strategies in comparison to standard MC codes in the context of SPECT imaging. GATE was considered the reference MC toolkit and used to evaluate the performance of newly developed GPU Geant4-based Monte Carlo Simulation (GGEMS) modules for SPECT imaging. Point source, uniform source and Jaszczak phantom acquisitions were simulated using a model of the GE Infinia II 3/8" Gamma Camera. Two different isotopes were considered: ^99mTc and ¹³¹I using a Low Energy High Resolution (LEHR) and a Medium Energy General Purpose (MEGP) collimator respectively. Overall image quality and associated computing performance of several implementation strategies were tested using same simulation set-up. Considering ^99mTc simulated acquisitions with the LEHR collimator, profiles across the GATE and GGEMS projections agreed well, with equivalent root mean squared difference (RMSD) scores. Both simulation platforms yielded a similar system sensitivity and image statistical quality. The ¹³¹I simulations using a MEGP collimator were also found to be equivalent in terms of system sensitivity and image statistical quality. Finally, the overall acceleration factor between GATE and GGEMS platform derived from the same Jaszczak phantom acquisitions was between one and two orders of magnitude depending on the isotope/collimator couple. From a range of several sources and collimators we have assessed the modelling quality of GPU based MC simulations in the context of SPECT imaging. The good agreement with reference codes and the obtained acceleration factors support the use of GPU implementation strategies for improving computational efficiency of SPECT imaging simulations.

We studied the effect of the presence of an air inhomogeneity on the dose distribution from an Iriduim-192 brachytherapy source. The source was modeled using Monte Carlo (MC) code MCNP5. The AAPM Task Group No. 43 (TG 43) parameters were calculated to validate our simulation. We also performed experimental measurements to measure the dose distribution using EBT3 Radiochromic film. The dose in water at distance 1 to 5 cm was scored in the presence of an air cavity of diameters (3, 5, and 7 mm) at 2 mm distance from the source. The MC dosimetry results agreed to within 2% with the AAPM ESTRO (AAPM report 229) consensus data. The MC dose distribution was within 8.2% agreement with radiochromic EBT3 film measurement. The dose to water at 2 cm from the source increased by 3%, 6%, and 9% in the presence of a 3, 5, and 7 mm diameter air cavity, respectively. This difference is clinically too significant to ignore.

SPECT is primarily used in the clinic for non-invasive cardiac myocardial perfusion imaging. However, sensitivity is impaired due to the need for collimation in SPECT. New second generation systems including those with multi-pinhole collimation has shown high sensitive acquisitions. This has enabled faster or lower dose ?stress-first? acquisitions, but not both. Here a third generation dedicated Cardiac SPECT system is proposed which yields higher sensitivity by a factor of ~3.1 compared to second generation systems (and more than order of magnitude over Generation I systems), by using hemi-Ellipsoid shaped detectors with pinhole collimation. This may ultimately enable ultra-low-dose imaging (~3mCi) in less than 4 min.

Radionuclides are commonly used as internal radiation sources for diagnostic and treatment purpose. This study uses the Monte Carlo (MC) method to simulate breast cone beam computed tomography (CBCT) by using ¹⁹²Ir as an external source. A user code named egs_cbct in EGSnrc provides fast scatter calculation for CBCT projections. Feldkamp-Davis-Kress (FDK) algorithm is applied for image reconstruction. After comparing the original CT and reconstructed data, we conclude that radionuclide is a suitable external source for breast imaging, and the imaging method has a potential to be used for image-guided radiation therapy.

The Small Animal Fast Insert for mRi (SAFIR) will be a PET insert for the Bruker BioSpin 70/30. It aims at applications where fast processes such as blood perfusion in the rodent brain are to be monitored comprehensively and non-invasively. Employing electronics originally designed for time-of-flight applications, coincidence resolving times of less than a nanosecond can be achieved allowing for short coincidence time windows. Consequently, random contributions to the coincidence events are suppressed, making short acquisition frames with very high tracer concentrations possible. This will allow collecting sufficient count statistics in a few seconds. Geant4-based Monte Carlo simulations were used to characterize the performance of the reference design consisting of 2.0 x 2.0 x 12.0 mm³ polished LSO-like crystals, one-to-one coupled to silicon photo-multipliers. Grouped into 8 x 8 matrices, they will form a cylindrical scanner with an inner diameter of approximately 12 cm. Similar methods as described in the NEMA NU 4-2008 standard were employed. The simulation results on NECR, sensitivity, and spatial resolution will be presented. Most notably, the NECR at 500 MBq is more than six times higher than at 50 MBq for the given scanner. Combined with a very good sensitivity, this allows for short acquisition times using these very high injected doses. In addition, essentially random-free measurements at standard activities below 50 MBq are possible.

In this study, we present a comparative simulation of a depth-of-interaction (DOI) positron emission tomography (PET) detector using SiPM with a digital and analog readouts. To measure DOI information from a mono-layer scintillation crystal using a single-ended readout, our group has previously devised a method based on light sharing by using triangular teeth shaped reflectors. We previously showed the experimental evaluation of digital SiPM-based DOI encoding detector with good DOI-encoding performance. In this study, we choose to use digital SiPM (dSiPM) as a photosensor due to several reasons, and the most important advantage was the simple individual pixel readout which increases the accuracy of light distribution measurement. To verify that dSiPM is truly advantageous in our DOI measurement, simulation was performed using GATE Monte Carlo simulation toolkit. SiPM characteristics were considered focusing on the impact of readout scheme, electronic noise, and dark count noise. The simulation results revealed that dSiPM is a suitable photosensor in our DOI measurement compared to analog devices.

The hexagonal prism shaped crystal shows potential to design a PET system for its higher ? ray blocking efficiency and equidistance with neighbors. This paper proposed a honeycomb PET based on hexagonal prism shaped scintillation crystals. Its performance was evaluated and compared with a traditional PET system based on rectangular prism shaped crystals by GEANT4 simulation. The results show that the sensitivity is enhanced from 2% to 5% in most positions of the field of view (FOV). Meanwhile, the axial, radial and tangential spatial resolution are both improved in the positions near the center of the FOV (0mm, 10mm radial offset), but deteriorated with the offset become larger.

With high specificity of tumor necrosis, hypoxia tissues and subclinical lesions, PET may have great potential for tumor surgery guidance. PET images in surgery will tell us how tissues change during the operation and whether the lesion has been cleared. We did research on panel PET guided therapy, and we have developed 3-dimensional reconstruction methods for panel PET systems. However we meet the requirement of a larger open space for operation, because a body covered with two enclosed panels could barely be operated in most tumor surgeries. And the weak performance of images perpendicular to panels leads to invalid guidance in some applications such as radiotherapy guidance. In order to apply the better images for guidance, we designed a panel PET with window for tumor surgery guidance, and it showed good performance of the images parallel to the panels. It is surprising that we took 12 seconds for scanning and acquired images parallel to the panels with good performance in the actual system experiment.

Nowadays, Single Photon Emission Computed Tomography (SPECT) has become an essential part of molecular imaging and nuclear medicine. Whether to establish a diagnosis or in the therapeutic monitoring, this modality present performance that continue to improve. For over 50 years, several collimation profiles have been proposed. The most known and used are the parallel-hole collimator, the (multi) pinhole collimator and the fan-beam/cone-beam collimator. All these collimation profiles present different resolution-sensitivity trade-offs, which are directly related to their geometries. Mainly governed by collimation parameters, the resolution-sensitivity trade-off is the factor determining the collimation profile the most suitable for an intended study. Although these collimation profiles enable a large number of studies, the sensitivity remains a characteristic, which can limit the use of SPECT systems. One alternative to the common approaches is the rotating slat collimator (RSC). In the present study, we studied the resolution-sensitivity trade-offs obtained by varying different collimation parameters: (i) the slats height (H), and (ii) the gap between two consecutive slats (g). One system matrix is generated for each set of collimation parameters (H,g) using our analytical method. Spatial resolutions and sensitivity corresponding to all the set of collimation parameters (H,g) were measured using the 2D projections reconstructed with ML-EM. Quantitative measurements were also obtained extracted from the 2D projections using a rods phantom.